mirror/freebsd
1
0
Fork 0
mirror of https://git.FreeBSD.org/src.git synced 2024-12-14 10:09:48 +00:00
Code Issues Releases Activity
5108879730
freebsd/sys/net/if_lagg.c

2744 lines
66 KiB
C
Raw Normal View History

Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
/* $OpenBSD: if_trunk.c,v 1.30 2007/01/31 06:20:19 reyk Exp $ */
/*
* Copyright (c) 2005, 2006 Reyk Floeter <reyk@openbsd.org>
Add myself to the copyright.
2007-12-17 18:49:44 +00:00
* Copyright (c) 2007 Andrew Thompson <thompsa@FreeBSD.org>
Add an IOCTL rr_limit to let users fine tuning the number of packets to be sent using roundrobin protocol and set a better granularity and distribution among the interfaces. Tuning the number of packages sent by interface can increase throughput and reduce unordered packets as well as reduce SACK. Example of usage: # ifconfig bge0 up # ifconfig bge1 up # ifconfig lagg0 create # ifconfig lagg0 laggproto roundrobin laggport bge0 laggport bge1 \ 192.168.1.1 netmask 255.255.255.0 # ifconfig lagg0 rr_limit 500 Reviewed by: thompsa, glebius, adrian (old patch) Approved by: bapt (mentor) Relnotes: Yes Differential Revision: https://reviews.freebsd.org/D540
2016-01-23 04:18:44 +00:00
* Copyright (c) 2014, 2016 Marcelo Araujo <araujo@FreeBSD.org>
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
*
* Permission to use, copy, modify, and distribute this software for any
* purpose with or without fee is hereby granted, provided that the above
* copyright notice and this permission notice appear in all copies.
*
* THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
* WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
* MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
* ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
* WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN
* ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
* OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
*/
#include <sys/cdefs.h>
__FBSDID("$FreeBSD$");
#include "opt_inet.h"
#include "opt_inet6.h"
Add kernel-side support for in-kernel TLS. KTLS adds support for in-kernel framing and encryption of Transport Layer Security (1.0-1.2) data on TCP sockets. KTLS only supports offload of TLS for transmitted data. Key negotation must still be performed in userland. Once completed, transmit session keys for a connection are provided to the kernel via a new TCP_TXTLS_ENABLE socket option. All subsequent data transmitted on the socket is placed into TLS frames and encrypted using the supplied keys. Any data written to a KTLS-enabled socket via write(2), aio_write(2), or sendfile(2) is assumed to be application data and is encoded in TLS frames with an application data type. Individual records can be sent with a custom type (e.g. handshake messages) via sendmsg(2) with a new control message (TLS_SET_RECORD_TYPE) specifying the record type. At present, rekeying is not supported though the in-kernel framework should support rekeying. KTLS makes use of the recently added unmapped mbufs to store TLS frames in the socket buffer. Each TLS frame is described by a single ext_pgs mbuf. The ext_pgs structure contains the header of the TLS record (and trailer for encrypted records) as well as references to the associated TLS session. KTLS supports two primary methods of encrypting TLS frames: software TLS and ifnet TLS. Software TLS marks mbufs holding socket data as not ready via M_NOTREADY similar to sendfile(2) when TLS framing information is added to an unmapped mbuf in ktls_frame(). ktls_enqueue() is then called to schedule TLS frames for encryption. In the case of sendfile_iodone() calls ktls_enqueue() instead of pru_ready() leaving the mbufs marked M_NOTREADY until encryption is completed. For other writes (vn_sendfile when pages are available, write(2), etc.), the PRUS_NOTREADY is set when invoking pru_send() along with invoking ktls_enqueue(). A pool of worker threads (the "KTLS" kernel process) encrypts TLS frames queued via ktls_enqueue(). Each TLS frame is temporarily mapped using the direct map and passed to a software encryption backend to perform the actual encryption. (Note: The use of PHYS_TO_DMAP could be replaced with sf_bufs if someone wished to make this work on architectures without a direct map.) KTLS supports pluggable software encryption backends. Internally, Netflix uses proprietary pure-software backends. This commit includes a simple backend in a new ktls_ocf.ko module that uses the kernel's OpenCrypto framework to provide AES-GCM encryption of TLS frames. As a result, software TLS is now a bit of a misnomer as it can make use of hardware crypto accelerators. Once software encryption has finished, the TLS frame mbufs are marked ready via pru_ready(). At this point, the encrypted data appears as regular payload to the TCP stack stored in unmapped mbufs. ifnet TLS permits a NIC to offload the TLS encryption and TCP segmentation. In this mode, a new send tag type (IF_SND_TAG_TYPE_TLS) is allocated on the interface a socket is routed over and associated with a TLS session. TLS records for a TLS session using ifnet TLS are not marked M_NOTREADY but are passed down the stack unencrypted. The ip_output_send() and ip6_output_send() helper functions that apply send tags to outbound IP packets verify that the send tag of the TLS record matches the outbound interface. If so, the packet is tagged with the TLS send tag and sent to the interface. The NIC device driver must recognize packets with the TLS send tag and schedule them for TLS encryption and TCP segmentation. If the the outbound interface does not match the interface in the TLS send tag, the packet is dropped. In addition, a task is scheduled to refresh the TLS send tag for the TLS session. If a new TLS send tag cannot be allocated, the connection is dropped. If a new TLS send tag is allocated, however, subsequent packets will be tagged with the correct TLS send tag. (This latter case has been tested by configuring both ports of a Chelsio T6 in a lagg and failing over from one port to another. As the connections migrated to the new port, new TLS send tags were allocated for the new port and connections resumed without being dropped.) ifnet TLS can be enabled and disabled on supported network interfaces via new '[-]txtls[46]' options to ifconfig(8). ifnet TLS is supported across both vlan devices and lagg interfaces using failover, lacp with flowid enabled, or lacp with flowid enabled. Applications may request the current KTLS mode of a connection via a new TCP_TXTLS_MODE socket option. They can also use this socket option to toggle between software and ifnet TLS modes. In addition, a testing tool is available in tools/tools/switch_tls. This is modeled on tcpdrop and uses similar syntax. However, instead of dropping connections, -s is used to force KTLS connections to switch to software TLS and -i is used to switch to ifnet TLS. Various sysctls and counters are available under the kern.ipc.tls sysctl node. The kern.ipc.tls.enable node must be set to true to enable KTLS (it is off by default). The use of unmapped mbufs must also be enabled via kern.ipc.mb_use_ext_pgs to enable KTLS. KTLS is enabled via the KERN_TLS kernel option. This patch is the culmination of years of work by several folks including Scott Long and Randall Stewart for the original design and implementation; Drew Gallatin for several optimizations including the use of ext_pgs mbufs, the M_NOTREADY mechanism for TLS records awaiting software encryption, and pluggable software crypto backends; and John Baldwin for modifications to support hardware TLS offload. Reviewed by: gallatin, hselasky, rrs Obtained from: Netflix Sponsored by: Netflix, Chelsio Communications Differential Revision: https://reviews.freebsd.org/D21277
2019-08-27 00:01:56 +00:00
#include "opt_kern_tls.h"
Implement kernel support for hardware rate limited sockets. - Add RATELIMIT kernel configuration keyword which must be set to enable the new functionality. - Add support for hardware driven, Receive Side Scaling, RSS aware, rate limited sendqueues and expose the functionality through the already established SO_MAX_PACING_RATE setsockopt(). The API support rates in the range from 1 to 4Gbytes/s which are suitable for regular TCP and UDP streams. The setsockopt(2) manual page has been updated. - Add rate limit function callback API to "struct ifnet" which supports the following operations: if_snd_tag_alloc(), if_snd_tag_modify(), if_snd_tag_query() and if_snd_tag_free(). - Add support to ifconfig to view, set and clear the IFCAP_TXRTLMT flag, which tells if a network driver supports rate limiting or not. - This patch also adds support for rate limiting through VLAN and LAGG intermediate network devices. - How rate limiting works: 1) The userspace application calls setsockopt() after accepting or making a new connection to set the rate which is then stored in the socket structure in the kernel. Later on when packets are transmitted a check is made in the transmit path for rate changes. A rate change implies a non-blocking ifp->if_snd_tag_alloc() call will be made to the destination network interface, which then sets up a custom sendqueue with the given rate limitation parameter. A "struct m_snd_tag" pointer is returned which serves as a "snd_tag" hint in the m_pkthdr for the subsequently transmitted mbufs. 2) When the network driver sees the "m->m_pkthdr.snd_tag" different from NULL, it will move the packets into a designated rate limited sendqueue given by the snd_tag pointer. It is up to the individual drivers how the rate limited traffic will be rate limited. 3) Route changes are detected by the NIC drivers in the ifp->if_transmit() routine when the ifnet pointer in the incoming snd_tag mismatches the one of the network interface. The network adapter frees the mbuf and returns EAGAIN which causes the ip_output() to release and clear the send tag. Upon next ip_output() a new "snd_tag" will be tried allocated. 4) When the PCB is detached the custom sendqueue will be released by a non-blocking ifp->if_snd_tag_free() call to the currently bound network interface. Reviewed by: wblock (manpages), adrian, gallatin, scottl (network) Differential Revision: https://reviews.freebsd.org/D3687 Sponsored by: Mellanox Technologies MFC after: 3 months
2017-01-18 13:31:17 +00:00
#include "opt_ratelimit.h"
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
#include <sys/param.h>
#include <sys/kernel.h>
#include <sys/malloc.h>
#include <sys/mbuf.h>
#include <sys/queue.h>
#include <sys/socket.h>
#include <sys/sockio.h>
#include <sys/sysctl.h>
#include <sys/module.h>
#include <sys/priv.h>
#include <sys/systm.h>
#include <sys/proc.h>
Change from a mutex to a read/write lock. This allows the tx port to be selected simultaneously by multiple senders and transmit/receive is not serialised between aggregated interfaces.
2007-05-15 07:41:46 +00:00
#include <sys/lock.h>
Convert the if_lagg rwlock to an rmlock. We've been seeing lots of cache line contention (but not lock contention!) in our workloads between the various TX and RX threads going on. The write lock is only grabbed when configuration changes are made - which are infrequent. With this patch, the contention and cycles spent waiting for updates disappear. Sponsored by: Netflix, Inc.
2013-08-29 19:35:14 +00:00
#include <sys/rmlock.h>
Introduce sleepable locks into if_lagg. Before this change if_lagg was using nonsleepable rmlocks to protect its internal state. This patch introduces another sx lock to protect code paths that require sleeping, while still uses old rmlock to protect hot nonsleepable data paths. This change allows to remove taskqueue decoupling used before to change interface addresses without holding the lock. Instead it uses sx lock to protect direct if_ioctl() calls. As another bonus, the new code synchronizes enabled capabilities of member interfaces, and allows to control them with ifconfig laggX, that was impossible before. This part should fix interoperation with if_bridge, that may need to disable some capabilities, such as TXCSUM or LRO, to allow bridging with noncapable interfaces. MFC after: 2 weeks Sponsored by: iXsystems, Inc. Differential Revision: https://reviews.freebsd.org/D10514
2017-05-02 19:09:11 +00:00
#include <sys/sx.h>
Call if_setlladdr() on the aggregation port from a taskqueue so the softc lock is not held. The short delay between aggregating the port and setting the MAC address is fine.
2007-05-07 00:35:15 +00:00
#include <sys/taskqueue.h>
Propagate the vlan eventis to the underlying interfaces/members so they can do initialization of hw related features. PR: kern/141646 Reviewed by: thompsa Approved by: thompsa(co-mentor) MFC after: 2 weeks
2010-02-06 13:49:35 +00:00
#include <sys/eventhandler.h>
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
#include <net/ethernet.h>
#include <net/if.h>
#include <net/if_clone.h>
#include <net/if_arp.h>
#include <net/if_dl.h>
#include <net/if_media.h>
#include <net/if_types.h>
#include <net/if_var.h>
#include <net/bpf.h>
Update route MTUs for bridge, lagg and vlan interfaces. Those interfaces may implicitly change their MTU on addition of parent interface in addition to normal SIOCSIFMTU ioctl path, where the route MTUs are updated normally. MFC after: 2 weeks Sponsored by: iXsystems, Inc.
2020-01-22 20:36:45 +00:00
#include <net/route.h>
Virtualize lagg(4) cloner. This change fixes a panic when tearing down if_lagg(4) interfaces which were cloned in a vnet jail. Sysctl nodes which are dynamically generated for each cloned interface (net.link.lagg.N.*) have been removed, and use_flowid and flowid_shift ifconfig(8) parameters have been added instead. Flags and per-interface statistics counters are displayed in "ifconfig -v". CR: D842
2014-10-01 21:37:32 +00:00
#include <net/vnet.h>
Add support for IP over infiniband, IPoIB, to lagg(4). Currently only the failover protocol is supported due to limitations in the IPoIB architecture. Refer to the lagg(4) manual page for how to configure and use this new feature. A new network interface type, IFT_INFINIBANDLAG, has been added, similar to the existing IFT_IEEE8023ADLAG . ifconfig(8) has been updated to accept a new laggtype argument when creating lagg(4) network interfaces. This new argument is used to distinguish between ethernet and infiniband type of lagg(4) network interface. The laggtype argument is optional and defaults to ethernet. The lagg(4) command line syntax is backwards compatible. Differential Revision: https://reviews.freebsd.org/D26254 Reviewed by: melifaro@ MFC after: 1 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2020-10-22 09:47:12 +00:00
#include <net/infiniband.h>
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
Make various (pseudo) interfaces compile without INET in the kernel adding appropriate #ifdefs. For module builds the framework needs adjustments for at least carp. Reviewed by: gnn Sponsored by: The FreeBSD Foundation Sponsored by: iXsystems MFC after: 4 days
2011-04-27 19:30:44 +00:00
#if defined(INET) || defined(INET6)
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
#include <netinet/in.h>
Fix build for non-INET that was broken by r264469. MFC after: 2 weeks
2014-04-15 13:28:54 +00:00
#include <netinet/ip.h>
Make various (pseudo) interfaces compile without INET in the kernel adding appropriate #ifdefs. For module builds the framework needs adjustments for at least carp. Reviewed by: gnn Sponsored by: The FreeBSD Foundation Sponsored by: iXsystems MFC after: 4 days
2011-04-27 19:30:44 +00:00
#endif
#ifdef INET
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
#include <netinet/in_systm.h>
#include <netinet/if_ether.h>
#endif
#ifdef INET6
#include <netinet/ip6.h>
- Allow ND6_IFF_AUTO_LINKLOCAL for IFT_BRIDGE. An interface with IFT_BRIDGE is initialized with !ND6_IFF_AUTO_LINKLOCAL && !ND6_IFF_ACCEPT_RTADV regardless of net.inet6.ip6.accept_rtadv and net.inet6.ip6.auto_linklocal. To configure an autoconfigured link-local address (RFC 4862), the following rc.conf(5) configuration can be used: ifconfig_bridge0_ipv6="inet6 auto_linklocal" - if_bridge(4) now removes IPv6 addresses on a member interface to be added when the parent interface or one of the existing member interfaces has an IPv6 address. if_bridge(4) merges each link-local scope zone which the member interfaces form respectively, so it causes address scope violation. Removal of the IPv6 addresses prevents it. - if_lagg(4) now removes IPv6 addresses on a member interfaces unconditionally. - Set reasonable flags to non-IPv6-capable interfaces. [*] Submitted by: rpaulo [*] MFC after: 1 week
2013-07-02 16:58:15 +00:00
#include <netinet6/in6_var.h>
#include <netinet6/in6_ifattach.h>
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
#endif
#include <net/if_vlan_var.h>
#include <net/if_lagg.h>
#include <net/ieee8023ad_lacp.h>
Update route MTUs for bridge, lagg and vlan interfaces. Those interfaces may implicitly change their MTU on addition of parent interface in addition to normal SIOCSIFMTU ioctl path, where the route MTUs are updated normally. MFC after: 2 weeks Sponsored by: iXsystems, Inc.
2020-01-22 20:36:45 +00:00
#ifdef INET6
/*
* XXX: declare here to avoid to include many inet6 related files..
* should be more generalized?
*/
extern void nd6_setmtu(struct ifnet *);
#endif
Replace rmlock with epoch in lagg Use the new epoch based reclamation API. Now the hot paths will not block at all, and the sx lock is used for the softc data. This fixes LORs reported where the rwlock was obtained when the sxlock was held. Submitted by: mmacy Reported by: Harry Schmalzbauer <freebsd@omnilan.de> Reviewed by: sbruno Sponsored by: Limelight Networks Differential Revision: https://reviews.freebsd.org/D15355
2018-05-14 20:06:49 +00:00
#define LAGG_SX_INIT(_sc) sx_init(&(_sc)->sc_sx, "if_lagg sx")
#define LAGG_SX_DESTROY(_sc) sx_destroy(&(_sc)->sc_sx)
#define LAGG_XLOCK(_sc) sx_xlock(&(_sc)->sc_sx)
#define LAGG_XUNLOCK(_sc) sx_xunlock(&(_sc)->sc_sx)
#define LAGG_SXLOCK_ASSERT(_sc) sx_assert(&(_sc)->sc_sx, SA_LOCKED)
#define LAGG_XLOCK_ASSERT(_sc) sx_assert(&(_sc)->sc_sx, SA_XLOCKED)
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
/* Special flags we should propagate to the lagg ports. */
static struct {
int flag;
int (*func)(struct ifnet *, int);
} lagg_pflags[] = {
{IFF_PROMISC, ifpromisc},
{IFF_ALLMULTI, if_allmulti},
{0, NULL}
};
Restructure mbuf send tags to provide stronger guarantees. - Perform ifp mismatch checks (to determine if a send tag is allocated for a different ifp than the one the packet is being output on), in ip_output() and ip6_output(). This avoids sending packets with send tags to ifnet drivers that don't support send tags. Since we are now checking for ifp mismatches before invoking if_output, we can now try to allocate a new tag before invoking if_output sending the original packet on the new tag if allocation succeeds. To avoid code duplication for the fragment and unfragmented cases, add ip_output_send() and ip6_output_send() as wrappers around if_output and nd6_output_ifp, respectively. All of the logic for setting send tags and dealing with send tag-related errors is done in these wrapper functions. For pseudo interfaces that wrap other network interfaces (vlan and lagg), wrapper send tags are now allocated so that ip*_output see the wrapper ifp as the ifp in the send tag. The if_transmit routines rewrite the send tags after performing an ifp mismatch check. If an ifp mismatch is detected, the transmit routines fail with EAGAIN. - To provide clearer life cycle management of send tags, especially in the presence of vlan and lagg wrapper tags, add a reference count to send tags managed via m_snd_tag_ref() and m_snd_tag_rele(). Provide a helper function (m_snd_tag_init()) for use by drivers supporting send tags. m_snd_tag_init() takes care of the if_ref on the ifp meaning that code alloating send tags via if_snd_tag_alloc no longer has to manage that manually. Similarly, m_snd_tag_rele drops the refcount on the ifp after invoking if_snd_tag_free when the last reference to a send tag is dropped. This also closes use after free races if there are pending packets in driver tx rings after the socket is closed (e.g. from tcpdrop). In order for m_free to work reliably, add a new CSUM_SND_TAG flag in csum_flags to indicate 'snd_tag' is set (rather than 'rcvif'). Drivers now also check this flag instead of checking snd_tag against NULL. This avoids false positive matches when a forwarded packet has a non-NULL rcvif that was treated as a send tag. - cxgbe was relying on snd_tag_free being called when the inp was detached so that it could kick the firmware to flush any pending work on the flow. This is because the driver doesn't require ACK messages from the firmware for every request, but instead does a kind of manual interrupt coalescing by only setting a flag to request a completion on a subset of requests. If all of the in-flight requests don't have the flag when the tag is detached from the inp, the flow might never return the credits. The current snd_tag_free command issues a flush command to force the credits to return. However, the credit return is what also frees the mbufs, and since those mbufs now hold references on the tag, this meant that snd_tag_free would never be called. To fix, explicitly drop the mbuf's reference on the snd tag when the mbuf is queued in the firmware work queue. This means that once the inp's reference on the tag goes away and all in-flight mbufs have been queued to the firmware, tag's refcount will drop to zero and snd_tag_free will kick in and send the flush request. Note that we need to avoid doing this in the middle of ethofld_tx(), so the driver grabs a temporary reference on the tag around that loop to defer the free to the end of the function in case it sends the last mbuf to the queue after the inp has dropped its reference on the tag. - mlx5 preallocates send tags and was using the ifp pointer even when the send tag wasn't in use. Explicitly use the ifp from other data structures instead. - Sprinkle some assertions in various places to assert that received packets don't have a send tag, and that other places that overwrite rcvif (e.g. 802.11 transmit) don't clobber a send tag pointer. Reviewed by: gallatin, hselasky, rgrimes, ae Sponsored by: Netflix Differential Revision: https://reviews.freebsd.org/D20117
2019-05-24 22:30:40 +00:00
struct lagg_snd_tag {
struct m_snd_tag com;
struct m_snd_tag *tag;
};
Virtualize lagg(4) cloner. This change fixes a panic when tearing down if_lagg(4) interfaces which were cloned in a vnet jail. Sysctl nodes which are dynamically generated for each cloned interface (net.link.lagg.N.*) have been removed, and use_flowid and flowid_shift ifconfig(8) parameters have been added instead. Flags and per-interface statistics counters are displayed in "ifconfig -v". CR: D842
2014-10-01 21:37:32 +00:00
VNET_DEFINE(SLIST_HEAD(__trhead, lagg_softc), lagg_list); /* list of laggs */
#define V_lagg_list VNET(lagg_list)
Use the new VNET_DEFINE_STATIC macro when we are defining static VNET variables. Reviewed by: bz Sponsored by: DARPA, AFRL Differential Revision: https://reviews.freebsd.org/D16147
2018-07-24 16:35:52 +00:00
VNET_DEFINE_STATIC(struct mtx, lagg_list_mtx);
Virtualize lagg(4) cloner. This change fixes a panic when tearing down if_lagg(4) interfaces which were cloned in a vnet jail. Sysctl nodes which are dynamically generated for each cloned interface (net.link.lagg.N.*) have been removed, and use_flowid and flowid_shift ifconfig(8) parameters have been added instead. Flags and per-interface statistics counters are displayed in "ifconfig -v". CR: D842
2014-10-01 21:37:32 +00:00
#define V_lagg_list_mtx VNET(lagg_list_mtx)
#define LAGG_LIST_LOCK_INIT(x) mtx_init(&V_lagg_list_mtx, \
"if_lagg list", NULL, MTX_DEF)
#define LAGG_LIST_LOCK_DESTROY(x) mtx_destroy(&V_lagg_list_mtx)
#define LAGG_LIST_LOCK(x) mtx_lock(&V_lagg_list_mtx)
#define LAGG_LIST_UNLOCK(x) mtx_unlock(&V_lagg_list_mtx)
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
eventhandler_tag lagg_detach_cookie = NULL;
if_clone: migrate some consumers to the new KPI. Convert most of the cloner customers who require custom params to the new if_clone KPI. Reviewed by: kp Differential Revision: https://reviews.freebsd.org/D36636 MFC after: 2 weeks
2022-09-22 12:30:09 +00:00
static int lagg_clone_create(struct if_clone *, char *, size_t,
struct ifc_data *, struct ifnet **);
static int lagg_clone_destroy(struct if_clone *, struct ifnet *, uint32_t);
Use the new VNET_DEFINE_STATIC macro when we are defining static VNET variables. Reviewed by: bz Sponsored by: DARPA, AFRL Differential Revision: https://reviews.freebsd.org/D16147
2018-07-24 16:35:52 +00:00
VNET_DEFINE_STATIC(struct if_clone *, lagg_cloner);
Virtualize lagg(4) cloner. This change fixes a panic when tearing down if_lagg(4) interfaces which were cloned in a vnet jail. Sysctl nodes which are dynamically generated for each cloned interface (net.link.lagg.N.*) have been removed, and use_flowid and flowid_shift ifconfig(8) parameters have been added instead. Flags and per-interface statistics counters are displayed in "ifconfig -v". CR: D842
2014-10-01 21:37:32 +00:00
#define V_lagg_cloner VNET(lagg_cloner)
Make the "struct if_clone" opaque to users of the cloning API. Users now use function calls: if_clone_simple() if_clone_advanced() to initialize a cloner, instead of macros that initialize if_clone structure. Discussed with: brooks, bz, 1 year ago
2012-10-16 13:37:54 +00:00
static const char laggname[] = "lagg";
Use a dedicated malloc type for lagg(4)'s structures. Reviewed by: gallatin MFC after: 1 month Sponsored by: Netflix Differential Revision: https://reviews.freebsd.org/D19719
2019-03-28 21:00:54 +00:00
static MALLOC_DEFINE(M_LAGG, laggname, "802.3AD Link Aggregation Interface");
Make the "struct if_clone" opaque to users of the cloning API. Users now use function calls: if_clone_simple() if_clone_advanced() to initialize a cloner, instead of macros that initialize if_clone structure. Discussed with: brooks, bz, 1 year ago
2012-10-16 13:37:54 +00:00
- Propagate the largest set of interface capabilities supported by all lagg ports to the lagg interface. - Use the MTU from the first interface as the lagg MTU, all extra interfaces must be the same. This fixes using a lagg interface for a vlan or enabling jumbo frames, etc. Approved by: re (kensmith) MFC After: 3 days
2007-07-30 20:17:22 +00:00
static void lagg_capabilities(struct lagg_softc *);
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
static int lagg_port_create(struct lagg_softc *, struct ifnet *);
static int lagg_port_destroy(struct lagg_port *, int);
Add support for IP over infiniband, IPoIB, to lagg(4). Currently only the failover protocol is supported due to limitations in the IPoIB architecture. Refer to the lagg(4) manual page for how to configure and use this new feature. A new network interface type, IFT_INFINIBANDLAG, has been added, similar to the existing IFT_IEEE8023ADLAG . ifconfig(8) has been updated to accept a new laggtype argument when creating lagg(4) network interfaces. This new argument is used to distinguish between ethernet and infiniband type of lagg(4) network interface. The laggtype argument is optional and defaults to ethernet. The lagg(4) command line syntax is backwards compatible. Differential Revision: https://reviews.freebsd.org/D26254 Reviewed by: melifaro@ MFC after: 1 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2020-10-22 09:47:12 +00:00
static struct mbuf *lagg_input_ethernet(struct ifnet *, struct mbuf *);
static struct mbuf *lagg_input_infiniband(struct ifnet *, struct mbuf *);
Have the lagg interface generate link up/down events, the interface is marked as up if at least one of its ports also has a link up. This fixes using carp+lagg together and any other system that relies on linkstate events. PR: kern/113956 MFC after: 3 days
2007-11-25 06:30:46 +00:00
static void lagg_linkstate(struct lagg_softc *);
Revert r292275 & r292379 glebius has concerns about these changes so reverting those can be discussed and addressed. Sponsored by: Multiplay
2015-12-17 14:41:30 +00:00
static void lagg_port_state(struct ifnet *, int);
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
static int lagg_port_ioctl(struct ifnet *, u_long, caddr_t);
static int lagg_port_output(struct ifnet *, struct mbuf *,
Add const qualifier to the dst parameter of the ifnet if_output method.
2013-04-26 12:50:32 +00:00
const struct sockaddr *, struct route *);
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
static void lagg_port_ifdetach(void *arg __unused, struct ifnet *);
Stop GCC from complaining about lagg_port_checkstacking() being unused.
2010-01-08 16:44:33 +00:00
#ifdef LAGG_PORT_STACKING
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
static int lagg_port_checkstacking(struct lagg_softc *);
Stop GCC from complaining about lagg_port_checkstacking() being unused.
2010-01-08 16:44:33 +00:00
#endif
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
static void lagg_port2req(struct lagg_port *, struct lagg_reqport *);
static void lagg_init(void *);
static void lagg_stop(struct lagg_softc *);
static int lagg_ioctl(struct ifnet *, u_long, caddr_t);
Add kernel-side support for in-kernel TLS. KTLS adds support for in-kernel framing and encryption of Transport Layer Security (1.0-1.2) data on TCP sockets. KTLS only supports offload of TLS for transmitted data. Key negotation must still be performed in userland. Once completed, transmit session keys for a connection are provided to the kernel via a new TCP_TXTLS_ENABLE socket option. All subsequent data transmitted on the socket is placed into TLS frames and encrypted using the supplied keys. Any data written to a KTLS-enabled socket via write(2), aio_write(2), or sendfile(2) is assumed to be application data and is encoded in TLS frames with an application data type. Individual records can be sent with a custom type (e.g. handshake messages) via sendmsg(2) with a new control message (TLS_SET_RECORD_TYPE) specifying the record type. At present, rekeying is not supported though the in-kernel framework should support rekeying. KTLS makes use of the recently added unmapped mbufs to store TLS frames in the socket buffer. Each TLS frame is described by a single ext_pgs mbuf. The ext_pgs structure contains the header of the TLS record (and trailer for encrypted records) as well as references to the associated TLS session. KTLS supports two primary methods of encrypting TLS frames: software TLS and ifnet TLS. Software TLS marks mbufs holding socket data as not ready via M_NOTREADY similar to sendfile(2) when TLS framing information is added to an unmapped mbuf in ktls_frame(). ktls_enqueue() is then called to schedule TLS frames for encryption. In the case of sendfile_iodone() calls ktls_enqueue() instead of pru_ready() leaving the mbufs marked M_NOTREADY until encryption is completed. For other writes (vn_sendfile when pages are available, write(2), etc.), the PRUS_NOTREADY is set when invoking pru_send() along with invoking ktls_enqueue(). A pool of worker threads (the "KTLS" kernel process) encrypts TLS frames queued via ktls_enqueue(). Each TLS frame is temporarily mapped using the direct map and passed to a software encryption backend to perform the actual encryption. (Note: The use of PHYS_TO_DMAP could be replaced with sf_bufs if someone wished to make this work on architectures without a direct map.) KTLS supports pluggable software encryption backends. Internally, Netflix uses proprietary pure-software backends. This commit includes a simple backend in a new ktls_ocf.ko module that uses the kernel's OpenCrypto framework to provide AES-GCM encryption of TLS frames. As a result, software TLS is now a bit of a misnomer as it can make use of hardware crypto accelerators. Once software encryption has finished, the TLS frame mbufs are marked ready via pru_ready(). At this point, the encrypted data appears as regular payload to the TCP stack stored in unmapped mbufs. ifnet TLS permits a NIC to offload the TLS encryption and TCP segmentation. In this mode, a new send tag type (IF_SND_TAG_TYPE_TLS) is allocated on the interface a socket is routed over and associated with a TLS session. TLS records for a TLS session using ifnet TLS are not marked M_NOTREADY but are passed down the stack unencrypted. The ip_output_send() and ip6_output_send() helper functions that apply send tags to outbound IP packets verify that the send tag of the TLS record matches the outbound interface. If so, the packet is tagged with the TLS send tag and sent to the interface. The NIC device driver must recognize packets with the TLS send tag and schedule them for TLS encryption and TCP segmentation. If the the outbound interface does not match the interface in the TLS send tag, the packet is dropped. In addition, a task is scheduled to refresh the TLS send tag for the TLS session. If a new TLS send tag cannot be allocated, the connection is dropped. If a new TLS send tag is allocated, however, subsequent packets will be tagged with the correct TLS send tag. (This latter case has been tested by configuring both ports of a Chelsio T6 in a lagg and failing over from one port to another. As the connections migrated to the new port, new TLS send tags were allocated for the new port and connections resumed without being dropped.) ifnet TLS can be enabled and disabled on supported network interfaces via new '[-]txtls[46]' options to ifconfig(8). ifnet TLS is supported across both vlan devices and lagg interfaces using failover, lacp with flowid enabled, or lacp with flowid enabled. Applications may request the current KTLS mode of a connection via a new TCP_TXTLS_MODE socket option. They can also use this socket option to toggle between software and ifnet TLS modes. In addition, a testing tool is available in tools/tools/switch_tls. This is modeled on tcpdrop and uses similar syntax. However, instead of dropping connections, -s is used to force KTLS connections to switch to software TLS and -i is used to switch to ifnet TLS. Various sysctls and counters are available under the kern.ipc.tls sysctl node. The kern.ipc.tls.enable node must be set to true to enable KTLS (it is off by default). The use of unmapped mbufs must also be enabled via kern.ipc.mb_use_ext_pgs to enable KTLS. KTLS is enabled via the KERN_TLS kernel option. This patch is the culmination of years of work by several folks including Scott Long and Randall Stewart for the original design and implementation; Drew Gallatin for several optimizations including the use of ext_pgs mbufs, the M_NOTREADY mechanism for TLS records awaiting software encryption, and pluggable software crypto backends; and John Baldwin for modifications to support hardware TLS offload. Reviewed by: gallatin, hselasky, rrs Obtained from: Netflix Sponsored by: Netflix, Chelsio Communications Differential Revision: https://reviews.freebsd.org/D21277
2019-08-27 00:01:56 +00:00
#if defined(KERN_TLS) || defined(RATELIMIT)
Implement kernel support for hardware rate limited sockets. - Add RATELIMIT kernel configuration keyword which must be set to enable the new functionality. - Add support for hardware driven, Receive Side Scaling, RSS aware, rate limited sendqueues and expose the functionality through the already established SO_MAX_PACING_RATE setsockopt(). The API support rates in the range from 1 to 4Gbytes/s which are suitable for regular TCP and UDP streams. The setsockopt(2) manual page has been updated. - Add rate limit function callback API to "struct ifnet" which supports the following operations: if_snd_tag_alloc(), if_snd_tag_modify(), if_snd_tag_query() and if_snd_tag_free(). - Add support to ifconfig to view, set and clear the IFCAP_TXRTLMT flag, which tells if a network driver supports rate limiting or not. - This patch also adds support for rate limiting through VLAN and LAGG intermediate network devices. - How rate limiting works: 1) The userspace application calls setsockopt() after accepting or making a new connection to set the rate which is then stored in the socket structure in the kernel. Later on when packets are transmitted a check is made in the transmit path for rate changes. A rate change implies a non-blocking ifp->if_snd_tag_alloc() call will be made to the destination network interface, which then sets up a custom sendqueue with the given rate limitation parameter. A "struct m_snd_tag" pointer is returned which serves as a "snd_tag" hint in the m_pkthdr for the subsequently transmitted mbufs. 2) When the network driver sees the "m->m_pkthdr.snd_tag" different from NULL, it will move the packets into a designated rate limited sendqueue given by the snd_tag pointer. It is up to the individual drivers how the rate limited traffic will be rate limited. 3) Route changes are detected by the NIC drivers in the ifp->if_transmit() routine when the ifnet pointer in the incoming snd_tag mismatches the one of the network interface. The network adapter frees the mbuf and returns EAGAIN which causes the ip_output() to release and clear the send tag. Upon next ip_output() a new "snd_tag" will be tried allocated. 4) When the PCB is detached the custom sendqueue will be released by a non-blocking ifp->if_snd_tag_free() call to the currently bound network interface. Reviewed by: wblock (manpages), adrian, gallatin, scottl (network) Differential Revision: https://reviews.freebsd.org/D3687 Sponsored by: Mellanox Technologies MFC after: 3 months
2017-01-18 13:31:17 +00:00
static int lagg_snd_tag_alloc(struct ifnet *,
union if_snd_tag_alloc_params *,
struct m_snd_tag **);
Restructure mbuf send tags to provide stronger guarantees. - Perform ifp mismatch checks (to determine if a send tag is allocated for a different ifp than the one the packet is being output on), in ip_output() and ip6_output(). This avoids sending packets with send tags to ifnet drivers that don't support send tags. Since we are now checking for ifp mismatches before invoking if_output, we can now try to allocate a new tag before invoking if_output sending the original packet on the new tag if allocation succeeds. To avoid code duplication for the fragment and unfragmented cases, add ip_output_send() and ip6_output_send() as wrappers around if_output and nd6_output_ifp, respectively. All of the logic for setting send tags and dealing with send tag-related errors is done in these wrapper functions. For pseudo interfaces that wrap other network interfaces (vlan and lagg), wrapper send tags are now allocated so that ip*_output see the wrapper ifp as the ifp in the send tag. The if_transmit routines rewrite the send tags after performing an ifp mismatch check. If an ifp mismatch is detected, the transmit routines fail with EAGAIN. - To provide clearer life cycle management of send tags, especially in the presence of vlan and lagg wrapper tags, add a reference count to send tags managed via m_snd_tag_ref() and m_snd_tag_rele(). Provide a helper function (m_snd_tag_init()) for use by drivers supporting send tags. m_snd_tag_init() takes care of the if_ref on the ifp meaning that code alloating send tags via if_snd_tag_alloc no longer has to manage that manually. Similarly, m_snd_tag_rele drops the refcount on the ifp after invoking if_snd_tag_free when the last reference to a send tag is dropped. This also closes use after free races if there are pending packets in driver tx rings after the socket is closed (e.g. from tcpdrop). In order for m_free to work reliably, add a new CSUM_SND_TAG flag in csum_flags to indicate 'snd_tag' is set (rather than 'rcvif'). Drivers now also check this flag instead of checking snd_tag against NULL. This avoids false positive matches when a forwarded packet has a non-NULL rcvif that was treated as a send tag. - cxgbe was relying on snd_tag_free being called when the inp was detached so that it could kick the firmware to flush any pending work on the flow. This is because the driver doesn't require ACK messages from the firmware for every request, but instead does a kind of manual interrupt coalescing by only setting a flag to request a completion on a subset of requests. If all of the in-flight requests don't have the flag when the tag is detached from the inp, the flow might never return the credits. The current snd_tag_free command issues a flush command to force the credits to return. However, the credit return is what also frees the mbufs, and since those mbufs now hold references on the tag, this meant that snd_tag_free would never be called. To fix, explicitly drop the mbuf's reference on the snd tag when the mbuf is queued in the firmware work queue. This means that once the inp's reference on the tag goes away and all in-flight mbufs have been queued to the firmware, tag's refcount will drop to zero and snd_tag_free will kick in and send the flush request. Note that we need to avoid doing this in the middle of ethofld_tx(), so the driver grabs a temporary reference on the tag around that loop to defer the free to the end of the function in case it sends the last mbuf to the queue after the inp has dropped its reference on the tag. - mlx5 preallocates send tags and was using the ifp pointer even when the send tag wasn't in use. Explicitly use the ifp from other data structures instead. - Sprinkle some assertions in various places to assert that received packets don't have a send tag, and that other places that overwrite rcvif (e.g. 802.11 transmit) don't clobber a send tag pointer. Reviewed by: gallatin, hselasky, rgrimes, ae Sponsored by: Netflix Differential Revision: https://reviews.freebsd.org/D20117
2019-05-24 22:30:40 +00:00
static int lagg_snd_tag_modify(struct m_snd_tag *,
union if_snd_tag_modify_params *);
static int lagg_snd_tag_query(struct m_snd_tag *,
union if_snd_tag_query_params *);
This commit adds the missing release mechanism for the ratelimiting code. The two modules (lagg and vlan) did have allocation routines, and even though they are indirect (and vector down to the underlying interfaces) they both need to have a free routine (that also vectors down to the actual interface). Sponsored by: Netflix Inc. Differential Revision: https://reviews.freebsd.org/D19032
2019-02-13 14:57:59 +00:00
static void lagg_snd_tag_free(struct m_snd_tag *);
This pulls over all the changes that are in the netflix tree that fix the ratelimit code. There were several bugs in tcp_ratelimit itself and we needed further work to support the multiple tag format coming for the joint TLS and Ratelimit dances. Sponsored by: Netflix Inc. Differential Revision: https://reviews.freebsd.org/D28357
2021-01-26 16:54:42 +00:00
static struct m_snd_tag *lagg_next_snd_tag(struct m_snd_tag *);
This adds the third step in getting BBR into the tree. BBR and an updated rack depend on having access to the new ratelimit api in this commit. Sponsored by: Netflix Inc. Differential Revision: https://reviews.freebsd.org/D20953
2019-08-01 14:17:31 +00:00
static void lagg_ratelimit_query(struct ifnet *,
struct if_ratelimit_query_results *);
Implement kernel support for hardware rate limited sockets. - Add RATELIMIT kernel configuration keyword which must be set to enable the new functionality. - Add support for hardware driven, Receive Side Scaling, RSS aware, rate limited sendqueues and expose the functionality through the already established SO_MAX_PACING_RATE setsockopt(). The API support rates in the range from 1 to 4Gbytes/s which are suitable for regular TCP and UDP streams. The setsockopt(2) manual page has been updated. - Add rate limit function callback API to "struct ifnet" which supports the following operations: if_snd_tag_alloc(), if_snd_tag_modify(), if_snd_tag_query() and if_snd_tag_free(). - Add support to ifconfig to view, set and clear the IFCAP_TXRTLMT flag, which tells if a network driver supports rate limiting or not. - This patch also adds support for rate limiting through VLAN and LAGG intermediate network devices. - How rate limiting works: 1) The userspace application calls setsockopt() after accepting or making a new connection to set the rate which is then stored in the socket structure in the kernel. Later on when packets are transmitted a check is made in the transmit path for rate changes. A rate change implies a non-blocking ifp->if_snd_tag_alloc() call will be made to the destination network interface, which then sets up a custom sendqueue with the given rate limitation parameter. A "struct m_snd_tag" pointer is returned which serves as a "snd_tag" hint in the m_pkthdr for the subsequently transmitted mbufs. 2) When the network driver sees the "m->m_pkthdr.snd_tag" different from NULL, it will move the packets into a designated rate limited sendqueue given by the snd_tag pointer. It is up to the individual drivers how the rate limited traffic will be rate limited. 3) Route changes are detected by the NIC drivers in the ifp->if_transmit() routine when the ifnet pointer in the incoming snd_tag mismatches the one of the network interface. The network adapter frees the mbuf and returns EAGAIN which causes the ip_output() to release and clear the send tag. Upon next ip_output() a new "snd_tag" will be tried allocated. 4) When the PCB is detached the custom sendqueue will be released by a non-blocking ifp->if_snd_tag_free() call to the currently bound network interface. Reviewed by: wblock (manpages), adrian, gallatin, scottl (network) Differential Revision: https://reviews.freebsd.org/D3687 Sponsored by: Mellanox Technologies MFC after: 3 months
2017-01-18 13:31:17 +00:00
#endif
Introduce sleepable locks into if_lagg. Before this change if_lagg was using nonsleepable rmlocks to protect its internal state. This patch introduces another sx lock to protect code paths that require sleeping, while still uses old rmlock to protect hot nonsleepable data paths. This change allows to remove taskqueue decoupling used before to change interface addresses without holding the lock. Instead it uses sx lock to protect direct if_ioctl() calls. As another bonus, the new code synchronizes enabled capabilities of member interfaces, and allows to control them with ifconfig laggX, that was impossible before. This part should fix interoperation with if_bridge, that may need to disable some capabilities, such as TXCSUM or LRO, to allow bridging with noncapable interfaces. MFC after: 2 weeks Sponsored by: iXsystems, Inc. Differential Revision: https://reviews.freebsd.org/D10514
2017-05-02 19:09:11 +00:00
static int lagg_setmulti(struct lagg_port *);
static int lagg_clrmulti(struct lagg_port *);
lagg: fix lagg ifioctl after SIOCSIFCAPNV Lagg was broken by SIOCSIFCAPNV when all underlying devices support SIOCSIFCAPNV. This change updates lagg to work with SIOCSIFCAPNV and if_capabilities2. Reviewed by: kib, hselasky Sponsored by: Netflix Differential Revision: https://reviews.freebsd.org/D35865
2022-07-28 14:36:22 +00:00
static void lagg_setcaps(struct lagg_port *, int cap, int cap2);
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
static int lagg_setflag(struct lagg_port *, int, int,
int (*func)(struct ifnet *, int));
static int lagg_setflags(struct lagg_port *, int status);
Use underlying ports counters to get lagg statistics instead of per-packet accounting. This introduce user-visible changes like aggregating error counters. Reviewed by: asomers (prev.version), glebius CR: D781 MFC after: 2 weeks Sponsored by: Yandex LLC
2014-09-27 13:57:48 +00:00
static uint64_t lagg_get_counter(struct ifnet *ifp, ift_counter cnt);
Add support for IP over infiniband, IPoIB, to lagg(4). Currently only the failover protocol is supported due to limitations in the IPoIB architecture. Refer to the lagg(4) manual page for how to configure and use this new feature. A new network interface type, IFT_INFINIBANDLAG, has been added, similar to the existing IFT_IEEE8023ADLAG . ifconfig(8) has been updated to accept a new laggtype argument when creating lagg(4) network interfaces. This new argument is used to distinguish between ethernet and infiniband type of lagg(4) network interface. The laggtype argument is optional and defaults to ethernet. The lagg(4) command line syntax is backwards compatible. Differential Revision: https://reviews.freebsd.org/D26254 Reviewed by: melifaro@ MFC after: 1 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2020-10-22 09:47:12 +00:00
static int lagg_transmit_ethernet(struct ifnet *, struct mbuf *);
static int lagg_transmit_infiniband(struct ifnet *, struct mbuf *);
Convert lagg(4) to use if_transmit instead of if_start. In collaboration with: thompsa, sbruno, fabient
2012-09-20 10:05:10 +00:00
static void lagg_qflush(struct ifnet *);
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
static int lagg_media_change(struct ifnet *);
static void lagg_media_status(struct ifnet *, struct ifmediareq *);
static struct lagg_port *lagg_link_active(struct lagg_softc *,
struct lagg_port *);
/* Simple round robin */
- When reconfiguring protocol on a lagg, first set it to LAGG_PROTO_NONE, then drop lock, run the attach routines, and then set it to specific proto. This removes tons of WITNESS warnings. - Make lagg protocol attach handlers not failing and allocate memory with M_WAITOK. Sponsored by: Netflix Sponsored by: Nginx, Inc.
2014-09-26 08:42:32 +00:00
static void lagg_rr_attach(struct lagg_softc *);
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
static int lagg_rr_start(struct lagg_softc *, struct mbuf *);
static struct mbuf *lagg_rr_input(struct lagg_softc *, struct lagg_port *,
struct mbuf *);
/* Active failover */
static int lagg_fail_start(struct lagg_softc *, struct mbuf *);
static struct mbuf *lagg_fail_input(struct lagg_softc *, struct lagg_port *,
struct mbuf *);
/* Loadbalancing */
- When reconfiguring protocol on a lagg, first set it to LAGG_PROTO_NONE, then drop lock, run the attach routines, and then set it to specific proto. This removes tons of WITNESS warnings. - Make lagg protocol attach handlers not failing and allocate memory with M_WAITOK. Sponsored by: Netflix Sponsored by: Nginx, Inc.
2014-09-26 08:42:32 +00:00
static void lagg_lb_attach(struct lagg_softc *);
Make lagg protocols detach methods returning void. Sponsored by: Netflix Sponsored by: Nginx, Inc.
2014-09-26 07:12:40 +00:00
static void lagg_lb_detach(struct lagg_softc *);
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
static int lagg_lb_port_create(struct lagg_port *);
static void lagg_lb_port_destroy(struct lagg_port *);
static int lagg_lb_start(struct lagg_softc *, struct mbuf *);
static struct mbuf *lagg_lb_input(struct lagg_softc *, struct lagg_port *,
struct mbuf *);
static int lagg_lb_porttable(struct lagg_softc *, struct lagg_port *);
Add laggproto broadcast, it allows sends frames to all ports of the lagg(4) group and receives frames on any port of the lagg(4). Phabric: D549 Reviewed by: glebius, thompsa Approved by: glebius Obtained from: OpenBSD Sponsored by: QNAP Systems Inc.
2014-09-18 02:12:48 +00:00
/* Broadcast */
static int lagg_bcast_start(struct lagg_softc *, struct mbuf *);
static struct mbuf *lagg_bcast_input(struct lagg_softc *, struct lagg_port *,
- Whitespace. - Remove caddr_t.
2014-09-26 12:35:58 +00:00
struct mbuf *);
Add laggproto broadcast, it allows sends frames to all ports of the lagg(4) group and receives frames on any port of the lagg(4). Phabric: D549 Reviewed by: glebius, thompsa Approved by: glebius Obtained from: OpenBSD Sponsored by: QNAP Systems Inc.
2014-09-18 02:12:48 +00:00
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
/* 802.3ad LACP */
- When reconfiguring protocol on a lagg, first set it to LAGG_PROTO_NONE, then drop lock, run the attach routines, and then set it to specific proto. This removes tons of WITNESS warnings. - Make lagg protocol attach handlers not failing and allocate memory with M_WAITOK. Sponsored by: Netflix Sponsored by: Nginx, Inc.
2014-09-26 08:42:32 +00:00
static void lagg_lacp_attach(struct lagg_softc *);
Make lagg protocols detach methods returning void. Sponsored by: Netflix Sponsored by: Nginx, Inc.
2014-09-26 07:12:40 +00:00
static void lagg_lacp_detach(struct lagg_softc *);
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
static int lagg_lacp_start(struct lagg_softc *, struct mbuf *);
static struct mbuf *lagg_lacp_input(struct lagg_softc *, struct lagg_port *,
struct mbuf *);
static void lagg_lacp_lladdr(struct lagg_softc *);
/* lagg protocol table */
- When reconfiguring protocol on a lagg, first set it to LAGG_PROTO_NONE, then drop lock, run the attach routines, and then set it to specific proto. This removes tons of WITNESS warnings. - Make lagg protocol attach handlers not failing and allocate memory with M_WAITOK. Sponsored by: Netflix Sponsored by: Nginx, Inc.
2014-09-26 08:42:32 +00:00
static const struct lagg_proto {
- Provide lagg_proto_attach(), lagg_proto_detach(). - Make detach a protocol method in lagg_protos. - Simplify code to lookup protocols. Sponsored by: Netflix Sponsored by: Nginx, Inc.
2014-09-26 11:01:04 +00:00
lagg_proto pr_num;
void (*pr_attach)(struct lagg_softc *);
void (*pr_detach)(struct lagg_softc *);
Make all lagg protocol methods live in lagg_protos, not in softc. All interfaces of a same protocol, use the same methods. Sponsored by: Netflix Sponsored by: Nginx, Inc.
2014-09-26 12:54:24 +00:00
int (*pr_start)(struct lagg_softc *, struct mbuf *);
struct mbuf * (*pr_input)(struct lagg_softc *, struct lagg_port *,
struct mbuf *);
int (*pr_addport)(struct lagg_port *);
void (*pr_delport)(struct lagg_port *);
void (*pr_linkstate)(struct lagg_port *);
void (*pr_init)(struct lagg_softc *);
void (*pr_stop)(struct lagg_softc *);
void (*pr_lladdr)(struct lagg_softc *);
void (*pr_request)(struct lagg_softc *, void *);
void (*pr_portreq)(struct lagg_port *, void *);
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
} lagg_protos[] = {
- Provide lagg_proto_attach(), lagg_proto_detach(). - Make detach a protocol method in lagg_protos. - Simplify code to lookup protocols. Sponsored by: Netflix Sponsored by: Nginx, Inc.
2014-09-26 11:01:04 +00:00
{
.pr_num = LAGG_PROTO_NONE
},
{
.pr_num = LAGG_PROTO_ROUNDROBIN,
.pr_attach = lagg_rr_attach,
Make all lagg protocol methods live in lagg_protos, not in softc. All interfaces of a same protocol, use the same methods. Sponsored by: Netflix Sponsored by: Nginx, Inc.
2014-09-26 12:54:24 +00:00
.pr_start = lagg_rr_start,
.pr_input = lagg_rr_input,
- Provide lagg_proto_attach(), lagg_proto_detach(). - Make detach a protocol method in lagg_protos. - Simplify code to lookup protocols. Sponsored by: Netflix Sponsored by: Nginx, Inc.
2014-09-26 11:01:04 +00:00
},
{
.pr_num = LAGG_PROTO_FAILOVER,
Make all lagg protocol methods live in lagg_protos, not in softc. All interfaces of a same protocol, use the same methods. Sponsored by: Netflix Sponsored by: Nginx, Inc.
2014-09-26 12:54:24 +00:00
.pr_start = lagg_fail_start,
.pr_input = lagg_fail_input,
- Provide lagg_proto_attach(), lagg_proto_detach(). - Make detach a protocol method in lagg_protos. - Simplify code to lookup protocols. Sponsored by: Netflix Sponsored by: Nginx, Inc.
2014-09-26 11:01:04 +00:00
},
{
.pr_num = LAGG_PROTO_LOADBALANCE,
.pr_attach = lagg_lb_attach,
.pr_detach = lagg_lb_detach,
Make all lagg protocol methods live in lagg_protos, not in softc. All interfaces of a same protocol, use the same methods. Sponsored by: Netflix Sponsored by: Nginx, Inc.
2014-09-26 12:54:24 +00:00
.pr_start = lagg_lb_start,
.pr_input = lagg_lb_input,
.pr_addport = lagg_lb_port_create,
.pr_delport = lagg_lb_port_destroy,
- Provide lagg_proto_attach(), lagg_proto_detach(). - Make detach a protocol method in lagg_protos. - Simplify code to lookup protocols. Sponsored by: Netflix Sponsored by: Nginx, Inc.
2014-09-26 11:01:04 +00:00
},
{
.pr_num = LAGG_PROTO_LACP,
.pr_attach = lagg_lacp_attach,
.pr_detach = lagg_lacp_detach,
Make all lagg protocol methods live in lagg_protos, not in softc. All interfaces of a same protocol, use the same methods. Sponsored by: Netflix Sponsored by: Nginx, Inc.
2014-09-26 12:54:24 +00:00
.pr_start = lagg_lacp_start,
.pr_input = lagg_lacp_input,
.pr_addport = lacp_port_create,
.pr_delport = lacp_port_destroy,
.pr_linkstate = lacp_linkstate,
.pr_init = lacp_init,
.pr_stop = lacp_stop,
.pr_lladdr = lagg_lacp_lladdr,
.pr_request = lacp_req,
.pr_portreq = lacp_portreq,
- Provide lagg_proto_attach(), lagg_proto_detach(). - Make detach a protocol method in lagg_protos. - Simplify code to lookup protocols. Sponsored by: Netflix Sponsored by: Nginx, Inc.
2014-09-26 11:01:04 +00:00
},
{
.pr_num = LAGG_PROTO_BROADCAST,
Make all lagg protocol methods live in lagg_protos, not in softc. All interfaces of a same protocol, use the same methods. Sponsored by: Netflix Sponsored by: Nginx, Inc.
2014-09-26 12:54:24 +00:00
.pr_start = lagg_bcast_start,
.pr_input = lagg_bcast_input,
- Provide lagg_proto_attach(), lagg_proto_detach(). - Make detach a protocol method in lagg_protos. - Simplify code to lookup protocols. Sponsored by: Netflix Sponsored by: Nginx, Inc.
2014-09-26 11:01:04 +00:00
},
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
};
Add a sysctl knob to accept input packets on any link in a failover lagg.
2010-09-01 16:53:38 +00:00
SYSCTL_DECL(_net_link);
Mark more nodes as CTLFLAG_MPSAFE or CTLFLAG_NEEDGIANT (17 of many) r357614 added CTLFLAG_NEEDGIANT to make it easier to find nodes that are still not MPSAFE (or already are but aren’t properly marked). Use it in preparation for a general review of all nodes. This is non-functional change that adds annotations to SYSCTL_NODE and SYSCTL_PROC nodes using one of the soon-to-be-required flags. Mark all obvious cases as MPSAFE. All entries that haven't been marked as MPSAFE before are by default marked as NEEDGIANT Approved by: kib (mentor, blanket) Commented by: kib, gallatin, melifaro Differential Revision: https://reviews.freebsd.org/D23718
2020-02-26 14:26:36 +00:00
SYSCTL_NODE(_net_link, OID_AUTO, lagg, CTLFLAG_RW | CTLFLAG_MPSAFE, 0,
Mark all SYSCTL_NODEs static that have no corresponding SYSCTL_DECLs. The SYSCTL_NODE macro defines a list that stores all child-elements of that node. If there's no SYSCTL_DECL macro anywhere else, there's no reason why it shouldn't be static.
2011-11-07 15:43:11 +00:00
"Link Aggregation");
Add a sysctl knob to accept input packets on any link in a failover lagg.
2010-09-01 16:53:38 +00:00
Virtualize lagg(4) cloner. This change fixes a panic when tearing down if_lagg(4) interfaces which were cloned in a vnet jail. Sysctl nodes which are dynamically generated for each cloned interface (net.link.lagg.N.*) have been removed, and use_flowid and flowid_shift ifconfig(8) parameters have been added instead. Flags and per-interface statistics counters are displayed in "ifconfig -v". CR: D842
2014-10-01 21:37:32 +00:00
/* Allow input on any failover links */
Use the new VNET_DEFINE_STATIC macro when we are defining static VNET variables. Reviewed by: bz Sponsored by: DARPA, AFRL Differential Revision: https://reviews.freebsd.org/D16147
2018-07-24 16:35:52 +00:00
VNET_DEFINE_STATIC(int, lagg_failover_rx_all);
Virtualize lagg(4) cloner. This change fixes a panic when tearing down if_lagg(4) interfaces which were cloned in a vnet jail. Sysctl nodes which are dynamically generated for each cloned interface (net.link.lagg.N.*) have been removed, and use_flowid and flowid_shift ifconfig(8) parameters have been added instead. Flags and per-interface statistics counters are displayed in "ifconfig -v". CR: D842
2014-10-01 21:37:32 +00:00
#define V_lagg_failover_rx_all VNET(lagg_failover_rx_all)
SYSCTL_INT(_net_link_lagg, OID_AUTO, failover_rx_all, CTLFLAG_RW | CTLFLAG_VNET,
&VNET_NAME(lagg_failover_rx_all), 0,
Add a sysctl knob to accept input packets on any link in a failover lagg.
2010-09-01 16:53:38 +00:00
"Accept input from any interface in a failover lagg");
Virtualize lagg(4) cloner. This change fixes a panic when tearing down if_lagg(4) interfaces which were cloned in a vnet jail. Sysctl nodes which are dynamically generated for each cloned interface (net.link.lagg.N.*) have been removed, and use_flowid and flowid_shift ifconfig(8) parameters have been added instead. Flags and per-interface statistics counters are displayed in "ifconfig -v". CR: D842
2014-10-01 21:37:32 +00:00
Start process of removing the use of the deprecated "M_FLOWID" flag from the FreeBSD network code. The flag is still kept around in the "sys/mbuf.h" header file, but does no longer have any users. Instead the "m_pkthdr.rsstype" field in the mbuf structure is now used to decide the meaning of the "m_pkthdr.flowid" field. To modify the "m_pkthdr.rsstype" field please use the existing "M_HASHTYPE_XXX" macros as defined in the "sys/mbuf.h" header file. This patch introduces new behaviour in the transmit direction. Previously network drivers checked if "M_FLOWID" was set in "m_flags" before using the "m_pkthdr.flowid" field. This check has now now been replaced by checking if "M_HASHTYPE_GET(m)" is different from "M_HASHTYPE_NONE". In the future more hashtypes will be added, for example hashtypes for hardware dedicated flows. "M_HASHTYPE_OPAQUE" indicates that the "m_pkthdr.flowid" value is valid and has no particular type. This change removes the need for an "if" statement in TCP transmit code checking for the presence of a valid flowid value. The "if" statement mentioned above is now a direct variable assignment which is then later checked by the respective network drivers like before. Additional notes: - The SCTP code changes will be committed as a separate patch. - Removal of the "M_FLOWID" flag will also be done separately. - The FreeBSD version has been bumped. MFC after: 1 month Sponsored by: Mellanox Technologies
2014-12-01 11:45:24 +00:00
/* Default value for using flowid */
Use the new VNET_DEFINE_STATIC macro when we are defining static VNET variables. Reviewed by: bz Sponsored by: DARPA, AFRL Differential Revision: https://reviews.freebsd.org/D16147
2018-07-24 16:35:52 +00:00
VNET_DEFINE_STATIC(int, def_use_flowid) = 0;
Virtualize lagg(4) cloner. This change fixes a panic when tearing down if_lagg(4) interfaces which were cloned in a vnet jail. Sysctl nodes which are dynamically generated for each cloned interface (net.link.lagg.N.*) have been removed, and use_flowid and flowid_shift ifconfig(8) parameters have been added instead. Flags and per-interface statistics counters are displayed in "ifconfig -v". CR: D842
2014-10-01 21:37:32 +00:00
#define V_def_use_flowid VNET(def_use_flowid)
Pull in r267961 and r267973 again. Fix for issues reported will follow.
2014-06-28 03:56:17 +00:00
SYSCTL_INT(_net_link_lagg, OID_AUTO, default_use_flowid, CTLFLAG_RWTUN,
Virtualize lagg(4) cloner. This change fixes a panic when tearing down if_lagg(4) interfaces which were cloned in a vnet jail. Sysctl nodes which are dynamically generated for each cloned interface (net.link.lagg.N.*) have been removed, and use_flowid and flowid_shift ifconfig(8) parameters have been added instead. Flags and per-interface statistics counters are displayed in "ifconfig -v". CR: D842
2014-10-01 21:37:32 +00:00
&VNET_NAME(def_use_flowid), 0,
Add a sysctl/tunable default value for the use_flowid sysctl in r232008.
2012-02-23 21:56:53 +00:00
"Default setting for using flow id for load sharing");
Virtualize lagg(4) cloner. This change fixes a panic when tearing down if_lagg(4) interfaces which were cloned in a vnet jail. Sysctl nodes which are dynamically generated for each cloned interface (net.link.lagg.N.*) have been removed, and use_flowid and flowid_shift ifconfig(8) parameters have been added instead. Flags and per-interface statistics counters are displayed in "ifconfig -v". CR: D842
2014-10-01 21:37:32 +00:00
Select lacp egress ports based on NUMA domain This change creates an array of port maps indexed by numa domain for lacp port selection. If we have lacp interfaces in more than one domain, then we select the egress port by indexing into the numa port maps and picking a port on the appropriate numa domain. This is behavior is controlled by the new ifconfig use_numa flag and net.link.lagg.use_numa sysctl/tunable (both modeled after the existing use_flowid), which default to enabled. Reviewed by: bz, hselasky, markj (and scottl, earlier version) Sponsored by: Netflix Differential Revision: https://reviews.freebsd.org/D20060
2019-05-03 14:43:21 +00:00
/* Default value for using numa */
VNET_DEFINE_STATIC(int, def_use_numa) = 1;
#define V_def_use_numa VNET(def_use_numa)
SYSCTL_INT(_net_link_lagg, OID_AUTO, default_use_numa, CTLFLAG_RWTUN,
&VNET_NAME(def_use_numa), 0,
"Use numa to steer flows");
Start process of removing the use of the deprecated "M_FLOWID" flag from the FreeBSD network code. The flag is still kept around in the "sys/mbuf.h" header file, but does no longer have any users. Instead the "m_pkthdr.rsstype" field in the mbuf structure is now used to decide the meaning of the "m_pkthdr.flowid" field. To modify the "m_pkthdr.rsstype" field please use the existing "M_HASHTYPE_XXX" macros as defined in the "sys/mbuf.h" header file. This patch introduces new behaviour in the transmit direction. Previously network drivers checked if "M_FLOWID" was set in "m_flags" before using the "m_pkthdr.flowid" field. This check has now now been replaced by checking if "M_HASHTYPE_GET(m)" is different from "M_HASHTYPE_NONE". In the future more hashtypes will be added, for example hashtypes for hardware dedicated flows. "M_HASHTYPE_OPAQUE" indicates that the "m_pkthdr.flowid" value is valid and has no particular type. This change removes the need for an "if" statement in TCP transmit code checking for the presence of a valid flowid value. The "if" statement mentioned above is now a direct variable assignment which is then later checked by the respective network drivers like before. Additional notes: - The SCTP code changes will be committed as a separate patch. - Removal of the "M_FLOWID" flag will also be done separately. - The FreeBSD version has been bumped. MFC after: 1 month Sponsored by: Mellanox Technologies
2014-12-01 11:45:24 +00:00
/* Default value for flowid shift */
Use the new VNET_DEFINE_STATIC macro when we are defining static VNET variables. Reviewed by: bz Sponsored by: DARPA, AFRL Differential Revision: https://reviews.freebsd.org/D16147
2018-07-24 16:35:52 +00:00
VNET_DEFINE_STATIC(int, def_flowid_shift) = 16;
Virtualize lagg(4) cloner. This change fixes a panic when tearing down if_lagg(4) interfaces which were cloned in a vnet jail. Sysctl nodes which are dynamically generated for each cloned interface (net.link.lagg.N.*) have been removed, and use_flowid and flowid_shift ifconfig(8) parameters have been added instead. Flags and per-interface statistics counters are displayed in "ifconfig -v". CR: D842
2014-10-01 21:37:32 +00:00
#define V_def_flowid_shift VNET(def_flowid_shift)
Pull in r267961 and r267973 again. Fix for issues reported will follow.
2014-06-28 03:56:17 +00:00
SYSCTL_INT(_net_link_lagg, OID_AUTO, default_flowid_shift, CTLFLAG_RWTUN,
Virtualize lagg(4) cloner. This change fixes a panic when tearing down if_lagg(4) interfaces which were cloned in a vnet jail. Sysctl nodes which are dynamically generated for each cloned interface (net.link.lagg.N.*) have been removed, and use_flowid and flowid_shift ifconfig(8) parameters have been added instead. Flags and per-interface statistics counters are displayed in "ifconfig -v". CR: D842
2014-10-01 21:37:32 +00:00
&VNET_NAME(def_flowid_shift), 0,
Multi-queue NIC drivers and multi-port lagg tend to use the same lower bits of the flowid as each other, resulting in a poor distribution of packets among queues in certain cases. Work around this by adding a set of sysctls for controlling a bit-shift on the flowid when doing multi-port aggrigation in lagg and lacp. By default, lagg/lacp will now use bits 16 and higher instead of 0 and higher. Reviewed by: max Obtained from: Netflix MFC after: 3 days
2013-12-30 01:32:17 +00:00
"Default setting for flowid shift for load sharing");
Add a sysctl knob to accept input packets on any link in a failover lagg.
2010-09-01 16:53:38 +00:00
Virtualize lagg(4) cloner. This change fixes a panic when tearing down if_lagg(4) interfaces which were cloned in a vnet jail. Sysctl nodes which are dynamically generated for each cloned interface (net.link.lagg.N.*) have been removed, and use_flowid and flowid_shift ifconfig(8) parameters have been added instead. Flags and per-interface statistics counters are displayed in "ifconfig -v". CR: D842
2014-10-01 21:37:32 +00:00
static void
vnet_lagg_init(const void *unused __unused)
{
LAGG_LIST_LOCK_INIT();
SLIST_INIT(&V_lagg_list);
if_clone: migrate some consumers to the new KPI. Convert most of the cloner customers who require custom params to the new if_clone KPI. Reviewed by: kp Differential Revision: https://reviews.freebsd.org/D36636 MFC after: 2 weeks
2022-09-22 12:30:09 +00:00
struct if_clone_addreq req = {
.create_f = lagg_clone_create,
.destroy_f = lagg_clone_destroy,
.flags = IFC_F_AUTOUNIT,
};
V_lagg_cloner = ifc_attach_cloner(laggname, &req);
Virtualize lagg(4) cloner. This change fixes a panic when tearing down if_lagg(4) interfaces which were cloned in a vnet jail. Sysctl nodes which are dynamically generated for each cloned interface (net.link.lagg.N.*) have been removed, and use_flowid and flowid_shift ifconfig(8) parameters have been added instead. Flags and per-interface statistics counters are displayed in "ifconfig -v". CR: D842
2014-10-01 21:37:32 +00:00
}
VNET_SYSINIT(vnet_lagg_init, SI_SUB_PROTO_IFATTACHDOMAIN, SI_ORDER_ANY,
vnet_lagg_init, NULL);
static void
vnet_lagg_uninit(const void *unused __unused)
{
if_clone: migrate some consumers to the new KPI. Convert most of the cloner customers who require custom params to the new if_clone KPI. Reviewed by: kp Differential Revision: https://reviews.freebsd.org/D36636 MFC after: 2 weeks
2022-09-22 12:30:09 +00:00
ifc_detach_cloner(V_lagg_cloner);
Virtualize lagg(4) cloner. This change fixes a panic when tearing down if_lagg(4) interfaces which were cloned in a vnet jail. Sysctl nodes which are dynamically generated for each cloned interface (net.link.lagg.N.*) have been removed, and use_flowid and flowid_shift ifconfig(8) parameters have been added instead. Flags and per-interface statistics counters are displayed in "ifconfig -v". CR: D842
2014-10-01 21:37:32 +00:00
LAGG_LIST_LOCK_DESTROY();
}
Get closer to a VIMAGE network stack teardown from top to bottom rather than removing the network interfaces first. This change is rather larger and convoluted as the ordering requirements cannot be separated. Move the pfil(9) framework to SI_SUB_PROTO_PFIL, move Firewalls and related modules to their own SI_SUB_PROTO_FIREWALL. Move initialization of "physical" interfaces to SI_SUB_DRIVERS, move virtual (cloned) interfaces to SI_SUB_PSEUDO. Move Multicast to SI_SUB_PROTO_MC. Re-work parts of multicast initialisation and teardown, not taking the huge amount of memory into account if used as a module yet. For interface teardown we try to do as many of them as we can on SI_SUB_INIT_IF, but for some this makes no sense, e.g., when tunnelling over a higher layer protocol such as IP. In that case the interface has to go along (or before) the higher layer protocol is shutdown. Kernel hhooks need to go last on teardown as they may be used at various higher layers and we cannot remove them before we cleaned up the higher layers. For interface teardown there are multiple paths: (a) a cloned interface is destroyed (inside a VIMAGE or in the base system), (b) any interface is moved from a virtual network stack to a different network stack ("vmove"), or (c) a virtual network stack is being shut down. All code paths go through if_detach_internal() where we, depending on the vmove flag or the vnet state, make a decision on how much to shut down; in case we are destroying a VNET the individual protocol layers will cleanup their own parts thus we cannot do so again for each interface as we end up with, e.g., double-frees, destroying locks twice or acquiring already destroyed locks. When calling into protocol cleanups we equally have to tell them whether they need to detach upper layer protocols ("ulp") or not (e.g., in6_ifdetach()). Provide or enahnce helper functions to do proper cleanup at a protocol rather than at an interface level. Approved by: re (hrs) Obtained from: projects/vnet Reviewed by: gnn, jhb Sponsored by: The FreeBSD Foundation MFC after: 2 weeks Differential Revision: https://reviews.freebsd.org/D6747
2016-06-21 13:48:49 +00:00
VNET_SYSUNINIT(vnet_lagg_uninit, SI_SUB_INIT_IF, SI_ORDER_ANY,
Virtualize lagg(4) cloner. This change fixes a panic when tearing down if_lagg(4) interfaces which were cloned in a vnet jail. Sysctl nodes which are dynamically generated for each cloned interface (net.link.lagg.N.*) have been removed, and use_flowid and flowid_shift ifconfig(8) parameters have been added instead. Flags and per-interface statistics counters are displayed in "ifconfig -v". CR: D842
2014-10-01 21:37:32 +00:00
vnet_lagg_uninit, NULL);
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
static int
lagg_modevent(module_t mod, int type, void *data)
{
switch (type) {
case MOD_LOAD:
Add support for IP over infiniband, IPoIB, to lagg(4). Currently only the failover protocol is supported due to limitations in the IPoIB architecture. Refer to the lagg(4) manual page for how to configure and use this new feature. A new network interface type, IFT_INFINIBANDLAG, has been added, similar to the existing IFT_IEEE8023ADLAG . ifconfig(8) has been updated to accept a new laggtype argument when creating lagg(4) network interfaces. This new argument is used to distinguish between ethernet and infiniband type of lagg(4) network interface. The laggtype argument is optional and defaults to ethernet. The lagg(4) command line syntax is backwards compatible. Differential Revision: https://reviews.freebsd.org/D26254 Reviewed by: melifaro@ MFC after: 1 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2020-10-22 09:47:12 +00:00
lagg_input_ethernet_p = lagg_input_ethernet;
lagg_input_infiniband_p = lagg_input_infiniband;
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
lagg_linkstate_p = lagg_port_state;
lagg_detach_cookie = EVENTHANDLER_REGISTER(
ifnet_departure_event, lagg_port_ifdetach, NULL,
EVENTHANDLER_PRI_ANY);
break;
case MOD_UNLOAD:
EVENTHANDLER_DEREGISTER(ifnet_departure_event,
lagg_detach_cookie);
Add support for IP over infiniband, IPoIB, to lagg(4). Currently only the failover protocol is supported due to limitations in the IPoIB architecture. Refer to the lagg(4) manual page for how to configure and use this new feature. A new network interface type, IFT_INFINIBANDLAG, has been added, similar to the existing IFT_IEEE8023ADLAG . ifconfig(8) has been updated to accept a new laggtype argument when creating lagg(4) network interfaces. This new argument is used to distinguish between ethernet and infiniband type of lagg(4) network interface. The laggtype argument is optional and defaults to ethernet. The lagg(4) command line syntax is backwards compatible. Differential Revision: https://reviews.freebsd.org/D26254 Reviewed by: melifaro@ MFC after: 1 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2020-10-22 09:47:12 +00:00
lagg_input_ethernet_p = NULL;
lagg_input_infiniband_p = NULL;
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
lagg_linkstate_p = NULL;
break;
default:
return (EOPNOTSUPP);
}
return (0);
}
static moduledata_t lagg_mod = {
"if_lagg",
lagg_modevent,
Revert previous commit... Pointyhat to: kevlo (myself)
2012-10-10 08:36:38 +00:00
0
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
};
DECLARE_MODULE(if_lagg, lagg_mod, SI_SUB_PSEUDO, SI_ORDER_ANY);
Add missing MODULE_VERSION() definition to protect against duplicating module loads. PR: kern/159345 Reported by: Eugene Grosbein <egrosbein att rdtc ru> Tested by: Eugene Grosbein <egrosbein att rdtc ru> Approved by: re (kib) MFC after: 1 week
2011-08-01 11:24:55 +00:00
MODULE_VERSION(if_lagg, 1);
Add support for IP over infiniband, IPoIB, to lagg(4). Currently only the failover protocol is supported due to limitations in the IPoIB architecture. Refer to the lagg(4) manual page for how to configure and use this new feature. A new network interface type, IFT_INFINIBANDLAG, has been added, similar to the existing IFT_IEEE8023ADLAG . ifconfig(8) has been updated to accept a new laggtype argument when creating lagg(4) network interfaces. This new argument is used to distinguish between ethernet and infiniband type of lagg(4) network interface. The laggtype argument is optional and defaults to ethernet. The lagg(4) command line syntax is backwards compatible. Differential Revision: https://reviews.freebsd.org/D26254 Reviewed by: melifaro@ MFC after: 1 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2020-10-22 09:47:12 +00:00
MODULE_DEPEND(if_lagg, if_infiniband, 1, 1, 1);
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
- Provide lagg_proto_attach(), lagg_proto_detach(). - Make detach a protocol method in lagg_protos. - Simplify code to lookup protocols. Sponsored by: Netflix Sponsored by: Nginx, Inc.
2014-09-26 11:01:04 +00:00
static void
lagg_proto_attach(struct lagg_softc *sc, lagg_proto pr)
{
Introduce sleepable locks into if_lagg. Before this change if_lagg was using nonsleepable rmlocks to protect its internal state. This patch introduces another sx lock to protect code paths that require sleeping, while still uses old rmlock to protect hot nonsleepable data paths. This change allows to remove taskqueue decoupling used before to change interface addresses without holding the lock. Instead it uses sx lock to protect direct if_ioctl() calls. As another bonus, the new code synchronizes enabled capabilities of member interfaces, and allows to control them with ifconfig laggX, that was impossible before. This part should fix interoperation with if_bridge, that may need to disable some capabilities, such as TXCSUM or LRO, to allow bridging with noncapable interfaces. MFC after: 2 weeks Sponsored by: iXsystems, Inc. Differential Revision: https://reviews.freebsd.org/D10514
2017-05-02 19:09:11 +00:00
LAGG_XLOCK_ASSERT(sc);
- Provide lagg_proto_attach(), lagg_proto_detach(). - Make detach a protocol method in lagg_protos. - Simplify code to lookup protocols. Sponsored by: Netflix Sponsored by: Nginx, Inc.
2014-09-26 11:01:04 +00:00
KASSERT(sc->sc_proto == LAGG_PROTO_NONE, ("%s: sc %p has proto",
__func__, sc));
if (sc->sc_ifflags & IFF_DEBUG)
if_printf(sc->sc_ifp, "using proto %u\n", pr);
Make all lagg protocol methods live in lagg_protos, not in softc. All interfaces of a same protocol, use the same methods. Sponsored by: Netflix Sponsored by: Nginx, Inc.
2014-09-26 12:54:24 +00:00
if (lagg_protos[pr].pr_attach != NULL)
lagg_protos[pr].pr_attach(sc);
- Provide lagg_proto_attach(), lagg_proto_detach(). - Make detach a protocol method in lagg_protos. - Simplify code to lookup protocols. Sponsored by: Netflix Sponsored by: Nginx, Inc.
2014-09-26 11:01:04 +00:00
sc->sc_proto = pr;
}
static void
lagg_proto_detach(struct lagg_softc *sc)
{
lagg_proto pr;
Introduce sleepable locks into if_lagg. Before this change if_lagg was using nonsleepable rmlocks to protect its internal state. This patch introduces another sx lock to protect code paths that require sleeping, while still uses old rmlock to protect hot nonsleepable data paths. This change allows to remove taskqueue decoupling used before to change interface addresses without holding the lock. Instead it uses sx lock to protect direct if_ioctl() calls. As another bonus, the new code synchronizes enabled capabilities of member interfaces, and allows to control them with ifconfig laggX, that was impossible before. This part should fix interoperation with if_bridge, that may need to disable some capabilities, such as TXCSUM or LRO, to allow bridging with noncapable interfaces. MFC after: 2 weeks Sponsored by: iXsystems, Inc. Differential Revision: https://reviews.freebsd.org/D10514
2017-05-02 19:09:11 +00:00
LAGG_XLOCK_ASSERT(sc);
- Provide lagg_proto_attach(), lagg_proto_detach(). - Make detach a protocol method in lagg_protos. - Simplify code to lookup protocols. Sponsored by: Netflix Sponsored by: Nginx, Inc.
2014-09-26 11:01:04 +00:00
pr = sc->sc_proto;
- Whitespace. - Remove caddr_t.
2014-09-26 12:35:58 +00:00
sc->sc_proto = LAGG_PROTO_NONE;
- Provide lagg_proto_attach(), lagg_proto_detach(). - Make detach a protocol method in lagg_protos. - Simplify code to lookup protocols. Sponsored by: Netflix Sponsored by: Nginx, Inc.
2014-09-26 11:01:04 +00:00
if (lagg_protos[pr].pr_detach != NULL)
lagg_protos[pr].pr_detach(sc);
}
Make all lagg protocol methods live in lagg_protos, not in softc. All interfaces of a same protocol, use the same methods. Sponsored by: Netflix Sponsored by: Nginx, Inc.
2014-09-26 12:54:24 +00:00
static int
lagg_proto_start(struct lagg_softc *sc, struct mbuf *m)
{
return (lagg_protos[sc->sc_proto].pr_start(sc, m));
}
static struct mbuf *
lagg_proto_input(struct lagg_softc *sc, struct lagg_port *lp, struct mbuf *m)
{
return (lagg_protos[sc->sc_proto].pr_input(sc, lp, m));
}
static int
lagg_proto_addport(struct lagg_softc *sc, struct lagg_port *lp)
{
if (lagg_protos[sc->sc_proto].pr_addport == NULL)
return (0);
else
return (lagg_protos[sc->sc_proto].pr_addport(lp));
}
static void
lagg_proto_delport(struct lagg_softc *sc, struct lagg_port *lp)
{
if (lagg_protos[sc->sc_proto].pr_delport != NULL)
lagg_protos[sc->sc_proto].pr_delport(lp);
}
static void
lagg_proto_linkstate(struct lagg_softc *sc, struct lagg_port *lp)
{
if (lagg_protos[sc->sc_proto].pr_linkstate != NULL)
lagg_protos[sc->sc_proto].pr_linkstate(lp);
}
static void
lagg_proto_init(struct lagg_softc *sc)
{
if (lagg_protos[sc->sc_proto].pr_init != NULL)
lagg_protos[sc->sc_proto].pr_init(sc);
}
static void
lagg_proto_stop(struct lagg_softc *sc)
{
if (lagg_protos[sc->sc_proto].pr_stop != NULL)
lagg_protos[sc->sc_proto].pr_stop(sc);
}
static void
lagg_proto_lladdr(struct lagg_softc *sc)
{
if (lagg_protos[sc->sc_proto].pr_lladdr != NULL)
lagg_protos[sc->sc_proto].pr_lladdr(sc);
}
static void
lagg_proto_request(struct lagg_softc *sc, void *v)
{
if (lagg_protos[sc->sc_proto].pr_request != NULL)
lagg_protos[sc->sc_proto].pr_request(sc, v);
}
static void
lagg_proto_portreq(struct lagg_softc *sc, struct lagg_port *lp, void *v)
{
if (lagg_protos[sc->sc_proto].pr_portreq != NULL)
lagg_protos[sc->sc_proto].pr_portreq(lp, v);
}
Propagate the vlan eventis to the underlying interfaces/members so they can do initialization of hw related features. PR: kern/141646 Reviewed by: thompsa Approved by: thompsa(co-mentor) MFC after: 2 weeks
2010-02-06 13:49:35 +00:00
/*
* This routine is run via an vlan
* config EVENT
*/
static void
lagg_register_vlan(void *arg, struct ifnet *ifp, u_int16_t vtag)
{
- Whitespace. - Remove caddr_t.
2014-09-26 12:35:58 +00:00
struct lagg_softc *sc = ifp->if_softc;
struct lagg_port *lp;
if (ifp->if_softc != arg) /* Not our event */
return;
The list of ports in configuration path shall be protected by locks, epoch shall be used only for fast path. Thus use LAGG_XLOCK() in lagg_[un]register_vlan. This fixes sleeping in epoch panic. PR: 240609
2020-12-08 16:46:00 +00:00
LAGG_XLOCK(sc);
Replace rmlock with epoch in lagg Use the new epoch based reclamation API. Now the hot paths will not block at all, and the sx lock is used for the softc data. This fixes LORs reported where the rwlock was obtained when the sxlock was held. Submitted by: mmacy Reported by: Harry Schmalzbauer <freebsd@omnilan.de> Reviewed by: sbruno Sponsored by: Limelight Networks Differential Revision: https://reviews.freebsd.org/D15355
2018-05-14 20:06:49 +00:00
CK_SLIST_FOREACH(lp, &sc->sc_ports, lp_entries)
Remove unneeded conditions. MFC after: 2 weeks
2017-04-22 08:38:49 +00:00
EVENTHANDLER_INVOKE(vlan_config, lp->lp_ifp, vtag);
The list of ports in configuration path shall be protected by locks, epoch shall be used only for fast path. Thus use LAGG_XLOCK() in lagg_[un]register_vlan. This fixes sleeping in epoch panic. PR: 240609
2020-12-08 16:46:00 +00:00
LAGG_XUNLOCK(sc);
Propagate the vlan eventis to the underlying interfaces/members so they can do initialization of hw related features. PR: kern/141646 Reviewed by: thompsa Approved by: thompsa(co-mentor) MFC after: 2 weeks
2010-02-06 13:49:35 +00:00
}
/*
* This routine is run via an vlan
* unconfig EVENT
*/
static void
lagg_unregister_vlan(void *arg, struct ifnet *ifp, u_int16_t vtag)
{
- Whitespace. - Remove caddr_t.
2014-09-26 12:35:58 +00:00
struct lagg_softc *sc = ifp->if_softc;
struct lagg_port *lp;
if (ifp->if_softc != arg) /* Not our event */
return;
The list of ports in configuration path shall be protected by locks, epoch shall be used only for fast path. Thus use LAGG_XLOCK() in lagg_[un]register_vlan. This fixes sleeping in epoch panic. PR: 240609
2020-12-08 16:46:00 +00:00
LAGG_XLOCK(sc);
Replace rmlock with epoch in lagg Use the new epoch based reclamation API. Now the hot paths will not block at all, and the sx lock is used for the softc data. This fixes LORs reported where the rwlock was obtained when the sxlock was held. Submitted by: mmacy Reported by: Harry Schmalzbauer <freebsd@omnilan.de> Reviewed by: sbruno Sponsored by: Limelight Networks Differential Revision: https://reviews.freebsd.org/D15355
2018-05-14 20:06:49 +00:00
CK_SLIST_FOREACH(lp, &sc->sc_ports, lp_entries)
Remove unneeded conditions. MFC after: 2 weeks
2017-04-22 08:38:49 +00:00
EVENTHANDLER_INVOKE(vlan_unconfig, lp->lp_ifp, vtag);
The list of ports in configuration path shall be protected by locks, epoch shall be used only for fast path. Thus use LAGG_XLOCK() in lagg_[un]register_vlan. This fixes sleeping in epoch panic. PR: 240609
2020-12-08 16:46:00 +00:00
LAGG_XUNLOCK(sc);
Propagate the vlan eventis to the underlying interfaces/members so they can do initialization of hw related features. PR: kern/141646 Reviewed by: thompsa Approved by: thompsa(co-mentor) MFC after: 2 weeks
2010-02-06 13:49:35 +00:00
}
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
static int
if_clone: migrate some consumers to the new KPI. Convert most of the cloner customers who require custom params to the new if_clone KPI. Reviewed by: kp Differential Revision: https://reviews.freebsd.org/D36636 MFC after: 2 weeks
2022-09-22 12:30:09 +00:00
lagg_clone_create(struct if_clone *ifc, char *name, size_t len,
struct ifc_data *ifd, struct ifnet **ifpp)
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
{
Add support for IP over infiniband, IPoIB, to lagg(4). Currently only the failover protocol is supported due to limitations in the IPoIB architecture. Refer to the lagg(4) manual page for how to configure and use this new feature. A new network interface type, IFT_INFINIBANDLAG, has been added, similar to the existing IFT_IEEE8023ADLAG . ifconfig(8) has been updated to accept a new laggtype argument when creating lagg(4) network interfaces. This new argument is used to distinguish between ethernet and infiniband type of lagg(4) network interface. The laggtype argument is optional and defaults to ethernet. The lagg(4) command line syntax is backwards compatible. Differential Revision: https://reviews.freebsd.org/D26254 Reviewed by: melifaro@ MFC after: 1 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2020-10-22 09:47:12 +00:00
struct iflaggparam iflp;
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
struct lagg_softc *sc;
struct ifnet *ifp;
Add support for IP over infiniband, IPoIB, to lagg(4). Currently only the failover protocol is supported due to limitations in the IPoIB architecture. Refer to the lagg(4) manual page for how to configure and use this new feature. A new network interface type, IFT_INFINIBANDLAG, has been added, similar to the existing IFT_IEEE8023ADLAG . ifconfig(8) has been updated to accept a new laggtype argument when creating lagg(4) network interfaces. This new argument is used to distinguish between ethernet and infiniband type of lagg(4) network interface. The laggtype argument is optional and defaults to ethernet. The lagg(4) command line syntax is backwards compatible. Differential Revision: https://reviews.freebsd.org/D26254 Reviewed by: melifaro@ MFC after: 1 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2020-10-22 09:47:12 +00:00
int if_type;
int error;
static const uint8_t eaddr[LAGG_ADDR_LEN];
if_clone: migrate some consumers to the new KPI. Convert most of the cloner customers who require custom params to the new if_clone KPI. Reviewed by: kp Differential Revision: https://reviews.freebsd.org/D36636 MFC after: 2 weeks
2022-09-22 12:30:09 +00:00
if (ifd->params != NULL) {
error = ifc_copyin(ifd, &iflp, sizeof(iflp));
Add support for IP over infiniband, IPoIB, to lagg(4). Currently only the failover protocol is supported due to limitations in the IPoIB architecture. Refer to the lagg(4) manual page for how to configure and use this new feature. A new network interface type, IFT_INFINIBANDLAG, has been added, similar to the existing IFT_IEEE8023ADLAG . ifconfig(8) has been updated to accept a new laggtype argument when creating lagg(4) network interfaces. This new argument is used to distinguish between ethernet and infiniband type of lagg(4) network interface. The laggtype argument is optional and defaults to ethernet. The lagg(4) command line syntax is backwards compatible. Differential Revision: https://reviews.freebsd.org/D26254 Reviewed by: melifaro@ MFC after: 1 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2020-10-22 09:47:12 +00:00
if (error)
return (error);
switch (iflp.lagg_type) {
case LAGG_TYPE_ETHERNET:
if_type = IFT_ETHER;
break;
case LAGG_TYPE_INFINIBAND:
if_type = IFT_INFINIBAND;
break;
default:
return (EINVAL);
}
} else {
if_type = IFT_ETHER;
}
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
Use a dedicated malloc type for lagg(4)'s structures. Reviewed by: gallatin MFC after: 1 month Sponsored by: Netflix Differential Revision: https://reviews.freebsd.org/D19719
2019-03-28 21:00:54 +00:00
sc = malloc(sizeof(*sc), M_LAGG, M_WAITOK|M_ZERO);
Add support for IP over infiniband, IPoIB, to lagg(4). Currently only the failover protocol is supported due to limitations in the IPoIB architecture. Refer to the lagg(4) manual page for how to configure and use this new feature. A new network interface type, IFT_INFINIBANDLAG, has been added, similar to the existing IFT_IEEE8023ADLAG . ifconfig(8) has been updated to accept a new laggtype argument when creating lagg(4) network interfaces. This new argument is used to distinguish between ethernet and infiniband type of lagg(4) network interface. The laggtype argument is optional and defaults to ethernet. The lagg(4) command line syntax is backwards compatible. Differential Revision: https://reviews.freebsd.org/D26254 Reviewed by: melifaro@ MFC after: 1 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2020-10-22 09:47:12 +00:00
ifp = sc->sc_ifp = if_alloc(if_type);
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
if (ifp == NULL) {
Use a dedicated malloc type for lagg(4)'s structures. Reviewed by: gallatin MFC after: 1 month Sponsored by: Netflix Differential Revision: https://reviews.freebsd.org/D19719
2019-03-28 21:00:54 +00:00
free(sc, M_LAGG);
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
return (ENOSPC);
}
Introduce sleepable locks into if_lagg. Before this change if_lagg was using nonsleepable rmlocks to protect its internal state. This patch introduces another sx lock to protect code paths that require sleeping, while still uses old rmlock to protect hot nonsleepable data paths. This change allows to remove taskqueue decoupling used before to change interface addresses without holding the lock. Instead it uses sx lock to protect direct if_ioctl() calls. As another bonus, the new code synchronizes enabled capabilities of member interfaces, and allows to control them with ifconfig laggX, that was impossible before. This part should fix interoperation with if_bridge, that may need to disable some capabilities, such as TXCSUM or LRO, to allow bridging with noncapable interfaces. MFC after: 2 weeks Sponsored by: iXsystems, Inc. Differential Revision: https://reviews.freebsd.org/D10514
2017-05-02 19:09:11 +00:00
LAGG_SX_INIT(sc);
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
Add support for IP over infiniband, IPoIB, to lagg(4). Currently only the failover protocol is supported due to limitations in the IPoIB architecture. Refer to the lagg(4) manual page for how to configure and use this new feature. A new network interface type, IFT_INFINIBANDLAG, has been added, similar to the existing IFT_IEEE8023ADLAG . ifconfig(8) has been updated to accept a new laggtype argument when creating lagg(4) network interfaces. This new argument is used to distinguish between ethernet and infiniband type of lagg(4) network interface. The laggtype argument is optional and defaults to ethernet. The lagg(4) command line syntax is backwards compatible. Differential Revision: https://reviews.freebsd.org/D26254 Reviewed by: melifaro@ MFC after: 1 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2020-10-22 09:47:12 +00:00
mtx_init(&sc->sc_mtx, "lagg-mtx", NULL, MTX_DEF);
callout_init_mtx(&sc->sc_watchdog, &sc->sc_mtx, 0);
Introduce sleepable locks into if_lagg. Before this change if_lagg was using nonsleepable rmlocks to protect its internal state. This patch introduces another sx lock to protect code paths that require sleeping, while still uses old rmlock to protect hot nonsleepable data paths. This change allows to remove taskqueue decoupling used before to change interface addresses without holding the lock. Instead it uses sx lock to protect direct if_ioctl() calls. As another bonus, the new code synchronizes enabled capabilities of member interfaces, and allows to control them with ifconfig laggX, that was impossible before. This part should fix interoperation with if_bridge, that may need to disable some capabilities, such as TXCSUM or LRO, to allow bridging with noncapable interfaces. MFC after: 2 weeks Sponsored by: iXsystems, Inc. Differential Revision: https://reviews.freebsd.org/D10514
2017-05-02 19:09:11 +00:00
LAGG_XLOCK(sc);
Virtualize lagg(4) cloner. This change fixes a panic when tearing down if_lagg(4) interfaces which were cloned in a vnet jail. Sysctl nodes which are dynamically generated for each cloned interface (net.link.lagg.N.*) have been removed, and use_flowid and flowid_shift ifconfig(8) parameters have been added instead. Flags and per-interface statistics counters are displayed in "ifconfig -v". CR: D842
2014-10-01 21:37:32 +00:00
if (V_def_use_flowid)
sc->sc_opts |= LAGG_OPT_USE_FLOWID;
Select lacp egress ports based on NUMA domain This change creates an array of port maps indexed by numa domain for lacp port selection. If we have lacp interfaces in more than one domain, then we select the egress port by indexing into the numa port maps and picking a port on the appropriate numa domain. This is behavior is controlled by the new ifconfig use_numa flag and net.link.lagg.use_numa sysctl/tunable (both modeled after the existing use_flowid), which default to enabled. Reviewed by: bz, hselasky, markj (and scottl, earlier version) Sponsored by: Netflix Differential Revision: https://reviews.freebsd.org/D20060
2019-05-03 14:43:21 +00:00
if (V_def_use_numa)
sc->sc_opts |= LAGG_OPT_USE_NUMA;
Virtualize lagg(4) cloner. This change fixes a panic when tearing down if_lagg(4) interfaces which were cloned in a vnet jail. Sysctl nodes which are dynamically generated for each cloned interface (net.link.lagg.N.*) have been removed, and use_flowid and flowid_shift ifconfig(8) parameters have been added instead. Flags and per-interface statistics counters are displayed in "ifconfig -v". CR: D842
2014-10-01 21:37:32 +00:00
sc->flowid_shift = V_def_flowid_shift;
Add the ability to set which packet layers are used for the load balance hash calculation.
2012-03-06 22:58:13 +00:00
/* Hash all layers by default */
Factor out mbuf hashing code from LAGG driver so that other network drivers can use it. This avoids some code duplication. Add missing default case to all switch statements while at it. Also move the hashing of the IPv6 flow field to layer 4 because the IPv6 flow field is constant on a per L4 connection basis and not on a per L3 network. Differential Revision: https://reviews.freebsd.org/D1987 Sponsored by: Mellanox Technologies MFC after: 1 month
2015-03-11 16:02:24 +00:00
sc->sc_flags = MBUF_HASHFLAG_L2|MBUF_HASHFLAG_L3|MBUF_HASHFLAG_L4;
Using the flowid in the mbuf assumes the network card is giving a good hash for the traffic flow, this may not be the case giving poor traffic distribution. Add a sysctl which allows us to fall back to our own flow hash code. PR: kern/164901 Submitted by: Eugene Grosbein MFC after: 1 week
2012-02-22 22:01:30 +00:00
- Provide lagg_proto_attach(), lagg_proto_detach(). - Make detach a protocol method in lagg_protos. - Simplify code to lookup protocols. Sponsored by: Netflix Sponsored by: Nginx, Inc.
2014-09-26 11:01:04 +00:00
lagg_proto_attach(sc, LAGG_PROTO_DEFAULT);
CK: update consumers to use CK macros across the board r334189 changed the fields to have names distinct from those in queue.h in order to expose the oversights as compile time errors
2018-05-24 23:21:23 +00:00
CK_SLIST_INIT(&sc->sc_ports);
Convert the if_lagg rwlock to an rmlock. We've been seeing lots of cache line contention (but not lock contention!) in our workloads between the various TX and RX threads going on. The write lock is only grabbed when configuration changes are made - which are infrequent. With this patch, the contention and cycles spent waiting for updates disappear. Sponsored by: Netflix, Inc.
2013-08-29 19:35:14 +00:00
Add support for IP over infiniband, IPoIB, to lagg(4). Currently only the failover protocol is supported due to limitations in the IPoIB architecture. Refer to the lagg(4) manual page for how to configure and use this new feature. A new network interface type, IFT_INFINIBANDLAG, has been added, similar to the existing IFT_IEEE8023ADLAG . ifconfig(8) has been updated to accept a new laggtype argument when creating lagg(4) network interfaces. This new argument is used to distinguish between ethernet and infiniband type of lagg(4) network interface. The laggtype argument is optional and defaults to ethernet. The lagg(4) command line syntax is backwards compatible. Differential Revision: https://reviews.freebsd.org/D26254 Reviewed by: melifaro@ MFC after: 1 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2020-10-22 09:47:12 +00:00
switch (if_type) {
case IFT_ETHER:
/* Initialise pseudo media types */
ifmedia_init(&sc->sc_media, 0, lagg_media_change,
lagg_media_status);
ifmedia_add(&sc->sc_media, IFM_ETHER | IFM_AUTO, 0, NULL);
ifmedia_set(&sc->sc_media, IFM_ETHER | IFM_AUTO);
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
if_clone: migrate some consumers to the new KPI. Convert most of the cloner customers who require custom params to the new if_clone KPI. Reviewed by: kp Differential Revision: https://reviews.freebsd.org/D36636 MFC after: 2 weeks
2022-09-22 12:30:09 +00:00
if_initname(ifp, laggname, ifd->unit);
Add support for IP over infiniband, IPoIB, to lagg(4). Currently only the failover protocol is supported due to limitations in the IPoIB architecture. Refer to the lagg(4) manual page for how to configure and use this new feature. A new network interface type, IFT_INFINIBANDLAG, has been added, similar to the existing IFT_IEEE8023ADLAG . ifconfig(8) has been updated to accept a new laggtype argument when creating lagg(4) network interfaces. This new argument is used to distinguish between ethernet and infiniband type of lagg(4) network interface. The laggtype argument is optional and defaults to ethernet. The lagg(4) command line syntax is backwards compatible. Differential Revision: https://reviews.freebsd.org/D26254 Reviewed by: melifaro@ MFC after: 1 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2020-10-22 09:47:12 +00:00
ifp->if_transmit = lagg_transmit_ethernet;
break;
case IFT_INFINIBAND:
if_clone: migrate some consumers to the new KPI. Convert most of the cloner customers who require custom params to the new if_clone KPI. Reviewed by: kp Differential Revision: https://reviews.freebsd.org/D36636 MFC after: 2 weeks
2022-09-22 12:30:09 +00:00
if_initname(ifp, laggname, ifd->unit);
Add support for IP over infiniband, IPoIB, to lagg(4). Currently only the failover protocol is supported due to limitations in the IPoIB architecture. Refer to the lagg(4) manual page for how to configure and use this new feature. A new network interface type, IFT_INFINIBANDLAG, has been added, similar to the existing IFT_IEEE8023ADLAG . ifconfig(8) has been updated to accept a new laggtype argument when creating lagg(4) network interfaces. This new argument is used to distinguish between ethernet and infiniband type of lagg(4) network interface. The laggtype argument is optional and defaults to ethernet. The lagg(4) command line syntax is backwards compatible. Differential Revision: https://reviews.freebsd.org/D26254 Reviewed by: melifaro@ MFC after: 1 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2020-10-22 09:47:12 +00:00
ifp->if_transmit = lagg_transmit_infiniband;
break;
default:
break;
}
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
ifp->if_softc = sc;
Convert lagg(4) to use if_transmit instead of if_start. In collaboration with: thompsa, sbruno, fabient
2012-09-20 10:05:10 +00:00
ifp->if_qflush = lagg_qflush;
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
ifp->if_init = lagg_init;
ifp->if_ioctl = lagg_ioctl;
Use underlying ports counters to get lagg statistics instead of per-packet accounting. This introduce user-visible changes like aggregating error counters. Reviewed by: asomers (prev.version), glebius CR: D781 MFC after: 2 weeks Sponsored by: Yandex LLC
2014-09-27 13:57:48 +00:00
ifp->if_get_counter = lagg_get_counter;
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
ifp->if_flags = IFF_SIMPLEX | IFF_BROADCAST | IFF_MULTICAST;
Add kernel-side support for in-kernel TLS. KTLS adds support for in-kernel framing and encryption of Transport Layer Security (1.0-1.2) data on TCP sockets. KTLS only supports offload of TLS for transmitted data. Key negotation must still be performed in userland. Once completed, transmit session keys for a connection are provided to the kernel via a new TCP_TXTLS_ENABLE socket option. All subsequent data transmitted on the socket is placed into TLS frames and encrypted using the supplied keys. Any data written to a KTLS-enabled socket via write(2), aio_write(2), or sendfile(2) is assumed to be application data and is encoded in TLS frames with an application data type. Individual records can be sent with a custom type (e.g. handshake messages) via sendmsg(2) with a new control message (TLS_SET_RECORD_TYPE) specifying the record type. At present, rekeying is not supported though the in-kernel framework should support rekeying. KTLS makes use of the recently added unmapped mbufs to store TLS frames in the socket buffer. Each TLS frame is described by a single ext_pgs mbuf. The ext_pgs structure contains the header of the TLS record (and trailer for encrypted records) as well as references to the associated TLS session. KTLS supports two primary methods of encrypting TLS frames: software TLS and ifnet TLS. Software TLS marks mbufs holding socket data as not ready via M_NOTREADY similar to sendfile(2) when TLS framing information is added to an unmapped mbuf in ktls_frame(). ktls_enqueue() is then called to schedule TLS frames for encryption. In the case of sendfile_iodone() calls ktls_enqueue() instead of pru_ready() leaving the mbufs marked M_NOTREADY until encryption is completed. For other writes (vn_sendfile when pages are available, write(2), etc.), the PRUS_NOTREADY is set when invoking pru_send() along with invoking ktls_enqueue(). A pool of worker threads (the "KTLS" kernel process) encrypts TLS frames queued via ktls_enqueue(). Each TLS frame is temporarily mapped using the direct map and passed to a software encryption backend to perform the actual encryption. (Note: The use of PHYS_TO_DMAP could be replaced with sf_bufs if someone wished to make this work on architectures without a direct map.) KTLS supports pluggable software encryption backends. Internally, Netflix uses proprietary pure-software backends. This commit includes a simple backend in a new ktls_ocf.ko module that uses the kernel's OpenCrypto framework to provide AES-GCM encryption of TLS frames. As a result, software TLS is now a bit of a misnomer as it can make use of hardware crypto accelerators. Once software encryption has finished, the TLS frame mbufs are marked ready via pru_ready(). At this point, the encrypted data appears as regular payload to the TCP stack stored in unmapped mbufs. ifnet TLS permits a NIC to offload the TLS encryption and TCP segmentation. In this mode, a new send tag type (IF_SND_TAG_TYPE_TLS) is allocated on the interface a socket is routed over and associated with a TLS session. TLS records for a TLS session using ifnet TLS are not marked M_NOTREADY but are passed down the stack unencrypted. The ip_output_send() and ip6_output_send() helper functions that apply send tags to outbound IP packets verify that the send tag of the TLS record matches the outbound interface. If so, the packet is tagged with the TLS send tag and sent to the interface. The NIC device driver must recognize packets with the TLS send tag and schedule them for TLS encryption and TCP segmentation. If the the outbound interface does not match the interface in the TLS send tag, the packet is dropped. In addition, a task is scheduled to refresh the TLS send tag for the TLS session. If a new TLS send tag cannot be allocated, the connection is dropped. If a new TLS send tag is allocated, however, subsequent packets will be tagged with the correct TLS send tag. (This latter case has been tested by configuring both ports of a Chelsio T6 in a lagg and failing over from one port to another. As the connections migrated to the new port, new TLS send tags were allocated for the new port and connections resumed without being dropped.) ifnet TLS can be enabled and disabled on supported network interfaces via new '[-]txtls[46]' options to ifconfig(8). ifnet TLS is supported across both vlan devices and lagg interfaces using failover, lacp with flowid enabled, or lacp with flowid enabled. Applications may request the current KTLS mode of a connection via a new TCP_TXTLS_MODE socket option. They can also use this socket option to toggle between software and ifnet TLS modes. In addition, a testing tool is available in tools/tools/switch_tls. This is modeled on tcpdrop and uses similar syntax. However, instead of dropping connections, -s is used to force KTLS connections to switch to software TLS and -i is used to switch to ifnet TLS. Various sysctls and counters are available under the kern.ipc.tls sysctl node. The kern.ipc.tls.enable node must be set to true to enable KTLS (it is off by default). The use of unmapped mbufs must also be enabled via kern.ipc.mb_use_ext_pgs to enable KTLS. KTLS is enabled via the KERN_TLS kernel option. This patch is the culmination of years of work by several folks including Scott Long and Randall Stewart for the original design and implementation; Drew Gallatin for several optimizations including the use of ext_pgs mbufs, the M_NOTREADY mechanism for TLS records awaiting software encryption, and pluggable software crypto backends; and John Baldwin for modifications to support hardware TLS offload. Reviewed by: gallatin, hselasky, rrs Obtained from: Netflix Sponsored by: Netflix, Chelsio Communications Differential Revision: https://reviews.freebsd.org/D21277
2019-08-27 00:01:56 +00:00
#if defined(KERN_TLS) || defined(RATELIMIT)
Implement kernel support for hardware rate limited sockets. - Add RATELIMIT kernel configuration keyword which must be set to enable the new functionality. - Add support for hardware driven, Receive Side Scaling, RSS aware, rate limited sendqueues and expose the functionality through the already established SO_MAX_PACING_RATE setsockopt(). The API support rates in the range from 1 to 4Gbytes/s which are suitable for regular TCP and UDP streams. The setsockopt(2) manual page has been updated. - Add rate limit function callback API to "struct ifnet" which supports the following operations: if_snd_tag_alloc(), if_snd_tag_modify(), if_snd_tag_query() and if_snd_tag_free(). - Add support to ifconfig to view, set and clear the IFCAP_TXRTLMT flag, which tells if a network driver supports rate limiting or not. - This patch also adds support for rate limiting through VLAN and LAGG intermediate network devices. - How rate limiting works: 1) The userspace application calls setsockopt() after accepting or making a new connection to set the rate which is then stored in the socket structure in the kernel. Later on when packets are transmitted a check is made in the transmit path for rate changes. A rate change implies a non-blocking ifp->if_snd_tag_alloc() call will be made to the destination network interface, which then sets up a custom sendqueue with the given rate limitation parameter. A "struct m_snd_tag" pointer is returned which serves as a "snd_tag" hint in the m_pkthdr for the subsequently transmitted mbufs. 2) When the network driver sees the "m->m_pkthdr.snd_tag" different from NULL, it will move the packets into a designated rate limited sendqueue given by the snd_tag pointer. It is up to the individual drivers how the rate limited traffic will be rate limited. 3) Route changes are detected by the NIC drivers in the ifp->if_transmit() routine when the ifnet pointer in the incoming snd_tag mismatches the one of the network interface. The network adapter frees the mbuf and returns EAGAIN which causes the ip_output() to release and clear the send tag. Upon next ip_output() a new "snd_tag" will be tried allocated. 4) When the PCB is detached the custom sendqueue will be released by a non-blocking ifp->if_snd_tag_free() call to the currently bound network interface. Reviewed by: wblock (manpages), adrian, gallatin, scottl (network) Differential Revision: https://reviews.freebsd.org/D3687 Sponsored by: Mellanox Technologies MFC after: 3 months
2017-01-18 13:31:17 +00:00
ifp->if_snd_tag_alloc = lagg_snd_tag_alloc;
This adds the third step in getting BBR into the tree. BBR and an updated rack depend on having access to the new ratelimit api in this commit. Sponsored by: Netflix Inc. Differential Revision: https://reviews.freebsd.org/D20953
2019-08-01 14:17:31 +00:00
ifp->if_ratelimit_query = lagg_ratelimit_query;
Implement kernel support for hardware rate limited sockets. - Add RATELIMIT kernel configuration keyword which must be set to enable the new functionality. - Add support for hardware driven, Receive Side Scaling, RSS aware, rate limited sendqueues and expose the functionality through the already established SO_MAX_PACING_RATE setsockopt(). The API support rates in the range from 1 to 4Gbytes/s which are suitable for regular TCP and UDP streams. The setsockopt(2) manual page has been updated. - Add rate limit function callback API to "struct ifnet" which supports the following operations: if_snd_tag_alloc(), if_snd_tag_modify(), if_snd_tag_query() and if_snd_tag_free(). - Add support to ifconfig to view, set and clear the IFCAP_TXRTLMT flag, which tells if a network driver supports rate limiting or not. - This patch also adds support for rate limiting through VLAN and LAGG intermediate network devices. - How rate limiting works: 1) The userspace application calls setsockopt() after accepting or making a new connection to set the rate which is then stored in the socket structure in the kernel. Later on when packets are transmitted a check is made in the transmit path for rate changes. A rate change implies a non-blocking ifp->if_snd_tag_alloc() call will be made to the destination network interface, which then sets up a custom sendqueue with the given rate limitation parameter. A "struct m_snd_tag" pointer is returned which serves as a "snd_tag" hint in the m_pkthdr for the subsequently transmitted mbufs. 2) When the network driver sees the "m->m_pkthdr.snd_tag" different from NULL, it will move the packets into a designated rate limited sendqueue given by the snd_tag pointer. It is up to the individual drivers how the rate limited traffic will be rate limited. 3) Route changes are detected by the NIC drivers in the ifp->if_transmit() routine when the ifnet pointer in the incoming snd_tag mismatches the one of the network interface. The network adapter frees the mbuf and returns EAGAIN which causes the ip_output() to release and clear the send tag. Upon next ip_output() a new "snd_tag" will be tried allocated. 4) When the PCB is detached the custom sendqueue will be released by a non-blocking ifp->if_snd_tag_free() call to the currently bound network interface. Reviewed by: wblock (manpages), adrian, gallatin, scottl (network) Differential Revision: https://reviews.freebsd.org/D3687 Sponsored by: Mellanox Technologies MFC after: 3 months
2017-01-18 13:31:17 +00:00
#endif
Don't set IFCAP_TXRTLMT during lagg_clone_create(). lagg_capabilities() will set the capability once interfaces supporting the feature are added to the lagg. Setting it on a lagg without any interfaces is pointless as the if_snd_tag_alloc call will always fail in that case. Reviewed by: hselasky, gallatin MFC after: 2 weeks Sponsored by: Netflix Differential Revision: https://reviews.freebsd.org/D19040
2019-01-31 21:35:37 +00:00
ifp->if_capenable = ifp->if_capabilities = IFCAP_HWSTATS;
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
/*
In r191367 the need for if_free_type() was removed and a new member if_alloctype was used to store the origional interface type. Take advantage of this change by removing all existing uses of if_free_type() in favor of if_free(). MFC after: 1 Month
2011-11-11 22:57:52 +00:00
* Attach as an ordinary ethernet device, children will be attached
Add support for IP over infiniband, IPoIB, to lagg(4). Currently only the failover protocol is supported due to limitations in the IPoIB architecture. Refer to the lagg(4) manual page for how to configure and use this new feature. A new network interface type, IFT_INFINIBANDLAG, has been added, similar to the existing IFT_IEEE8023ADLAG . ifconfig(8) has been updated to accept a new laggtype argument when creating lagg(4) network interfaces. This new argument is used to distinguish between ethernet and infiniband type of lagg(4) network interface. The laggtype argument is optional and defaults to ethernet. The lagg(4) command line syntax is backwards compatible. Differential Revision: https://reviews.freebsd.org/D26254 Reviewed by: melifaro@ MFC after: 1 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2020-10-22 09:47:12 +00:00
* as special device IFT_IEEE8023ADLAG or IFT_INFINIBANDLAG.
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
*/
Add support for IP over infiniband, IPoIB, to lagg(4). Currently only the failover protocol is supported due to limitations in the IPoIB architecture. Refer to the lagg(4) manual page for how to configure and use this new feature. A new network interface type, IFT_INFINIBANDLAG, has been added, similar to the existing IFT_IEEE8023ADLAG . ifconfig(8) has been updated to accept a new laggtype argument when creating lagg(4) network interfaces. This new argument is used to distinguish between ethernet and infiniband type of lagg(4) network interface. The laggtype argument is optional and defaults to ethernet. The lagg(4) command line syntax is backwards compatible. Differential Revision: https://reviews.freebsd.org/D26254 Reviewed by: melifaro@ MFC after: 1 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2020-10-22 09:47:12 +00:00
switch (if_type) {
case IFT_ETHER:
ether_ifattach(ifp, eaddr);
break;
case IFT_INFINIBAND:
Fix for IPoIB over lagg(4). Need to update both link layer address and broadcast address when active link changes for IP over infiniband. This is because the broadcast address contains the so-called P-key, which is interface dependent. Reviewed by: kib @ Differential Revision: https://reviews.freebsd.org/D27658 MFC after: 1 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2020-12-29 16:34:01 +00:00
infiniband_ifattach(ifp, eaddr, sc->sc_bcast_addr);
Add support for IP over infiniband, IPoIB, to lagg(4). Currently only the failover protocol is supported due to limitations in the IPoIB architecture. Refer to the lagg(4) manual page for how to configure and use this new feature. A new network interface type, IFT_INFINIBANDLAG, has been added, similar to the existing IFT_IEEE8023ADLAG . ifconfig(8) has been updated to accept a new laggtype argument when creating lagg(4) network interfaces. This new argument is used to distinguish between ethernet and infiniband type of lagg(4) network interface. The laggtype argument is optional and defaults to ethernet. The lagg(4) command line syntax is backwards compatible. Differential Revision: https://reviews.freebsd.org/D26254 Reviewed by: melifaro@ MFC after: 1 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2020-10-22 09:47:12 +00:00
break;
default:
break;
}
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
Propagate the vlan eventis to the underlying interfaces/members so they can do initialization of hw related features. PR: kern/141646 Reviewed by: thompsa Approved by: thompsa(co-mentor) MFC after: 2 weeks
2010-02-06 13:49:35 +00:00
sc->vlan_attach = EVENTHANDLER_REGISTER(vlan_config,
lagg_register_vlan, sc, EVENTHANDLER_PRI_FIRST);
sc->vlan_detach = EVENTHANDLER_REGISTER(vlan_unconfig,
lagg_unregister_vlan, sc, EVENTHANDLER_PRI_FIRST);
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
/* Insert into the global list of laggs */
Virtualize lagg(4) cloner. This change fixes a panic when tearing down if_lagg(4) interfaces which were cloned in a vnet jail. Sysctl nodes which are dynamically generated for each cloned interface (net.link.lagg.N.*) have been removed, and use_flowid and flowid_shift ifconfig(8) parameters have been added instead. Flags and per-interface statistics counters are displayed in "ifconfig -v". CR: D842
2014-10-01 21:37:32 +00:00
LAGG_LIST_LOCK();
SLIST_INSERT_HEAD(&V_lagg_list, sc, sc_entries);
LAGG_LIST_UNLOCK();
Introduce sleepable locks into if_lagg. Before this change if_lagg was using nonsleepable rmlocks to protect its internal state. This patch introduces another sx lock to protect code paths that require sleeping, while still uses old rmlock to protect hot nonsleepable data paths. This change allows to remove taskqueue decoupling used before to change interface addresses without holding the lock. Instead it uses sx lock to protect direct if_ioctl() calls. As another bonus, the new code synchronizes enabled capabilities of member interfaces, and allows to control them with ifconfig laggX, that was impossible before. This part should fix interoperation with if_bridge, that may need to disable some capabilities, such as TXCSUM or LRO, to allow bridging with noncapable interfaces. MFC after: 2 weeks Sponsored by: iXsystems, Inc. Differential Revision: https://reviews.freebsd.org/D10514
2017-05-02 19:09:11 +00:00
LAGG_XUNLOCK(sc);
if_clone: migrate some consumers to the new KPI. Convert most of the cloner customers who require custom params to the new if_clone KPI. Reviewed by: kp Differential Revision: https://reviews.freebsd.org/D36636 MFC after: 2 weeks
2022-09-22 12:30:09 +00:00
*ifpp = ifp;
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
return (0);
}
if_clone: migrate some consumers to the new KPI. Convert most of the cloner customers who require custom params to the new if_clone KPI. Reviewed by: kp Differential Revision: https://reviews.freebsd.org/D36636 MFC after: 2 weeks
2022-09-22 12:30:09 +00:00
static int
lagg_clone_destroy(struct if_clone *ifc, struct ifnet *ifp, uint32_t flags)
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
{
struct lagg_softc *sc = (struct lagg_softc *)ifp->if_softc;
struct lagg_port *lp;
Introduce sleepable locks into if_lagg. Before this change if_lagg was using nonsleepable rmlocks to protect its internal state. This patch introduces another sx lock to protect code paths that require sleeping, while still uses old rmlock to protect hot nonsleepable data paths. This change allows to remove taskqueue decoupling used before to change interface addresses without holding the lock. Instead it uses sx lock to protect direct if_ioctl() calls. As another bonus, the new code synchronizes enabled capabilities of member interfaces, and allows to control them with ifconfig laggX, that was impossible before. This part should fix interoperation with if_bridge, that may need to disable some capabilities, such as TXCSUM or LRO, to allow bridging with noncapable interfaces. MFC after: 2 weeks Sponsored by: iXsystems, Inc. Differential Revision: https://reviews.freebsd.org/D10514
2017-05-02 19:09:11 +00:00
LAGG_XLOCK(sc);
sc->sc_destroying = 1;
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
lagg_stop(sc);
ifp->if_flags &= ~IFF_UP;
Propagate the vlan eventis to the underlying interfaces/members so they can do initialization of hw related features. PR: kern/141646 Reviewed by: thompsa Approved by: thompsa(co-mentor) MFC after: 2 weeks
2010-02-06 13:49:35 +00:00
EVENTHANDLER_DEREGISTER(vlan_config, sc->vlan_attach);
EVENTHANDLER_DEREGISTER(vlan_unconfig, sc->vlan_detach);
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
/* Shutdown and remove lagg ports */
CK: update consumers to use CK macros across the board r334189 changed the fields to have names distinct from those in queue.h in order to expose the oversights as compile time errors
2018-05-24 23:21:23 +00:00
while ((lp = CK_SLIST_FIRST(&sc->sc_ports)) != NULL)
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
lagg_port_destroy(lp, 1);
Introduce sleepable locks into if_lagg. Before this change if_lagg was using nonsleepable rmlocks to protect its internal state. This patch introduces another sx lock to protect code paths that require sleeping, while still uses old rmlock to protect hot nonsleepable data paths. This change allows to remove taskqueue decoupling used before to change interface addresses without holding the lock. Instead it uses sx lock to protect direct if_ioctl() calls. As another bonus, the new code synchronizes enabled capabilities of member interfaces, and allows to control them with ifconfig laggX, that was impossible before. This part should fix interoperation with if_bridge, that may need to disable some capabilities, such as TXCSUM or LRO, to allow bridging with noncapable interfaces. MFC after: 2 weeks Sponsored by: iXsystems, Inc. Differential Revision: https://reviews.freebsd.org/D10514
2017-05-02 19:09:11 +00:00
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
/* Unhook the aggregation protocol */
- Provide lagg_proto_attach(), lagg_proto_detach(). - Make detach a protocol method in lagg_protos. - Simplify code to lookup protocols. Sponsored by: Netflix Sponsored by: Nginx, Inc.
2014-09-26 11:01:04 +00:00
lagg_proto_detach(sc);
Relax r317696 locking to not drain taskqueue under the lock. MFC after: 11 days
2017-05-05 16:51:53 +00:00
LAGG_XUNLOCK(sc);
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
Add support for IP over infiniband, IPoIB, to lagg(4). Currently only the failover protocol is supported due to limitations in the IPoIB architecture. Refer to the lagg(4) manual page for how to configure and use this new feature. A new network interface type, IFT_INFINIBANDLAG, has been added, similar to the existing IFT_IEEE8023ADLAG . ifconfig(8) has been updated to accept a new laggtype argument when creating lagg(4) network interfaces. This new argument is used to distinguish between ethernet and infiniband type of lagg(4) network interface. The laggtype argument is optional and defaults to ethernet. The lagg(4) command line syntax is backwards compatible. Differential Revision: https://reviews.freebsd.org/D26254 Reviewed by: melifaro@ MFC after: 1 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2020-10-22 09:47:12 +00:00
switch (ifp->if_type) {
case IFT_ETHER:
ifmedia_removeall(&sc->sc_media);
ether_ifdetach(ifp);
break;
case IFT_INFINIBAND:
infiniband_ifdetach(ifp);
break;
default:
break;
}
In r191367 the need for if_free_type() was removed and a new member if_alloctype was used to store the origional interface type. Take advantage of this change by removing all existing uses of if_free_type() in favor of if_free(). MFC after: 1 Month
2011-11-11 22:57:52 +00:00
if_free(ifp);
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
Virtualize lagg(4) cloner. This change fixes a panic when tearing down if_lagg(4) interfaces which were cloned in a vnet jail. Sysctl nodes which are dynamically generated for each cloned interface (net.link.lagg.N.*) have been removed, and use_flowid and flowid_shift ifconfig(8) parameters have been added instead. Flags and per-interface statistics counters are displayed in "ifconfig -v". CR: D842
2014-10-01 21:37:32 +00:00
LAGG_LIST_LOCK();
SLIST_REMOVE(&V_lagg_list, sc, lagg_softc, sc_entries);
LAGG_LIST_UNLOCK();
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
Add support for IP over infiniband, IPoIB, to lagg(4). Currently only the failover protocol is supported due to limitations in the IPoIB architecture. Refer to the lagg(4) manual page for how to configure and use this new feature. A new network interface type, IFT_INFINIBANDLAG, has been added, similar to the existing IFT_IEEE8023ADLAG . ifconfig(8) has been updated to accept a new laggtype argument when creating lagg(4) network interfaces. This new argument is used to distinguish between ethernet and infiniband type of lagg(4) network interface. The laggtype argument is optional and defaults to ethernet. The lagg(4) command line syntax is backwards compatible. Differential Revision: https://reviews.freebsd.org/D26254 Reviewed by: melifaro@ MFC after: 1 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2020-10-22 09:47:12 +00:00
mtx_destroy(&sc->sc_mtx);
Introduce sleepable locks into if_lagg. Before this change if_lagg was using nonsleepable rmlocks to protect its internal state. This patch introduces another sx lock to protect code paths that require sleeping, while still uses old rmlock to protect hot nonsleepable data paths. This change allows to remove taskqueue decoupling used before to change interface addresses without holding the lock. Instead it uses sx lock to protect direct if_ioctl() calls. As another bonus, the new code synchronizes enabled capabilities of member interfaces, and allows to control them with ifconfig laggX, that was impossible before. This part should fix interoperation with if_bridge, that may need to disable some capabilities, such as TXCSUM or LRO, to allow bridging with noncapable interfaces. MFC after: 2 weeks Sponsored by: iXsystems, Inc. Differential Revision: https://reviews.freebsd.org/D10514
2017-05-02 19:09:11 +00:00
LAGG_SX_DESTROY(sc);
Use a dedicated malloc type for lagg(4)'s structures. Reviewed by: gallatin MFC after: 1 month Sponsored by: Netflix Differential Revision: https://reviews.freebsd.org/D19719
2019-03-28 21:00:54 +00:00
free(sc, M_LAGG);
if_clone: migrate some consumers to the new KPI. Convert most of the cloner customers who require custom params to the new if_clone KPI. Reviewed by: kp Differential Revision: https://reviews.freebsd.org/D36636 MFC after: 2 weeks
2022-09-22 12:30:09 +00:00
return (0);
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
}
- Propagate the largest set of interface capabilities supported by all lagg ports to the lagg interface. - Use the MTU from the first interface as the lagg MTU, all extra interfaces must be the same. This fixes using a lagg interface for a vlan or enabling jumbo frames, etc. Approved by: re (kensmith) MFC After: 3 days
2007-07-30 20:17:22 +00:00
static void
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
lagg_capabilities(struct lagg_softc *sc)
{
struct lagg_port *lp;
lagg: fix lagg ifioctl after SIOCSIFCAPNV Lagg was broken by SIOCSIFCAPNV when all underlying devices support SIOCSIFCAPNV. This change updates lagg to work with SIOCSIFCAPNV and if_capabilities2. Reviewed by: kib, hselasky Sponsored by: Netflix Differential Revision: https://reviews.freebsd.org/D35865
2022-07-28 14:36:22 +00:00
int cap, cap2, ena, ena2, pena, pena2;
Improve applying unified capabilities to the lagg ports. Some NICs have some capabilities dependent, so that disabling one require disabling some other (TXCSUM/RXCSUM on em). This code tries to reach the consensus more insistently. PR: 219453 MFC after: 1 week
2017-05-26 20:15:33 +00:00
uint64_t hwa;
Improve transmit sending offload, TSO, algorithm in general. The current TSO limitation feature only takes the total number of bytes in an mbuf chain into account and does not limit by the number of mbufs in a chain. Some kinds of hardware is limited by two factors. One is the fragment length and the second is the fragment count. Both of these limits need to be taken into account when doing TSO. Else some kinds of hardware might have to drop completely valid mbuf chains because they cannot loaded into the given hardware's DMA engine. The new way of doing TSO limitation has been made backwards compatible as input from other FreeBSD developers and will use defaults for values not set. Reviewed by: adrian, rmacklem Sponsored by: Mellanox Technologies MFC after: 1 week
2014-09-22 08:27:27 +00:00
struct ifnet_hw_tsomax hw_tsomax;
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
Introduce sleepable locks into if_lagg. Before this change if_lagg was using nonsleepable rmlocks to protect its internal state. This patch introduces another sx lock to protect code paths that require sleeping, while still uses old rmlock to protect hot nonsleepable data paths. This change allows to remove taskqueue decoupling used before to change interface addresses without holding the lock. Instead it uses sx lock to protect direct if_ioctl() calls. As another bonus, the new code synchronizes enabled capabilities of member interfaces, and allows to control them with ifconfig laggX, that was impossible before. This part should fix interoperation with if_bridge, that may need to disable some capabilities, such as TXCSUM or LRO, to allow bridging with noncapable interfaces. MFC after: 2 weeks Sponsored by: iXsystems, Inc. Differential Revision: https://reviews.freebsd.org/D10514
2017-05-02 19:09:11 +00:00
LAGG_XLOCK_ASSERT(sc);
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
Improve applying unified capabilities to the lagg ports. Some NICs have some capabilities dependent, so that disabling one require disabling some other (TXCSUM/RXCSUM on em). This code tries to reach the consensus more insistently. PR: 219453 MFC after: 1 week
2017-05-26 20:15:33 +00:00
/* Get common enabled capabilities for the lagg ports */
lagg: fix lagg ifioctl after SIOCSIFCAPNV Lagg was broken by SIOCSIFCAPNV when all underlying devices support SIOCSIFCAPNV. This change updates lagg to work with SIOCSIFCAPNV and if_capabilities2. Reviewed by: kib, hselasky Sponsored by: Netflix Differential Revision: https://reviews.freebsd.org/D35865
2022-07-28 14:36:22 +00:00
ena = ena2 = ~0;
CK_SLIST_FOREACH(lp, &sc->sc_ports, lp_entries) {
Improve applying unified capabilities to the lagg ports. Some NICs have some capabilities dependent, so that disabling one require disabling some other (TXCSUM/RXCSUM on em). This code tries to reach the consensus more insistently. PR: 219453 MFC after: 1 week
2017-05-26 20:15:33 +00:00
ena &= lp->lp_ifp->if_capenable;
lagg: fix lagg ifioctl after SIOCSIFCAPNV Lagg was broken by SIOCSIFCAPNV when all underlying devices support SIOCSIFCAPNV. This change updates lagg to work with SIOCSIFCAPNV and if_capabilities2. Reviewed by: kib, hselasky Sponsored by: Netflix Differential Revision: https://reviews.freebsd.org/D35865
2022-07-28 14:36:22 +00:00
ena2 &= lp->lp_ifp->if_capenable2;
}
if (CK_SLIST_FIRST(&sc->sc_ports) == NULL)
ena = ena2 = 0;
Improve transmit sending offload, TSO, algorithm in general. The current TSO limitation feature only takes the total number of bytes in an mbuf chain into account and does not limit by the number of mbufs in a chain. Some kinds of hardware is limited by two factors. One is the fragment length and the second is the fragment count. Both of these limits need to be taken into account when doing TSO. Else some kinds of hardware might have to drop completely valid mbuf chains because they cannot loaded into the given hardware's DMA engine. The new way of doing TSO limitation has been made backwards compatible as input from other FreeBSD developers and will use defaults for values not set. Reviewed by: adrian, rmacklem Sponsored by: Mellanox Technologies MFC after: 1 week
2014-09-22 08:27:27 +00:00
Improve applying unified capabilities to the lagg ports. Some NICs have some capabilities dependent, so that disabling one require disabling some other (TXCSUM/RXCSUM on em). This code tries to reach the consensus more insistently. PR: 219453 MFC after: 1 week
2017-05-26 20:15:33 +00:00
/*
* Apply common enabled capabilities back to the lagg ports.
* May require several iterations if they are dependent.
*/
do {
pena = ena;
lagg: fix lagg ifioctl after SIOCSIFCAPNV Lagg was broken by SIOCSIFCAPNV when all underlying devices support SIOCSIFCAPNV. This change updates lagg to work with SIOCSIFCAPNV and if_capabilities2. Reviewed by: kib, hselasky Sponsored by: Netflix Differential Revision: https://reviews.freebsd.org/D35865
2022-07-28 14:36:22 +00:00
pena2 = ena2;
Replace rmlock with epoch in lagg Use the new epoch based reclamation API. Now the hot paths will not block at all, and the sx lock is used for the softc data. This fixes LORs reported where the rwlock was obtained when the sxlock was held. Submitted by: mmacy Reported by: Harry Schmalzbauer <freebsd@omnilan.de> Reviewed by: sbruno Sponsored by: Limelight Networks Differential Revision: https://reviews.freebsd.org/D15355
2018-05-14 20:06:49 +00:00
CK_SLIST_FOREACH(lp, &sc->sc_ports, lp_entries) {
lagg: fix lagg ifioctl after SIOCSIFCAPNV Lagg was broken by SIOCSIFCAPNV when all underlying devices support SIOCSIFCAPNV. This change updates lagg to work with SIOCSIFCAPNV and if_capabilities2. Reviewed by: kib, hselasky Sponsored by: Netflix Differential Revision: https://reviews.freebsd.org/D35865
2022-07-28 14:36:22 +00:00
lagg_setcaps(lp, ena, ena2);
Improve applying unified capabilities to the lagg ports. Some NICs have some capabilities dependent, so that disabling one require disabling some other (TXCSUM/RXCSUM on em). This code tries to reach the consensus more insistently. PR: 219453 MFC after: 1 week
2017-05-26 20:15:33 +00:00
ena &= lp->lp_ifp->if_capenable;
lagg: fix lagg ifioctl after SIOCSIFCAPNV Lagg was broken by SIOCSIFCAPNV when all underlying devices support SIOCSIFCAPNV. This change updates lagg to work with SIOCSIFCAPNV and if_capabilities2. Reviewed by: kib, hselasky Sponsored by: Netflix Differential Revision: https://reviews.freebsd.org/D35865
2022-07-28 14:36:22 +00:00
ena2 &= lp->lp_ifp->if_capenable2;
Improve applying unified capabilities to the lagg ports. Some NICs have some capabilities dependent, so that disabling one require disabling some other (TXCSUM/RXCSUM on em). This code tries to reach the consensus more insistently. PR: 219453 MFC after: 1 week
2017-05-26 20:15:33 +00:00
}
lagg: fix lagg ifioctl after SIOCSIFCAPNV Lagg was broken by SIOCSIFCAPNV when all underlying devices support SIOCSIFCAPNV. This change updates lagg to work with SIOCSIFCAPNV and if_capabilities2. Reviewed by: kib, hselasky Sponsored by: Netflix Differential Revision: https://reviews.freebsd.org/D35865
2022-07-28 14:36:22 +00:00
} while (pena != ena || pena2 != ena2);
Improve applying unified capabilities to the lagg ports. Some NICs have some capabilities dependent, so that disabling one require disabling some other (TXCSUM/RXCSUM on em). This code tries to reach the consensus more insistently. PR: 219453 MFC after: 1 week
2017-05-26 20:15:33 +00:00
/* Get other capabilities from the lagg ports */
lagg: fix lagg ifioctl after SIOCSIFCAPNV Lagg was broken by SIOCSIFCAPNV when all underlying devices support SIOCSIFCAPNV. This change updates lagg to work with SIOCSIFCAPNV and if_capabilities2. Reviewed by: kib, hselasky Sponsored by: Netflix Differential Revision: https://reviews.freebsd.org/D35865
2022-07-28 14:36:22 +00:00
cap = cap2 = ~0;
Improve applying unified capabilities to the lagg ports. Some NICs have some capabilities dependent, so that disabling one require disabling some other (TXCSUM/RXCSUM on em). This code tries to reach the consensus more insistently. PR: 219453 MFC after: 1 week
2017-05-26 20:15:33 +00:00
hwa = ~(uint64_t)0;
memset(&hw_tsomax, 0, sizeof(hw_tsomax));
Replace rmlock with epoch in lagg Use the new epoch based reclamation API. Now the hot paths will not block at all, and the sx lock is used for the softc data. This fixes LORs reported where the rwlock was obtained when the sxlock was held. Submitted by: mmacy Reported by: Harry Schmalzbauer <freebsd@omnilan.de> Reviewed by: sbruno Sponsored by: Limelight Networks Differential Revision: https://reviews.freebsd.org/D15355
2018-05-14 20:06:49 +00:00
CK_SLIST_FOREACH(lp, &sc->sc_ports, lp_entries) {
- Propagate the largest set of interface capabilities supported by all lagg ports to the lagg interface. - Use the MTU from the first interface as the lagg MTU, all extra interfaces must be the same. This fixes using a lagg interface for a vlan or enabling jumbo frames, etc. Approved by: re (kensmith) MFC After: 3 days
2007-07-30 20:17:22 +00:00
cap &= lp->lp_ifp->if_capabilities;
lagg: fix lagg ifioctl after SIOCSIFCAPNV Lagg was broken by SIOCSIFCAPNV when all underlying devices support SIOCSIFCAPNV. This change updates lagg to work with SIOCSIFCAPNV and if_capabilities2. Reviewed by: kib, hselasky Sponsored by: Netflix Differential Revision: https://reviews.freebsd.org/D35865
2022-07-28 14:36:22 +00:00
cap2 &= lp->lp_ifp->if_capabilities2;
Also propagate the if_hwassist value to the parent so that cksum offload works. Submitted by: Tom Hicks (thicks_averesys.com)
2008-12-16 22:16:34 +00:00
hwa &= lp->lp_ifp->if_hwassist;
Improve transmit sending offload, TSO, algorithm in general. The current TSO limitation feature only takes the total number of bytes in an mbuf chain into account and does not limit by the number of mbufs in a chain. Some kinds of hardware is limited by two factors. One is the fragment length and the second is the fragment count. Both of these limits need to be taken into account when doing TSO. Else some kinds of hardware might have to drop completely valid mbuf chains because they cannot loaded into the given hardware's DMA engine. The new way of doing TSO limitation has been made backwards compatible as input from other FreeBSD developers and will use defaults for values not set. Reviewed by: adrian, rmacklem Sponsored by: Mellanox Technologies MFC after: 1 week
2014-09-22 08:27:27 +00:00
if_hw_tsomax_common(lp->lp_ifp, &hw_tsomax);
- Propagate the largest set of interface capabilities supported by all lagg ports to the lagg interface. - Use the MTU from the first interface as the lagg MTU, all extra interfaces must be the same. This fixes using a lagg interface for a vlan or enabling jumbo frames, etc. Approved by: re (kensmith) MFC After: 3 days
2007-07-30 20:17:22 +00:00
}
lagg: fix lagg ifioctl after SIOCSIFCAPNV Lagg was broken by SIOCSIFCAPNV when all underlying devices support SIOCSIFCAPNV. This change updates lagg to work with SIOCSIFCAPNV and if_capabilities2. Reviewed by: kib, hselasky Sponsored by: Netflix Differential Revision: https://reviews.freebsd.org/D35865
2022-07-28 14:36:22 +00:00
if (CK_SLIST_FIRST(&sc->sc_ports) == NULL)
cap = cap2 = hwa = 0;
- Propagate the largest set of interface capabilities supported by all lagg ports to the lagg interface. - Use the MTU from the first interface as the lagg MTU, all extra interfaces must be the same. This fixes using a lagg interface for a vlan or enabling jumbo frames, etc. Approved by: re (kensmith) MFC After: 3 days
2007-07-30 20:17:22 +00:00
if (sc->sc_ifp->if_capabilities != cap ||
Also propagate the if_hwassist value to the parent so that cksum offload works. Submitted by: Tom Hicks (thicks_averesys.com)
2008-12-16 22:16:34 +00:00
sc->sc_ifp->if_capenable != ena ||
lagg: fix lagg ifioctl after SIOCSIFCAPNV Lagg was broken by SIOCSIFCAPNV when all underlying devices support SIOCSIFCAPNV. This change updates lagg to work with SIOCSIFCAPNV and if_capabilities2. Reviewed by: kib, hselasky Sponsored by: Netflix Differential Revision: https://reviews.freebsd.org/D35865
2022-07-28 14:36:22 +00:00
sc->sc_ifp->if_capenable2 != ena2 ||
Lagg did not set the value of if_hw_tsomax, so when lagg was stacked on top of network interfaces that set if_hw_tsomax, tcp_output() would see the default value instead of the value set by the network interface(s). This patch modifies lagg so that it sets if_hw_tsomax to the minimum of the value(s) for the underlying network interfaces. Reviewed by: glebius MFC after: 2 weeks
2014-04-14 20:34:48 +00:00
sc->sc_ifp->if_hwassist != hwa ||
Improve transmit sending offload, TSO, algorithm in general. The current TSO limitation feature only takes the total number of bytes in an mbuf chain into account and does not limit by the number of mbufs in a chain. Some kinds of hardware is limited by two factors. One is the fragment length and the second is the fragment count. Both of these limits need to be taken into account when doing TSO. Else some kinds of hardware might have to drop completely valid mbuf chains because they cannot loaded into the given hardware's DMA engine. The new way of doing TSO limitation has been made backwards compatible as input from other FreeBSD developers and will use defaults for values not set. Reviewed by: adrian, rmacklem Sponsored by: Mellanox Technologies MFC after: 1 week
2014-09-22 08:27:27 +00:00
if_hw_tsomax_update(sc->sc_ifp, &hw_tsomax) != 0) {
- Propagate the largest set of interface capabilities supported by all lagg ports to the lagg interface. - Use the MTU from the first interface as the lagg MTU, all extra interfaces must be the same. This fixes using a lagg interface for a vlan or enabling jumbo frames, etc. Approved by: re (kensmith) MFC After: 3 days
2007-07-30 20:17:22 +00:00
sc->sc_ifp->if_capabilities = cap;
lagg: fix lagg ifioctl after SIOCSIFCAPNV Lagg was broken by SIOCSIFCAPNV when all underlying devices support SIOCSIFCAPNV. This change updates lagg to work with SIOCSIFCAPNV and if_capabilities2. Reviewed by: kib, hselasky Sponsored by: Netflix Differential Revision: https://reviews.freebsd.org/D35865
2022-07-28 14:36:22 +00:00
sc->sc_ifp->if_capabilities2 = cap2;
- Propagate the largest set of interface capabilities supported by all lagg ports to the lagg interface. - Use the MTU from the first interface as the lagg MTU, all extra interfaces must be the same. This fixes using a lagg interface for a vlan or enabling jumbo frames, etc. Approved by: re (kensmith) MFC After: 3 days
2007-07-30 20:17:22 +00:00
sc->sc_ifp->if_capenable = ena;
lagg: fix lagg ifioctl after SIOCSIFCAPNV Lagg was broken by SIOCSIFCAPNV when all underlying devices support SIOCSIFCAPNV. This change updates lagg to work with SIOCSIFCAPNV and if_capabilities2. Reviewed by: kib, hselasky Sponsored by: Netflix Differential Revision: https://reviews.freebsd.org/D35865
2022-07-28 14:36:22 +00:00
sc->sc_ifp->if_capenable2 = ena2;
Also propagate the if_hwassist value to the parent so that cksum offload works. Submitted by: Tom Hicks (thicks_averesys.com)
2008-12-16 22:16:34 +00:00
sc->sc_ifp->if_hwassist = hwa;
- Propagate the largest set of interface capabilities supported by all lagg ports to the lagg interface. - Use the MTU from the first interface as the lagg MTU, all extra interfaces must be the same. This fixes using a lagg interface for a vlan or enabling jumbo frames, etc. Approved by: re (kensmith) MFC After: 3 days
2007-07-30 20:17:22 +00:00
getmicrotime(&sc->sc_ifp->if_lastchange);
if (sc->sc_ifflags & IFF_DEBUG)
if_printf(sc->sc_ifp,
"capabilities 0x%08x enabled 0x%08x\n", cap, ena);
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
}
}
static int
lagg_port_create(struct lagg_softc *sc, struct ifnet *ifp)
{
struct lagg_softc *sc_ptr;
Keep list of lagg ports sorted by if_index. Obtained from: Yandex LLC MFC after: 1 week Sponsored by: Yandex LLC
2014-09-26 12:42:06 +00:00
struct lagg_port *lp, *tlp;
Allow changing lagg(4) MTU. Previously, changing the MTU would require destroying the lagg and creating a new one. Now it is allowed to change the MTU of the lagg interface and the MTU of the ports will be set to match. If any port cannot set the new MTU, all ports are reverted to the original MTU of the lagg. Additionally, when adding ports, the MTU of a port will be automatically set to the MTU of the lagg. As always, the MTU of the lagg is initially determined by the MTU of the first port added. If adding an interface as a port for some reason fails, that interface is reverted to its original MTU. Submitted by: Ryan Moeller <ryan@freqlabs.com> Reviewed by: mav Relnotes: Yes Sponsored by: iXsystems Inc. Differential Revision: https://reviews.freebsd.org/D17576
2018-10-30 09:53:57 +00:00
struct ifreq ifr;
int error, i, oldmtu;
Add support for IP over infiniband, IPoIB, to lagg(4). Currently only the failover protocol is supported due to limitations in the IPoIB architecture. Refer to the lagg(4) manual page for how to configure and use this new feature. A new network interface type, IFT_INFINIBANDLAG, has been added, similar to the existing IFT_IEEE8023ADLAG . ifconfig(8) has been updated to accept a new laggtype argument when creating lagg(4) network interfaces. This new argument is used to distinguish between ethernet and infiniband type of lagg(4) network interface. The laggtype argument is optional and defaults to ethernet. The lagg(4) command line syntax is backwards compatible. Differential Revision: https://reviews.freebsd.org/D26254 Reviewed by: melifaro@ MFC after: 1 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2020-10-22 09:47:12 +00:00
int if_type;
Use underlying ports counters to get lagg statistics instead of per-packet accounting. This introduce user-visible changes like aggregating error counters. Reviewed by: asomers (prev.version), glebius CR: D781 MFC after: 2 weeks Sponsored by: Yandex LLC
2014-09-27 13:57:48 +00:00
uint64_t *pval;
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
Introduce sleepable locks into if_lagg. Before this change if_lagg was using nonsleepable rmlocks to protect its internal state. This patch introduces another sx lock to protect code paths that require sleeping, while still uses old rmlock to protect hot nonsleepable data paths. This change allows to remove taskqueue decoupling used before to change interface addresses without holding the lock. Instead it uses sx lock to protect direct if_ioctl() calls. As another bonus, the new code synchronizes enabled capabilities of member interfaces, and allows to control them with ifconfig laggX, that was impossible before. This part should fix interoperation with if_bridge, that may need to disable some capabilities, such as TXCSUM or LRO, to allow bridging with noncapable interfaces. MFC after: 2 weeks Sponsored by: iXsystems, Inc. Differential Revision: https://reviews.freebsd.org/D10514
2017-05-02 19:09:11 +00:00
LAGG_XLOCK_ASSERT(sc);
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
Allow changing lagg(4) MTU. Previously, changing the MTU would require destroying the lagg and creating a new one. Now it is allowed to change the MTU of the lagg interface and the MTU of the ports will be set to match. If any port cannot set the new MTU, all ports are reverted to the original MTU of the lagg. Additionally, when adding ports, the MTU of a port will be automatically set to the MTU of the lagg. As always, the MTU of the lagg is initially determined by the MTU of the first port added. If adding an interface as a port for some reason fails, that interface is reverted to its original MTU. Submitted by: Ryan Moeller <ryan@freqlabs.com> Reviewed by: mav Relnotes: Yes Sponsored by: iXsystems Inc. Differential Revision: https://reviews.freebsd.org/D17576
2018-10-30 09:53:57 +00:00
if (sc->sc_ifp == ifp) {
if_printf(sc->sc_ifp,
"cannot add a lagg to itself as a port\n");
return (EINVAL);
}
lagg: Avoid adding a port to a lagg device being destroyed. The lagg_clone_destroy() handles detach and waiting for ifconfig callers to drain already. This narrows the race for 2 panics that the tests triggered. Both were a consequence of adding a port to the lagg device after it had already detached from all of its ports. The link state task would run after lagg_clone_destroy() free'd the lagg softc. kernel:trap_fatal+0xa4 kernel:trap_pfault+0x61 kernel:trap+0x316 kernel:witness_checkorder+0x6d kernel:_sx_xlock+0x72 if_lagg.ko:lagg_port_state+0x3b kernel:if_down+0x144 kernel:if_detach+0x659 if_tap.ko:tap_destroy+0x46 kernel:if_clone_destroyif+0x1b7 kernel:if_clone_destroy+0x8d kernel:ifioctl+0x29c kernel:kern_ioctl+0x2bd kernel:sys_ioctl+0x16d kernel:amd64_syscall+0x337 kernel:trap_fatal+0xa4 kernel:trap_pfault+0x61 kernel:trap+0x316 kernel:witness_checkorder+0x6d kernel:_sx_xlock+0x72 if_lagg.ko:lagg_port_state+0x3b kernel:do_link_state_change+0x9b kernel:taskqueue_run_locked+0x10b kernel:taskqueue_run+0x49 kernel:ithread_loop+0x19c kernel:fork_exit+0x83 PR: 244168 Reviewed by: markj MFC after: 2 weeks Sponsored by: Dell EMC Differential Revision: https://reviews.freebsd.org/D25284
2020-08-13 22:06:27 +00:00
if (sc->sc_destroying == 1)
return (ENXIO);
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
/* Limit the maximal number of lagg ports */
if (sc->sc_count >= LAGG_MAX_PORTS)
return (ENOSPC);
/* Check if port has already been associated to a lagg */
if_lagg: allow to invoke SIOCSLAGGPORT multiple times in a row Currently, 'ifconfig laggX down' does not remove members from this lagg(4) interface. So, 'service netif stop laggX' followed by 'service netif start laggX' will choke, because "stop" will leave interfaces attached to the laggX and ifconfig from the "start" will refuse to add already-existing interfaces. The real-world case is when I am bundling together my Ethernet and WiFi interfaces and using multiple profiles for accessing network in different places: system being booted up with one profile, but later this profile being exchanged to another one, followed by 'service netif restart' will not add WiFi interface back to the lagg: the "stop" action from 'service netif restart' will shut down my main WiFi interface, so wlan0 that exists in the lagg0 will be destroyed and purged from lagg0; the "start" action will try to re-add both interfaces, but since Ethernet one is already in lagg0, ifconfig will refuse to add the wlan0 from WiFi interface. Since adding the interface to the lagg(4) when it is already here should be an idempotent action: we're really not changing anything, so this fix doesn't change the semantics of interface addition. Approved by: thompsa Reviewed by: emaste MFC after: 1 week
2012-05-28 12:13:04 +00:00
if (ifp->if_lagg != NULL) {
/* Port is already in the current lagg? */
lp = (struct lagg_port *)ifp->if_lagg;
if (lp->lp_softc == sc)
return (EEXIST);
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
return (EBUSY);
if_lagg: allow to invoke SIOCSLAGGPORT multiple times in a row Currently, 'ifconfig laggX down' does not remove members from this lagg(4) interface. So, 'service netif stop laggX' followed by 'service netif start laggX' will choke, because "stop" will leave interfaces attached to the laggX and ifconfig from the "start" will refuse to add already-existing interfaces. The real-world case is when I am bundling together my Ethernet and WiFi interfaces and using multiple profiles for accessing network in different places: system being booted up with one profile, but later this profile being exchanged to another one, followed by 'service netif restart' will not add WiFi interface back to the lagg: the "stop" action from 'service netif restart' will shut down my main WiFi interface, so wlan0 that exists in the lagg0 will be destroyed and purged from lagg0; the "start" action will try to re-add both interfaces, but since Ethernet one is already in lagg0, ifconfig will refuse to add the wlan0 from WiFi interface. Since adding the interface to the lagg(4) when it is already here should be an idempotent action: we're really not changing anything, so this fix doesn't change the semantics of interface addition. Approved by: thompsa Reviewed by: emaste MFC after: 1 week
2012-05-28 12:13:04 +00:00
}
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
Add support for IP over infiniband, IPoIB, to lagg(4). Currently only the failover protocol is supported due to limitations in the IPoIB architecture. Refer to the lagg(4) manual page for how to configure and use this new feature. A new network interface type, IFT_INFINIBANDLAG, has been added, similar to the existing IFT_IEEE8023ADLAG . ifconfig(8) has been updated to accept a new laggtype argument when creating lagg(4) network interfaces. This new argument is used to distinguish between ethernet and infiniband type of lagg(4) network interface. The laggtype argument is optional and defaults to ethernet. The lagg(4) command line syntax is backwards compatible. Differential Revision: https://reviews.freebsd.org/D26254 Reviewed by: melifaro@ MFC after: 1 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2020-10-22 09:47:12 +00:00
switch (sc->sc_ifp->if_type) {
case IFT_ETHER:
/* XXX Disallow non-ethernet interfaces (this should be any of 802) */
if (ifp->if_type != IFT_ETHER && ifp->if_type != IFT_L2VLAN)
return (EPROTONOSUPPORT);
if_type = IFT_IEEE8023ADLAG;
break;
case IFT_INFINIBAND:
/* XXX Disallow non-infiniband interfaces */
if (ifp->if_type != IFT_INFINIBAND)
return (EPROTONOSUPPORT);
if_type = IFT_INFINIBANDLAG;
break;
default:
break;
}
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
- Propagate the largest set of interface capabilities supported by all lagg ports to the lagg interface. - Use the MTU from the first interface as the lagg MTU, all extra interfaces must be the same. This fixes using a lagg interface for a vlan or enabling jumbo frames, etc. Approved by: re (kensmith) MFC After: 3 days
2007-07-30 20:17:22 +00:00
/* Allow the first Ethernet member to define the MTU */
Allow changing lagg(4) MTU. Previously, changing the MTU would require destroying the lagg and creating a new one. Now it is allowed to change the MTU of the lagg interface and the MTU of the ports will be set to match. If any port cannot set the new MTU, all ports are reverted to the original MTU of the lagg. Additionally, when adding ports, the MTU of a port will be automatically set to the MTU of the lagg. As always, the MTU of the lagg is initially determined by the MTU of the first port added. If adding an interface as a port for some reason fails, that interface is reverted to its original MTU. Submitted by: Ryan Moeller <ryan@freqlabs.com> Reviewed by: mav Relnotes: Yes Sponsored by: iXsystems Inc. Differential Revision: https://reviews.freebsd.org/D17576
2018-10-30 09:53:57 +00:00
oldmtu = -1;
if (CK_SLIST_EMPTY(&sc->sc_ports)) {
- Propagate the largest set of interface capabilities supported by all lagg ports to the lagg interface. - Use the MTU from the first interface as the lagg MTU, all extra interfaces must be the same. This fixes using a lagg interface for a vlan or enabling jumbo frames, etc. Approved by: re (kensmith) MFC After: 3 days
2007-07-30 20:17:22 +00:00
sc->sc_ifp->if_mtu = ifp->if_mtu;
Allow changing lagg(4) MTU. Previously, changing the MTU would require destroying the lagg and creating a new one. Now it is allowed to change the MTU of the lagg interface and the MTU of the ports will be set to match. If any port cannot set the new MTU, all ports are reverted to the original MTU of the lagg. Additionally, when adding ports, the MTU of a port will be automatically set to the MTU of the lagg. As always, the MTU of the lagg is initially determined by the MTU of the first port added. If adding an interface as a port for some reason fails, that interface is reverted to its original MTU. Submitted by: Ryan Moeller <ryan@freqlabs.com> Reviewed by: mav Relnotes: Yes Sponsored by: iXsystems Inc. Differential Revision: https://reviews.freebsd.org/D17576
2018-10-30 09:53:57 +00:00
} else if (sc->sc_ifp->if_mtu != ifp->if_mtu) {
if (ifp->if_ioctl == NULL) {
if_printf(sc->sc_ifp, "cannot change MTU for %s\n",
ifp->if_xname);
return (EINVAL);
}
oldmtu = ifp->if_mtu;
strlcpy(ifr.ifr_name, ifp->if_xname, sizeof(ifr.ifr_name));
ifr.ifr_mtu = sc->sc_ifp->if_mtu;
error = (*ifp->if_ioctl)(ifp, SIOCSIFMTU, (caddr_t)&ifr);
if (error != 0) {
if_printf(sc->sc_ifp, "invalid MTU for %s\n",
ifp->if_xname);
return (error);
}
ifr.ifr_mtu = oldmtu;
- Propagate the largest set of interface capabilities supported by all lagg ports to the lagg interface. - Use the MTU from the first interface as the lagg MTU, all extra interfaces must be the same. This fixes using a lagg interface for a vlan or enabling jumbo frames, etc. Approved by: re (kensmith) MFC After: 3 days
2007-07-30 20:17:22 +00:00
}
Use a dedicated malloc type for lagg(4)'s structures. Reviewed by: gallatin MFC after: 1 month Sponsored by: Netflix Differential Revision: https://reviews.freebsd.org/D19719
2019-03-28 21:00:54 +00:00
lp = malloc(sizeof(struct lagg_port), M_LAGG, M_WAITOK|M_ZERO);
Introduce sleepable locks into if_lagg. Before this change if_lagg was using nonsleepable rmlocks to protect its internal state. This patch introduces another sx lock to protect code paths that require sleeping, while still uses old rmlock to protect hot nonsleepable data paths. This change allows to remove taskqueue decoupling used before to change interface addresses without holding the lock. Instead it uses sx lock to protect direct if_ioctl() calls. As another bonus, the new code synchronizes enabled capabilities of member interfaces, and allows to control them with ifconfig laggX, that was impossible before. This part should fix interoperation with if_bridge, that may need to disable some capabilities, such as TXCSUM or LRO, to allow bridging with noncapable interfaces. MFC after: 2 weeks Sponsored by: iXsystems, Inc. Differential Revision: https://reviews.freebsd.org/D10514
2017-05-02 19:09:11 +00:00
lp->lp_softc = sc;
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
/* Check if port is a stacked lagg */
Virtualize lagg(4) cloner. This change fixes a panic when tearing down if_lagg(4) interfaces which were cloned in a vnet jail. Sysctl nodes which are dynamically generated for each cloned interface (net.link.lagg.N.*) have been removed, and use_flowid and flowid_shift ifconfig(8) parameters have been added instead. Flags and per-interface statistics counters are displayed in "ifconfig -v". CR: D842
2014-10-01 21:37:32 +00:00
LAGG_LIST_LOCK();
SLIST_FOREACH(sc_ptr, &V_lagg_list, sc_entries) {
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
if (ifp == sc_ptr->sc_ifp) {
Virtualize lagg(4) cloner. This change fixes a panic when tearing down if_lagg(4) interfaces which were cloned in a vnet jail. Sysctl nodes which are dynamically generated for each cloned interface (net.link.lagg.N.*) have been removed, and use_flowid and flowid_shift ifconfig(8) parameters have been added instead. Flags and per-interface statistics counters are displayed in "ifconfig -v". CR: D842
2014-10-01 21:37:32 +00:00
LAGG_LIST_UNLOCK();
Use a dedicated malloc type for lagg(4)'s structures. Reviewed by: gallatin MFC after: 1 month Sponsored by: Netflix Differential Revision: https://reviews.freebsd.org/D19719
2019-03-28 21:00:54 +00:00
free(lp, M_LAGG);
Allow changing lagg(4) MTU. Previously, changing the MTU would require destroying the lagg and creating a new one. Now it is allowed to change the MTU of the lagg interface and the MTU of the ports will be set to match. If any port cannot set the new MTU, all ports are reverted to the original MTU of the lagg. Additionally, when adding ports, the MTU of a port will be automatically set to the MTU of the lagg. As always, the MTU of the lagg is initially determined by the MTU of the first port added. If adding an interface as a port for some reason fails, that interface is reverted to its original MTU. Submitted by: Ryan Moeller <ryan@freqlabs.com> Reviewed by: mav Relnotes: Yes Sponsored by: iXsystems Inc. Differential Revision: https://reviews.freebsd.org/D17576
2018-10-30 09:53:57 +00:00
if (oldmtu != -1)
(*ifp->if_ioctl)(ifp, SIOCSIFMTU,
(caddr_t)&ifr);
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
return (EINVAL);
Stop GCC from complaining about lagg_port_checkstacking() being unused.
2010-01-08 16:44:33 +00:00
/* XXX disable stacking for the moment, its untested */
#ifdef LAGG_PORT_STACKING
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
lp->lp_flags |= LAGG_PORT_STACK;
if (lagg_port_checkstacking(sc_ptr) >=
LAGG_MAX_STACKING) {
Virtualize lagg(4) cloner. This change fixes a panic when tearing down if_lagg(4) interfaces which were cloned in a vnet jail. Sysctl nodes which are dynamically generated for each cloned interface (net.link.lagg.N.*) have been removed, and use_flowid and flowid_shift ifconfig(8) parameters have been added instead. Flags and per-interface statistics counters are displayed in "ifconfig -v". CR: D842
2014-10-01 21:37:32 +00:00
LAGG_LIST_UNLOCK();
Use a dedicated malloc type for lagg(4)'s structures. Reviewed by: gallatin MFC after: 1 month Sponsored by: Netflix Differential Revision: https://reviews.freebsd.org/D19719
2019-03-28 21:00:54 +00:00
free(lp, M_LAGG);
Allow changing lagg(4) MTU. Previously, changing the MTU would require destroying the lagg and creating a new one. Now it is allowed to change the MTU of the lagg interface and the MTU of the ports will be set to match. If any port cannot set the new MTU, all ports are reverted to the original MTU of the lagg. Additionally, when adding ports, the MTU of a port will be automatically set to the MTU of the lagg. As always, the MTU of the lagg is initially determined by the MTU of the first port added. If adding an interface as a port for some reason fails, that interface is reverted to its original MTU. Submitted by: Ryan Moeller <ryan@freqlabs.com> Reviewed by: mav Relnotes: Yes Sponsored by: iXsystems Inc. Differential Revision: https://reviews.freebsd.org/D17576
2018-10-30 09:53:57 +00:00
if (oldmtu != -1)
(*ifp->if_ioctl)(ifp, SIOCSIFMTU,
(caddr_t)&ifr);
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
return (E2BIG);
}
Stop GCC from complaining about lagg_port_checkstacking() being unused.
2010-01-08 16:44:33 +00:00
#endif
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
}
}
Virtualize lagg(4) cloner. This change fixes a panic when tearing down if_lagg(4) interfaces which were cloned in a vnet jail. Sysctl nodes which are dynamically generated for each cloned interface (net.link.lagg.N.*) have been removed, and use_flowid and flowid_shift ifconfig(8) parameters have been added instead. Flags and per-interface statistics counters are displayed in "ifconfig -v". CR: D842
2014-10-01 21:37:32 +00:00
LAGG_LIST_UNLOCK();
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
Introduce sleepable locks into if_lagg. Before this change if_lagg was using nonsleepable rmlocks to protect its internal state. This patch introduces another sx lock to protect code paths that require sleeping, while still uses old rmlock to protect hot nonsleepable data paths. This change allows to remove taskqueue decoupling used before to change interface addresses without holding the lock. Instead it uses sx lock to protect direct if_ioctl() calls. As another bonus, the new code synchronizes enabled capabilities of member interfaces, and allows to control them with ifconfig laggX, that was impossible before. This part should fix interoperation with if_bridge, that may need to disable some capabilities, such as TXCSUM or LRO, to allow bridging with noncapable interfaces. MFC after: 2 weeks Sponsored by: iXsystems, Inc. Differential Revision: https://reviews.freebsd.org/D10514
2017-05-02 19:09:11 +00:00
if_ref(ifp);
lp->lp_ifp = ifp;
Add support for IP over infiniband, IPoIB, to lagg(4). Currently only the failover protocol is supported due to limitations in the IPoIB architecture. Refer to the lagg(4) manual page for how to configure and use this new feature. A new network interface type, IFT_INFINIBANDLAG, has been added, similar to the existing IFT_IEEE8023ADLAG . ifconfig(8) has been updated to accept a new laggtype argument when creating lagg(4) network interfaces. This new argument is used to distinguish between ethernet and infiniband type of lagg(4) network interface. The laggtype argument is optional and defaults to ethernet. The lagg(4) command line syntax is backwards compatible. Differential Revision: https://reviews.freebsd.org/D26254 Reviewed by: melifaro@ MFC after: 1 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2020-10-22 09:47:12 +00:00
bcopy(IF_LLADDR(ifp), lp->lp_lladdr, ifp->if_addrlen);
Introduce sleepable locks into if_lagg. Before this change if_lagg was using nonsleepable rmlocks to protect its internal state. This patch introduces another sx lock to protect code paths that require sleeping, while still uses old rmlock to protect hot nonsleepable data paths. This change allows to remove taskqueue decoupling used before to change interface addresses without holding the lock. Instead it uses sx lock to protect direct if_ioctl() calls. As another bonus, the new code synchronizes enabled capabilities of member interfaces, and allows to control them with ifconfig laggX, that was impossible before. This part should fix interoperation with if_bridge, that may need to disable some capabilities, such as TXCSUM or LRO, to allow bridging with noncapable interfaces. MFC after: 2 weeks Sponsored by: iXsystems, Inc. Differential Revision: https://reviews.freebsd.org/D10514
2017-05-02 19:09:11 +00:00
lp->lp_ifcapenable = ifp->if_capenable;
CK: update consumers to use CK macros across the board r334189 changed the fields to have names distinct from those in queue.h in order to expose the oversights as compile time errors
2018-05-24 23:21:23 +00:00
if (CK_SLIST_EMPTY(&sc->sc_ports)) {
Add support for IP over infiniband, IPoIB, to lagg(4). Currently only the failover protocol is supported due to limitations in the IPoIB architecture. Refer to the lagg(4) manual page for how to configure and use this new feature. A new network interface type, IFT_INFINIBANDLAG, has been added, similar to the existing IFT_IEEE8023ADLAG . ifconfig(8) has been updated to accept a new laggtype argument when creating lagg(4) network interfaces. This new argument is used to distinguish between ethernet and infiniband type of lagg(4) network interface. The laggtype argument is optional and defaults to ethernet. The lagg(4) command line syntax is backwards compatible. Differential Revision: https://reviews.freebsd.org/D26254 Reviewed by: melifaro@ MFC after: 1 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2020-10-22 09:47:12 +00:00
bcopy(IF_LLADDR(ifp), IF_LLADDR(sc->sc_ifp), ifp->if_addrlen);
Introduce sleepable locks into if_lagg. Before this change if_lagg was using nonsleepable rmlocks to protect its internal state. This patch introduces another sx lock to protect code paths that require sleeping, while still uses old rmlock to protect hot nonsleepable data paths. This change allows to remove taskqueue decoupling used before to change interface addresses without holding the lock. Instead it uses sx lock to protect direct if_ioctl() calls. As another bonus, the new code synchronizes enabled capabilities of member interfaces, and allows to control them with ifconfig laggX, that was impossible before. This part should fix interoperation with if_bridge, that may need to disable some capabilities, such as TXCSUM or LRO, to allow bridging with noncapable interfaces. MFC after: 2 weeks Sponsored by: iXsystems, Inc. Differential Revision: https://reviews.freebsd.org/D10514
2017-05-02 19:09:11 +00:00
lagg_proto_lladdr(sc);
EVENTHANDLER_INVOKE(iflladdr_event, sc->sc_ifp);
} else {
Add support for IP over infiniband, IPoIB, to lagg(4). Currently only the failover protocol is supported due to limitations in the IPoIB architecture. Refer to the lagg(4) manual page for how to configure and use this new feature. A new network interface type, IFT_INFINIBANDLAG, has been added, similar to the existing IFT_IEEE8023ADLAG . ifconfig(8) has been updated to accept a new laggtype argument when creating lagg(4) network interfaces. This new argument is used to distinguish between ethernet and infiniband type of lagg(4) network interface. The laggtype argument is optional and defaults to ethernet. The lagg(4) command line syntax is backwards compatible. Differential Revision: https://reviews.freebsd.org/D26254 Reviewed by: melifaro@ MFC after: 1 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2020-10-22 09:47:12 +00:00
if_setlladdr(ifp, IF_LLADDR(sc->sc_ifp), ifp->if_addrlen);
Introduce sleepable locks into if_lagg. Before this change if_lagg was using nonsleepable rmlocks to protect its internal state. This patch introduces another sx lock to protect code paths that require sleeping, while still uses old rmlock to protect hot nonsleepable data paths. This change allows to remove taskqueue decoupling used before to change interface addresses without holding the lock. Instead it uses sx lock to protect direct if_ioctl() calls. As another bonus, the new code synchronizes enabled capabilities of member interfaces, and allows to control them with ifconfig laggX, that was impossible before. This part should fix interoperation with if_bridge, that may need to disable some capabilities, such as TXCSUM or LRO, to allow bridging with noncapable interfaces. MFC after: 2 weeks Sponsored by: iXsystems, Inc. Differential Revision: https://reviews.freebsd.org/D10514
2017-05-02 19:09:11 +00:00
}
lagg_setflags(lp, 1);
CK: update consumers to use CK macros across the board r334189 changed the fields to have names distinct from those in queue.h in order to expose the oversights as compile time errors
2018-05-24 23:21:23 +00:00
if (CK_SLIST_EMPTY(&sc->sc_ports))
Introduce sleepable locks into if_lagg. Before this change if_lagg was using nonsleepable rmlocks to protect its internal state. This patch introduces another sx lock to protect code paths that require sleeping, while still uses old rmlock to protect hot nonsleepable data paths. This change allows to remove taskqueue decoupling used before to change interface addresses without holding the lock. Instead it uses sx lock to protect direct if_ioctl() calls. As another bonus, the new code synchronizes enabled capabilities of member interfaces, and allows to control them with ifconfig laggX, that was impossible before. This part should fix interoperation with if_bridge, that may need to disable some capabilities, such as TXCSUM or LRO, to allow bridging with noncapable interfaces. MFC after: 2 weeks Sponsored by: iXsystems, Inc. Differential Revision: https://reviews.freebsd.org/D10514
2017-05-02 19:09:11 +00:00
sc->sc_primary = lp;
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
/* Change the interface type */
lp->lp_iftype = ifp->if_type;
Add support for IP over infiniband, IPoIB, to lagg(4). Currently only the failover protocol is supported due to limitations in the IPoIB architecture. Refer to the lagg(4) manual page for how to configure and use this new feature. A new network interface type, IFT_INFINIBANDLAG, has been added, similar to the existing IFT_IEEE8023ADLAG . ifconfig(8) has been updated to accept a new laggtype argument when creating lagg(4) network interfaces. This new argument is used to distinguish between ethernet and infiniband type of lagg(4) network interface. The laggtype argument is optional and defaults to ethernet. The lagg(4) command line syntax is backwards compatible. Differential Revision: https://reviews.freebsd.org/D26254 Reviewed by: melifaro@ MFC after: 1 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2020-10-22 09:47:12 +00:00
ifp->if_type = if_type;
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
ifp->if_lagg = lp;
lp->lp_ioctl = ifp->if_ioctl;
ifp->if_ioctl = lagg_port_ioctl;
lp->lp_output = ifp->if_output;
ifp->if_output = lagg_port_output;
Introduce sleepable locks into if_lagg. Before this change if_lagg was using nonsleepable rmlocks to protect its internal state. This patch introduces another sx lock to protect code paths that require sleeping, while still uses old rmlock to protect hot nonsleepable data paths. This change allows to remove taskqueue decoupling used before to change interface addresses without holding the lock. Instead it uses sx lock to protect direct if_ioctl() calls. As another bonus, the new code synchronizes enabled capabilities of member interfaces, and allows to control them with ifconfig laggX, that was impossible before. This part should fix interoperation with if_bridge, that may need to disable some capabilities, such as TXCSUM or LRO, to allow bridging with noncapable interfaces. MFC after: 2 weeks Sponsored by: iXsystems, Inc. Differential Revision: https://reviews.freebsd.org/D10514
2017-05-02 19:09:11 +00:00
/* Read port counters */
pval = lp->port_counters.val;
for (i = 0; i < IFCOUNTERS; i++, pval++)
*pval = ifp->if_get_counter(ifp, i);
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
Fix condition and really sort ports. Also add comment describing the intent of this code. Reported by: sbruno MFC after: 1 week Sponsored by: Yandex LLC
2015-01-17 11:32:09 +00:00
/*
* Insert into the list of ports.
* Keep ports sorted by if_index. It is handy, when configuration
* is predictable and `ifconfig laggN create ...` command
* will lead to the same result each time.
*/
Replace rmlock with epoch in lagg Use the new epoch based reclamation API. Now the hot paths will not block at all, and the sx lock is used for the softc data. This fixes LORs reported where the rwlock was obtained when the sxlock was held. Submitted by: mmacy Reported by: Harry Schmalzbauer <freebsd@omnilan.de> Reviewed by: sbruno Sponsored by: Limelight Networks Differential Revision: https://reviews.freebsd.org/D15355
2018-05-14 20:06:49 +00:00
CK_SLIST_FOREACH(tlp, &sc->sc_ports, lp_entries) {
Keep list of lagg ports sorted by if_index. Obtained from: Yandex LLC MFC after: 1 week Sponsored by: Yandex LLC
2014-09-26 12:42:06 +00:00
if (tlp->lp_ifp->if_index < ifp->if_index && (
CK: update consumers to use CK macros across the board r334189 changed the fields to have names distinct from those in queue.h in order to expose the oversights as compile time errors
2018-05-24 23:21:23 +00:00
CK_SLIST_NEXT(tlp, lp_entries) == NULL ||
((struct lagg_port*)CK_SLIST_NEXT(tlp, lp_entries))->lp_ifp->if_index >
Keep list of lagg ports sorted by if_index. Obtained from: Yandex LLC MFC after: 1 week Sponsored by: Yandex LLC
2014-09-26 12:42:06 +00:00
ifp->if_index))
break;
}
if (tlp != NULL)
CK: update consumers to use CK macros across the board r334189 changed the fields to have names distinct from those in queue.h in order to expose the oversights as compile time errors
2018-05-24 23:21:23 +00:00
CK_SLIST_INSERT_AFTER(tlp, lp, lp_entries);
Keep list of lagg ports sorted by if_index. Obtained from: Yandex LLC MFC after: 1 week Sponsored by: Yandex LLC
2014-09-26 12:42:06 +00:00
else
CK: update consumers to use CK macros across the board r334189 changed the fields to have names distinct from those in queue.h in order to expose the oversights as compile time errors
2018-05-24 23:21:23 +00:00
CK_SLIST_INSERT_HEAD(&sc->sc_ports, lp, lp_entries);
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
sc->sc_count++;
Introduce sleepable locks into if_lagg. Before this change if_lagg was using nonsleepable rmlocks to protect its internal state. This patch introduces another sx lock to protect code paths that require sleeping, while still uses old rmlock to protect hot nonsleepable data paths. This change allows to remove taskqueue decoupling used before to change interface addresses without holding the lock. Instead it uses sx lock to protect direct if_ioctl() calls. As another bonus, the new code synchronizes enabled capabilities of member interfaces, and allows to control them with ifconfig laggX, that was impossible before. This part should fix interoperation with if_bridge, that may need to disable some capabilities, such as TXCSUM or LRO, to allow bridging with noncapable interfaces. MFC after: 2 weeks Sponsored by: iXsystems, Inc. Differential Revision: https://reviews.freebsd.org/D10514
2017-05-02 19:09:11 +00:00
lagg_setmulti(lp);
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
Make all lagg protocol methods live in lagg_protos, not in softc. All interfaces of a same protocol, use the same methods. Sponsored by: Netflix Sponsored by: Nginx, Inc.
2014-09-26 12:54:24 +00:00
if ((error = lagg_proto_addport(sc, lp)) != 0) {
/* Remove the port, without calling pr_delport. */
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
lagg_port_destroy(lp, 0);
Allow changing lagg(4) MTU. Previously, changing the MTU would require destroying the lagg and creating a new one. Now it is allowed to change the MTU of the lagg interface and the MTU of the ports will be set to match. If any port cannot set the new MTU, all ports are reverted to the original MTU of the lagg. Additionally, when adding ports, the MTU of a port will be automatically set to the MTU of the lagg. As always, the MTU of the lagg is initially determined by the MTU of the first port added. If adding an interface as a port for some reason fails, that interface is reverted to its original MTU. Submitted by: Ryan Moeller <ryan@freqlabs.com> Reviewed by: mav Relnotes: Yes Sponsored by: iXsystems Inc. Differential Revision: https://reviews.freebsd.org/D17576
2018-10-30 09:53:57 +00:00
if (oldmtu != -1)
(*ifp->if_ioctl)(ifp, SIOCSIFMTU, (caddr_t)&ifr);
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
return (error);
}
Introduce sleepable locks into if_lagg. Before this change if_lagg was using nonsleepable rmlocks to protect its internal state. This patch introduces another sx lock to protect code paths that require sleeping, while still uses old rmlock to protect hot nonsleepable data paths. This change allows to remove taskqueue decoupling used before to change interface addresses without holding the lock. Instead it uses sx lock to protect direct if_ioctl() calls. As another bonus, the new code synchronizes enabled capabilities of member interfaces, and allows to control them with ifconfig laggX, that was impossible before. This part should fix interoperation with if_bridge, that may need to disable some capabilities, such as TXCSUM or LRO, to allow bridging with noncapable interfaces. MFC after: 2 weeks Sponsored by: iXsystems, Inc. Differential Revision: https://reviews.freebsd.org/D10514
2017-05-02 19:09:11 +00:00
/* Update lagg capabilities */
lagg_capabilities(sc);
lagg_linkstate(sc);
Make all lagg protocol methods live in lagg_protos, not in softc. All interfaces of a same protocol, use the same methods. Sponsored by: Netflix Sponsored by: Nginx, Inc.
2014-09-26 12:54:24 +00:00
return (0);
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
}
Stop GCC from complaining about lagg_port_checkstacking() being unused.
2010-01-08 16:44:33 +00:00
#ifdef LAGG_PORT_STACKING
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
static int
lagg_port_checkstacking(struct lagg_softc *sc)
{
struct lagg_softc *sc_ptr;
struct lagg_port *lp;
int m = 0;
Introduce sleepable locks into if_lagg. Before this change if_lagg was using nonsleepable rmlocks to protect its internal state. This patch introduces another sx lock to protect code paths that require sleeping, while still uses old rmlock to protect hot nonsleepable data paths. This change allows to remove taskqueue decoupling used before to change interface addresses without holding the lock. Instead it uses sx lock to protect direct if_ioctl() calls. As another bonus, the new code synchronizes enabled capabilities of member interfaces, and allows to control them with ifconfig laggX, that was impossible before. This part should fix interoperation with if_bridge, that may need to disable some capabilities, such as TXCSUM or LRO, to allow bridging with noncapable interfaces. MFC after: 2 weeks Sponsored by: iXsystems, Inc. Differential Revision: https://reviews.freebsd.org/D10514
2017-05-02 19:09:11 +00:00
LAGG_SXLOCK_ASSERT(sc);
Replace rmlock with epoch in lagg Use the new epoch based reclamation API. Now the hot paths will not block at all, and the sx lock is used for the softc data. This fixes LORs reported where the rwlock was obtained when the sxlock was held. Submitted by: mmacy Reported by: Harry Schmalzbauer <freebsd@omnilan.de> Reviewed by: sbruno Sponsored by: Limelight Networks Differential Revision: https://reviews.freebsd.org/D15355
2018-05-14 20:06:49 +00:00
CK_SLIST_FOREACH(lp, &sc->sc_ports, lp_entries) {
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
if (lp->lp_flags & LAGG_PORT_STACK) {
sc_ptr = (struct lagg_softc *)lp->lp_ifp->if_softc;
m = MAX(m, lagg_port_checkstacking(sc_ptr));
}
}
return (m + 1);
}
Stop GCC from complaining about lagg_port_checkstacking() being unused.
2010-01-08 16:44:33 +00:00
#endif
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
Replace rmlock with epoch in lagg Use the new epoch based reclamation API. Now the hot paths will not block at all, and the sx lock is used for the softc data. This fixes LORs reported where the rwlock was obtained when the sxlock was held. Submitted by: mmacy Reported by: Harry Schmalzbauer <freebsd@omnilan.de> Reviewed by: sbruno Sponsored by: Limelight Networks Differential Revision: https://reviews.freebsd.org/D15355
2018-05-14 20:06:49 +00:00
static void
lagg_port_destroy_cb(epoch_context_t ec)
{
struct lagg_port *lp;
struct ifnet *ifp;
lp = __containerof(ec, struct lagg_port, lp_epoch_ctx);
ifp = lp->lp_ifp;
if_rele(ifp);
Use a dedicated malloc type for lagg(4)'s structures. Reviewed by: gallatin MFC after: 1 month Sponsored by: Netflix Differential Revision: https://reviews.freebsd.org/D19719
2019-03-28 21:00:54 +00:00
free(lp, M_LAGG);
Replace rmlock with epoch in lagg Use the new epoch based reclamation API. Now the hot paths will not block at all, and the sx lock is used for the softc data. This fixes LORs reported where the rwlock was obtained when the sxlock was held. Submitted by: mmacy Reported by: Harry Schmalzbauer <freebsd@omnilan.de> Reviewed by: sbruno Sponsored by: Limelight Networks Differential Revision: https://reviews.freebsd.org/D15355
2018-05-14 20:06:49 +00:00
}
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
static int
Make all lagg protocol methods live in lagg_protos, not in softc. All interfaces of a same protocol, use the same methods. Sponsored by: Netflix Sponsored by: Nginx, Inc.
2014-09-26 12:54:24 +00:00
lagg_port_destroy(struct lagg_port *lp, int rundelport)
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
{
non-functional cleanup - remove dead code - use consistent variable names - gc unused defines - whitespace cleanup
2007-06-12 07:29:11 +00:00
struct lagg_softc *sc = lp->lp_softc;
Fix a bug that caused reinitialization failure of MAC addresses on the lagg interface when removing the primary port. PR: 201916 Differential Revision: https://reviews.freebsd.org/D3301
2015-10-07 06:32:34 +00:00
struct lagg_port *lp_ptr, *lp0;
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
struct ifnet *ifp = lp->lp_ifp;
Use underlying ports counters to get lagg statistics instead of per-packet accounting. This introduce user-visible changes like aggregating error counters. Reviewed by: asomers (prev.version), glebius CR: D781 MFC after: 2 weeks Sponsored by: Yandex LLC
2014-09-27 13:57:48 +00:00
uint64_t *pval, vdiff;
int i;
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
Introduce sleepable locks into if_lagg. Before this change if_lagg was using nonsleepable rmlocks to protect its internal state. This patch introduces another sx lock to protect code paths that require sleeping, while still uses old rmlock to protect hot nonsleepable data paths. This change allows to remove taskqueue decoupling used before to change interface addresses without holding the lock. Instead it uses sx lock to protect direct if_ioctl() calls. As another bonus, the new code synchronizes enabled capabilities of member interfaces, and allows to control them with ifconfig laggX, that was impossible before. This part should fix interoperation with if_bridge, that may need to disable some capabilities, such as TXCSUM or LRO, to allow bridging with noncapable interfaces. MFC after: 2 weeks Sponsored by: iXsystems, Inc. Differential Revision: https://reviews.freebsd.org/D10514
2017-05-02 19:09:11 +00:00
LAGG_XLOCK_ASSERT(sc);
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
Replace rmlock with epoch in lagg Use the new epoch based reclamation API. Now the hot paths will not block at all, and the sx lock is used for the softc data. This fixes LORs reported where the rwlock was obtained when the sxlock was held. Submitted by: mmacy Reported by: Harry Schmalzbauer <freebsd@omnilan.de> Reviewed by: sbruno Sponsored by: Limelight Networks Differential Revision: https://reviews.freebsd.org/D15355
2018-05-14 20:06:49 +00:00
if (rundelport)
Make all lagg protocol methods live in lagg_protos, not in softc. All interfaces of a same protocol, use the same methods. Sponsored by: Netflix Sponsored by: Nginx, Inc.
2014-09-26 12:54:24 +00:00
lagg_proto_delport(sc, lp);
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
Introduce sleepable locks into if_lagg. Before this change if_lagg was using nonsleepable rmlocks to protect its internal state. This patch introduces another sx lock to protect code paths that require sleeping, while still uses old rmlock to protect hot nonsleepable data paths. This change allows to remove taskqueue decoupling used before to change interface addresses without holding the lock. Instead it uses sx lock to protect direct if_ioctl() calls. As another bonus, the new code synchronizes enabled capabilities of member interfaces, and allows to control them with ifconfig laggX, that was impossible before. This part should fix interoperation with if_bridge, that may need to disable some capabilities, such as TXCSUM or LRO, to allow bridging with noncapable interfaces. MFC after: 2 weeks Sponsored by: iXsystems, Inc. Differential Revision: https://reviews.freebsd.org/D10514
2017-05-02 19:09:11 +00:00
if (lp->lp_detaching == 0)
lagg_clrmulti(lp);
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
/* Restore interface */
ifp->if_type = lp->lp_iftype;
ifp->if_ioctl = lp->lp_ioctl;
ifp->if_output = lp->lp_output;
ifp->if_lagg = NULL;
Use underlying ports counters to get lagg statistics instead of per-packet accounting. This introduce user-visible changes like aggregating error counters. Reviewed by: asomers (prev.version), glebius CR: D781 MFC after: 2 weeks Sponsored by: Yandex LLC
2014-09-27 13:57:48 +00:00
/* Update detached port counters */
pval = lp->port_counters.val;
Fix off by one in lagg_port_destroy(). Reported by: "Max N. Boyarov" <zotrix bsd.by>
2014-10-01 11:23:54 +00:00
for (i = 0; i < IFCOUNTERS; i++, pval++) {
Use underlying ports counters to get lagg statistics instead of per-packet accounting. This introduce user-visible changes like aggregating error counters. Reviewed by: asomers (prev.version), glebius CR: D781 MFC after: 2 weeks Sponsored by: Yandex LLC
2014-09-27 13:57:48 +00:00
vdiff = ifp->if_get_counter(ifp, i) - *pval;
Finally, convert counters in struct ifnet to counter(9). Sponsored by: Netflix Sponsored by: Nginx, Inc.
2014-09-28 08:57:07 +00:00
sc->detached_counters.val[i] += vdiff;
Use underlying ports counters to get lagg statistics instead of per-packet accounting. This introduce user-visible changes like aggregating error counters. Reviewed by: asomers (prev.version), glebius CR: D781 MFC after: 2 weeks Sponsored by: Yandex LLC
2014-09-27 13:57:48 +00:00
}
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
/* Finally, remove the port from the lagg */
Replace rmlock with epoch in lagg Use the new epoch based reclamation API. Now the hot paths will not block at all, and the sx lock is used for the softc data. This fixes LORs reported where the rwlock was obtained when the sxlock was held. Submitted by: mmacy Reported by: Harry Schmalzbauer <freebsd@omnilan.de> Reviewed by: sbruno Sponsored by: Limelight Networks Differential Revision: https://reviews.freebsd.org/D15355
2018-05-14 20:06:49 +00:00
CK_SLIST_REMOVE(&sc->sc_ports, lp, lagg_port, lp_entries);
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
sc->sc_count--;
/* Update the primary interface */
if (lp == sc->sc_primary) {
Add support for IP over infiniband, IPoIB, to lagg(4). Currently only the failover protocol is supported due to limitations in the IPoIB architecture. Refer to the lagg(4) manual page for how to configure and use this new feature. A new network interface type, IFT_INFINIBANDLAG, has been added, similar to the existing IFT_IEEE8023ADLAG . ifconfig(8) has been updated to accept a new laggtype argument when creating lagg(4) network interfaces. This new argument is used to distinguish between ethernet and infiniband type of lagg(4) network interface. The laggtype argument is optional and defaults to ethernet. The lagg(4) command line syntax is backwards compatible. Differential Revision: https://reviews.freebsd.org/D26254 Reviewed by: melifaro@ MFC after: 1 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2020-10-22 09:47:12 +00:00
uint8_t lladdr[LAGG_ADDR_LEN];
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
CK: update consumers to use CK macros across the board r334189 changed the fields to have names distinct from those in queue.h in order to expose the oversights as compile time errors
2018-05-24 23:21:23 +00:00
if ((lp0 = CK_SLIST_FIRST(&sc->sc_ports)) == NULL)
Add support for IP over infiniband, IPoIB, to lagg(4). Currently only the failover protocol is supported due to limitations in the IPoIB architecture. Refer to the lagg(4) manual page for how to configure and use this new feature. A new network interface type, IFT_INFINIBANDLAG, has been added, similar to the existing IFT_IEEE8023ADLAG . ifconfig(8) has been updated to accept a new laggtype argument when creating lagg(4) network interfaces. This new argument is used to distinguish between ethernet and infiniband type of lagg(4) network interface. The laggtype argument is optional and defaults to ethernet. The lagg(4) command line syntax is backwards compatible. Differential Revision: https://reviews.freebsd.org/D26254 Reviewed by: melifaro@ MFC after: 1 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2020-10-22 09:47:12 +00:00
bzero(&lladdr, LAGG_ADDR_LEN);
Introduce sleepable locks into if_lagg. Before this change if_lagg was using nonsleepable rmlocks to protect its internal state. This patch introduces another sx lock to protect code paths that require sleeping, while still uses old rmlock to protect hot nonsleepable data paths. This change allows to remove taskqueue decoupling used before to change interface addresses without holding the lock. Instead it uses sx lock to protect direct if_ioctl() calls. As another bonus, the new code synchronizes enabled capabilities of member interfaces, and allows to control them with ifconfig laggX, that was impossible before. This part should fix interoperation with if_bridge, that may need to disable some capabilities, such as TXCSUM or LRO, to allow bridging with noncapable interfaces. MFC after: 2 weeks Sponsored by: iXsystems, Inc. Differential Revision: https://reviews.freebsd.org/D10514
2017-05-02 19:09:11 +00:00
else
Add support for IP over infiniband, IPoIB, to lagg(4). Currently only the failover protocol is supported due to limitations in the IPoIB architecture. Refer to the lagg(4) manual page for how to configure and use this new feature. A new network interface type, IFT_INFINIBANDLAG, has been added, similar to the existing IFT_IEEE8023ADLAG . ifconfig(8) has been updated to accept a new laggtype argument when creating lagg(4) network interfaces. This new argument is used to distinguish between ethernet and infiniband type of lagg(4) network interface. The laggtype argument is optional and defaults to ethernet. The lagg(4) command line syntax is backwards compatible. Differential Revision: https://reviews.freebsd.org/D26254 Reviewed by: melifaro@ MFC after: 1 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2020-10-22 09:47:12 +00:00
bcopy(lp0->lp_lladdr, lladdr, LAGG_ADDR_LEN);
Fix lladdr change propagation for on vlans on top of it. Fix lladdr update when setting mac address manually. Fix lladdr_event for slave ports addition. MFC after: 4 weeks Sponsored by: Yandex LLC Differential Revision: https://reviews.freebsd.org/D4004
2015-11-01 19:59:04 +00:00
sc->sc_primary = lp0;
Introduce sleepable locks into if_lagg. Before this change if_lagg was using nonsleepable rmlocks to protect its internal state. This patch introduces another sx lock to protect code paths that require sleeping, while still uses old rmlock to protect hot nonsleepable data paths. This change allows to remove taskqueue decoupling used before to change interface addresses without holding the lock. Instead it uses sx lock to protect direct if_ioctl() calls. As another bonus, the new code synchronizes enabled capabilities of member interfaces, and allows to control them with ifconfig laggX, that was impossible before. This part should fix interoperation with if_bridge, that may need to disable some capabilities, such as TXCSUM or LRO, to allow bridging with noncapable interfaces. MFC after: 2 weeks Sponsored by: iXsystems, Inc. Differential Revision: https://reviews.freebsd.org/D10514
2017-05-02 19:09:11 +00:00
if (sc->sc_destroying == 0) {
Add support for IP over infiniband, IPoIB, to lagg(4). Currently only the failover protocol is supported due to limitations in the IPoIB architecture. Refer to the lagg(4) manual page for how to configure and use this new feature. A new network interface type, IFT_INFINIBANDLAG, has been added, similar to the existing IFT_IEEE8023ADLAG . ifconfig(8) has been updated to accept a new laggtype argument when creating lagg(4) network interfaces. This new argument is used to distinguish between ethernet and infiniband type of lagg(4) network interface. The laggtype argument is optional and defaults to ethernet. The lagg(4) command line syntax is backwards compatible. Differential Revision: https://reviews.freebsd.org/D26254 Reviewed by: melifaro@ MFC after: 1 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2020-10-22 09:47:12 +00:00
bcopy(lladdr, IF_LLADDR(sc->sc_ifp), sc->sc_ifp->if_addrlen);
Introduce sleepable locks into if_lagg. Before this change if_lagg was using nonsleepable rmlocks to protect its internal state. This patch introduces another sx lock to protect code paths that require sleeping, while still uses old rmlock to protect hot nonsleepable data paths. This change allows to remove taskqueue decoupling used before to change interface addresses without holding the lock. Instead it uses sx lock to protect direct if_ioctl() calls. As another bonus, the new code synchronizes enabled capabilities of member interfaces, and allows to control them with ifconfig laggX, that was impossible before. This part should fix interoperation with if_bridge, that may need to disable some capabilities, such as TXCSUM or LRO, to allow bridging with noncapable interfaces. MFC after: 2 weeks Sponsored by: iXsystems, Inc. Differential Revision: https://reviews.freebsd.org/D10514
2017-05-02 19:09:11 +00:00
lagg_proto_lladdr(sc);
EVENTHANDLER_INVOKE(iflladdr_event, sc->sc_ifp);
Fix lladdr change propagation for on vlans on top of it. Fix lladdr update when setting mac address manually. Fix lladdr_event for slave ports addition. MFC after: 4 weeks Sponsored by: Yandex LLC Differential Revision: https://reviews.freebsd.org/D4004
2015-11-01 19:59:04 +00:00
lagg: don't update link layer addresses on destroy When the lagg is being destroyed it is not necessary update the lladdr of all the lagg members every time we update the primary interface. Reviewed by: scottl Obtained from: pfSense MFC after: 1 week Sponsored by: Rubicon Communications, LLC ("Netgate") Differential Revision: https://reviews.freebsd.org/D31586
2021-08-17 14:23:50 +00:00
/*
* Update lladdr for each port (new primary needs update
* as well, to switch from old lladdr to its 'real' one).
* We can skip this if the lagg is being destroyed.
*/
CK_SLIST_FOREACH(lp_ptr, &sc->sc_ports, lp_entries)
if_setlladdr(lp_ptr->lp_ifp, lladdr,
lp_ptr->lp_ifp->if_addrlen);
}
Replace rmlock with epoch in lagg Use the new epoch based reclamation API. Now the hot paths will not block at all, and the sx lock is used for the softc data. This fixes LORs reported where the rwlock was obtained when the sxlock was held. Submitted by: mmacy Reported by: Harry Schmalzbauer <freebsd@omnilan.de> Reviewed by: sbruno Sponsored by: Limelight Networks Differential Revision: https://reviews.freebsd.org/D15355
2018-05-14 20:06:49 +00:00
}
Call if_setlladdr() on the aggregation port from a taskqueue so the softc lock is not held. The short delay between aggregating the port and setting the MAC address is fine.
2007-05-07 00:35:15 +00:00
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
if (lp->lp_ifflags)
if_printf(ifp, "%s: lp_ifflags unclean\n", __func__);
Introduce sleepable locks into if_lagg. Before this change if_lagg was using nonsleepable rmlocks to protect its internal state. This patch introduces another sx lock to protect code paths that require sleeping, while still uses old rmlock to protect hot nonsleepable data paths. This change allows to remove taskqueue decoupling used before to change interface addresses without holding the lock. Instead it uses sx lock to protect direct if_ioctl() calls. As another bonus, the new code synchronizes enabled capabilities of member interfaces, and allows to control them with ifconfig laggX, that was impossible before. This part should fix interoperation with if_bridge, that may need to disable some capabilities, such as TXCSUM or LRO, to allow bridging with noncapable interfaces. MFC after: 2 weeks Sponsored by: iXsystems, Inc. Differential Revision: https://reviews.freebsd.org/D10514
2017-05-02 19:09:11 +00:00
if (lp->lp_detaching == 0) {
lagg_setflags(lp, 0);
lagg: fix lagg ifioctl after SIOCSIFCAPNV Lagg was broken by SIOCSIFCAPNV when all underlying devices support SIOCSIFCAPNV. This change updates lagg to work with SIOCSIFCAPNV and if_capabilities2. Reviewed by: kib, hselasky Sponsored by: Netflix Differential Revision: https://reviews.freebsd.org/D35865
2022-07-28 14:36:22 +00:00
lagg_setcaps(lp, lp->lp_ifcapenable, lp->lp_ifcapenable2);
Add support for IP over infiniband, IPoIB, to lagg(4). Currently only the failover protocol is supported due to limitations in the IPoIB architecture. Refer to the lagg(4) manual page for how to configure and use this new feature. A new network interface type, IFT_INFINIBANDLAG, has been added, similar to the existing IFT_IEEE8023ADLAG . ifconfig(8) has been updated to accept a new laggtype argument when creating lagg(4) network interfaces. This new argument is used to distinguish between ethernet and infiniband type of lagg(4) network interface. The laggtype argument is optional and defaults to ethernet. The lagg(4) command line syntax is backwards compatible. Differential Revision: https://reviews.freebsd.org/D26254 Reviewed by: melifaro@ MFC after: 1 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2020-10-22 09:47:12 +00:00
if_setlladdr(ifp, lp->lp_lladdr, ifp->if_addrlen);
Introduce sleepable locks into if_lagg. Before this change if_lagg was using nonsleepable rmlocks to protect its internal state. This patch introduces another sx lock to protect code paths that require sleeping, while still uses old rmlock to protect hot nonsleepable data paths. This change allows to remove taskqueue decoupling used before to change interface addresses without holding the lock. Instead it uses sx lock to protect direct if_ioctl() calls. As another bonus, the new code synchronizes enabled capabilities of member interfaces, and allows to control them with ifconfig laggX, that was impossible before. This part should fix interoperation with if_bridge, that may need to disable some capabilities, such as TXCSUM or LRO, to allow bridging with noncapable interfaces. MFC after: 2 weeks Sponsored by: iXsystems, Inc. Differential Revision: https://reviews.freebsd.org/D10514
2017-05-02 19:09:11 +00:00
}
Replace rmlock with epoch in lagg Use the new epoch based reclamation API. Now the hot paths will not block at all, and the sx lock is used for the softc data. This fixes LORs reported where the rwlock was obtained when the sxlock was held. Submitted by: mmacy Reported by: Harry Schmalzbauer <freebsd@omnilan.de> Reviewed by: sbruno Sponsored by: Limelight Networks Differential Revision: https://reviews.freebsd.org/D15355
2018-05-14 20:06:49 +00:00
/*
* free port and release it's ifnet reference after a grace period has
* elapsed.
*/
Introduce NET_EPOCH_CALL() macro and use it everywhere where we free data based on the network epoch. The macro reverses the argument order of epoch_call(9) - first function, then its argument. NFC
2020-01-15 06:05:20 +00:00
NET_EPOCH_CALL(lagg_port_destroy_cb, &lp->lp_epoch_ctx);
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
/* Update lagg capabilities */
- Propagate the largest set of interface capabilities supported by all lagg ports to the lagg interface. - Use the MTU from the first interface as the lagg MTU, all extra interfaces must be the same. This fixes using a lagg interface for a vlan or enabling jumbo frames, etc. Approved by: re (kensmith) MFC After: 3 days
2007-07-30 20:17:22 +00:00
lagg_capabilities(sc);
Have the lagg interface generate link up/down events, the interface is marked as up if at least one of its ports also has a link up. This fixes using carp+lagg together and any other system that relies on linkstate events. PR: kern/113956 MFC after: 3 days
2007-11-25 06:30:46 +00:00
lagg_linkstate(sc);
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
return (0);
}
static int
lagg_port_ioctl(struct ifnet *ifp, u_long cmd, caddr_t data)
{
Convert LAGG_RLOCK() to NET_EPOCH_ENTER(). No functional changes.
2020-12-08 16:36:46 +00:00
struct epoch_tracker et;
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
struct lagg_reqport *rp = (struct lagg_reqport *)data;
struct lagg_softc *sc;
struct lagg_port *lp = NULL;
int error = 0;
/* Should be checked by the caller */
Add support for IP over infiniband, IPoIB, to lagg(4). Currently only the failover protocol is supported due to limitations in the IPoIB architecture. Refer to the lagg(4) manual page for how to configure and use this new feature. A new network interface type, IFT_INFINIBANDLAG, has been added, similar to the existing IFT_IEEE8023ADLAG . ifconfig(8) has been updated to accept a new laggtype argument when creating lagg(4) network interfaces. This new argument is used to distinguish between ethernet and infiniband type of lagg(4) network interface. The laggtype argument is optional and defaults to ethernet. The lagg(4) command line syntax is backwards compatible. Differential Revision: https://reviews.freebsd.org/D26254 Reviewed by: melifaro@ MFC after: 1 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2020-10-22 09:47:12 +00:00
switch (ifp->if_type) {
case IFT_IEEE8023ADLAG:
case IFT_INFINIBANDLAG:
if ((lp = ifp->if_lagg) == NULL || (sc = lp->lp_softc) == NULL)
goto fallback;
break;
default:
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
goto fallback;
Add support for IP over infiniband, IPoIB, to lagg(4). Currently only the failover protocol is supported due to limitations in the IPoIB architecture. Refer to the lagg(4) manual page for how to configure and use this new feature. A new network interface type, IFT_INFINIBANDLAG, has been added, similar to the existing IFT_IEEE8023ADLAG . ifconfig(8) has been updated to accept a new laggtype argument when creating lagg(4) network interfaces. This new argument is used to distinguish between ethernet and infiniband type of lagg(4) network interface. The laggtype argument is optional and defaults to ethernet. The lagg(4) command line syntax is backwards compatible. Differential Revision: https://reviews.freebsd.org/D26254 Reviewed by: melifaro@ MFC after: 1 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2020-10-22 09:47:12 +00:00
}
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
switch (cmd) {
case SIOCGLAGGPORT:
if (rp->rp_portname[0] == '\0' ||
ifunit(rp->rp_portname) != ifp) {
error = EINVAL;
break;
}
Convert LAGG_RLOCK() to NET_EPOCH_ENTER(). No functional changes.
2020-12-08 16:36:46 +00:00
NET_EPOCH_ENTER(et);
Avoid holding the softc lock when using copyout(). Reported by: dfr Approved by: re (rwatson)
2007-07-26 20:30:18 +00:00
if ((lp = ifp->if_lagg) == NULL || lp->lp_softc != sc) {
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
error = ENOENT;
Convert LAGG_RLOCK() to NET_EPOCH_ENTER(). No functional changes.
2020-12-08 16:36:46 +00:00
NET_EPOCH_EXIT(et);
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
break;
}
lagg_port2req(lp, rp);
Convert LAGG_RLOCK() to NET_EPOCH_ENTER(). No functional changes.
2020-12-08 16:36:46 +00:00
NET_EPOCH_EXIT(et);
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
break;
- Propagate the largest set of interface capabilities supported by all lagg ports to the lagg interface. - Use the MTU from the first interface as the lagg MTU, all extra interfaces must be the same. This fixes using a lagg interface for a vlan or enabling jumbo frames, etc. Approved by: re (kensmith) MFC After: 3 days
2007-07-30 20:17:22 +00:00
case SIOCSIFCAP:
lagg: fix lagg ifioctl after SIOCSIFCAPNV Lagg was broken by SIOCSIFCAPNV when all underlying devices support SIOCSIFCAPNV. This change updates lagg to work with SIOCSIFCAPNV and if_capabilities2. Reviewed by: kib, hselasky Sponsored by: Netflix Differential Revision: https://reviews.freebsd.org/D35865
2022-07-28 14:36:22 +00:00
case SIOCSIFCAPNV:
- Propagate the largest set of interface capabilities supported by all lagg ports to the lagg interface. - Use the MTU from the first interface as the lagg MTU, all extra interfaces must be the same. This fixes using a lagg interface for a vlan or enabling jumbo frames, etc. Approved by: re (kensmith) MFC After: 3 days
2007-07-30 20:17:22 +00:00
if (lp->lp_ioctl == NULL) {
error = EINVAL;
break;
}
error = (*lp->lp_ioctl)(ifp, cmd, data);
if (error)
break;
/* Update lagg interface capabilities */
Introduce sleepable locks into if_lagg. Before this change if_lagg was using nonsleepable rmlocks to protect its internal state. This patch introduces another sx lock to protect code paths that require sleeping, while still uses old rmlock to protect hot nonsleepable data paths. This change allows to remove taskqueue decoupling used before to change interface addresses without holding the lock. Instead it uses sx lock to protect direct if_ioctl() calls. As another bonus, the new code synchronizes enabled capabilities of member interfaces, and allows to control them with ifconfig laggX, that was impossible before. This part should fix interoperation with if_bridge, that may need to disable some capabilities, such as TXCSUM or LRO, to allow bridging with noncapable interfaces. MFC after: 2 weeks Sponsored by: iXsystems, Inc. Differential Revision: https://reviews.freebsd.org/D10514
2017-05-02 19:09:11 +00:00
LAGG_XLOCK(sc);
- Propagate the largest set of interface capabilities supported by all lagg ports to the lagg interface. - Use the MTU from the first interface as the lagg MTU, all extra interfaces must be the same. This fixes using a lagg interface for a vlan or enabling jumbo frames, etc. Approved by: re (kensmith) MFC After: 3 days
2007-07-30 20:17:22 +00:00
lagg_capabilities(sc);
Introduce sleepable locks into if_lagg. Before this change if_lagg was using nonsleepable rmlocks to protect its internal state. This patch introduces another sx lock to protect code paths that require sleeping, while still uses old rmlock to protect hot nonsleepable data paths. This change allows to remove taskqueue decoupling used before to change interface addresses without holding the lock. Instead it uses sx lock to protect direct if_ioctl() calls. As another bonus, the new code synchronizes enabled capabilities of member interfaces, and allows to control them with ifconfig laggX, that was impossible before. This part should fix interoperation with if_bridge, that may need to disable some capabilities, such as TXCSUM or LRO, to allow bridging with noncapable interfaces. MFC after: 2 weeks Sponsored by: iXsystems, Inc. Differential Revision: https://reviews.freebsd.org/D10514
2017-05-02 19:09:11 +00:00
LAGG_XUNLOCK(sc);
Call VLAN_CAPABILITIES() when LAGG capabilities change. This makes VLAN on top of LAGG to expose proper capabilities if they are changed after creation. MFC after: 1 week
2017-05-26 22:22:48 +00:00
VLAN_CAPABILITIES(sc->sc_ifp);
- Propagate the largest set of interface capabilities supported by all lagg ports to the lagg interface. - Use the MTU from the first interface as the lagg MTU, all extra interfaces must be the same. This fixes using a lagg interface for a vlan or enabling jumbo frames, etc. Approved by: re (kensmith) MFC After: 3 days
2007-07-30 20:17:22 +00:00
break;
case SIOCSIFMTU:
/* Do not allow the MTU to be changed once joined */
error = EINVAL;
break;
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
default:
goto fallback;
}
return (error);
fallback:
Fix panic during lagg destruction with simultaneous status check If you run "ifconfig lagg0 destroy" and "ifconfig lagg0" at the same time a page fault may result. The first process will destroy ifp->if_lagg in lagg_clone_destroy (called by if_clone_destroy). Then the second process will observe that ifp->if_lagg is NULL at the top of lagg_port_ioctl and goto fallback: where it will promptly dereference ifp->if_lagg anyway. The solution is to repeat the NULL check for ifp->if_lagg MFC after: 4 weeks Sponsored by: Spectra Logic Corp Differential Revision: https://reviews.freebsd.org/D8512
2016-12-16 22:39:30 +00:00
if (lp != NULL && lp->lp_ioctl != NULL)
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
return ((*lp->lp_ioctl)(ifp, cmd, data));
return (EINVAL);
}
Use underlying ports counters to get lagg statistics instead of per-packet accounting. This introduce user-visible changes like aggregating error counters. Reviewed by: asomers (prev.version), glebius CR: D781 MFC after: 2 weeks Sponsored by: Yandex LLC
2014-09-27 13:57:48 +00:00
/*
* Requests counter @cnt data.
*
* Counter value is calculated the following way:
* 1) for each port, sum difference between current and "initial" measurements.
* 2) add lagg logical interface counters.
* 3) add data from detached_counters array.
*
* We also do the following things on ports attach/detach:
* 1) On port attach we store all counters it has into port_counter array.
* 2) On port detach we add the different between "initial" and
* current counters data to detached_counters array.
*/
static uint64_t
lagg_get_counter(struct ifnet *ifp, ift_counter cnt)
{
Convert LAGG_RLOCK() to NET_EPOCH_ENTER(). No functional changes.
2020-12-08 16:36:46 +00:00
struct epoch_tracker et;
Use underlying ports counters to get lagg statistics instead of per-packet accounting. This introduce user-visible changes like aggregating error counters. Reviewed by: asomers (prev.version), glebius CR: D781 MFC after: 2 weeks Sponsored by: Yandex LLC
2014-09-27 13:57:48 +00:00
struct lagg_softc *sc;
struct lagg_port *lp;
struct ifnet *lpifp;
uint64_t newval, oldval, vsum;
Finally, convert counters in struct ifnet to counter(9). Sponsored by: Netflix Sponsored by: Nginx, Inc.
2014-09-28 08:57:07 +00:00
/* Revise this when we've got non-generic counters. */
KASSERT(cnt < IFCOUNTERS, ("%s: invalid cnt %d", __func__, cnt));
Use underlying ports counters to get lagg statistics instead of per-packet accounting. This introduce user-visible changes like aggregating error counters. Reviewed by: asomers (prev.version), glebius CR: D781 MFC after: 2 weeks Sponsored by: Yandex LLC
2014-09-27 13:57:48 +00:00
sc = (struct lagg_softc *)ifp->if_softc;
vsum = 0;
Convert LAGG_RLOCK() to NET_EPOCH_ENTER(). No functional changes.
2020-12-08 16:36:46 +00:00
NET_EPOCH_ENTER(et);
Replace rmlock with epoch in lagg Use the new epoch based reclamation API. Now the hot paths will not block at all, and the sx lock is used for the softc data. This fixes LORs reported where the rwlock was obtained when the sxlock was held. Submitted by: mmacy Reported by: Harry Schmalzbauer <freebsd@omnilan.de> Reviewed by: sbruno Sponsored by: Limelight Networks Differential Revision: https://reviews.freebsd.org/D15355
2018-05-14 20:06:49 +00:00
CK_SLIST_FOREACH(lp, &sc->sc_ports, lp_entries) {
Use underlying ports counters to get lagg statistics instead of per-packet accounting. This introduce user-visible changes like aggregating error counters. Reviewed by: asomers (prev.version), glebius CR: D781 MFC after: 2 weeks Sponsored by: Yandex LLC
2014-09-27 13:57:48 +00:00
/* Saved attached value */
Finally, convert counters in struct ifnet to counter(9). Sponsored by: Netflix Sponsored by: Nginx, Inc.
2014-09-28 08:57:07 +00:00
oldval = lp->port_counters.val[cnt];
Use underlying ports counters to get lagg statistics instead of per-packet accounting. This introduce user-visible changes like aggregating error counters. Reviewed by: asomers (prev.version), glebius CR: D781 MFC after: 2 weeks Sponsored by: Yandex LLC
2014-09-27 13:57:48 +00:00
/* current value */
lpifp = lp->lp_ifp;
newval = lpifp->if_get_counter(lpifp, cnt);
/* Calculate diff and save new */
vsum += newval - oldval;
}
Convert LAGG_RLOCK() to NET_EPOCH_ENTER(). No functional changes.
2020-12-08 16:36:46 +00:00
NET_EPOCH_EXIT(et);
Use underlying ports counters to get lagg statistics instead of per-packet accounting. This introduce user-visible changes like aggregating error counters. Reviewed by: asomers (prev.version), glebius CR: D781 MFC after: 2 weeks Sponsored by: Yandex LLC
2014-09-27 13:57:48 +00:00
/*
* Add counter data which might be added by upper
* layer protocols operating on logical interface.
*/
vsum += if_get_counter_default(ifp, cnt);
/*
* Add counter data from detached ports counters
*/
Finally, convert counters in struct ifnet to counter(9). Sponsored by: Netflix Sponsored by: Nginx, Inc.
2014-09-28 08:57:07 +00:00
vsum += sc->detached_counters.val[cnt];
Use underlying ports counters to get lagg statistics instead of per-packet accounting. This introduce user-visible changes like aggregating error counters. Reviewed by: asomers (prev.version), glebius CR: D781 MFC after: 2 weeks Sponsored by: Yandex LLC
2014-09-27 13:57:48 +00:00
return (vsum);
}
Relax restriction on direct tx to child ports Lagg(4) restricts the type of packet that may be sent directly to a child port, to avoid undesired output from accidental misconfiguration. Previously only ETHERTYPE_PAE was permitted. BPF writes to a lagg(4) child port are presumably intentional, so just allow them, while still blocking other packets that should take the aggregation path. PR: kern/138620 Approved by: thompsa@
2012-05-03 01:41:12 +00:00
/*
* For direct output to child ports.
*/
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
static int
lagg_port_output(struct ifnet *ifp, struct mbuf *m,
Add const qualifier to the dst parameter of the ifnet if_output method.
2013-04-26 12:50:32 +00:00
const struct sockaddr *dst, struct route *ro)
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
{
struct lagg_port *lp = ifp->if_lagg;
switch (dst->sa_family) {
case pseudo_AF_HDRCMPLT:
case AF_UNSPEC:
Fix possible NULL pointer dereference. lagg(4) replaces if_output method of its child interfaces and expects that this method can be called only by child interfaces. But it is possible that lagg_port_output() could be called by children of child interfaces. In this case ifnet's if_lagg field is NULL. Add check that lp is not NULL. Obtained from: Yandex LLC MFC after: 1 week Sponsored by: Yandex LLC
2020-11-11 15:53:36 +00:00
if (lp != NULL)
return ((*lp->lp_output)(ifp, m, dst, ro));
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
}
/* drop any other frames */
m_freem(m);
If lagg(4) can't forward a packet due to underlying port problems, return much more meaningful ENETDOWN to the stack, instead of EBUSY.
2013-01-21 08:59:31 +00:00
return (ENETDOWN);
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
}
static void
lagg_port_ifdetach(void *arg __unused, struct ifnet *ifp)
{
struct lagg_port *lp;
struct lagg_softc *sc;
if ((lp = ifp->if_lagg) == NULL)
return;
Add the same check as vlan(4) where we ignore the ifnet departure event if the interface is just being renamed. PR: kern/169557 Submitted by: Mark Johnston MFC after: 3 days
2012-06-30 19:09:02 +00:00
/* If the ifnet is just being renamed, don't do anything. */
if (ifp->if_flags & IFF_RENAMING)
return;
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
non-functional cleanup - remove dead code - use consistent variable names - gc unused defines - whitespace cleanup
2007-06-12 07:29:11 +00:00
sc = lp->lp_softc;
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
Introduce sleepable locks into if_lagg. Before this change if_lagg was using nonsleepable rmlocks to protect its internal state. This patch introduces another sx lock to protect code paths that require sleeping, while still uses old rmlock to protect hot nonsleepable data paths. This change allows to remove taskqueue decoupling used before to change interface addresses without holding the lock. Instead it uses sx lock to protect direct if_ioctl() calls. As another bonus, the new code synchronizes enabled capabilities of member interfaces, and allows to control them with ifconfig laggX, that was impossible before. This part should fix interoperation with if_bridge, that may need to disable some capabilities, such as TXCSUM or LRO, to allow bridging with noncapable interfaces. MFC after: 2 weeks Sponsored by: iXsystems, Inc. Differential Revision: https://reviews.freebsd.org/D10514
2017-05-02 19:09:11 +00:00
LAGG_XLOCK(sc);
lp->lp_detaching = 1;
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
lagg_port_destroy(lp, 1);
Introduce sleepable locks into if_lagg. Before this change if_lagg was using nonsleepable rmlocks to protect its internal state. This patch introduces another sx lock to protect code paths that require sleeping, while still uses old rmlock to protect hot nonsleepable data paths. This change allows to remove taskqueue decoupling used before to change interface addresses without holding the lock. Instead it uses sx lock to protect direct if_ioctl() calls. As another bonus, the new code synchronizes enabled capabilities of member interfaces, and allows to control them with ifconfig laggX, that was impossible before. This part should fix interoperation with if_bridge, that may need to disable some capabilities, such as TXCSUM or LRO, to allow bridging with noncapable interfaces. MFC after: 2 weeks Sponsored by: iXsystems, Inc. Differential Revision: https://reviews.freebsd.org/D10514
2017-05-02 19:09:11 +00:00
LAGG_XUNLOCK(sc);
Call VLAN_CAPABILITIES() when LAGG capabilities change. This makes VLAN on top of LAGG to expose proper capabilities if they are changed after creation. MFC after: 1 week
2017-05-26 22:22:48 +00:00
VLAN_CAPABILITIES(sc->sc_ifp);
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
}
static void
lagg_port2req(struct lagg_port *lp, struct lagg_reqport *rp)
{
non-functional cleanup - remove dead code - use consistent variable names - gc unused defines - whitespace cleanup
2007-06-12 07:29:11 +00:00
struct lagg_softc *sc = lp->lp_softc;
Show the ACTIVE flag in ifconfig for the single interface that is actaully active in failover mode rather than all interfaces with a link. This makes it clear if the master interface is in use or one of the backup links. Found by: Writing the Handbook section Approved by: re (kensmith)
2007-08-30 19:12:10 +00:00
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
strlcpy(rp->rp_ifname, sc->sc_ifname, sizeof(rp->rp_ifname));
strlcpy(rp->rp_portname, lp->lp_ifp->if_xname, sizeof(rp->rp_portname));
rp->rp_prio = lp->lp_prio;
rp->rp_flags = lp->lp_flags;
Make all lagg protocol methods live in lagg_protos, not in softc. All interfaces of a same protocol, use the same methods. Sponsored by: Netflix Sponsored by: Nginx, Inc.
2014-09-26 12:54:24 +00:00
lagg_proto_portreq(sc, lp, &rp->rp_psc);
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
/* Add protocol specific flags */
switch (sc->sc_proto) {
case LAGG_PROTO_FAILOVER:
if (lp == sc->sc_primary)
Set the master flag on the right variable.
2007-05-02 08:58:28 +00:00
rp->rp_flags |= LAGG_PORT_MASTER;
Show the ACTIVE flag in ifconfig for the single interface that is actaully active in failover mode rather than all interfaces with a link. This makes it clear if the master interface is in use or one of the backup links. Found by: Writing the Handbook section Approved by: re (kensmith)
2007-08-30 19:12:10 +00:00
if (lp == lagg_link_active(sc, sc->sc_primary))
rp->rp_flags |= LAGG_PORT_ACTIVE;
break;
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
case LAGG_PROTO_ROUNDROBIN:
case LAGG_PROTO_LOADBALANCE:
- Whitespace. - Remove caddr_t.
2014-09-26 12:35:58 +00:00
case LAGG_PROTO_BROADCAST:
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
if (LAGG_PORTACTIVE(lp))
rp->rp_flags |= LAGG_PORT_ACTIVE;
break;
case LAGG_PROTO_LACP:
/* LACP has a different definition of active */
Switch the LACP state machine over to its own mutex to protect the internals, this means that it no longer grabs the lagg rwlock. Use two port table arrays which list the active ports for Tx and switch between them with an atomic op. Now the lagg rwlock is only exclusively locked for management (ioctls) and queuing of lacp control frames isnt needed.
2008-03-16 19:25:30 +00:00
if (lacp_isactive(lp))
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
rp->rp_flags |= LAGG_PORT_ACTIVE;
Switch the LACP state machine over to its own mutex to protect the internals, this means that it no longer grabs the lagg rwlock. Use two port table arrays which list the active ports for Tx and switch between them with an atomic op. Now the lagg rwlock is only exclusively locked for management (ioctls) and queuing of lacp control frames isnt needed.
2008-03-16 19:25:30 +00:00
if (lacp_iscollecting(lp))
rp->rp_flags |= LAGG_PORT_COLLECTING;
if (lacp_isdistributing(lp))
rp->rp_flags |= LAGG_PORT_DISTRIBUTING;
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
break;
}
}
Add support for IP over infiniband, IPoIB, to lagg(4). Currently only the failover protocol is supported due to limitations in the IPoIB architecture. Refer to the lagg(4) manual page for how to configure and use this new feature. A new network interface type, IFT_INFINIBANDLAG, has been added, similar to the existing IFT_IEEE8023ADLAG . ifconfig(8) has been updated to accept a new laggtype argument when creating lagg(4) network interfaces. This new argument is used to distinguish between ethernet and infiniband type of lagg(4) network interface. The laggtype argument is optional and defaults to ethernet. The lagg(4) command line syntax is backwards compatible. Differential Revision: https://reviews.freebsd.org/D26254 Reviewed by: melifaro@ MFC after: 1 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2020-10-22 09:47:12 +00:00
static void
lagg_watchdog_infiniband(void *arg)
{
Convert LAGG_RLOCK() to NET_EPOCH_ENTER(). No functional changes.
2020-12-08 16:36:46 +00:00
struct epoch_tracker et;
Add support for IP over infiniband, IPoIB, to lagg(4). Currently only the failover protocol is supported due to limitations in the IPoIB architecture. Refer to the lagg(4) manual page for how to configure and use this new feature. A new network interface type, IFT_INFINIBANDLAG, has been added, similar to the existing IFT_IEEE8023ADLAG . ifconfig(8) has been updated to accept a new laggtype argument when creating lagg(4) network interfaces. This new argument is used to distinguish between ethernet and infiniband type of lagg(4) network interface. The laggtype argument is optional and defaults to ethernet. The lagg(4) command line syntax is backwards compatible. Differential Revision: https://reviews.freebsd.org/D26254 Reviewed by: melifaro@ MFC after: 1 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2020-10-22 09:47:12 +00:00
struct lagg_softc *sc;
struct lagg_port *lp;
struct ifnet *ifp;
struct ifnet *lp_ifp;
sc = arg;
/*
* Because infiniband nodes have a fixed MAC address, which is
* generated by the so-called GID, we need to regularly update
* the link level address of the parent lagg<N> device when
* the active port changes. Possibly we could piggy-back on
* link up/down events aswell, but using a timer also provides
* a guarantee against too frequent events. This operation
* does not have to be atomic.
*/
Convert LAGG_RLOCK() to NET_EPOCH_ENTER(). No functional changes.
2020-12-08 16:36:46 +00:00
NET_EPOCH_ENTER(et);
Add support for IP over infiniband, IPoIB, to lagg(4). Currently only the failover protocol is supported due to limitations in the IPoIB architecture. Refer to the lagg(4) manual page for how to configure and use this new feature. A new network interface type, IFT_INFINIBANDLAG, has been added, similar to the existing IFT_IEEE8023ADLAG . ifconfig(8) has been updated to accept a new laggtype argument when creating lagg(4) network interfaces. This new argument is used to distinguish between ethernet and infiniband type of lagg(4) network interface. The laggtype argument is optional and defaults to ethernet. The lagg(4) command line syntax is backwards compatible. Differential Revision: https://reviews.freebsd.org/D26254 Reviewed by: melifaro@ MFC after: 1 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2020-10-22 09:47:12 +00:00
lp = lagg_link_active(sc, sc->sc_primary);
if (lp != NULL) {
ifp = sc->sc_ifp;
lp_ifp = lp->lp_ifp;
if (ifp != NULL && lp_ifp != NULL &&
Fix for IPoIB over lagg(4). Need to update both link layer address and broadcast address when active link changes for IP over infiniband. This is because the broadcast address contains the so-called P-key, which is interface dependent. Reviewed by: kib @ Differential Revision: https://reviews.freebsd.org/D27658 MFC after: 1 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2020-12-29 16:34:01 +00:00
(memcmp(IF_LLADDR(ifp), IF_LLADDR(lp_ifp), ifp->if_addrlen) != 0 ||
memcmp(sc->sc_bcast_addr, lp_ifp->if_broadcastaddr, ifp->if_addrlen) != 0)) {
Add support for IP over infiniband, IPoIB, to lagg(4). Currently only the failover protocol is supported due to limitations in the IPoIB architecture. Refer to the lagg(4) manual page for how to configure and use this new feature. A new network interface type, IFT_INFINIBANDLAG, has been added, similar to the existing IFT_IEEE8023ADLAG . ifconfig(8) has been updated to accept a new laggtype argument when creating lagg(4) network interfaces. This new argument is used to distinguish between ethernet and infiniband type of lagg(4) network interface. The laggtype argument is optional and defaults to ethernet. The lagg(4) command line syntax is backwards compatible. Differential Revision: https://reviews.freebsd.org/D26254 Reviewed by: melifaro@ MFC after: 1 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2020-10-22 09:47:12 +00:00
memcpy(IF_LLADDR(ifp), IF_LLADDR(lp_ifp), ifp->if_addrlen);
Fix for IPoIB over lagg(4). Need to update both link layer address and broadcast address when active link changes for IP over infiniband. This is because the broadcast address contains the so-called P-key, which is interface dependent. Reviewed by: kib @ Differential Revision: https://reviews.freebsd.org/D27658 MFC after: 1 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2020-12-29 16:34:01 +00:00
memcpy(sc->sc_bcast_addr, lp_ifp->if_broadcastaddr, ifp->if_addrlen);
Add support for IP over infiniband, IPoIB, to lagg(4). Currently only the failover protocol is supported due to limitations in the IPoIB architecture. Refer to the lagg(4) manual page for how to configure and use this new feature. A new network interface type, IFT_INFINIBANDLAG, has been added, similar to the existing IFT_IEEE8023ADLAG . ifconfig(8) has been updated to accept a new laggtype argument when creating lagg(4) network interfaces. This new argument is used to distinguish between ethernet and infiniband type of lagg(4) network interface. The laggtype argument is optional and defaults to ethernet. The lagg(4) command line syntax is backwards compatible. Differential Revision: https://reviews.freebsd.org/D26254 Reviewed by: melifaro@ MFC after: 1 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2020-10-22 09:47:12 +00:00
CURVNET_SET(ifp->if_vnet);
EVENTHANDLER_INVOKE(iflladdr_event, ifp);
CURVNET_RESTORE();
}
}
Convert LAGG_RLOCK() to NET_EPOCH_ENTER(). No functional changes.
2020-12-08 16:36:46 +00:00
NET_EPOCH_EXIT(et);
Add support for IP over infiniband, IPoIB, to lagg(4). Currently only the failover protocol is supported due to limitations in the IPoIB architecture. Refer to the lagg(4) manual page for how to configure and use this new feature. A new network interface type, IFT_INFINIBANDLAG, has been added, similar to the existing IFT_IEEE8023ADLAG . ifconfig(8) has been updated to accept a new laggtype argument when creating lagg(4) network interfaces. This new argument is used to distinguish between ethernet and infiniband type of lagg(4) network interface. The laggtype argument is optional and defaults to ethernet. The lagg(4) command line syntax is backwards compatible. Differential Revision: https://reviews.freebsd.org/D26254 Reviewed by: melifaro@ MFC after: 1 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2020-10-22 09:47:12 +00:00
callout_reset(&sc->sc_watchdog, hz, &lagg_watchdog_infiniband, arg);
}
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
static void
lagg_init(void *xsc)
{
struct lagg_softc *sc = (struct lagg_softc *)xsc;
struct ifnet *ifp = sc->sc_ifp;
Fix lladdr change propagation for on vlans on top of it. Fix lladdr update when setting mac address manually. Fix lladdr_event for slave ports addition. MFC after: 4 weeks Sponsored by: Yandex LLC Differential Revision: https://reviews.freebsd.org/D4004
2015-11-01 19:59:04 +00:00
struct lagg_port *lp;
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
Introduce sleepable locks into if_lagg. Before this change if_lagg was using nonsleepable rmlocks to protect its internal state. This patch introduces another sx lock to protect code paths that require sleeping, while still uses old rmlock to protect hot nonsleepable data paths. This change allows to remove taskqueue decoupling used before to change interface addresses without holding the lock. Instead it uses sx lock to protect direct if_ioctl() calls. As another bonus, the new code synchronizes enabled capabilities of member interfaces, and allows to control them with ifconfig laggX, that was impossible before. This part should fix interoperation with if_bridge, that may need to disable some capabilities, such as TXCSUM or LRO, to allow bridging with noncapable interfaces. MFC after: 2 weeks Sponsored by: iXsystems, Inc. Differential Revision: https://reviews.freebsd.org/D10514
2017-05-02 19:09:11 +00:00
LAGG_XLOCK(sc);
if (ifp->if_drv_flags & IFF_DRV_RUNNING) {
LAGG_XUNLOCK(sc);
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
return;
Introduce sleepable locks into if_lagg. Before this change if_lagg was using nonsleepable rmlocks to protect its internal state. This patch introduces another sx lock to protect code paths that require sleeping, while still uses old rmlock to protect hot nonsleepable data paths. This change allows to remove taskqueue decoupling used before to change interface addresses without holding the lock. Instead it uses sx lock to protect direct if_ioctl() calls. As another bonus, the new code synchronizes enabled capabilities of member interfaces, and allows to control them with ifconfig laggX, that was impossible before. This part should fix interoperation with if_bridge, that may need to disable some capabilities, such as TXCSUM or LRO, to allow bridging with noncapable interfaces. MFC after: 2 weeks Sponsored by: iXsystems, Inc. Differential Revision: https://reviews.freebsd.org/D10514
2017-05-02 19:09:11 +00:00
}
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
ifp->if_drv_flags |= IFF_DRV_RUNNING;
Fix lladdr change propagation for on vlans on top of it. Fix lladdr update when setting mac address manually. Fix lladdr_event for slave ports addition. MFC after: 4 weeks Sponsored by: Yandex LLC Differential Revision: https://reviews.freebsd.org/D4004
2015-11-01 19:59:04 +00:00
/*
* Update the port lladdrs if needed.
* This might be if_setlladdr() notification
* that lladdr has been changed.
*/
Replace rmlock with epoch in lagg Use the new epoch based reclamation API. Now the hot paths will not block at all, and the sx lock is used for the softc data. This fixes LORs reported where the rwlock was obtained when the sxlock was held. Submitted by: mmacy Reported by: Harry Schmalzbauer <freebsd@omnilan.de> Reviewed by: sbruno Sponsored by: Limelight Networks Differential Revision: https://reviews.freebsd.org/D15355
2018-05-14 20:06:49 +00:00
CK_SLIST_FOREACH(lp, &sc->sc_ports, lp_entries) {
Introduce sleepable locks into if_lagg. Before this change if_lagg was using nonsleepable rmlocks to protect its internal state. This patch introduces another sx lock to protect code paths that require sleeping, while still uses old rmlock to protect hot nonsleepable data paths. This change allows to remove taskqueue decoupling used before to change interface addresses without holding the lock. Instead it uses sx lock to protect direct if_ioctl() calls. As another bonus, the new code synchronizes enabled capabilities of member interfaces, and allows to control them with ifconfig laggX, that was impossible before. This part should fix interoperation with if_bridge, that may need to disable some capabilities, such as TXCSUM or LRO, to allow bridging with noncapable interfaces. MFC after: 2 weeks Sponsored by: iXsystems, Inc. Differential Revision: https://reviews.freebsd.org/D10514
2017-05-02 19:09:11 +00:00
if (memcmp(IF_LLADDR(ifp), IF_LLADDR(lp->lp_ifp),
Add support for IP over infiniband, IPoIB, to lagg(4). Currently only the failover protocol is supported due to limitations in the IPoIB architecture. Refer to the lagg(4) manual page for how to configure and use this new feature. A new network interface type, IFT_INFINIBANDLAG, has been added, similar to the existing IFT_IEEE8023ADLAG . ifconfig(8) has been updated to accept a new laggtype argument when creating lagg(4) network interfaces. This new argument is used to distinguish between ethernet and infiniband type of lagg(4) network interface. The laggtype argument is optional and defaults to ethernet. The lagg(4) command line syntax is backwards compatible. Differential Revision: https://reviews.freebsd.org/D26254 Reviewed by: melifaro@ MFC after: 1 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2020-10-22 09:47:12 +00:00
ifp->if_addrlen) != 0)
if_setlladdr(lp->lp_ifp, IF_LLADDR(ifp), ifp->if_addrlen);
Introduce sleepable locks into if_lagg. Before this change if_lagg was using nonsleepable rmlocks to protect its internal state. This patch introduces another sx lock to protect code paths that require sleeping, while still uses old rmlock to protect hot nonsleepable data paths. This change allows to remove taskqueue decoupling used before to change interface addresses without holding the lock. Instead it uses sx lock to protect direct if_ioctl() calls. As another bonus, the new code synchronizes enabled capabilities of member interfaces, and allows to control them with ifconfig laggX, that was impossible before. This part should fix interoperation with if_bridge, that may need to disable some capabilities, such as TXCSUM or LRO, to allow bridging with noncapable interfaces. MFC after: 2 weeks Sponsored by: iXsystems, Inc. Differential Revision: https://reviews.freebsd.org/D10514
2017-05-02 19:09:11 +00:00
}
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
Make all lagg protocol methods live in lagg_protos, not in softc. All interfaces of a same protocol, use the same methods. Sponsored by: Netflix Sponsored by: Nginx, Inc.
2014-09-26 12:54:24 +00:00
lagg_proto_init(sc);
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
Add support for IP over infiniband, IPoIB, to lagg(4). Currently only the failover protocol is supported due to limitations in the IPoIB architecture. Refer to the lagg(4) manual page for how to configure and use this new feature. A new network interface type, IFT_INFINIBANDLAG, has been added, similar to the existing IFT_IEEE8023ADLAG . ifconfig(8) has been updated to accept a new laggtype argument when creating lagg(4) network interfaces. This new argument is used to distinguish between ethernet and infiniband type of lagg(4) network interface. The laggtype argument is optional and defaults to ethernet. The lagg(4) command line syntax is backwards compatible. Differential Revision: https://reviews.freebsd.org/D26254 Reviewed by: melifaro@ MFC after: 1 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2020-10-22 09:47:12 +00:00
if (ifp->if_type == IFT_INFINIBAND) {
mtx_lock(&sc->sc_mtx);
lagg_watchdog_infiniband(sc);
mtx_unlock(&sc->sc_mtx);
}
Introduce sleepable locks into if_lagg. Before this change if_lagg was using nonsleepable rmlocks to protect its internal state. This patch introduces another sx lock to protect code paths that require sleeping, while still uses old rmlock to protect hot nonsleepable data paths. This change allows to remove taskqueue decoupling used before to change interface addresses without holding the lock. Instead it uses sx lock to protect direct if_ioctl() calls. As another bonus, the new code synchronizes enabled capabilities of member interfaces, and allows to control them with ifconfig laggX, that was impossible before. This part should fix interoperation with if_bridge, that may need to disable some capabilities, such as TXCSUM or LRO, to allow bridging with noncapable interfaces. MFC after: 2 weeks Sponsored by: iXsystems, Inc. Differential Revision: https://reviews.freebsd.org/D10514
2017-05-02 19:09:11 +00:00
LAGG_XUNLOCK(sc);
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
}
static void
lagg_stop(struct lagg_softc *sc)
{
struct ifnet *ifp = sc->sc_ifp;
Introduce sleepable locks into if_lagg. Before this change if_lagg was using nonsleepable rmlocks to protect its internal state. This patch introduces another sx lock to protect code paths that require sleeping, while still uses old rmlock to protect hot nonsleepable data paths. This change allows to remove taskqueue decoupling used before to change interface addresses without holding the lock. Instead it uses sx lock to protect direct if_ioctl() calls. As another bonus, the new code synchronizes enabled capabilities of member interfaces, and allows to control them with ifconfig laggX, that was impossible before. This part should fix interoperation with if_bridge, that may need to disable some capabilities, such as TXCSUM or LRO, to allow bridging with noncapable interfaces. MFC after: 2 weeks Sponsored by: iXsystems, Inc. Differential Revision: https://reviews.freebsd.org/D10514
2017-05-02 19:09:11 +00:00
LAGG_XLOCK_ASSERT(sc);
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
if ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0)
return;
ifp->if_drv_flags &= ~IFF_DRV_RUNNING;
Make all lagg protocol methods live in lagg_protos, not in softc. All interfaces of a same protocol, use the same methods. Sponsored by: Netflix Sponsored by: Nginx, Inc.
2014-09-26 12:54:24 +00:00
lagg_proto_stop(sc);
Add support for IP over infiniband, IPoIB, to lagg(4). Currently only the failover protocol is supported due to limitations in the IPoIB architecture. Refer to the lagg(4) manual page for how to configure and use this new feature. A new network interface type, IFT_INFINIBANDLAG, has been added, similar to the existing IFT_IEEE8023ADLAG . ifconfig(8) has been updated to accept a new laggtype argument when creating lagg(4) network interfaces. This new argument is used to distinguish between ethernet and infiniband type of lagg(4) network interface. The laggtype argument is optional and defaults to ethernet. The lagg(4) command line syntax is backwards compatible. Differential Revision: https://reviews.freebsd.org/D26254 Reviewed by: melifaro@ MFC after: 1 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2020-10-22 09:47:12 +00:00
mtx_lock(&sc->sc_mtx);
callout_stop(&sc->sc_watchdog);
mtx_unlock(&sc->sc_mtx);
callout_drain(&sc->sc_watchdog);
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
}
static int
lagg_ioctl(struct ifnet *ifp, u_long cmd, caddr_t data)
{
Convert LAGG_RLOCK() to NET_EPOCH_ENTER(). No functional changes.
2020-12-08 16:36:46 +00:00
struct epoch_tracker et;
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
struct lagg_softc *sc = (struct lagg_softc *)ifp->if_softc;
struct lagg_reqall *ra = (struct lagg_reqall *)data;
Separate option handling from SIOC[SG]LAGG to SIOC[SG]LAGGOPTS for backward compatibility with old ifconfig(8).
2014-10-02 20:01:13 +00:00
struct lagg_reqopts *ro = (struct lagg_reqopts *)data;
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
struct lagg_reqport *rp = (struct lagg_reqport *)data, rpbuf;
Add the ability to set which packet layers are used for the load balance hash calculation.
2012-03-06 22:58:13 +00:00
struct lagg_reqflags *rf = (struct lagg_reqflags *)data;
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
struct ifreq *ifr = (struct ifreq *)data;
struct lagg_port *lp;
struct ifnet *tpif;
struct thread *td = curthread;
Avoid holding the softc lock when using copyout(). Reported by: dfr Approved by: re (rwatson)
2007-07-26 20:30:18 +00:00
char *buf, *outbuf;
Update route MTUs for bridge, lagg and vlan interfaces. Those interfaces may implicitly change their MTU on addition of parent interface in addition to normal SIOCSIFMTU ioctl path, where the route MTUs are updated normally. MFC after: 2 weeks Sponsored by: iXsystems, Inc.
2020-01-22 20:36:45 +00:00
int count, buflen, len, error = 0, oldmtu;
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
bzero(&rpbuf, sizeof(rpbuf));
lagg: Avoid adding a port to a lagg device being destroyed. The lagg_clone_destroy() handles detach and waiting for ifconfig callers to drain already. This narrows the race for 2 panics that the tests triggered. Both were a consequence of adding a port to the lagg device after it had already detached from all of its ports. The link state task would run after lagg_clone_destroy() free'd the lagg softc. kernel:trap_fatal+0xa4 kernel:trap_pfault+0x61 kernel:trap+0x316 kernel:witness_checkorder+0x6d kernel:_sx_xlock+0x72 if_lagg.ko:lagg_port_state+0x3b kernel:if_down+0x144 kernel:if_detach+0x659 if_tap.ko:tap_destroy+0x46 kernel:if_clone_destroyif+0x1b7 kernel:if_clone_destroy+0x8d kernel:ifioctl+0x29c kernel:kern_ioctl+0x2bd kernel:sys_ioctl+0x16d kernel:amd64_syscall+0x337 kernel:trap_fatal+0xa4 kernel:trap_pfault+0x61 kernel:trap+0x316 kernel:witness_checkorder+0x6d kernel:_sx_xlock+0x72 if_lagg.ko:lagg_port_state+0x3b kernel:do_link_state_change+0x9b kernel:taskqueue_run_locked+0x10b kernel:taskqueue_run+0x49 kernel:ithread_loop+0x19c kernel:fork_exit+0x83 PR: 244168 Reviewed by: markj MFC after: 2 weeks Sponsored by: Dell EMC Differential Revision: https://reviews.freebsd.org/D25284
2020-08-13 22:06:27 +00:00
/* XXX: This can race with lagg_clone_destroy. */
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
switch (cmd) {
case SIOCGLAGG:
Replace rmlock with epoch in lagg Use the new epoch based reclamation API. Now the hot paths will not block at all, and the sx lock is used for the softc data. This fixes LORs reported where the rwlock was obtained when the sxlock was held. Submitted by: mmacy Reported by: Harry Schmalzbauer <freebsd@omnilan.de> Reviewed by: sbruno Sponsored by: Limelight Networks Differential Revision: https://reviews.freebsd.org/D15355
2018-05-14 20:06:49 +00:00
LAGG_XLOCK(sc);
Introduce sleepable locks into if_lagg. Before this change if_lagg was using nonsleepable rmlocks to protect its internal state. This patch introduces another sx lock to protect code paths that require sleeping, while still uses old rmlock to protect hot nonsleepable data paths. This change allows to remove taskqueue decoupling used before to change interface addresses without holding the lock. Instead it uses sx lock to protect direct if_ioctl() calls. As another bonus, the new code synchronizes enabled capabilities of member interfaces, and allows to control them with ifconfig laggX, that was impossible before. This part should fix interoperation with if_bridge, that may need to disable some capabilities, such as TXCSUM or LRO, to allow bridging with noncapable interfaces. MFC after: 2 weeks Sponsored by: iXsystems, Inc. Differential Revision: https://reviews.freebsd.org/D10514
2017-05-02 19:09:11 +00:00
buflen = sc->sc_count * sizeof(struct lagg_reqport);
Avoid holding the softc lock when using copyout(). Reported by: dfr Approved by: re (rwatson)
2007-07-26 20:30:18 +00:00
outbuf = malloc(buflen, M_TEMP, M_WAITOK | M_ZERO);
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
ra->ra_proto = sc->sc_proto;
Make all lagg protocol methods live in lagg_protos, not in softc. All interfaces of a same protocol, use the same methods. Sponsored by: Netflix Sponsored by: Nginx, Inc.
2014-09-26 12:54:24 +00:00
lagg_proto_request(sc, &ra->ra_psc);
Avoid holding the softc lock when using copyout(). Reported by: dfr Approved by: re (rwatson)
2007-07-26 20:30:18 +00:00
count = 0;
buf = outbuf;
len = min(ra->ra_size, buflen);
Replace rmlock with epoch in lagg Use the new epoch based reclamation API. Now the hot paths will not block at all, and the sx lock is used for the softc data. This fixes LORs reported where the rwlock was obtained when the sxlock was held. Submitted by: mmacy Reported by: Harry Schmalzbauer <freebsd@omnilan.de> Reviewed by: sbruno Sponsored by: Limelight Networks Differential Revision: https://reviews.freebsd.org/D15355
2018-05-14 20:06:49 +00:00
CK_SLIST_FOREACH(lp, &sc->sc_ports, lp_entries) {
Avoid holding the softc lock when using copyout(). Reported by: dfr Approved by: re (rwatson)
2007-07-26 20:30:18 +00:00
if (len < sizeof(rpbuf))
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
break;
Avoid holding the softc lock when using copyout(). Reported by: dfr Approved by: re (rwatson)
2007-07-26 20:30:18 +00:00
lagg_port2req(lp, &rpbuf);
memcpy(buf, &rpbuf, sizeof(rpbuf));
count++;
buf += sizeof(rpbuf);
len -= sizeof(rpbuf);
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
}
Replace rmlock with epoch in lagg Use the new epoch based reclamation API. Now the hot paths will not block at all, and the sx lock is used for the softc data. This fixes LORs reported where the rwlock was obtained when the sxlock was held. Submitted by: mmacy Reported by: Harry Schmalzbauer <freebsd@omnilan.de> Reviewed by: sbruno Sponsored by: Limelight Networks Differential Revision: https://reviews.freebsd.org/D15355
2018-05-14 20:06:49 +00:00
LAGG_XUNLOCK(sc);
Avoid holding the softc lock when using copyout(). Reported by: dfr Approved by: re (rwatson)
2007-07-26 20:30:18 +00:00
ra->ra_ports = count;
ra->ra_size = count * sizeof(rpbuf);
error = copyout(outbuf, ra->ra_port, ra->ra_size);
free(outbuf, M_TEMP);
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
break;
case SIOCSLAGG:
error = priv_check(td, PRIV_NET_LAGG);
if (error)
break;
Fix regression introduced on 272446r. lagg(4) supports the protocol none, where it disables any traffic without disabling the lagg(4) interface itself. PR: 206921 Submitted by: Pushkar Kothavade <pushkarbk@gmail.com> Reviewed by: rpokala Approved by: bapt (mentor) MFC after: 3 weeks Sponsored by: gandi.net Differential Revision: https://reviews.freebsd.org/D5076
2016-02-19 06:35:53 +00:00
if (ra->ra_proto >= LAGG_PROTO_MAX) {
Separate option handling from SIOC[SG]LAGG to SIOC[SG]LAGGOPTS for backward compatibility with old ifconfig(8).
2014-10-02 20:01:13 +00:00
error = EPROTONOSUPPORT;
break;
}
Add support for IP over infiniband, IPoIB, to lagg(4). Currently only the failover protocol is supported due to limitations in the IPoIB architecture. Refer to the lagg(4) manual page for how to configure and use this new feature. A new network interface type, IFT_INFINIBANDLAG, has been added, similar to the existing IFT_IEEE8023ADLAG . ifconfig(8) has been updated to accept a new laggtype argument when creating lagg(4) network interfaces. This new argument is used to distinguish between ethernet and infiniband type of lagg(4) network interface. The laggtype argument is optional and defaults to ethernet. The lagg(4) command line syntax is backwards compatible. Differential Revision: https://reviews.freebsd.org/D26254 Reviewed by: melifaro@ MFC after: 1 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2020-10-22 09:47:12 +00:00
/* Infiniband only supports the failover protocol. */
if (ra->ra_proto != LAGG_PROTO_FAILOVER &&
ifp->if_type == IFT_INFINIBAND) {
error = EPROTONOSUPPORT;
break;
}
Introduce sleepable locks into if_lagg. Before this change if_lagg was using nonsleepable rmlocks to protect its internal state. This patch introduces another sx lock to protect code paths that require sleeping, while still uses old rmlock to protect hot nonsleepable data paths. This change allows to remove taskqueue decoupling used before to change interface addresses without holding the lock. Instead it uses sx lock to protect direct if_ioctl() calls. As another bonus, the new code synchronizes enabled capabilities of member interfaces, and allows to control them with ifconfig laggX, that was impossible before. This part should fix interoperation with if_bridge, that may need to disable some capabilities, such as TXCSUM or LRO, to allow bridging with noncapable interfaces. MFC after: 2 weeks Sponsored by: iXsystems, Inc. Differential Revision: https://reviews.freebsd.org/D10514
2017-05-02 19:09:11 +00:00
LAGG_XLOCK(sc);
Separate option handling from SIOC[SG]LAGG to SIOC[SG]LAGGOPTS for backward compatibility with old ifconfig(8).
2014-10-02 20:01:13 +00:00
lagg_proto_detach(sc);
lagg_proto_attach(sc, ra->ra_proto);
Introduce sleepable locks into if_lagg. Before this change if_lagg was using nonsleepable rmlocks to protect its internal state. This patch introduces another sx lock to protect code paths that require sleeping, while still uses old rmlock to protect hot nonsleepable data paths. This change allows to remove taskqueue decoupling used before to change interface addresses without holding the lock. Instead it uses sx lock to protect direct if_ioctl() calls. As another bonus, the new code synchronizes enabled capabilities of member interfaces, and allows to control them with ifconfig laggX, that was impossible before. This part should fix interoperation with if_bridge, that may need to disable some capabilities, such as TXCSUM or LRO, to allow bridging with noncapable interfaces. MFC after: 2 weeks Sponsored by: iXsystems, Inc. Differential Revision: https://reviews.freebsd.org/D10514
2017-05-02 19:09:11 +00:00
LAGG_XUNLOCK(sc);
Separate option handling from SIOC[SG]LAGG to SIOC[SG]LAGGOPTS for backward compatibility with old ifconfig(8).
2014-10-02 20:01:13 +00:00
break;
case SIOCGLAGGOPTS:
Replace rmlock with epoch in lagg Use the new epoch based reclamation API. Now the hot paths will not block at all, and the sx lock is used for the softc data. This fixes LORs reported where the rwlock was obtained when the sxlock was held. Submitted by: mmacy Reported by: Harry Schmalzbauer <freebsd@omnilan.de> Reviewed by: sbruno Sponsored by: Limelight Networks Differential Revision: https://reviews.freebsd.org/D15355
2018-05-14 20:06:49 +00:00
LAGG_XLOCK(sc);
Separate option handling from SIOC[SG]LAGG to SIOC[SG]LAGGOPTS for backward compatibility with old ifconfig(8).
2014-10-02 20:01:13 +00:00
ro->ro_opts = sc->sc_opts;
if (sc->sc_proto == LAGG_PROTO_LACP) {
struct lacp_softc *lsc;
lsc = (struct lacp_softc *)sc->sc_psc;
if (lsc->lsc_debug.lsc_tx_test != 0)
ro->ro_opts |= LAGG_OPT_LACP_TXTEST;
if (lsc->lsc_debug.lsc_rx_test != 0)
ro->ro_opts |= LAGG_OPT_LACP_RXTEST;
if (lsc->lsc_strict_mode != 0)
ro->ro_opts |= LAGG_OPT_LACP_STRICT;
Make LAG LACP fast timeout tunable through IOCTL. Differential Revision: D3300 Submitted by: LN Sundararajan <lakshmi.n at msystechnologies> Reviewed by: wblock, smh, gnn, hiren, rpokala at panasas MFC after: 2 weeks Sponsored by: Panasas
2015-08-12 20:21:04 +00:00
if (lsc->lsc_fast_timeout != 0)
Decode the "LACP Fast Timeout" LAGG option flag r286700 added the "lacp_fast_timeout" option to `ifconfig', but we forgot to include the new option in the string used to decode the option bits. Add "LACP_FAST_TIMO" to LAGG_OPT_BITS. Also, s/LAGG_OPT_LACP_TIMEOUT/LAGG_OPT_LACP_FAST_TIMO/g , to be clearer that the flag indicates "Fast Timeout" mode. Reported by: Greg Foster <gfoster at panasas dot com> Reviewed by: jpaetzel MFC after: 1 week Sponsored by: Panasas Differential Revision: https://reviews.freebsd.org/D25239
2020-06-11 22:46:08 +00:00
ro->ro_opts |= LAGG_OPT_LACP_FAST_TIMO;
Separate option handling from SIOC[SG]LAGG to SIOC[SG]LAGGOPTS for backward compatibility with old ifconfig(8).
2014-10-02 20:01:13 +00:00
ro->ro_active = sc->sc_active;
} else {
ro->ro_active = 0;
Replace rmlock with epoch in lagg Use the new epoch based reclamation API. Now the hot paths will not block at all, and the sx lock is used for the softc data. This fixes LORs reported where the rwlock was obtained when the sxlock was held. Submitted by: mmacy Reported by: Harry Schmalzbauer <freebsd@omnilan.de> Reviewed by: sbruno Sponsored by: Limelight Networks Differential Revision: https://reviews.freebsd.org/D15355
2018-05-14 20:06:49 +00:00
CK_SLIST_FOREACH(lp, &sc->sc_ports, lp_entries)
Separate option handling from SIOC[SG]LAGG to SIOC[SG]LAGGOPTS for backward compatibility with old ifconfig(8).
2014-10-02 20:01:13 +00:00
ro->ro_active += LAGG_PORTACTIVE(lp);
}
lagg: Clean up handling of the rr_limit option. - Don't allow an unprivileged user to set the stride. [1] - Only set the stride under the softc lock. - Rename the internal fields to accurately reflect their use. Keep ro_bkt to avoid changing the user API. - Simplify the implementation. The port index is just sc_seq / stride. - Document rr_limit in ifconfig.8. Reported by: Ilja Van Sprundel <ivansprundel@ioactive.com> [1] MFC after: 1 week Sponsored by: The FreeBSD Foundation Differential Revision: https://reviews.freebsd.org/D22857
2019-12-22 21:56:47 +00:00
ro->ro_bkt = sc->sc_stride;
Separate option handling from SIOC[SG]LAGG to SIOC[SG]LAGGOPTS for backward compatibility with old ifconfig(8).
2014-10-02 20:01:13 +00:00
ro->ro_flapping = sc->sc_flapping;
ro->ro_flowid_shift = sc->flowid_shift;
Replace rmlock with epoch in lagg Use the new epoch based reclamation API. Now the hot paths will not block at all, and the sx lock is used for the softc data. This fixes LORs reported where the rwlock was obtained when the sxlock was held. Submitted by: mmacy Reported by: Harry Schmalzbauer <freebsd@omnilan.de> Reviewed by: sbruno Sponsored by: Limelight Networks Differential Revision: https://reviews.freebsd.org/D15355
2018-05-14 20:06:49 +00:00
LAGG_XUNLOCK(sc);
Separate option handling from SIOC[SG]LAGG to SIOC[SG]LAGGOPTS for backward compatibility with old ifconfig(8).
2014-10-02 20:01:13 +00:00
break;
case SIOCSLAGGOPTS:
error = priv_check(td, PRIV_NET_LAGG);
if (error)
break;
lagg: Clean up handling of the rr_limit option. - Don't allow an unprivileged user to set the stride. [1] - Only set the stride under the softc lock. - Rename the internal fields to accurately reflect their use. Keep ro_bkt to avoid changing the user API. - Simplify the implementation. The port index is just sc_seq / stride. - Document rr_limit in ifconfig.8. Reported by: Ilja Van Sprundel <ivansprundel@ioactive.com> [1] MFC after: 1 week Sponsored by: The FreeBSD Foundation Differential Revision: https://reviews.freebsd.org/D22857
2019-12-22 21:56:47 +00:00
/*
* The stride option was added without defining a corresponding
lagg: Further cleanup of the rr_limit option. Add an option flag so that arbitrary updates to a lagg's configuration do not clear sc_stride. Preseve compatibility for old ifconfig binaries. Update ifconfig to use the new flag and improve the casting used when parsing the option parameter. Modify the RR transmit function to avoid locklessly reading sc_stride twice. Ensure that sc_stride is always 1 or greater. Reviewed by: hselasky MFC after: 1 week Sponsored by: The FreeBSD Foundation Differential Revision: https://reviews.freebsd.org/D23092
2020-01-09 14:58:41 +00:00
* LAGG_OPT flag, so handle a non-zero value before checking
* anything else to preserve compatibility.
lagg: Clean up handling of the rr_limit option. - Don't allow an unprivileged user to set the stride. [1] - Only set the stride under the softc lock. - Rename the internal fields to accurately reflect their use. Keep ro_bkt to avoid changing the user API. - Simplify the implementation. The port index is just sc_seq / stride. - Document rr_limit in ifconfig.8. Reported by: Ilja Van Sprundel <ivansprundel@ioactive.com> [1] MFC after: 1 week Sponsored by: The FreeBSD Foundation Differential Revision: https://reviews.freebsd.org/D22857
2019-12-22 21:56:47 +00:00
*/
LAGG_XLOCK(sc);
lagg: Further cleanup of the rr_limit option. Add an option flag so that arbitrary updates to a lagg's configuration do not clear sc_stride. Preseve compatibility for old ifconfig binaries. Update ifconfig to use the new flag and improve the casting used when parsing the option parameter. Modify the RR transmit function to avoid locklessly reading sc_stride twice. Ensure that sc_stride is always 1 or greater. Reviewed by: hselasky MFC after: 1 week Sponsored by: The FreeBSD Foundation Differential Revision: https://reviews.freebsd.org/D23092
2020-01-09 14:58:41 +00:00
if (ro->ro_opts == 0 && ro->ro_bkt != 0) {
lagg: Clean up handling of the rr_limit option. - Don't allow an unprivileged user to set the stride. [1] - Only set the stride under the softc lock. - Rename the internal fields to accurately reflect their use. Keep ro_bkt to avoid changing the user API. - Simplify the implementation. The port index is just sc_seq / stride. - Document rr_limit in ifconfig.8. Reported by: Ilja Van Sprundel <ivansprundel@ioactive.com> [1] MFC after: 1 week Sponsored by: The FreeBSD Foundation Differential Revision: https://reviews.freebsd.org/D22857
2019-12-22 21:56:47 +00:00
if (sc->sc_proto != LAGG_PROTO_ROUNDROBIN) {
LAGG_XUNLOCK(sc);
error = EINVAL;
break;
}
sc->sc_stride = ro->ro_bkt;
}
if (ro->ro_opts == 0) {
LAGG_XUNLOCK(sc);
Separate option handling from SIOC[SG]LAGG to SIOC[SG]LAGGOPTS for backward compatibility with old ifconfig(8).
2014-10-02 20:01:13 +00:00
break;
lagg: Clean up handling of the rr_limit option. - Don't allow an unprivileged user to set the stride. [1] - Only set the stride under the softc lock. - Rename the internal fields to accurately reflect their use. Keep ro_bkt to avoid changing the user API. - Simplify the implementation. The port index is just sc_seq / stride. - Document rr_limit in ifconfig.8. Reported by: Ilja Van Sprundel <ivansprundel@ioactive.com> [1] MFC after: 1 week Sponsored by: The FreeBSD Foundation Differential Revision: https://reviews.freebsd.org/D22857
2019-12-22 21:56:47 +00:00
}
Separate option handling from SIOC[SG]LAGG to SIOC[SG]LAGGOPTS for backward compatibility with old ifconfig(8).
2014-10-02 20:01:13 +00:00
/*
* Set options. LACP options are stored in sc->sc_psc,
* not in sc_opts.
*/
int valid, lacp;
switch (ro->ro_opts) {
case LAGG_OPT_USE_FLOWID:
case -LAGG_OPT_USE_FLOWID:
Select lacp egress ports based on NUMA domain This change creates an array of port maps indexed by numa domain for lacp port selection. If we have lacp interfaces in more than one domain, then we select the egress port by indexing into the numa port maps and picking a port on the appropriate numa domain. This is behavior is controlled by the new ifconfig use_numa flag and net.link.lagg.use_numa sysctl/tunable (both modeled after the existing use_flowid), which default to enabled. Reviewed by: bz, hselasky, markj (and scottl, earlier version) Sponsored by: Netflix Differential Revision: https://reviews.freebsd.org/D20060
2019-05-03 14:43:21 +00:00
case LAGG_OPT_USE_NUMA:
case -LAGG_OPT_USE_NUMA:
Separate option handling from SIOC[SG]LAGG to SIOC[SG]LAGGOPTS for backward compatibility with old ifconfig(8).
2014-10-02 20:01:13 +00:00
case LAGG_OPT_FLOWIDSHIFT:
lagg: Further cleanup of the rr_limit option. Add an option flag so that arbitrary updates to a lagg's configuration do not clear sc_stride. Preseve compatibility for old ifconfig binaries. Update ifconfig to use the new flag and improve the casting used when parsing the option parameter. Modify the RR transmit function to avoid locklessly reading sc_stride twice. Ensure that sc_stride is always 1 or greater. Reviewed by: hselasky MFC after: 1 week Sponsored by: The FreeBSD Foundation Differential Revision: https://reviews.freebsd.org/D23092
2020-01-09 14:58:41 +00:00
case LAGG_OPT_RR_LIMIT:
Separate option handling from SIOC[SG]LAGG to SIOC[SG]LAGGOPTS for backward compatibility with old ifconfig(8).
2014-10-02 20:01:13 +00:00
valid = 1;
lacp = 0;
break;
case LAGG_OPT_LACP_TXTEST:
case -LAGG_OPT_LACP_TXTEST:
case LAGG_OPT_LACP_RXTEST:
case -LAGG_OPT_LACP_RXTEST:
case LAGG_OPT_LACP_STRICT:
case -LAGG_OPT_LACP_STRICT:
Decode the "LACP Fast Timeout" LAGG option flag r286700 added the "lacp_fast_timeout" option to `ifconfig', but we forgot to include the new option in the string used to decode the option bits. Add "LACP_FAST_TIMO" to LAGG_OPT_BITS. Also, s/LAGG_OPT_LACP_TIMEOUT/LAGG_OPT_LACP_FAST_TIMO/g , to be clearer that the flag indicates "Fast Timeout" mode. Reported by: Greg Foster <gfoster at panasas dot com> Reviewed by: jpaetzel MFC after: 1 week Sponsored by: Panasas Differential Revision: https://reviews.freebsd.org/D25239
2020-06-11 22:46:08 +00:00
case LAGG_OPT_LACP_FAST_TIMO:
case -LAGG_OPT_LACP_FAST_TIMO:
Separate option handling from SIOC[SG]LAGG to SIOC[SG]LAGGOPTS for backward compatibility with old ifconfig(8).
2014-10-02 20:01:13 +00:00
valid = lacp = 1;
break;
default:
valid = lacp = 0;
break;
}
if (valid == 0 ||
(lacp == 1 && sc->sc_proto != LAGG_PROTO_LACP)) {
/* Invalid combination of options specified. */
error = EINVAL;
Introduce sleepable locks into if_lagg. Before this change if_lagg was using nonsleepable rmlocks to protect its internal state. This patch introduces another sx lock to protect code paths that require sleeping, while still uses old rmlock to protect hot nonsleepable data paths. This change allows to remove taskqueue decoupling used before to change interface addresses without holding the lock. Instead it uses sx lock to protect direct if_ioctl() calls. As another bonus, the new code synchronizes enabled capabilities of member interfaces, and allows to control them with ifconfig laggX, that was impossible before. This part should fix interoperation with if_bridge, that may need to disable some capabilities, such as TXCSUM or LRO, to allow bridging with noncapable interfaces. MFC after: 2 weeks Sponsored by: iXsystems, Inc. Differential Revision: https://reviews.freebsd.org/D10514
2017-05-02 19:09:11 +00:00
LAGG_XUNLOCK(sc);
Separate option handling from SIOC[SG]LAGG to SIOC[SG]LAGGOPTS for backward compatibility with old ifconfig(8).
2014-10-02 20:01:13 +00:00
break; /* Return from SIOCSLAGGOPTS. */
}
lagg: Further cleanup of the rr_limit option. Add an option flag so that arbitrary updates to a lagg's configuration do not clear sc_stride. Preseve compatibility for old ifconfig binaries. Update ifconfig to use the new flag and improve the casting used when parsing the option parameter. Modify the RR transmit function to avoid locklessly reading sc_stride twice. Ensure that sc_stride is always 1 or greater. Reviewed by: hselasky MFC after: 1 week Sponsored by: The FreeBSD Foundation Differential Revision: https://reviews.freebsd.org/D23092
2020-01-09 14:58:41 +00:00
Separate option handling from SIOC[SG]LAGG to SIOC[SG]LAGGOPTS for backward compatibility with old ifconfig(8).
2014-10-02 20:01:13 +00:00
/*
* Store new options into sc->sc_opts except for
lagg: Further cleanup of the rr_limit option. Add an option flag so that arbitrary updates to a lagg's configuration do not clear sc_stride. Preseve compatibility for old ifconfig binaries. Update ifconfig to use the new flag and improve the casting used when parsing the option parameter. Modify the RR transmit function to avoid locklessly reading sc_stride twice. Ensure that sc_stride is always 1 or greater. Reviewed by: hselasky MFC after: 1 week Sponsored by: The FreeBSD Foundation Differential Revision: https://reviews.freebsd.org/D23092
2020-01-09 14:58:41 +00:00
* FLOWIDSHIFT, RR and LACP options.
Separate option handling from SIOC[SG]LAGG to SIOC[SG]LAGGOPTS for backward compatibility with old ifconfig(8).
2014-10-02 20:01:13 +00:00
*/
if (lacp == 0) {
if (ro->ro_opts == LAGG_OPT_FLOWIDSHIFT)
sc->flowid_shift = ro->ro_flowid_shift;
lagg: Further cleanup of the rr_limit option. Add an option flag so that arbitrary updates to a lagg's configuration do not clear sc_stride. Preseve compatibility for old ifconfig binaries. Update ifconfig to use the new flag and improve the casting used when parsing the option parameter. Modify the RR transmit function to avoid locklessly reading sc_stride twice. Ensure that sc_stride is always 1 or greater. Reviewed by: hselasky MFC after: 1 week Sponsored by: The FreeBSD Foundation Differential Revision: https://reviews.freebsd.org/D23092
2020-01-09 14:58:41 +00:00
else if (ro->ro_opts == LAGG_OPT_RR_LIMIT) {
if (sc->sc_proto != LAGG_PROTO_ROUNDROBIN ||
ro->ro_bkt == 0) {
error = EINVAL;
LAGG_XUNLOCK(sc);
break;
}
sc->sc_stride = ro->ro_bkt;
} else if (ro->ro_opts > 0)
Separate option handling from SIOC[SG]LAGG to SIOC[SG]LAGGOPTS for backward compatibility with old ifconfig(8).
2014-10-02 20:01:13 +00:00
sc->sc_opts |= ro->ro_opts;
else
sc->sc_opts &= ~ro->ro_opts;
} else {
struct lacp_softc *lsc;
Make LAG LACP fast timeout tunable through IOCTL. Differential Revision: D3300 Submitted by: LN Sundararajan <lakshmi.n at msystechnologies> Reviewed by: wblock, smh, gnn, hiren, rpokala at panasas MFC after: 2 weeks Sponsored by: Panasas
2015-08-12 20:21:04 +00:00
struct lacp_port *lp;
Separate option handling from SIOC[SG]LAGG to SIOC[SG]LAGGOPTS for backward compatibility with old ifconfig(8).
2014-10-02 20:01:13 +00:00
lsc = (struct lacp_softc *)sc->sc_psc;
switch (ro->ro_opts) {
Virtualize lagg(4) cloner. This change fixes a panic when tearing down if_lagg(4) interfaces which were cloned in a vnet jail. Sysctl nodes which are dynamically generated for each cloned interface (net.link.lagg.N.*) have been removed, and use_flowid and flowid_shift ifconfig(8) parameters have been added instead. Flags and per-interface statistics counters are displayed in "ifconfig -v". CR: D842
2014-10-01 21:37:32 +00:00
case LAGG_OPT_LACP_TXTEST:
Separate option handling from SIOC[SG]LAGG to SIOC[SG]LAGGOPTS for backward compatibility with old ifconfig(8).
2014-10-02 20:01:13 +00:00
lsc->lsc_debug.lsc_tx_test = 1;
break;
Virtualize lagg(4) cloner. This change fixes a panic when tearing down if_lagg(4) interfaces which were cloned in a vnet jail. Sysctl nodes which are dynamically generated for each cloned interface (net.link.lagg.N.*) have been removed, and use_flowid and flowid_shift ifconfig(8) parameters have been added instead. Flags and per-interface statistics counters are displayed in "ifconfig -v". CR: D842
2014-10-01 21:37:32 +00:00
case -LAGG_OPT_LACP_TXTEST:
Separate option handling from SIOC[SG]LAGG to SIOC[SG]LAGGOPTS for backward compatibility with old ifconfig(8).
2014-10-02 20:01:13 +00:00
lsc->lsc_debug.lsc_tx_test = 0;
break;
Virtualize lagg(4) cloner. This change fixes a panic when tearing down if_lagg(4) interfaces which were cloned in a vnet jail. Sysctl nodes which are dynamically generated for each cloned interface (net.link.lagg.N.*) have been removed, and use_flowid and flowid_shift ifconfig(8) parameters have been added instead. Flags and per-interface statistics counters are displayed in "ifconfig -v". CR: D842
2014-10-01 21:37:32 +00:00
case LAGG_OPT_LACP_RXTEST:
Separate option handling from SIOC[SG]LAGG to SIOC[SG]LAGGOPTS for backward compatibility with old ifconfig(8).
2014-10-02 20:01:13 +00:00
lsc->lsc_debug.lsc_rx_test = 1;
break;
Virtualize lagg(4) cloner. This change fixes a panic when tearing down if_lagg(4) interfaces which were cloned in a vnet jail. Sysctl nodes which are dynamically generated for each cloned interface (net.link.lagg.N.*) have been removed, and use_flowid and flowid_shift ifconfig(8) parameters have been added instead. Flags and per-interface statistics counters are displayed in "ifconfig -v". CR: D842
2014-10-01 21:37:32 +00:00
case -LAGG_OPT_LACP_RXTEST:
Separate option handling from SIOC[SG]LAGG to SIOC[SG]LAGGOPTS for backward compatibility with old ifconfig(8).
2014-10-02 20:01:13 +00:00
lsc->lsc_debug.lsc_rx_test = 0;
break;
Virtualize lagg(4) cloner. This change fixes a panic when tearing down if_lagg(4) interfaces which were cloned in a vnet jail. Sysctl nodes which are dynamically generated for each cloned interface (net.link.lagg.N.*) have been removed, and use_flowid and flowid_shift ifconfig(8) parameters have been added instead. Flags and per-interface statistics counters are displayed in "ifconfig -v". CR: D842
2014-10-01 21:37:32 +00:00
case LAGG_OPT_LACP_STRICT:
Separate option handling from SIOC[SG]LAGG to SIOC[SG]LAGGOPTS for backward compatibility with old ifconfig(8).
2014-10-02 20:01:13 +00:00
lsc->lsc_strict_mode = 1;
Virtualize lagg(4) cloner. This change fixes a panic when tearing down if_lagg(4) interfaces which were cloned in a vnet jail. Sysctl nodes which are dynamically generated for each cloned interface (net.link.lagg.N.*) have been removed, and use_flowid and flowid_shift ifconfig(8) parameters have been added instead. Flags and per-interface statistics counters are displayed in "ifconfig -v". CR: D842
2014-10-01 21:37:32 +00:00
break;
Separate option handling from SIOC[SG]LAGG to SIOC[SG]LAGGOPTS for backward compatibility with old ifconfig(8).
2014-10-02 20:01:13 +00:00
case -LAGG_OPT_LACP_STRICT:
lsc->lsc_strict_mode = 0;
Virtualize lagg(4) cloner. This change fixes a panic when tearing down if_lagg(4) interfaces which were cloned in a vnet jail. Sysctl nodes which are dynamically generated for each cloned interface (net.link.lagg.N.*) have been removed, and use_flowid and flowid_shift ifconfig(8) parameters have been added instead. Flags and per-interface statistics counters are displayed in "ifconfig -v". CR: D842
2014-10-01 21:37:32 +00:00
break;
Decode the "LACP Fast Timeout" LAGG option flag r286700 added the "lacp_fast_timeout" option to `ifconfig', but we forgot to include the new option in the string used to decode the option bits. Add "LACP_FAST_TIMO" to LAGG_OPT_BITS. Also, s/LAGG_OPT_LACP_TIMEOUT/LAGG_OPT_LACP_FAST_TIMO/g , to be clearer that the flag indicates "Fast Timeout" mode. Reported by: Greg Foster <gfoster at panasas dot com> Reviewed by: jpaetzel MFC after: 1 week Sponsored by: Panasas Differential Revision: https://reviews.freebsd.org/D25239
2020-06-11 22:46:08 +00:00
case LAGG_OPT_LACP_FAST_TIMO:
Make LAG LACP fast timeout tunable through IOCTL. Differential Revision: D3300 Submitted by: LN Sundararajan <lakshmi.n at msystechnologies> Reviewed by: wblock, smh, gnn, hiren, rpokala at panasas MFC after: 2 weeks Sponsored by: Panasas
2015-08-12 20:21:04 +00:00
LACP_LOCK(lsc);
LIST_FOREACH(lp, &lsc->lsc_ports, lp_next)
lp->lp_state |= LACP_STATE_TIMEOUT;
LACP_UNLOCK(lsc);
lsc->lsc_fast_timeout = 1;
break;
Decode the "LACP Fast Timeout" LAGG option flag r286700 added the "lacp_fast_timeout" option to `ifconfig', but we forgot to include the new option in the string used to decode the option bits. Add "LACP_FAST_TIMO" to LAGG_OPT_BITS. Also, s/LAGG_OPT_LACP_TIMEOUT/LAGG_OPT_LACP_FAST_TIMO/g , to be clearer that the flag indicates "Fast Timeout" mode. Reported by: Greg Foster <gfoster at panasas dot com> Reviewed by: jpaetzel MFC after: 1 week Sponsored by: Panasas Differential Revision: https://reviews.freebsd.org/D25239
2020-06-11 22:46:08 +00:00
case -LAGG_OPT_LACP_FAST_TIMO:
Make LAG LACP fast timeout tunable through IOCTL. Differential Revision: D3300 Submitted by: LN Sundararajan <lakshmi.n at msystechnologies> Reviewed by: wblock, smh, gnn, hiren, rpokala at panasas MFC after: 2 weeks Sponsored by: Panasas
2015-08-12 20:21:04 +00:00
LACP_LOCK(lsc);
LIST_FOREACH(lp, &lsc->lsc_ports, lp_next)
lp->lp_state &= ~LACP_STATE_TIMEOUT;
LACP_UNLOCK(lsc);
lsc->lsc_fast_timeout = 0;
break;
Virtualize lagg(4) cloner. This change fixes a panic when tearing down if_lagg(4) interfaces which were cloned in a vnet jail. Sysctl nodes which are dynamically generated for each cloned interface (net.link.lagg.N.*) have been removed, and use_flowid and flowid_shift ifconfig(8) parameters have been added instead. Flags and per-interface statistics counters are displayed in "ifconfig -v". CR: D842
2014-10-01 21:37:32 +00:00
}
}
Introduce sleepable locks into if_lagg. Before this change if_lagg was using nonsleepable rmlocks to protect its internal state. This patch introduces another sx lock to protect code paths that require sleeping, while still uses old rmlock to protect hot nonsleepable data paths. This change allows to remove taskqueue decoupling used before to change interface addresses without holding the lock. Instead it uses sx lock to protect direct if_ioctl() calls. As another bonus, the new code synchronizes enabled capabilities of member interfaces, and allows to control them with ifconfig laggX, that was impossible before. This part should fix interoperation with if_bridge, that may need to disable some capabilities, such as TXCSUM or LRO, to allow bridging with noncapable interfaces. MFC after: 2 weeks Sponsored by: iXsystems, Inc. Differential Revision: https://reviews.freebsd.org/D10514
2017-05-02 19:09:11 +00:00
LAGG_XUNLOCK(sc);
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
break;
Add the ability to set which packet layers are used for the load balance hash calculation.
2012-03-06 22:58:13 +00:00
case SIOCGLAGGFLAGS:
Factor out mbuf hashing code from LAGG driver so that other network drivers can use it. This avoids some code duplication. Add missing default case to all switch statements while at it. Also move the hashing of the IPv6 flow field to layer 4 because the IPv6 flow field is constant on a per L4 connection basis and not on a per L3 network. Differential Revision: https://reviews.freebsd.org/D1987 Sponsored by: Mellanox Technologies MFC after: 1 month
2015-03-11 16:02:24 +00:00
rf->rf_flags = 0;
Replace rmlock with epoch in lagg Use the new epoch based reclamation API. Now the hot paths will not block at all, and the sx lock is used for the softc data. This fixes LORs reported where the rwlock was obtained when the sxlock was held. Submitted by: mmacy Reported by: Harry Schmalzbauer <freebsd@omnilan.de> Reviewed by: sbruno Sponsored by: Limelight Networks Differential Revision: https://reviews.freebsd.org/D15355
2018-05-14 20:06:49 +00:00
LAGG_XLOCK(sc);
Factor out mbuf hashing code from LAGG driver so that other network drivers can use it. This avoids some code duplication. Add missing default case to all switch statements while at it. Also move the hashing of the IPv6 flow field to layer 4 because the IPv6 flow field is constant on a per L4 connection basis and not on a per L3 network. Differential Revision: https://reviews.freebsd.org/D1987 Sponsored by: Mellanox Technologies MFC after: 1 month
2015-03-11 16:02:24 +00:00
if (sc->sc_flags & MBUF_HASHFLAG_L2)
rf->rf_flags |= LAGG_F_HASHL2;
if (sc->sc_flags & MBUF_HASHFLAG_L3)
rf->rf_flags |= LAGG_F_HASHL3;
if (sc->sc_flags & MBUF_HASHFLAG_L4)
rf->rf_flags |= LAGG_F_HASHL4;
Replace rmlock with epoch in lagg Use the new epoch based reclamation API. Now the hot paths will not block at all, and the sx lock is used for the softc data. This fixes LORs reported where the rwlock was obtained when the sxlock was held. Submitted by: mmacy Reported by: Harry Schmalzbauer <freebsd@omnilan.de> Reviewed by: sbruno Sponsored by: Limelight Networks Differential Revision: https://reviews.freebsd.org/D15355
2018-05-14 20:06:49 +00:00
LAGG_XUNLOCK(sc);
Add the ability to set which packet layers are used for the load balance hash calculation.
2012-03-06 22:58:13 +00:00
break;
case SIOCSLAGGHASH:
error = priv_check(td, PRIV_NET_LAGG);
if (error)
break;
if ((rf->rf_flags & LAGG_F_HASHMASK) == 0) {
error = EINVAL;
break;
}
Introduce sleepable locks into if_lagg. Before this change if_lagg was using nonsleepable rmlocks to protect its internal state. This patch introduces another sx lock to protect code paths that require sleeping, while still uses old rmlock to protect hot nonsleepable data paths. This change allows to remove taskqueue decoupling used before to change interface addresses without holding the lock. Instead it uses sx lock to protect direct if_ioctl() calls. As another bonus, the new code synchronizes enabled capabilities of member interfaces, and allows to control them with ifconfig laggX, that was impossible before. This part should fix interoperation with if_bridge, that may need to disable some capabilities, such as TXCSUM or LRO, to allow bridging with noncapable interfaces. MFC after: 2 weeks Sponsored by: iXsystems, Inc. Differential Revision: https://reviews.freebsd.org/D10514
2017-05-02 19:09:11 +00:00
LAGG_XLOCK(sc);
Factor out mbuf hashing code from LAGG driver so that other network drivers can use it. This avoids some code duplication. Add missing default case to all switch statements while at it. Also move the hashing of the IPv6 flow field to layer 4 because the IPv6 flow field is constant on a per L4 connection basis and not on a per L3 network. Differential Revision: https://reviews.freebsd.org/D1987 Sponsored by: Mellanox Technologies MFC after: 1 month
2015-03-11 16:02:24 +00:00
sc->sc_flags = 0;
if (rf->rf_flags & LAGG_F_HASHL2)
sc->sc_flags |= MBUF_HASHFLAG_L2;
if (rf->rf_flags & LAGG_F_HASHL3)
sc->sc_flags |= MBUF_HASHFLAG_L3;
if (rf->rf_flags & LAGG_F_HASHL4)
sc->sc_flags |= MBUF_HASHFLAG_L4;
Introduce sleepable locks into if_lagg. Before this change if_lagg was using nonsleepable rmlocks to protect its internal state. This patch introduces another sx lock to protect code paths that require sleeping, while still uses old rmlock to protect hot nonsleepable data paths. This change allows to remove taskqueue decoupling used before to change interface addresses without holding the lock. Instead it uses sx lock to protect direct if_ioctl() calls. As another bonus, the new code synchronizes enabled capabilities of member interfaces, and allows to control them with ifconfig laggX, that was impossible before. This part should fix interoperation with if_bridge, that may need to disable some capabilities, such as TXCSUM or LRO, to allow bridging with noncapable interfaces. MFC after: 2 weeks Sponsored by: iXsystems, Inc. Differential Revision: https://reviews.freebsd.org/D10514
2017-05-02 19:09:11 +00:00
LAGG_XUNLOCK(sc);
Add the ability to set which packet layers are used for the load balance hash calculation.
2012-03-06 22:58:13 +00:00
break;
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
case SIOCGLAGGPORT:
if (rp->rp_portname[0] == '\0' ||
Add interface reference counting to if_lagg. Using plain ifunit() looks like request for troubles. MFC after: 2 weeks
2017-04-21 13:45:01 +00:00
(tpif = ifunit_ref(rp->rp_portname)) == NULL) {
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
error = EINVAL;
break;
}
Convert LAGG_RLOCK() to NET_EPOCH_ENTER(). No functional changes.
2020-12-08 16:36:46 +00:00
NET_EPOCH_ENTER(et);
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
if ((lp = (struct lagg_port *)tpif->if_lagg) == NULL ||
non-functional cleanup - remove dead code - use consistent variable names - gc unused defines - whitespace cleanup
2007-06-12 07:29:11 +00:00
lp->lp_softc != sc) {
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
error = ENOENT;
Convert LAGG_RLOCK() to NET_EPOCH_ENTER(). No functional changes.
2020-12-08 16:36:46 +00:00
NET_EPOCH_EXIT(et);
Add interface reference counting to if_lagg. Using plain ifunit() looks like request for troubles. MFC after: 2 weeks
2017-04-21 13:45:01 +00:00
if_rele(tpif);
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
break;
}
lagg_port2req(lp, rp);
Convert LAGG_RLOCK() to NET_EPOCH_ENTER(). No functional changes.
2020-12-08 16:36:46 +00:00
NET_EPOCH_EXIT(et);
Add interface reference counting to if_lagg. Using plain ifunit() looks like request for troubles. MFC after: 2 weeks
2017-04-21 13:45:01 +00:00
if_rele(tpif);
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
break;
case SIOCSLAGGPORT:
error = priv_check(td, PRIV_NET_LAGG);
if (error)
break;
if (rp->rp_portname[0] == '\0' ||
Add interface reference counting to if_lagg. Using plain ifunit() looks like request for troubles. MFC after: 2 weeks
2017-04-21 13:45:01 +00:00
(tpif = ifunit_ref(rp->rp_portname)) == NULL) {
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
error = EINVAL;
break;
}
- Fix lladdr configuration which could prevent LACP mode from working. - Fix LORs when a laggport interface has an IPv6 LLA. PR: 194321
2014-10-17 09:08:44 +00:00
#ifdef INET6
/*
* A laggport interface should not have inet6 address
* because two interfaces with a valid link-local
* scope zone must not be merged in any form. This
* restriction is needed to prevent violation of
* link-local scope zone. Attempts to add a laggport
* interface which has inet6 addresses triggers
* removal of all inet6 addresses on the member
* interface.
*/
if (in6ifa_llaonifp(tpif)) {
in6_ifdetach(tpif);
if_printf(sc->sc_ifp,
"IPv6 addresses on %s have been removed "
"before adding it as a member to prevent "
"IPv6 address scope violation.\n",
tpif->if_xname);
}
#endif
Update route MTUs for bridge, lagg and vlan interfaces. Those interfaces may implicitly change their MTU on addition of parent interface in addition to normal SIOCSIFMTU ioctl path, where the route MTUs are updated normally. MFC after: 2 weeks Sponsored by: iXsystems, Inc.
2020-01-22 20:36:45 +00:00
oldmtu = ifp->if_mtu;
Introduce sleepable locks into if_lagg. Before this change if_lagg was using nonsleepable rmlocks to protect its internal state. This patch introduces another sx lock to protect code paths that require sleeping, while still uses old rmlock to protect hot nonsleepable data paths. This change allows to remove taskqueue decoupling used before to change interface addresses without holding the lock. Instead it uses sx lock to protect direct if_ioctl() calls. As another bonus, the new code synchronizes enabled capabilities of member interfaces, and allows to control them with ifconfig laggX, that was impossible before. This part should fix interoperation with if_bridge, that may need to disable some capabilities, such as TXCSUM or LRO, to allow bridging with noncapable interfaces. MFC after: 2 weeks Sponsored by: iXsystems, Inc. Differential Revision: https://reviews.freebsd.org/D10514
2017-05-02 19:09:11 +00:00
LAGG_XLOCK(sc);
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
error = lagg_port_create(sc, tpif);
Introduce sleepable locks into if_lagg. Before this change if_lagg was using nonsleepable rmlocks to protect its internal state. This patch introduces another sx lock to protect code paths that require sleeping, while still uses old rmlock to protect hot nonsleepable data paths. This change allows to remove taskqueue decoupling used before to change interface addresses without holding the lock. Instead it uses sx lock to protect direct if_ioctl() calls. As another bonus, the new code synchronizes enabled capabilities of member interfaces, and allows to control them with ifconfig laggX, that was impossible before. This part should fix interoperation with if_bridge, that may need to disable some capabilities, such as TXCSUM or LRO, to allow bridging with noncapable interfaces. MFC after: 2 weeks Sponsored by: iXsystems, Inc. Differential Revision: https://reviews.freebsd.org/D10514
2017-05-02 19:09:11 +00:00
LAGG_XUNLOCK(sc);
Add interface reference counting to if_lagg. Using plain ifunit() looks like request for troubles. MFC after: 2 weeks
2017-04-21 13:45:01 +00:00
if_rele(tpif);
Update route MTUs for bridge, lagg and vlan interfaces. Those interfaces may implicitly change their MTU on addition of parent interface in addition to normal SIOCSIFMTU ioctl path, where the route MTUs are updated normally. MFC after: 2 weeks Sponsored by: iXsystems, Inc.
2020-01-22 20:36:45 +00:00
/*
* LAGG MTU may change during addition of the first port.
* If it did, do network layer specific procedure.
*/
if (ifp->if_mtu != oldmtu) {
#ifdef INET6
nd6_setmtu(ifp);
#endif
rt_updatemtu(ifp);
}
Call VLAN_CAPABILITIES() when LAGG capabilities change. This makes VLAN on top of LAGG to expose proper capabilities if they are changed after creation. MFC after: 1 week
2017-05-26 22:22:48 +00:00
VLAN_CAPABILITIES(ifp);
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
break;
case SIOCSLAGGDELPORT:
error = priv_check(td, PRIV_NET_LAGG);
if (error)
break;
if (rp->rp_portname[0] == '\0' ||
Add interface reference counting to if_lagg. Using plain ifunit() looks like request for troubles. MFC after: 2 weeks
2017-04-21 13:45:01 +00:00
(tpif = ifunit_ref(rp->rp_portname)) == NULL) {
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
error = EINVAL;
break;
}
Introduce sleepable locks into if_lagg. Before this change if_lagg was using nonsleepable rmlocks to protect its internal state. This patch introduces another sx lock to protect code paths that require sleeping, while still uses old rmlock to protect hot nonsleepable data paths. This change allows to remove taskqueue decoupling used before to change interface addresses without holding the lock. Instead it uses sx lock to protect direct if_ioctl() calls. As another bonus, the new code synchronizes enabled capabilities of member interfaces, and allows to control them with ifconfig laggX, that was impossible before. This part should fix interoperation with if_bridge, that may need to disable some capabilities, such as TXCSUM or LRO, to allow bridging with noncapable interfaces. MFC after: 2 weeks Sponsored by: iXsystems, Inc. Differential Revision: https://reviews.freebsd.org/D10514
2017-05-02 19:09:11 +00:00
LAGG_XLOCK(sc);
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
if ((lp = (struct lagg_port *)tpif->if_lagg) == NULL ||
non-functional cleanup - remove dead code - use consistent variable names - gc unused defines - whitespace cleanup
2007-06-12 07:29:11 +00:00
lp->lp_softc != sc) {
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
error = ENOENT;
Introduce sleepable locks into if_lagg. Before this change if_lagg was using nonsleepable rmlocks to protect its internal state. This patch introduces another sx lock to protect code paths that require sleeping, while still uses old rmlock to protect hot nonsleepable data paths. This change allows to remove taskqueue decoupling used before to change interface addresses without holding the lock. Instead it uses sx lock to protect direct if_ioctl() calls. As another bonus, the new code synchronizes enabled capabilities of member interfaces, and allows to control them with ifconfig laggX, that was impossible before. This part should fix interoperation with if_bridge, that may need to disable some capabilities, such as TXCSUM or LRO, to allow bridging with noncapable interfaces. MFC after: 2 weeks Sponsored by: iXsystems, Inc. Differential Revision: https://reviews.freebsd.org/D10514
2017-05-02 19:09:11 +00:00
LAGG_XUNLOCK(sc);
Add interface reference counting to if_lagg. Using plain ifunit() looks like request for troubles. MFC after: 2 weeks
2017-04-21 13:45:01 +00:00
if_rele(tpif);
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
break;
}
error = lagg_port_destroy(lp, 1);
Introduce sleepable locks into if_lagg. Before this change if_lagg was using nonsleepable rmlocks to protect its internal state. This patch introduces another sx lock to protect code paths that require sleeping, while still uses old rmlock to protect hot nonsleepable data paths. This change allows to remove taskqueue decoupling used before to change interface addresses without holding the lock. Instead it uses sx lock to protect direct if_ioctl() calls. As another bonus, the new code synchronizes enabled capabilities of member interfaces, and allows to control them with ifconfig laggX, that was impossible before. This part should fix interoperation with if_bridge, that may need to disable some capabilities, such as TXCSUM or LRO, to allow bridging with noncapable interfaces. MFC after: 2 weeks Sponsored by: iXsystems, Inc. Differential Revision: https://reviews.freebsd.org/D10514
2017-05-02 19:09:11 +00:00
LAGG_XUNLOCK(sc);
Add interface reference counting to if_lagg. Using plain ifunit() looks like request for troubles. MFC after: 2 weeks
2017-04-21 13:45:01 +00:00
if_rele(tpif);
Call VLAN_CAPABILITIES() when LAGG capabilities change. This makes VLAN on top of LAGG to expose proper capabilities if they are changed after creation. MFC after: 1 week
2017-05-26 22:22:48 +00:00
VLAN_CAPABILITIES(ifp);
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
break;
case SIOCSIFFLAGS:
/* Set flags on ports too */
Introduce sleepable locks into if_lagg. Before this change if_lagg was using nonsleepable rmlocks to protect its internal state. This patch introduces another sx lock to protect code paths that require sleeping, while still uses old rmlock to protect hot nonsleepable data paths. This change allows to remove taskqueue decoupling used before to change interface addresses without holding the lock. Instead it uses sx lock to protect direct if_ioctl() calls. As another bonus, the new code synchronizes enabled capabilities of member interfaces, and allows to control them with ifconfig laggX, that was impossible before. This part should fix interoperation with if_bridge, that may need to disable some capabilities, such as TXCSUM or LRO, to allow bridging with noncapable interfaces. MFC after: 2 weeks Sponsored by: iXsystems, Inc. Differential Revision: https://reviews.freebsd.org/D10514
2017-05-02 19:09:11 +00:00
LAGG_XLOCK(sc);
Replace rmlock with epoch in lagg Use the new epoch based reclamation API. Now the hot paths will not block at all, and the sx lock is used for the softc data. This fixes LORs reported where the rwlock was obtained when the sxlock was held. Submitted by: mmacy Reported by: Harry Schmalzbauer <freebsd@omnilan.de> Reviewed by: sbruno Sponsored by: Limelight Networks Differential Revision: https://reviews.freebsd.org/D15355
2018-05-14 20:06:49 +00:00
CK_SLIST_FOREACH(lp, &sc->sc_ports, lp_entries) {
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
lagg_setflags(lp, 1);
}
if (!(ifp->if_flags & IFF_UP) &&
(ifp->if_drv_flags & IFF_DRV_RUNNING)) {
/*
* If interface is marked down and it is running,
* then stop and disable it.
*/
lagg_stop(sc);
Introduce sleepable locks into if_lagg. Before this change if_lagg was using nonsleepable rmlocks to protect its internal state. This patch introduces another sx lock to protect code paths that require sleeping, while still uses old rmlock to protect hot nonsleepable data paths. This change allows to remove taskqueue decoupling used before to change interface addresses without holding the lock. Instead it uses sx lock to protect direct if_ioctl() calls. As another bonus, the new code synchronizes enabled capabilities of member interfaces, and allows to control them with ifconfig laggX, that was impossible before. This part should fix interoperation with if_bridge, that may need to disable some capabilities, such as TXCSUM or LRO, to allow bridging with noncapable interfaces. MFC after: 2 weeks Sponsored by: iXsystems, Inc. Differential Revision: https://reviews.freebsd.org/D10514
2017-05-02 19:09:11 +00:00
LAGG_XUNLOCK(sc);
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
} else if ((ifp->if_flags & IFF_UP) &&
!(ifp->if_drv_flags & IFF_DRV_RUNNING)) {
/*
* If interface is marked up and it is stopped, then
* start it.
*/
Introduce sleepable locks into if_lagg. Before this change if_lagg was using nonsleepable rmlocks to protect its internal state. This patch introduces another sx lock to protect code paths that require sleeping, while still uses old rmlock to protect hot nonsleepable data paths. This change allows to remove taskqueue decoupling used before to change interface addresses without holding the lock. Instead it uses sx lock to protect direct if_ioctl() calls. As another bonus, the new code synchronizes enabled capabilities of member interfaces, and allows to control them with ifconfig laggX, that was impossible before. This part should fix interoperation with if_bridge, that may need to disable some capabilities, such as TXCSUM or LRO, to allow bridging with noncapable interfaces. MFC after: 2 weeks Sponsored by: iXsystems, Inc. Differential Revision: https://reviews.freebsd.org/D10514
2017-05-02 19:09:11 +00:00
LAGG_XUNLOCK(sc);
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
(*ifp->if_init)(sc);
Introduce sleepable locks into if_lagg. Before this change if_lagg was using nonsleepable rmlocks to protect its internal state. This patch introduces another sx lock to protect code paths that require sleeping, while still uses old rmlock to protect hot nonsleepable data paths. This change allows to remove taskqueue decoupling used before to change interface addresses without holding the lock. Instead it uses sx lock to protect direct if_ioctl() calls. As another bonus, the new code synchronizes enabled capabilities of member interfaces, and allows to control them with ifconfig laggX, that was impossible before. This part should fix interoperation with if_bridge, that may need to disable some capabilities, such as TXCSUM or LRO, to allow bridging with noncapable interfaces. MFC after: 2 weeks Sponsored by: iXsystems, Inc. Differential Revision: https://reviews.freebsd.org/D10514
2017-05-02 19:09:11 +00:00
} else
LAGG_XUNLOCK(sc);
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
break;
case SIOCADDMULTI:
case SIOCDELMULTI:
Replace rmlock with epoch in lagg Use the new epoch based reclamation API. Now the hot paths will not block at all, and the sx lock is used for the softc data. This fixes LORs reported where the rwlock was obtained when the sxlock was held. Submitted by: mmacy Reported by: Harry Schmalzbauer <freebsd@omnilan.de> Reviewed by: sbruno Sponsored by: Limelight Networks Differential Revision: https://reviews.freebsd.org/D15355
2018-05-14 20:06:49 +00:00
LAGG_XLOCK(sc);
CK_SLIST_FOREACH(lp, &sc->sc_ports, lp_entries) {
Introduce sleepable locks into if_lagg. Before this change if_lagg was using nonsleepable rmlocks to protect its internal state. This patch introduces another sx lock to protect code paths that require sleeping, while still uses old rmlock to protect hot nonsleepable data paths. This change allows to remove taskqueue decoupling used before to change interface addresses without holding the lock. Instead it uses sx lock to protect direct if_ioctl() calls. As another bonus, the new code synchronizes enabled capabilities of member interfaces, and allows to control them with ifconfig laggX, that was impossible before. This part should fix interoperation with if_bridge, that may need to disable some capabilities, such as TXCSUM or LRO, to allow bridging with noncapable interfaces. MFC after: 2 weeks Sponsored by: iXsystems, Inc. Differential Revision: https://reviews.freebsd.org/D10514
2017-05-02 19:09:11 +00:00
lagg_clrmulti(lp);
lagg_setmulti(lp);
}
Replace rmlock with epoch in lagg Use the new epoch based reclamation API. Now the hot paths will not block at all, and the sx lock is used for the softc data. This fixes LORs reported where the rwlock was obtained when the sxlock was held. Submitted by: mmacy Reported by: Harry Schmalzbauer <freebsd@omnilan.de> Reviewed by: sbruno Sponsored by: Limelight Networks Differential Revision: https://reviews.freebsd.org/D15355
2018-05-14 20:06:49 +00:00
LAGG_XUNLOCK(sc);
Introduce sleepable locks into if_lagg. Before this change if_lagg was using nonsleepable rmlocks to protect its internal state. This patch introduces another sx lock to protect code paths that require sleeping, while still uses old rmlock to protect hot nonsleepable data paths. This change allows to remove taskqueue decoupling used before to change interface addresses without holding the lock. Instead it uses sx lock to protect direct if_ioctl() calls. As another bonus, the new code synchronizes enabled capabilities of member interfaces, and allows to control them with ifconfig laggX, that was impossible before. This part should fix interoperation with if_bridge, that may need to disable some capabilities, such as TXCSUM or LRO, to allow bridging with noncapable interfaces. MFC after: 2 weeks Sponsored by: iXsystems, Inc. Differential Revision: https://reviews.freebsd.org/D10514
2017-05-02 19:09:11 +00:00
error = 0;
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
break;
case SIOCSIFMEDIA:
case SIOCGIFMEDIA:
Add support for IP over infiniband, IPoIB, to lagg(4). Currently only the failover protocol is supported due to limitations in the IPoIB architecture. Refer to the lagg(4) manual page for how to configure and use this new feature. A new network interface type, IFT_INFINIBANDLAG, has been added, similar to the existing IFT_IEEE8023ADLAG . ifconfig(8) has been updated to accept a new laggtype argument when creating lagg(4) network interfaces. This new argument is used to distinguish between ethernet and infiniband type of lagg(4) network interface. The laggtype argument is optional and defaults to ethernet. The lagg(4) command line syntax is backwards compatible. Differential Revision: https://reviews.freebsd.org/D26254 Reviewed by: melifaro@ MFC after: 1 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2020-10-22 09:47:12 +00:00
if (ifp->if_type == IFT_INFINIBAND)
error = EINVAL;
else
error = ifmedia_ioctl(ifp, ifr, &sc->sc_media, cmd);
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
break;
- Propagate the largest set of interface capabilities supported by all lagg ports to the lagg interface. - Use the MTU from the first interface as the lagg MTU, all extra interfaces must be the same. This fixes using a lagg interface for a vlan or enabling jumbo frames, etc. Approved by: re (kensmith) MFC After: 3 days
2007-07-30 20:17:22 +00:00
case SIOCSIFCAP:
lagg: fix lagg ifioctl after SIOCSIFCAPNV Lagg was broken by SIOCSIFCAPNV when all underlying devices support SIOCSIFCAPNV. This change updates lagg to work with SIOCSIFCAPNV and if_capabilities2. Reviewed by: kib, hselasky Sponsored by: Netflix Differential Revision: https://reviews.freebsd.org/D35865
2022-07-28 14:36:22 +00:00
case SIOCSIFCAPNV:
Introduce sleepable locks into if_lagg. Before this change if_lagg was using nonsleepable rmlocks to protect its internal state. This patch introduces another sx lock to protect code paths that require sleeping, while still uses old rmlock to protect hot nonsleepable data paths. This change allows to remove taskqueue decoupling used before to change interface addresses without holding the lock. Instead it uses sx lock to protect direct if_ioctl() calls. As another bonus, the new code synchronizes enabled capabilities of member interfaces, and allows to control them with ifconfig laggX, that was impossible before. This part should fix interoperation with if_bridge, that may need to disable some capabilities, such as TXCSUM or LRO, to allow bridging with noncapable interfaces. MFC after: 2 weeks Sponsored by: iXsystems, Inc. Differential Revision: https://reviews.freebsd.org/D10514
2017-05-02 19:09:11 +00:00
LAGG_XLOCK(sc);
Replace rmlock with epoch in lagg Use the new epoch based reclamation API. Now the hot paths will not block at all, and the sx lock is used for the softc data. This fixes LORs reported where the rwlock was obtained when the sxlock was held. Submitted by: mmacy Reported by: Harry Schmalzbauer <freebsd@omnilan.de> Reviewed by: sbruno Sponsored by: Limelight Networks Differential Revision: https://reviews.freebsd.org/D15355
2018-05-14 20:06:49 +00:00
CK_SLIST_FOREACH(lp, &sc->sc_ports, lp_entries) {
Introduce sleepable locks into if_lagg. Before this change if_lagg was using nonsleepable rmlocks to protect its internal state. This patch introduces another sx lock to protect code paths that require sleeping, while still uses old rmlock to protect hot nonsleepable data paths. This change allows to remove taskqueue decoupling used before to change interface addresses without holding the lock. Instead it uses sx lock to protect direct if_ioctl() calls. As another bonus, the new code synchronizes enabled capabilities of member interfaces, and allows to control them with ifconfig laggX, that was impossible before. This part should fix interoperation with if_bridge, that may need to disable some capabilities, such as TXCSUM or LRO, to allow bridging with noncapable interfaces. MFC after: 2 weeks Sponsored by: iXsystems, Inc. Differential Revision: https://reviews.freebsd.org/D10514
2017-05-02 19:09:11 +00:00
if (lp->lp_ioctl != NULL)
(*lp->lp_ioctl)(lp->lp_ifp, cmd, data);
}
lagg_capabilities(sc);
LAGG_XUNLOCK(sc);
Call VLAN_CAPABILITIES() when LAGG capabilities change. This makes VLAN on top of LAGG to expose proper capabilities if they are changed after creation. MFC after: 1 week
2017-05-26 22:22:48 +00:00
VLAN_CAPABILITIES(ifp);
Introduce sleepable locks into if_lagg. Before this change if_lagg was using nonsleepable rmlocks to protect its internal state. This patch introduces another sx lock to protect code paths that require sleeping, while still uses old rmlock to protect hot nonsleepable data paths. This change allows to remove taskqueue decoupling used before to change interface addresses without holding the lock. Instead it uses sx lock to protect direct if_ioctl() calls. As another bonus, the new code synchronizes enabled capabilities of member interfaces, and allows to control them with ifconfig laggX, that was impossible before. This part should fix interoperation with if_bridge, that may need to disable some capabilities, such as TXCSUM or LRO, to allow bridging with noncapable interfaces. MFC after: 2 weeks Sponsored by: iXsystems, Inc. Differential Revision: https://reviews.freebsd.org/D10514
2017-05-02 19:09:11 +00:00
error = 0;
break;
lagg: fix lagg ifioctl after SIOCSIFCAPNV Lagg was broken by SIOCSIFCAPNV when all underlying devices support SIOCSIFCAPNV. This change updates lagg to work with SIOCSIFCAPNV and if_capabilities2. Reviewed by: kib, hselasky Sponsored by: Netflix Differential Revision: https://reviews.freebsd.org/D35865
2022-07-28 14:36:22 +00:00
case SIOCGIFCAPNV:
error = 0;
break;
Storage Controller Interface driver - typo in unimplemented macro in scic_sds_controller_registers.h s/contoller/controller/ PR: 207336 Submitted by: Tony Narlock <tony @ git-pull.com>
2016-04-06 04:50:28 +00:00
case SIOCSIFMTU:
Allow changing lagg(4) MTU. Previously, changing the MTU would require destroying the lagg and creating a new one. Now it is allowed to change the MTU of the lagg interface and the MTU of the ports will be set to match. If any port cannot set the new MTU, all ports are reverted to the original MTU of the lagg. Additionally, when adding ports, the MTU of a port will be automatically set to the MTU of the lagg. As always, the MTU of the lagg is initially determined by the MTU of the first port added. If adding an interface as a port for some reason fails, that interface is reverted to its original MTU. Submitted by: Ryan Moeller <ryan@freqlabs.com> Reviewed by: mav Relnotes: Yes Sponsored by: iXsystems Inc. Differential Revision: https://reviews.freebsd.org/D17576
2018-10-30 09:53:57 +00:00
LAGG_XLOCK(sc);
CK_SLIST_FOREACH(lp, &sc->sc_ports, lp_entries) {
if (lp->lp_ioctl != NULL)
error = (*lp->lp_ioctl)(lp->lp_ifp, cmd, data);
else
error = EINVAL;
if (error != 0) {
if_printf(ifp,
"failed to change MTU to %d on port %s, "
"reverting all ports to original MTU (%d)\n",
ifr->ifr_mtu, lp->lp_ifp->if_xname, ifp->if_mtu);
break;
}
}
if (error == 0) {
ifp->if_mtu = ifr->ifr_mtu;
} else {
/* set every port back to the original MTU */
ifr->ifr_mtu = ifp->if_mtu;
CK_SLIST_FOREACH(lp, &sc->sc_ports, lp_entries) {
if (lp->lp_ioctl != NULL)
(*lp->lp_ioctl)(lp->lp_ifp, cmd, data);
}
}
lagg: update capabilites on SIOCSIFMTU Some NICs might have limited capabilities when Jumbo frames are used. For exampe some neta interfaces only support TX csum offload when the packet size is lower than a value specified in DT. Fix it by re-reading capabilities of children interfaces after MTU has been successfully changed. Found by: Jerome Tomczyk <jerome.tomczyk@stormshield.eu> Reviewed by: jhb Obtained from: Semihalf Sponsored by: Stormshield Differential revision: https://reviews.freebsd.org/D32724
2021-11-06 09:41:43 +00:00
lagg_capabilities(sc);
Allow changing lagg(4) MTU. Previously, changing the MTU would require destroying the lagg and creating a new one. Now it is allowed to change the MTU of the lagg interface and the MTU of the ports will be set to match. If any port cannot set the new MTU, all ports are reverted to the original MTU of the lagg. Additionally, when adding ports, the MTU of a port will be automatically set to the MTU of the lagg. As always, the MTU of the lagg is initially determined by the MTU of the first port added. If adding an interface as a port for some reason fails, that interface is reverted to its original MTU. Submitted by: Ryan Moeller <ryan@freqlabs.com> Reviewed by: mav Relnotes: Yes Sponsored by: iXsystems Inc. Differential Revision: https://reviews.freebsd.org/D17576
2018-10-30 09:53:57 +00:00
LAGG_XUNLOCK(sc);
lagg: update capabilites on SIOCSIFMTU Some NICs might have limited capabilities when Jumbo frames are used. For exampe some neta interfaces only support TX csum offload when the packet size is lower than a value specified in DT. Fix it by re-reading capabilities of children interfaces after MTU has been successfully changed. Found by: Jerome Tomczyk <jerome.tomczyk@stormshield.eu> Reviewed by: jhb Obtained from: Semihalf Sponsored by: Stormshield Differential revision: https://reviews.freebsd.org/D32724
2021-11-06 09:41:43 +00:00
VLAN_CAPABILITIES(ifp);
Storage Controller Interface driver - typo in unimplemented macro in scic_sds_controller_registers.h s/contoller/controller/ PR: 207336 Submitted by: Tony Narlock <tony @ git-pull.com>
2016-04-06 04:50:28 +00:00
break;
- Propagate the largest set of interface capabilities supported by all lagg ports to the lagg interface. - Use the MTU from the first interface as the lagg MTU, all extra interfaces must be the same. This fixes using a lagg interface for a vlan or enabling jumbo frames, etc. Approved by: re (kensmith) MFC After: 3 days
2007-07-30 20:17:22 +00:00
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
default:
error = ether_ioctl(ifp, cmd, data);
break;
}
return (error);
}
Add kernel-side support for in-kernel TLS. KTLS adds support for in-kernel framing and encryption of Transport Layer Security (1.0-1.2) data on TCP sockets. KTLS only supports offload of TLS for transmitted data. Key negotation must still be performed in userland. Once completed, transmit session keys for a connection are provided to the kernel via a new TCP_TXTLS_ENABLE socket option. All subsequent data transmitted on the socket is placed into TLS frames and encrypted using the supplied keys. Any data written to a KTLS-enabled socket via write(2), aio_write(2), or sendfile(2) is assumed to be application data and is encoded in TLS frames with an application data type. Individual records can be sent with a custom type (e.g. handshake messages) via sendmsg(2) with a new control message (TLS_SET_RECORD_TYPE) specifying the record type. At present, rekeying is not supported though the in-kernel framework should support rekeying. KTLS makes use of the recently added unmapped mbufs to store TLS frames in the socket buffer. Each TLS frame is described by a single ext_pgs mbuf. The ext_pgs structure contains the header of the TLS record (and trailer for encrypted records) as well as references to the associated TLS session. KTLS supports two primary methods of encrypting TLS frames: software TLS and ifnet TLS. Software TLS marks mbufs holding socket data as not ready via M_NOTREADY similar to sendfile(2) when TLS framing information is added to an unmapped mbuf in ktls_frame(). ktls_enqueue() is then called to schedule TLS frames for encryption. In the case of sendfile_iodone() calls ktls_enqueue() instead of pru_ready() leaving the mbufs marked M_NOTREADY until encryption is completed. For other writes (vn_sendfile when pages are available, write(2), etc.), the PRUS_NOTREADY is set when invoking pru_send() along with invoking ktls_enqueue(). A pool of worker threads (the "KTLS" kernel process) encrypts TLS frames queued via ktls_enqueue(). Each TLS frame is temporarily mapped using the direct map and passed to a software encryption backend to perform the actual encryption. (Note: The use of PHYS_TO_DMAP could be replaced with sf_bufs if someone wished to make this work on architectures without a direct map.) KTLS supports pluggable software encryption backends. Internally, Netflix uses proprietary pure-software backends. This commit includes a simple backend in a new ktls_ocf.ko module that uses the kernel's OpenCrypto framework to provide AES-GCM encryption of TLS frames. As a result, software TLS is now a bit of a misnomer as it can make use of hardware crypto accelerators. Once software encryption has finished, the TLS frame mbufs are marked ready via pru_ready(). At this point, the encrypted data appears as regular payload to the TCP stack stored in unmapped mbufs. ifnet TLS permits a NIC to offload the TLS encryption and TCP segmentation. In this mode, a new send tag type (IF_SND_TAG_TYPE_TLS) is allocated on the interface a socket is routed over and associated with a TLS session. TLS records for a TLS session using ifnet TLS are not marked M_NOTREADY but are passed down the stack unencrypted. The ip_output_send() and ip6_output_send() helper functions that apply send tags to outbound IP packets verify that the send tag of the TLS record matches the outbound interface. If so, the packet is tagged with the TLS send tag and sent to the interface. The NIC device driver must recognize packets with the TLS send tag and schedule them for TLS encryption and TCP segmentation. If the the outbound interface does not match the interface in the TLS send tag, the packet is dropped. In addition, a task is scheduled to refresh the TLS send tag for the TLS session. If a new TLS send tag cannot be allocated, the connection is dropped. If a new TLS send tag is allocated, however, subsequent packets will be tagged with the correct TLS send tag. (This latter case has been tested by configuring both ports of a Chelsio T6 in a lagg and failing over from one port to another. As the connections migrated to the new port, new TLS send tags were allocated for the new port and connections resumed without being dropped.) ifnet TLS can be enabled and disabled on supported network interfaces via new '[-]txtls[46]' options to ifconfig(8). ifnet TLS is supported across both vlan devices and lagg interfaces using failover, lacp with flowid enabled, or lacp with flowid enabled. Applications may request the current KTLS mode of a connection via a new TCP_TXTLS_MODE socket option. They can also use this socket option to toggle between software and ifnet TLS modes. In addition, a testing tool is available in tools/tools/switch_tls. This is modeled on tcpdrop and uses similar syntax. However, instead of dropping connections, -s is used to force KTLS connections to switch to software TLS and -i is used to switch to ifnet TLS. Various sysctls and counters are available under the kern.ipc.tls sysctl node. The kern.ipc.tls.enable node must be set to true to enable KTLS (it is off by default). The use of unmapped mbufs must also be enabled via kern.ipc.mb_use_ext_pgs to enable KTLS. KTLS is enabled via the KERN_TLS kernel option. This patch is the culmination of years of work by several folks including Scott Long and Randall Stewart for the original design and implementation; Drew Gallatin for several optimizations including the use of ext_pgs mbufs, the M_NOTREADY mechanism for TLS records awaiting software encryption, and pluggable software crypto backends; and John Baldwin for modifications to support hardware TLS offload. Reviewed by: gallatin, hselasky, rrs Obtained from: Netflix Sponsored by: Netflix, Chelsio Communications Differential Revision: https://reviews.freebsd.org/D21277
2019-08-27 00:01:56 +00:00
#if defined(KERN_TLS) || defined(RATELIMIT)
Add a switch structure for send tags. Move the type and function pointers for operations on existing send tags (modify, query, next, free) out of 'struct ifnet' and into a new 'struct if_snd_tag_sw'. A pointer to this structure is added to the generic part of send tags and is initialized by m_snd_tag_init() (which now accepts a switch structure as a new argument in place of the type). Previously, device driver ifnet methods switched on the type to call type-specific functions. Now, those type-specific functions are saved in the switch structure and invoked directly. In addition, this more gracefully permits multiple implementations of the same tag within a driver. In particular, NIC TLS for future Chelsio adapters will use a different implementation than the existing NIC TLS support for T6 adapters. Reviewed by: gallatin, hselasky, kib (older version) Sponsored by: Chelsio Communications Differential Revision: https://reviews.freebsd.org/D31572
2021-09-14 18:43:41 +00:00
#ifdef RATELIMIT
static const struct if_snd_tag_sw lagg_snd_tag_ul_sw = {
.snd_tag_modify = lagg_snd_tag_modify,
.snd_tag_query = lagg_snd_tag_query,
.snd_tag_free = lagg_snd_tag_free,
.next_snd_tag = lagg_next_snd_tag,
.type = IF_SND_TAG_TYPE_UNLIMITED
};
static const struct if_snd_tag_sw lagg_snd_tag_rl_sw = {
.snd_tag_modify = lagg_snd_tag_modify,
.snd_tag_query = lagg_snd_tag_query,
.snd_tag_free = lagg_snd_tag_free,
.next_snd_tag = lagg_next_snd_tag,
.type = IF_SND_TAG_TYPE_RATE_LIMIT
};
#endif
#ifdef KERN_TLS
static const struct if_snd_tag_sw lagg_snd_tag_tls_sw = {
.snd_tag_modify = lagg_snd_tag_modify,
.snd_tag_query = lagg_snd_tag_query,
.snd_tag_free = lagg_snd_tag_free,
.next_snd_tag = lagg_next_snd_tag,
.type = IF_SND_TAG_TYPE_TLS
};
#ifdef RATELIMIT
static const struct if_snd_tag_sw lagg_snd_tag_tls_rl_sw = {
.snd_tag_modify = lagg_snd_tag_modify,
.snd_tag_query = lagg_snd_tag_query,
.snd_tag_free = lagg_snd_tag_free,
.next_snd_tag = lagg_next_snd_tag,
.type = IF_SND_TAG_TYPE_TLS_RATE_LIMIT
};
#endif
#endif
Restructure mbuf send tags to provide stronger guarantees. - Perform ifp mismatch checks (to determine if a send tag is allocated for a different ifp than the one the packet is being output on), in ip_output() and ip6_output(). This avoids sending packets with send tags to ifnet drivers that don't support send tags. Since we are now checking for ifp mismatches before invoking if_output, we can now try to allocate a new tag before invoking if_output sending the original packet on the new tag if allocation succeeds. To avoid code duplication for the fragment and unfragmented cases, add ip_output_send() and ip6_output_send() as wrappers around if_output and nd6_output_ifp, respectively. All of the logic for setting send tags and dealing with send tag-related errors is done in these wrapper functions. For pseudo interfaces that wrap other network interfaces (vlan and lagg), wrapper send tags are now allocated so that ip*_output see the wrapper ifp as the ifp in the send tag. The if_transmit routines rewrite the send tags after performing an ifp mismatch check. If an ifp mismatch is detected, the transmit routines fail with EAGAIN. - To provide clearer life cycle management of send tags, especially in the presence of vlan and lagg wrapper tags, add a reference count to send tags managed via m_snd_tag_ref() and m_snd_tag_rele(). Provide a helper function (m_snd_tag_init()) for use by drivers supporting send tags. m_snd_tag_init() takes care of the if_ref on the ifp meaning that code alloating send tags via if_snd_tag_alloc no longer has to manage that manually. Similarly, m_snd_tag_rele drops the refcount on the ifp after invoking if_snd_tag_free when the last reference to a send tag is dropped. This also closes use after free races if there are pending packets in driver tx rings after the socket is closed (e.g. from tcpdrop). In order for m_free to work reliably, add a new CSUM_SND_TAG flag in csum_flags to indicate 'snd_tag' is set (rather than 'rcvif'). Drivers now also check this flag instead of checking snd_tag against NULL. This avoids false positive matches when a forwarded packet has a non-NULL rcvif that was treated as a send tag. - cxgbe was relying on snd_tag_free being called when the inp was detached so that it could kick the firmware to flush any pending work on the flow. This is because the driver doesn't require ACK messages from the firmware for every request, but instead does a kind of manual interrupt coalescing by only setting a flag to request a completion on a subset of requests. If all of the in-flight requests don't have the flag when the tag is detached from the inp, the flow might never return the credits. The current snd_tag_free command issues a flush command to force the credits to return. However, the credit return is what also frees the mbufs, and since those mbufs now hold references on the tag, this meant that snd_tag_free would never be called. To fix, explicitly drop the mbuf's reference on the snd tag when the mbuf is queued in the firmware work queue. This means that once the inp's reference on the tag goes away and all in-flight mbufs have been queued to the firmware, tag's refcount will drop to zero and snd_tag_free will kick in and send the flush request. Note that we need to avoid doing this in the middle of ethofld_tx(), so the driver grabs a temporary reference on the tag around that loop to defer the free to the end of the function in case it sends the last mbuf to the queue after the inp has dropped its reference on the tag. - mlx5 preallocates send tags and was using the ifp pointer even when the send tag wasn't in use. Explicitly use the ifp from other data structures instead. - Sprinkle some assertions in various places to assert that received packets don't have a send tag, and that other places that overwrite rcvif (e.g. 802.11 transmit) don't clobber a send tag pointer. Reviewed by: gallatin, hselasky, rgrimes, ae Sponsored by: Netflix Differential Revision: https://reviews.freebsd.org/D20117
2019-05-24 22:30:40 +00:00
static inline struct lagg_snd_tag *
mst_to_lst(struct m_snd_tag *mst)
Implement kernel support for hardware rate limited sockets. - Add RATELIMIT kernel configuration keyword which must be set to enable the new functionality. - Add support for hardware driven, Receive Side Scaling, RSS aware, rate limited sendqueues and expose the functionality through the already established SO_MAX_PACING_RATE setsockopt(). The API support rates in the range from 1 to 4Gbytes/s which are suitable for regular TCP and UDP streams. The setsockopt(2) manual page has been updated. - Add rate limit function callback API to "struct ifnet" which supports the following operations: if_snd_tag_alloc(), if_snd_tag_modify(), if_snd_tag_query() and if_snd_tag_free(). - Add support to ifconfig to view, set and clear the IFCAP_TXRTLMT flag, which tells if a network driver supports rate limiting or not. - This patch also adds support for rate limiting through VLAN and LAGG intermediate network devices. - How rate limiting works: 1) The userspace application calls setsockopt() after accepting or making a new connection to set the rate which is then stored in the socket structure in the kernel. Later on when packets are transmitted a check is made in the transmit path for rate changes. A rate change implies a non-blocking ifp->if_snd_tag_alloc() call will be made to the destination network interface, which then sets up a custom sendqueue with the given rate limitation parameter. A "struct m_snd_tag" pointer is returned which serves as a "snd_tag" hint in the m_pkthdr for the subsequently transmitted mbufs. 2) When the network driver sees the "m->m_pkthdr.snd_tag" different from NULL, it will move the packets into a designated rate limited sendqueue given by the snd_tag pointer. It is up to the individual drivers how the rate limited traffic will be rate limited. 3) Route changes are detected by the NIC drivers in the ifp->if_transmit() routine when the ifnet pointer in the incoming snd_tag mismatches the one of the network interface. The network adapter frees the mbuf and returns EAGAIN which causes the ip_output() to release and clear the send tag. Upon next ip_output() a new "snd_tag" will be tried allocated. 4) When the PCB is detached the custom sendqueue will be released by a non-blocking ifp->if_snd_tag_free() call to the currently bound network interface. Reviewed by: wblock (manpages), adrian, gallatin, scottl (network) Differential Revision: https://reviews.freebsd.org/D3687 Sponsored by: Mellanox Technologies MFC after: 3 months
2017-01-18 13:31:17 +00:00
{
Restructure mbuf send tags to provide stronger guarantees. - Perform ifp mismatch checks (to determine if a send tag is allocated for a different ifp than the one the packet is being output on), in ip_output() and ip6_output(). This avoids sending packets with send tags to ifnet drivers that don't support send tags. Since we are now checking for ifp mismatches before invoking if_output, we can now try to allocate a new tag before invoking if_output sending the original packet on the new tag if allocation succeeds. To avoid code duplication for the fragment and unfragmented cases, add ip_output_send() and ip6_output_send() as wrappers around if_output and nd6_output_ifp, respectively. All of the logic for setting send tags and dealing with send tag-related errors is done in these wrapper functions. For pseudo interfaces that wrap other network interfaces (vlan and lagg), wrapper send tags are now allocated so that ip*_output see the wrapper ifp as the ifp in the send tag. The if_transmit routines rewrite the send tags after performing an ifp mismatch check. If an ifp mismatch is detected, the transmit routines fail with EAGAIN. - To provide clearer life cycle management of send tags, especially in the presence of vlan and lagg wrapper tags, add a reference count to send tags managed via m_snd_tag_ref() and m_snd_tag_rele(). Provide a helper function (m_snd_tag_init()) for use by drivers supporting send tags. m_snd_tag_init() takes care of the if_ref on the ifp meaning that code alloating send tags via if_snd_tag_alloc no longer has to manage that manually. Similarly, m_snd_tag_rele drops the refcount on the ifp after invoking if_snd_tag_free when the last reference to a send tag is dropped. This also closes use after free races if there are pending packets in driver tx rings after the socket is closed (e.g. from tcpdrop). In order for m_free to work reliably, add a new CSUM_SND_TAG flag in csum_flags to indicate 'snd_tag' is set (rather than 'rcvif'). Drivers now also check this flag instead of checking snd_tag against NULL. This avoids false positive matches when a forwarded packet has a non-NULL rcvif that was treated as a send tag. - cxgbe was relying on snd_tag_free being called when the inp was detached so that it could kick the firmware to flush any pending work on the flow. This is because the driver doesn't require ACK messages from the firmware for every request, but instead does a kind of manual interrupt coalescing by only setting a flag to request a completion on a subset of requests. If all of the in-flight requests don't have the flag when the tag is detached from the inp, the flow might never return the credits. The current snd_tag_free command issues a flush command to force the credits to return. However, the credit return is what also frees the mbufs, and since those mbufs now hold references on the tag, this meant that snd_tag_free would never be called. To fix, explicitly drop the mbuf's reference on the snd tag when the mbuf is queued in the firmware work queue. This means that once the inp's reference on the tag goes away and all in-flight mbufs have been queued to the firmware, tag's refcount will drop to zero and snd_tag_free will kick in and send the flush request. Note that we need to avoid doing this in the middle of ethofld_tx(), so the driver grabs a temporary reference on the tag around that loop to defer the free to the end of the function in case it sends the last mbuf to the queue after the inp has dropped its reference on the tag. - mlx5 preallocates send tags and was using the ifp pointer even when the send tag wasn't in use. Explicitly use the ifp from other data structures instead. - Sprinkle some assertions in various places to assert that received packets don't have a send tag, and that other places that overwrite rcvif (e.g. 802.11 transmit) don't clobber a send tag pointer. Reviewed by: gallatin, hselasky, rgrimes, ae Sponsored by: Netflix Differential Revision: https://reviews.freebsd.org/D20117
2019-05-24 22:30:40 +00:00
return (__containerof(mst, struct lagg_snd_tag, com));
}
/*
* Look up the port used by a specific flow. This only works for lagg
* protocols with deterministic port mappings (e.g. not roundrobin).
* In addition protocols which use a hash to map flows to ports must
* be configured to use the mbuf flowid rather than hashing packet
* contents.
*/
static struct lagg_port *
make lacp's use_numa hashing aware of send tags When I did the use_numa support, I missed the fact that there is a separate hash function for send tag nic selection. So when use_numa is enabled, ktls offload does not work properly, as it does not reliably allocate a send tag on the proper egress nic since different egress nics are selected for send-tag allocation and packet transmit. To fix this, this change: - refectors lacp_select_tx_port_by_hash() and lacp_select_tx_port() to make lacp_select_tx_port_by_hash() always called by lacp_select_tx_port() - pre-shifts flowids to convert them to hashes when calling lacp_select_tx_port_by_hash() - adds a numa_domain field to if_snd_tag_alloc_params - plumbs the numa domain into places where we allocate send tags In testing with NIC TLS setup on a NUMA machine, I see thousands of output errors before the change when enabling kern.ipc.tls.ifnet.permitted=1. After the change, I see no errors, and I see the NIC sysctl counters showing active TLS offload sessions. Reviewed by: rrs, hselasky, jhb Sponsored by: Netflix
2020-03-09 13:44:51 +00:00
lookup_snd_tag_port(struct ifnet *ifp, uint32_t flowid, uint32_t flowtype,
uint8_t numa_domain)
Restructure mbuf send tags to provide stronger guarantees. - Perform ifp mismatch checks (to determine if a send tag is allocated for a different ifp than the one the packet is being output on), in ip_output() and ip6_output(). This avoids sending packets with send tags to ifnet drivers that don't support send tags. Since we are now checking for ifp mismatches before invoking if_output, we can now try to allocate a new tag before invoking if_output sending the original packet on the new tag if allocation succeeds. To avoid code duplication for the fragment and unfragmented cases, add ip_output_send() and ip6_output_send() as wrappers around if_output and nd6_output_ifp, respectively. All of the logic for setting send tags and dealing with send tag-related errors is done in these wrapper functions. For pseudo interfaces that wrap other network interfaces (vlan and lagg), wrapper send tags are now allocated so that ip*_output see the wrapper ifp as the ifp in the send tag. The if_transmit routines rewrite the send tags after performing an ifp mismatch check. If an ifp mismatch is detected, the transmit routines fail with EAGAIN. - To provide clearer life cycle management of send tags, especially in the presence of vlan and lagg wrapper tags, add a reference count to send tags managed via m_snd_tag_ref() and m_snd_tag_rele(). Provide a helper function (m_snd_tag_init()) for use by drivers supporting send tags. m_snd_tag_init() takes care of the if_ref on the ifp meaning that code alloating send tags via if_snd_tag_alloc no longer has to manage that manually. Similarly, m_snd_tag_rele drops the refcount on the ifp after invoking if_snd_tag_free when the last reference to a send tag is dropped. This also closes use after free races if there are pending packets in driver tx rings after the socket is closed (e.g. from tcpdrop). In order for m_free to work reliably, add a new CSUM_SND_TAG flag in csum_flags to indicate 'snd_tag' is set (rather than 'rcvif'). Drivers now also check this flag instead of checking snd_tag against NULL. This avoids false positive matches when a forwarded packet has a non-NULL rcvif that was treated as a send tag. - cxgbe was relying on snd_tag_free being called when the inp was detached so that it could kick the firmware to flush any pending work on the flow. This is because the driver doesn't require ACK messages from the firmware for every request, but instead does a kind of manual interrupt coalescing by only setting a flag to request a completion on a subset of requests. If all of the in-flight requests don't have the flag when the tag is detached from the inp, the flow might never return the credits. The current snd_tag_free command issues a flush command to force the credits to return. However, the credit return is what also frees the mbufs, and since those mbufs now hold references on the tag, this meant that snd_tag_free would never be called. To fix, explicitly drop the mbuf's reference on the snd tag when the mbuf is queued in the firmware work queue. This means that once the inp's reference on the tag goes away and all in-flight mbufs have been queued to the firmware, tag's refcount will drop to zero and snd_tag_free will kick in and send the flush request. Note that we need to avoid doing this in the middle of ethofld_tx(), so the driver grabs a temporary reference on the tag around that loop to defer the free to the end of the function in case it sends the last mbuf to the queue after the inp has dropped its reference on the tag. - mlx5 preallocates send tags and was using the ifp pointer even when the send tag wasn't in use. Explicitly use the ifp from other data structures instead. - Sprinkle some assertions in various places to assert that received packets don't have a send tag, and that other places that overwrite rcvif (e.g. 802.11 transmit) don't clobber a send tag pointer. Reviewed by: gallatin, hselasky, rgrimes, ae Sponsored by: Netflix Differential Revision: https://reviews.freebsd.org/D20117
2019-05-24 22:30:40 +00:00
{
struct lagg_softc *sc;
Implement kernel support for hardware rate limited sockets. - Add RATELIMIT kernel configuration keyword which must be set to enable the new functionality. - Add support for hardware driven, Receive Side Scaling, RSS aware, rate limited sendqueues and expose the functionality through the already established SO_MAX_PACING_RATE setsockopt(). The API support rates in the range from 1 to 4Gbytes/s which are suitable for regular TCP and UDP streams. The setsockopt(2) manual page has been updated. - Add rate limit function callback API to "struct ifnet" which supports the following operations: if_snd_tag_alloc(), if_snd_tag_modify(), if_snd_tag_query() and if_snd_tag_free(). - Add support to ifconfig to view, set and clear the IFCAP_TXRTLMT flag, which tells if a network driver supports rate limiting or not. - This patch also adds support for rate limiting through VLAN and LAGG intermediate network devices. - How rate limiting works: 1) The userspace application calls setsockopt() after accepting or making a new connection to set the rate which is then stored in the socket structure in the kernel. Later on when packets are transmitted a check is made in the transmit path for rate changes. A rate change implies a non-blocking ifp->if_snd_tag_alloc() call will be made to the destination network interface, which then sets up a custom sendqueue with the given rate limitation parameter. A "struct m_snd_tag" pointer is returned which serves as a "snd_tag" hint in the m_pkthdr for the subsequently transmitted mbufs. 2) When the network driver sees the "m->m_pkthdr.snd_tag" different from NULL, it will move the packets into a designated rate limited sendqueue given by the snd_tag pointer. It is up to the individual drivers how the rate limited traffic will be rate limited. 3) Route changes are detected by the NIC drivers in the ifp->if_transmit() routine when the ifnet pointer in the incoming snd_tag mismatches the one of the network interface. The network adapter frees the mbuf and returns EAGAIN which causes the ip_output() to release and clear the send tag. Upon next ip_output() a new "snd_tag" will be tried allocated. 4) When the PCB is detached the custom sendqueue will be released by a non-blocking ifp->if_snd_tag_free() call to the currently bound network interface. Reviewed by: wblock (manpages), adrian, gallatin, scottl (network) Differential Revision: https://reviews.freebsd.org/D3687 Sponsored by: Mellanox Technologies MFC after: 3 months
2017-01-18 13:31:17 +00:00
struct lagg_port *lp;
struct lagg_lb *lb;
make lacp's use_numa hashing aware of send tags When I did the use_numa support, I missed the fact that there is a separate hash function for send tag nic selection. So when use_numa is enabled, ktls offload does not work properly, as it does not reliably allocate a send tag on the proper egress nic since different egress nics are selected for send-tag allocation and packet transmit. To fix this, this change: - refectors lacp_select_tx_port_by_hash() and lacp_select_tx_port() to make lacp_select_tx_port_by_hash() always called by lacp_select_tx_port() - pre-shifts flowids to convert them to hashes when calling lacp_select_tx_port_by_hash() - adds a numa_domain field to if_snd_tag_alloc_params - plumbs the numa domain into places where we allocate send tags In testing with NIC TLS setup on a NUMA machine, I see thousands of output errors before the change when enabling kern.ipc.tls.ifnet.permitted=1. After the change, I see no errors, and I see the NIC sysctl counters showing active TLS offload sessions. Reviewed by: rrs, hselasky, jhb Sponsored by: Netflix
2020-03-09 13:44:51 +00:00
uint32_t hash, p;
LACP: When suppressing distributing, return ENOBUFS When links come and go, lacp goes into a "suppress distributing" mode where it drops traffic for 3 seconds. When in this mode, lagg/lacp historiclally drops traffic with ENETDOWN. That return value causes TCP to close any connection where it gets that value back from the lower parts of the stack. This means that any TCP connection with active traffic during a 3-second windown when an LACP link comes or goes would get closed. TCP treats return values of ENOBUFS as transient errors, and re-schedules transmission later. So rather than returning ENETDOWN, lets return ENOBUFS instead. This allows TCP connections to be preserved. I've tested this by repeatedly bouncing links on a Netlfix CDN server under a moderate (20Gb/s) load and overved ENOBUFS reported back to the TCP stack (as reported by a RACK TCP sysctl). Reviewed by: jhb, jtl, rrs Sponsored by: Netflix Differential Revision: https://reviews.freebsd.org/D27188
2020-11-18 14:55:49 +00:00
int err;
Implement kernel support for hardware rate limited sockets. - Add RATELIMIT kernel configuration keyword which must be set to enable the new functionality. - Add support for hardware driven, Receive Side Scaling, RSS aware, rate limited sendqueues and expose the functionality through the already established SO_MAX_PACING_RATE setsockopt(). The API support rates in the range from 1 to 4Gbytes/s which are suitable for regular TCP and UDP streams. The setsockopt(2) manual page has been updated. - Add rate limit function callback API to "struct ifnet" which supports the following operations: if_snd_tag_alloc(), if_snd_tag_modify(), if_snd_tag_query() and if_snd_tag_free(). - Add support to ifconfig to view, set and clear the IFCAP_TXRTLMT flag, which tells if a network driver supports rate limiting or not. - This patch also adds support for rate limiting through VLAN and LAGG intermediate network devices. - How rate limiting works: 1) The userspace application calls setsockopt() after accepting or making a new connection to set the rate which is then stored in the socket structure in the kernel. Later on when packets are transmitted a check is made in the transmit path for rate changes. A rate change implies a non-blocking ifp->if_snd_tag_alloc() call will be made to the destination network interface, which then sets up a custom sendqueue with the given rate limitation parameter. A "struct m_snd_tag" pointer is returned which serves as a "snd_tag" hint in the m_pkthdr for the subsequently transmitted mbufs. 2) When the network driver sees the "m->m_pkthdr.snd_tag" different from NULL, it will move the packets into a designated rate limited sendqueue given by the snd_tag pointer. It is up to the individual drivers how the rate limited traffic will be rate limited. 3) Route changes are detected by the NIC drivers in the ifp->if_transmit() routine when the ifnet pointer in the incoming snd_tag mismatches the one of the network interface. The network adapter frees the mbuf and returns EAGAIN which causes the ip_output() to release and clear the send tag. Upon next ip_output() a new "snd_tag" will be tried allocated. 4) When the PCB is detached the custom sendqueue will be released by a non-blocking ifp->if_snd_tag_free() call to the currently bound network interface. Reviewed by: wblock (manpages), adrian, gallatin, scottl (network) Differential Revision: https://reviews.freebsd.org/D3687 Sponsored by: Mellanox Technologies MFC after: 3 months
2017-01-18 13:31:17 +00:00
Restructure mbuf send tags to provide stronger guarantees. - Perform ifp mismatch checks (to determine if a send tag is allocated for a different ifp than the one the packet is being output on), in ip_output() and ip6_output(). This avoids sending packets with send tags to ifnet drivers that don't support send tags. Since we are now checking for ifp mismatches before invoking if_output, we can now try to allocate a new tag before invoking if_output sending the original packet on the new tag if allocation succeeds. To avoid code duplication for the fragment and unfragmented cases, add ip_output_send() and ip6_output_send() as wrappers around if_output and nd6_output_ifp, respectively. All of the logic for setting send tags and dealing with send tag-related errors is done in these wrapper functions. For pseudo interfaces that wrap other network interfaces (vlan and lagg), wrapper send tags are now allocated so that ip*_output see the wrapper ifp as the ifp in the send tag. The if_transmit routines rewrite the send tags after performing an ifp mismatch check. If an ifp mismatch is detected, the transmit routines fail with EAGAIN. - To provide clearer life cycle management of send tags, especially in the presence of vlan and lagg wrapper tags, add a reference count to send tags managed via m_snd_tag_ref() and m_snd_tag_rele(). Provide a helper function (m_snd_tag_init()) for use by drivers supporting send tags. m_snd_tag_init() takes care of the if_ref on the ifp meaning that code alloating send tags via if_snd_tag_alloc no longer has to manage that manually. Similarly, m_snd_tag_rele drops the refcount on the ifp after invoking if_snd_tag_free when the last reference to a send tag is dropped. This also closes use after free races if there are pending packets in driver tx rings after the socket is closed (e.g. from tcpdrop). In order for m_free to work reliably, add a new CSUM_SND_TAG flag in csum_flags to indicate 'snd_tag' is set (rather than 'rcvif'). Drivers now also check this flag instead of checking snd_tag against NULL. This avoids false positive matches when a forwarded packet has a non-NULL rcvif that was treated as a send tag. - cxgbe was relying on snd_tag_free being called when the inp was detached so that it could kick the firmware to flush any pending work on the flow. This is because the driver doesn't require ACK messages from the firmware for every request, but instead does a kind of manual interrupt coalescing by only setting a flag to request a completion on a subset of requests. If all of the in-flight requests don't have the flag when the tag is detached from the inp, the flow might never return the credits. The current snd_tag_free command issues a flush command to force the credits to return. However, the credit return is what also frees the mbufs, and since those mbufs now hold references on the tag, this meant that snd_tag_free would never be called. To fix, explicitly drop the mbuf's reference on the snd tag when the mbuf is queued in the firmware work queue. This means that once the inp's reference on the tag goes away and all in-flight mbufs have been queued to the firmware, tag's refcount will drop to zero and snd_tag_free will kick in and send the flush request. Note that we need to avoid doing this in the middle of ethofld_tx(), so the driver grabs a temporary reference on the tag around that loop to defer the free to the end of the function in case it sends the last mbuf to the queue after the inp has dropped its reference on the tag. - mlx5 preallocates send tags and was using the ifp pointer even when the send tag wasn't in use. Explicitly use the ifp from other data structures instead. - Sprinkle some assertions in various places to assert that received packets don't have a send tag, and that other places that overwrite rcvif (e.g. 802.11 transmit) don't clobber a send tag pointer. Reviewed by: gallatin, hselasky, rgrimes, ae Sponsored by: Netflix Differential Revision: https://reviews.freebsd.org/D20117
2019-05-24 22:30:40 +00:00
sc = ifp->if_softc;
Implement kernel support for hardware rate limited sockets. - Add RATELIMIT kernel configuration keyword which must be set to enable the new functionality. - Add support for hardware driven, Receive Side Scaling, RSS aware, rate limited sendqueues and expose the functionality through the already established SO_MAX_PACING_RATE setsockopt(). The API support rates in the range from 1 to 4Gbytes/s which are suitable for regular TCP and UDP streams. The setsockopt(2) manual page has been updated. - Add rate limit function callback API to "struct ifnet" which supports the following operations: if_snd_tag_alloc(), if_snd_tag_modify(), if_snd_tag_query() and if_snd_tag_free(). - Add support to ifconfig to view, set and clear the IFCAP_TXRTLMT flag, which tells if a network driver supports rate limiting or not. - This patch also adds support for rate limiting through VLAN and LAGG intermediate network devices. - How rate limiting works: 1) The userspace application calls setsockopt() after accepting or making a new connection to set the rate which is then stored in the socket structure in the kernel. Later on when packets are transmitted a check is made in the transmit path for rate changes. A rate change implies a non-blocking ifp->if_snd_tag_alloc() call will be made to the destination network interface, which then sets up a custom sendqueue with the given rate limitation parameter. A "struct m_snd_tag" pointer is returned which serves as a "snd_tag" hint in the m_pkthdr for the subsequently transmitted mbufs. 2) When the network driver sees the "m->m_pkthdr.snd_tag" different from NULL, it will move the packets into a designated rate limited sendqueue given by the snd_tag pointer. It is up to the individual drivers how the rate limited traffic will be rate limited. 3) Route changes are detected by the NIC drivers in the ifp->if_transmit() routine when the ifnet pointer in the incoming snd_tag mismatches the one of the network interface. The network adapter frees the mbuf and returns EAGAIN which causes the ip_output() to release and clear the send tag. Upon next ip_output() a new "snd_tag" will be tried allocated. 4) When the PCB is detached the custom sendqueue will be released by a non-blocking ifp->if_snd_tag_free() call to the currently bound network interface. Reviewed by: wblock (manpages), adrian, gallatin, scottl (network) Differential Revision: https://reviews.freebsd.org/D3687 Sponsored by: Mellanox Technologies MFC after: 3 months
2017-01-18 13:31:17 +00:00
switch (sc->sc_proto) {
case LAGG_PROTO_FAILOVER:
Restructure mbuf send tags to provide stronger guarantees. - Perform ifp mismatch checks (to determine if a send tag is allocated for a different ifp than the one the packet is being output on), in ip_output() and ip6_output(). This avoids sending packets with send tags to ifnet drivers that don't support send tags. Since we are now checking for ifp mismatches before invoking if_output, we can now try to allocate a new tag before invoking if_output sending the original packet on the new tag if allocation succeeds. To avoid code duplication for the fragment and unfragmented cases, add ip_output_send() and ip6_output_send() as wrappers around if_output and nd6_output_ifp, respectively. All of the logic for setting send tags and dealing with send tag-related errors is done in these wrapper functions. For pseudo interfaces that wrap other network interfaces (vlan and lagg), wrapper send tags are now allocated so that ip*_output see the wrapper ifp as the ifp in the send tag. The if_transmit routines rewrite the send tags after performing an ifp mismatch check. If an ifp mismatch is detected, the transmit routines fail with EAGAIN. - To provide clearer life cycle management of send tags, especially in the presence of vlan and lagg wrapper tags, add a reference count to send tags managed via m_snd_tag_ref() and m_snd_tag_rele(). Provide a helper function (m_snd_tag_init()) for use by drivers supporting send tags. m_snd_tag_init() takes care of the if_ref on the ifp meaning that code alloating send tags via if_snd_tag_alloc no longer has to manage that manually. Similarly, m_snd_tag_rele drops the refcount on the ifp after invoking if_snd_tag_free when the last reference to a send tag is dropped. This also closes use after free races if there are pending packets in driver tx rings after the socket is closed (e.g. from tcpdrop). In order for m_free to work reliably, add a new CSUM_SND_TAG flag in csum_flags to indicate 'snd_tag' is set (rather than 'rcvif'). Drivers now also check this flag instead of checking snd_tag against NULL. This avoids false positive matches when a forwarded packet has a non-NULL rcvif that was treated as a send tag. - cxgbe was relying on snd_tag_free being called when the inp was detached so that it could kick the firmware to flush any pending work on the flow. This is because the driver doesn't require ACK messages from the firmware for every request, but instead does a kind of manual interrupt coalescing by only setting a flag to request a completion on a subset of requests. If all of the in-flight requests don't have the flag when the tag is detached from the inp, the flow might never return the credits. The current snd_tag_free command issues a flush command to force the credits to return. However, the credit return is what also frees the mbufs, and since those mbufs now hold references on the tag, this meant that snd_tag_free would never be called. To fix, explicitly drop the mbuf's reference on the snd tag when the mbuf is queued in the firmware work queue. This means that once the inp's reference on the tag goes away and all in-flight mbufs have been queued to the firmware, tag's refcount will drop to zero and snd_tag_free will kick in and send the flush request. Note that we need to avoid doing this in the middle of ethofld_tx(), so the driver grabs a temporary reference on the tag around that loop to defer the free to the end of the function in case it sends the last mbuf to the queue after the inp has dropped its reference on the tag. - mlx5 preallocates send tags and was using the ifp pointer even when the send tag wasn't in use. Explicitly use the ifp from other data structures instead. - Sprinkle some assertions in various places to assert that received packets don't have a send tag, and that other places that overwrite rcvif (e.g. 802.11 transmit) don't clobber a send tag pointer. Reviewed by: gallatin, hselasky, rgrimes, ae Sponsored by: Netflix Differential Revision: https://reviews.freebsd.org/D20117
2019-05-24 22:30:40 +00:00
return (lagg_link_active(sc, sc->sc_primary));
Implement kernel support for hardware rate limited sockets. - Add RATELIMIT kernel configuration keyword which must be set to enable the new functionality. - Add support for hardware driven, Receive Side Scaling, RSS aware, rate limited sendqueues and expose the functionality through the already established SO_MAX_PACING_RATE setsockopt(). The API support rates in the range from 1 to 4Gbytes/s which are suitable for regular TCP and UDP streams. The setsockopt(2) manual page has been updated. - Add rate limit function callback API to "struct ifnet" which supports the following operations: if_snd_tag_alloc(), if_snd_tag_modify(), if_snd_tag_query() and if_snd_tag_free(). - Add support to ifconfig to view, set and clear the IFCAP_TXRTLMT flag, which tells if a network driver supports rate limiting or not. - This patch also adds support for rate limiting through VLAN and LAGG intermediate network devices. - How rate limiting works: 1) The userspace application calls setsockopt() after accepting or making a new connection to set the rate which is then stored in the socket structure in the kernel. Later on when packets are transmitted a check is made in the transmit path for rate changes. A rate change implies a non-blocking ifp->if_snd_tag_alloc() call will be made to the destination network interface, which then sets up a custom sendqueue with the given rate limitation parameter. A "struct m_snd_tag" pointer is returned which serves as a "snd_tag" hint in the m_pkthdr for the subsequently transmitted mbufs. 2) When the network driver sees the "m->m_pkthdr.snd_tag" different from NULL, it will move the packets into a designated rate limited sendqueue given by the snd_tag pointer. It is up to the individual drivers how the rate limited traffic will be rate limited. 3) Route changes are detected by the NIC drivers in the ifp->if_transmit() routine when the ifnet pointer in the incoming snd_tag mismatches the one of the network interface. The network adapter frees the mbuf and returns EAGAIN which causes the ip_output() to release and clear the send tag. Upon next ip_output() a new "snd_tag" will be tried allocated. 4) When the PCB is detached the custom sendqueue will be released by a non-blocking ifp->if_snd_tag_free() call to the currently bound network interface. Reviewed by: wblock (manpages), adrian, gallatin, scottl (network) Differential Revision: https://reviews.freebsd.org/D3687 Sponsored by: Mellanox Technologies MFC after: 3 months
2017-01-18 13:31:17 +00:00
case LAGG_PROTO_LOADBALANCE:
if ((sc->sc_opts & LAGG_OPT_USE_FLOWID) == 0 ||
Restructure mbuf send tags to provide stronger guarantees. - Perform ifp mismatch checks (to determine if a send tag is allocated for a different ifp than the one the packet is being output on), in ip_output() and ip6_output(). This avoids sending packets with send tags to ifnet drivers that don't support send tags. Since we are now checking for ifp mismatches before invoking if_output, we can now try to allocate a new tag before invoking if_output sending the original packet on the new tag if allocation succeeds. To avoid code duplication for the fragment and unfragmented cases, add ip_output_send() and ip6_output_send() as wrappers around if_output and nd6_output_ifp, respectively. All of the logic for setting send tags and dealing with send tag-related errors is done in these wrapper functions. For pseudo interfaces that wrap other network interfaces (vlan and lagg), wrapper send tags are now allocated so that ip*_output see the wrapper ifp as the ifp in the send tag. The if_transmit routines rewrite the send tags after performing an ifp mismatch check. If an ifp mismatch is detected, the transmit routines fail with EAGAIN. - To provide clearer life cycle management of send tags, especially in the presence of vlan and lagg wrapper tags, add a reference count to send tags managed via m_snd_tag_ref() and m_snd_tag_rele(). Provide a helper function (m_snd_tag_init()) for use by drivers supporting send tags. m_snd_tag_init() takes care of the if_ref on the ifp meaning that code alloating send tags via if_snd_tag_alloc no longer has to manage that manually. Similarly, m_snd_tag_rele drops the refcount on the ifp after invoking if_snd_tag_free when the last reference to a send tag is dropped. This also closes use after free races if there are pending packets in driver tx rings after the socket is closed (e.g. from tcpdrop). In order for m_free to work reliably, add a new CSUM_SND_TAG flag in csum_flags to indicate 'snd_tag' is set (rather than 'rcvif'). Drivers now also check this flag instead of checking snd_tag against NULL. This avoids false positive matches when a forwarded packet has a non-NULL rcvif that was treated as a send tag. - cxgbe was relying on snd_tag_free being called when the inp was detached so that it could kick the firmware to flush any pending work on the flow. This is because the driver doesn't require ACK messages from the firmware for every request, but instead does a kind of manual interrupt coalescing by only setting a flag to request a completion on a subset of requests. If all of the in-flight requests don't have the flag when the tag is detached from the inp, the flow might never return the credits. The current snd_tag_free command issues a flush command to force the credits to return. However, the credit return is what also frees the mbufs, and since those mbufs now hold references on the tag, this meant that snd_tag_free would never be called. To fix, explicitly drop the mbuf's reference on the snd tag when the mbuf is queued in the firmware work queue. This means that once the inp's reference on the tag goes away and all in-flight mbufs have been queued to the firmware, tag's refcount will drop to zero and snd_tag_free will kick in and send the flush request. Note that we need to avoid doing this in the middle of ethofld_tx(), so the driver grabs a temporary reference on the tag around that loop to defer the free to the end of the function in case it sends the last mbuf to the queue after the inp has dropped its reference on the tag. - mlx5 preallocates send tags and was using the ifp pointer even when the send tag wasn't in use. Explicitly use the ifp from other data structures instead. - Sprinkle some assertions in various places to assert that received packets don't have a send tag, and that other places that overwrite rcvif (e.g. 802.11 transmit) don't clobber a send tag pointer. Reviewed by: gallatin, hselasky, rgrimes, ae Sponsored by: Netflix Differential Revision: https://reviews.freebsd.org/D20117
2019-05-24 22:30:40 +00:00
flowtype == M_HASHTYPE_NONE)
return (NULL);
p = flowid >> sc->flowid_shift;
Implement kernel support for hardware rate limited sockets. - Add RATELIMIT kernel configuration keyword which must be set to enable the new functionality. - Add support for hardware driven, Receive Side Scaling, RSS aware, rate limited sendqueues and expose the functionality through the already established SO_MAX_PACING_RATE setsockopt(). The API support rates in the range from 1 to 4Gbytes/s which are suitable for regular TCP and UDP streams. The setsockopt(2) manual page has been updated. - Add rate limit function callback API to "struct ifnet" which supports the following operations: if_snd_tag_alloc(), if_snd_tag_modify(), if_snd_tag_query() and if_snd_tag_free(). - Add support to ifconfig to view, set and clear the IFCAP_TXRTLMT flag, which tells if a network driver supports rate limiting or not. - This patch also adds support for rate limiting through VLAN and LAGG intermediate network devices. - How rate limiting works: 1) The userspace application calls setsockopt() after accepting or making a new connection to set the rate which is then stored in the socket structure in the kernel. Later on when packets are transmitted a check is made in the transmit path for rate changes. A rate change implies a non-blocking ifp->if_snd_tag_alloc() call will be made to the destination network interface, which then sets up a custom sendqueue with the given rate limitation parameter. A "struct m_snd_tag" pointer is returned which serves as a "snd_tag" hint in the m_pkthdr for the subsequently transmitted mbufs. 2) When the network driver sees the "m->m_pkthdr.snd_tag" different from NULL, it will move the packets into a designated rate limited sendqueue given by the snd_tag pointer. It is up to the individual drivers how the rate limited traffic will be rate limited. 3) Route changes are detected by the NIC drivers in the ifp->if_transmit() routine when the ifnet pointer in the incoming snd_tag mismatches the one of the network interface. The network adapter frees the mbuf and returns EAGAIN which causes the ip_output() to release and clear the send tag. Upon next ip_output() a new "snd_tag" will be tried allocated. 4) When the PCB is detached the custom sendqueue will be released by a non-blocking ifp->if_snd_tag_free() call to the currently bound network interface. Reviewed by: wblock (manpages), adrian, gallatin, scottl (network) Differential Revision: https://reviews.freebsd.org/D3687 Sponsored by: Mellanox Technologies MFC after: 3 months
2017-01-18 13:31:17 +00:00
p %= sc->sc_count;
lb = (struct lagg_lb *)sc->sc_psc;
lp = lb->lb_ports[p];
Restructure mbuf send tags to provide stronger guarantees. - Perform ifp mismatch checks (to determine if a send tag is allocated for a different ifp than the one the packet is being output on), in ip_output() and ip6_output(). This avoids sending packets with send tags to ifnet drivers that don't support send tags. Since we are now checking for ifp mismatches before invoking if_output, we can now try to allocate a new tag before invoking if_output sending the original packet on the new tag if allocation succeeds. To avoid code duplication for the fragment and unfragmented cases, add ip_output_send() and ip6_output_send() as wrappers around if_output and nd6_output_ifp, respectively. All of the logic for setting send tags and dealing with send tag-related errors is done in these wrapper functions. For pseudo interfaces that wrap other network interfaces (vlan and lagg), wrapper send tags are now allocated so that ip*_output see the wrapper ifp as the ifp in the send tag. The if_transmit routines rewrite the send tags after performing an ifp mismatch check. If an ifp mismatch is detected, the transmit routines fail with EAGAIN. - To provide clearer life cycle management of send tags, especially in the presence of vlan and lagg wrapper tags, add a reference count to send tags managed via m_snd_tag_ref() and m_snd_tag_rele(). Provide a helper function (m_snd_tag_init()) for use by drivers supporting send tags. m_snd_tag_init() takes care of the if_ref on the ifp meaning that code alloating send tags via if_snd_tag_alloc no longer has to manage that manually. Similarly, m_snd_tag_rele drops the refcount on the ifp after invoking if_snd_tag_free when the last reference to a send tag is dropped. This also closes use after free races if there are pending packets in driver tx rings after the socket is closed (e.g. from tcpdrop). In order for m_free to work reliably, add a new CSUM_SND_TAG flag in csum_flags to indicate 'snd_tag' is set (rather than 'rcvif'). Drivers now also check this flag instead of checking snd_tag against NULL. This avoids false positive matches when a forwarded packet has a non-NULL rcvif that was treated as a send tag. - cxgbe was relying on snd_tag_free being called when the inp was detached so that it could kick the firmware to flush any pending work on the flow. This is because the driver doesn't require ACK messages from the firmware for every request, but instead does a kind of manual interrupt coalescing by only setting a flag to request a completion on a subset of requests. If all of the in-flight requests don't have the flag when the tag is detached from the inp, the flow might never return the credits. The current snd_tag_free command issues a flush command to force the credits to return. However, the credit return is what also frees the mbufs, and since those mbufs now hold references on the tag, this meant that snd_tag_free would never be called. To fix, explicitly drop the mbuf's reference on the snd tag when the mbuf is queued in the firmware work queue. This means that once the inp's reference on the tag goes away and all in-flight mbufs have been queued to the firmware, tag's refcount will drop to zero and snd_tag_free will kick in and send the flush request. Note that we need to avoid doing this in the middle of ethofld_tx(), so the driver grabs a temporary reference on the tag around that loop to defer the free to the end of the function in case it sends the last mbuf to the queue after the inp has dropped its reference on the tag. - mlx5 preallocates send tags and was using the ifp pointer even when the send tag wasn't in use. Explicitly use the ifp from other data structures instead. - Sprinkle some assertions in various places to assert that received packets don't have a send tag, and that other places that overwrite rcvif (e.g. 802.11 transmit) don't clobber a send tag pointer. Reviewed by: gallatin, hselasky, rgrimes, ae Sponsored by: Netflix Differential Revision: https://reviews.freebsd.org/D20117
2019-05-24 22:30:40 +00:00
return (lagg_link_active(sc, lp));
Implement kernel support for hardware rate limited sockets. - Add RATELIMIT kernel configuration keyword which must be set to enable the new functionality. - Add support for hardware driven, Receive Side Scaling, RSS aware, rate limited sendqueues and expose the functionality through the already established SO_MAX_PACING_RATE setsockopt(). The API support rates in the range from 1 to 4Gbytes/s which are suitable for regular TCP and UDP streams. The setsockopt(2) manual page has been updated. - Add rate limit function callback API to "struct ifnet" which supports the following operations: if_snd_tag_alloc(), if_snd_tag_modify(), if_snd_tag_query() and if_snd_tag_free(). - Add support to ifconfig to view, set and clear the IFCAP_TXRTLMT flag, which tells if a network driver supports rate limiting or not. - This patch also adds support for rate limiting through VLAN and LAGG intermediate network devices. - How rate limiting works: 1) The userspace application calls setsockopt() after accepting or making a new connection to set the rate which is then stored in the socket structure in the kernel. Later on when packets are transmitted a check is made in the transmit path for rate changes. A rate change implies a non-blocking ifp->if_snd_tag_alloc() call will be made to the destination network interface, which then sets up a custom sendqueue with the given rate limitation parameter. A "struct m_snd_tag" pointer is returned which serves as a "snd_tag" hint in the m_pkthdr for the subsequently transmitted mbufs. 2) When the network driver sees the "m->m_pkthdr.snd_tag" different from NULL, it will move the packets into a designated rate limited sendqueue given by the snd_tag pointer. It is up to the individual drivers how the rate limited traffic will be rate limited. 3) Route changes are detected by the NIC drivers in the ifp->if_transmit() routine when the ifnet pointer in the incoming snd_tag mismatches the one of the network interface. The network adapter frees the mbuf and returns EAGAIN which causes the ip_output() to release and clear the send tag. Upon next ip_output() a new "snd_tag" will be tried allocated. 4) When the PCB is detached the custom sendqueue will be released by a non-blocking ifp->if_snd_tag_free() call to the currently bound network interface. Reviewed by: wblock (manpages), adrian, gallatin, scottl (network) Differential Revision: https://reviews.freebsd.org/D3687 Sponsored by: Mellanox Technologies MFC after: 3 months
2017-01-18 13:31:17 +00:00
case LAGG_PROTO_LACP:
if ((sc->sc_opts & LAGG_OPT_USE_FLOWID) == 0 ||
Restructure mbuf send tags to provide stronger guarantees. - Perform ifp mismatch checks (to determine if a send tag is allocated for a different ifp than the one the packet is being output on), in ip_output() and ip6_output(). This avoids sending packets with send tags to ifnet drivers that don't support send tags. Since we are now checking for ifp mismatches before invoking if_output, we can now try to allocate a new tag before invoking if_output sending the original packet on the new tag if allocation succeeds. To avoid code duplication for the fragment and unfragmented cases, add ip_output_send() and ip6_output_send() as wrappers around if_output and nd6_output_ifp, respectively. All of the logic for setting send tags and dealing with send tag-related errors is done in these wrapper functions. For pseudo interfaces that wrap other network interfaces (vlan and lagg), wrapper send tags are now allocated so that ip*_output see the wrapper ifp as the ifp in the send tag. The if_transmit routines rewrite the send tags after performing an ifp mismatch check. If an ifp mismatch is detected, the transmit routines fail with EAGAIN. - To provide clearer life cycle management of send tags, especially in the presence of vlan and lagg wrapper tags, add a reference count to send tags managed via m_snd_tag_ref() and m_snd_tag_rele(). Provide a helper function (m_snd_tag_init()) for use by drivers supporting send tags. m_snd_tag_init() takes care of the if_ref on the ifp meaning that code alloating send tags via if_snd_tag_alloc no longer has to manage that manually. Similarly, m_snd_tag_rele drops the refcount on the ifp after invoking if_snd_tag_free when the last reference to a send tag is dropped. This also closes use after free races if there are pending packets in driver tx rings after the socket is closed (e.g. from tcpdrop). In order for m_free to work reliably, add a new CSUM_SND_TAG flag in csum_flags to indicate 'snd_tag' is set (rather than 'rcvif'). Drivers now also check this flag instead of checking snd_tag against NULL. This avoids false positive matches when a forwarded packet has a non-NULL rcvif that was treated as a send tag. - cxgbe was relying on snd_tag_free being called when the inp was detached so that it could kick the firmware to flush any pending work on the flow. This is because the driver doesn't require ACK messages from the firmware for every request, but instead does a kind of manual interrupt coalescing by only setting a flag to request a completion on a subset of requests. If all of the in-flight requests don't have the flag when the tag is detached from the inp, the flow might never return the credits. The current snd_tag_free command issues a flush command to force the credits to return. However, the credit return is what also frees the mbufs, and since those mbufs now hold references on the tag, this meant that snd_tag_free would never be called. To fix, explicitly drop the mbuf's reference on the snd tag when the mbuf is queued in the firmware work queue. This means that once the inp's reference on the tag goes away and all in-flight mbufs have been queued to the firmware, tag's refcount will drop to zero and snd_tag_free will kick in and send the flush request. Note that we need to avoid doing this in the middle of ethofld_tx(), so the driver grabs a temporary reference on the tag around that loop to defer the free to the end of the function in case it sends the last mbuf to the queue after the inp has dropped its reference on the tag. - mlx5 preallocates send tags and was using the ifp pointer even when the send tag wasn't in use. Explicitly use the ifp from other data structures instead. - Sprinkle some assertions in various places to assert that received packets don't have a send tag, and that other places that overwrite rcvif (e.g. 802.11 transmit) don't clobber a send tag pointer. Reviewed by: gallatin, hselasky, rgrimes, ae Sponsored by: Netflix Differential Revision: https://reviews.freebsd.org/D20117
2019-05-24 22:30:40 +00:00
flowtype == M_HASHTYPE_NONE)
return (NULL);
make lacp's use_numa hashing aware of send tags When I did the use_numa support, I missed the fact that there is a separate hash function for send tag nic selection. So when use_numa is enabled, ktls offload does not work properly, as it does not reliably allocate a send tag on the proper egress nic since different egress nics are selected for send-tag allocation and packet transmit. To fix this, this change: - refectors lacp_select_tx_port_by_hash() and lacp_select_tx_port() to make lacp_select_tx_port_by_hash() always called by lacp_select_tx_port() - pre-shifts flowids to convert them to hashes when calling lacp_select_tx_port_by_hash() - adds a numa_domain field to if_snd_tag_alloc_params - plumbs the numa domain into places where we allocate send tags In testing with NIC TLS setup on a NUMA machine, I see thousands of output errors before the change when enabling kern.ipc.tls.ifnet.permitted=1. After the change, I see no errors, and I see the NIC sysctl counters showing active TLS offload sessions. Reviewed by: rrs, hselasky, jhb Sponsored by: Netflix
2020-03-09 13:44:51 +00:00
hash = flowid >> sc->flowid_shift;
LACP: When suppressing distributing, return ENOBUFS When links come and go, lacp goes into a "suppress distributing" mode where it drops traffic for 3 seconds. When in this mode, lagg/lacp historiclally drops traffic with ENETDOWN. That return value causes TCP to close any connection where it gets that value back from the lower parts of the stack. This means that any TCP connection with active traffic during a 3-second windown when an LACP link comes or goes would get closed. TCP treats return values of ENOBUFS as transient errors, and re-schedules transmission later. So rather than returning ENETDOWN, lets return ENOBUFS instead. This allows TCP connections to be preserved. I've tested this by repeatedly bouncing links on a Netlfix CDN server under a moderate (20Gb/s) load and overved ENOBUFS reported back to the TCP stack (as reported by a RACK TCP sysctl). Reviewed by: jhb, jtl, rrs Sponsored by: Netflix Differential Revision: https://reviews.freebsd.org/D27188
2020-11-18 14:55:49 +00:00
return (lacp_select_tx_port_by_hash(sc, hash, numa_domain, &err));
Implement kernel support for hardware rate limited sockets. - Add RATELIMIT kernel configuration keyword which must be set to enable the new functionality. - Add support for hardware driven, Receive Side Scaling, RSS aware, rate limited sendqueues and expose the functionality through the already established SO_MAX_PACING_RATE setsockopt(). The API support rates in the range from 1 to 4Gbytes/s which are suitable for regular TCP and UDP streams. The setsockopt(2) manual page has been updated. - Add rate limit function callback API to "struct ifnet" which supports the following operations: if_snd_tag_alloc(), if_snd_tag_modify(), if_snd_tag_query() and if_snd_tag_free(). - Add support to ifconfig to view, set and clear the IFCAP_TXRTLMT flag, which tells if a network driver supports rate limiting or not. - This patch also adds support for rate limiting through VLAN and LAGG intermediate network devices. - How rate limiting works: 1) The userspace application calls setsockopt() after accepting or making a new connection to set the rate which is then stored in the socket structure in the kernel. Later on when packets are transmitted a check is made in the transmit path for rate changes. A rate change implies a non-blocking ifp->if_snd_tag_alloc() call will be made to the destination network interface, which then sets up a custom sendqueue with the given rate limitation parameter. A "struct m_snd_tag" pointer is returned which serves as a "snd_tag" hint in the m_pkthdr for the subsequently transmitted mbufs. 2) When the network driver sees the "m->m_pkthdr.snd_tag" different from NULL, it will move the packets into a designated rate limited sendqueue given by the snd_tag pointer. It is up to the individual drivers how the rate limited traffic will be rate limited. 3) Route changes are detected by the NIC drivers in the ifp->if_transmit() routine when the ifnet pointer in the incoming snd_tag mismatches the one of the network interface. The network adapter frees the mbuf and returns EAGAIN which causes the ip_output() to release and clear the send tag. Upon next ip_output() a new "snd_tag" will be tried allocated. 4) When the PCB is detached the custom sendqueue will be released by a non-blocking ifp->if_snd_tag_free() call to the currently bound network interface. Reviewed by: wblock (manpages), adrian, gallatin, scottl (network) Differential Revision: https://reviews.freebsd.org/D3687 Sponsored by: Mellanox Technologies MFC after: 3 months
2017-01-18 13:31:17 +00:00
default:
Restructure mbuf send tags to provide stronger guarantees. - Perform ifp mismatch checks (to determine if a send tag is allocated for a different ifp than the one the packet is being output on), in ip_output() and ip6_output(). This avoids sending packets with send tags to ifnet drivers that don't support send tags. Since we are now checking for ifp mismatches before invoking if_output, we can now try to allocate a new tag before invoking if_output sending the original packet on the new tag if allocation succeeds. To avoid code duplication for the fragment and unfragmented cases, add ip_output_send() and ip6_output_send() as wrappers around if_output and nd6_output_ifp, respectively. All of the logic for setting send tags and dealing with send tag-related errors is done in these wrapper functions. For pseudo interfaces that wrap other network interfaces (vlan and lagg), wrapper send tags are now allocated so that ip*_output see the wrapper ifp as the ifp in the send tag. The if_transmit routines rewrite the send tags after performing an ifp mismatch check. If an ifp mismatch is detected, the transmit routines fail with EAGAIN. - To provide clearer life cycle management of send tags, especially in the presence of vlan and lagg wrapper tags, add a reference count to send tags managed via m_snd_tag_ref() and m_snd_tag_rele(). Provide a helper function (m_snd_tag_init()) for use by drivers supporting send tags. m_snd_tag_init() takes care of the if_ref on the ifp meaning that code alloating send tags via if_snd_tag_alloc no longer has to manage that manually. Similarly, m_snd_tag_rele drops the refcount on the ifp after invoking if_snd_tag_free when the last reference to a send tag is dropped. This also closes use after free races if there are pending packets in driver tx rings after the socket is closed (e.g. from tcpdrop). In order for m_free to work reliably, add a new CSUM_SND_TAG flag in csum_flags to indicate 'snd_tag' is set (rather than 'rcvif'). Drivers now also check this flag instead of checking snd_tag against NULL. This avoids false positive matches when a forwarded packet has a non-NULL rcvif that was treated as a send tag. - cxgbe was relying on snd_tag_free being called when the inp was detached so that it could kick the firmware to flush any pending work on the flow. This is because the driver doesn't require ACK messages from the firmware for every request, but instead does a kind of manual interrupt coalescing by only setting a flag to request a completion on a subset of requests. If all of the in-flight requests don't have the flag when the tag is detached from the inp, the flow might never return the credits. The current snd_tag_free command issues a flush command to force the credits to return. However, the credit return is what also frees the mbufs, and since those mbufs now hold references on the tag, this meant that snd_tag_free would never be called. To fix, explicitly drop the mbuf's reference on the snd tag when the mbuf is queued in the firmware work queue. This means that once the inp's reference on the tag goes away and all in-flight mbufs have been queued to the firmware, tag's refcount will drop to zero and snd_tag_free will kick in and send the flush request. Note that we need to avoid doing this in the middle of ethofld_tx(), so the driver grabs a temporary reference on the tag around that loop to defer the free to the end of the function in case it sends the last mbuf to the queue after the inp has dropped its reference on the tag. - mlx5 preallocates send tags and was using the ifp pointer even when the send tag wasn't in use. Explicitly use the ifp from other data structures instead. - Sprinkle some assertions in various places to assert that received packets don't have a send tag, and that other places that overwrite rcvif (e.g. 802.11 transmit) don't clobber a send tag pointer. Reviewed by: gallatin, hselasky, rgrimes, ae Sponsored by: Netflix Differential Revision: https://reviews.freebsd.org/D20117
2019-05-24 22:30:40 +00:00
return (NULL);
}
}
static int
lagg_snd_tag_alloc(struct ifnet *ifp,
union if_snd_tag_alloc_params *params,
struct m_snd_tag **ppmt)
{
Convert LAGG_RLOCK() to NET_EPOCH_ENTER(). No functional changes.
2020-12-08 16:36:46 +00:00
struct epoch_tracker et;
Add a switch structure for send tags. Move the type and function pointers for operations on existing send tags (modify, query, next, free) out of 'struct ifnet' and into a new 'struct if_snd_tag_sw'. A pointer to this structure is added to the generic part of send tags and is initialized by m_snd_tag_init() (which now accepts a switch structure as a new argument in place of the type). Previously, device driver ifnet methods switched on the type to call type-specific functions. Now, those type-specific functions are saved in the switch structure and invoked directly. In addition, this more gracefully permits multiple implementations of the same tag within a driver. In particular, NIC TLS for future Chelsio adapters will use a different implementation than the existing NIC TLS support for T6 adapters. Reviewed by: gallatin, hselasky, kib (older version) Sponsored by: Chelsio Communications Differential Revision: https://reviews.freebsd.org/D31572
2021-09-14 18:43:41 +00:00
const struct if_snd_tag_sw *sw;
Restructure mbuf send tags to provide stronger guarantees. - Perform ifp mismatch checks (to determine if a send tag is allocated for a different ifp than the one the packet is being output on), in ip_output() and ip6_output(). This avoids sending packets with send tags to ifnet drivers that don't support send tags. Since we are now checking for ifp mismatches before invoking if_output, we can now try to allocate a new tag before invoking if_output sending the original packet on the new tag if allocation succeeds. To avoid code duplication for the fragment and unfragmented cases, add ip_output_send() and ip6_output_send() as wrappers around if_output and nd6_output_ifp, respectively. All of the logic for setting send tags and dealing with send tag-related errors is done in these wrapper functions. For pseudo interfaces that wrap other network interfaces (vlan and lagg), wrapper send tags are now allocated so that ip*_output see the wrapper ifp as the ifp in the send tag. The if_transmit routines rewrite the send tags after performing an ifp mismatch check. If an ifp mismatch is detected, the transmit routines fail with EAGAIN. - To provide clearer life cycle management of send tags, especially in the presence of vlan and lagg wrapper tags, add a reference count to send tags managed via m_snd_tag_ref() and m_snd_tag_rele(). Provide a helper function (m_snd_tag_init()) for use by drivers supporting send tags. m_snd_tag_init() takes care of the if_ref on the ifp meaning that code alloating send tags via if_snd_tag_alloc no longer has to manage that manually. Similarly, m_snd_tag_rele drops the refcount on the ifp after invoking if_snd_tag_free when the last reference to a send tag is dropped. This also closes use after free races if there are pending packets in driver tx rings after the socket is closed (e.g. from tcpdrop). In order for m_free to work reliably, add a new CSUM_SND_TAG flag in csum_flags to indicate 'snd_tag' is set (rather than 'rcvif'). Drivers now also check this flag instead of checking snd_tag against NULL. This avoids false positive matches when a forwarded packet has a non-NULL rcvif that was treated as a send tag. - cxgbe was relying on snd_tag_free being called when the inp was detached so that it could kick the firmware to flush any pending work on the flow. This is because the driver doesn't require ACK messages from the firmware for every request, but instead does a kind of manual interrupt coalescing by only setting a flag to request a completion on a subset of requests. If all of the in-flight requests don't have the flag when the tag is detached from the inp, the flow might never return the credits. The current snd_tag_free command issues a flush command to force the credits to return. However, the credit return is what also frees the mbufs, and since those mbufs now hold references on the tag, this meant that snd_tag_free would never be called. To fix, explicitly drop the mbuf's reference on the snd tag when the mbuf is queued in the firmware work queue. This means that once the inp's reference on the tag goes away and all in-flight mbufs have been queued to the firmware, tag's refcount will drop to zero and snd_tag_free will kick in and send the flush request. Note that we need to avoid doing this in the middle of ethofld_tx(), so the driver grabs a temporary reference on the tag around that loop to defer the free to the end of the function in case it sends the last mbuf to the queue after the inp has dropped its reference on the tag. - mlx5 preallocates send tags and was using the ifp pointer even when the send tag wasn't in use. Explicitly use the ifp from other data structures instead. - Sprinkle some assertions in various places to assert that received packets don't have a send tag, and that other places that overwrite rcvif (e.g. 802.11 transmit) don't clobber a send tag pointer. Reviewed by: gallatin, hselasky, rgrimes, ae Sponsored by: Netflix Differential Revision: https://reviews.freebsd.org/D20117
2019-05-24 22:30:40 +00:00
struct lagg_snd_tag *lst;
struct lagg_port *lp;
struct ifnet *lp_ifp;
lagg(4): Add support for allocating TLS receive tags. The TLS receive tags are allocated directly from the receiving interface, because mbufs are flowing in the opposite direction and then route change checks are not useful, because they only work for outgoing traffic. Differential revision: https://reviews.freebsd.org/D32356 Sponsored by: NVIDIA Networking
2022-05-25 10:38:30 +00:00
struct m_snd_tag *mst;
Restructure mbuf send tags to provide stronger guarantees. - Perform ifp mismatch checks (to determine if a send tag is allocated for a different ifp than the one the packet is being output on), in ip_output() and ip6_output(). This avoids sending packets with send tags to ifnet drivers that don't support send tags. Since we are now checking for ifp mismatches before invoking if_output, we can now try to allocate a new tag before invoking if_output sending the original packet on the new tag if allocation succeeds. To avoid code duplication for the fragment and unfragmented cases, add ip_output_send() and ip6_output_send() as wrappers around if_output and nd6_output_ifp, respectively. All of the logic for setting send tags and dealing with send tag-related errors is done in these wrapper functions. For pseudo interfaces that wrap other network interfaces (vlan and lagg), wrapper send tags are now allocated so that ip*_output see the wrapper ifp as the ifp in the send tag. The if_transmit routines rewrite the send tags after performing an ifp mismatch check. If an ifp mismatch is detected, the transmit routines fail with EAGAIN. - To provide clearer life cycle management of send tags, especially in the presence of vlan and lagg wrapper tags, add a reference count to send tags managed via m_snd_tag_ref() and m_snd_tag_rele(). Provide a helper function (m_snd_tag_init()) for use by drivers supporting send tags. m_snd_tag_init() takes care of the if_ref on the ifp meaning that code alloating send tags via if_snd_tag_alloc no longer has to manage that manually. Similarly, m_snd_tag_rele drops the refcount on the ifp after invoking if_snd_tag_free when the last reference to a send tag is dropped. This also closes use after free races if there are pending packets in driver tx rings after the socket is closed (e.g. from tcpdrop). In order for m_free to work reliably, add a new CSUM_SND_TAG flag in csum_flags to indicate 'snd_tag' is set (rather than 'rcvif'). Drivers now also check this flag instead of checking snd_tag against NULL. This avoids false positive matches when a forwarded packet has a non-NULL rcvif that was treated as a send tag. - cxgbe was relying on snd_tag_free being called when the inp was detached so that it could kick the firmware to flush any pending work on the flow. This is because the driver doesn't require ACK messages from the firmware for every request, but instead does a kind of manual interrupt coalescing by only setting a flag to request a completion on a subset of requests. If all of the in-flight requests don't have the flag when the tag is detached from the inp, the flow might never return the credits. The current snd_tag_free command issues a flush command to force the credits to return. However, the credit return is what also frees the mbufs, and since those mbufs now hold references on the tag, this meant that snd_tag_free would never be called. To fix, explicitly drop the mbuf's reference on the snd tag when the mbuf is queued in the firmware work queue. This means that once the inp's reference on the tag goes away and all in-flight mbufs have been queued to the firmware, tag's refcount will drop to zero and snd_tag_free will kick in and send the flush request. Note that we need to avoid doing this in the middle of ethofld_tx(), so the driver grabs a temporary reference on the tag around that loop to defer the free to the end of the function in case it sends the last mbuf to the queue after the inp has dropped its reference on the tag. - mlx5 preallocates send tags and was using the ifp pointer even when the send tag wasn't in use. Explicitly use the ifp from other data structures instead. - Sprinkle some assertions in various places to assert that received packets don't have a send tag, and that other places that overwrite rcvif (e.g. 802.11 transmit) don't clobber a send tag pointer. Reviewed by: gallatin, hselasky, rgrimes, ae Sponsored by: Netflix Differential Revision: https://reviews.freebsd.org/D20117
2019-05-24 22:30:40 +00:00
int error;
Add a switch structure for send tags. Move the type and function pointers for operations on existing send tags (modify, query, next, free) out of 'struct ifnet' and into a new 'struct if_snd_tag_sw'. A pointer to this structure is added to the generic part of send tags and is initialized by m_snd_tag_init() (which now accepts a switch structure as a new argument in place of the type). Previously, device driver ifnet methods switched on the type to call type-specific functions. Now, those type-specific functions are saved in the switch structure and invoked directly. In addition, this more gracefully permits multiple implementations of the same tag within a driver. In particular, NIC TLS for future Chelsio adapters will use a different implementation than the existing NIC TLS support for T6 adapters. Reviewed by: gallatin, hselasky, kib (older version) Sponsored by: Chelsio Communications Differential Revision: https://reviews.freebsd.org/D31572
2021-09-14 18:43:41 +00:00
switch (params->hdr.type) {
#ifdef RATELIMIT
case IF_SND_TAG_TYPE_UNLIMITED:
sw = &lagg_snd_tag_ul_sw;
break;
case IF_SND_TAG_TYPE_RATE_LIMIT:
sw = &lagg_snd_tag_rl_sw;
break;
#endif
#ifdef KERN_TLS
case IF_SND_TAG_TYPE_TLS:
sw = &lagg_snd_tag_tls_sw;
break;
lagg(4): Add support for allocating TLS receive tags. The TLS receive tags are allocated directly from the receiving interface, because mbufs are flowing in the opposite direction and then route change checks are not useful, because they only work for outgoing traffic. Differential revision: https://reviews.freebsd.org/D32356 Sponsored by: NVIDIA Networking
2022-05-25 10:38:30 +00:00
case IF_SND_TAG_TYPE_TLS_RX:
/* Return tag from port interface directly. */
sw = NULL;
break;
Add a switch structure for send tags. Move the type and function pointers for operations on existing send tags (modify, query, next, free) out of 'struct ifnet' and into a new 'struct if_snd_tag_sw'. A pointer to this structure is added to the generic part of send tags and is initialized by m_snd_tag_init() (which now accepts a switch structure as a new argument in place of the type). Previously, device driver ifnet methods switched on the type to call type-specific functions. Now, those type-specific functions are saved in the switch structure and invoked directly. In addition, this more gracefully permits multiple implementations of the same tag within a driver. In particular, NIC TLS for future Chelsio adapters will use a different implementation than the existing NIC TLS support for T6 adapters. Reviewed by: gallatin, hselasky, kib (older version) Sponsored by: Chelsio Communications Differential Revision: https://reviews.freebsd.org/D31572
2021-09-14 18:43:41 +00:00
#ifdef RATELIMIT
case IF_SND_TAG_TYPE_TLS_RATE_LIMIT:
sw = &lagg_snd_tag_tls_rl_sw;
break;
#endif
#endif
default:
return (EOPNOTSUPP);
}
Convert LAGG_RLOCK() to NET_EPOCH_ENTER(). No functional changes.
2020-12-08 16:36:46 +00:00
NET_EPOCH_ENTER(et);
make lacp's use_numa hashing aware of send tags When I did the use_numa support, I missed the fact that there is a separate hash function for send tag nic selection. So when use_numa is enabled, ktls offload does not work properly, as it does not reliably allocate a send tag on the proper egress nic since different egress nics are selected for send-tag allocation and packet transmit. To fix this, this change: - refectors lacp_select_tx_port_by_hash() and lacp_select_tx_port() to make lacp_select_tx_port_by_hash() always called by lacp_select_tx_port() - pre-shifts flowids to convert them to hashes when calling lacp_select_tx_port_by_hash() - adds a numa_domain field to if_snd_tag_alloc_params - plumbs the numa domain into places where we allocate send tags In testing with NIC TLS setup on a NUMA machine, I see thousands of output errors before the change when enabling kern.ipc.tls.ifnet.permitted=1. After the change, I see no errors, and I see the NIC sysctl counters showing active TLS offload sessions. Reviewed by: rrs, hselasky, jhb Sponsored by: Netflix
2020-03-09 13:44:51 +00:00
lp = lookup_snd_tag_port(ifp, params->hdr.flowid,
params->hdr.flowtype, params->hdr.numa_domain);
Restructure mbuf send tags to provide stronger guarantees. - Perform ifp mismatch checks (to determine if a send tag is allocated for a different ifp than the one the packet is being output on), in ip_output() and ip6_output(). This avoids sending packets with send tags to ifnet drivers that don't support send tags. Since we are now checking for ifp mismatches before invoking if_output, we can now try to allocate a new tag before invoking if_output sending the original packet on the new tag if allocation succeeds. To avoid code duplication for the fragment and unfragmented cases, add ip_output_send() and ip6_output_send() as wrappers around if_output and nd6_output_ifp, respectively. All of the logic for setting send tags and dealing with send tag-related errors is done in these wrapper functions. For pseudo interfaces that wrap other network interfaces (vlan and lagg), wrapper send tags are now allocated so that ip*_output see the wrapper ifp as the ifp in the send tag. The if_transmit routines rewrite the send tags after performing an ifp mismatch check. If an ifp mismatch is detected, the transmit routines fail with EAGAIN. - To provide clearer life cycle management of send tags, especially in the presence of vlan and lagg wrapper tags, add a reference count to send tags managed via m_snd_tag_ref() and m_snd_tag_rele(). Provide a helper function (m_snd_tag_init()) for use by drivers supporting send tags. m_snd_tag_init() takes care of the if_ref on the ifp meaning that code alloating send tags via if_snd_tag_alloc no longer has to manage that manually. Similarly, m_snd_tag_rele drops the refcount on the ifp after invoking if_snd_tag_free when the last reference to a send tag is dropped. This also closes use after free races if there are pending packets in driver tx rings after the socket is closed (e.g. from tcpdrop). In order for m_free to work reliably, add a new CSUM_SND_TAG flag in csum_flags to indicate 'snd_tag' is set (rather than 'rcvif'). Drivers now also check this flag instead of checking snd_tag against NULL. This avoids false positive matches when a forwarded packet has a non-NULL rcvif that was treated as a send tag. - cxgbe was relying on snd_tag_free being called when the inp was detached so that it could kick the firmware to flush any pending work on the flow. This is because the driver doesn't require ACK messages from the firmware for every request, but instead does a kind of manual interrupt coalescing by only setting a flag to request a completion on a subset of requests. If all of the in-flight requests don't have the flag when the tag is detached from the inp, the flow might never return the credits. The current snd_tag_free command issues a flush command to force the credits to return. However, the credit return is what also frees the mbufs, and since those mbufs now hold references on the tag, this meant that snd_tag_free would never be called. To fix, explicitly drop the mbuf's reference on the snd tag when the mbuf is queued in the firmware work queue. This means that once the inp's reference on the tag goes away and all in-flight mbufs have been queued to the firmware, tag's refcount will drop to zero and snd_tag_free will kick in and send the flush request. Note that we need to avoid doing this in the middle of ethofld_tx(), so the driver grabs a temporary reference on the tag around that loop to defer the free to the end of the function in case it sends the last mbuf to the queue after the inp has dropped its reference on the tag. - mlx5 preallocates send tags and was using the ifp pointer even when the send tag wasn't in use. Explicitly use the ifp from other data structures instead. - Sprinkle some assertions in various places to assert that received packets don't have a send tag, and that other places that overwrite rcvif (e.g. 802.11 transmit) don't clobber a send tag pointer. Reviewed by: gallatin, hselasky, rgrimes, ae Sponsored by: Netflix Differential Revision: https://reviews.freebsd.org/D20117
2019-05-24 22:30:40 +00:00
if (lp == NULL) {
Fixup r368446 with KERN_TLS.
2020-12-08 23:54:09 +00:00
NET_EPOCH_EXIT(et);
Implement kernel support for hardware rate limited sockets. - Add RATELIMIT kernel configuration keyword which must be set to enable the new functionality. - Add support for hardware driven, Receive Side Scaling, RSS aware, rate limited sendqueues and expose the functionality through the already established SO_MAX_PACING_RATE setsockopt(). The API support rates in the range from 1 to 4Gbytes/s which are suitable for regular TCP and UDP streams. The setsockopt(2) manual page has been updated. - Add rate limit function callback API to "struct ifnet" which supports the following operations: if_snd_tag_alloc(), if_snd_tag_modify(), if_snd_tag_query() and if_snd_tag_free(). - Add support to ifconfig to view, set and clear the IFCAP_TXRTLMT flag, which tells if a network driver supports rate limiting or not. - This patch also adds support for rate limiting through VLAN and LAGG intermediate network devices. - How rate limiting works: 1) The userspace application calls setsockopt() after accepting or making a new connection to set the rate which is then stored in the socket structure in the kernel. Later on when packets are transmitted a check is made in the transmit path for rate changes. A rate change implies a non-blocking ifp->if_snd_tag_alloc() call will be made to the destination network interface, which then sets up a custom sendqueue with the given rate limitation parameter. A "struct m_snd_tag" pointer is returned which serves as a "snd_tag" hint in the m_pkthdr for the subsequently transmitted mbufs. 2) When the network driver sees the "m->m_pkthdr.snd_tag" different from NULL, it will move the packets into a designated rate limited sendqueue given by the snd_tag pointer. It is up to the individual drivers how the rate limited traffic will be rate limited. 3) Route changes are detected by the NIC drivers in the ifp->if_transmit() routine when the ifnet pointer in the incoming snd_tag mismatches the one of the network interface. The network adapter frees the mbuf and returns EAGAIN which causes the ip_output() to release and clear the send tag. Upon next ip_output() a new "snd_tag" will be tried allocated. 4) When the PCB is detached the custom sendqueue will be released by a non-blocking ifp->if_snd_tag_free() call to the currently bound network interface. Reviewed by: wblock (manpages), adrian, gallatin, scottl (network) Differential Revision: https://reviews.freebsd.org/D3687 Sponsored by: Mellanox Technologies MFC after: 3 months
2017-01-18 13:31:17 +00:00
return (EOPNOTSUPP);
}
Add m_snd_tag_alloc() as a wrapper around if_snd_tag_alloc(). This gives a more uniform API for send tag life cycle management. Reviewed by: gallatin, hselasky Sponsored by: Netflix Differential Revision: https://reviews.freebsd.org/D27000
2020-10-29 23:28:39 +00:00
if (lp->lp_ifp == NULL) {
Fixup r368446 with KERN_TLS.
2020-12-08 23:54:09 +00:00
NET_EPOCH_EXIT(et);
Implement kernel support for hardware rate limited sockets. - Add RATELIMIT kernel configuration keyword which must be set to enable the new functionality. - Add support for hardware driven, Receive Side Scaling, RSS aware, rate limited sendqueues and expose the functionality through the already established SO_MAX_PACING_RATE setsockopt(). The API support rates in the range from 1 to 4Gbytes/s which are suitable for regular TCP and UDP streams. The setsockopt(2) manual page has been updated. - Add rate limit function callback API to "struct ifnet" which supports the following operations: if_snd_tag_alloc(), if_snd_tag_modify(), if_snd_tag_query() and if_snd_tag_free(). - Add support to ifconfig to view, set and clear the IFCAP_TXRTLMT flag, which tells if a network driver supports rate limiting or not. - This patch also adds support for rate limiting through VLAN and LAGG intermediate network devices. - How rate limiting works: 1) The userspace application calls setsockopt() after accepting or making a new connection to set the rate which is then stored in the socket structure in the kernel. Later on when packets are transmitted a check is made in the transmit path for rate changes. A rate change implies a non-blocking ifp->if_snd_tag_alloc() call will be made to the destination network interface, which then sets up a custom sendqueue with the given rate limitation parameter. A "struct m_snd_tag" pointer is returned which serves as a "snd_tag" hint in the m_pkthdr for the subsequently transmitted mbufs. 2) When the network driver sees the "m->m_pkthdr.snd_tag" different from NULL, it will move the packets into a designated rate limited sendqueue given by the snd_tag pointer. It is up to the individual drivers how the rate limited traffic will be rate limited. 3) Route changes are detected by the NIC drivers in the ifp->if_transmit() routine when the ifnet pointer in the incoming snd_tag mismatches the one of the network interface. The network adapter frees the mbuf and returns EAGAIN which causes the ip_output() to release and clear the send tag. Upon next ip_output() a new "snd_tag" will be tried allocated. 4) When the PCB is detached the custom sendqueue will be released by a non-blocking ifp->if_snd_tag_free() call to the currently bound network interface. Reviewed by: wblock (manpages), adrian, gallatin, scottl (network) Differential Revision: https://reviews.freebsd.org/D3687 Sponsored by: Mellanox Technologies MFC after: 3 months
2017-01-18 13:31:17 +00:00
return (EOPNOTSUPP);
Remove nested epochs from lagg(4). lagg_bcast_start appeared to have a bug in that was using the last lagg port structure after exiting the epoch that was keeping that structure alive. However, upon further inspection, the epoch was already entered by the caller (lagg_transmit), so the epoch enter/exit in lagg_bcast_start was actually unnecessary. This commit generally removes uses of the net epoch via LAGG_RLOCK to protect the list of ports when the list of ports was already protected by an existing LAGG_RLOCK in a caller, or the LAGG_XLOCK. It also adds a missing epoch enter/exit in lagg_snd_tag_alloc while accessing the lagg port structures. An ifp is still accessed via an unsafe reference after the epoch is exited, but that is true in the current code and will be fixed in a future change. Reviewed by: gallatin MFC after: 1 month Sponsored by: Netflix Differential Revision: https://reviews.freebsd.org/D19718
2019-03-28 20:25:36 +00:00
}
Restructure mbuf send tags to provide stronger guarantees. - Perform ifp mismatch checks (to determine if a send tag is allocated for a different ifp than the one the packet is being output on), in ip_output() and ip6_output(). This avoids sending packets with send tags to ifnet drivers that don't support send tags. Since we are now checking for ifp mismatches before invoking if_output, we can now try to allocate a new tag before invoking if_output sending the original packet on the new tag if allocation succeeds. To avoid code duplication for the fragment and unfragmented cases, add ip_output_send() and ip6_output_send() as wrappers around if_output and nd6_output_ifp, respectively. All of the logic for setting send tags and dealing with send tag-related errors is done in these wrapper functions. For pseudo interfaces that wrap other network interfaces (vlan and lagg), wrapper send tags are now allocated so that ip*_output see the wrapper ifp as the ifp in the send tag. The if_transmit routines rewrite the send tags after performing an ifp mismatch check. If an ifp mismatch is detected, the transmit routines fail with EAGAIN. - To provide clearer life cycle management of send tags, especially in the presence of vlan and lagg wrapper tags, add a reference count to send tags managed via m_snd_tag_ref() and m_snd_tag_rele(). Provide a helper function (m_snd_tag_init()) for use by drivers supporting send tags. m_snd_tag_init() takes care of the if_ref on the ifp meaning that code alloating send tags via if_snd_tag_alloc no longer has to manage that manually. Similarly, m_snd_tag_rele drops the refcount on the ifp after invoking if_snd_tag_free when the last reference to a send tag is dropped. This also closes use after free races if there are pending packets in driver tx rings after the socket is closed (e.g. from tcpdrop). In order for m_free to work reliably, add a new CSUM_SND_TAG flag in csum_flags to indicate 'snd_tag' is set (rather than 'rcvif'). Drivers now also check this flag instead of checking snd_tag against NULL. This avoids false positive matches when a forwarded packet has a non-NULL rcvif that was treated as a send tag. - cxgbe was relying on snd_tag_free being called when the inp was detached so that it could kick the firmware to flush any pending work on the flow. This is because the driver doesn't require ACK messages from the firmware for every request, but instead does a kind of manual interrupt coalescing by only setting a flag to request a completion on a subset of requests. If all of the in-flight requests don't have the flag when the tag is detached from the inp, the flow might never return the credits. The current snd_tag_free command issues a flush command to force the credits to return. However, the credit return is what also frees the mbufs, and since those mbufs now hold references on the tag, this meant that snd_tag_free would never be called. To fix, explicitly drop the mbuf's reference on the snd tag when the mbuf is queued in the firmware work queue. This means that once the inp's reference on the tag goes away and all in-flight mbufs have been queued to the firmware, tag's refcount will drop to zero and snd_tag_free will kick in and send the flush request. Note that we need to avoid doing this in the middle of ethofld_tx(), so the driver grabs a temporary reference on the tag around that loop to defer the free to the end of the function in case it sends the last mbuf to the queue after the inp has dropped its reference on the tag. - mlx5 preallocates send tags and was using the ifp pointer even when the send tag wasn't in use. Explicitly use the ifp from other data structures instead. - Sprinkle some assertions in various places to assert that received packets don't have a send tag, and that other places that overwrite rcvif (e.g. 802.11 transmit) don't clobber a send tag pointer. Reviewed by: gallatin, hselasky, rgrimes, ae Sponsored by: Netflix Differential Revision: https://reviews.freebsd.org/D20117
2019-05-24 22:30:40 +00:00
lp_ifp = lp->lp_ifp;
if_ref(lp_ifp);
Convert LAGG_RLOCK() to NET_EPOCH_ENTER(). No functional changes.
2020-12-08 16:36:46 +00:00
NET_EPOCH_EXIT(et);
Implement kernel support for hardware rate limited sockets. - Add RATELIMIT kernel configuration keyword which must be set to enable the new functionality. - Add support for hardware driven, Receive Side Scaling, RSS aware, rate limited sendqueues and expose the functionality through the already established SO_MAX_PACING_RATE setsockopt(). The API support rates in the range from 1 to 4Gbytes/s which are suitable for regular TCP and UDP streams. The setsockopt(2) manual page has been updated. - Add rate limit function callback API to "struct ifnet" which supports the following operations: if_snd_tag_alloc(), if_snd_tag_modify(), if_snd_tag_query() and if_snd_tag_free(). - Add support to ifconfig to view, set and clear the IFCAP_TXRTLMT flag, which tells if a network driver supports rate limiting or not. - This patch also adds support for rate limiting through VLAN and LAGG intermediate network devices. - How rate limiting works: 1) The userspace application calls setsockopt() after accepting or making a new connection to set the rate which is then stored in the socket structure in the kernel. Later on when packets are transmitted a check is made in the transmit path for rate changes. A rate change implies a non-blocking ifp->if_snd_tag_alloc() call will be made to the destination network interface, which then sets up a custom sendqueue with the given rate limitation parameter. A "struct m_snd_tag" pointer is returned which serves as a "snd_tag" hint in the m_pkthdr for the subsequently transmitted mbufs. 2) When the network driver sees the "m->m_pkthdr.snd_tag" different from NULL, it will move the packets into a designated rate limited sendqueue given by the snd_tag pointer. It is up to the individual drivers how the rate limited traffic will be rate limited. 3) Route changes are detected by the NIC drivers in the ifp->if_transmit() routine when the ifnet pointer in the incoming snd_tag mismatches the one of the network interface. The network adapter frees the mbuf and returns EAGAIN which causes the ip_output() to release and clear the send tag. Upon next ip_output() a new "snd_tag" will be tried allocated. 4) When the PCB is detached the custom sendqueue will be released by a non-blocking ifp->if_snd_tag_free() call to the currently bound network interface. Reviewed by: wblock (manpages), adrian, gallatin, scottl (network) Differential Revision: https://reviews.freebsd.org/D3687 Sponsored by: Mellanox Technologies MFC after: 3 months
2017-01-18 13:31:17 +00:00
lagg(4): Add support for allocating TLS receive tags. The TLS receive tags are allocated directly from the receiving interface, because mbufs are flowing in the opposite direction and then route change checks are not useful, because they only work for outgoing traffic. Differential revision: https://reviews.freebsd.org/D32356 Sponsored by: NVIDIA Networking
2022-05-25 10:38:30 +00:00
if (sw != NULL) {
lst = malloc(sizeof(*lst), M_LAGG, M_NOWAIT);
if (lst == NULL) {
if_rele(lp_ifp);
return (ENOMEM);
}
} else
lst = NULL;
Restructure mbuf send tags to provide stronger guarantees. - Perform ifp mismatch checks (to determine if a send tag is allocated for a different ifp than the one the packet is being output on), in ip_output() and ip6_output(). This avoids sending packets with send tags to ifnet drivers that don't support send tags. Since we are now checking for ifp mismatches before invoking if_output, we can now try to allocate a new tag before invoking if_output sending the original packet on the new tag if allocation succeeds. To avoid code duplication for the fragment and unfragmented cases, add ip_output_send() and ip6_output_send() as wrappers around if_output and nd6_output_ifp, respectively. All of the logic for setting send tags and dealing with send tag-related errors is done in these wrapper functions. For pseudo interfaces that wrap other network interfaces (vlan and lagg), wrapper send tags are now allocated so that ip*_output see the wrapper ifp as the ifp in the send tag. The if_transmit routines rewrite the send tags after performing an ifp mismatch check. If an ifp mismatch is detected, the transmit routines fail with EAGAIN. - To provide clearer life cycle management of send tags, especially in the presence of vlan and lagg wrapper tags, add a reference count to send tags managed via m_snd_tag_ref() and m_snd_tag_rele(). Provide a helper function (m_snd_tag_init()) for use by drivers supporting send tags. m_snd_tag_init() takes care of the if_ref on the ifp meaning that code alloating send tags via if_snd_tag_alloc no longer has to manage that manually. Similarly, m_snd_tag_rele drops the refcount on the ifp after invoking if_snd_tag_free when the last reference to a send tag is dropped. This also closes use after free races if there are pending packets in driver tx rings after the socket is closed (e.g. from tcpdrop). In order for m_free to work reliably, add a new CSUM_SND_TAG flag in csum_flags to indicate 'snd_tag' is set (rather than 'rcvif'). Drivers now also check this flag instead of checking snd_tag against NULL. This avoids false positive matches when a forwarded packet has a non-NULL rcvif that was treated as a send tag. - cxgbe was relying on snd_tag_free being called when the inp was detached so that it could kick the firmware to flush any pending work on the flow. This is because the driver doesn't require ACK messages from the firmware for every request, but instead does a kind of manual interrupt coalescing by only setting a flag to request a completion on a subset of requests. If all of the in-flight requests don't have the flag when the tag is detached from the inp, the flow might never return the credits. The current snd_tag_free command issues a flush command to force the credits to return. However, the credit return is what also frees the mbufs, and since those mbufs now hold references on the tag, this meant that snd_tag_free would never be called. To fix, explicitly drop the mbuf's reference on the snd tag when the mbuf is queued in the firmware work queue. This means that once the inp's reference on the tag goes away and all in-flight mbufs have been queued to the firmware, tag's refcount will drop to zero and snd_tag_free will kick in and send the flush request. Note that we need to avoid doing this in the middle of ethofld_tx(), so the driver grabs a temporary reference on the tag around that loop to defer the free to the end of the function in case it sends the last mbuf to the queue after the inp has dropped its reference on the tag. - mlx5 preallocates send tags and was using the ifp pointer even when the send tag wasn't in use. Explicitly use the ifp from other data structures instead. - Sprinkle some assertions in various places to assert that received packets don't have a send tag, and that other places that overwrite rcvif (e.g. 802.11 transmit) don't clobber a send tag pointer. Reviewed by: gallatin, hselasky, rgrimes, ae Sponsored by: Netflix Differential Revision: https://reviews.freebsd.org/D20117
2019-05-24 22:30:40 +00:00
lagg(4): Add support for allocating TLS receive tags. The TLS receive tags are allocated directly from the receiving interface, because mbufs are flowing in the opposite direction and then route change checks are not useful, because they only work for outgoing traffic. Differential revision: https://reviews.freebsd.org/D32356 Sponsored by: NVIDIA Networking
2022-05-25 10:38:30 +00:00
error = m_snd_tag_alloc(lp_ifp, params, &mst);
Restructure mbuf send tags to provide stronger guarantees. - Perform ifp mismatch checks (to determine if a send tag is allocated for a different ifp than the one the packet is being output on), in ip_output() and ip6_output(). This avoids sending packets with send tags to ifnet drivers that don't support send tags. Since we are now checking for ifp mismatches before invoking if_output, we can now try to allocate a new tag before invoking if_output sending the original packet on the new tag if allocation succeeds. To avoid code duplication for the fragment and unfragmented cases, add ip_output_send() and ip6_output_send() as wrappers around if_output and nd6_output_ifp, respectively. All of the logic for setting send tags and dealing with send tag-related errors is done in these wrapper functions. For pseudo interfaces that wrap other network interfaces (vlan and lagg), wrapper send tags are now allocated so that ip*_output see the wrapper ifp as the ifp in the send tag. The if_transmit routines rewrite the send tags after performing an ifp mismatch check. If an ifp mismatch is detected, the transmit routines fail with EAGAIN. - To provide clearer life cycle management of send tags, especially in the presence of vlan and lagg wrapper tags, add a reference count to send tags managed via m_snd_tag_ref() and m_snd_tag_rele(). Provide a helper function (m_snd_tag_init()) for use by drivers supporting send tags. m_snd_tag_init() takes care of the if_ref on the ifp meaning that code alloating send tags via if_snd_tag_alloc no longer has to manage that manually. Similarly, m_snd_tag_rele drops the refcount on the ifp after invoking if_snd_tag_free when the last reference to a send tag is dropped. This also closes use after free races if there are pending packets in driver tx rings after the socket is closed (e.g. from tcpdrop). In order for m_free to work reliably, add a new CSUM_SND_TAG flag in csum_flags to indicate 'snd_tag' is set (rather than 'rcvif'). Drivers now also check this flag instead of checking snd_tag against NULL. This avoids false positive matches when a forwarded packet has a non-NULL rcvif that was treated as a send tag. - cxgbe was relying on snd_tag_free being called when the inp was detached so that it could kick the firmware to flush any pending work on the flow. This is because the driver doesn't require ACK messages from the firmware for every request, but instead does a kind of manual interrupt coalescing by only setting a flag to request a completion on a subset of requests. If all of the in-flight requests don't have the flag when the tag is detached from the inp, the flow might never return the credits. The current snd_tag_free command issues a flush command to force the credits to return. However, the credit return is what also frees the mbufs, and since those mbufs now hold references on the tag, this meant that snd_tag_free would never be called. To fix, explicitly drop the mbuf's reference on the snd tag when the mbuf is queued in the firmware work queue. This means that once the inp's reference on the tag goes away and all in-flight mbufs have been queued to the firmware, tag's refcount will drop to zero and snd_tag_free will kick in and send the flush request. Note that we need to avoid doing this in the middle of ethofld_tx(), so the driver grabs a temporary reference on the tag around that loop to defer the free to the end of the function in case it sends the last mbuf to the queue after the inp has dropped its reference on the tag. - mlx5 preallocates send tags and was using the ifp pointer even when the send tag wasn't in use. Explicitly use the ifp from other data structures instead. - Sprinkle some assertions in various places to assert that received packets don't have a send tag, and that other places that overwrite rcvif (e.g. 802.11 transmit) don't clobber a send tag pointer. Reviewed by: gallatin, hselasky, rgrimes, ae Sponsored by: Netflix Differential Revision: https://reviews.freebsd.org/D20117
2019-05-24 22:30:40 +00:00
if_rele(lp_ifp);
if (error) {
free(lst, M_LAGG);
return (error);
}
lagg(4): Add support for allocating TLS receive tags. The TLS receive tags are allocated directly from the receiving interface, because mbufs are flowing in the opposite direction and then route change checks are not useful, because they only work for outgoing traffic. Differential revision: https://reviews.freebsd.org/D32356 Sponsored by: NVIDIA Networking
2022-05-25 10:38:30 +00:00
if (sw != NULL) {
m_snd_tag_init(&lst->com, ifp, sw);
lst->tag = mst;
*ppmt = &lst->com;
} else
*ppmt = mst;
Restructure mbuf send tags to provide stronger guarantees. - Perform ifp mismatch checks (to determine if a send tag is allocated for a different ifp than the one the packet is being output on), in ip_output() and ip6_output(). This avoids sending packets with send tags to ifnet drivers that don't support send tags. Since we are now checking for ifp mismatches before invoking if_output, we can now try to allocate a new tag before invoking if_output sending the original packet on the new tag if allocation succeeds. To avoid code duplication for the fragment and unfragmented cases, add ip_output_send() and ip6_output_send() as wrappers around if_output and nd6_output_ifp, respectively. All of the logic for setting send tags and dealing with send tag-related errors is done in these wrapper functions. For pseudo interfaces that wrap other network interfaces (vlan and lagg), wrapper send tags are now allocated so that ip*_output see the wrapper ifp as the ifp in the send tag. The if_transmit routines rewrite the send tags after performing an ifp mismatch check. If an ifp mismatch is detected, the transmit routines fail with EAGAIN. - To provide clearer life cycle management of send tags, especially in the presence of vlan and lagg wrapper tags, add a reference count to send tags managed via m_snd_tag_ref() and m_snd_tag_rele(). Provide a helper function (m_snd_tag_init()) for use by drivers supporting send tags. m_snd_tag_init() takes care of the if_ref on the ifp meaning that code alloating send tags via if_snd_tag_alloc no longer has to manage that manually. Similarly, m_snd_tag_rele drops the refcount on the ifp after invoking if_snd_tag_free when the last reference to a send tag is dropped. This also closes use after free races if there are pending packets in driver tx rings after the socket is closed (e.g. from tcpdrop). In order for m_free to work reliably, add a new CSUM_SND_TAG flag in csum_flags to indicate 'snd_tag' is set (rather than 'rcvif'). Drivers now also check this flag instead of checking snd_tag against NULL. This avoids false positive matches when a forwarded packet has a non-NULL rcvif that was treated as a send tag. - cxgbe was relying on snd_tag_free being called when the inp was detached so that it could kick the firmware to flush any pending work on the flow. This is because the driver doesn't require ACK messages from the firmware for every request, but instead does a kind of manual interrupt coalescing by only setting a flag to request a completion on a subset of requests. If all of the in-flight requests don't have the flag when the tag is detached from the inp, the flow might never return the credits. The current snd_tag_free command issues a flush command to force the credits to return. However, the credit return is what also frees the mbufs, and since those mbufs now hold references on the tag, this meant that snd_tag_free would never be called. To fix, explicitly drop the mbuf's reference on the snd tag when the mbuf is queued in the firmware work queue. This means that once the inp's reference on the tag goes away and all in-flight mbufs have been queued to the firmware, tag's refcount will drop to zero and snd_tag_free will kick in and send the flush request. Note that we need to avoid doing this in the middle of ethofld_tx(), so the driver grabs a temporary reference on the tag around that loop to defer the free to the end of the function in case it sends the last mbuf to the queue after the inp has dropped its reference on the tag. - mlx5 preallocates send tags and was using the ifp pointer even when the send tag wasn't in use. Explicitly use the ifp from other data structures instead. - Sprinkle some assertions in various places to assert that received packets don't have a send tag, and that other places that overwrite rcvif (e.g. 802.11 transmit) don't clobber a send tag pointer. Reviewed by: gallatin, hselasky, rgrimes, ae Sponsored by: Netflix Differential Revision: https://reviews.freebsd.org/D20117
2019-05-24 22:30:40 +00:00
return (0);
}
This pulls over all the changes that are in the netflix tree that fix the ratelimit code. There were several bugs in tcp_ratelimit itself and we needed further work to support the multiple tag format coming for the joint TLS and Ratelimit dances. Sponsored by: Netflix Inc. Differential Revision: https://reviews.freebsd.org/D28357
2021-01-26 16:54:42 +00:00
static struct m_snd_tag *
lagg_next_snd_tag(struct m_snd_tag *mst)
{
struct lagg_snd_tag *lst;
lst = mst_to_lst(mst);
return (lst->tag);
}
Restructure mbuf send tags to provide stronger guarantees. - Perform ifp mismatch checks (to determine if a send tag is allocated for a different ifp than the one the packet is being output on), in ip_output() and ip6_output(). This avoids sending packets with send tags to ifnet drivers that don't support send tags. Since we are now checking for ifp mismatches before invoking if_output, we can now try to allocate a new tag before invoking if_output sending the original packet on the new tag if allocation succeeds. To avoid code duplication for the fragment and unfragmented cases, add ip_output_send() and ip6_output_send() as wrappers around if_output and nd6_output_ifp, respectively. All of the logic for setting send tags and dealing with send tag-related errors is done in these wrapper functions. For pseudo interfaces that wrap other network interfaces (vlan and lagg), wrapper send tags are now allocated so that ip*_output see the wrapper ifp as the ifp in the send tag. The if_transmit routines rewrite the send tags after performing an ifp mismatch check. If an ifp mismatch is detected, the transmit routines fail with EAGAIN. - To provide clearer life cycle management of send tags, especially in the presence of vlan and lagg wrapper tags, add a reference count to send tags managed via m_snd_tag_ref() and m_snd_tag_rele(). Provide a helper function (m_snd_tag_init()) for use by drivers supporting send tags. m_snd_tag_init() takes care of the if_ref on the ifp meaning that code alloating send tags via if_snd_tag_alloc no longer has to manage that manually. Similarly, m_snd_tag_rele drops the refcount on the ifp after invoking if_snd_tag_free when the last reference to a send tag is dropped. This also closes use after free races if there are pending packets in driver tx rings after the socket is closed (e.g. from tcpdrop). In order for m_free to work reliably, add a new CSUM_SND_TAG flag in csum_flags to indicate 'snd_tag' is set (rather than 'rcvif'). Drivers now also check this flag instead of checking snd_tag against NULL. This avoids false positive matches when a forwarded packet has a non-NULL rcvif that was treated as a send tag. - cxgbe was relying on snd_tag_free being called when the inp was detached so that it could kick the firmware to flush any pending work on the flow. This is because the driver doesn't require ACK messages from the firmware for every request, but instead does a kind of manual interrupt coalescing by only setting a flag to request a completion on a subset of requests. If all of the in-flight requests don't have the flag when the tag is detached from the inp, the flow might never return the credits. The current snd_tag_free command issues a flush command to force the credits to return. However, the credit return is what also frees the mbufs, and since those mbufs now hold references on the tag, this meant that snd_tag_free would never be called. To fix, explicitly drop the mbuf's reference on the snd tag when the mbuf is queued in the firmware work queue. This means that once the inp's reference on the tag goes away and all in-flight mbufs have been queued to the firmware, tag's refcount will drop to zero and snd_tag_free will kick in and send the flush request. Note that we need to avoid doing this in the middle of ethofld_tx(), so the driver grabs a temporary reference on the tag around that loop to defer the free to the end of the function in case it sends the last mbuf to the queue after the inp has dropped its reference on the tag. - mlx5 preallocates send tags and was using the ifp pointer even when the send tag wasn't in use. Explicitly use the ifp from other data structures instead. - Sprinkle some assertions in various places to assert that received packets don't have a send tag, and that other places that overwrite rcvif (e.g. 802.11 transmit) don't clobber a send tag pointer. Reviewed by: gallatin, hselasky, rgrimes, ae Sponsored by: Netflix Differential Revision: https://reviews.freebsd.org/D20117
2019-05-24 22:30:40 +00:00
static int
lagg_snd_tag_modify(struct m_snd_tag *mst,
union if_snd_tag_modify_params *params)
{
struct lagg_snd_tag *lst;
lst = mst_to_lst(mst);
Add a switch structure for send tags. Move the type and function pointers for operations on existing send tags (modify, query, next, free) out of 'struct ifnet' and into a new 'struct if_snd_tag_sw'. A pointer to this structure is added to the generic part of send tags and is initialized by m_snd_tag_init() (which now accepts a switch structure as a new argument in place of the type). Previously, device driver ifnet methods switched on the type to call type-specific functions. Now, those type-specific functions are saved in the switch structure and invoked directly. In addition, this more gracefully permits multiple implementations of the same tag within a driver. In particular, NIC TLS for future Chelsio adapters will use a different implementation than the existing NIC TLS support for T6 adapters. Reviewed by: gallatin, hselasky, kib (older version) Sponsored by: Chelsio Communications Differential Revision: https://reviews.freebsd.org/D31572
2021-09-14 18:43:41 +00:00
return (lst->tag->sw->snd_tag_modify(lst->tag, params));
Restructure mbuf send tags to provide stronger guarantees. - Perform ifp mismatch checks (to determine if a send tag is allocated for a different ifp than the one the packet is being output on), in ip_output() and ip6_output(). This avoids sending packets with send tags to ifnet drivers that don't support send tags. Since we are now checking for ifp mismatches before invoking if_output, we can now try to allocate a new tag before invoking if_output sending the original packet on the new tag if allocation succeeds. To avoid code duplication for the fragment and unfragmented cases, add ip_output_send() and ip6_output_send() as wrappers around if_output and nd6_output_ifp, respectively. All of the logic for setting send tags and dealing with send tag-related errors is done in these wrapper functions. For pseudo interfaces that wrap other network interfaces (vlan and lagg), wrapper send tags are now allocated so that ip*_output see the wrapper ifp as the ifp in the send tag. The if_transmit routines rewrite the send tags after performing an ifp mismatch check. If an ifp mismatch is detected, the transmit routines fail with EAGAIN. - To provide clearer life cycle management of send tags, especially in the presence of vlan and lagg wrapper tags, add a reference count to send tags managed via m_snd_tag_ref() and m_snd_tag_rele(). Provide a helper function (m_snd_tag_init()) for use by drivers supporting send tags. m_snd_tag_init() takes care of the if_ref on the ifp meaning that code alloating send tags via if_snd_tag_alloc no longer has to manage that manually. Similarly, m_snd_tag_rele drops the refcount on the ifp after invoking if_snd_tag_free when the last reference to a send tag is dropped. This also closes use after free races if there are pending packets in driver tx rings after the socket is closed (e.g. from tcpdrop). In order for m_free to work reliably, add a new CSUM_SND_TAG flag in csum_flags to indicate 'snd_tag' is set (rather than 'rcvif'). Drivers now also check this flag instead of checking snd_tag against NULL. This avoids false positive matches when a forwarded packet has a non-NULL rcvif that was treated as a send tag. - cxgbe was relying on snd_tag_free being called when the inp was detached so that it could kick the firmware to flush any pending work on the flow. This is because the driver doesn't require ACK messages from the firmware for every request, but instead does a kind of manual interrupt coalescing by only setting a flag to request a completion on a subset of requests. If all of the in-flight requests don't have the flag when the tag is detached from the inp, the flow might never return the credits. The current snd_tag_free command issues a flush command to force the credits to return. However, the credit return is what also frees the mbufs, and since those mbufs now hold references on the tag, this meant that snd_tag_free would never be called. To fix, explicitly drop the mbuf's reference on the snd tag when the mbuf is queued in the firmware work queue. This means that once the inp's reference on the tag goes away and all in-flight mbufs have been queued to the firmware, tag's refcount will drop to zero and snd_tag_free will kick in and send the flush request. Note that we need to avoid doing this in the middle of ethofld_tx(), so the driver grabs a temporary reference on the tag around that loop to defer the free to the end of the function in case it sends the last mbuf to the queue after the inp has dropped its reference on the tag. - mlx5 preallocates send tags and was using the ifp pointer even when the send tag wasn't in use. Explicitly use the ifp from other data structures instead. - Sprinkle some assertions in various places to assert that received packets don't have a send tag, and that other places that overwrite rcvif (e.g. 802.11 transmit) don't clobber a send tag pointer. Reviewed by: gallatin, hselasky, rgrimes, ae Sponsored by: Netflix Differential Revision: https://reviews.freebsd.org/D20117
2019-05-24 22:30:40 +00:00
}
static int
lagg_snd_tag_query(struct m_snd_tag *mst,
union if_snd_tag_query_params *params)
{
struct lagg_snd_tag *lst;
lst = mst_to_lst(mst);
Add a switch structure for send tags. Move the type and function pointers for operations on existing send tags (modify, query, next, free) out of 'struct ifnet' and into a new 'struct if_snd_tag_sw'. A pointer to this structure is added to the generic part of send tags and is initialized by m_snd_tag_init() (which now accepts a switch structure as a new argument in place of the type). Previously, device driver ifnet methods switched on the type to call type-specific functions. Now, those type-specific functions are saved in the switch structure and invoked directly. In addition, this more gracefully permits multiple implementations of the same tag within a driver. In particular, NIC TLS for future Chelsio adapters will use a different implementation than the existing NIC TLS support for T6 adapters. Reviewed by: gallatin, hselasky, kib (older version) Sponsored by: Chelsio Communications Differential Revision: https://reviews.freebsd.org/D31572
2021-09-14 18:43:41 +00:00
return (lst->tag->sw->snd_tag_query(lst->tag, params));
Implement kernel support for hardware rate limited sockets. - Add RATELIMIT kernel configuration keyword which must be set to enable the new functionality. - Add support for hardware driven, Receive Side Scaling, RSS aware, rate limited sendqueues and expose the functionality through the already established SO_MAX_PACING_RATE setsockopt(). The API support rates in the range from 1 to 4Gbytes/s which are suitable for regular TCP and UDP streams. The setsockopt(2) manual page has been updated. - Add rate limit function callback API to "struct ifnet" which supports the following operations: if_snd_tag_alloc(), if_snd_tag_modify(), if_snd_tag_query() and if_snd_tag_free(). - Add support to ifconfig to view, set and clear the IFCAP_TXRTLMT flag, which tells if a network driver supports rate limiting or not. - This patch also adds support for rate limiting through VLAN and LAGG intermediate network devices. - How rate limiting works: 1) The userspace application calls setsockopt() after accepting or making a new connection to set the rate which is then stored in the socket structure in the kernel. Later on when packets are transmitted a check is made in the transmit path for rate changes. A rate change implies a non-blocking ifp->if_snd_tag_alloc() call will be made to the destination network interface, which then sets up a custom sendqueue with the given rate limitation parameter. A "struct m_snd_tag" pointer is returned which serves as a "snd_tag" hint in the m_pkthdr for the subsequently transmitted mbufs. 2) When the network driver sees the "m->m_pkthdr.snd_tag" different from NULL, it will move the packets into a designated rate limited sendqueue given by the snd_tag pointer. It is up to the individual drivers how the rate limited traffic will be rate limited. 3) Route changes are detected by the NIC drivers in the ifp->if_transmit() routine when the ifnet pointer in the incoming snd_tag mismatches the one of the network interface. The network adapter frees the mbuf and returns EAGAIN which causes the ip_output() to release and clear the send tag. Upon next ip_output() a new "snd_tag" will be tried allocated. 4) When the PCB is detached the custom sendqueue will be released by a non-blocking ifp->if_snd_tag_free() call to the currently bound network interface. Reviewed by: wblock (manpages), adrian, gallatin, scottl (network) Differential Revision: https://reviews.freebsd.org/D3687 Sponsored by: Mellanox Technologies MFC after: 3 months
2017-01-18 13:31:17 +00:00
}
This commit adds the missing release mechanism for the ratelimiting code. The two modules (lagg and vlan) did have allocation routines, and even though they are indirect (and vector down to the underlying interfaces) they both need to have a free routine (that also vectors down to the actual interface). Sponsored by: Netflix Inc. Differential Revision: https://reviews.freebsd.org/D19032
2019-02-13 14:57:59 +00:00
static void
Restructure mbuf send tags to provide stronger guarantees. - Perform ifp mismatch checks (to determine if a send tag is allocated for a different ifp than the one the packet is being output on), in ip_output() and ip6_output(). This avoids sending packets with send tags to ifnet drivers that don't support send tags. Since we are now checking for ifp mismatches before invoking if_output, we can now try to allocate a new tag before invoking if_output sending the original packet on the new tag if allocation succeeds. To avoid code duplication for the fragment and unfragmented cases, add ip_output_send() and ip6_output_send() as wrappers around if_output and nd6_output_ifp, respectively. All of the logic for setting send tags and dealing with send tag-related errors is done in these wrapper functions. For pseudo interfaces that wrap other network interfaces (vlan and lagg), wrapper send tags are now allocated so that ip*_output see the wrapper ifp as the ifp in the send tag. The if_transmit routines rewrite the send tags after performing an ifp mismatch check. If an ifp mismatch is detected, the transmit routines fail with EAGAIN. - To provide clearer life cycle management of send tags, especially in the presence of vlan and lagg wrapper tags, add a reference count to send tags managed via m_snd_tag_ref() and m_snd_tag_rele(). Provide a helper function (m_snd_tag_init()) for use by drivers supporting send tags. m_snd_tag_init() takes care of the if_ref on the ifp meaning that code alloating send tags via if_snd_tag_alloc no longer has to manage that manually. Similarly, m_snd_tag_rele drops the refcount on the ifp after invoking if_snd_tag_free when the last reference to a send tag is dropped. This also closes use after free races if there are pending packets in driver tx rings after the socket is closed (e.g. from tcpdrop). In order for m_free to work reliably, add a new CSUM_SND_TAG flag in csum_flags to indicate 'snd_tag' is set (rather than 'rcvif'). Drivers now also check this flag instead of checking snd_tag against NULL. This avoids false positive matches when a forwarded packet has a non-NULL rcvif that was treated as a send tag. - cxgbe was relying on snd_tag_free being called when the inp was detached so that it could kick the firmware to flush any pending work on the flow. This is because the driver doesn't require ACK messages from the firmware for every request, but instead does a kind of manual interrupt coalescing by only setting a flag to request a completion on a subset of requests. If all of the in-flight requests don't have the flag when the tag is detached from the inp, the flow might never return the credits. The current snd_tag_free command issues a flush command to force the credits to return. However, the credit return is what also frees the mbufs, and since those mbufs now hold references on the tag, this meant that snd_tag_free would never be called. To fix, explicitly drop the mbuf's reference on the snd tag when the mbuf is queued in the firmware work queue. This means that once the inp's reference on the tag goes away and all in-flight mbufs have been queued to the firmware, tag's refcount will drop to zero and snd_tag_free will kick in and send the flush request. Note that we need to avoid doing this in the middle of ethofld_tx(), so the driver grabs a temporary reference on the tag around that loop to defer the free to the end of the function in case it sends the last mbuf to the queue after the inp has dropped its reference on the tag. - mlx5 preallocates send tags and was using the ifp pointer even when the send tag wasn't in use. Explicitly use the ifp from other data structures instead. - Sprinkle some assertions in various places to assert that received packets don't have a send tag, and that other places that overwrite rcvif (e.g. 802.11 transmit) don't clobber a send tag pointer. Reviewed by: gallatin, hselasky, rgrimes, ae Sponsored by: Netflix Differential Revision: https://reviews.freebsd.org/D20117
2019-05-24 22:30:40 +00:00
lagg_snd_tag_free(struct m_snd_tag *mst)
This commit adds the missing release mechanism for the ratelimiting code. The two modules (lagg and vlan) did have allocation routines, and even though they are indirect (and vector down to the underlying interfaces) they both need to have a free routine (that also vectors down to the actual interface). Sponsored by: Netflix Inc. Differential Revision: https://reviews.freebsd.org/D19032
2019-02-13 14:57:59 +00:00
{
Restructure mbuf send tags to provide stronger guarantees. - Perform ifp mismatch checks (to determine if a send tag is allocated for a different ifp than the one the packet is being output on), in ip_output() and ip6_output(). This avoids sending packets with send tags to ifnet drivers that don't support send tags. Since we are now checking for ifp mismatches before invoking if_output, we can now try to allocate a new tag before invoking if_output sending the original packet on the new tag if allocation succeeds. To avoid code duplication for the fragment and unfragmented cases, add ip_output_send() and ip6_output_send() as wrappers around if_output and nd6_output_ifp, respectively. All of the logic for setting send tags and dealing with send tag-related errors is done in these wrapper functions. For pseudo interfaces that wrap other network interfaces (vlan and lagg), wrapper send tags are now allocated so that ip*_output see the wrapper ifp as the ifp in the send tag. The if_transmit routines rewrite the send tags after performing an ifp mismatch check. If an ifp mismatch is detected, the transmit routines fail with EAGAIN. - To provide clearer life cycle management of send tags, especially in the presence of vlan and lagg wrapper tags, add a reference count to send tags managed via m_snd_tag_ref() and m_snd_tag_rele(). Provide a helper function (m_snd_tag_init()) for use by drivers supporting send tags. m_snd_tag_init() takes care of the if_ref on the ifp meaning that code alloating send tags via if_snd_tag_alloc no longer has to manage that manually. Similarly, m_snd_tag_rele drops the refcount on the ifp after invoking if_snd_tag_free when the last reference to a send tag is dropped. This also closes use after free races if there are pending packets in driver tx rings after the socket is closed (e.g. from tcpdrop). In order for m_free to work reliably, add a new CSUM_SND_TAG flag in csum_flags to indicate 'snd_tag' is set (rather than 'rcvif'). Drivers now also check this flag instead of checking snd_tag against NULL. This avoids false positive matches when a forwarded packet has a non-NULL rcvif that was treated as a send tag. - cxgbe was relying on snd_tag_free being called when the inp was detached so that it could kick the firmware to flush any pending work on the flow. This is because the driver doesn't require ACK messages from the firmware for every request, but instead does a kind of manual interrupt coalescing by only setting a flag to request a completion on a subset of requests. If all of the in-flight requests don't have the flag when the tag is detached from the inp, the flow might never return the credits. The current snd_tag_free command issues a flush command to force the credits to return. However, the credit return is what also frees the mbufs, and since those mbufs now hold references on the tag, this meant that snd_tag_free would never be called. To fix, explicitly drop the mbuf's reference on the snd tag when the mbuf is queued in the firmware work queue. This means that once the inp's reference on the tag goes away and all in-flight mbufs have been queued to the firmware, tag's refcount will drop to zero and snd_tag_free will kick in and send the flush request. Note that we need to avoid doing this in the middle of ethofld_tx(), so the driver grabs a temporary reference on the tag around that loop to defer the free to the end of the function in case it sends the last mbuf to the queue after the inp has dropped its reference on the tag. - mlx5 preallocates send tags and was using the ifp pointer even when the send tag wasn't in use. Explicitly use the ifp from other data structures instead. - Sprinkle some assertions in various places to assert that received packets don't have a send tag, and that other places that overwrite rcvif (e.g. 802.11 transmit) don't clobber a send tag pointer. Reviewed by: gallatin, hselasky, rgrimes, ae Sponsored by: Netflix Differential Revision: https://reviews.freebsd.org/D20117
2019-05-24 22:30:40 +00:00
struct lagg_snd_tag *lst;
lst = mst_to_lst(mst);
m_snd_tag_rele(lst->tag);
free(lst, M_LAGG);
This commit adds the missing release mechanism for the ratelimiting code. The two modules (lagg and vlan) did have allocation routines, and even though they are indirect (and vector down to the underlying interfaces) they both need to have a free routine (that also vectors down to the actual interface). Sponsored by: Netflix Inc. Differential Revision: https://reviews.freebsd.org/D19032
2019-02-13 14:57:59 +00:00
}
This adds the third step in getting BBR into the tree. BBR and an updated rack depend on having access to the new ratelimit api in this commit. Sponsored by: Netflix Inc. Differential Revision: https://reviews.freebsd.org/D20953
2019-08-01 14:17:31 +00:00
static void
lagg_ratelimit_query(struct ifnet *ifp __unused, struct if_ratelimit_query_results *q)
{
/*
* For lagg, we have an indirect
* interface. The caller needs to
* get a ratelimit tag on the actual
* interface the flow will go on.
*/
q->rate_table = NULL;
q->flags = RT_IS_INDIRECT;
q->max_flows = 0;
q->number_of_rates = 0;
}
Implement kernel support for hardware rate limited sockets. - Add RATELIMIT kernel configuration keyword which must be set to enable the new functionality. - Add support for hardware driven, Receive Side Scaling, RSS aware, rate limited sendqueues and expose the functionality through the already established SO_MAX_PACING_RATE setsockopt(). The API support rates in the range from 1 to 4Gbytes/s which are suitable for regular TCP and UDP streams. The setsockopt(2) manual page has been updated. - Add rate limit function callback API to "struct ifnet" which supports the following operations: if_snd_tag_alloc(), if_snd_tag_modify(), if_snd_tag_query() and if_snd_tag_free(). - Add support to ifconfig to view, set and clear the IFCAP_TXRTLMT flag, which tells if a network driver supports rate limiting or not. - This patch also adds support for rate limiting through VLAN and LAGG intermediate network devices. - How rate limiting works: 1) The userspace application calls setsockopt() after accepting or making a new connection to set the rate which is then stored in the socket structure in the kernel. Later on when packets are transmitted a check is made in the transmit path for rate changes. A rate change implies a non-blocking ifp->if_snd_tag_alloc() call will be made to the destination network interface, which then sets up a custom sendqueue with the given rate limitation parameter. A "struct m_snd_tag" pointer is returned which serves as a "snd_tag" hint in the m_pkthdr for the subsequently transmitted mbufs. 2) When the network driver sees the "m->m_pkthdr.snd_tag" different from NULL, it will move the packets into a designated rate limited sendqueue given by the snd_tag pointer. It is up to the individual drivers how the rate limited traffic will be rate limited. 3) Route changes are detected by the NIC drivers in the ifp->if_transmit() routine when the ifnet pointer in the incoming snd_tag mismatches the one of the network interface. The network adapter frees the mbuf and returns EAGAIN which causes the ip_output() to release and clear the send tag. Upon next ip_output() a new "snd_tag" will be tried allocated. 4) When the PCB is detached the custom sendqueue will be released by a non-blocking ifp->if_snd_tag_free() call to the currently bound network interface. Reviewed by: wblock (manpages), adrian, gallatin, scottl (network) Differential Revision: https://reviews.freebsd.org/D3687 Sponsored by: Mellanox Technologies MFC after: 3 months
2017-01-18 13:31:17 +00:00
#endif
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
static int
Introduce sleepable locks into if_lagg. Before this change if_lagg was using nonsleepable rmlocks to protect its internal state. This patch introduces another sx lock to protect code paths that require sleeping, while still uses old rmlock to protect hot nonsleepable data paths. This change allows to remove taskqueue decoupling used before to change interface addresses without holding the lock. Instead it uses sx lock to protect direct if_ioctl() calls. As another bonus, the new code synchronizes enabled capabilities of member interfaces, and allows to control them with ifconfig laggX, that was impossible before. This part should fix interoperation with if_bridge, that may need to disable some capabilities, such as TXCSUM or LRO, to allow bridging with noncapable interfaces. MFC after: 2 weeks Sponsored by: iXsystems, Inc. Differential Revision: https://reviews.freebsd.org/D10514
2017-05-02 19:09:11 +00:00
lagg_setmulti(struct lagg_port *lp)
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
{
Introduce sleepable locks into if_lagg. Before this change if_lagg was using nonsleepable rmlocks to protect its internal state. This patch introduces another sx lock to protect code paths that require sleeping, while still uses old rmlock to protect hot nonsleepable data paths. This change allows to remove taskqueue decoupling used before to change interface addresses without holding the lock. Instead it uses sx lock to protect direct if_ioctl() calls. As another bonus, the new code synchronizes enabled capabilities of member interfaces, and allows to control them with ifconfig laggX, that was impossible before. This part should fix interoperation with if_bridge, that may need to disable some capabilities, such as TXCSUM or LRO, to allow bridging with noncapable interfaces. MFC after: 2 weeks Sponsored by: iXsystems, Inc. Differential Revision: https://reviews.freebsd.org/D10514
2017-05-02 19:09:11 +00:00
struct lagg_softc *sc = lp->lp_softc;
struct ifnet *ifp = lp->lp_ifp;
struct ifnet *scifp = sc->sc_ifp;
struct lagg_mc *mc;
struct ifmultiaddr *ifma;
int error;
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
Introduce sleepable locks into if_lagg. Before this change if_lagg was using nonsleepable rmlocks to protect its internal state. This patch introduces another sx lock to protect code paths that require sleeping, while still uses old rmlock to protect hot nonsleepable data paths. This change allows to remove taskqueue decoupling used before to change interface addresses without holding the lock. Instead it uses sx lock to protect direct if_ioctl() calls. As another bonus, the new code synchronizes enabled capabilities of member interfaces, and allows to control them with ifconfig laggX, that was impossible before. This part should fix interoperation with if_bridge, that may need to disable some capabilities, such as TXCSUM or LRO, to allow bridging with noncapable interfaces. MFC after: 2 weeks Sponsored by: iXsystems, Inc. Differential Revision: https://reviews.freebsd.org/D10514
2017-05-02 19:09:11 +00:00
IF_ADDR_WLOCK(scifp);
ifnet: Replace if_addr_lock rwlock with epoch + mutex Run on LLNW canaries and tested by pho@ gallatin: Using a 14-core, 28-HTT single socket E5-2697 v3 with a 40GbE MLX5 based ConnectX 4-LX NIC, I see an almost 12% improvement in received packet rate, and a larger improvement in bytes delivered all the way to userspace. When the host receiving 64 streams of netperf -H $DUT -t UDP_STREAM -- -m 1, I see, using nstat -I mce0 1 before the patch: InMpps OMpps InGbs OGbs err TCP Est %CPU syscalls csw irq GBfree 4.98 0.00 4.42 0.00 4235592 33 83.80 4720653 2149771 1235 247.32 4.73 0.00 4.20 0.00 4025260 33 82.99 4724900 2139833 1204 247.32 4.72 0.00 4.20 0.00 4035252 33 82.14 4719162 2132023 1264 247.32 4.71 0.00 4.21 0.00 4073206 33 83.68 4744973 2123317 1347 247.32 4.72 0.00 4.21 0.00 4061118 33 80.82 4713615 2188091 1490 247.32 4.72 0.00 4.21 0.00 4051675 33 85.29 4727399 2109011 1205 247.32 4.73 0.00 4.21 0.00 4039056 33 84.65 4724735 2102603 1053 247.32 After the patch InMpps OMpps InGbs OGbs err TCP Est %CPU syscalls csw irq GBfree 5.43 0.00 4.20 0.00 3313143 33 84.96 5434214 1900162 2656 245.51 5.43 0.00 4.20 0.00 3308527 33 85.24 5439695 1809382 2521 245.51 5.42 0.00 4.19 0.00 3316778 33 87.54 5416028 1805835 2256 245.51 5.42 0.00 4.19 0.00 3317673 33 90.44 5426044 1763056 2332 245.51 5.42 0.00 4.19 0.00 3314839 33 88.11 5435732 1792218 2499 245.52 5.44 0.00 4.19 0.00 3293228 33 91.84 5426301 1668597 2121 245.52 Similarly, netperf reports 230Mb/s before the patch, and 270Mb/s after the patch Reviewed by: gallatin Sponsored by: Limelight Networks Differential Revision: https://reviews.freebsd.org/D15366
2018-05-18 20:13:34 +00:00
CK_STAILQ_FOREACH(ifma, &scifp->if_multiaddrs, ifma_link) {
Introduce sleepable locks into if_lagg. Before this change if_lagg was using nonsleepable rmlocks to protect its internal state. This patch introduces another sx lock to protect code paths that require sleeping, while still uses old rmlock to protect hot nonsleepable data paths. This change allows to remove taskqueue decoupling used before to change interface addresses without holding the lock. Instead it uses sx lock to protect direct if_ioctl() calls. As another bonus, the new code synchronizes enabled capabilities of member interfaces, and allows to control them with ifconfig laggX, that was impossible before. This part should fix interoperation with if_bridge, that may need to disable some capabilities, such as TXCSUM or LRO, to allow bridging with noncapable interfaces. MFC after: 2 weeks Sponsored by: iXsystems, Inc. Differential Revision: https://reviews.freebsd.org/D10514
2017-05-02 19:09:11 +00:00
if (ifma->ifma_addr->sa_family != AF_LINK)
continue;
Use a dedicated malloc type for lagg(4)'s structures. Reviewed by: gallatin MFC after: 1 month Sponsored by: Netflix Differential Revision: https://reviews.freebsd.org/D19719
2019-03-28 21:00:54 +00:00
mc = malloc(sizeof(struct lagg_mc), M_LAGG, M_NOWAIT);
Introduce sleepable locks into if_lagg. Before this change if_lagg was using nonsleepable rmlocks to protect its internal state. This patch introduces another sx lock to protect code paths that require sleeping, while still uses old rmlock to protect hot nonsleepable data paths. This change allows to remove taskqueue decoupling used before to change interface addresses without holding the lock. Instead it uses sx lock to protect direct if_ioctl() calls. As another bonus, the new code synchronizes enabled capabilities of member interfaces, and allows to control them with ifconfig laggX, that was impossible before. This part should fix interoperation with if_bridge, that may need to disable some capabilities, such as TXCSUM or LRO, to allow bridging with noncapable interfaces. MFC after: 2 weeks Sponsored by: iXsystems, Inc. Differential Revision: https://reviews.freebsd.org/D10514
2017-05-02 19:09:11 +00:00
if (mc == NULL) {
IF_ADDR_WUNLOCK(scifp);
return (ENOMEM);
}
bcopy(ifma->ifma_addr, &mc->mc_addr,
ifma->ifma_addr->sa_len);
mc->mc_addr.sdl_index = ifp->if_index;
mc->mc_ifma = NULL;
SLIST_INSERT_HEAD(&lp->lp_mc_head, mc, mc_entries);
}
IF_ADDR_WUNLOCK(scifp);
SLIST_FOREACH (mc, &lp->lp_mc_head, mc_entries) {
error = if_addmulti(ifp,
(struct sockaddr *)&mc->mc_addr, &mc->mc_ifma);
if (error)
return (error);
- Correctly check if lp_ioctl is null - Remove lagg_ether_purgemulti as its no longer needed - Mark the interface as up if any ports are active rather than just the primary
2007-05-07 09:53:02 +00:00
}
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
return (0);
}
static int
Introduce sleepable locks into if_lagg. Before this change if_lagg was using nonsleepable rmlocks to protect its internal state. This patch introduces another sx lock to protect code paths that require sleeping, while still uses old rmlock to protect hot nonsleepable data paths. This change allows to remove taskqueue decoupling used before to change interface addresses without holding the lock. Instead it uses sx lock to protect direct if_ioctl() calls. As another bonus, the new code synchronizes enabled capabilities of member interfaces, and allows to control them with ifconfig laggX, that was impossible before. This part should fix interoperation with if_bridge, that may need to disable some capabilities, such as TXCSUM or LRO, to allow bridging with noncapable interfaces. MFC after: 2 weeks Sponsored by: iXsystems, Inc. Differential Revision: https://reviews.freebsd.org/D10514
2017-05-02 19:09:11 +00:00
lagg_clrmulti(struct lagg_port *lp)
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
{
Change from using if_delmulti() to if_delmulti_ifma() as it simplifies the code and is safe to use if the ifp has disappeared. Suggested by: bms
2007-05-07 00:18:56 +00:00
struct lagg_mc *mc;
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
Replace rmlock with epoch in lagg Use the new epoch based reclamation API. Now the hot paths will not block at all, and the sx lock is used for the softc data. This fixes LORs reported where the rwlock was obtained when the sxlock was held. Submitted by: mmacy Reported by: Harry Schmalzbauer <freebsd@omnilan.de> Reviewed by: sbruno Sponsored by: Limelight Networks Differential Revision: https://reviews.freebsd.org/D15355
2018-05-14 20:06:49 +00:00
LAGG_XLOCK_ASSERT(lp->lp_softc);
Introduce sleepable locks into if_lagg. Before this change if_lagg was using nonsleepable rmlocks to protect its internal state. This patch introduces another sx lock to protect code paths that require sleeping, while still uses old rmlock to protect hot nonsleepable data paths. This change allows to remove taskqueue decoupling used before to change interface addresses without holding the lock. Instead it uses sx lock to protect direct if_ioctl() calls. As another bonus, the new code synchronizes enabled capabilities of member interfaces, and allows to control them with ifconfig laggX, that was impossible before. This part should fix interoperation with if_bridge, that may need to disable some capabilities, such as TXCSUM or LRO, to allow bridging with noncapable interfaces. MFC after: 2 weeks Sponsored by: iXsystems, Inc. Differential Revision: https://reviews.freebsd.org/D10514
2017-05-02 19:09:11 +00:00
while ((mc = SLIST_FIRST(&lp->lp_mc_head)) != NULL) {
SLIST_REMOVE(&lp->lp_mc_head, mc, lagg_mc, mc_entries);
if (mc->mc_ifma && lp->lp_detaching == 0)
if_delmulti_ifma(mc->mc_ifma);
Use a dedicated malloc type for lagg(4)'s structures. Reviewed by: gallatin MFC after: 1 month Sponsored by: Netflix Differential Revision: https://reviews.freebsd.org/D19719
2019-03-28 21:00:54 +00:00
free(mc, M_LAGG);
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
}
return (0);
}
lagg: fix lagg ifioctl after SIOCSIFCAPNV Lagg was broken by SIOCSIFCAPNV when all underlying devices support SIOCSIFCAPNV. This change updates lagg to work with SIOCSIFCAPNV and if_capabilities2. Reviewed by: kib, hselasky Sponsored by: Netflix Differential Revision: https://reviews.freebsd.org/D35865
2022-07-28 14:36:22 +00:00
static void
lagg_setcaps(struct lagg_port *lp, int cap, int cap2)
Introduce sleepable locks into if_lagg. Before this change if_lagg was using nonsleepable rmlocks to protect its internal state. This patch introduces another sx lock to protect code paths that require sleeping, while still uses old rmlock to protect hot nonsleepable data paths. This change allows to remove taskqueue decoupling used before to change interface addresses without holding the lock. Instead it uses sx lock to protect direct if_ioctl() calls. As another bonus, the new code synchronizes enabled capabilities of member interfaces, and allows to control them with ifconfig laggX, that was impossible before. This part should fix interoperation with if_bridge, that may need to disable some capabilities, such as TXCSUM or LRO, to allow bridging with noncapable interfaces. MFC after: 2 weeks Sponsored by: iXsystems, Inc. Differential Revision: https://reviews.freebsd.org/D10514
2017-05-02 19:09:11 +00:00
{
struct ifreq ifr;
lagg: fix lagg ifioctl after SIOCSIFCAPNV Lagg was broken by SIOCSIFCAPNV when all underlying devices support SIOCSIFCAPNV. This change updates lagg to work with SIOCSIFCAPNV and if_capabilities2. Reviewed by: kib, hselasky Sponsored by: Netflix Differential Revision: https://reviews.freebsd.org/D35865
2022-07-28 14:36:22 +00:00
struct siocsifcapnv_driver_data drv_ioctl_data;
Introduce sleepable locks into if_lagg. Before this change if_lagg was using nonsleepable rmlocks to protect its internal state. This patch introduces another sx lock to protect code paths that require sleeping, while still uses old rmlock to protect hot nonsleepable data paths. This change allows to remove taskqueue decoupling used before to change interface addresses without holding the lock. Instead it uses sx lock to protect direct if_ioctl() calls. As another bonus, the new code synchronizes enabled capabilities of member interfaces, and allows to control them with ifconfig laggX, that was impossible before. This part should fix interoperation with if_bridge, that may need to disable some capabilities, such as TXCSUM or LRO, to allow bridging with noncapable interfaces. MFC after: 2 weeks Sponsored by: iXsystems, Inc. Differential Revision: https://reviews.freebsd.org/D10514
2017-05-02 19:09:11 +00:00
lagg: fix lagg ifioctl after SIOCSIFCAPNV Lagg was broken by SIOCSIFCAPNV when all underlying devices support SIOCSIFCAPNV. This change updates lagg to work with SIOCSIFCAPNV and if_capabilities2. Reviewed by: kib, hselasky Sponsored by: Netflix Differential Revision: https://reviews.freebsd.org/D35865
2022-07-28 14:36:22 +00:00
if (lp->lp_ifp->if_capenable == cap &&
lp->lp_ifp->if_capenable2 == cap2)
return;
Introduce sleepable locks into if_lagg. Before this change if_lagg was using nonsleepable rmlocks to protect its internal state. This patch introduces another sx lock to protect code paths that require sleeping, while still uses old rmlock to protect hot nonsleepable data paths. This change allows to remove taskqueue decoupling used before to change interface addresses without holding the lock. Instead it uses sx lock to protect direct if_ioctl() calls. As another bonus, the new code synchronizes enabled capabilities of member interfaces, and allows to control them with ifconfig laggX, that was impossible before. This part should fix interoperation with if_bridge, that may need to disable some capabilities, such as TXCSUM or LRO, to allow bridging with noncapable interfaces. MFC after: 2 weeks Sponsored by: iXsystems, Inc. Differential Revision: https://reviews.freebsd.org/D10514
2017-05-02 19:09:11 +00:00
if (lp->lp_ioctl == NULL)
lagg: fix lagg ifioctl after SIOCSIFCAPNV Lagg was broken by SIOCSIFCAPNV when all underlying devices support SIOCSIFCAPNV. This change updates lagg to work with SIOCSIFCAPNV and if_capabilities2. Reviewed by: kib, hselasky Sponsored by: Netflix Differential Revision: https://reviews.freebsd.org/D35865
2022-07-28 14:36:22 +00:00
return;
/* XXX */
if ((lp->lp_ifp->if_capabilities & IFCAP_NV) != 0) {
drv_ioctl_data.reqcap = cap;
drv_ioctl_data.reqcap2 = cap2;
drv_ioctl_data.nvcap = NULL;
(*lp->lp_ioctl)(lp->lp_ifp, SIOCSIFCAPNV,
(caddr_t)&drv_ioctl_data);
} else {
ifr.ifr_reqcap = cap;
(*lp->lp_ioctl)(lp->lp_ifp, SIOCSIFCAP, (caddr_t)&ifr);
}
Introduce sleepable locks into if_lagg. Before this change if_lagg was using nonsleepable rmlocks to protect its internal state. This patch introduces another sx lock to protect code paths that require sleeping, while still uses old rmlock to protect hot nonsleepable data paths. This change allows to remove taskqueue decoupling used before to change interface addresses without holding the lock. Instead it uses sx lock to protect direct if_ioctl() calls. As another bonus, the new code synchronizes enabled capabilities of member interfaces, and allows to control them with ifconfig laggX, that was impossible before. This part should fix interoperation with if_bridge, that may need to disable some capabilities, such as TXCSUM or LRO, to allow bridging with noncapable interfaces. MFC after: 2 weeks Sponsored by: iXsystems, Inc. Differential Revision: https://reviews.freebsd.org/D10514
2017-05-02 19:09:11 +00:00
}
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
/* Handle a ref counted flag that should be set on the lagg port as well */
static int
lagg_setflag(struct lagg_port *lp, int flag, int status,
- Whitespace. - Remove caddr_t.
2014-09-26 12:35:58 +00:00
int (*func)(struct ifnet *, int))
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
{
non-functional cleanup - remove dead code - use consistent variable names - gc unused defines - whitespace cleanup
2007-06-12 07:29:11 +00:00
struct lagg_softc *sc = lp->lp_softc;
struct ifnet *scifp = sc->sc_ifp;
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
struct ifnet *ifp = lp->lp_ifp;
int error;
Introduce sleepable locks into if_lagg. Before this change if_lagg was using nonsleepable rmlocks to protect its internal state. This patch introduces another sx lock to protect code paths that require sleeping, while still uses old rmlock to protect hot nonsleepable data paths. This change allows to remove taskqueue decoupling used before to change interface addresses without holding the lock. Instead it uses sx lock to protect direct if_ioctl() calls. As another bonus, the new code synchronizes enabled capabilities of member interfaces, and allows to control them with ifconfig laggX, that was impossible before. This part should fix interoperation with if_bridge, that may need to disable some capabilities, such as TXCSUM or LRO, to allow bridging with noncapable interfaces. MFC after: 2 weeks Sponsored by: iXsystems, Inc. Differential Revision: https://reviews.freebsd.org/D10514
2017-05-02 19:09:11 +00:00
LAGG_XLOCK_ASSERT(sc);
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
non-functional cleanup - remove dead code - use consistent variable names - gc unused defines - whitespace cleanup
2007-06-12 07:29:11 +00:00
status = status ? (scifp->if_flags & flag) : 0;
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
/* Now "status" contains the flag value or 0 */
/*
* See if recorded ports status is different from what
* we want it to be. If it is, flip it. We record ports
* status in lp_ifflags so that we won't clear ports flag
* we haven't set. In fact, we don't clear or set ports
* flags directly, but get or release references to them.
* That's why we can be sure that recorded flags still are
* in accord with actual ports flags.
*/
if (status != (lp->lp_ifflags & flag)) {
error = (*func)(ifp, status);
if (error)
return (error);
lp->lp_ifflags &= ~flag;
lp->lp_ifflags |= status;
}
return (0);
}
/*
* Handle IFF_* flags that require certain changes on the lagg port
* if "status" is true, update ports flags respective to the lagg
* if "status" is false, forcedly clear the flags set on port.
*/
static int
lagg_setflags(struct lagg_port *lp, int status)
{
int error, i;
non-functional cleanup - remove dead code - use consistent variable names - gc unused defines - whitespace cleanup
2007-06-12 07:29:11 +00:00
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
for (i = 0; lagg_pflags[i].flag; i++) {
error = lagg_setflag(lp, lagg_pflags[i].flag,
status, lagg_pflags[i].func);
if (error)
return (error);
}
return (0);
}
Convert lagg(4) to use if_transmit instead of if_start. In collaboration with: thompsa, sbruno, fabient
2012-09-20 10:05:10 +00:00
static int
Add support for IP over infiniband, IPoIB, to lagg(4). Currently only the failover protocol is supported due to limitations in the IPoIB architecture. Refer to the lagg(4) manual page for how to configure and use this new feature. A new network interface type, IFT_INFINIBANDLAG, has been added, similar to the existing IFT_IEEE8023ADLAG . ifconfig(8) has been updated to accept a new laggtype argument when creating lagg(4) network interfaces. This new argument is used to distinguish between ethernet and infiniband type of lagg(4) network interface. The laggtype argument is optional and defaults to ethernet. The lagg(4) command line syntax is backwards compatible. Differential Revision: https://reviews.freebsd.org/D26254 Reviewed by: melifaro@ MFC after: 1 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2020-10-22 09:47:12 +00:00
lagg_transmit_ethernet(struct ifnet *ifp, struct mbuf *m)
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
{
Convert LAGG_RLOCK() to NET_EPOCH_ENTER(). No functional changes.
2020-12-08 16:36:46 +00:00
struct epoch_tracker et;
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
struct lagg_softc *sc = (struct lagg_softc *)ifp->if_softc;
net: fix set but not used
2018-05-19 05:27:49 +00:00
int error;
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
Add kernel-side support for in-kernel TLS. KTLS adds support for in-kernel framing and encryption of Transport Layer Security (1.0-1.2) data on TCP sockets. KTLS only supports offload of TLS for transmitted data. Key negotation must still be performed in userland. Once completed, transmit session keys for a connection are provided to the kernel via a new TCP_TXTLS_ENABLE socket option. All subsequent data transmitted on the socket is placed into TLS frames and encrypted using the supplied keys. Any data written to a KTLS-enabled socket via write(2), aio_write(2), or sendfile(2) is assumed to be application data and is encoded in TLS frames with an application data type. Individual records can be sent with a custom type (e.g. handshake messages) via sendmsg(2) with a new control message (TLS_SET_RECORD_TYPE) specifying the record type. At present, rekeying is not supported though the in-kernel framework should support rekeying. KTLS makes use of the recently added unmapped mbufs to store TLS frames in the socket buffer. Each TLS frame is described by a single ext_pgs mbuf. The ext_pgs structure contains the header of the TLS record (and trailer for encrypted records) as well as references to the associated TLS session. KTLS supports two primary methods of encrypting TLS frames: software TLS and ifnet TLS. Software TLS marks mbufs holding socket data as not ready via M_NOTREADY similar to sendfile(2) when TLS framing information is added to an unmapped mbuf in ktls_frame(). ktls_enqueue() is then called to schedule TLS frames for encryption. In the case of sendfile_iodone() calls ktls_enqueue() instead of pru_ready() leaving the mbufs marked M_NOTREADY until encryption is completed. For other writes (vn_sendfile when pages are available, write(2), etc.), the PRUS_NOTREADY is set when invoking pru_send() along with invoking ktls_enqueue(). A pool of worker threads (the "KTLS" kernel process) encrypts TLS frames queued via ktls_enqueue(). Each TLS frame is temporarily mapped using the direct map and passed to a software encryption backend to perform the actual encryption. (Note: The use of PHYS_TO_DMAP could be replaced with sf_bufs if someone wished to make this work on architectures without a direct map.) KTLS supports pluggable software encryption backends. Internally, Netflix uses proprietary pure-software backends. This commit includes a simple backend in a new ktls_ocf.ko module that uses the kernel's OpenCrypto framework to provide AES-GCM encryption of TLS frames. As a result, software TLS is now a bit of a misnomer as it can make use of hardware crypto accelerators. Once software encryption has finished, the TLS frame mbufs are marked ready via pru_ready(). At this point, the encrypted data appears as regular payload to the TCP stack stored in unmapped mbufs. ifnet TLS permits a NIC to offload the TLS encryption and TCP segmentation. In this mode, a new send tag type (IF_SND_TAG_TYPE_TLS) is allocated on the interface a socket is routed over and associated with a TLS session. TLS records for a TLS session using ifnet TLS are not marked M_NOTREADY but are passed down the stack unencrypted. The ip_output_send() and ip6_output_send() helper functions that apply send tags to outbound IP packets verify that the send tag of the TLS record matches the outbound interface. If so, the packet is tagged with the TLS send tag and sent to the interface. The NIC device driver must recognize packets with the TLS send tag and schedule them for TLS encryption and TCP segmentation. If the the outbound interface does not match the interface in the TLS send tag, the packet is dropped. In addition, a task is scheduled to refresh the TLS send tag for the TLS session. If a new TLS send tag cannot be allocated, the connection is dropped. If a new TLS send tag is allocated, however, subsequent packets will be tagged with the correct TLS send tag. (This latter case has been tested by configuring both ports of a Chelsio T6 in a lagg and failing over from one port to another. As the connections migrated to the new port, new TLS send tags were allocated for the new port and connections resumed without being dropped.) ifnet TLS can be enabled and disabled on supported network interfaces via new '[-]txtls[46]' options to ifconfig(8). ifnet TLS is supported across both vlan devices and lagg interfaces using failover, lacp with flowid enabled, or lacp with flowid enabled. Applications may request the current KTLS mode of a connection via a new TCP_TXTLS_MODE socket option. They can also use this socket option to toggle between software and ifnet TLS modes. In addition, a testing tool is available in tools/tools/switch_tls. This is modeled on tcpdrop and uses similar syntax. However, instead of dropping connections, -s is used to force KTLS connections to switch to software TLS and -i is used to switch to ifnet TLS. Various sysctls and counters are available under the kern.ipc.tls sysctl node. The kern.ipc.tls.enable node must be set to true to enable KTLS (it is off by default). The use of unmapped mbufs must also be enabled via kern.ipc.mb_use_ext_pgs to enable KTLS. KTLS is enabled via the KERN_TLS kernel option. This patch is the culmination of years of work by several folks including Scott Long and Randall Stewart for the original design and implementation; Drew Gallatin for several optimizations including the use of ext_pgs mbufs, the M_NOTREADY mechanism for TLS records awaiting software encryption, and pluggable software crypto backends; and John Baldwin for modifications to support hardware TLS offload. Reviewed by: gallatin, hselasky, rrs Obtained from: Netflix Sponsored by: Netflix, Chelsio Communications Differential Revision: https://reviews.freebsd.org/D21277
2019-08-27 00:01:56 +00:00
#if defined(KERN_TLS) || defined(RATELIMIT)
Restructure mbuf send tags to provide stronger guarantees. - Perform ifp mismatch checks (to determine if a send tag is allocated for a different ifp than the one the packet is being output on), in ip_output() and ip6_output(). This avoids sending packets with send tags to ifnet drivers that don't support send tags. Since we are now checking for ifp mismatches before invoking if_output, we can now try to allocate a new tag before invoking if_output sending the original packet on the new tag if allocation succeeds. To avoid code duplication for the fragment and unfragmented cases, add ip_output_send() and ip6_output_send() as wrappers around if_output and nd6_output_ifp, respectively. All of the logic for setting send tags and dealing with send tag-related errors is done in these wrapper functions. For pseudo interfaces that wrap other network interfaces (vlan and lagg), wrapper send tags are now allocated so that ip*_output see the wrapper ifp as the ifp in the send tag. The if_transmit routines rewrite the send tags after performing an ifp mismatch check. If an ifp mismatch is detected, the transmit routines fail with EAGAIN. - To provide clearer life cycle management of send tags, especially in the presence of vlan and lagg wrapper tags, add a reference count to send tags managed via m_snd_tag_ref() and m_snd_tag_rele(). Provide a helper function (m_snd_tag_init()) for use by drivers supporting send tags. m_snd_tag_init() takes care of the if_ref on the ifp meaning that code alloating send tags via if_snd_tag_alloc no longer has to manage that manually. Similarly, m_snd_tag_rele drops the refcount on the ifp after invoking if_snd_tag_free when the last reference to a send tag is dropped. This also closes use after free races if there are pending packets in driver tx rings after the socket is closed (e.g. from tcpdrop). In order for m_free to work reliably, add a new CSUM_SND_TAG flag in csum_flags to indicate 'snd_tag' is set (rather than 'rcvif'). Drivers now also check this flag instead of checking snd_tag against NULL. This avoids false positive matches when a forwarded packet has a non-NULL rcvif that was treated as a send tag. - cxgbe was relying on snd_tag_free being called when the inp was detached so that it could kick the firmware to flush any pending work on the flow. This is because the driver doesn't require ACK messages from the firmware for every request, but instead does a kind of manual interrupt coalescing by only setting a flag to request a completion on a subset of requests. If all of the in-flight requests don't have the flag when the tag is detached from the inp, the flow might never return the credits. The current snd_tag_free command issues a flush command to force the credits to return. However, the credit return is what also frees the mbufs, and since those mbufs now hold references on the tag, this meant that snd_tag_free would never be called. To fix, explicitly drop the mbuf's reference on the snd tag when the mbuf is queued in the firmware work queue. This means that once the inp's reference on the tag goes away and all in-flight mbufs have been queued to the firmware, tag's refcount will drop to zero and snd_tag_free will kick in and send the flush request. Note that we need to avoid doing this in the middle of ethofld_tx(), so the driver grabs a temporary reference on the tag around that loop to defer the free to the end of the function in case it sends the last mbuf to the queue after the inp has dropped its reference on the tag. - mlx5 preallocates send tags and was using the ifp pointer even when the send tag wasn't in use. Explicitly use the ifp from other data structures instead. - Sprinkle some assertions in various places to assert that received packets don't have a send tag, and that other places that overwrite rcvif (e.g. 802.11 transmit) don't clobber a send tag pointer. Reviewed by: gallatin, hselasky, rgrimes, ae Sponsored by: Netflix Differential Revision: https://reviews.freebsd.org/D20117
2019-05-24 22:30:40 +00:00
if (m->m_pkthdr.csum_flags & CSUM_SND_TAG)
MPASS(m->m_pkthdr.snd_tag->ifp == ifp);
#endif
Convert LAGG_RLOCK() to NET_EPOCH_ENTER(). No functional changes.
2020-12-08 16:36:46 +00:00
NET_EPOCH_ENTER(et);
Move the protocol and port count checks to outside the loop, these conditions can not change while we have the lock so no point retesting.
2008-09-18 20:56:35 +00:00
/* We need a Tx algorithm and at least one port */
if (sc->sc_proto == LAGG_PROTO_NONE || sc->sc_count == 0) {
Convert LAGG_RLOCK() to NET_EPOCH_ENTER(). No functional changes.
2020-12-08 16:36:46 +00:00
NET_EPOCH_EXIT(et);
Convert lagg(4) to use if_transmit instead of if_start. In collaboration with: thompsa, sbruno, fabient
2012-09-20 10:05:10 +00:00
m_freem(m);
Convert to if_inc_counter() last remnantes of bare access to struct ifnet counters.
2014-09-28 07:43:38 +00:00
if_inc_counter(ifp, IFCOUNTER_OERRORS, 1);
Convert lagg(4) to use if_transmit instead of if_start. In collaboration with: thompsa, sbruno, fabient
2012-09-20 10:05:10 +00:00
return (ENXIO);
Move the protocol and port count checks to outside the loop, these conditions can not change while we have the lock so no point retesting.
2008-09-18 20:56:35 +00:00
}
Convert lagg(4) to use if_transmit instead of if_start. In collaboration with: thompsa, sbruno, fabient
2012-09-20 10:05:10 +00:00
ETHER_BPF_MTAP(ifp, m);
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
Make all lagg protocol methods live in lagg_protos, not in softc. All interfaces of a same protocol, use the same methods. Sponsored by: Netflix Sponsored by: Nginx, Inc.
2014-09-26 12:54:24 +00:00
error = lagg_proto_start(sc, m);
Convert LAGG_RLOCK() to NET_EPOCH_ENTER(). No functional changes.
2020-12-08 16:36:46 +00:00
NET_EPOCH_EXIT(et);
Convert lagg(4) to use if_transmit instead of if_start. In collaboration with: thompsa, sbruno, fabient
2012-09-20 10:05:10 +00:00
return (error);
}
Add support for IP over infiniband, IPoIB, to lagg(4). Currently only the failover protocol is supported due to limitations in the IPoIB architecture. Refer to the lagg(4) manual page for how to configure and use this new feature. A new network interface type, IFT_INFINIBANDLAG, has been added, similar to the existing IFT_IEEE8023ADLAG . ifconfig(8) has been updated to accept a new laggtype argument when creating lagg(4) network interfaces. This new argument is used to distinguish between ethernet and infiniband type of lagg(4) network interface. The laggtype argument is optional and defaults to ethernet. The lagg(4) command line syntax is backwards compatible. Differential Revision: https://reviews.freebsd.org/D26254 Reviewed by: melifaro@ MFC after: 1 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2020-10-22 09:47:12 +00:00
static int
lagg_transmit_infiniband(struct ifnet *ifp, struct mbuf *m)
{
Convert LAGG_RLOCK() to NET_EPOCH_ENTER(). No functional changes.
2020-12-08 16:36:46 +00:00
struct epoch_tracker et;
Add support for IP over infiniband, IPoIB, to lagg(4). Currently only the failover protocol is supported due to limitations in the IPoIB architecture. Refer to the lagg(4) manual page for how to configure and use this new feature. A new network interface type, IFT_INFINIBANDLAG, has been added, similar to the existing IFT_IEEE8023ADLAG . ifconfig(8) has been updated to accept a new laggtype argument when creating lagg(4) network interfaces. This new argument is used to distinguish between ethernet and infiniband type of lagg(4) network interface. The laggtype argument is optional and defaults to ethernet. The lagg(4) command line syntax is backwards compatible. Differential Revision: https://reviews.freebsd.org/D26254 Reviewed by: melifaro@ MFC after: 1 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2020-10-22 09:47:12 +00:00
struct lagg_softc *sc = (struct lagg_softc *)ifp->if_softc;
int error;
#if defined(KERN_TLS) || defined(RATELIMIT)
if (m->m_pkthdr.csum_flags & CSUM_SND_TAG)
MPASS(m->m_pkthdr.snd_tag->ifp == ifp);
#endif
Convert LAGG_RLOCK() to NET_EPOCH_ENTER(). No functional changes.
2020-12-08 16:36:46 +00:00
NET_EPOCH_ENTER(et);
Add support for IP over infiniband, IPoIB, to lagg(4). Currently only the failover protocol is supported due to limitations in the IPoIB architecture. Refer to the lagg(4) manual page for how to configure and use this new feature. A new network interface type, IFT_INFINIBANDLAG, has been added, similar to the existing IFT_IEEE8023ADLAG . ifconfig(8) has been updated to accept a new laggtype argument when creating lagg(4) network interfaces. This new argument is used to distinguish between ethernet and infiniband type of lagg(4) network interface. The laggtype argument is optional and defaults to ethernet. The lagg(4) command line syntax is backwards compatible. Differential Revision: https://reviews.freebsd.org/D26254 Reviewed by: melifaro@ MFC after: 1 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2020-10-22 09:47:12 +00:00
/* We need a Tx algorithm and at least one port */
if (sc->sc_proto == LAGG_PROTO_NONE || sc->sc_count == 0) {
Convert LAGG_RLOCK() to NET_EPOCH_ENTER(). No functional changes.
2020-12-08 16:36:46 +00:00
NET_EPOCH_EXIT(et);
Add support for IP over infiniband, IPoIB, to lagg(4). Currently only the failover protocol is supported due to limitations in the IPoIB architecture. Refer to the lagg(4) manual page for how to configure and use this new feature. A new network interface type, IFT_INFINIBANDLAG, has been added, similar to the existing IFT_IEEE8023ADLAG . ifconfig(8) has been updated to accept a new laggtype argument when creating lagg(4) network interfaces. This new argument is used to distinguish between ethernet and infiniband type of lagg(4) network interface. The laggtype argument is optional and defaults to ethernet. The lagg(4) command line syntax is backwards compatible. Differential Revision: https://reviews.freebsd.org/D26254 Reviewed by: melifaro@ MFC after: 1 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2020-10-22 09:47:12 +00:00
m_freem(m);
if_inc_counter(ifp, IFCOUNTER_OERRORS, 1);
return (ENXIO);
}
INFINIBAND_BPF_MTAP(ifp, m);
error = lagg_proto_start(sc, m);
Convert LAGG_RLOCK() to NET_EPOCH_ENTER(). No functional changes.
2020-12-08 16:36:46 +00:00
NET_EPOCH_EXIT(et);
Add support for IP over infiniband, IPoIB, to lagg(4). Currently only the failover protocol is supported due to limitations in the IPoIB architecture. Refer to the lagg(4) manual page for how to configure and use this new feature. A new network interface type, IFT_INFINIBANDLAG, has been added, similar to the existing IFT_IEEE8023ADLAG . ifconfig(8) has been updated to accept a new laggtype argument when creating lagg(4) network interfaces. This new argument is used to distinguish between ethernet and infiniband type of lagg(4) network interface. The laggtype argument is optional and defaults to ethernet. The lagg(4) command line syntax is backwards compatible. Differential Revision: https://reviews.freebsd.org/D26254 Reviewed by: melifaro@ MFC after: 1 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2020-10-22 09:47:12 +00:00
return (error);
}
Convert lagg(4) to use if_transmit instead of if_start. In collaboration with: thompsa, sbruno, fabient
2012-09-20 10:05:10 +00:00
/*
* The ifp->if_qflush entry point for lagg(4) is no-op.
*/
static void
lagg_qflush(struct ifnet *ifp __unused)
{
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
}
static struct mbuf *
Add support for IP over infiniband, IPoIB, to lagg(4). Currently only the failover protocol is supported due to limitations in the IPoIB architecture. Refer to the lagg(4) manual page for how to configure and use this new feature. A new network interface type, IFT_INFINIBANDLAG, has been added, similar to the existing IFT_IEEE8023ADLAG . ifconfig(8) has been updated to accept a new laggtype argument when creating lagg(4) network interfaces. This new argument is used to distinguish between ethernet and infiniband type of lagg(4) network interface. The laggtype argument is optional and defaults to ethernet. The lagg(4) command line syntax is backwards compatible. Differential Revision: https://reviews.freebsd.org/D26254 Reviewed by: melifaro@ MFC after: 1 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2020-10-22 09:47:12 +00:00
lagg_input_ethernet(struct ifnet *ifp, struct mbuf *m)
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
{
Convert LAGG_RLOCK() to NET_EPOCH_ENTER(). No functional changes.
2020-12-08 16:36:46 +00:00
struct epoch_tracker et;
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
struct lagg_port *lp = ifp->if_lagg;
non-functional cleanup - remove dead code - use consistent variable names - gc unused defines - whitespace cleanup
2007-06-12 07:29:11 +00:00
struct lagg_softc *sc = lp->lp_softc;
struct ifnet *scifp = sc->sc_ifp;
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
Convert LAGG_RLOCK() to NET_EPOCH_ENTER(). No functional changes.
2020-12-08 16:36:46 +00:00
NET_EPOCH_ENTER(et);
non-functional cleanup - remove dead code - use consistent variable names - gc unused defines - whitespace cleanup
2007-06-12 07:29:11 +00:00
if ((scifp->if_drv_flags & IFF_DRV_RUNNING) == 0 ||
Simplify lagg_input(). No functional change intended. MFC after: 2 weeks
2018-05-22 15:35:38 +00:00
lp->lp_detaching != 0 ||
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
sc->sc_proto == LAGG_PROTO_NONE) {
Convert LAGG_RLOCK() to NET_EPOCH_ENTER(). No functional changes.
2020-12-08 16:36:46 +00:00
NET_EPOCH_EXIT(et);
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
m_freem(m);
return (NULL);
}
Use ETHER_BPF_MTAP so that the vlan tags are visible to bpf(4) when stacked under a vlan. MFC after: 3 days
2007-10-20 02:43:23 +00:00
ETHER_BPF_MTAP(scifp, m);
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
Simplify lagg_input(). No functional change intended. MFC after: 2 weeks
2018-05-22 15:35:38 +00:00
m = lagg_proto_input(sc, lp, m);
if (m != NULL && (scifp->if_flags & IFF_MONITOR) != 0) {
Fix a possible mbuf leak on interface departure. Reported by: Alexandre Martins
2015-03-26 23:40:22 +00:00
m_freem(m);
m = NULL;
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
}
Convert LAGG_RLOCK() to NET_EPOCH_ENTER(). No functional changes.
2020-12-08 16:36:46 +00:00
NET_EPOCH_EXIT(et);
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
return (m);
}
Add support for IP over infiniband, IPoIB, to lagg(4). Currently only the failover protocol is supported due to limitations in the IPoIB architecture. Refer to the lagg(4) manual page for how to configure and use this new feature. A new network interface type, IFT_INFINIBANDLAG, has been added, similar to the existing IFT_IEEE8023ADLAG . ifconfig(8) has been updated to accept a new laggtype argument when creating lagg(4) network interfaces. This new argument is used to distinguish between ethernet and infiniband type of lagg(4) network interface. The laggtype argument is optional and defaults to ethernet. The lagg(4) command line syntax is backwards compatible. Differential Revision: https://reviews.freebsd.org/D26254 Reviewed by: melifaro@ MFC after: 1 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2020-10-22 09:47:12 +00:00
static struct mbuf *
lagg_input_infiniband(struct ifnet *ifp, struct mbuf *m)
{
Convert LAGG_RLOCK() to NET_EPOCH_ENTER(). No functional changes.
2020-12-08 16:36:46 +00:00
struct epoch_tracker et;
Add support for IP over infiniband, IPoIB, to lagg(4). Currently only the failover protocol is supported due to limitations in the IPoIB architecture. Refer to the lagg(4) manual page for how to configure and use this new feature. A new network interface type, IFT_INFINIBANDLAG, has been added, similar to the existing IFT_IEEE8023ADLAG . ifconfig(8) has been updated to accept a new laggtype argument when creating lagg(4) network interfaces. This new argument is used to distinguish between ethernet and infiniband type of lagg(4) network interface. The laggtype argument is optional and defaults to ethernet. The lagg(4) command line syntax is backwards compatible. Differential Revision: https://reviews.freebsd.org/D26254 Reviewed by: melifaro@ MFC after: 1 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2020-10-22 09:47:12 +00:00
struct lagg_port *lp = ifp->if_lagg;
struct lagg_softc *sc = lp->lp_softc;
struct ifnet *scifp = sc->sc_ifp;
Convert LAGG_RLOCK() to NET_EPOCH_ENTER(). No functional changes.
2020-12-08 16:36:46 +00:00
NET_EPOCH_ENTER(et);
Add support for IP over infiniband, IPoIB, to lagg(4). Currently only the failover protocol is supported due to limitations in the IPoIB architecture. Refer to the lagg(4) manual page for how to configure and use this new feature. A new network interface type, IFT_INFINIBANDLAG, has been added, similar to the existing IFT_IEEE8023ADLAG . ifconfig(8) has been updated to accept a new laggtype argument when creating lagg(4) network interfaces. This new argument is used to distinguish between ethernet and infiniband type of lagg(4) network interface. The laggtype argument is optional and defaults to ethernet. The lagg(4) command line syntax is backwards compatible. Differential Revision: https://reviews.freebsd.org/D26254 Reviewed by: melifaro@ MFC after: 1 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2020-10-22 09:47:12 +00:00
if ((scifp->if_drv_flags & IFF_DRV_RUNNING) == 0 ||
lp->lp_detaching != 0 ||
sc->sc_proto == LAGG_PROTO_NONE) {
Convert LAGG_RLOCK() to NET_EPOCH_ENTER(). No functional changes.
2020-12-08 16:36:46 +00:00
NET_EPOCH_EXIT(et);
Add support for IP over infiniband, IPoIB, to lagg(4). Currently only the failover protocol is supported due to limitations in the IPoIB architecture. Refer to the lagg(4) manual page for how to configure and use this new feature. A new network interface type, IFT_INFINIBANDLAG, has been added, similar to the existing IFT_IEEE8023ADLAG . ifconfig(8) has been updated to accept a new laggtype argument when creating lagg(4) network interfaces. This new argument is used to distinguish between ethernet and infiniband type of lagg(4) network interface. The laggtype argument is optional and defaults to ethernet. The lagg(4) command line syntax is backwards compatible. Differential Revision: https://reviews.freebsd.org/D26254 Reviewed by: melifaro@ MFC after: 1 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2020-10-22 09:47:12 +00:00
m_freem(m);
return (NULL);
}
INFINIBAND_BPF_MTAP(scifp, m);
m = lagg_proto_input(sc, lp, m);
if (m != NULL && (scifp->if_flags & IFF_MONITOR) != 0) {
m_freem(m);
m = NULL;
}
Convert LAGG_RLOCK() to NET_EPOCH_ENTER(). No functional changes.
2020-12-08 16:36:46 +00:00
NET_EPOCH_EXIT(et);
Add support for IP over infiniband, IPoIB, to lagg(4). Currently only the failover protocol is supported due to limitations in the IPoIB architecture. Refer to the lagg(4) manual page for how to configure and use this new feature. A new network interface type, IFT_INFINIBANDLAG, has been added, similar to the existing IFT_IEEE8023ADLAG . ifconfig(8) has been updated to accept a new laggtype argument when creating lagg(4) network interfaces. This new argument is used to distinguish between ethernet and infiniband type of lagg(4) network interface. The laggtype argument is optional and defaults to ethernet. The lagg(4) command line syntax is backwards compatible. Differential Revision: https://reviews.freebsd.org/D26254 Reviewed by: melifaro@ MFC after: 1 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2020-10-22 09:47:12 +00:00
return (m);
}
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
static int
lagg_media_change(struct ifnet *ifp)
{
struct lagg_softc *sc = (struct lagg_softc *)ifp->if_softc;
if (sc->sc_ifflags & IFF_DEBUG)
printf("%s\n", __func__);
/* Ignore */
return (0);
}
static void
lagg_media_status(struct ifnet *ifp, struct ifmediareq *imr)
{
Convert LAGG_RLOCK() to NET_EPOCH_ENTER(). No functional changes.
2020-12-08 16:36:46 +00:00
struct epoch_tracker et;
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
struct lagg_softc *sc = (struct lagg_softc *)ifp->if_softc;
struct lagg_port *lp;
imr->ifm_status = IFM_AVALID;
imr->ifm_active = IFM_ETHER | IFM_AUTO;
Convert LAGG_RLOCK() to NET_EPOCH_ENTER(). No functional changes.
2020-12-08 16:36:46 +00:00
NET_EPOCH_ENTER(et);
Replace rmlock with epoch in lagg Use the new epoch based reclamation API. Now the hot paths will not block at all, and the sx lock is used for the softc data. This fixes LORs reported where the rwlock was obtained when the sxlock was held. Submitted by: mmacy Reported by: Harry Schmalzbauer <freebsd@omnilan.de> Reviewed by: sbruno Sponsored by: Limelight Networks Differential Revision: https://reviews.freebsd.org/D15355
2018-05-14 20:06:49 +00:00
CK_SLIST_FOREACH(lp, &sc->sc_ports, lp_entries) {
- Correctly check if lp_ioctl is null - Remove lagg_ether_purgemulti as its no longer needed - Mark the interface as up if any ports are active rather than just the primary
2007-05-07 09:53:02 +00:00
if (LAGG_PORTACTIVE(lp))
imr->ifm_status |= IFM_ACTIVE;
}
Convert LAGG_RLOCK() to NET_EPOCH_ENTER(). No functional changes.
2020-12-08 16:36:46 +00:00
NET_EPOCH_EXIT(et);
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
}
Have the lagg interface generate link up/down events, the interface is marked as up if at least one of its ports also has a link up. This fixes using carp+lagg together and any other system that relies on linkstate events. PR: kern/113956 MFC after: 3 days
2007-11-25 06:30:46 +00:00
static void
lagg_linkstate(struct lagg_softc *sc)
{
Convert LAGG_RLOCK() to NET_EPOCH_ENTER(). No functional changes.
2020-12-08 16:36:46 +00:00
struct epoch_tracker et;
Have the lagg interface generate link up/down events, the interface is marked as up if at least one of its ports also has a link up. This fixes using carp+lagg together and any other system that relies on linkstate events. PR: kern/113956 MFC after: 3 days
2007-11-25 06:30:46 +00:00
struct lagg_port *lp;
int new_link = LINK_STATE_DOWN;
- Protect against sc->sc_primary being null - Initialise speed where its used
2008-12-17 21:04:43 +00:00
uint64_t speed;
Have the lagg interface generate link up/down events, the interface is marked as up if at least one of its ports also has a link up. This fixes using carp+lagg together and any other system that relies on linkstate events. PR: kern/113956 MFC after: 3 days
2007-11-25 06:30:46 +00:00
Introduce sleepable locks into if_lagg. Before this change if_lagg was using nonsleepable rmlocks to protect its internal state. This patch introduces another sx lock to protect code paths that require sleeping, while still uses old rmlock to protect hot nonsleepable data paths. This change allows to remove taskqueue decoupling used before to change interface addresses without holding the lock. Instead it uses sx lock to protect direct if_ioctl() calls. As another bonus, the new code synchronizes enabled capabilities of member interfaces, and allows to control them with ifconfig laggX, that was impossible before. This part should fix interoperation with if_bridge, that may need to disable some capabilities, such as TXCSUM or LRO, to allow bridging with noncapable interfaces. MFC after: 2 weeks Sponsored by: iXsystems, Inc. Differential Revision: https://reviews.freebsd.org/D10514
2017-05-02 19:09:11 +00:00
LAGG_XLOCK_ASSERT(sc);
lagg: allow lacp to manage the link state Lacp needs to manage the link state itself. Unlike other lagg protocols, the ability of lacp to pass traffic depends not only on the lagg members having link, but also on the lacp protocol converging to a distributing state with the link partner. If we prematurely mark the link as up, then we will send a gratuitous arp (via arp_handle_ifllchange()) before the lacp interface is capable of passing traffic. When this happens, the gratuitous arp is lost, and our link partner may cache a stale mac address (eg, when the base mac address for the lagg bundle changes, due to a BIOS change re-ordering NIC unit numbers) Reviewed by: jtl, hselasky Sponsored by: Netflix
2018-08-13 14:13:25 +00:00
/* LACP handles link state itself */
if (sc->sc_proto == LAGG_PROTO_LACP)
return;
Have the lagg interface generate link up/down events, the interface is marked as up if at least one of its ports also has a link up. This fixes using carp+lagg together and any other system that relies on linkstate events. PR: kern/113956 MFC after: 3 days
2007-11-25 06:30:46 +00:00
/* Our link is considered up if at least one of our ports is active */
Convert LAGG_RLOCK() to NET_EPOCH_ENTER(). No functional changes.
2020-12-08 16:36:46 +00:00
NET_EPOCH_ENTER(et);
Replace rmlock with epoch in lagg Use the new epoch based reclamation API. Now the hot paths will not block at all, and the sx lock is used for the softc data. This fixes LORs reported where the rwlock was obtained when the sxlock was held. Submitted by: mmacy Reported by: Harry Schmalzbauer <freebsd@omnilan.de> Reviewed by: sbruno Sponsored by: Limelight Networks Differential Revision: https://reviews.freebsd.org/D15355
2018-05-14 20:06:49 +00:00
CK_SLIST_FOREACH(lp, &sc->sc_ports, lp_entries) {
Remove macros that hide access to struct ifnet fields.
2014-09-26 13:02:29 +00:00
if (lp->lp_ifp->if_link_state == LINK_STATE_UP) {
Have the lagg interface generate link up/down events, the interface is marked as up if at least one of its ports also has a link up. This fixes using carp+lagg together and any other system that relies on linkstate events. PR: kern/113956 MFC after: 3 days
2007-11-25 06:30:46 +00:00
new_link = LINK_STATE_UP;
break;
}
}
Convert LAGG_RLOCK() to NET_EPOCH_ENTER(). No functional changes.
2020-12-08 16:36:46 +00:00
NET_EPOCH_EXIT(et);
Revert r292275 & r292379 glebius has concerns about these changes so reverting those can be discussed and addressed. Sponsored by: Multiplay
2015-12-17 14:41:30 +00:00
if_link_state_change(sc->sc_ifp, new_link);
Update the interface baudrate taking into account the max speed for the different aggregation protocols.
2008-12-17 20:58:10 +00:00
/* Update if_baudrate to reflect the max possible speed */
switch (sc->sc_proto) {
case LAGG_PROTO_FAILOVER:
- Protect against sc->sc_primary being null - Initialise speed where its used
2008-12-17 21:04:43 +00:00
sc->sc_ifp->if_baudrate = sc->sc_primary != NULL ?
sc->sc_primary->lp_ifp->if_baudrate : 0;
Update the interface baudrate taking into account the max speed for the different aggregation protocols.
2008-12-17 20:58:10 +00:00
break;
case LAGG_PROTO_ROUNDROBIN:
case LAGG_PROTO_LOADBALANCE:
- Whitespace. - Remove caddr_t.
2014-09-26 12:35:58 +00:00
case LAGG_PROTO_BROADCAST:
- Protect against sc->sc_primary being null - Initialise speed where its used
2008-12-17 21:04:43 +00:00
speed = 0;
Convert LAGG_RLOCK() to NET_EPOCH_ENTER(). No functional changes.
2020-12-08 16:36:46 +00:00
NET_EPOCH_ENTER(et);
Replace rmlock with epoch in lagg Use the new epoch based reclamation API. Now the hot paths will not block at all, and the sx lock is used for the softc data. This fixes LORs reported where the rwlock was obtained when the sxlock was held. Submitted by: mmacy Reported by: Harry Schmalzbauer <freebsd@omnilan.de> Reviewed by: sbruno Sponsored by: Limelight Networks Differential Revision: https://reviews.freebsd.org/D15355
2018-05-14 20:06:49 +00:00
CK_SLIST_FOREACH(lp, &sc->sc_ports, lp_entries)
Update the interface baudrate taking into account the max speed for the different aggregation protocols.
2008-12-17 20:58:10 +00:00
speed += lp->lp_ifp->if_baudrate;
Convert LAGG_RLOCK() to NET_EPOCH_ENTER(). No functional changes.
2020-12-08 16:36:46 +00:00
NET_EPOCH_EXIT(et);
Update the interface baudrate taking into account the max speed for the different aggregation protocols.
2008-12-17 20:58:10 +00:00
sc->sc_ifp->if_baudrate = speed;
break;
case LAGG_PROTO_LACP:
/* LACP updates if_baudrate itself */
break;
}
Have the lagg interface generate link up/down events, the interface is marked as up if at least one of its ports also has a link up. This fixes using carp+lagg together and any other system that relies on linkstate events. PR: kern/113956 MFC after: 3 days
2007-11-25 06:30:46 +00:00
}
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
static void
Revert r292275 & r292379 glebius has concerns about these changes so reverting those can be discussed and addressed. Sponsored by: Multiplay
2015-12-17 14:41:30 +00:00
lagg_port_state(struct ifnet *ifp, int state)
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
{
struct lagg_port *lp = (struct lagg_port *)ifp->if_lagg;
struct lagg_softc *sc = NULL;
if (lp != NULL)
non-functional cleanup - remove dead code - use consistent variable names - gc unused defines - whitespace cleanup
2007-06-12 07:29:11 +00:00
sc = lp->lp_softc;
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
if (sc == NULL)
return;
Introduce sleepable locks into if_lagg. Before this change if_lagg was using nonsleepable rmlocks to protect its internal state. This patch introduces another sx lock to protect code paths that require sleeping, while still uses old rmlock to protect hot nonsleepable data paths. This change allows to remove taskqueue decoupling used before to change interface addresses without holding the lock. Instead it uses sx lock to protect direct if_ioctl() calls. As another bonus, the new code synchronizes enabled capabilities of member interfaces, and allows to control them with ifconfig laggX, that was impossible before. This part should fix interoperation with if_bridge, that may need to disable some capabilities, such as TXCSUM or LRO, to allow bridging with noncapable interfaces. MFC after: 2 weeks Sponsored by: iXsystems, Inc. Differential Revision: https://reviews.freebsd.org/D10514
2017-05-02 19:09:11 +00:00
LAGG_XLOCK(sc);
Have the lagg interface generate link up/down events, the interface is marked as up if at least one of its ports also has a link up. This fixes using carp+lagg together and any other system that relies on linkstate events. PR: kern/113956 MFC after: 3 days
2007-11-25 06:30:46 +00:00
lagg_linkstate(sc);
Make all lagg protocol methods live in lagg_protos, not in softc. All interfaces of a same protocol, use the same methods. Sponsored by: Netflix Sponsored by: Nginx, Inc.
2014-09-26 12:54:24 +00:00
lagg_proto_linkstate(sc, lp);
Introduce sleepable locks into if_lagg. Before this change if_lagg was using nonsleepable rmlocks to protect its internal state. This patch introduces another sx lock to protect code paths that require sleeping, while still uses old rmlock to protect hot nonsleepable data paths. This change allows to remove taskqueue decoupling used before to change interface addresses without holding the lock. Instead it uses sx lock to protect direct if_ioctl() calls. As another bonus, the new code synchronizes enabled capabilities of member interfaces, and allows to control them with ifconfig laggX, that was impossible before. This part should fix interoperation with if_bridge, that may need to disable some capabilities, such as TXCSUM or LRO, to allow bridging with noncapable interfaces. MFC after: 2 weeks Sponsored by: iXsystems, Inc. Differential Revision: https://reviews.freebsd.org/D10514
2017-05-02 19:09:11 +00:00
LAGG_XUNLOCK(sc);
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
}
Revert r292275 & r292379 glebius has concerns about these changes so reverting those can be discussed and addressed. Sponsored by: Multiplay
2015-12-17 14:41:30 +00:00
struct lagg_port *
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
lagg_link_active(struct lagg_softc *sc, struct lagg_port *lp)
{
struct lagg_port *lp_next, *rval = NULL;
/*
* Search a port which reports an active link state.
*/
Conditionalize an in_epoch() call on INVARIANTS. Its result is only used to determine whether to perform further INVARIANTS-only checks. Remove a stale comment while here. Submitted by: Sebastian Huber <sebastian.huber@embedded-brains.de> MFC after: 1 week
2019-06-06 16:22:29 +00:00
#ifdef INVARIANTS
Restore the comment removed in r348745. LAGG_RLOCK() enters an epoch section, so the comment wasn't stale. Reported by: jhb MFC with: r348745
2019-06-06 17:20:35 +00:00
/*
Convert LAGG_RLOCK() to NET_EPOCH_ENTER(). No functional changes.
2020-12-08 16:36:46 +00:00
* This is called with either in the network epoch
* or with LAGG_XLOCK(sc) held.
Restore the comment removed in r348745. LAGG_RLOCK() enters an epoch section, so the comment wasn't stale. Reported by: jhb MFC with: r348745
2019-06-06 17:20:35 +00:00
*/
Remove nested epochs from lagg(4). lagg_bcast_start appeared to have a bug in that was using the last lagg port structure after exiting the epoch that was keeping that structure alive. However, upon further inspection, the epoch was already entered by the caller (lagg_transmit), so the epoch enter/exit in lagg_bcast_start was actually unnecessary. This commit generally removes uses of the net epoch via LAGG_RLOCK to protect the list of ports when the list of ports was already protected by an existing LAGG_RLOCK in a caller, or the LAGG_XLOCK. It also adds a missing epoch enter/exit in lagg_snd_tag_alloc while accessing the lagg port structures. An ifp is still accessed via an unsafe reference after the epoch is exited, but that is true in the current code and will be fixed in a future change. Reviewed by: gallatin MFC after: 1 month Sponsored by: Netflix Differential Revision: https://reviews.freebsd.org/D19718
2019-03-28 20:25:36 +00:00
if (!in_epoch(net_epoch_preempt))
LAGG_XLOCK_ASSERT(sc);
Conditionalize an in_epoch() call on INVARIANTS. Its result is only used to determine whether to perform further INVARIANTS-only checks. Remove a stale comment while here. Submitted by: Sebastian Huber <sebastian.huber@embedded-brains.de> MFC after: 1 week
2019-06-06 16:22:29 +00:00
#endif
Remove nested epochs from lagg(4). lagg_bcast_start appeared to have a bug in that was using the last lagg port structure after exiting the epoch that was keeping that structure alive. However, upon further inspection, the epoch was already entered by the caller (lagg_transmit), so the epoch enter/exit in lagg_bcast_start was actually unnecessary. This commit generally removes uses of the net epoch via LAGG_RLOCK to protect the list of ports when the list of ports was already protected by an existing LAGG_RLOCK in a caller, or the LAGG_XLOCK. It also adds a missing epoch enter/exit in lagg_snd_tag_alloc while accessing the lagg port structures. An ifp is still accessed via an unsafe reference after the epoch is exited, but that is true in the current code and will be fixed in a future change. Reviewed by: gallatin MFC after: 1 month Sponsored by: Netflix Differential Revision: https://reviews.freebsd.org/D19718
2019-03-28 20:25:36 +00:00
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
if (lp == NULL)
goto search;
if (LAGG_PORTACTIVE(lp)) {
rval = lp;
goto found;
}
CK: update consumers to use CK macros across the board r334189 changed the fields to have names distinct from those in queue.h in order to expose the oversights as compile time errors
2018-05-24 23:21:23 +00:00
if ((lp_next = CK_SLIST_NEXT(lp, lp_entries)) != NULL &&
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
LAGG_PORTACTIVE(lp_next)) {
rval = lp_next;
goto found;
}
Remove nested epochs from lagg(4). lagg_bcast_start appeared to have a bug in that was using the last lagg port structure after exiting the epoch that was keeping that structure alive. However, upon further inspection, the epoch was already entered by the caller (lagg_transmit), so the epoch enter/exit in lagg_bcast_start was actually unnecessary. This commit generally removes uses of the net epoch via LAGG_RLOCK to protect the list of ports when the list of ports was already protected by an existing LAGG_RLOCK in a caller, or the LAGG_XLOCK. It also adds a missing epoch enter/exit in lagg_snd_tag_alloc while accessing the lagg port structures. An ifp is still accessed via an unsafe reference after the epoch is exited, but that is true in the current code and will be fixed in a future change. Reviewed by: gallatin MFC after: 1 month Sponsored by: Netflix Differential Revision: https://reviews.freebsd.org/D19718
2019-03-28 20:25:36 +00:00
search:
Replace rmlock with epoch in lagg Use the new epoch based reclamation API. Now the hot paths will not block at all, and the sx lock is used for the softc data. This fixes LORs reported where the rwlock was obtained when the sxlock was held. Submitted by: mmacy Reported by: Harry Schmalzbauer <freebsd@omnilan.de> Reviewed by: sbruno Sponsored by: Limelight Networks Differential Revision: https://reviews.freebsd.org/D15355
2018-05-14 20:06:49 +00:00
CK_SLIST_FOREACH(lp_next, &sc->sc_ports, lp_entries) {
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
if (LAGG_PORTACTIVE(lp_next)) {
epoch(9): allow preemptible epochs to compose - Add tracker argument to preemptible epochs - Inline epoch read path in kernel and tied modules - Change in_epoch to take an epoch as argument - Simplify tfb_tcp_do_segment to not take a ti_locked argument, there's no longer any benefit to dropping the pcbinfo lock and trying to do so just adds an error prone branchfest to these functions - Remove cases of same function recursion on the epoch as recursing is no longer free. - Remove the the TAILQ_ENTRY and epoch_section from struct thread as the tracker field is now stack or heap allocated as appropriate. Tested by: pho and Limelight Networks Reviewed by: kbowling at llnw dot com Sponsored by: Limelight Networks Differential Revision: https://reviews.freebsd.org/D16066
2018-07-04 02:47:16 +00:00
return (lp_next);
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
}
}
found:
return (rval);
}
int
lagg_enqueue(struct ifnet *ifp, struct mbuf *m)
{
Add kernel-side support for in-kernel TLS. KTLS adds support for in-kernel framing and encryption of Transport Layer Security (1.0-1.2) data on TCP sockets. KTLS only supports offload of TLS for transmitted data. Key negotation must still be performed in userland. Once completed, transmit session keys for a connection are provided to the kernel via a new TCP_TXTLS_ENABLE socket option. All subsequent data transmitted on the socket is placed into TLS frames and encrypted using the supplied keys. Any data written to a KTLS-enabled socket via write(2), aio_write(2), or sendfile(2) is assumed to be application data and is encoded in TLS frames with an application data type. Individual records can be sent with a custom type (e.g. handshake messages) via sendmsg(2) with a new control message (TLS_SET_RECORD_TYPE) specifying the record type. At present, rekeying is not supported though the in-kernel framework should support rekeying. KTLS makes use of the recently added unmapped mbufs to store TLS frames in the socket buffer. Each TLS frame is described by a single ext_pgs mbuf. The ext_pgs structure contains the header of the TLS record (and trailer for encrypted records) as well as references to the associated TLS session. KTLS supports two primary methods of encrypting TLS frames: software TLS and ifnet TLS. Software TLS marks mbufs holding socket data as not ready via M_NOTREADY similar to sendfile(2) when TLS framing information is added to an unmapped mbuf in ktls_frame(). ktls_enqueue() is then called to schedule TLS frames for encryption. In the case of sendfile_iodone() calls ktls_enqueue() instead of pru_ready() leaving the mbufs marked M_NOTREADY until encryption is completed. For other writes (vn_sendfile when pages are available, write(2), etc.), the PRUS_NOTREADY is set when invoking pru_send() along with invoking ktls_enqueue(). A pool of worker threads (the "KTLS" kernel process) encrypts TLS frames queued via ktls_enqueue(). Each TLS frame is temporarily mapped using the direct map and passed to a software encryption backend to perform the actual encryption. (Note: The use of PHYS_TO_DMAP could be replaced with sf_bufs if someone wished to make this work on architectures without a direct map.) KTLS supports pluggable software encryption backends. Internally, Netflix uses proprietary pure-software backends. This commit includes a simple backend in a new ktls_ocf.ko module that uses the kernel's OpenCrypto framework to provide AES-GCM encryption of TLS frames. As a result, software TLS is now a bit of a misnomer as it can make use of hardware crypto accelerators. Once software encryption has finished, the TLS frame mbufs are marked ready via pru_ready(). At this point, the encrypted data appears as regular payload to the TCP stack stored in unmapped mbufs. ifnet TLS permits a NIC to offload the TLS encryption and TCP segmentation. In this mode, a new send tag type (IF_SND_TAG_TYPE_TLS) is allocated on the interface a socket is routed over and associated with a TLS session. TLS records for a TLS session using ifnet TLS are not marked M_NOTREADY but are passed down the stack unencrypted. The ip_output_send() and ip6_output_send() helper functions that apply send tags to outbound IP packets verify that the send tag of the TLS record matches the outbound interface. If so, the packet is tagged with the TLS send tag and sent to the interface. The NIC device driver must recognize packets with the TLS send tag and schedule them for TLS encryption and TCP segmentation. If the the outbound interface does not match the interface in the TLS send tag, the packet is dropped. In addition, a task is scheduled to refresh the TLS send tag for the TLS session. If a new TLS send tag cannot be allocated, the connection is dropped. If a new TLS send tag is allocated, however, subsequent packets will be tagged with the correct TLS send tag. (This latter case has been tested by configuring both ports of a Chelsio T6 in a lagg and failing over from one port to another. As the connections migrated to the new port, new TLS send tags were allocated for the new port and connections resumed without being dropped.) ifnet TLS can be enabled and disabled on supported network interfaces via new '[-]txtls[46]' options to ifconfig(8). ifnet TLS is supported across both vlan devices and lagg interfaces using failover, lacp with flowid enabled, or lacp with flowid enabled. Applications may request the current KTLS mode of a connection via a new TCP_TXTLS_MODE socket option. They can also use this socket option to toggle between software and ifnet TLS modes. In addition, a testing tool is available in tools/tools/switch_tls. This is modeled on tcpdrop and uses similar syntax. However, instead of dropping connections, -s is used to force KTLS connections to switch to software TLS and -i is used to switch to ifnet TLS. Various sysctls and counters are available under the kern.ipc.tls sysctl node. The kern.ipc.tls.enable node must be set to true to enable KTLS (it is off by default). The use of unmapped mbufs must also be enabled via kern.ipc.mb_use_ext_pgs to enable KTLS. KTLS is enabled via the KERN_TLS kernel option. This patch is the culmination of years of work by several folks including Scott Long and Randall Stewart for the original design and implementation; Drew Gallatin for several optimizations including the use of ext_pgs mbufs, the M_NOTREADY mechanism for TLS records awaiting software encryption, and pluggable software crypto backends; and John Baldwin for modifications to support hardware TLS offload. Reviewed by: gallatin, hselasky, rrs Obtained from: Netflix Sponsored by: Netflix, Chelsio Communications Differential Revision: https://reviews.freebsd.org/D21277
2019-08-27 00:01:56 +00:00
#if defined(KERN_TLS) || defined(RATELIMIT)
Restructure mbuf send tags to provide stronger guarantees. - Perform ifp mismatch checks (to determine if a send tag is allocated for a different ifp than the one the packet is being output on), in ip_output() and ip6_output(). This avoids sending packets with send tags to ifnet drivers that don't support send tags. Since we are now checking for ifp mismatches before invoking if_output, we can now try to allocate a new tag before invoking if_output sending the original packet on the new tag if allocation succeeds. To avoid code duplication for the fragment and unfragmented cases, add ip_output_send() and ip6_output_send() as wrappers around if_output and nd6_output_ifp, respectively. All of the logic for setting send tags and dealing with send tag-related errors is done in these wrapper functions. For pseudo interfaces that wrap other network interfaces (vlan and lagg), wrapper send tags are now allocated so that ip*_output see the wrapper ifp as the ifp in the send tag. The if_transmit routines rewrite the send tags after performing an ifp mismatch check. If an ifp mismatch is detected, the transmit routines fail with EAGAIN. - To provide clearer life cycle management of send tags, especially in the presence of vlan and lagg wrapper tags, add a reference count to send tags managed via m_snd_tag_ref() and m_snd_tag_rele(). Provide a helper function (m_snd_tag_init()) for use by drivers supporting send tags. m_snd_tag_init() takes care of the if_ref on the ifp meaning that code alloating send tags via if_snd_tag_alloc no longer has to manage that manually. Similarly, m_snd_tag_rele drops the refcount on the ifp after invoking if_snd_tag_free when the last reference to a send tag is dropped. This also closes use after free races if there are pending packets in driver tx rings after the socket is closed (e.g. from tcpdrop). In order for m_free to work reliably, add a new CSUM_SND_TAG flag in csum_flags to indicate 'snd_tag' is set (rather than 'rcvif'). Drivers now also check this flag instead of checking snd_tag against NULL. This avoids false positive matches when a forwarded packet has a non-NULL rcvif that was treated as a send tag. - cxgbe was relying on snd_tag_free being called when the inp was detached so that it could kick the firmware to flush any pending work on the flow. This is because the driver doesn't require ACK messages from the firmware for every request, but instead does a kind of manual interrupt coalescing by only setting a flag to request a completion on a subset of requests. If all of the in-flight requests don't have the flag when the tag is detached from the inp, the flow might never return the credits. The current snd_tag_free command issues a flush command to force the credits to return. However, the credit return is what also frees the mbufs, and since those mbufs now hold references on the tag, this meant that snd_tag_free would never be called. To fix, explicitly drop the mbuf's reference on the snd tag when the mbuf is queued in the firmware work queue. This means that once the inp's reference on the tag goes away and all in-flight mbufs have been queued to the firmware, tag's refcount will drop to zero and snd_tag_free will kick in and send the flush request. Note that we need to avoid doing this in the middle of ethofld_tx(), so the driver grabs a temporary reference on the tag around that loop to defer the free to the end of the function in case it sends the last mbuf to the queue after the inp has dropped its reference on the tag. - mlx5 preallocates send tags and was using the ifp pointer even when the send tag wasn't in use. Explicitly use the ifp from other data structures instead. - Sprinkle some assertions in various places to assert that received packets don't have a send tag, and that other places that overwrite rcvif (e.g. 802.11 transmit) don't clobber a send tag pointer. Reviewed by: gallatin, hselasky, rgrimes, ae Sponsored by: Netflix Differential Revision: https://reviews.freebsd.org/D20117
2019-05-24 22:30:40 +00:00
if (m->m_pkthdr.csum_flags & CSUM_SND_TAG) {
struct lagg_snd_tag *lst;
struct m_snd_tag *mst;
mst = m->m_pkthdr.snd_tag;
lst = mst_to_lst(mst);
if (lst->tag->ifp != ifp) {
m_freem(m);
return (EAGAIN);
}
m->m_pkthdr.snd_tag = m_snd_tag_ref(lst->tag);
m_snd_tag_rele(mst);
}
#endif
convert calls to IFQ_HANDOFF to if_transmit
2008-11-22 07:35:45 +00:00
return (ifp->if_transmit)(ifp, m);
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
}
/*
* Simple round robin aggregation
*/
- When reconfiguring protocol on a lagg, first set it to LAGG_PROTO_NONE, then drop lock, run the attach routines, and then set it to specific proto. This removes tons of WITNESS warnings. - Make lagg protocol attach handlers not failing and allocate memory with M_WAITOK. Sponsored by: Netflix Sponsored by: Nginx, Inc.
2014-09-26 08:42:32 +00:00
static void
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
lagg_rr_attach(struct lagg_softc *sc)
{
Fix two panics in lagg. 1. The locking was changed to shared but roundrobin mode still updated a pointer in the softc with the next tx interface to use. This will panic under high load. Change this to an atomically incremented sequence number in order to choose the tx port in round robin. 2. IFQ_HANDOFF will free the mbuf if the queue is full, this will then be freed again by lagg_start() and panic. Reorganised the error handling and freeing to fix this. MFC after: 3 days
2007-10-12 03:03:16 +00:00
sc->sc_seq = 0;
lagg: Further cleanup of the rr_limit option. Add an option flag so that arbitrary updates to a lagg's configuration do not clear sc_stride. Preseve compatibility for old ifconfig binaries. Update ifconfig to use the new flag and improve the casting used when parsing the option parameter. Modify the RR transmit function to avoid locklessly reading sc_stride twice. Ensure that sc_stride is always 1 or greater. Reviewed by: hselasky MFC after: 1 week Sponsored by: The FreeBSD Foundation Differential Revision: https://reviews.freebsd.org/D23092
2020-01-09 14:58:41 +00:00
sc->sc_stride = 1;
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
}
static int
lagg_rr_start(struct lagg_softc *sc, struct mbuf *m)
{
Fix two panics in lagg. 1. The locking was changed to shared but roundrobin mode still updated a pointer in the softc with the next tx interface to use. This will panic under high load. Change this to an atomically incremented sequence number in order to choose the tx port in round robin. 2. IFQ_HANDOFF will free the mbuf if the queue is full, this will then be freed again by lagg_start() and panic. Reorganised the error handling and freeing to fix this. MFC after: 3 days
2007-10-12 03:03:16 +00:00
struct lagg_port *lp;
uint32_t p;
lagg: Clean up handling of the rr_limit option. - Don't allow an unprivileged user to set the stride. [1] - Only set the stride under the softc lock. - Rename the internal fields to accurately reflect their use. Keep ro_bkt to avoid changing the user API. - Simplify the implementation. The port index is just sc_seq / stride. - Document rr_limit in ifconfig.8. Reported by: Ilja Van Sprundel <ivansprundel@ioactive.com> [1] MFC after: 1 week Sponsored by: The FreeBSD Foundation Differential Revision: https://reviews.freebsd.org/D22857
2019-12-22 21:56:47 +00:00
p = atomic_fetchadd_32(&sc->sc_seq, 1);
lagg: Further cleanup of the rr_limit option. Add an option flag so that arbitrary updates to a lagg's configuration do not clear sc_stride. Preseve compatibility for old ifconfig binaries. Update ifconfig to use the new flag and improve the casting used when parsing the option parameter. Modify the RR transmit function to avoid locklessly reading sc_stride twice. Ensure that sc_stride is always 1 or greater. Reviewed by: hselasky MFC after: 1 week Sponsored by: The FreeBSD Foundation Differential Revision: https://reviews.freebsd.org/D23092
2020-01-09 14:58:41 +00:00
p /= sc->sc_stride;
Fix two panics in lagg. 1. The locking was changed to shared but roundrobin mode still updated a pointer in the softc with the next tx interface to use. This will panic under high load. Change this to an atomically incremented sequence number in order to choose the tx port in round robin. 2. IFQ_HANDOFF will free the mbuf if the queue is full, this will then be freed again by lagg_start() and panic. Reorganised the error handling and freeing to fix this. MFC after: 3 days
2007-10-12 03:03:16 +00:00
p %= sc->sc_count;
CK: update consumers to use CK macros across the board r334189 changed the fields to have names distinct from those in queue.h in order to expose the oversights as compile time errors
2018-05-24 23:21:23 +00:00
lp = CK_SLIST_FIRST(&sc->sc_ports);
Add an IOCTL rr_limit to let users fine tuning the number of packets to be sent using roundrobin protocol and set a better granularity and distribution among the interfaces. Tuning the number of packages sent by interface can increase throughput and reduce unordered packets as well as reduce SACK. Example of usage: # ifconfig bge0 up # ifconfig bge1 up # ifconfig lagg0 create # ifconfig lagg0 laggproto roundrobin laggport bge0 laggport bge1 \ 192.168.1.1 netmask 255.255.255.0 # ifconfig lagg0 rr_limit 500 Reviewed by: thompsa, glebius, adrian (old patch) Approved by: bapt (mentor) Relnotes: Yes Differential Revision: https://reviews.freebsd.org/D540
2016-01-23 04:18:44 +00:00
Fix two panics in lagg. 1. The locking was changed to shared but roundrobin mode still updated a pointer in the softc with the next tx interface to use. This will panic under high load. Change this to an atomically incremented sequence number in order to choose the tx port in round robin. 2. IFQ_HANDOFF will free the mbuf if the queue is full, this will then be freed again by lagg_start() and panic. Reorganised the error handling and freeing to fix this. MFC after: 3 days
2007-10-12 03:03:16 +00:00
while (p--)
CK: update consumers to use CK macros across the board r334189 changed the fields to have names distinct from those in queue.h in order to expose the oversights as compile time errors
2018-05-24 23:21:23 +00:00
lp = CK_SLIST_NEXT(lp, lp_entries);
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
Fix two panics in lagg. 1. The locking was changed to shared but roundrobin mode still updated a pointer in the softc with the next tx interface to use. This will panic under high load. Change this to an atomically incremented sequence number in order to choose the tx port in round robin. 2. IFQ_HANDOFF will free the mbuf if the queue is full, this will then be freed again by lagg_start() and panic. Reorganised the error handling and freeing to fix this. MFC after: 3 days
2007-10-12 03:03:16 +00:00
/*
* Check the port's link state. This will return the next active
* port if the link is down or the port is NULL.
*/
if ((lp = lagg_link_active(sc, lp)) == NULL) {
lagg: stop double-counting output errors and counting drops as errors Before this change, lagg double-counted errors from lagg members, and counted every drop by a lagg member as an error. Eg, if lagg sent a packet, and the underlying hardware driver dropped it, a counter would be incremented by both lagg and the underlying driver. This change attempts to fix that by incrementing lagg's counters only for errors that do not come from underlying drivers. Reviewed by: hselasky, jhb Sponsored by: Netflix Differential Revision: https://reviews.freebsd.org/D24331
2020-04-13 23:06:56 +00:00
if_inc_counter(sc->sc_ifp, IFCOUNTER_OERRORS, 1);
Fix two panics in lagg. 1. The locking was changed to shared but roundrobin mode still updated a pointer in the softc with the next tx interface to use. This will panic under high load. Change this to an atomically incremented sequence number in order to choose the tx port in round robin. 2. IFQ_HANDOFF will free the mbuf if the queue is full, this will then be freed again by lagg_start() and panic. Reorganised the error handling and freeing to fix this. MFC after: 3 days
2007-10-12 03:03:16 +00:00
m_freem(m);
Return ENETDOWN instead of ENOENT when all lagg(4) links are inactive when upper layer tries to transmit packet. This gives better feedback and meaningful errors for applications. MFC after: 2 weeks Reviewed by: thompsa
2013-06-17 19:31:03 +00:00
return (ENETDOWN);
Fix two panics in lagg. 1. The locking was changed to shared but roundrobin mode still updated a pointer in the softc with the next tx interface to use. This will panic under high load. Change this to an atomically incremented sequence number in order to choose the tx port in round robin. 2. IFQ_HANDOFF will free the mbuf if the queue is full, this will then be freed again by lagg_start() and panic. Reorganised the error handling and freeing to fix this. MFC after: 3 days
2007-10-12 03:03:16 +00:00
}
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
/* Send mbuf */
Fix two panics in lagg. 1. The locking was changed to shared but roundrobin mode still updated a pointer in the softc with the next tx interface to use. This will panic under high load. Change this to an atomically incremented sequence number in order to choose the tx port in round robin. 2. IFQ_HANDOFF will free the mbuf if the queue is full, this will then be freed again by lagg_start() and panic. Reorganised the error handling and freeing to fix this. MFC after: 3 days
2007-10-12 03:03:16 +00:00
return (lagg_enqueue(lp->lp_ifp, m));
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
}
static struct mbuf *
lagg_rr_input(struct lagg_softc *sc, struct lagg_port *lp, struct mbuf *m)
{
struct ifnet *ifp = sc->sc_ifp;
/* Just pass in the packet to our lagg device */
m->m_pkthdr.rcvif = ifp;
return (m);
}
Add laggproto broadcast, it allows sends frames to all ports of the lagg(4) group and receives frames on any port of the lagg(4). Phabric: D549 Reviewed by: glebius, thompsa Approved by: glebius Obtained from: OpenBSD Sponsored by: QNAP Systems Inc.
2014-09-18 02:12:48 +00:00
/*
* Broadcast mode
*/
static int
lagg_bcast_start(struct lagg_softc *sc, struct mbuf *m)
{
- Whitespace. - Remove caddr_t.
2014-09-26 12:35:58 +00:00
int errors = 0;
int ret;
struct lagg_port *lp, *last = NULL;
struct mbuf *m0;
Convert LAGG_RLOCK() to NET_EPOCH_ENTER(). No functional changes.
2020-12-08 16:36:46 +00:00
NET_EPOCH_ASSERT();
Replace rmlock with epoch in lagg Use the new epoch based reclamation API. Now the hot paths will not block at all, and the sx lock is used for the softc data. This fixes LORs reported where the rwlock was obtained when the sxlock was held. Submitted by: mmacy Reported by: Harry Schmalzbauer <freebsd@omnilan.de> Reviewed by: sbruno Sponsored by: Limelight Networks Differential Revision: https://reviews.freebsd.org/D15355
2018-05-14 20:06:49 +00:00
CK_SLIST_FOREACH(lp, &sc->sc_ports, lp_entries) {
- Whitespace. - Remove caddr_t.
2014-09-26 12:35:58 +00:00
if (!LAGG_PORTACTIVE(lp))
continue;
if (last != NULL) {
m0 = m_copym(m, 0, M_COPYALL, M_NOWAIT);
if (m0 == NULL) {
ret = ENOBUFS;
errors++;
break;
}
lagg: stop double-counting output errors and counting drops as errors Before this change, lagg double-counted errors from lagg members, and counted every drop by a lagg member as an error. Eg, if lagg sent a packet, and the underlying hardware driver dropped it, a counter would be incremented by both lagg and the underlying driver. This change attempts to fix that by incrementing lagg's counters only for errors that do not come from underlying drivers. Reviewed by: hselasky, jhb Sponsored by: Netflix Differential Revision: https://reviews.freebsd.org/D24331
2020-04-13 23:06:56 +00:00
lagg_enqueue(last->lp_ifp, m0);
- Whitespace. - Remove caddr_t.
2014-09-26 12:35:58 +00:00
}
last = lp;
}
Replace rmlock with epoch in lagg Use the new epoch based reclamation API. Now the hot paths will not block at all, and the sx lock is used for the softc data. This fixes LORs reported where the rwlock was obtained when the sxlock was held. Submitted by: mmacy Reported by: Harry Schmalzbauer <freebsd@omnilan.de> Reviewed by: sbruno Sponsored by: Limelight Networks Differential Revision: https://reviews.freebsd.org/D15355
2018-05-14 20:06:49 +00:00
- Whitespace. - Remove caddr_t.
2014-09-26 12:35:58 +00:00
if (last == NULL) {
lagg: stop double-counting output errors and counting drops as errors Before this change, lagg double-counted errors from lagg members, and counted every drop by a lagg member as an error. Eg, if lagg sent a packet, and the underlying hardware driver dropped it, a counter would be incremented by both lagg and the underlying driver. This change attempts to fix that by incrementing lagg's counters only for errors that do not come from underlying drivers. Reviewed by: hselasky, jhb Sponsored by: Netflix Differential Revision: https://reviews.freebsd.org/D24331
2020-04-13 23:06:56 +00:00
if_inc_counter(sc->sc_ifp, IFCOUNTER_OERRORS, 1);
- Whitespace. - Remove caddr_t.
2014-09-26 12:35:58 +00:00
m_freem(m);
return (ENOENT);
}
if ((last = lagg_link_active(sc, last)) == NULL) {
lagg: stop double-counting output errors and counting drops as errors Before this change, lagg double-counted errors from lagg members, and counted every drop by a lagg member as an error. Eg, if lagg sent a packet, and the underlying hardware driver dropped it, a counter would be incremented by both lagg and the underlying driver. This change attempts to fix that by incrementing lagg's counters only for errors that do not come from underlying drivers. Reviewed by: hselasky, jhb Sponsored by: Netflix Differential Revision: https://reviews.freebsd.org/D24331
2020-04-13 23:06:56 +00:00
errors++;
if_inc_counter(sc->sc_ifp, IFCOUNTER_OERRORS, errors);
- Whitespace. - Remove caddr_t.
2014-09-26 12:35:58 +00:00
m_freem(m);
return (ENETDOWN);
}
ret = lagg_enqueue(last->lp_ifp, m);
lagg: stop double-counting output errors and counting drops as errors Before this change, lagg double-counted errors from lagg members, and counted every drop by a lagg member as an error. Eg, if lagg sent a packet, and the underlying hardware driver dropped it, a counter would be incremented by both lagg and the underlying driver. This change attempts to fix that by incrementing lagg's counters only for errors that do not come from underlying drivers. Reviewed by: hselasky, jhb Sponsored by: Netflix Differential Revision: https://reviews.freebsd.org/D24331
2020-04-13 23:06:56 +00:00
if (errors != 0)
if_inc_counter(sc->sc_ifp, IFCOUNTER_OERRORS, errors);
- Whitespace. - Remove caddr_t.
2014-09-26 12:35:58 +00:00
lagg: stop double-counting output errors and counting drops as errors Before this change, lagg double-counted errors from lagg members, and counted every drop by a lagg member as an error. Eg, if lagg sent a packet, and the underlying hardware driver dropped it, a counter would be incremented by both lagg and the underlying driver. This change attempts to fix that by incrementing lagg's counters only for errors that do not come from underlying drivers. Reviewed by: hselasky, jhb Sponsored by: Netflix Differential Revision: https://reviews.freebsd.org/D24331
2020-04-13 23:06:56 +00:00
return (ret);
Add laggproto broadcast, it allows sends frames to all ports of the lagg(4) group and receives frames on any port of the lagg(4). Phabric: D549 Reviewed by: glebius, thompsa Approved by: glebius Obtained from: OpenBSD Sponsored by: QNAP Systems Inc.
2014-09-18 02:12:48 +00:00
}
static struct mbuf*
- Whitespace. - Remove caddr_t.
2014-09-26 12:35:58 +00:00
lagg_bcast_input(struct lagg_softc *sc, struct lagg_port *lp, struct mbuf *m)
Add laggproto broadcast, it allows sends frames to all ports of the lagg(4) group and receives frames on any port of the lagg(4). Phabric: D549 Reviewed by: glebius, thompsa Approved by: glebius Obtained from: OpenBSD Sponsored by: QNAP Systems Inc.
2014-09-18 02:12:48 +00:00
{
- Whitespace. - Remove caddr_t.
2014-09-26 12:35:58 +00:00
struct ifnet *ifp = sc->sc_ifp;
Add laggproto broadcast, it allows sends frames to all ports of the lagg(4) group and receives frames on any port of the lagg(4). Phabric: D549 Reviewed by: glebius, thompsa Approved by: glebius Obtained from: OpenBSD Sponsored by: QNAP Systems Inc.
2014-09-18 02:12:48 +00:00
- Whitespace. - Remove caddr_t.
2014-09-26 12:35:58 +00:00
/* Just pass in the packet to our lagg device */
m->m_pkthdr.rcvif = ifp;
return (m);
Add laggproto broadcast, it allows sends frames to all ports of the lagg(4) group and receives frames on any port of the lagg(4). Phabric: D549 Reviewed by: glebius, thompsa Approved by: glebius Obtained from: OpenBSD Sponsored by: QNAP Systems Inc.
2014-09-18 02:12:48 +00:00
}
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
/*
* Active failover
*/
static int
lagg_fail_start(struct lagg_softc *sc, struct mbuf *m)
{
struct lagg_port *lp;
/* Use the master port if active or the next available port */
Fix two panics in lagg. 1. The locking was changed to shared but roundrobin mode still updated a pointer in the softc with the next tx interface to use. This will panic under high load. Change this to an atomically incremented sequence number in order to choose the tx port in round robin. 2. IFQ_HANDOFF will free the mbuf if the queue is full, this will then be freed again by lagg_start() and panic. Reorganised the error handling and freeing to fix this. MFC after: 3 days
2007-10-12 03:03:16 +00:00
if ((lp = lagg_link_active(sc, sc->sc_primary)) == NULL) {
lagg: stop double-counting output errors and counting drops as errors Before this change, lagg double-counted errors from lagg members, and counted every drop by a lagg member as an error. Eg, if lagg sent a packet, and the underlying hardware driver dropped it, a counter would be incremented by both lagg and the underlying driver. This change attempts to fix that by incrementing lagg's counters only for errors that do not come from underlying drivers. Reviewed by: hselasky, jhb Sponsored by: Netflix Differential Revision: https://reviews.freebsd.org/D24331
2020-04-13 23:06:56 +00:00
if_inc_counter(sc->sc_ifp, IFCOUNTER_OERRORS, 1);
Fix two panics in lagg. 1. The locking was changed to shared but roundrobin mode still updated a pointer in the softc with the next tx interface to use. This will panic under high load. Change this to an atomically incremented sequence number in order to choose the tx port in round robin. 2. IFQ_HANDOFF will free the mbuf if the queue is full, this will then be freed again by lagg_start() and panic. Reorganised the error handling and freeing to fix this. MFC after: 3 days
2007-10-12 03:03:16 +00:00
m_freem(m);
Return ENETDOWN instead of ENOENT when all lagg(4) links are inactive when upper layer tries to transmit packet. This gives better feedback and meaningful errors for applications. MFC after: 2 weeks Reviewed by: thompsa
2013-06-17 19:31:03 +00:00
return (ENETDOWN);
Fix two panics in lagg. 1. The locking was changed to shared but roundrobin mode still updated a pointer in the softc with the next tx interface to use. This will panic under high load. Change this to an atomically incremented sequence number in order to choose the tx port in round robin. 2. IFQ_HANDOFF will free the mbuf if the queue is full, this will then be freed again by lagg_start() and panic. Reorganised the error handling and freeing to fix this. MFC after: 3 days
2007-10-12 03:03:16 +00:00
}
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
/* Send mbuf */
return (lagg_enqueue(lp->lp_ifp, m));
}
static struct mbuf *
lagg_fail_input(struct lagg_softc *sc, struct lagg_port *lp, struct mbuf *m)
{
struct ifnet *ifp = sc->sc_ifp;
struct lagg_port *tmp_tp;
Virtualize lagg(4) cloner. This change fixes a panic when tearing down if_lagg(4) interfaces which were cloned in a vnet jail. Sysctl nodes which are dynamically generated for each cloned interface (net.link.lagg.N.*) have been removed, and use_flowid and flowid_shift ifconfig(8) parameters have been added instead. Flags and per-interface statistics counters are displayed in "ifconfig -v". CR: D842
2014-10-01 21:37:32 +00:00
if (lp == sc->sc_primary || V_lagg_failover_rx_all) {
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
m->m_pkthdr.rcvif = ifp;
return (m);
}
- Use the macro to check the port status has it will also test if its administratively down (!IFF_UP) - Use the same parameters to lagg_link_active() to get the backup port as in the output path, this didnt actually matter in practice as sc_primary is always the first on the port list. MFC after: 3 days
2007-12-18 02:12:03 +00:00
if (!LAGG_PORTACTIVE(sc->sc_primary)) {
tmp_tp = lagg_link_active(sc, sc->sc_primary);
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
/*
sys/net*: minor spelling fixes. No functional change.
2016-05-03 18:05:43 +00:00
* If tmp_tp is null, we've received a packet when all
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
* our links are down. Weird, but process it anyways.
*/
if ((tmp_tp == NULL || tmp_tp == lp)) {
m->m_pkthdr.rcvif = ifp;
return (m);
}
}
m_freem(m);
return (NULL);
}
/*
* Loadbalancing
*/
- When reconfiguring protocol on a lagg, first set it to LAGG_PROTO_NONE, then drop lock, run the attach routines, and then set it to specific proto. This removes tons of WITNESS warnings. - Make lagg protocol attach handlers not failing and allocate memory with M_WAITOK. Sponsored by: Netflix Sponsored by: Nginx, Inc.
2014-09-26 08:42:32 +00:00
static void
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
lagg_lb_attach(struct lagg_softc *sc)
{
struct lagg_port *lp;
struct lagg_lb *lb;
Replace rmlock with epoch in lagg Use the new epoch based reclamation API. Now the hot paths will not block at all, and the sx lock is used for the softc data. This fixes LORs reported where the rwlock was obtained when the sxlock was held. Submitted by: mmacy Reported by: Harry Schmalzbauer <freebsd@omnilan.de> Reviewed by: sbruno Sponsored by: Limelight Networks Differential Revision: https://reviews.freebsd.org/D15355
2018-05-14 20:06:49 +00:00
LAGG_XLOCK_ASSERT(sc);
Use a dedicated malloc type for lagg(4)'s structures. Reviewed by: gallatin MFC after: 1 month Sponsored by: Netflix Differential Revision: https://reviews.freebsd.org/D19719
2019-03-28 21:00:54 +00:00
lb = malloc(sizeof(struct lagg_lb), M_LAGG, M_WAITOK | M_ZERO);
Factor out mbuf hashing code from LAGG driver so that other network drivers can use it. This avoids some code duplication. Add missing default case to all switch statements while at it. Also move the hashing of the IPv6 flow field to layer 4 because the IPv6 flow field is constant on a per L4 connection basis and not on a per L3 network. Differential Revision: https://reviews.freebsd.org/D1987 Sponsored by: Mellanox Technologies MFC after: 1 month
2015-03-11 16:02:24 +00:00
lb->lb_key = m_ether_tcpip_hash_init();
- Whitespace. - Remove caddr_t.
2014-09-26 12:35:58 +00:00
sc->sc_psc = lb;
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
Replace rmlock with epoch in lagg Use the new epoch based reclamation API. Now the hot paths will not block at all, and the sx lock is used for the softc data. This fixes LORs reported where the rwlock was obtained when the sxlock was held. Submitted by: mmacy Reported by: Harry Schmalzbauer <freebsd@omnilan.de> Reviewed by: sbruno Sponsored by: Limelight Networks Differential Revision: https://reviews.freebsd.org/D15355
2018-05-14 20:06:49 +00:00
CK_SLIST_FOREACH(lp, &sc->sc_ports, lp_entries)
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
lagg_lb_port_create(lp);
}
Make lagg protocols detach methods returning void. Sponsored by: Netflix Sponsored by: Nginx, Inc.
2014-09-26 07:12:40 +00:00
static void
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
lagg_lb_detach(struct lagg_softc *sc)
{
Make lagg protocols detach methods returning void. Sponsored by: Netflix Sponsored by: Nginx, Inc.
2014-09-26 07:12:40 +00:00
struct lagg_lb *lb;
lb = (struct lagg_lb *)sc->sc_psc;
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
if (lb != NULL)
Use a dedicated malloc type for lagg(4)'s structures. Reviewed by: gallatin MFC after: 1 month Sponsored by: Netflix Differential Revision: https://reviews.freebsd.org/D19719
2019-03-28 21:00:54 +00:00
free(lb, M_LAGG);
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
}
static int
lagg_lb_porttable(struct lagg_softc *sc, struct lagg_port *lp)
{
struct lagg_lb *lb = (struct lagg_lb *)sc->sc_psc;
struct lagg_port *lp_next;
There are three places where we return from a function which entered an epoch section without exiting that epoch section. This is bad for two reasons: the epoch section won't exit, and we will leave the epoch tracker from the stack on the epoch list. Fix the epoch leak by making sure we exit epoch sections before returning. Reviewed by: ae, gallatin, mmacy Approved by: re (gjb, kib) Sponsored by: Netflix Differential Revision: https://reviews.freebsd.org/D17450
2018-10-09 13:26:06 +00:00
int i = 0, rv;
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
There are three places where we return from a function which entered an epoch section without exiting that epoch section. This is bad for two reasons: the epoch section won't exit, and we will leave the epoch tracker from the stack on the epoch list. Fix the epoch leak by making sure we exit epoch sections before returning. Reviewed by: ae, gallatin, mmacy Approved by: re (gjb, kib) Sponsored by: Netflix Differential Revision: https://reviews.freebsd.org/D17450
2018-10-09 13:26:06 +00:00
rv = 0;
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
bzero(&lb->lb_ports, sizeof(lb->lb_ports));
Remove nested epochs from lagg(4). lagg_bcast_start appeared to have a bug in that was using the last lagg port structure after exiting the epoch that was keeping that structure alive. However, upon further inspection, the epoch was already entered by the caller (lagg_transmit), so the epoch enter/exit in lagg_bcast_start was actually unnecessary. This commit generally removes uses of the net epoch via LAGG_RLOCK to protect the list of ports when the list of ports was already protected by an existing LAGG_RLOCK in a caller, or the LAGG_XLOCK. It also adds a missing epoch enter/exit in lagg_snd_tag_alloc while accessing the lagg port structures. An ifp is still accessed via an unsafe reference after the epoch is exited, but that is true in the current code and will be fixed in a future change. Reviewed by: gallatin MFC after: 1 month Sponsored by: Netflix Differential Revision: https://reviews.freebsd.org/D19718
2019-03-28 20:25:36 +00:00
LAGG_XLOCK_ASSERT(sc);
Replace rmlock with epoch in lagg Use the new epoch based reclamation API. Now the hot paths will not block at all, and the sx lock is used for the softc data. This fixes LORs reported where the rwlock was obtained when the sxlock was held. Submitted by: mmacy Reported by: Harry Schmalzbauer <freebsd@omnilan.de> Reviewed by: sbruno Sponsored by: Limelight Networks Differential Revision: https://reviews.freebsd.org/D15355
2018-05-14 20:06:49 +00:00
CK_SLIST_FOREACH(lp_next, &sc->sc_ports, lp_entries) {
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
if (lp_next == lp)
continue;
There are three places where we return from a function which entered an epoch section without exiting that epoch section. This is bad for two reasons: the epoch section won't exit, and we will leave the epoch tracker from the stack on the epoch list. Fix the epoch leak by making sure we exit epoch sections before returning. Reviewed by: ae, gallatin, mmacy Approved by: re (gjb, kib) Sponsored by: Netflix Differential Revision: https://reviews.freebsd.org/D17450
2018-10-09 13:26:06 +00:00
if (i >= LAGG_MAX_PORTS) {
rv = EINVAL;
break;
}
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
if (sc->sc_ifflags & IFF_DEBUG)
printf("%s: port %s at index %d\n",
Remove macros that hide access to struct ifnet fields.
2014-09-26 13:02:29 +00:00
sc->sc_ifname, lp_next->lp_ifp->if_xname, i);
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
lb->lb_ports[i++] = lp_next;
}
There are three places where we return from a function which entered an epoch section without exiting that epoch section. This is bad for two reasons: the epoch section won't exit, and we will leave the epoch tracker from the stack on the epoch list. Fix the epoch leak by making sure we exit epoch sections before returning. Reviewed by: ae, gallatin, mmacy Approved by: re (gjb, kib) Sponsored by: Netflix Differential Revision: https://reviews.freebsd.org/D17450
2018-10-09 13:26:06 +00:00
return (rv);
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
}
static int
lagg_lb_port_create(struct lagg_port *lp)
{
non-functional cleanup - remove dead code - use consistent variable names - gc unused defines - whitespace cleanup
2007-06-12 07:29:11 +00:00
struct lagg_softc *sc = lp->lp_softc;
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
return (lagg_lb_porttable(sc, NULL));
}
static void
lagg_lb_port_destroy(struct lagg_port *lp)
{
non-functional cleanup - remove dead code - use consistent variable names - gc unused defines - whitespace cleanup
2007-06-12 07:29:11 +00:00
struct lagg_softc *sc = lp->lp_softc;
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
lagg_lb_porttable(sc, lp);
}
static int
lagg_lb_start(struct lagg_softc *sc, struct mbuf *m)
{
struct lagg_lb *lb = (struct lagg_lb *)sc->sc_psc;
struct lagg_port *lp = NULL;
uint32_t p = 0;
Start process of removing the use of the deprecated "M_FLOWID" flag from the FreeBSD network code. The flag is still kept around in the "sys/mbuf.h" header file, but does no longer have any users. Instead the "m_pkthdr.rsstype" field in the mbuf structure is now used to decide the meaning of the "m_pkthdr.flowid" field. To modify the "m_pkthdr.rsstype" field please use the existing "M_HASHTYPE_XXX" macros as defined in the "sys/mbuf.h" header file. This patch introduces new behaviour in the transmit direction. Previously network drivers checked if "M_FLOWID" was set in "m_flags" before using the "m_pkthdr.flowid" field. This check has now now been replaced by checking if "M_HASHTYPE_GET(m)" is different from "M_HASHTYPE_NONE". In the future more hashtypes will be added, for example hashtypes for hardware dedicated flows. "M_HASHTYPE_OPAQUE" indicates that the "m_pkthdr.flowid" value is valid and has no particular type. This change removes the need for an "if" statement in TCP transmit code checking for the presence of a valid flowid value. The "if" statement mentioned above is now a direct variable assignment which is then later checked by the respective network drivers like before. Additional notes: - The SCTP code changes will be committed as a separate patch. - Removal of the "M_FLOWID" flag will also be done separately. - The FreeBSD version has been bumped. MFC after: 1 month Sponsored by: Mellanox Technologies
2014-12-01 11:45:24 +00:00
if ((sc->sc_opts & LAGG_OPT_USE_FLOWID) &&
M_HASHTYPE_GET(m) != M_HASHTYPE_NONE)
Multi-queue NIC drivers and multi-port lagg tend to use the same lower bits of the flowid as each other, resulting in a poor distribution of packets among queues in certain cases. Work around this by adding a set of sysctls for controlling a bit-shift on the flowid when doing multi-port aggrigation in lagg and lacp. By default, lagg/lacp will now use bits 16 and higher instead of 0 and higher. Reviewed by: max Obtained from: Netflix MFC after: 3 days
2013-12-30 01:32:17 +00:00
p = m->m_pkthdr.flowid >> sc->flowid_shift;
Use the flowid if its available for selecting the tx port.
2009-04-30 14:25:44 +00:00
else
Factor out mbuf hashing code from LAGG driver so that other network drivers can use it. This avoids some code duplication. Add missing default case to all switch statements while at it. Also move the hashing of the IPv6 flow field to layer 4 because the IPv6 flow field is constant on a per L4 connection basis and not on a per L3 network. Differential Revision: https://reviews.freebsd.org/D1987 Sponsored by: Mellanox Technologies MFC after: 1 month
2015-03-11 16:02:24 +00:00
p = m_ether_tcpip_hash(sc->sc_flags, m, lb->lb_key);
port % count will never be greater than LAGG_MAX_PORTS so nuke the test.
2008-07-04 05:33:58 +00:00
p %= sc->sc_count;
lp = lb->lb_ports[p];
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
/*
* Check the port's link state. This will return the next active
* port if the link is down or the port is NULL.
*/
Fix two panics in lagg. 1. The locking was changed to shared but roundrobin mode still updated a pointer in the softc with the next tx interface to use. This will panic under high load. Change this to an atomically incremented sequence number in order to choose the tx port in round robin. 2. IFQ_HANDOFF will free the mbuf if the queue is full, this will then be freed again by lagg_start() and panic. Reorganised the error handling and freeing to fix this. MFC after: 3 days
2007-10-12 03:03:16 +00:00
if ((lp = lagg_link_active(sc, lp)) == NULL) {
lagg: stop double-counting output errors and counting drops as errors Before this change, lagg double-counted errors from lagg members, and counted every drop by a lagg member as an error. Eg, if lagg sent a packet, and the underlying hardware driver dropped it, a counter would be incremented by both lagg and the underlying driver. This change attempts to fix that by incrementing lagg's counters only for errors that do not come from underlying drivers. Reviewed by: hselasky, jhb Sponsored by: Netflix Differential Revision: https://reviews.freebsd.org/D24331
2020-04-13 23:06:56 +00:00
if_inc_counter(sc->sc_ifp, IFCOUNTER_OERRORS, 1);
Fix two panics in lagg. 1. The locking was changed to shared but roundrobin mode still updated a pointer in the softc with the next tx interface to use. This will panic under high load. Change this to an atomically incremented sequence number in order to choose the tx port in round robin. 2. IFQ_HANDOFF will free the mbuf if the queue is full, this will then be freed again by lagg_start() and panic. Reorganised the error handling and freeing to fix this. MFC after: 3 days
2007-10-12 03:03:16 +00:00
m_freem(m);
Return ENETDOWN instead of ENOENT when all lagg(4) links are inactive when upper layer tries to transmit packet. This gives better feedback and meaningful errors for applications. MFC after: 2 weeks Reviewed by: thompsa
2013-06-17 19:31:03 +00:00
return (ENETDOWN);
Fix two panics in lagg. 1. The locking was changed to shared but roundrobin mode still updated a pointer in the softc with the next tx interface to use. This will panic under high load. Change this to an atomically incremented sequence number in order to choose the tx port in round robin. 2. IFQ_HANDOFF will free the mbuf if the queue is full, this will then be freed again by lagg_start() and panic. Reorganised the error handling and freeing to fix this. MFC after: 3 days
2007-10-12 03:03:16 +00:00
}
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
/* Send mbuf */
return (lagg_enqueue(lp->lp_ifp, m));
}
static struct mbuf *
lagg_lb_input(struct lagg_softc *sc, struct lagg_port *lp, struct mbuf *m)
{
struct ifnet *ifp = sc->sc_ifp;
/* Just pass in the packet to our lagg device */
m->m_pkthdr.rcvif = ifp;
return (m);
}
/*
* 802.3ad LACP
*/
- When reconfiguring protocol on a lagg, first set it to LAGG_PROTO_NONE, then drop lock, run the attach routines, and then set it to specific proto. This removes tons of WITNESS warnings. - Make lagg protocol attach handlers not failing and allocate memory with M_WAITOK. Sponsored by: Netflix Sponsored by: Nginx, Inc.
2014-09-26 08:42:32 +00:00
static void
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
lagg_lacp_attach(struct lagg_softc *sc)
{
struct lagg_port *lp;
- When reconfiguring protocol on a lagg, first set it to LAGG_PROTO_NONE, then drop lock, run the attach routines, and then set it to specific proto. This removes tons of WITNESS warnings. - Make lagg protocol attach handlers not failing and allocate memory with M_WAITOK. Sponsored by: Netflix Sponsored by: Nginx, Inc.
2014-09-26 08:42:32 +00:00
lacp_attach(sc);
Replace rmlock with epoch in lagg Use the new epoch based reclamation API. Now the hot paths will not block at all, and the sx lock is used for the softc data. This fixes LORs reported where the rwlock was obtained when the sxlock was held. Submitted by: mmacy Reported by: Harry Schmalzbauer <freebsd@omnilan.de> Reviewed by: sbruno Sponsored by: Limelight Networks Differential Revision: https://reviews.freebsd.org/D15355
2018-05-14 20:06:49 +00:00
LAGG_XLOCK_ASSERT(sc);
CK_SLIST_FOREACH(lp, &sc->sc_ports, lp_entries)
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
lacp_port_create(lp);
}
Make lagg protocols detach methods returning void. Sponsored by: Netflix Sponsored by: Nginx, Inc.
2014-09-26 07:12:40 +00:00
static void
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
lagg_lacp_detach(struct lagg_softc *sc)
{
struct lagg_port *lp;
- When reconfiguring protocol on a lagg, first set it to LAGG_PROTO_NONE, then drop lock, run the attach routines, and then set it to specific proto. This removes tons of WITNESS warnings. - Make lagg protocol attach handlers not failing and allocate memory with M_WAITOK. Sponsored by: Netflix Sponsored by: Nginx, Inc.
2014-09-26 08:42:32 +00:00
void *psc;
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
Replace rmlock with epoch in lagg Use the new epoch based reclamation API. Now the hot paths will not block at all, and the sx lock is used for the softc data. This fixes LORs reported where the rwlock was obtained when the sxlock was held. Submitted by: mmacy Reported by: Harry Schmalzbauer <freebsd@omnilan.de> Reviewed by: sbruno Sponsored by: Limelight Networks Differential Revision: https://reviews.freebsd.org/D15355
2018-05-14 20:06:49 +00:00
LAGG_XLOCK_ASSERT(sc);
CK_SLIST_FOREACH(lp, &sc->sc_ports, lp_entries)
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
lacp_port_destroy(lp);
- When reconfiguring protocol on a lagg, first set it to LAGG_PROTO_NONE, then drop lock, run the attach routines, and then set it to specific proto. This removes tons of WITNESS warnings. - Make lagg protocol attach handlers not failing and allocate memory with M_WAITOK. Sponsored by: Netflix Sponsored by: Nginx, Inc.
2014-09-26 08:42:32 +00:00
psc = sc->sc_psc;
sc->sc_psc = NULL;
lacp_detach(psc);
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
}
static void
lagg_lacp_lladdr(struct lagg_softc *sc)
{
struct lagg_port *lp;
Introduce sleepable locks into if_lagg. Before this change if_lagg was using nonsleepable rmlocks to protect its internal state. This patch introduces another sx lock to protect code paths that require sleeping, while still uses old rmlock to protect hot nonsleepable data paths. This change allows to remove taskqueue decoupling used before to change interface addresses without holding the lock. Instead it uses sx lock to protect direct if_ioctl() calls. As another bonus, the new code synchronizes enabled capabilities of member interfaces, and allows to control them with ifconfig laggX, that was impossible before. This part should fix interoperation with if_bridge, that may need to disable some capabilities, such as TXCSUM or LRO, to allow bridging with noncapable interfaces. MFC after: 2 weeks Sponsored by: iXsystems, Inc. Differential Revision: https://reviews.freebsd.org/D10514
2017-05-02 19:09:11 +00:00
LAGG_SXLOCK_ASSERT(sc);
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
/* purge all the lacp ports */
Replace rmlock with epoch in lagg Use the new epoch based reclamation API. Now the hot paths will not block at all, and the sx lock is used for the softc data. This fixes LORs reported where the rwlock was obtained when the sxlock was held. Submitted by: mmacy Reported by: Harry Schmalzbauer <freebsd@omnilan.de> Reviewed by: sbruno Sponsored by: Limelight Networks Differential Revision: https://reviews.freebsd.org/D15355
2018-05-14 20:06:49 +00:00
CK_SLIST_FOREACH(lp, &sc->sc_ports, lp_entries)
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
lacp_port_destroy(lp);
/* add them back in */
Replace rmlock with epoch in lagg Use the new epoch based reclamation API. Now the hot paths will not block at all, and the sx lock is used for the softc data. This fixes LORs reported where the rwlock was obtained when the sxlock was held. Submitted by: mmacy Reported by: Harry Schmalzbauer <freebsd@omnilan.de> Reviewed by: sbruno Sponsored by: Limelight Networks Differential Revision: https://reviews.freebsd.org/D15355
2018-05-14 20:06:49 +00:00
CK_SLIST_FOREACH(lp, &sc->sc_ports, lp_entries)
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
lacp_port_create(lp);
}
static int
lagg_lacp_start(struct lagg_softc *sc, struct mbuf *m)
{
struct lagg_port *lp;
LACP: When suppressing distributing, return ENOBUFS When links come and go, lacp goes into a "suppress distributing" mode where it drops traffic for 3 seconds. When in this mode, lagg/lacp historiclally drops traffic with ENETDOWN. That return value causes TCP to close any connection where it gets that value back from the lower parts of the stack. This means that any TCP connection with active traffic during a 3-second windown when an LACP link comes or goes would get closed. TCP treats return values of ENOBUFS as transient errors, and re-schedules transmission later. So rather than returning ENETDOWN, lets return ENOBUFS instead. This allows TCP connections to be preserved. I've tested this by repeatedly bouncing links on a Netlfix CDN server under a moderate (20Gb/s) load and overved ENOBUFS reported back to the TCP stack (as reported by a RACK TCP sysctl). Reviewed by: jhb, jtl, rrs Sponsored by: Netflix Differential Revision: https://reviews.freebsd.org/D27188
2020-11-18 14:55:49 +00:00
int err;
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
LACP: When suppressing distributing, return ENOBUFS When links come and go, lacp goes into a "suppress distributing" mode where it drops traffic for 3 seconds. When in this mode, lagg/lacp historiclally drops traffic with ENETDOWN. That return value causes TCP to close any connection where it gets that value back from the lower parts of the stack. This means that any TCP connection with active traffic during a 3-second windown when an LACP link comes or goes would get closed. TCP treats return values of ENOBUFS as transient errors, and re-schedules transmission later. So rather than returning ENETDOWN, lets return ENOBUFS instead. This allows TCP connections to be preserved. I've tested this by repeatedly bouncing links on a Netlfix CDN server under a moderate (20Gb/s) load and overved ENOBUFS reported back to the TCP stack (as reported by a RACK TCP sysctl). Reviewed by: jhb, jtl, rrs Sponsored by: Netflix Differential Revision: https://reviews.freebsd.org/D27188
2020-11-18 14:55:49 +00:00
lp = lacp_select_tx_port(sc, m, &err);
Fix two panics in lagg. 1. The locking was changed to shared but roundrobin mode still updated a pointer in the softc with the next tx interface to use. This will panic under high load. Change this to an atomically incremented sequence number in order to choose the tx port in round robin. 2. IFQ_HANDOFF will free the mbuf if the queue is full, this will then be freed again by lagg_start() and panic. Reorganised the error handling and freeing to fix this. MFC after: 3 days
2007-10-12 03:03:16 +00:00
if (lp == NULL) {
lagg: stop double-counting output errors and counting drops as errors Before this change, lagg double-counted errors from lagg members, and counted every drop by a lagg member as an error. Eg, if lagg sent a packet, and the underlying hardware driver dropped it, a counter would be incremented by both lagg and the underlying driver. This change attempts to fix that by incrementing lagg's counters only for errors that do not come from underlying drivers. Reviewed by: hselasky, jhb Sponsored by: Netflix Differential Revision: https://reviews.freebsd.org/D24331
2020-04-13 23:06:56 +00:00
if_inc_counter(sc->sc_ifp, IFCOUNTER_OERRORS, 1);
Fix two panics in lagg. 1. The locking was changed to shared but roundrobin mode still updated a pointer in the softc with the next tx interface to use. This will panic under high load. Change this to an atomically incremented sequence number in order to choose the tx port in round robin. 2. IFQ_HANDOFF will free the mbuf if the queue is full, this will then be freed again by lagg_start() and panic. Reorganised the error handling and freeing to fix this. MFC after: 3 days
2007-10-12 03:03:16 +00:00
m_freem(m);
LACP: When suppressing distributing, return ENOBUFS When links come and go, lacp goes into a "suppress distributing" mode where it drops traffic for 3 seconds. When in this mode, lagg/lacp historiclally drops traffic with ENETDOWN. That return value causes TCP to close any connection where it gets that value back from the lower parts of the stack. This means that any TCP connection with active traffic during a 3-second windown when an LACP link comes or goes would get closed. TCP treats return values of ENOBUFS as transient errors, and re-schedules transmission later. So rather than returning ENETDOWN, lets return ENOBUFS instead. This allows TCP connections to be preserved. I've tested this by repeatedly bouncing links on a Netlfix CDN server under a moderate (20Gb/s) load and overved ENOBUFS reported back to the TCP stack (as reported by a RACK TCP sysctl). Reviewed by: jhb, jtl, rrs Sponsored by: Netflix Differential Revision: https://reviews.freebsd.org/D27188
2020-11-18 14:55:49 +00:00
return (err);
Fix two panics in lagg. 1. The locking was changed to shared but roundrobin mode still updated a pointer in the softc with the next tx interface to use. This will panic under high load. Change this to an atomically incremented sequence number in order to choose the tx port in round robin. 2. IFQ_HANDOFF will free the mbuf if the queue is full, this will then be freed again by lagg_start() and panic. Reorganised the error handling and freeing to fix this. MFC after: 3 days
2007-10-12 03:03:16 +00:00
}
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
/* Send mbuf */
return (lagg_enqueue(lp->lp_ifp, m));
}
static struct mbuf *
lagg_lacp_input(struct lagg_softc *sc, struct lagg_port *lp, struct mbuf *m)
{
struct ifnet *ifp = sc->sc_ifp;
struct ether_header *eh;
u_short etype;
eh = mtod(m, struct ether_header *);
etype = ntohs(eh->ether_type);
/* Tap off LACP control messages */
LACP frames must not be send VLAN-tagged, check for that before processing. PR: kern/156743 Submitted by: Dmitrij Tejblum MFC after: 1 week
2011-04-30 20:34:52 +00:00
if ((m->m_flags & M_VLANTAG) == 0 && etype == ETHERTYPE_SLOW) {
Pass any unmatched slowprotocols frames up the stack instead of dropping them, there are more subtypes than just LACP.
2007-12-31 01:16:35 +00:00
m = lacp_input(lp, m);
if (m == NULL)
return (NULL);
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
}
/*
* If the port is not collecting or not in the active aggregator then
* free and return.
*/
Switch the LACP state machine over to its own mutex to protect the internals, this means that it no longer grabs the lagg rwlock. Use two port table arrays which list the active ports for Tx and switch between them with an atomic op. Now the lagg rwlock is only exclusively locked for management (ioctls) and queuing of lacp control frames isnt needed.
2008-03-16 19:25:30 +00:00
if (lacp_iscollecting(lp) == 0 || lacp_isactive(lp) == 0) {
Rename the trunk(4) driver to lagg(4) as it is too similar to vlan trunking. The name trunk is misused as the networking term trunk means carrying multiple VLANs over a single connection. The IEEE standard for link aggregation (802.3 section 3) does not talk about 'trunk' at all while it is used throughout IEEE 802.1Q in describing vlans. The lagg(4) driver provides link aggregation, failover and fault tolerance. Discussed on: current@
2007-04-17 00:35:11 +00:00
m_freem(m);
return (NULL);
}
m->m_pkthdr.rcvif = ifp;
return (m);
}
Reference in New Issue Copy Permalink