1
0
mirror of https://git.FreeBSD.org/src.git synced 2024-12-25 11:37:56 +00:00
freebsd/sys/net
Robert Watson 7527624efa Several years after initial development, merge prototype support for
linking NIC Receive Side Scaling (RSS) to the network stack's
connection-group implementation.  This prototype (and derived patches)
are in use at Juniper and several other FreeBSD-using companies, so
despite some reservations about its maturity, merge the patch to the
base tree so that it can be iteratively refined in collaboration rather
than maintained as a set of gradually diverging patch sets.

(1) Merge a software implementation of the Toeplitz hash specified in
    RSS implemented by David Malone.  This is used to allow suitable
    pcbgroup placement of connections before the first packet is
    received from the NIC.  Software hashing is generally avoided,
    however, due to high cost of the hash on general-purpose CPUs.

(2) In in_rss.c, maintain authoritative versions of RSS state intended
    to be pushed to each NIC, including keying material, hash
    algorithm/ configuration, and buckets.  Provide software-facing
    interfaces to hash 2- and 4-tuples for IPv4 and IPv6 using both
    the RSS standardised Toeplitz and a 'naive' variation with a hash
    efficient in software but with poor distribution properties.
    Implement rss_m2cpuid()to be used by netisr and other load
    balancing code to look up the CPU on which an mbuf should be
    processed.

(3) In the Ethernet link layer, allow netisr distribution using RSS as
    a source of policy as an alternative to source ordering; continue
    to default to direct dispatch (i.e., don't try and requeue packets
    for processing on the 'right' CPU if they arrive in a directly
    dispatchable context).

(4) Allow RSS to control tuning of connection groups in order to align
    groups with RSS buckets.  If a packet arrives on a protocol using
    connection groups, and contains a suitable hardware-generated
    hash, use that hash value to select the connection group for pcb
    lookup for both IPv4 and IPv6.  If no hardware-generated Toeplitz
    hash is available, we fall back on regular PCB lookup risking
    contention rather than pay the cost of Toeplitz in software --
    this is a less scalable but, at my last measurement, faster
    approach.  As core counts go up, we may want to revise this
    strategy despite CPU overhead.

Where device drivers suitably configure NICs, and connection groups /
RSS are enabled, this should avoid both lock and line contention during
connection lookup for TCP.  This commit does not modify any device
drivers to tune device RSS configuration to the global RSS
configuration; patches are in circulation to do this for at least
Chelsio T3 and Intel 1G/10G drivers.  Currently, the KPI for device
drivers is not particularly robust, nor aware of more advanced features
such as runtime reconfiguration/rebalancing.  This will hopefully prove
a useful starting point for refinement.

No MFC is scheduled as we will first want to nail down a more mature
and maintainable KPI/KBI for device drivers.

Sponsored by:   Juniper Networks (original work)
Sponsored by:   EMC/Isilon (patch update and merge)
2014-03-15 00:57:50 +00:00
..
bpf_buffer.c
bpf_buffer.h
bpf_filter.c
bpf_jitter.c
bpf_jitter.h
bpf_zerocopy.c
bpf_zerocopy.h
bpf.c Convert the random entropy harvesting code to use a const void * pointer 2013-11-01 20:53:49 +00:00
bpf.h Style: s/SYS_EVENTHANDLER_H/_SYS_EVENTHANDLER_H_/g 2013-10-28 20:32:05 +00:00
bpfdesc.h
bridgestp.c The r48589 promised to remove implicit inclusion of if_var.h soon. Prepare 2013-10-26 17:58:36 +00:00
bridgestp.h
ethernet.h
fddi.h
firewire.h
flowtable.c Pacify gcc. 2014-03-05 02:35:15 +00:00
flowtable.h Add my copyright to flowtable. 2014-02-17 12:07:17 +00:00
ieee8023ad_lacp.c Simplify filling sockaddr_dl structure for if_resolvemulti() 2014-01-18 23:24:51 +00:00
ieee8023ad_lacp.h Break out the static, global LACP debug options into a per-lagg unit 2013-07-26 19:41:13 +00:00
ieee_oui.h Clean up the macros to avoid using casts. 2013-11-15 16:03:32 +00:00
if_arc.h
if_arcsubr.c Remove IPX support. 2014-03-14 02:58:48 +00:00
if_arp.h
if_atm.h
if_atmsubr.c The r48589 promised to remove implicit inclusion of if_var.h soon. Prepare 2013-10-26 17:58:36 +00:00
if_bridge.c Since 32-bit if_baudrate isn't enough to describe a baud rate of a 10 Gbit 2014-03-13 03:42:24 +00:00
if_bridgevar.h
if_clone.c Include necessary headers that now are available due to pollution 2013-10-28 07:29:16 +00:00
if_clone.h Style: s/SYS_EVENTHANDLER_H/_SYS_EVENTHANDLER_H_/g 2013-10-28 20:32:05 +00:00
if_dead.c
if_debug.c
if_disc.c - Remove rt_metrics_lite and simply put its members into rtentry. 2014-03-05 01:17:47 +00:00
if_dl.h Simplify filling sockaddr_dl structure for if_resolvemulti() 2014-01-18 23:24:51 +00:00
if_edsc.c The r48589 promised to remove implicit inclusion of if_var.h soon. Prepare 2013-10-26 17:58:36 +00:00
if_enc.c The r48589 promised to remove implicit inclusion of if_var.h soon. Prepare 2013-10-26 17:58:36 +00:00
if_enc.h
if_epair.c Since 32-bit if_baudrate isn't enough to describe a baud rate of a 10 Gbit 2014-03-13 03:42:24 +00:00
if_ethersubr.c Several years after initial development, merge prototype support for 2014-03-15 00:57:50 +00:00
if_faith.c - Remove rt_metrics_lite and simply put its members into rtentry. 2014-03-05 01:17:47 +00:00
if_fddisubr.c Remove AppleTalk support. 2014-03-14 06:29:43 +00:00
if_fwsubr.c The r48589 promised to remove implicit inclusion of if_var.h soon. Prepare 2013-10-26 17:58:36 +00:00
if_gif.c ANSIfy function defintions. 2013-11-15 12:12:50 +00:00
if_gif.h
if_gre.c Remove AppleTalk support. 2014-03-14 06:29:43 +00:00
if_gre.h
if_iso88025subr.c Remove IPX support. 2014-03-14 02:58:48 +00:00
if_lagg.c Simplify filling sockaddr_dl structure for if_resolvemulti() 2014-01-18 23:24:51 +00:00
if_lagg.h Multi-queue NIC drivers and multi-port lagg tend to use the same lower 2013-12-30 01:32:17 +00:00
if_llatbl.c
if_llatbl.h Remove include opt_ofed.h since OFED is unifdef'd. 2013-08-27 16:45:00 +00:00
if_llc.h
if_loop.c Remove AppleTalk support. 2014-03-14 06:29:43 +00:00
if_media.c
if_media.h Provide forward declaration for struct ifnet. Consumers 2013-10-27 17:27:06 +00:00
if_mib.c The r48589 promised to remove implicit inclusion of if_var.h soon. Prepare 2013-10-26 17:58:36 +00:00
if_mib.h
if_pflog.h
if_pfsync.h Remove more constants related to static sysctl nodes. The MAXID constants 2014-02-25 18:44:33 +00:00
if_sppp.h
if_spppfr.c Remove AppleTalk support. 2014-03-14 06:29:43 +00:00
if_spppsubr.c Remove IPX support. 2014-03-14 02:58:48 +00:00
if_stf.c - Remove rt_metrics_lite and simply put its members into rtentry. 2014-03-05 01:17:47 +00:00
if_stf.h
if_tap.c Teach every SIOCGIFSTATUS provider to fill in ifs->ascii anyway. 2014-01-07 15:59:33 +00:00
if_tap.h
if_tapvar.h
if_tun.c Remove AppleTalk support. 2014-03-14 06:29:43 +00:00
if_tun.h
if_types.h Remove IPX support. 2014-03-14 02:58:48 +00:00
if_var.h Since 32-bit if_baudrate isn't enough to describe a baud rate of a 10 Gbit 2014-03-13 03:42:24 +00:00
if_vlan_var.h Style: s/SYS_EVENTHANDLER_H/_SYS_EVENTHANDLER_H_/g 2013-10-28 20:32:05 +00:00
if_vlan.c Include necessary headers that now are available due to pollution 2013-10-28 07:29:16 +00:00
if.c Simplify filling sockaddr_dl structure for if_resolvemulti() 2014-01-18 23:24:51 +00:00
if.h Since 32-bit if_baudrate isn't enough to describe a baud rate of a 10 Gbit 2014-03-13 03:42:24 +00:00
ifq.h Move all interface queue related structures, macros and definitions 2013-10-29 17:48:08 +00:00
iso88025.h
netisr_internal.h
netisr.c Fix macro name in comment. 2013-11-26 15:23:56 +00:00
netisr.h Remove AppleTalk support. 2014-03-14 06:29:43 +00:00
netmap_user.h This new version of netmap brings you the following: 2014-02-15 04:53:04 +00:00
netmap.h This new version of netmap brings you the following: 2014-02-15 04:53:04 +00:00
pfil.c The r48589 promised to remove implicit inclusion of if_var.h soon. Prepare 2013-10-26 17:58:36 +00:00
pfil.h Whitespace, style cleanups, and improved comments. 2013-08-24 12:03:24 +00:00
pfkeyv2.h
pfvar.h Instead of playing games with casts simply add 3 more members to the 2014-03-05 00:40:03 +00:00
ppp_defs.h
radix_mpath.c - Remove rt_metrics_lite and simply put its members into rtentry. 2014-03-05 01:17:47 +00:00
radix_mpath.h
radix.c Change semantics for rnh_lookup() function: now 2014-01-04 22:25:26 +00:00
radix.h Revert previous commit (262727) and bounce patch back to the 2014-03-04 23:55:04 +00:00
raw_cb.c vnet.h needs to be included before raw_cb.h. Now it compiles due to 2013-10-25 19:49:03 +00:00
raw_cb.h
raw_usrreq.c vnet.h needs to be included before raw_cb.h. Now it compiles due to 2013-10-25 19:49:03 +00:00
route.c The route code used to mtx_destroy() a locked mutex before rtentry free. Now, 2014-03-05 21:16:46 +00:00
route.h The route code used to mtx_destroy() a locked mutex before rtentry free. Now, 2014-03-05 21:16:46 +00:00
rtsock.c Remove IPX support. 2014-03-14 02:58:48 +00:00
sff8472.h Convert #defines to enums so that the values are visible in the debugger. 2014-01-02 21:30:59 +00:00
slcompress.c
slcompress.h
vnet.c Remove identical vnet sysctl handlers, and handle CTLFLAG_VNET 2014-02-07 13:47:33 +00:00
vnet.h Fixup for r261590 (vnet sysctl handlers cleanup). 2014-02-09 08:13:17 +00:00
zlib.c
zlib.h
zutil.h