net.inet.tcp.rexmit_min (default 3 ticks equiv)
This sysctl is the retransmit timer RTO minimum,
specified in milliseconds. This value is
designed for algorithmic stability only.
net.inet.tcp.rexmit_slop (default 200ms)
This sysctl is the retransmit timer RTO slop
which is added to every retransmit timeout and
is designed to handle protocol stack overheads
and delayed ack issues.
Note that the *original* code applied a 1-second
RTO minimum but never applied real slop to the RTO
calculation, so any RTO calculation over one second
would have no slop and thus not account for
protocol stack overheads (TCP timestamps are not
a measure of protocol turnaround!). Essentially,
the original code made the RTO calculation almost
completely irrelevant.
Please note that the 200ms slop is debateable.
This commit is not meant to be a line in the sand,
and if the community winds up deciding that increasing
it is the correct solution then it's easy to do.
Note that larger values will destroy performance
on lossy networks while smaller values may result in
a greater number of unnecessary retransmits.
one second but it badly breaks throughput on networks with minor packet
loss.
Complaints by: at least two people tracked down to this.
MFC after: 3 days
just because you leave your session idle.
Also, put in a fix for 64-bit architectures (to be revised).
In detail:
ip_fw.h
* Reorder fields in struct ip_fw to avoid alignment problems on
64-bit machines. This only masks the problem, I am still not
sure whether I am doing something wrong in the code or there
is a problem elsewhere (e.g. different aligmnent of structures
between userland and kernel because of pragmas etc.)
* added fields in dyn_rule to store ack numbers, so we can
generate keepalives when the dynamic rule is about to expire
ip_fw2.c
* use a local function, send_pkt(), to generate TCP RST for Reset rules;
* save about 250 bytes by cleaning up the various snprintf()
in ipfw_log() ...
* ... and use twice as many bytes to implement keepalives
(this seems to be working, but i have not tested it extensively).
Keepalives are generated once every 5 seconds for the last 20 seconds
of the lifetime of a dynamic rule for an established TCP flow. The
packets are sent to both sides, so if at least one of the endpoints
is responding, the timeout is refreshed and the rule will not expire.
You can disable this feature with
sysctl net.inet.ip.fw.dyn_keepalive=0
(the default is 1, to have them enabled).
MFC after: 1 day
(just kidding... I will supply an updated version of ipfw2 for
RELENG_4 tomorrow).
This was always broken in HEAD (the offending statement was introduced
in rev. 1.123 for HEAD, while RELENG_4 included this fix (in rev.
1.99.2.12 for RELENG_4) and I inadvertently deleted it in 1.99.2.30.
So I am also restoring these two lines in RELENG_4 now.
We might need another few things from 1.99.2.30.
no punch_fw was used.
Fix another couple of bugs which prevented rules from being
installed properly.
On passing, use IPFW2 instead of NEW_IPFW to compile the new code,
and slightly simplify the instruction generation code.
Following Darren's suggestion, make Dijkstra happy and rewrite the
ipfw_chk() main loop removing a lot of goto's and using instead a
variable to store match status.
Add a lot of comments to explain what instructions are supposed to
do and how -- this should ease auditing of the code and make people
more confident with it.
In terms of code size: the entire file takes about 12700 bytes of text,
about 3K of which are for the main function, ipfw_chk(), and 2K (ouch!)
for ipfw_log().
now it should support all the instructions of the old ipfw.
Fix some bugs in the user interface, /sbin/ipfw.
Please check this code against your rulesets, so i can fix the
remaining bugs (if any, i think they will be mostly in /sbin/ipfw).
Once we have done a bit of testing, this code is ready to be MFC'ed,
together with a bunch of other changes (glue to ipfw, and also the
removal of some global variables) which have been in -current for
a couple of weeks now.
MFC after: 7 days
so that, if we recieve a ICMP "time to live exceeded in transit",
(type 11, code 0) for a TCP connection on SYN-SENT state, close
the connection.
MFC after: 2 weeks
syncache_respond(A), ip_output(), ip_input(), tcp_input(), syncache_badack(B)
Which winds up deleting a different entry from the syncache. Handle
this by not utilizing the next entry in the timer chain until after
syncache_respond() completes. The case of A == B should not be possible.
Problem found by: Don Bowman <don@sandvine.com>
This code makes use of variable-size kernel representation of rules
(exactly the same concept of BPF instructions, as used in the BSDI's
firewall), which makes firewall operation a lot faster, and the
code more readable and easier to extend and debug.
The interface with the rest of the system is unchanged, as witnessed
by this commit. The only extra kernel files that I am touching
are if_fw.h and ip_dummynet.c, which is quite tied to ipfw. In
userland I only had to touch those programs which manipulate the
internal representation of firewall rules).
The code is almost entirely new (and I believe I have written the
vast majority of those sections which were taken from the former
ip_fw.c), so rather than modifying the old ip_fw.c I decided to
create a new file, sys/netinet/ip_fw2.c . Same for the user
interface, which is in sbin/ipfw/ipfw2.c (it still compiles to
/sbin/ipfw). The old files are still there, and will be removed
in due time.
I have not renamed the header file because it would have required
touching a one-line change to a number of kernel files.
In terms of user interface, the new "ipfw" is supposed to accepts
the old syntax for ipfw rules (and produce the same output with
"ipfw show". Only a couple of the old options (out of some 30 of
them) has not been implemented, but they will be soon.
On the other hand, the new code has some very powerful extensions.
First, you can put "or" connectives between match fields (and soon
also between options), and write things like
ipfw add allow ip from { 1.2.3.4/27 or 5.6.7.8/30 } 10-23,25,1024-3000 to any
This should make rulesets slightly more compact (and lines longer!),
by condensing 2 or more of the old rules into single ones.
Also, as an example of how easy the rules can be extended, I have
implemented an 'address set' match pattern, where you can specify
an IP address in a format like this:
10.20.30.0/26{18,44,33,22,9}
which will match the set of hosts listed in braces belonging to the
subnet 10.20.30.0/26 . The match is done using a bitmap, so it is
essentially a constant time operation requiring a handful of CPU
instructions (and a very small amount of memmory -- for a full /24
subnet, the instruction only consumes 40 bytes).
Again, in this commit I have focused on functionality and tried
to minimize changes to the other parts of the system. Some performance
improvement can be achieved with minor changes to the interface of
ip_fw_chk_t. This will be done later when this code is settled.
The code is meant to compile unmodified on RELENG_4 (once the
PACKET_TAG_* changes have been merged), for this reason
you will see #ifdef __FreeBSD_version in a couple of places.
This should minimize errors when (hopefully soon) it will be time
to do the MFC.
MAKEDEV: Add MAKEDEV glue for the ti(4) device nodes.
ti.4: Update the ti(4) man page to include information on the
TI_JUMBO_HDRSPLIT and TI_PRIVATE_JUMBOS kernel options,
and also include information about the new character
device interface and the associated ioctls.
man9/Makefile: Add jumbo.9 and zero_copy.9 man pages and associated
links.
jumbo.9: New man page describing the jumbo buffer allocator
interface and operation.
zero_copy.9: New man page describing the general characteristics of
the zero copy send and receive code, and what an
application author should do to take advantage of the
zero copy functionality.
NOTES: Add entries for ZERO_COPY_SOCKETS, TI_PRIVATE_JUMBOS,
TI_JUMBO_HDRSPLIT, MSIZE, and MCLSHIFT.
conf/files: Add uipc_jumbo.c and uipc_cow.c.
conf/options: Add the 5 options mentioned above.
kern_subr.c: Receive side zero copy implementation. This takes
"disposable" pages attached to an mbuf, gives them to
a user process, and then recycles the user's page.
This is only active when ZERO_COPY_SOCKETS is turned on
and the kern.ipc.zero_copy.receive sysctl variable is
set to 1.
uipc_cow.c: Send side zero copy functions. Takes a page written
by the user and maps it copy on write and assigns it
kernel virtual address space. Removes copy on write
mapping once the buffer has been freed by the network
stack.
uipc_jumbo.c: Jumbo disposable page allocator code. This allocates
(optionally) disposable pages for network drivers that
want to give the user the option of doing zero copy
receive.
uipc_socket.c: Add kern.ipc.zero_copy.{send,receive} sysctls that are
enabled if ZERO_COPY_SOCKETS is turned on.
Add zero copy send support to sosend() -- pages get
mapped into the kernel instead of getting copied if
they meet size and alignment restrictions.
uipc_syscalls.c:Un-staticize some of the sf* functions so that they
can be used elsewhere. (uipc_cow.c)
if_media.c: In the SIOCGIFMEDIA ioctl in ifmedia_ioctl(), avoid
calling malloc() with M_WAITOK. Return an error if
the M_NOWAIT malloc fails.
The ti(4) driver and the wi(4) driver, at least, call
this with a mutex held. This causes witness warnings
for 'ifconfig -a' with a wi(4) or ti(4) board in the
system. (I've only verified for ti(4)).
ip_output.c: Fragment large datagrams so that each segment contains
a multiple of PAGE_SIZE amount of data plus headers.
This allows the receiver to potentially do page
flipping on receives.
if_ti.c: Add zero copy receive support to the ti(4) driver. If
TI_PRIVATE_JUMBOS is not defined, it now uses the
jumbo(9) buffer allocator for jumbo receive buffers.
Add a new character device interface for the ti(4)
driver for the new debugging interface. This allows
(a patched version of) gdb to talk to the Tigon board
and debug the firmware. There are also a few additional
debugging ioctls available through this interface.
Add header splitting support to the ti(4) driver.
Tweak some of the default interrupt coalescing
parameters to more useful defaults.
Add hooks for supporting transmit flow control, but
leave it turned off with a comment describing why it
is turned off.
if_tireg.h: Change the firmware rev to 12.4.11, since we're really
at 12.4.11 plus fixes from 12.4.13.
Add defines needed for debugging.
Remove the ti_stats structure, it is now defined in
sys/tiio.h.
ti_fw.h: 12.4.11 firmware.
ti_fw2.h: 12.4.11 firmware, plus selected fixes from 12.4.13,
and my header splitting patches. Revision 12.4.13
doesn't handle 10/100 negotiation properly. (This
firmware is the same as what was in the tree previously,
with the addition of header splitting support.)
sys/jumbo.h: Jumbo buffer allocator interface.
sys/mbuf.h: Add a new external mbuf type, EXT_DISPOSABLE, to
indicate that the payload buffer can be thrown away /
flipped to a userland process.
socketvar.h: Add prototype for socow_setup.
tiio.h: ioctl interface to the character portion of the ti(4)
driver, plus associated structure/type definitions.
uio.h: Change prototype for uiomoveco() so that we'll know
whether the source page is disposable.
ufs_readwrite.c:Update for new prototype of uiomoveco().
vm_fault.c: In vm_fault(), check to see whether we need to do a page
based copy on write fault.
vm_object.c: Add a new function, vm_object_allocate_wait(). This
does the same thing that vm_object allocate does, except
that it gives the caller the opportunity to specify whether
it should wait on the uma_zalloc() of the object structre.
This allows vm objects to be allocated while holding a
mutex. (Without generating WITNESS warnings.)
vm_object_allocate() is implemented as a call to
vm_object_allocate_wait() with the malloc flag set to
M_WAITOK.
vm_object.h: Add prototype for vm_object_allocate_wait().
vm_page.c: Add page-based copy on write setup, clear and fault
routines.
vm_page.h: Add page based COW function prototypes and variable in
the vm_page structure.
Many thanks to Drew Gallatin, who wrote the zero copy send and receive
code, and to all the other folks who have tested and reviewed this code
over the years.
Add XXX comments to mark places which need to be taken care of
if we want to remove this part of the kernel from Giant.
Add a comment on a potential performance problem with ip_forward()
packet forwarding state ("annotations") during ip processing.
The code is considerably cleaner now.
The variables removed by this change are:
ip_divert_cookie used by divert sockets
ip_fw_fwd_addr used for transparent ip redirection
last_pkt used by dynamic pipes in dummynet
Removal of the first two has been done by carrying the annotations
into volatile structs prepended to the mbuf chains, and adding
appropriate code to add/remove annotations in the routines which
make use of them, i.e. ip_input(), ip_output(), tcp_input(),
bdg_forward(), ether_demux(), ether_output_frame(), div_output().
On passing, remove a bug in divert handling of fragmented packet.
Now it is the fragment at offset 0 which sets the divert status of
the whole packet, whereas formerly it was the last incoming fragment
to decide.
Removal of last_pkt required a change in the interface of ip_fw_chk()
and dummynet_io(). On passing, use the same mechanism for dummynet
annotations and for divert/forward annotations.
option IPFIREWALL_FORWARD is effectively useless, the code to
implement it is very small and is now in by default to avoid the
obfuscation of conditionally compiled code.
NOTES:
* there is at least one global variable left, sro_fwd, in ip_output().
I am not sure if/how this can be removed.
* I have deliberately avoided gratuitous style changes in this commit
to avoid cluttering the diffs. Minor stule cleanup will likely be
necessary
* this commit only focused on the IP layer. I am sure there is a
number of global variables used in the TCP and maybe UDP stack.
* despite the number of files touched, there are absolutely no API's
or data structures changed by this commit (except the interfaces of
ip_fw_chk() and dummynet_io(), which are internal anyways), so
an MFC is quite safe and unintrusive (and desirable, given the
improved readability of the code).
MFC after: 10 days
Register the ISR early, but do not actually kick off the timer until we
see some activity. This still saves us from running the arp timers on
a system with no network cards.
indication of whether this happenned so the calling function
knows whether or not to unlock the pcb.
Submitted by: Jennifer Yang (yangjihui@yahoo.com)
Bug reported by: Sid Carter (sidcarter@symonds.net)
Ensure that the syn cache's syn-ack packets contain the same
ip_tos, ip_ttl, and DF bits as all other tcp packets.
PR: 39141
MFC after: 2 weeks
This time, make sure that ipv4 specific code (aka all of the above)
is only run in the ipv4 case.