walks the list of connections in TIME_WAIT closing expired connections
due to contention on the global TCP pcbinfo lock.
To remediate, introduce a new global lock to protect the list of
connections in TIME_WAIT. Only acquire the TCP pcbinfo lock when
closing an expired connection. This limits the window of time when
TCP input processing is stopped to the amount of time needed to close
a single connection.
Submitted by: Julien Charbon <jcharbon@verisign.com>
Reviewed by: rwatson, rrs, adrian
MFC after: 2 months
TCP_KEEPCNT, that allow to control initial timeout, idle time, idle
re-send interval and idle send count on a per-socket basis.
Reviewed by: andre, bz, lstewart
to give way for the pluggable congestion control framework. It is
the task of the congestion control algorithm to set the congestion
window and amount of inflight data without external interference.
In 'struct tcpcb' the variables previously used by the inflight
limiter are renamed to spares to keep the ABI intact and to have
some more space for future extensions.
In 'struct tcp_info' the variable 'tcpi_snd_bwnd' is not removed to
preserve the ABI. It is always set to 0.
In siftr.c in 'struct pkt_node' the variable 'snd_bwnd' is not removed
to preserve the ABI. It is always set to 0.
These unused variable in the various structures may be reused in the
future or garbage collected before the next release or at some other
point when an ABI change happens anyway for other reasons.
No MFC is planned. The inflight bandwidth limiter stays disabled by
default in the other branches but remains available.
feature when you have a seemingly stuck socket and want to figure
out why it has not been closed yet.
No plans to MFC this, as it changes the netstat sysctl ABI.
Reviewed by: andre, rwatson, Eric Van Gyzen
- Reintegrate the ANSI C function declaration change
from tcp_timer.c rev 1.92
- Reorganize the tcpcb structure so that it has a single
pointer to the "tcp_timer" structure which contains all
of the tcp timer callouts. This change means that when
the single tcp timer change is reintegrated, tcpcb will
not change in size, and therefore the ABI between
netstat and the kernel will not change.
Neither of these changes should have any functional
impact.
Reviewed by: bmah, rrs
Approved by: re (bmah)
TCP timers as a single timer, but retain the API changes necessary to
reintroduce this change. This will back out the source of at least two
reported problems: lock leaks in certain timer edge cases, and TCP timers
continuing to fire after a connection has closed (a bug previously fixed and
then reintroduced with the timer rewrite).
In a follow-up commit, some minor restylings and comment changes performed
after the TCP timer rewrite will be reapplied, and a further change to allow
the TCP timer rewrite to be added back without disturbing the ABI. The new
design is believed to be a good thing, but the outstanding issues are
leading to significant stability/correctness problems that are holding
up 7.0.
This patch was generated by silby, but is being committed by proxy due to
poor network connectivity for silby this week.
Approved by: re (kensmith)
Submitted by: silby
Tested by: rwatson, kris
Problems reported by: peter, kris, others
be in ticks "for algorithm stability" when originally committed, it turns
out that it has a significant impact in timing out connections. When we
changed HZ from 100 to 1000, this had a big effect on reducing the time
before dropping connections.
To demonstrate, boot with kern.hz=100. ssh to a box on local ethernet
and establish a reliable round-trip-time (ie: type a few commands).
Then unplug the ethernet and press a key. Time how long it takes to
drop the connection.
The old behavior (with hz=100) caused the connection to typically drop
between 90 and 110 seconds of getting no response.
Now boot with kern.hz=1000 (default). The same test causes the ssh session
to drop after just 9-10 seconds. This is a big deal on a wifi connection.
With kern.hz=1000, change sysctl net.inet.tcp.rexmit_min from 3 to 30.
Note how it behaves the same as when HZ was 100. Also, note that when
booting with hz=100, net.inet.tcp.rexmit_min *used* to be 30.
This commit changes TCPTV_MIN to be scaled with hz. rexmit_min should
always be about 30. If you set hz to Really Slow(TM), there is a safety
feature to prevent a value of 0 being used.
This may be revised in the future, but for the time being, it restores the
old, pre-hz=1000 behavior, which is significantly less annoying.
As a workaround, to avoid rebooting or rebuilding a kernel, you can run
"sysctl net.inet.tcp.rexmit_min=30" and add "net.inet.tcp.rexmit_min=30"
to /etc/sysctl.conf. This is safe to run from 6.0 onwards.
Approved by: re (rwatson)
Reviewed by: andre, silby
other than repo copied tcp_subr.c into tcp_timewait.c#1.284:
tcp_input.c#1.350 tcp_timewait() -> tcp_twcheck()
tcp_timer.c#1.92 tcp_timer_2msl_reset() -> tcp_tw_2msl_reset()
tcp_timer.c#1.92 tcp_timer_2msl_stop() -> tcp_tw_2msl_stop()
tcp_timer.c#1.92 tcp_timer_2msl_tw() -> tcp_tw_2msl_scan()
This is a mechanical move with appropriate renames and making
them static if used only locally.
The tcp_tw_2msl_scan() cleanup function is still run from the
tcp_slowtimo() in tcp_timer.c.
directly to a merged model where only one callout, the next to fire,
is registered.
Instead of callout_reset(9) and callout_stop(9) the new function
tcp_timer_activate() is used which then internally manages the callout.
The single new callout is a mutex callout on inpcb simplifying the
locking a bit.
tcp_timer() is the called function which handles all race conditions
in one place and then dispatches the individual timer functions.
Reviewed by: rwatson (earlier version)
potential issues where the peer does not close, potentially leaving
thousands of connections in FIN_WAIT_2. This is controlled by a new sysctl
fast_finwait2_recycle, which is disabled by default.
Reviewed by: gnn, silby.
timeouts for TCP and T/TCP connections in the TIME_WAIT
state, and we had two separate timed wait queues for them.
Now that is has gone, the timeout is always 2*MSL again,
and there is no reason to keep two queues (the first was
unused anyway!).
Also, reimplement the remaining queue using a TAILQ (it
was technically impossible before, with two queues).
threshold. Inflight doesn't make sense on a LAN as it has
trouble figuring out the maximal bandwidth because of the coarse
tick granularity.
The sysctl net.inet.tcp.inflight.rttthresh specifies the threshold
in milliseconds below which inflight will disengage. It defaults
to 10ms.
Tested by: Joao Barros <joao.barros-at-gmail.com>,
Rich Murphey <rich-at-whiteoaklabs.com>
Sponsored by: TCP/IP Optimization Fundraise 2005
- Trailing tab/space cleanup
- Remove spurious spaces between or before tabs
This change avoids touching files that Andre likely has in his working
set for PFIL hooks changes for IPFW/DUMMYNET.
Approved by: re (scottl)
Submitted by: Xin LI <delphij@frontfree.net>
control block. Allow the socket and tcpcb structures to be freed
earlier than inpcb. Update code to understand an inp w/o a socket.
Reviewed by: hsu, silby, jayanth
Sponsored by: DARPA, NAI Labs
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
is an application space macro and the applications are supposed to be free
to use it as they please (but cannot). This is consistant with the other
BSD's who made this change quite some time ago. More commits to come.
This results in closer behavior to earlier versions, where the fixed
200ms timer actually resulted in a delay anywhere from 1..200ms, with
the average delay being 100ms.
Pointed out by: dg
- eliminate the fast/slow timeout lists for TCP and instead use a
callout entry for each timer.
- increase the TCP timer granularity to HZ
- implement "bad retransmit" recovery, as presented in
"On Estimating End-to-End Network Path Properties", by Allman and Paxson.
Submitted by: jlemon, wollmann
This will make a number of things easier in the future, as well as (finally!)
avoiding the Id-smashing problem which has plagued developers for so long.
Boy, I'm glad we're not using sup anymore. This update would have been
insane otherwise.
time, in seconds, that state for non-established TCP sessions stays about)
a sysctl modifyable variable.
[part 1 of two commits, I just realized I can't play with the indices as
I was typing this commit message.]
