Page MenuHomeFreeBSD
Differential D24826

Hung sessions
ClosedPublic
Actions

Authored by rrs on May 13 2020, 3:18 PM.
Tags
None
Referenced Files
F103584641: D24826.diff
Tue, Nov 26, 6:51 PM
Unknown Object (File)
Sat, Nov 23, 7:09 AM
Unknown Object (File)
Sat, Nov 23, 6:58 AM
Unknown Object (File)
Oct 25 2024, 11:49 PM
Unknown Object (File)
Oct 11 2024, 3:39 PM
Unknown Object (File)
Oct 10 2024, 12:57 AM
Unknown Object (File)
Oct 7 2024, 11:24 PM
Unknown Object (File)
Sep 25 2024, 4:10 AM
View All 61 Files
Subscribers
melifaro

Details

Reviewers
tuexen
Group Reviewers
transport
Summary

Michael Tuexen as came up with a test case where rack ends with a stuck session. It rotates
around the TCP-FO. The problem is a return(0) and not a walk through the normal just-return
path.. special for SYN-Sent/SYN-Rcvd and FO.

This change fixes that. The attached program is a reproducer that Michael sent me. With the
fix the session now is no longer stuck.

Test Plan

michaels_stuck_tcp.c870 BDownload

Diff Detail

Lint
Lint Skipped
Unit
Tests Skipped

Event Timeline

rrs created this revision.May 13 2020, 3:18 PM
Herald added 1 blocking reviewer(s): transport. · View Herald TranscriptMay 13 2020, 3:18 PM
Herald added a subscriber: melifaro. · View Herald Transcript
rrs requested review of this revision.May 13 2020, 3:18 PM
tuexen added a comment.May 13 2020, 5:11 PM

Except for the whitespace issue, I'm fine with it.

FreeBSD/sys/netinet/tcp_stacks/rack.c
12122 ↗(On Diff #71719)

Are the trailing whitespaces intended?

rrs updated this revision to Diff 71730.May 13 2020, 7:03 PM

Looks like BBR has the same issue, and we need to strip trailing emacs presents too :)

rrs updated this revision to Diff 71761.May 14 2020, 11:20 AM

Opps used the wrong goto label...

tuexen accepted this revision.May 14 2020, 6:09 PM
This revision is now accepted and ready to land.May 14 2020, 6:09 PM
rrs updated this revision to Diff 71807.May 15 2020, 4:48 AM

Let's also fix the accounting panic that skyzaller is finding here too. When we are in a front
state and have a SYN outstanding that should count in the ctf_outstanding size. Also if
we have sent a FIN same thing.

This revision now requires review to proceed.May 15 2020, 4:48 AM
rrs updated this revision to Diff 71814.May 15 2020, 11:13 AM

And even yet another issue thanks to skyzaller in bbr.. the rc_lost when doing TCP-FO was incorrect in rxt timers.

rrs updated this revision to Diff 71816.May 15 2020, 12:01 PM

One little other niggle that I noticed, we should not do a goto again if the SYN bit is set ...

rrs updated this revision to Diff 71819.May 15 2020, 12:39 PM

We should not ever do GP measurements until we are established

tuexen accepted this revision.May 15 2020, 1:07 PM
This revision is now accepted and ready to land.May 15 2020, 1:07 PM
tuexen added a comment.May 15 2020, 2:17 PM

This change is committed in r361080.

tuexen closed this revision.May 15 2020, 2:18 PM
rscheff mentioned this in D24852: Invalid addresses in sending or connecting.May 15 2020, 7:42 PM

Revision Contents
Changeset List

PathSize
tcp_stacks/
13 lines
6 lines
9 lines

Diff 71816

View Options

tcp_stacks/bbr.c

Loading...
View Options

tcp_stacks/rack.c

Loading...
View Options

tcp_stacks/rack_bbr_common.c

Loading...