Page MenuHomeFreeBSD
Differential D57155

tcp: remove ddb(4) support
AbandonedPublic
Actions

Authored by glebius on Thu, May 21, 8:12 PM.
Tags
None
Referenced Files
F158017207: D57155.diff
Wed, May 27, 1:03 PM
Unknown Object (File)
Tue, May 26, 11:15 PM
Unknown Object (File)
Tue, May 26, 3:31 PM
Unknown Object (File)
Tue, May 26, 1:01 AM
Unknown Object (File)
Mon, May 25, 8:05 PM
Unknown Object (File)
Sun, May 24, 2:56 PM
Unknown Object (File)
Sun, May 24, 10:28 AM
Unknown Object (File)
Sun, May 24, 10:28 AM
View All 9 Files
Subscribers
bz
imp
melifaro
tuexen

Details

Reviewers
pouria
Group Reviewers
transport
Summary

In the modern kernel we have enough tools for TCP debugging. I personally
never used ddb(4) to debug TCP problems and never heard of anybody using
it in the last 15+ years. The code have always been just a maintenance
burden.

Diff Detail

Repository
rG FreeBSD src repository
Lint
Lint Skipped
Unit
Tests Skipped
Build Status
Buildable 73324
Build 70207: arc lint + arc unit

Event Timeline

glebius created this revision.Thu, May 21, 8:12 PM
Herald added 1 blocking reviewer(s): transport. · View Herald TranscriptThu, May 21, 8:12 PM
Herald added subscribers: melifaro, imp. · View Herald Transcript
glebius requested review of this revision.Thu, May 21, 8:12 PM
Harbormaster completed remote builds in B73324: Diff 178344.Thu, May 21, 8:12 PM
pouria accepted this revision as: pouria.Thu, May 21, 8:50 PM
tuexen added a subscriber: tuexen.Thu, May 21, 8:52 PM

I actually used and extended it. For example, syzkaller is dumping the information for all locked tcpcbs when panicing. This gives at least some information in a situation where you cannot access a core. This was good enough for me to fix some bugs.

I am happy to do some maintenance work, if required.

glebius abandoned this revision.Thu, May 21, 8:53 PM
bz added a subscriber: bz.Thu, May 21, 8:56 PM

Unrelated but FYI I recently started to extend show ifnet again which helped me to fix two bugs so far; just in case that's on anyone's list as well.

glebius added a comment.Thu, May 21, 8:59 PM
In D57155#1309850, @bz wrote:

Unrelated but FYI I recently started to extend show ifnet again which helped me to fix two bugs so far; just in case that's on anyone's list as well.

Bjoern, what was your use case? syzkaller, too?

I personally find very little value in ddb today. Large changes to ifnet I did recently were all debugged with bhyve + kgdb.

bz added a comment.Thu, May 21, 9:13 PM
In D57155#1309855, @glebius wrote:
In D57155#1309850, @bz wrote:

Unrelated but FYI I recently started to extend show ifnet again which helped me to fix two bugs so far; just in case that's on anyone's list as well.

Bjoern, what was your use case? syzkaller, too?

I personally find very little value in ddb today. Large changes to ifnet I did recently were all debugged with bhyve + kgdb.

Bringup on someone's HW/architecture where bhyve isn't ready to be used yet (especially not with passthru).
And in these cases it was M_TEMP memory corruption on if_groups.
There is still another one to be debugged at least.

If bhyve one day works with passthru on arm64 or riscv debugging may be different but then the bare metal would still not boot to actually get there... and typing in a few commands into ddb and setting hwatch is so much faster then transfering a dump to be used offline. Maybe is netdump would support IPv6 but ...

Revision Contents
Changeset List

PathSize
sys/
netinet/
293 lines

Diff 178344

View Options

sys/netinet/tcp_usrreq.c

Loading...