Re: crash with head as of 2h ago (in_pcblbgroup_insert)

[Alexander Leidinger]

Am 2025-02-21 22:31, schrieb Mark Johnston:

On Thu, Feb 20, 2025 at 10:27:28AM +0100, Alexander Leidinger wrote:

Hi,

I get this backtrace:
---snip---
[102] panic: invalid local group size 16 and count 16
[102] cpuid = 17
[102] time = 1740041984
[102] KDB: stack backtrace:
[102] db_trace_self_wrapper() at db_trace_self_wrapper+0x2b/frame
0xfffffe07453a5b80
[102] vpanic() at vpanic+0x136/frame 0xfffffe07453a5cb0
[102] panic() at panic+0x43/frame 0xfffffe07453a5d10
[102] in_pcblbgroup_insert() at in_pcblbgroup_insert+0xaa/frame
0xfffffe07453a5d30
[102] in_pcblisten() at in_pcblisten+0x92/frame 0xfffffe07453a5d50

[102] solisten() at solisten+0x47/frame 0xfffffe07453a5dd0
[102] kern_listen() at kern_listen+0x3f/frame 0xfffffe07453a5e00
[102] amd64_syscall() at amd64_syscall+0x15b/frame 0xfffffe07453a5f30
[102] fast_syscall_common() at fast_syscall_common+0xf8/frame
0xfffffe07453a5f30
---snip---


from 1 month ago (2025-01-22-120655).

[102] tcp_usr_listen() at tcp_usr_listen+0x18a/frame 0xfffffe07453a5db0 This is with head updated at 2025-02-20-080119 CET. The previous world is

I believe this would be fixed by https://reviews.freebsd.org/D49100

So I was just luck that this works ATM... or unlucky that I got a panic on the previous try.

Please note that I haven't tested it yet (currently running it through
the regression test suite to start).

Bye,
Alexander.

--
http://www.Leidinger.net alexan...@leidinger.net: PGP 0x8F31830F9F2772BF
http://www.FreeBSD.org    netch...@freebsd.org  : PGP 0x8F31830F9F2772BF

As the config and number of applications/jails/services started didn't change, any idea how to provoke this race condition on a system with some real applications (instead of test cases; if you need an independent validation)? A reboot loop until it happens is not really something I want to do on this system.

signature.asc
Description: OpenPGP digital signature