Re: panic: tcp_do_segment: sent too much

> On 23. Oct 2024, at 09:42, Peter Holm <[email protected]> wrote:
> 
> On Wed, Oct 23, 2024 at 08:32:05AM +0200, Michael Tuexen wrote:
>>> On 23. Oct 2024, at 07:51, Peter Holm <[email protected]> wrote:
>>> 
>>> + export 'tcpINCARNATIONS=16678'
>>> + [ 16678 -le 0 ]
>>> + export 'TESTPROGS= ./testcases/tcp/tcp'
>>> + sleep 900
>>> + su test -c '(cd ..; ./testcases/run/run $TESTPROGS)'
>>>    run: run time  0+00:15:00, incarnations   1, load 100, verbose 1
>>> 23:59:21 Loop #1
>>>    tcp: run time  0+00:02:00, incarnations 3202, load 100, verbose 1
>>> witness_lock_list_get: witness exhausted
>>> panic: tcp_do_segment: sent too much
>>> cpuid = 6
>>> time = 1729634436
>>> KDB: stack backtrace:
>>> db_trace_self_wrapper() at db_trace_self_wrapper+0x2b/frame 0xfffffe01081bd9b0
>>> vpanic() at vpanic+0x13f/frame 0xfffffe01081bdae0
>>> panic() at panic+0x43/frame 0xfffffe01081bdb40
>>> tcp_do_segment() at tcp_do_segment+0x29a2/frame 0xfffffe01081bdc20
>>> tcp_input_with_port() at tcp_input_with_port+0x10e5/frame 0xfffffe01081bdd70
>>> tcp_input() at tcp_input+0xb/frame 0xfffffe01081bdd80
>>> ip_input() at ip_input+0x28f/frame 0xfffffe01081bdde0
>>> swi_net() at swi_net+0x19b/frame 0xfffffe01081bde60
>>> ithread_loop() at ithread_loop+0x266/frame 0xfffffe01081bdef0
>>> fork_exit() at fork_exit+0x82/frame 0xfffffe01081bdf30
>>> fork_trampoline() at fork_trampoline+0xe/frame 0xfffffe01081bdf30
>>> --- trap 0, rip = 0, rsp = 0, rbp = 0 ---
>>> 
>>> Details @ https://people.freebsd.org/~pho/stress/log/log0557.txt
>> Hi Peter,
>> 
>> thank you very much for the report!
>> Is this reproducible? I'm CC-ing Warner, since he has also reported this to
>> me privately, but wasn't able to reproduce it.
>> 
>> Best regards
>> Michael
>>> 
> 
> Hi Michael,
> 
> I have have been able to reproduce the panic with several different
> tcp stress tests (on real hardware).
> 
> Here's how to reproduce this one:
> 
> cd src/tools/test/stress2/misc
> ./all.sh -o tcp.sh
Thank you. I will look into this.

Best regards
Michael
> 
> - Peter