Re: Soft-Lockup/Race in networking in 2.6.31-rc1+195 (possibly caused by netem)

[Andres Freund <andres@anarazel.de>]

On 07/01/2009 11:22 PM, Andres Freund wrote:
> On 07/01/2009 08:39 PM, Jarek Poplawski wrote:
>> Andres Freund wrote, On 07/01/2009 01:20 AM:
>>> While playing around with netem (time, not packet count based loss-
>>> bursts) I experienced soft lockups several times - to exclude it was my
>>> modifications causing this I recompiled with the original and it is
>>> still locking up.
>>> I captured several of those traces via the thankfully
>>> still working netconsole.
>>> The simplest policy I could reproduce the error with was:
>>> tc qdisc add dev eth0 root handle 1: netem delay 10ms loss 0
>>>
>>> I could not reproduce the error without delay - but that may only be a
>>> timing issue, as the host I was mainly transferring data to was on a
>>> local network.
>>> I could not reproduce the issue on lo.
>>>
>>> The time to reproduce the error varied from seconds after executing tc
>>> to several minutes.
>>>
>>> Traces 5+6 are made with vanilla
>>> 52989765629e7d182b4f146050ebba0abf2cb0b7
>>>
>>> The earlier traces are made with parts of my patches applied, and only
>>> included for completeness as I don't believe my modifications were
>>> causing this and all traces are different, so it may give some clues.
>>>
>>> Lockdep was enabled but did not diagnose anything relevant (one dvb
>>> warning during bootup).
>>>
>>> Any ideas for debugging?
>>
>> Maybe these traces will be enough, but lockdep report could save time.
>> If dvb warning triggers every time then lockdep probably turns off
>> just after (it works this way, unless something was changed). So,
>> could you try to repeat this without dvb? Btw., did you try this on
>> some earlier kernel?
> Yes. Today I could not manage to reproduce it on 2.6.30 but could on
> current git...
> Will try without dvb.
So I tried - and I did not catch any lockdep output before the crash. 
Unfortunately I do not have another machine on the same local network to 
catch any messages after the crash... So I could be missing some warning 
(I did synchronous logging though).
Will check with netconsole tomorrow.

Andres