Re: Panic with RACCT on a jailed process

Wiadomość napisana przez Jeremie Le Hen <jlh_at_FreeBSD.org> w dniu 19 cze 2013, o godz. 00:08:
> Hi,
> 
> I've been bit by a panic three times over the last month.
> 
> I'm currently running:
>    FreeBSD obiwan.piupiu.local 10.0-CURRENT FreeBSD 10.0-CURRENT #7 r251519: Sun Jun  9 22:37:09 CEST 2013 root_at_obiwan.piupiu.local:/usr/obj/usr/src/sys/OBIWAN  amd64
> 
> The panic message is:
>    panic: destroying non-empty racct: 1007616 allocated for resource 4
> 
> I quicky tried to track this down, but in half an hour I didn't get very
> far.  Resource 4 is RACCT_RSS.  The faulty process seems to be 66555
> which looks like a 32-bits cron(8) running in a jail.

I've seen a similar report in the past, but I was unable to figure out what could
be causing this.

Do you have a way to figure out exactly what cron job causes this?  Also,
could you try a workaround of adding any kind of rctl rule for that user
(this will cause the uidinfo to be "held" by the rctl rule, so uifree() won't try
to destroy its racct), e.g. "user:XXX:maxproc:log=1000000000", and see
if it panics in some other place?

Thanks!