Unkillable process
Vladimir Grebenschikov
vova at sw.ru
Thu Oct 28 01:19:55 PDT 2004
В ср, 27/10/2004 в 15:31 -0400, Ken Smith пишет:
> On Wed, Oct 27, 2004 at 04:37:27PM +0400, Vladimir Grebenschikov wrote:
>
> > Looks like I should upgrade to 1.251:
> >
> > diff -u -r1.251 -r1.250
> > --- sys/kern/kern_exit.c 23 Oct 2004 11:20:26 -0000 1.251
> > +++ sys/kern/kern_exit.c 5 Oct 2004 18:51:11 -0000 1.250
> > @@ -403,7 +403,7 @@
> > * since their existence means someone is screwing up.
> > */
> > if (q->p_flag & P_TRACED) {
> > - q->p_flag &= ~(P_TRACED | P_STOPPED_TRACE);
> > + q->p_flag &= ~P_TRACED;
> > psignal(q, SIGKILL);
> > }
>
> Yes, but before you do... :-)
>
> If it's not too late can you do:
>
> ps -o f -l
>
> Or if anyone "succeeds" at wedging processes (especially if you
> already have the above patch applied) can you try this?
Sorry, it is too late, also I had no patch applied.
I guess my case can be reproduced by killing -KILL gdb attached to
stopped process. With updated kernel process being debugged disapper
with gdb.
> I have a trivial procedure that produces unkillable processes
> before the above patch, and this patch fixes that case. I've got
> one person reporting that he can still get wedged processes even
> with the above patch though so I'm looking for a bit more information.
>
> Thanks...
>
--
Vladimir B. Grebenschikov
SWsoft Inc. vova at sw-soft.com
More information about the freebsd-current
mailing list