New upstream sysvinit tarball version 2.87dsf created
Scott James Remnant
scott at canonical.com
Tue Jul 14 11:23:03 UTC 2009
On Tue, 2009-07-14 at 08:03 -0300, Henrique de Moraes Holschuh wrote:
> On Tue, 14 Jul 2009, Scott James Remnant wrote:
> > > > I noticed a strange case with killall, it seems to cause the Upstart
> > > > rc.conf job to get SIGSTOP/SIGCONT repeatedly?
> > >
> > > killall5 signals all processes with SIGSTOP before deciding what to
> > > kill, to avoid new processes popping up while killall5 is running.
> > >
> > Interesting ... Upstart will send SIGCONT to any supervised process that
> > raises SIGSTOP, so this is likely working against you here.
>
> I'd get pretty mad if something kept sending SIGCONT to stuff I stopped for
> debugging, etc.
>
> > I'm not entirely sure why I send SIGCONT tbh.
>
> Unless you use SIGSTOP yourself in upstart, or you're trying to work around
> some weird userspace util (I've seen one, once, that used SIGSTOP/CONT to
> throttle down CPU usage) -- in which case it should be an optional feature
> IMHO -- I'd advice you to not send SIGCONT.
>
Sorry I was mistaken.
Upstart doesn't send SIGCONT, I was confusing it with the ptrace code
where it has to re-deliver the SIGSTOP signal.
It actually looks like the sendsigs signal itself is sending SIGCONT,
ie. it repeatedly sends SIGSTOP/SIGCONT to all processes over and over
again. Is that right?
Scott
--
Scott James Remnant
scott at canonical.com
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 197 bytes
Desc: This is a digitally signed message part
URL: <http://lists.alioth.debian.org/pipermail/initscripts-ng-devel/attachments/20090714/072129af/attachment.pgp>
More information about the initscripts-ng-devel
mailing list