Re: kern/37933: very frequent crashes after suspend/resume

["Steven M. Bellovin" <smb%cs.columbia.edu@localhost>]

On Sun, 10 Feb 2008 11:12:27 -0500
"Jared D. McNeill" <jmcneill%invisible.ca@localhost> wrote:

> Steven M. Bellovin wrote:
> > On Sat, 09 Feb 2008 21:16:56 -0500
> > "Jared D. McNeill" <jmcneill%invisible.ca@localhost> wrote:
> > 
> >> Are you using cpuctl to offline all APs before you suspend?
> >>
> > I hadn't -- should I have?  It does seem to prevent the problem; at
> > least, I wasn't able to cause any crashes when I tried that.  (To be
> > sure, I got another crash when I shut down the machine to try that
> > test; I haven't looked at that dump yet....)
> > 
> > For now, I'm going to put a cpuctl in my suspend/resume scripts --
> > it had reached the point where I was shutting down rather than
> > risking a second suspend.
> 
> I can't remember if I raised a PR on this or not, but there are
> issues with spinning up the APs as early as we do.
> 
> My powerd scripts simply do:
> 
>    /usr/sbin/cpuctl offline 1
>    [..]
>    sysctl -w machdep.sleep_state=3
>    [..]
>    /usr/sbin/cpuctl online 1
> 
That's what I'm doing now, too.  I'll run that way for a week -- call
it data-gathering -- and see if it improves the situation.  (I do have
a 'sleep 1' after the 'offline' command, to make sure there are no
problems with asynchrony.)


                --Steve Bellovin, http://www.cs.columbia.edu/~smb