On Mon, 20 Sep 2010, tom60 wrote:
> Hi Randy,
>
> Thank you for your reply. I enabled the core dump saving and I got two crash
> dumps.
>
> The two x4270 all have the same OS and were installed with the same OS
> release, snv143.
Could you please email one or both of them to me (if not, we will
need to work out how to better get relevant information)?
> I see the machine where panic occurred has "S3-support enable" in its
> /etc/power.conf, which was set by a test script doing Suspend & Resume test on
> a network driver on this machine. But no matter what the test script runs, the
> "mutex not owner" panic should not happen.
This is correct. This isn't a suspend/resume problem, but a lock
error on configuring CPUPM. AFAICT, this would only occur if there
were a code path that failed to acquire the lock, a code path that
could release the lock twice, or some corruption path (though
corruption rarely includes zeroing data). The corefile *should* help
me determine which of those paths is the likely culprit (and will be
needed for a solution, unless I can find one of the paths first).
>
> Attached is the extracted information from the core dump and the message log.
> Let me know if you need other help.
>
> Also, please let me know the CR number after you filed the bug.
Sure will. I will also put an email reference in the bug, so let me
know (privately, if desired) what email you would choose to use.
Cheers!
---- Randy
>
> Thanks.
>
> Tom
>
> --------------------------------------------------
> From: "Randy Fishel" <[email protected]>
> Sent: Saturday, September 18, 2010 2:49 PM
> To: "Tom Chen" <[email protected]>
> Cc: <[email protected]>
> Subject: Re: [pm-discuss] what is this panic?
>
> >
> >
> >
> > On Sat, 18 Sep 2010, Tom Chen wrote:
> >
> > > Hello,
> > >
> > > One of my x4270 with snv143 OS always panic after reboot. See the
> > > log below. What is this panic? it seems an error in power
> > > management. However, the other x4270 never has this panic. It is
> > > very strange. Is this a bad setup in this x86 machine's BIOS or one
> > > device driver not good? how do I disable pm?
> > >
> > > I failed to make a core dump since recent Solaris Express release
> > > apparently does not enable creation of core dump when panic happens.
> > > How do I enable?
> > >
> > > Tom
> > >
> >
> > To start, I would like to get a dump, as various values will be of
> > use in determining a solution, but this problem appears to be the
> > result of releasing the lock after setting CPU's to their max power
> > setting. Setting this level should not have occured withough first
> > acquiring the lock, so there is either a path where it isn't acquired,
> > or there is a path where it is released more than once (a debug kernel
> > would likely have tripped on an ASSERT). As another machine of the
> > same architecture is not doing this, the trigger is something to do
> > with a difference in OS (are they both running the same release?), or
> > some difference in settings between the two (does the failing machine
> > run from the CD or a previous release?).
> >
> > In any account, I see this as being a bug in some part of the CPU
> > power management path, and we desire a dump to check it out further.
> > The messages file, though, does indicate that a dump was taken, but it
> > is possible that it wasn't extracted because an extraction location
> > was not specified or doesn't exist. Use dumpadm(1m) to change or view
> > the settings. Note that the directory that is defined by 'Savecore
> > directory' must also exist. You can also extract this dump if you
> > have a different BE that will allow the machine to boot.
> >
> > Send me whatever you might be able to extract and I will see about
> > getting a bug filed. Also, I don't see offhand that there will be a
> > way to disable PM in the OS (if this is an always-panic-at-boot
> > problem), so you might have to check BIOS settings for CPU power
> > management settings to get the machine to stop panicing (check the
> > difference in BIOS PM settings between the machines).
> >
> > Cheers!
> >
> > ---- Randy
> >
_______________________________________________
pm-discuss mailing list
[email protected]
http://mail.opensolaris.org/mailman/listinfo/pm-discuss