Re: 6.0-BETA2 DRM/witness panic: Assertion j < 1000 failed at /usr/src/sys/kern/subr_witness.c:1513

[posted to -current rather than stable as the rest of the thread started
before 6.0 was released]

On Thu, 2005-10-06 at 13:55 -0400, John Baldwin wrote:
> On Thursday 06 October 2005 07:34 am, Gavin Atkinson wrote:
> > On Thu, 2005-09-29 at 12:10 -0400, John Baldwin wrote:
> > > On Tuesday 09 August 2005 12:02 pm, Gavin Atkinson wrote:
> > > > On Tue, 2005-08-09 at 10:50 -0400, John Baldwin wrote:
> > > > > On Tuesday 09 August 2005 07:57 am, Gavin Atkinson wrote:
> > > > > > Hi,
> > > > > >
> > > > > > I have a regular panic (3 times a day) on a system running
> > > > > > 6.0-BETA2 which was very stable under 5.4. dmesg available at
> > > > > > http://www.devrandom.co.uk/freebsd/dmesg-buffy-20050809 . The panic
> > > > > > seems to relate to DRM, I have a ATI Radeon QY RV100 7000/VE card.
> > > > >
> > > > > Do a 'show witness', it looks like witness has a cycle somehow. 
> > > > > Normally these can only occur if there is a cycle in the static lock
> > > > > order.
> > > >
> > > > http://www.devrandom.co.uk/freebsd/sh-witness-buffy-20050809
> > > >
> > > > Gavin
> > >
> > > Do you still get this panic on more recent 6.0?
> >
> > I've been running a 6.0 from Friday for a week now without seeing this
> > panic, so I suspect it's fixed.  However, I don't remember seeing any
> > commits which look like they were responsible for fixing it, is it
> > possible that it's just been masked by some other commit?  I'm happy to
> > step back and find the responsible commit, unless you're happy that it
> > was fixed for certain.
> >
> > Gavin
> 
> Well, I'm not sure how you could even have triggered the panic in the first 
> place.  If you can't reproduce it that is fine with me for now.


Sadly, it looks like I spoke too soon.  Since this last email, I was
happily running with the kernel mentioned above without any problems,
and an uptime of 40 days with the screensavers that used to trigger it
running all night.  I rebooted my machine a couple of days ago, and
since then (and with the exact same kernel/world) the panic is back.
I've updated the machine to 6-STABLE but the panic remains.

I guess I was just lucky in my month without panics - they definitely
still exist.  I wonder if the fact that it survived so long means
anything - e.g. is it a failure to initialise something that just by
chance was initialised correctly that time? I'm happy to do any more
investigating that you want.

Gavin