Xorg 7.4 crashes on exit

[Kevin Oberman]

> Date: Tue, 31 Mar 2009 16:16:57 +0100
> From: "N.J. Mann" <njm at njm.me.uk>
> 
> Hi Robert,
> 
> 
> In message <1238445388.8491.358.camel at balrog.2hip.net>,
> 	Robert Noland (rnoland at FreeBSD.org) wrote:
> > On Mon, 2009-03-30 at 13:11 -0700, Kevin Oberman wrote:
> > > > From: Robert Noland <rnoland at FreeBSD.org>
> > > > Date: Mon, 30 Mar 2009 14:54:45 -0500
> [snip]
> > > > 
> > > > Is this on CURRENT or STABLE and do you know at least approximately when
> > > > you started seeing the issue?
> > > > 
> > > 
> > > It is on stable and I THINK it started when I updated my system to
> > > 7.2-PRERELEASE last week. My previous Stable was on February 7 and it
> > > pre-dated the Xorg-1.5.3 server update on the system.
> > > 
> > > I normally update ports daily, but I did delay the Xorg update for a
> > > week or two to let dust settle.
> > 
> > There are a lot of changed in that window...  I really hope that this
> > doesn't fix it, but could you try reverting:
> > 
> > balrog% svn diff -c r189669
> > Index: drm_drv.c
> > ===================================================================
> > --- drm_drv.c   (revision 189668)
> > +++ drm_drv.c   (revision 189669)
> > @@ -131,7 +131,7 @@
> >         .d_poll =       drm_poll,
> >         .d_mmap =       drm_mmap,
> >         .d_name =       "drm",
> > -       .d_flags =      D_TRACKCLOSE | D_NEEDGIANT
> > +       .d_flags =      D_TRACKCLOSE
> >  };
> >  
> >  static struct drm_msi_blacklist_entry drm_msi_blacklist[] = {
> 
> I was not sure whether you wanted Kevin, me or both of us to try the
> above, so I tried it anyway. :-)  It has no visible affect here at all.
> I double checked I had patched the right file and even re-compiled
> (after doing a rm -rf /usr/obj and a make clean) and re-installed the
> resulting kernel.  No change.
> 
> Are there any debug messages I can enable to help you out here?

For me and my Radeon, the change seems to have "fixed" the problem. 

I have rebooted several times, including after loading the system with a
variety of graphics oriented tasks (Flash, MPEGs, WMV, ...), I have been
unable to trigger the problem with the NEEDGIANT. I don't have witness
in my kernel and really would have problems with having it there on this
system this week as I am on-call. Yuck!
-- 
R. Kevin Oberman, Network Engineer
Energy Sciences Network (ESnet)
Ernest O. Lawrence Berkeley National Laboratory (Berkeley Lab)
E-mail: oberman at es.net			Phone: +1 510 486-8634
Key fingerprint:059B 2DDF 031C 9BA3 14A4  EADA 927D EBB3 987B 3751