[Mesa-dev] [RFC 0/3] Fix several segmentation faults related to depth buffer

[Alon Levy]

On Tue, Sep 06, 2011 at 03:11:25PM -0300, Eugeni Dodonov wrote:
> On Mon, Sep 5, 2011 at 16:37, Alon Levy <alevy at redhat.com> wrote:
> 
> > Fixes three segmentation faults where it is wrongly assumed the depth
> > buffer's
> > renderbuffer's region is not NULL.
> >
> > 735794 - verified to be fixed by these three patches (these three apply
> > to 7.11 almost cleanly, just the second requires minor changes).
> >
> > Related RHBZs (maybe solved as well, didn't reproduce):
> >  734183 - [abrt] gnome-shell-3.1.4-2.gite7b9933.fc16: prepare_depthbuffer
> >  717140 -[abrt] compiz-0.9.4-2.fc15: prepare_depthbuffer
> >
> > Alon Levy (3):
> >  i965: prepare_depthbuffer: fix segfault, rhbz#735794
> >  i965: prepare_depthbuffer: don't update NULL region'ed surface,
> >    rhbz#735794
> >  i965: emit_depthbuffer: fix segfault, rhbz#735794
> >
> >  src/mesa/drivers/dri/i965/brw_misc_state.c       |    7 +++++--
> >  src/mesa/drivers/dri/i965/brw_wm_surface_state.c |    3 +++
> >  2 files changed, 8 insertions(+), 2 deletions(-)
> >
> >
> Hi Alon,
> 
> those patches fix the crash you are having, but I am wondering why the crash
> happened in the first place. E.g., we should not get to the situation where
> region is NULL under normal cases.
> 
> From what it looks like, it could happen when GPU was hung for example.
> 
> Could you open a bug on freedesktop.org bugzilla and attach dmesg and
> /sys/kernel/debug/dri/0/i915_error_state, so we could look into it further?
> 

I haven't tried to reproduce, sorry, I will probably not get to it this week either.

I do have several other display related issues (reverted my patches since, running with
the recently updated mesa package in F16), but have not gotten to tracing them.

> -- 
> Eugeni Dodonov