Bug 770

Summary:	Hangs with chromium using CVS as of 20040531
Product:	Mesa	Reporter:	Eric Anholt <eta>
Component:	Drivers/DRI/r200	Assignee:	Eric Anholt <eric>
Status:	RESOLVED FIXED	QA Contact:
Severity:	normal
Priority:	high
Version:	git
Hardware:	x86 (IA32)
OS:	FreeBSD
Whiteboard:
i915 platform:		i915 features:
Bug Depends on:
Bug Blocks:	351, 1065, 1066

Description Eric Anholt 2004-06-20 15:08:32 UTC

I'm using a system with the following:
- X.Org release 6.7.0
- libGL from X.Org CVS head 20040620 (to support new DRI drivers due to bug 709)
- DRI drivers from Mesa CVS 20040531.
- radeon.ko 1.11.0 from FreeBSD-current

While most things work, I was able to hang the machine twice using chromium. 
Once was while testing render acceleration with Chromium running and doing an ls
-lR in a terminal, and the following messages were logged.

/var/log/messages:Jun 19 12:41:58 anholt kernel: error:
[drm:pid1368:radeon_check_and_fixup_packets] *ERROR* Invalid R200 cubic texture
offset
/var/log/messages-Jun 19 12:41:58 anholt kernel: error:
[drm:pid1368:radeon_emit_packets] *ERROR* Packet verification failed
/var/log/messages-Jun 19 12:41:58 anholt kernel: error:
[drm:pid1368:radeon_cp_cmdbuf] *ERROR* radeon_emit_packets failed
/var/log/messages:Jun 19 12:42:00 anholt kernel: error:
[drm:pid1369:radeon_check_and_fixup_packets] *ERROR* Invalid R200 cubic texture
offset
/var/log/messages-Jun 19 12:42:00 anholt kernel: error:
[drm:pid1369:radeon_emit_packets] *ERROR* Packet verification failed
/var/log/messages-Jun 19 12:42:00 anholt kernel: error:
[drm:pid1369:radeon_cp_cmdbuf] *ERROR* radeon_emit_packets failed

Then I tried with Render accel disabled, and was able to hang the system, this
time without the messages, by running two chromiums and a fire demo at once.

Any clues?

Comment 1 Alex Deucher 2004-06-21 06:13:38 UTC

multiple GL apps are currently busted on R200.  check the dri-devel archives for
Roland's thread about this. render probably triggers this as well since it uses
the 3d engine.

Comment 2 Eric Anholt 2004-08-11 15:15:54 UTC

Take this bug as mine, since I've been looking at it.  It's a major issue, which
anecdotes say sprouted between XFree86 4.3.0 and X.Org 6.7.0.  Mark it as a blocker.

Comment 3 Eric Anholt 2004-08-15 02:47:03 UTC

Keith and I spent tonight debugging this issue.  The problem is that we need to
make EmitState+LOAD_VBPTR+DRAW_VBUF sets atomic (and other EmitState+something
setes, as well, most likely) in terms of not dropping the lock in the middle due
to full buffers.   We've implemented kludges that have dealt with the problem
(testing with ipers) and a more elegant solution will come soon.

Comment 4 Eric Anholt 2004-08-16 18:48:27 UTC

A fix has been committed to Mesa CVS which fixes the hangs and rendering
glitches on R200 with multiple clients (which included Render acceleration) and
on rendering glitches on R100 with multiple clients.  This should be merged for
6.8.0.

Comment 5 Kevin E. Martin 2004-08-16 19:09:25 UTC

(In reply to comment #4)
> A fix has been committed to Mesa CVS which fixes the hangs and rendering
> glitches on R200 with multiple clients (which included Render acceleration) and
> on rendering glitches on R100 with multiple clients.  This should be merged for
> 6.8.0.

This issue was discussed on the release wranglers call this morning and was on
the blocker list prior to the code freeze.  Please go ahead and merge the
changes from Mesa CVS.

Comment 6 Eric Anholt 2004-08-17 13:23:34 UTC

Changes have been pushed in to X.Org CVS.

Comment 7 Adam Jackson 2009-08-24 12:22:30 UTC

Mass version move, cvs -> git

Use of freedesktop.org services, including Bugzilla, is subject to our Code of Conduct. How we collect and use information is described in our Privacy Policy.