Bug 85379

Summary:

[snb][drm] GPU HANG: ecode 0:0x85fffffd, in Xorg.bin

Product:

xorg

Reporter:

gedgon

Component:

Driver/intel

Assignee:

Chris Wilson <chris>

Status:

RESOLVED FIXED

QA Contact:

Intel GFX Bugs mailing list <intel-gfx-bugs>

Severity:

normal

Priority:

medium

Version:

unspecified

Hardware:

x86-64 (AMD64)

OS:

Linux (All)

Whiteboard:

i915 platform:

i915 features:

Attachments:

Description	Flags
error	none
dmesg	none
Xorg.0.log	none

Description gedgon 2014-10-23 18:23:59 UTC

Created attachment 108311 [details]
error

GPU hang at login since:

From de96fbf66cf9a0df85cfd1dad8cf786a1d6f8325
From: Chris Wilson <chris@chris-wilson.co.uk>
Date: Tue, 21 Oct 2014 12:20:29 +0100
Subject: sna/trapezoids: Prefer GPU for AddTraps()

linux 3.17.1
xorg-server 1.16.1
mesa 10.3.1
login manager: sddm 0.10
wm: compiz-bzr 3889

Unfortunately, there is no gpu hang with --enable-debug=full

Comment 1 gedgon 2014-10-23 18:24:23 UTC

Created attachment 108312 [details]
dmesg

Comment 2 gedgon 2014-10-23 18:24:51 UTC

Created attachment 108313 [details]
Xorg.0.log

Comment 3 Chris Wilson 2014-10-23 20:11:24 UTC

Your Xorg.0.log doesn't correspond with an upstream commit. Could you please verify the hang with xf86-video-intel.git, currently at:

commit e46f26b9e2671468e9accd2f69ffc22fb96b5849
Author: Chris Wilson <chris@chris-wilson.co.uk>
Date:   Wed Oct 22 19:37:42 2014 +0100

    sna/gen8: Load the sampler result directly into the fb write message

Comment 4 gedgon 2014-10-23 20:20:08 UTC

(In reply to Chris Wilson from comment #3)
> Your Xorg.0.log doesn't correspond with an upstream commit. Could you please
> verify the hang with xf86-video-intel.git, currently at:
> 
> commit e46f26b9e2671468e9accd2f69ffc22fb96b5849
> Author: Chris Wilson <chris@chris-wilson.co.uk>
> Date:   Wed Oct 22 19:37:42 2014 +0100
> 
>     sna/gen8: Load the sampler result directly into the fb write message

All logs coming from
From a8d1eaf6ee1fed0aeff681d33864e17cb904ddd3
From: Chris Wilson <chris@chris-wilson.co.uk>
Date: Wed, 22 Oct 2014 21:12:56 +0100
Subject: sna: Fix use of uninitialized value when handling migration for PRIME

also reproducible in e46f26b, bisected to

From de96fbf66cf9a0df85cfd1dad8cf786a1d6f8325
From: Chris Wilson <chris@chris-wilson.co.uk>
Date: Tue, 21 Oct 2014 12:20:29 +0100
Subject: sna/trapezoids: Prefer GPU for AddTraps()

Comment 5 Chris Wilson 2014-10-24 12:28:51 UTC

I suspect the actual culprit was b4872f74c05a92c91d7b57146eb35ed2ecc86d97 since that matches the error state. I thought it was fixed by

commit 2114b54a0b8a561dd7db1a38a87654571b4aa853
Author: Chris Wilson <chris@chris-wilson.co.uk>
Date:   Wed Oct 22 18:51:39 2014 +0100

    sna: Fix double application of batch shrinking to relocation targets

but requires

commit 4df0052a21efd744c4b8cb2409139ded6e45f5c8
Author: Chris Wilson <chris@chris-wilson.co.uk>
Date:   Fri Oct 24 08:08:23 2014 +0100

    sna: Update the reloc->delta for the shrunk batch

as well.

Comment 6 gedgon 2014-10-24 14:49:47 UTC

Thank you very much!

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.