Bug 72104

Summary:

[BDW]igt/gem_pipe_control_store_loop causes gpu hangs

Product:

DRI

Reporter:

Guo Jinxian <jinxianx.guo>

Component:

DRM/Intel

Assignee:

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

Status:

CLOSED FIXED

QA Contact:

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

Severity:

major

Priority:

high

CC:

intel-gfx-bugs

Version:

XOrg git

Hardware:

Other

OS:

All

Whiteboard:

i915 platform:

i915 features:

Attachments:

Description	Flags
dmesg	none
error state	none
error state on IGT commit 8ac7b93b	none

Description Guo Jinxian 2013-11-28 09:16:21 UTC

Created attachment 89949 [details]
dmesg

System Environment:
--------------------------
Platform: Broadwell
kernel   (drm-intel-nexe-queued)8b4f49e03901e82898540bd1189333edf9588d74
Some additional commit info:
Author: Jesse Barnes <jbarnes@virtuousgeek.org>
Date:   Mon Nov 25 15:51:16 2013 -0800

    drm/i915: split fb allocation and initialization v2

    If we use a stolen buffer, our probe callback shouldn't allocate a new
    buffer; we should re-use the one from the BIOS instead if possible.

    v2: fix locking (Jesse)

Bug detailed description:
-----------------------------
Igt/gem_pipe_control_store_loop causes gpu hangs, and the os unable to reboot.

This is the first time to run igt test on Broadwell.

Error log:
[  531.524816] [drm] stuck on render ring
[  531.524864] [drm] GPU crash dump saved to /sys/class/drm/card0/error
[  531.524928] [drm] GPU hangs can indicate a bug anywhere in the entire gfx stack, including userspace.
[  531.525026] [drm] Please file a _new_ bug report on bugs.freedesktop.org against DRI -> DRM/Intel
[  531.525109] [drm] drm/i915 developers can then reassign to the right component if it's not a kernel issue.
[  531.525201] [drm] The gpu crash dump is required to analyze gpu hangs, so please always attach it.
Steps:
---------------------------
./gem_pipe_control_store_loop

Comment 1 Daniel Vetter 2013-11-28 13:18:31 UTC

Please attach the error state.

Comment 2 Guo Jinxian 2013-11-29 03:20:33 UTC

Created attachment 89969 [details]
error state

attached error state

Comment 3 Chris Wilson 2013-11-29 14:35:29 UTC

The blit and PIPE_CONTROL commands have not been updated for gen8 in gem_pipe_control_store_loop.c

Comment 4 Ben Widawsky 2013-12-05 23:46:13 UTC

Please retest with latest IGT

Comment 5 Guo Jinxian 2013-12-06 03:36:06 UTC

Retested on latest IGT(commit 8ac7b93b218af1544988331cd68c69ba4102c29f),this bug still cause gpu gangs.

Comment 6 Ben Widawsky 2013-12-06 04:17:52 UTC

(In reply to comment #5)
> Retested on latest IGT(commit 8ac7b93b218af1544988331cd68c69ba4102c29f),this
> bug still cause gpu gangs.

Can you attach the error state for the hang on 8ac7b93b218af1544988331cd68c69ba4102c29f

Comment 7 Guo Jinxian 2013-12-06 05:54:41 UTC

Created attachment 90337 [details]
error state on IGT commit 8ac7b93b

Comment 8 Ben Widawsky 2013-12-06 17:22:22 UTC

(In reply to comment #7)
> Created attachment 90337 [details]
> error state on IGT commit 8ac7b93b

Are you absolutely certain to properly rebuilt IGT? The error state indicates either you did not, or the code somehow isn't taking the gen8 path.s

Comment 9 Guo Jinxian 2013-12-10 06:38:26 UTC

Update test result on IGT commit(3c5423b9a74fe056cae1fce41fd3ad16584e9c8d), the test is passed, thanks.

[root@x-bdw01 tests]# ./gem_pipe_control_store_loop
Subtest fresh-buffer: SUCCESS
Subtest reused-buffer: SUCCESS

Comment 10 Daniel Vetter 2013-12-10 07:09:51 UTC

Yay!

Comment 11 Guo Jinxian 2013-12-10 07:22:15 UTC

test passed on IGT commit(3c5423b9a74fe056cae1fce41fd3ad16584e9c8d)

Comment 12 Elizabeth 2017-10-06 14:41:49 UTC

Closing old verified.

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.