Bug 52945

Summary: [SNB]Piglit texturing_depthstencil-render-miplevels cases cause system hang with calltrace on nightly testing
Product: DRI Reporter: lu hua <huax.lu>
Component: DRM/IntelAssignee: Daniel Vetter <daniel>
Status: CLOSED FIXED QA Contact: Intel GFX Bugs mailing list <intel-gfx-bugs>
Severity: major    
Priority: medium CC: ben, chris, daniel, jbarnes, kenneth, xunx.fang
Version: unspecified   
Hardware: All   
OS: Linux (All)   
Whiteboard:
i915 platform: i915 features:
Attachments:
Description Flags
dmesg
none
netconsole log
none
netconsole
none
netconsole log none

Description lu hua 2012-07-30 08:53:14 UTC 
Created attachment 64940 [details]
dmesg

System Environment:
--------------------------
Arch:             i386
Platform:         Sandybridge
Libdrm:	(master)libdrm-2.4.37-20-gebd7904877d08525beb5039e4ea2f5b6c0a7c23f
Mesa:	(8.0)539a02b9a73276a6d012436e296cfab55d5257a3
Xserver:(server-1.12-branch)xorg-server-1.12.3
Xf86_video_intel:(master)2.20.2-4-g3d45f0affe263985f440e144203ed7cbb3803696
Libva:	(master)8b2835e883936d28388b72b92d3d492d0c32c57e
Libva_intel_driver:(master)6c408cf08906992b602d7f8a0d9a3c53a71c99df
Kernel:	(drm-intel-fixes) 6af2d180f82151cf3d58952e35a4f96e45bc453a

Bug detailed description:
---------------------
It happens on sandybridge with mesa 8.0 branch.It doesn't happen on mesa master branch.
This case has another bug(Bug 50271).
Calltrace:
[  222.581554]  [<ffffffffa00bef6a>] intel_unpin_fb_obj+0x38/0x3a [i915]
[  222.581597]  [<ffffffffa00c1a40>] intel_unpin_work_fn+0x1f/0x74 [i915]
[  222.581635]  [<ffffffff81041b98>] process_one_work+0x1da/0x33c
[  222.581667]  [<ffffffff81041fdf>] worker_thread+0x152/0x210
[  222.581698]  [<ffffffff81041e8d>] ? rescuer_thread+0x169/0x169
[  222.581731]  [<ffffffff81045d07>] kthread+0x86/0x8e
[  222.581761]  [<ffffffff813cf794>] kernel_thread_helper+0x4/0x10
[  222.581794]  [<ffffffff81045c81>] ? kthread_freezable_should_stop+0x52/0x52
[  222.581831]  [<ffffffff813cf790>] ? gs_change+0xb/0xb
[  222.581876] Code: c4 10 8b 45 d8 49 83 3c 24 00 eb db 41 59 41 5a 5b 41 5c 41 5d 41 5e 5d c3 8b 97 f0 00 00 00 55 48 89 e5 f7 c2 00 e0 01 00 75 02 <0f> 0b 48 83 bf 98 00 00 00 00 75 02 0f 0b 89 d0 81 e2 ff 1f fe
[  222.581979] RIP  [<ffffffffa00aff21>] i915_gem_object_unpin+0x12/0x4e [i915]
[  222.582017]  RSP <ffff880234899df0>

Reproduce steps:
----------------------------
1.   xinit
2.  ./bin/depthstencil-render-miplevels 146 s=z24_s8_d=z24 -auto
Comment 1 Chris Wilson 2012-07-30 09:22:07 UTC 
unpin leak following GPU hang, demoting priority as it is a secondary bug.

Can you please look at whether this is a kernel regression, or whether this bug has been dormant for some time? It looks that pageflipping over the hang might be connected to the leak.
Comment 2 lu hua 2012-07-31 06:17:48 UTC 
Created attachment 64977 [details]
netconsole log
Comment 3 lu hua 2012-08-01 07:00:53 UTC 
When nightly testing run texturing_depthstencil-render-miplevels cases, system hangs, and has calltrace in dmesg.I can't reproduce it via manually test.
This issue blocks nightly testing.
Comment 4 Chris Wilson 2012-08-04 11:39:25 UTC 
The netconsole log indicates that the kernel bug is not blocking your testing, but that the mesa bug is killing the GPU beyond recovery. You need to escalate the priority of *that* bug as that is stopping the testing.
Comment 5 Chris Wilson 2012-09-15 09:30:13 UTC 
Worksforme on dinq with the i965_dri.so blorp fixes.
Comment 6 lu hua 2012-09-17 07:35:44 UTC 
It still happens on -fixes kernel(3d840a13a2f9605) and -queued kernel(commit:a0db295dcd).
Comment 7 lu hua 2012-09-17 07:36:43 UTC 
Created attachment 67269 [details]
netconsole
Comment 8 Jesse Barnes 2012-12-11 19:20:38 UTC 
Is this still an issue with current kernels and mesa 8.0?
Comment 9 Daniel Vetter 2012-12-11 19:29:23 UTC 
Wasn't miplevel rendering broken rather often in mesa itself? Add random mesa victim to clarify the situation here ... Ken, which version do we need to avoid all the hangs due to alignment constraints for miplevel rendering?
Comment 10 Kenneth Graunke 2012-12-11 20:15:23 UTC 
Mesa 8.0.5 or 9.0 should be sufficient to avoid hangs, but you will get some failures.  Eric fixed a number of failures in master.
Comment 11 lu hua 2012-12-12 02:26:12 UTC 
Created attachment 71369 [details]
netconsole log

It still happens on -fixes kernel(commit 539526b4137bc0e7a8806c38c8522f226814a0e6)

Disable RC6, this issue goes away.
Comment 12 Chris Wilson 2012-12-12 09:28:29 UTC 
One thing that would be very useful is seeing if a change in mesa to use a different code path makes any difference (i.e. 8.0 vs 9.0). That would help narrow the search for a workaround.

*** This bug has been marked as a duplicate of bug 50619 ***
Comment 13 Elizabeth 2017-10-06 14:48:48 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.