Bug 86425 - [BYT]igt/gem_exec_big fails
Summary: [BYT]igt/gem_exec_big fails
Status: CLOSED FIXED
Alias: None
Product: DRI
Classification: Unclassified
Component: DRM/Intel (show other bugs)
Version: XOrg git
Hardware: Other All
: medium normal
Assignee: Intel GFX Bugs mailing list
QA Contact: Intel GFX Bugs mailing list
URL:
Whiteboard:
Keywords:
Depends on:
Blocks:
 
Reported: 2014-11-18 08:41 UTC by Guo Jinxian
Modified: 2015-04-08 08:26 UTC (History)
1 user (show)

See Also:
i915 platform:
i915 features:


Attachments
dmesg (103.36 KB, text/plain)
2014-11-18 08:41 UTC, Guo Jinxian
no flags Details
make error (185.90 KB, text/plain)
2014-11-20 02:53 UTC, Guo Jinxian
no flags Details
patch rebased onto -fixes (3.19 KB, patch)
2014-11-20 14:53 UTC, Daniel Vetter
no flags Details | Splinter Review
dmesg on nightly branch 0403 (113.95 KB, text/plain)
2015-04-03 07:29 UTC, Ding Heng
no flags Details

Description Guo Jinxian 2014-11-18 08:41:30 UTC
Created attachment 109658 [details]
dmesg

==System Environment==
--------------------------
Regression: Not sure,
I didn't find good commit on -fixes branch

Non-working platforms: BDW

==kernel==
--------------------------
origin/drm-intel-nightly: 2728bc8199164de1a27586f8c94e37d03120f8cd(works)
    drm-intel-nightly: 2014y-11m-17d-20h-21m-51s UTC integration manifest
origin/drm-intel-next-queued: 6ac0483b35bfc7546398d1198b65ed7be17f3cc5(works)
    drm/i915: Propagate invalid setcrtc cloning errors back to userspace
origin/drm-intel-fixes: 0485c9dc24ec0939b42ca5104c0373297506b555(fails)
    drm/i915: Kick fbdev before vgacon

==Bug detailed description==
igt/gem_exec_big fails

Output:
[root@x-byt05 tests]# ./gem_exec_big
IGT-Version: 1.8-gaa63fc7 (x86_64) (Linux: 3.18.0-rc5_drm-intel-fixes_0485c9_20141118+ x86_64)
Test assertion failure function exec, file gem_exec_big.c:97:
Failed assertion: tmp == gem_reloc[0].presumed_offset
error: 0 != 274432
[root@x-byt05 tests]# echo $?
99


==Reproduce steps==
---------------------------- 
1. ./gem_exec_big
Comment 1 Daniel Vetter 2014-11-18 09:00:08 UTC
Please run the testcase with debugging enabled, i.e.

# gem_exec_big --debug

and attach the complete output.
Comment 2 Chris Wilson 2014-11-18 09:12:00 UTC
Another obj->map_and_fenceable testcase.
Comment 3 Daniel Vetter 2014-11-18 09:22:42 UTC
Ok, I'm confused: Is this an issue on -fixes only or does this also happen on -nightly?

You're testing on -fixes but there's not mention of -nightly.
Comment 4 Daniel Vetter 2014-11-18 09:23:23 UTC
Again is this an issue with -fixes only or does it also happen in -nightly?
Comment 5 Daniel Vetter 2014-11-18 09:29:03 UTC
Please try to revert

commit e9d784d535e43777f9ff44d7e582946f4776c000
Author: Chris Wilson <chris@chris-wilson.co.uk>
Date:   Thu Nov 6 08:40:35 2014 +0000

    drm/i915: Fix obj->map_and_fenceable across tiling changes

on -fixes. We might have applied the thing to the wrong tree ...

This assumes my guess in comments 3/4 is right and this is only an issue on -fixes.
Comment 6 Guo Jinxian 2014-11-19 05:05:56 UTC
(In reply to Daniel Vetter from comment #4)
> Again is this an issue with -fixes only or does it also happen in -nightly?

It's a -fixes only issue. I reported the status on all branches. Please check this below. Thanks

origin/drm-intel-nightly: 2728bc8199164de1a27586f8c94e37d03120f8cd(works)
    drm-intel-nightly: 2014y-11m-17d-20h-21m-51s UTC integration manifest
origin/drm-intel-next-queued: 6ac0483b35bfc7546398d1198b65ed7be17f3cc5(works)
    drm/i915: Propagate invalid setcrtc cloning errors back to userspace
origin/drm-intel-fixes: 0485c9dc24ec0939b42ca5104c0373297506b555(fails)
    drm/i915: Kick fbdev before vgacon
Comment 7 Guo Jinxian 2014-11-19 05:44:36 UTC
(In reply to Daniel Vetter from comment #5)
> Please try to revert
> 
> commit e9d784d535e43777f9ff44d7e582946f4776c000
> Author: Chris Wilson <chris@chris-wilson.co.uk>
> Date:   Thu Nov 6 08:40:35 2014 +0000
> 
>     drm/i915: Fix obj->map_and_fenceable across tiling changes
> 
> on -fixes. We might have applied the thing to the wrong tree ...
> 
> This assumes my guess in comments 3/4 is right and this is only an issue on
> -fixes.
I reverted commit 0485c9dc24ec0939b42ca5104c0373297506b555 on latest -fixes(0485c9dc24ec0939b42ca5104c0373297506b555), the failure still able to reproduce.

[root@x-byt05 tests]# ./gem_exec_big
IGT-Version: 1.8-gaa63fc7 (x86_64) (Linux: 3.18.0-rc5_kcloud_3cf816_20141119+ x86_64)
Test assertion failure function exec, file gem_exec_big.c:97:
Failed assertion: tmp == gem_reloc[0].presumed_offset
error: 0 != 3592192
Comment 8 Daniel Vetter 2014-11-19 09:17:59 UTC
Same as with Bug 86426

Please cherry-pick

Author: Chris Wilson <chris@chris-wilson.co.uk>
Date:   Fri Oct 31 13:53:52 2014 +0000

    drm/i915: Only mark as map-and-fenceable when bound into the GGTT

on top of -fixes (without the revert) and retest.
Comment 9 Guo Jinxian 2014-11-20 02:53:37 UTC
Created attachment 109748 [details]
make error

(In reply to Daniel Vetter from comment #8)
> Same as with Bug 86426
> 
> Please cherry-pick
> 
> Author: Chris Wilson <chris@chris-wilson.co.uk>
> Date:   Fri Oct 31 13:53:52 2014 +0000
> 
>     drm/i915: Only mark as map-and-fenceable when bound into the GGTT
> 
> on top of -fixes (without the revert) and retest.

cherry-pick commit above on top of -fixes. the commit makes error. Please check error info in attachment. 


commit ca99bbb19e58d3c10af06096332cf41548a08f2e
Author:     Chris Wilson <chris@chris-wilson.co.uk>
AuthorDate: Fri Oct 31 13:53:52 2014 +0000
Commit:     Jinxian <jinxianx.guo@intel.com>
CommitDate: Thu Nov 20 08:56:07 2014 +0800

    drm/i915: Only mark as map-and-fenceable when bound into the GGTT

    We use the obj->map_and_fenceable hint for when we already have a
    valid mapping of this object in the aperture. This hint can only apply
    to the GGTT and not to the aliasing-ppGTT. One user of the hint is
    execbuffer relocation, which began to fail when it tried to follow the
    hint and perform the relocate through the non-existent GGTT mapping.

    Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=85671
    Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
    Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

commit 0485c9dc24ec0939b42ca5104c0373297506b555
Author:     Daniel Vetter <daniel.vetter@ffwll.ch>
AuthorDate: Fri Nov 14 10:09:49 2014 +0100
Commit:     Jani Nikula <jani.nikula@intel.com>
CommitDate: Mon Nov 17 10:18:21 2014 +0200

    drm/i915: Kick fbdev before vgacon
Comment 10 Daniel Vetter 2014-11-20 14:53:55 UTC
Created attachment 109756 [details] [review]
patch rebased onto -fixes

Sorry for missing the compile fail, rebased patch attached.
Comment 11 Chris Wilson 2014-12-21 19:56:17 UTC
Patch already upstream; presuming fixed.
Comment 12 Yi Sun 2014-12-25 00:54:34 UTC
(In reply to Chris Wilson from comment #11)
> Patch already upstream; presuming fixed.

It's still failed on latest -fixes branch, as Bug 86426
BTW, can we mark this bug a duplicated one as bug 86426 if they're totally the same.

root@x-bytm02:/GFX/Test/Intel_gpu_tools/intel-gpu-tools/tests# ./gem_exec_big
IGT-Version: 1.9-geb799b2 (x86_64) (Linux: 3.18.0-rc7_drm-intel-fixes_b0616c_20141224+ x86_64)
Test assertion failure function exec, file gem_exec_big.c:100:
Failed assertion: tmp == gem_reloc[0].presumed_offset
error: 0 != 8904704
Comment 13 Chris Wilson 2014-12-25 09:42:31 UTC
Can you please test with

commit e9d784d535e43777f9ff44d7e582946f4776c000
Author: Chris Wilson <chris@chris-wilson.co.uk>
Date:   Thu Nov 6 08:40:35 2014 +0000

    drm/i915: Fix obj->map_and_fenceable across tiling changes

and if that passes perform a bisection from that point.
Comment 14 Ding Heng 2014-12-29 07:25:21 UTC
(In reply to Chris Wilson from comment #13)
> Can you please test with

commit e9d784d535e43777f9ff44d7e582946f4776c000
> Author: Chris Wilson <chris@chris-wilson.co.uk>
Date:   Thu Nov 6 08:40:35
> 2014 +0000

    drm/i915: Fix obj->map_and_fenceable across tiling changes
> and if that passes perform a bisection from that point.

It still fails on BYT:
root@x-bytm02:/GFX/Test/Intel_gpu_tools/intel-gpu-tools/tests# ./gem_exec_big
IGT-Version: 1.9-geb799b2 (x86_64) (Linux: 3.18.0-rc3_heng_e9d784_20141229+ x86_64)
Test assertion failure function exec, file gem_exec_big.c:100:
Failed assertion: tmp == gem_reloc[0].presumed_offset
error: 0 != 8912896
root@x-bytm02:/GFX/Test/Intel_gpu_tools/intel-gpu-tools/tests# ./gem_exec_big
IGT-Version: 1.9-geb799b2 (x86_64) (Linux: 3.18.0-rc3_heng_e9d784_20141229+ x86_64)
Test assertion failure function exec, file gem_exec_big.c:100:
Failed assertion: tmp == gem_reloc[0].presumed_offset
error: 0 != 8904704
Comment 15 Jesse Barnes 2015-04-02 17:25:07 UTC
This one has me confused; first it only failed in -fixes but now it fails everywhere, but the commit that caused it to fail in -fixes, if reverted, does not fix things?  Can you re-test on the latest bits and see what happens?
Comment 16 Ding Heng 2015-04-03 07:28:45 UTC
(In reply to Jesse Barnes from comment #15)
> This one has me confused; first it only failed in -fixes but now it fails
> everywhere, but the commit that caused it to fail in -fixes, if reverted,
> does not fix things?  Can you re-test on the latest bits and see what
> happens?

This case now cause call trace on BYT with nightly branch latest commit 333cf6eed7cae0ee2d6bb1b4c4d421b94f84fc13(2015-04-03)
Comment 17 Ding Heng 2015-04-03 07:29:52 UTC
Created attachment 114843 [details]
dmesg on nightly branch 0403
Comment 18 Chris Wilson 2015-04-03 07:55:57 UTC
(In reply to Ding Heng from comment #16)
> (In reply to Jesse Barnes from comment #15)
> > This one has me confused; first it only failed in -fixes but now it fails
> > everywhere, but the commit that caused it to fail in -fixes, if reverted,
> > does not fix things?  Can you re-test on the latest bits and see what
> > happens?
> 
> This case now cause call trace on BYT with nightly branch latest commit
> 333cf6eed7cae0ee2d6bb1b4c4d421b94f84fc13(2015-04-03)

That's bug 88308.
Comment 19 Jesse Barnes 2015-04-03 16:11:19 UTC
So I guess this one is fixed now, but 88308 applies to BYT too.
Comment 20 Ding Heng 2015-04-08 08:26:22 UTC
(In reply to Chris Wilson from comment #18)
> (In reply to Ding Heng from comment #16)
> (In reply to Jesse Barnes from
> comment #15)
> > This one has me confused; first it only failed in -fixes
> but now it fails
> > everywhere, but the commit that caused it to fail in
> -fixes, if reverted,
> > does not fix things?  Can you re-test on the latest
> bits and see what
> > happens?
> 
> This case now cause call trace on BYT
> with nightly branch latest commit
>
> 333cf6eed7cae0ee2d6bb1b4c4d421b94f84fc13(2015-04-03)

That's bug 88308.

since we have bug 88308 track the call trace issue. close this bug.


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.