Bug 84207

It impacts SNB+ platforms.

The good commit sha does't exist on current -nightly 
So please paste the commit subject along with sha.

Also, the bad commit is a Docbook integration, so I assume there was still a gap beween bad and good commit right?

good commit:9c787942907face82da505c2c5493998b56cfc5a

Yeah, I big gap between those 2. Could you please go a bit deep on this bisect, trying to find the offending commit?

The bisect is difficult. I bisect it and run each step 10 rounds, shows  9430dfa67d7 is the first bad commit. Then test on commit b478e336b3e755057, it also has this issue, fail rate: 1/25.

Could you run with CONFIG_DEBUG_PAGEALLOC, CONFIG_DEBUG_OBJECTS, CONFIG_SLUB_DEBUG_ON, CONFIG_PROVE_LOCKING and CONFIG_DEBUG_LIST enabled?

Created attachment 106898 [details]
dmesg

(In reply to comment #6)
> Could you run with CONFIG_DEBUG_PAGEALLOC, CONFIG_DEBUG_OBJECTS,
> CONFIG_SLUB_DEBUG_ON, CONFIG_PROVE_LOCKING and CONFIG_DEBUG_LIST enabled?

Attached the dmesg with these settings.

Hmm, that shows no signs of poisoning, so I think we can rule out a double-free of the i915_mm_struct. Could you gdb vmlinux (it will be in the root directory of your build tree) and "list * mmu_notifier_unregister+0x18"

You run nothing else after boot? (i.e. you boot the machine and the first thing you run is gem_render_copy_redux in a loop?)

I can't see how that would trigger userptr and creation of the i915_mm_struct.

Oh, the userptr is from libdrm. And the invalid dereference is an error pointer.

*** Bug 84358 has been marked as a duplicate of this bug. ***

commit f2775039b1d2f3c24876622e4528604496de8abc
Author: Chris Wilson <chris@chris-wilson.co.uk>
Date:   Fri Sep 26 10:22:33 2014 +0100

    igt/gem_userptr_blits: Test interruptible create-destroy
    
    In order to exercise https://bugs.freedesktop.org/show_bug.cgi?id=84207
    we need to interrupt the mmu_notifier_register with a signal. This is
    likely to be quite difficult, but let's just try running the
    create-destroy test in an interruptible loop for 5s.
    
    Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

http://patchwork.freedesktop.org/patch/34166/

(In reply to comment #13)
> http://patchwork.freedesktop.org/patch/34166/

Apply this patch and test on latest igt, run 30 cycles, it works well.

OOO on Sep 28 – Sep 30. Sorry for no mail access.

Best Regards,
Shuo

Test below cause system hang on HSW IVB

[root@x-hsw27 tests]# time ./gem_userptr_blits --run-subtest create-destroy-sync
IGT-Version: 1.8-g32a0308 (x86_64) (Linux: 3.17.0-rc6_drm-intel-fixes_c84db7_20140929+ x86_64)
Aperture size is 2048 MiB
Total RAM is 7669 MiB
Testing unsynchronized mappings...
Testing synchronized mappings...




^C

[root@x-ivb9 tests]# time ./gem_userptr_blits --run-subtest create-destroy-sync
IGT-Version: 1.8-g32a0308 (x86_64) (Linux: 3.17.0-rc6_drm-intel-nightly_7101d8_20140929+ x86_64)
Aperture size is 2048 MiB
Total RAM is 3836 MiB
Testing unsynchronized mappings...
Testing synchronized mappings...






^C^C

(In reply to comment #14)
> (In reply to comment #13)
> > http://patchwork.freedesktop.org/patch/34166/
> 
> Apply this patch and test on latest igt, run 30 cycles, it works well.

Pushed to drm-intel-fixes as

commit 72e59c89131606106452f1773a316b90d9f54423
Author: Chris Wilson <chris@chris-wilson.co.uk>
Date:   Fri Sep 26 10:31:02 2014 +0100

    drm/i915: Do not store the error pointer for a failed userptr registration

(In reply to comment #16)
> Test below cause system hang on HSW IVB

Please file new reports for them, AFAICT these are separate issues.

(In reply to comment #17) 
> (In reply to comment #16)
> > Test below cause system hang on HSW IVB
> 
> Please file new reports for them, AFAICT these are separate issues.

Nope. They were test failures added to explicitly reproduce this bug, see comment 12.

(In reply to comment #17)
> (In reply to comment #14)
> > (In reply to comment #13)
> > > http://patchwork.freedesktop.org/patch/34166/
> > 
> > Apply this patch and test on latest igt, run 30 cycles, it works well.
> 
> Pushed to drm-intel-fixes as
> 
> commit 72e59c89131606106452f1773a316b90d9f54423
> Author: Chris Wilson <chris@chris-wilson.co.uk>
> Date:   Fri Sep 26 10:31:02 2014 +0100
> 
>     drm/i915: Do not store the error pointer for a failed userptr
> registration
> 

scratch that, it's drm-intel-next-fixes as

commit e9681366ea9e76ab8f75e84351f2f3ca63ee542c
Author: Chris Wilson <chris@chris-wilson.co.uk>
Date:   Fri Sep 26 10:31:02 2014 +0100

    drm/i915: Do not store the error pointer for a failed userptr registration



> (In reply to comment #16)
> > Test below cause system hang on HSW IVB
> 
> Please file new reports for them, AFAICT these are separate issues.

(In reply to comment #18)
> Nope. They were test failures added to explicitly reproduce this bug, see
> comment 12.

Reopen?

commit e9681366ea9e76ab8f75e84351f2f3ca63ee542c
Author: Chris Wilson <chris@chris-wilson.co.uk>
Date:   Fri Sep 26 10:31:02 2014 +0100

    drm/i915: Do not store the error pointer for a failed userptr registration

    Fixes regression from commit ad46cb533d586fdb256855437af876617c6cf609
    Author: Chris Wilson <chris@chris-wilson.co.uk>
    Date:   Thu Aug 7 14:20:40 2014 +0100
    
        drm/i915: Prevent recursive deadlock on releasing a busy userptr
    
    Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=84207
    Testcase: igt/gem_render_copy_redux
    Testcase: igt/gem_userptr_blits/create-destroy-sync
    Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
    Cc: Jacek Danecki <jacek.danecki@intel.com>
    Cc: "Gong, Zhipeng" <zhipeng.gong@intel.com>
    Cc: Jacek Danecki <jacek.danecki@intel.com>
    Cc: "Ursulin, Tvrtko" <tvrtko.ursulin@intel.com>
    Cc: stable@vger.kernel.org
    Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
    Signed-off-by: Daniel Vetter <daniel.vetter@intel.com>

Note that the commit references both this bugzilla and the new test case from c12 that qa reported failure for in c17.

Verified.Fixed.

Closing old verified+fixed.