[Intel-gfx] [PATCH v2] Fix pointer tests in error-handling paths

[Maarten Lankhorst]

Hey,

Op 27-01-16 om 13:41 schreef Tvrtko Ursulin:
>
> On 27/01/16 12:33, Maarten Lankhorst wrote:
>> Hey,
>>
>> Op 25-01-16 om 18:57 schreef Dave Gordon:
>>> In the error-handling paths of i915_gem_do_execbuffer() and
>>> intel_crtc_page_flip(), the local pointer-to-request variables
>>> were expected to be either valid pointers or NULL. Since
>>>
>>>    2682708 drm/i915: simplify allocation of driver-internal requests
>>>
>>> they could also be ERR_PTR() values, so the tests need to be
>>> updated to accommodate this case.
>>>
>>> v2:    Added testcase (Chris Wilson)
>>>
>>> Testcase: igt/gem_close_race
>>> Signed-off-by: Dave Gordon <david.s.gordon at intel.com>
>>> Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin at intel.com>
>> The testcase you mentioned doesn't work, it crashes on my machine instead of running to completion or triggering the bug.
>
> It worked for me but I think I was running with GuC enabled at the time so that might be the differenct.
>
> gem_close_race in general worked for me just yesterday.
It crashes in __lll_unlock_elision called from pthread_mutex_unlock, but I don't see anything weird there that could cause it.
Broadwell with the exact same config works.

It's tempting to blame libc, it looks like a bug in its tsx stuff.

~Maarten