Nicolai Hähnle a8d81b3626 drm/amdgpu: hold reference to fences in amdgpu_sa_bo_new (v2)
An arbitrary amount of time can pass between spin_unlock and
fence_wait_any_timeout, so we need to ensure that nobody frees the
fences from under us.

A stress test (rapidly starting and killing hundreds of glxgears
instances) ran into a deadlock in fence_wait_any_timeout after
about an hour, and this race condition appears to be a plausible
cause.

v2: agd: rebase on upstream

Signed-off-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Cc: stable@vger.kernel.org
2016-02-10 14:07:31 -05:00
..
2016-01-31 16:55:04 -08:00
2016-01-20 18:42:30 -08:00
2016-01-20 18:16:29 -08:00
2016-01-20 10:15:21 -08:00
2016-01-20 17:30:20 -08:00
2016-01-13 10:23:36 -08:00
2016-01-31 17:00:27 -08:00
2016-01-12 12:17:28 +01:00
2016-01-20 18:42:30 -08:00
2016-01-22 12:04:21 -08:00
2016-01-22 18:04:28 -05:00
2016-01-20 17:09:18 -08:00
2016-01-12 01:12:40 +01:00
2016-01-20 17:09:18 -08:00
2016-01-15 12:14:47 -08:00
2016-01-18 12:10:45 -08:00
2016-01-15 13:18:47 -08:00
2016-01-31 17:09:39 -08:00
2016-01-31 17:09:39 -08:00
2016-01-28 08:32:42 -08:00
2016-01-22 18:04:28 -05:00