On 2017-12-13 08:44 PM, Andrey Grodzovsky wrote:
> With introduction of amdgpu_gpu_recovery we don't need any more
> to rely on amdgpu_lockup_timeout == 0 for disabling GPU reset.
>
> Signed-off-by: Andrey Grodzovsky <andrey.grodzovsky at amd.com>
Since this change landed, I'm once again unable to finish a piglit run
on my development machine, see the attached dmesg output (happens pretty
quickly, after ~5% of piglit tests have run). I realized that with
lockup_timeout != 0, the
WARN_ON_ONCE(bo->tbo.mem.mem_type == TTM_PL_SYSTEM);
at the top of amdgpu_bo_gpu_offset has been triggering since the 4.15
development cycle. See the bisection result below. Note that I'm not
100% sure this is the correct guilty commit, since it's probably been
the most painful bisection I've ever done so far (14 skips, had to
revert 4 commits causing other regressions). But I'm quite sure this
regression happened in the
84d43463a2d09c28c9222fbb7d1082c078e2523a..3f3333f8a0e90ac26f84ed7b0aa344efce695c08
range.
3f3333f8a0e90ac26f84ed7b0aa344efce695c08 is the first bad commit
commit 3f3333f8a0e90ac26f84ed7b0aa344efce695c08
Author: Christian König <christian.koenig at amd.com>
Date: Thu Aug 3 14:02:13 2017 +0200
drm/amdgpu: track evicted page tables v2
Instead of validating all page tables when one was evicted,
track which one needs a validation.
v2: simplify amdgpu_vm_ready as well
Signed-off-by: Christian König <christian.koenig at amd.com>
Reviewed-by: Alex Deucher <alexander.deucher at amd.com> (v1)
Reviewed-by: Chunming Zhou <david1.zhou at amd.com>
--
Earthling Michel Dänzer | http://www.amd.com
Libre software enthusiast | Mesa and X developer
-------------- next part --------------
A non-text attachment was scrubbed...
Name: amdgpu_bo_gpu_offset-WARN.diff
Type: text/x-patch
Size: 28301 bytes
Desc: not available
URL: <https://lists.freedesktop.org/archives/amd-gfx/attachments/20171219/426d4056/attachment-0001.bin>
