Bug 106707

The warning is that it took so long to set the test up, the fence expired before we had done so; thus the order of execution was indeterminate.

It usually times out during the earlier construction and skips.

https://intel-gfx-ci.01.org/tree/drm-tip/drmtip_56/fi-bxt-dsi/igt@gem_exec_schedule@deep-bsd.html

(gem_exec_schedule:2059) igt_vgem-CRITICAL: Test assertion failure function vgem_fence_signal, file ../lib/igt_vgem.c:193:
(gem_exec_schedule:2059) igt_vgem-CRITICAL: Failed assertion: __vgem_fence_signal(fd, fence) == 0
(gem_exec_schedule:2059) igt_vgem-CRITICAL: error: -110 != 0
Subtest deep-bsd failed.

Seen on BSW: https://intel-gfx-ci.01.org/tree/drm-tip/drmtip_67/fi-bsw-n3050/igt@gem_exec_schedule@deep-render.html

(gem_exec_schedule:1455) igt_vgem-CRITICAL: Test assertion failure function vgem_fence_signal, file ../lib/igt_vgem.c:193:
(gem_exec_schedule:1455) igt_vgem-CRITICAL: Failed assertion: __vgem_fence_signal(fd, fence) == 0
(gem_exec_schedule:1455) igt_vgem-CRITICAL: error: -110 != 0
Subtest deep-render failed.

Also seen on SKL: https://intel-gfx-ci.01.org/tree/drm-tip/drmtip_69/fi-skl-gvtdvm/igt@gem_exec_schedule@deep-vebox.html

(gem_exec_schedule:1459) igt_vgem-CRITICAL: Test assertion failure function vgem_fence_signal, file ../lib/igt_vgem.c:193:
(gem_exec_schedule:1459) igt_vgem-CRITICAL: Failed assertion: __vgem_fence_signal(fd, fence) == 0
(gem_exec_schedule:1459) igt_vgem-CRITICAL: error: -110 != 0
Subtest deep-vebox failed.

https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_4519/fi-kbl-8809g/igt@amdgpu_amd_prime@amd-to-i915.html

(amd_prime:4831) igt_vgem-CRITICAL: Test assertion failure function vgem_fence_signal, file ../lib/igt_vgem.c:193:
(amd_prime:4831) igt_vgem-CRITICAL: Failed assertion: __vgem_fence_signal(fd, fence) == 0
(amd_prime:4831) igt_vgem-CRITICAL: error: -110 != 0
Subtest amd-to-i915 failed.

(In reply to Martin Peres from comment #5)
> https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_4519/fi-kbl-8809g/
> igt@amdgpu_amd_prime@amd-to-i915.html
> 
> (amd_prime:4831) igt_vgem-CRITICAL: Test assertion failure function
> vgem_fence_signal, file ../lib/igt_vgem.c:193:
> (amd_prime:4831) igt_vgem-CRITICAL: Failed assertion:
> __vgem_fence_signal(fd, fence) == 0
> (amd_prime:4831) igt_vgem-CRITICAL: error: -110 != 0
> Subtest amd-to-i915 failed.

Not a common problem. These are individual test setups that are just too slow on CI.

(In reply to Chris Wilson from comment #6)
> (In reply to Martin Peres from comment #5)
> > https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_4519/fi-kbl-8809g/
> > igt@amdgpu_amd_prime@amd-to-i915.html
> > 
> > (amd_prime:4831) igt_vgem-CRITICAL: Test assertion failure function
> > vgem_fence_signal, file ../lib/igt_vgem.c:193:
> > (amd_prime:4831) igt_vgem-CRITICAL: Failed assertion:
> > __vgem_fence_signal(fd, fence) == 0
> > (amd_prime:4831) igt_vgem-CRITICAL: error: -110 != 0
> > Subtest amd-to-i915 failed.
> 
> Not a common problem. These are individual test setups that are just too
> slow on CI.

Not sure I understand your explanation. Can you rephrase?

Also, what would you suggest I do for this issue? Should I create a new bug?

(In reply to Martin Peres from comment #7)
> (In reply to Chris Wilson from comment #6)
> > (In reply to Martin Peres from comment #5)
> > > https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_4519/fi-kbl-8809g/
> > > igt@amdgpu_amd_prime@amd-to-i915.html
> > > 
> > > (amd_prime:4831) igt_vgem-CRITICAL: Test assertion failure function
> > > vgem_fence_signal, file ../lib/igt_vgem.c:193:
> > > (amd_prime:4831) igt_vgem-CRITICAL: Failed assertion:
> > > __vgem_fence_signal(fd, fence) == 0
> > > (amd_prime:4831) igt_vgem-CRITICAL: error: -110 != 0
> > > Subtest amd-to-i915 failed.
> > 
> > Not a common problem. These are individual test setups that are just too
> > slow on CI.
> 
> Not sure I understand your explanation. Can you rephrase?
> 
> Also, what would you suggest I do for this issue? Should I create a new bug?

It's a separate issue for the amdgpu_amd_prime test. We need to tune the setup to run within 10s or use a sw_sync fence instead of a vgem plug.

(In reply to Chris Wilson from comment #8)
> (In reply to Martin Peres from comment #7)
> > (In reply to Chris Wilson from comment #6)
> > > (In reply to Martin Peres from comment #5)
> > > > https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_4519/fi-kbl-8809g/
> > > > igt@amdgpu_amd_prime@amd-to-i915.html
> > > > 
> > > > (amd_prime:4831) igt_vgem-CRITICAL: Test assertion failure function
> > > > vgem_fence_signal, file ../lib/igt_vgem.c:193:
> > > > (amd_prime:4831) igt_vgem-CRITICAL: Failed assertion:
> > > > __vgem_fence_signal(fd, fence) == 0
> > > > (amd_prime:4831) igt_vgem-CRITICAL: error: -110 != 0
> > > > Subtest amd-to-i915 failed.
> > > 
> > > Not a common problem. These are individual test setups that are just too
> > > slow on CI.
> > 
> > Not sure I understand your explanation. Can you rephrase?
> > 
> > Also, what would you suggest I do for this issue? Should I create a new bug?
> 
> It's a separate issue for the amdgpu_amd_prime test. We need to tune the
> setup to run within 10s or use a sw_sync fence instead of a vgem plug.

Thanks! I will create a new IGT bug then :)

New bug created: https://bugs.freedesktop.org/show_bug.cgi?id=107341

However, I would also like to documented that we can also see this bug in GLK:

https://intel-gfx-ci.01.org/tree/drm-tip/drmtip_74/fi-glk-dsi/igt@gem_exec_schedule@deep-render.html

https://intel-gfx-ci.01.org/tree/drm-tip/drmtip_76/fi-glk-dsi/igt@gem_exec_schedule@deep-vebox.html

https://intel-gfx-ci.01.org/tree/drm-tip/drmtip_76/fi-glk-dsi/igt@gem_exec_schedule@deep-bsd.html

(gem_exec_schedule:1158) igt_vgem-CRITICAL: Test assertion failure function vgem_fence_signal, file ../lib/igt_vgem.c:193:
(gem_exec_schedule:1158) igt_vgem-CRITICAL: Failed assertion: __vgem_fence_signal(fd, fence) == 0
(gem_exec_schedule:1158) igt_vgem-CRITICAL: error: -110 != 0
Subtest deep-bsd failed.

commit 2884f91dd6d7682ea738ef6f0943cc591643dda2 (HEAD, upstream/master)
Author: Chris Wilson <chris@chris-wilson.co.uk>
Date:   Fri Jul 20 12:29:30 2018 +0100

    igt/gem_exec_schedule: Trim deep runtime
    
    Time the runtime for emitting deep dependency tree, while keeping it
    full of umpteen thousand requests.
    
    Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=106707
    Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
    Reviewed-by: Katarzyna Dec <katarzyna.dec@intel.com>

Closing this as this bug seen  more than 1 month ago.

This issue used o occur almost every round of CI DRM execution. Not seen this issue since ~200 rounds. Closing this defect.