Re: [PATCH v1] Fix: SYNCOBJ TIMELINE Test failed.

[Christian König]

Hi Jesse,

yes, I know that's a well known bug.

The Intel guys have already narrowed it down to a missing 
dma_fence_enable_signaling() in the syncobj code path.

I strongly suggest to work together with them to find where that needs 
to be added instead.

Regards,
Christian.

Am 30.06.22 um 17:26 schrieb Zhang, Jesse(Jie):
> [AMD Official Use Only - General]
>
>
> Hi  Christian,
> If we remove the following patch, the  "syncobj timeline test" can pass.
> https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=721255b52700b320c4ae2e23d57f7d9ad1db50b9
>
>
> The following log is provided by AMD CQE team. They run the amdgpu_test tool on ubuntu22 (kernel version 5.15.0-39)
> Suite: SYNCOBJ TIMELINE Tests
>    Test: syncobj timeline test ...FAILED
>      1. sources/drm/tests/amdgpu/syncobj_tests.c:299  - CU_ASSERT_EQUAL(payload,18)
>      2. sources/drm/tests/amdgpu/syncobj_tests.c:309  - CU_ASSERT_EQUAL(payload,20)
> You can get more detail information by the attachment.
>
> So we need fix this issue. And if you have any better solution to solve the issue, please let me know.
>
> Thanks
> Jesse
>
> -----Original Message-----
> From: Koenig, Christian <Christian.Koenig@xxxxxxx>
> Sent: Wednesday, 29 June 2022 5:12 pm
> To: Zhang, Jesse(Jie) <Jesse.Zhang@xxxxxxx>; broonie@xxxxxxxxxx; alsa-devel@xxxxxxxxxxxxxxxx
> Cc: Mukunda, Vijendar <Vijendar.Mukunda@xxxxxxx>; Hiregoudar, Basavaraj <Basavaraj.Hiregoudar@xxxxxxx>; Dommati, Sunil-kumar <Sunil-kumar.Dommati@xxxxxxx>; Pandey, Ajit Kumar <AjitKumar.Pandey@xxxxxxx>; Nirmoy Das <nirmoy.das@xxxxxxxxxxxxxxx>; Maarten Lankhorst <maarten.lankhorst@xxxxxxxxxxxxxxx>; Maxime Ripard <mripard@xxxxxxxxxx>; Thomas Zimmermann <tzimmermann@xxxxxxx>; David Airlie <airlied@xxxxxxxx>; Daniel Vetter <daniel@xxxxxxxx>; Sumit Semwal <sumit.semwal@xxxxxxxxxx>; open list:DRM DRIVERS <dri-devel@xxxxxxxxxxxxxxxxxxxxx>; open list <linux-kernel@xxxxxxxxxxxxxxx>; open list:DMA BUFFER SHARING FRAMEWORK <linux-media@xxxxxxxxxxxxxxx>; moderated list:DMA BUFFER SHARING FRAMEWORK <linaro-mm-sig@xxxxxxxxxxxxxxxx>
> Subject: Re: [PATCH v1] Fix: SYNCOBJ TIMELINE Test failed.
>
> Am 29.06.22 um 08:02 schrieb jie1zhan:
> >    The issue cause by the commit :
> >
> > 721255b527(drm/syncobj: flatten dma_fence_chains on transfer).
> >
> > Because it use the point of dma_fence incorrectly
> >
> > Correct the point of dma_fence by fence array
> Well that patch is just utterly nonsense as far as I can see.
>
> > Signed-off-by: jie1zhan <jesse.zhang@xxxxxxx>
> >
> > Reviewed-by: Christian König <christian.koenig@xxxxxxx>
> >
> > Reviewed-by: Nirmoy Das <nirmoy.das@xxxxxxxxxxxxxxx>
> I have strong doubts that Nirmoy has reviewed this and I certainly haven't reviewed it.
>
> Christian.
>
> > ---
> >    drivers/gpu/drm/drm_syncobj.c | 2 +-
> >    1 file changed, 1 insertion(+), 1 deletion(-)
> >
> > diff --git a/drivers/gpu/drm/drm_syncobj.c
> > b/drivers/gpu/drm/drm_syncobj.c index 7e48dcd1bee4..d5db818f1c76
> > 100644
> > --- a/drivers/gpu/drm/drm_syncobj.c
> > +++ b/drivers/gpu/drm/drm_syncobj.c
> > @@ -887,7 +887,7 @@ static int drm_syncobj_flatten_chain(struct dma_fence **f)
> >    		goto free_fences;
> >    
> >    	dma_fence_put(*f);
> > -	*f = &array->base;
> > +	*f = array->fences[0];
> >    	return 0;
> >    
> >    free_fences: