drm-intel-fixes CI issues

drm-intel-fixes CI issues

[Rodrigo Vivi]

Hi all,

I need your help to decide what to do with this round of fixes.

I have collected these patches this week:

commit b43e8916172a ("drm/i915/dp: Link train Fallback on eDP only if fallback link BW can fit panel's native mode")
commit 5abb01e541ed ("drm/i915: Fix intel_dp_mst_best_encoder()")
commit 02713246296d ("drm/i915: Skip vcpi allocation for MSTB ports that are gone")
commit cc6e027f5f50 ("drm/i915: Don't unset intel_connector->mst_port")
commit f5aec50ba21e ("drm/i915: Use the correct crtc when sanitizing plane mapping")
commit 6547684bf50a ("drm/i915: Restore vblank interrupts earlier")

CI_DIF_309 represents Greg's v4.19-rc7 and it is clean.

However 2 following CI runs are kind of strange.

There's few underruns here and there, but those looks flip-flops.

My biggest concern is specially around:

igt@kms_plane@pixel-format-pipe-a-planes:
https://intel-gfx-ci.01.org/tree/drm-intel-fixes/shards.html

https://intel-gfx-ci.01.org/tree/drm-intel-fixes/CI_DIF_311/shard-glk8/igt@kms_plane@pixel-format-pipe-c-planes.html

Thoughts?

I'm holding the pull request for now and will try to do some local tests here
to see if I can identify a culprit.

Thanks,
Rodrigo.
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx

Re: drm-intel-fixes CI issues

[David Airlie]

On Thu, Oct 11, 2018 at 8:53 AM Rodrigo Vivi <rodrigo.vivi@intel.com> wrote:
>
> Hi all,
>
> I need your help to decide what to do with this round of fixes.
>
> I have collected these patches this week:
>
> commit b43e8916172a ("drm/i915/dp: Link train Fallback on eDP only if fallback link BW can fit panel's native mode")
> commit 5abb01e541ed ("drm/i915: Fix intel_dp_mst_best_encoder()")
> commit 02713246296d ("drm/i915: Skip vcpi allocation for MSTB ports that are gone")
> commit cc6e027f5f50 ("drm/i915: Don't unset intel_connector->mst_port")
> commit f5aec50ba21e ("drm/i915: Use the correct crtc when sanitizing plane mapping")
> commit 6547684bf50a ("drm/i915: Restore vblank interrupts earlier")
>
> CI_DIF_309 represents Greg's v4.19-rc7 and it is clean.
>
> However 2 following CI runs are kind of strange.
>
> There's few underruns here and there, but those looks flip-flops.
>
> My biggest concern is specially around:
>
> igt@kms_plane@pixel-format-pipe-a-planes:
> https://intel-gfx-ci.01.org/tree/drm-intel-fixes/shards.html
>
> https://intel-gfx-ci.01.org/tree/drm-intel-fixes/CI_DIF_311/shard-glk8/igt@kms_plane@pixel-format-pipe-c-planes.html
>
> Thoughts?
>
> I'm holding the pull request for now and will try to do some local tests here
> to see if I can identify a culprit.

At this late in the game for rc8, unless these fix a major regression
in the current tree, I'd say drop them until -next.

Dave.
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx

Re: drm-intel-fixes CI issues

[Lyude Paul]

speaking for my own patches here I'm fine with that!

On Thu, 2018-10-11 at 09:17 +1000, David Airlie wrote:
> On Thu, Oct 11, 2018 at 8:53 AM Rodrigo Vivi <rodrigo.vivi@intel.com> wrote:
> > 
> > Hi all,
> > 
> > I need your help to decide what to do with this round of fixes.
> > 
> > I have collected these patches this week:
> > 
> > commit b43e8916172a ("drm/i915/dp: Link train Fallback on eDP only if
> > fallback link BW can fit panel's native mode")
> > commit 5abb01e541ed ("drm/i915: Fix intel_dp_mst_best_encoder()")
> > commit 02713246296d ("drm/i915: Skip vcpi allocation for MSTB ports that
> > are gone")
> > commit cc6e027f5f50 ("drm/i915: Don't unset intel_connector->mst_port")
> > commit f5aec50ba21e ("drm/i915: Use the correct crtc when sanitizing plane
> > mapping")
> > commit 6547684bf50a ("drm/i915: Restore vblank interrupts earlier")
> > 
> > CI_DIF_309 represents Greg's v4.19-rc7 and it is clean.
> > 
> > However 2 following CI runs are kind of strange.
> > 
> > There's few underruns here and there, but those looks flip-flops.
> > 
> > My biggest concern is specially around:
> > 
> > igt@kms_plane@pixel-format-pipe-a-planes:
> > https://intel-gfx-ci.01.org/tree/drm-intel-fixes/shards.html
> > 
> > 
https://intel-gfx-ci.01.org/tree/drm-intel-fixes/CI_DIF_311/shard-glk8/igt@kms_plane@pixel-format-pipe-c-planes.html
> > 
> > Thoughts?
> > 
> > I'm holding the pull request for now and will try to do some local tests
> > here
> > to see if I can identify a culprit.
> 
> At this late in the game for rc8, unless these fix a major regression
> in the current tree, I'd say drop them until -next.
> 
> Dave.
-- 
Cheers,
	Lyude Paul

_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx

Re: drm-intel-fixes CI issues

[Manasi Navare]

On Thu, Oct 11, 2018 at 09:17:52AM +1000, David Airlie wrote:
> On Thu, Oct 11, 2018 at 8:53 AM Rodrigo Vivi <rodrigo.vivi@intel.com> wrote:
> >
> > Hi all,
> >
> > I need your help to decide what to do with this round of fixes.
> >
> > I have collected these patches this week:
> >
> > commit b43e8916172a ("drm/i915/dp: Link train Fallback on eDP only if fallback link BW can fit panel's native mode")
> > commit 5abb01e541ed ("drm/i915: Fix intel_dp_mst_best_encoder()")
> > commit 02713246296d ("drm/i915: Skip vcpi allocation for MSTB ports that are gone")
> > commit cc6e027f5f50 ("drm/i915: Don't unset intel_connector->mst_port")
> > commit f5aec50ba21e ("drm/i915: Use the correct crtc when sanitizing plane mapping")
> > commit 6547684bf50a ("drm/i915: Restore vblank interrupts earlier")
> >
> > CI_DIF_309 represents Greg's v4.19-rc7 and it is clean.
> >
> > However 2 following CI runs are kind of strange.
> >
> > There's few underruns here and there, but those looks flip-flops.
> >
> > My biggest concern is specially around:
> >
> > igt@kms_plane@pixel-format-pipe-a-planes:
> > https://intel-gfx-ci.01.org/tree/drm-intel-fixes/shards.html
> >
> > https://intel-gfx-ci.01.org/tree/drm-intel-fixes/CI_DIF_311/shard-glk8/igt@kms_plane@pixel-format-pipe-c-planes.html
> >
> > Thoughts?
> >
> > I'm holding the pull request for now and will try to do some local tests here
> > to see if I can identify a culprit.
> 
> At this late in the game for rc8, unless these fix a major regression
> in the current tree, I'd say drop them until -next.
>

I am okay with that w.r.t my patch ("drm/i915/dp: Link train Fallback on eDP only if fallback link BW can fit panel's native mode")

Manasi
 
> Dave.
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx

Re: drm-intel-fixes CI issues

[Lyude Paul]

Poke: still wondering what we should do about the patch in these fixes that
came up a little later which got Cc'd to stable, despite it apparently not
being a patch we want in stable (mentioned this over IRC):

https://patchwork.freedesktop.org/patch/255428/ and
https://patchwork.freedesktop.org/series/50770/

On Thu, 2018-10-11 at 09:17 +1000, David Airlie wrote:
> On Thu, Oct 11, 2018 at 8:53 AM Rodrigo Vivi <rodrigo.vivi@intel.com> wrote:
> > 
> > Hi all,
> > 
> > I need your help to decide what to do with this round of fixes.
> > 
> > I have collected these patches this week:
> > 
> > commit b43e8916172a ("drm/i915/dp: Link train Fallback on eDP only if
> > fallback link BW can fit panel's native mode")
> > commit 5abb01e541ed ("drm/i915: Fix intel_dp_mst_best_encoder()")
> > commit 02713246296d ("drm/i915: Skip vcpi allocation for MSTB ports that
> > are gone")
> > commit cc6e027f5f50 ("drm/i915: Don't unset intel_connector->mst_port")
> > commit f5aec50ba21e ("drm/i915: Use the correct crtc when sanitizing plane
> > mapping")
> > commit 6547684bf50a ("drm/i915: Restore vblank interrupts earlier")
> > 
> > CI_DIF_309 represents Greg's v4.19-rc7 and it is clean.
> > 
> > However 2 following CI runs are kind of strange.
> > 
> > There's few underruns here and there, but those looks flip-flops.
> > 
> > My biggest concern is specially around:
> > 
> > igt@kms_plane@pixel-format-pipe-a-planes:
> > https://intel-gfx-ci.01.org/tree/drm-intel-fixes/shards.html
> > 
> > 
https://intel-gfx-ci.01.org/tree/drm-intel-fixes/CI_DIF_311/shard-glk8/igt@kms_plane@pixel-format-pipe-c-planes.html
> > 
> > Thoughts?
> > 
> > I'm holding the pull request for now and will try to do some local tests
> > here
> > to see if I can identify a culprit.
> 
> At this late in the game for rc8, unless these fix a major regression
> in the current tree, I'd say drop them until -next.
> 
> Dave.
-- 
Cheers,
	Lyude Paul

_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx

Re: drm-intel-fixes CI issues

[Rodrigo Vivi]

On Mon, Oct 15, 2018 at 03:59:36PM -0400, Lyude Paul wrote:
> Poke: still wondering what we should do about the patch in these fixes that
> came up a little later which got Cc'd to stable, despite it apparently not
> being a patch we want in stable (mentioned this over IRC):
> 
> https://patchwork.freedesktop.org/patch/255428/ and
> https://patchwork.freedesktop.org/series/50770/

Thanks for bringing this up.

So, These 2 patches are merged on the tree, but we don't want
them propagated to stable tree anymore? Do we have a fix for those?
Or are we reverting them?

We probably want to raise this up to Greg so he doesn't pick
them when running his stable scripts.

But also I was wondering another aspect of these patches I had here
on this dropped list.

In the end most of patches that was here will be picked by Greg's
stable scripts anyways.

Also other patches with cc:stable that were on that round
but got removed:

commit 1e712535c51a ("drm/i915/dp: Link train Fallback on eDP only if fallback link BW can fit panel's native mode")
commit 62358aa4ee86 ("drm/i915: Use the correct crtc when sanitizing plane mapping")
commit 68bc30deac62 ("drm/i915: Restore vblank interrupts earlier")

Should I add back at least these patches this week?
Or we should really wait for 4.19 to be released?

Thanks,
Rodrigo.

> 
> On Thu, 2018-10-11 at 09:17 +1000, David Airlie wrote:
> > On Thu, Oct 11, 2018 at 8:53 AM Rodrigo Vivi <rodrigo.vivi@intel.com> wrote:
> > > 
> > > Hi all,
> > > 
> > > I need your help to decide what to do with this round of fixes.
> > > 
> > > I have collected these patches this week:
> > > 
> > > commit b43e8916172a ("drm/i915/dp: Link train Fallback on eDP only if
> > > fallback link BW can fit panel's native mode")
> > > commit 5abb01e541ed ("drm/i915: Fix intel_dp_mst_best_encoder()")
> > > commit 02713246296d ("drm/i915: Skip vcpi allocation for MSTB ports that
> > > are gone")
> > > commit cc6e027f5f50 ("drm/i915: Don't unset intel_connector->mst_port")
> > > commit f5aec50ba21e ("drm/i915: Use the correct crtc when sanitizing plane
> > > mapping")
> > > commit 6547684bf50a ("drm/i915: Restore vblank interrupts earlier")
> > > 
> > > CI_DIF_309 represents Greg's v4.19-rc7 and it is clean.
> > > 
> > > However 2 following CI runs are kind of strange.
> > > 
> > > There's few underruns here and there, but those looks flip-flops.
> > > 
> > > My biggest concern is specially around:
> > > 
> > > igt@kms_plane@pixel-format-pipe-a-planes:
> > > https://intel-gfx-ci.01.org/tree/drm-intel-fixes/shards.html
> > > 
> > > 
> https://intel-gfx-ci.01.org/tree/drm-intel-fixes/CI_DIF_311/shard-glk8/igt@kms_plane@pixel-format-pipe-c-planes.html
> > > 
> > > Thoughts?
> > > 
> > > I'm holding the pull request for now and will try to do some local tests
> > > here
> > > to see if I can identify a culprit.
> > 
> > At this late in the game for rc8, unless these fix a major regression
> > in the current tree, I'd say drop them until -next.
> > 
> > Dave.
> -- 
> Cheers,
> 	Lyude Paul
> 
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx

Re: drm-intel-fixes CI issues

[Lyude Paul]

On Mon, 2018-10-15 at 14:20 -0700, Rodrigo Vivi wrote:
> On Mon, Oct 15, 2018 at 03:59:36PM -0400, Lyude Paul wrote:
> > Poke: still wondering what we should do about the patch in these fixes
> > that
> > came up a little later which got Cc'd to stable, despite it apparently not
> > being a patch we want in stable (mentioned this over IRC):
> > 
> > https://patchwork.freedesktop.org/patch/255428/ and
> > https://patchwork.freedesktop.org/series/50770/
> 
> Thanks for bringing this up.
> 
> So, These 2 patches are merged on the tree, but we don't want
> them propagated to stable tree anymore? Do we have a fix for those?
> Or are we reverting them?
I am in the process of making a fix and am just about finished; just had to
clear some corner cases up with danvet. I think we should be fine if we do
either one of two things:

 * Just cc the third patch (once I've got it on the ML today) to stable, since
   I did want the problem these were intended to fix to also be fixed
   downstream, and while messy it should take care of fixing the problems that
   were introduced.
 * Don't cc the two patches to stable, and I can just handle backporting the
   third and final fix to stable

This is a new situation for me, so I'm honestly not sure which of those two
would be the best approach.
[more below]
> 
> We probably want to raise this up to Greg so he doesn't pick
> them when running his stable scripts.
> 
> But also I was wondering another aspect of these patches I had here
> on this dropped list.
> 
> In the end most of patches that was here will be picked by Greg's
> stable scripts anyways.
> 
> Also other patches with cc:stable that were on that round
> but got removed:
> 
> commit 1e712535c51a ("drm/i915/dp: Link train Fallback on eDP only if
> fallback link BW can fit panel's native mode")
According to Manasi earlier in the thread:

"I am okay with that w.r.t my patch ("drm/i915/dp: Link train Fallback on eDP
only if fallback link BW can fit panel's native mode")"

So I think waiting until v4.19 on that one should be fine.

> commit 62358aa4ee86 ("drm/i915: Use the correct crtc when sanitizing plane
> mapping")
> commit 68bc30deac62 ("drm/i915: Restore vblank interrupts earlier")
> 
Can't speak for these two though

> Should I add back at least these patches this week?
> Or we should really wait for 4.19 to be released?
> 
> Thanks,
> Rodrigo.
> 
> > 
> > On Thu, 2018-10-11 at 09:17 +1000, David Airlie wrote:
> > > On Thu, Oct 11, 2018 at 8:53 AM Rodrigo Vivi <rodrigo.vivi@intel.com>
> > > wrote:
> > > > 
> > > > Hi all,
> > > > 
> > > > I need your help to decide what to do with this round of fixes.
> > > > 
> > > > I have collected these patches this week:
> > > > 
> > > > commit b43e8916172a ("drm/i915/dp: Link train Fallback on eDP only if
> > > > fallback link BW can fit panel's native mode")
> > > > commit 5abb01e541ed ("drm/i915: Fix intel_dp_mst_best_encoder()")
> > > > commit 02713246296d ("drm/i915: Skip vcpi allocation for MSTB ports
> > > > that
> > > > are gone")
> > > > commit cc6e027f5f50 ("drm/i915: Don't unset intel_connector-
> > > > >mst_port")
> > > > commit f5aec50ba21e ("drm/i915: Use the correct crtc when sanitizing
> > > > plane
> > > > mapping")
> > > > commit 6547684bf50a ("drm/i915: Restore vblank interrupts earlier")
> > > > 
> > > > CI_DIF_309 represents Greg's v4.19-rc7 and it is clean.
> > > > 
> > > > However 2 following CI runs are kind of strange.
> > > > 
> > > > There's few underruns here and there, but those looks flip-flops.
> > > > 
> > > > My biggest concern is specially around:
> > > > 
> > > > igt@kms_plane@pixel-format-pipe-a-planes:
> > > > https://intel-gfx-ci.01.org/tree/drm-intel-fixes/shards.html
> > > > 
> > > > 
> > 
> > 
https://intel-gfx-ci.01.org/tree/drm-intel-fixes/CI_DIF_311/shard-glk8/igt@kms_plane@pixel-format-pipe-c-planes.html
> > > > 
> > > > Thoughts?
> > > > 
> > > > I'm holding the pull request for now and will try to do some local
> > > > tests
> > > > here
> > > > to see if I can identify a culprit.
> > > 
> > > At this late in the game for rc8, unless these fix a major regression
> > > in the current tree, I'd say drop them until -next.
> > > 
> > > Dave.
> > 
> > -- 
> > Cheers,
> > 	Lyude Paul
> > 
-- 
Cheers,
	Lyude Paul

_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx