intel-xe.lists.freedesktop.org archive mirror
 help / color / mirror / Atom feed
From: "Ville Syrjälä" <ville.syrjala@linux.intel.com>
To: Tvrtko Ursulin <tvrtko.ursulin@igalia.com>
Cc: intel-xe@lists.freedesktop.org
Subject: Re: ✗ Xe.CI.Full: failure for series starting with [1/2] drm/xe: Fix ggtt fb alignment
Date: Wed, 24 Sep 2025 14:56:25 +0300	[thread overview]
Message-ID: <aNPcaaAn0-MXCv0U@intel.com> (raw)
In-Reply-To: <90ac93f9-90c4-4283-9dd7-f274cda34615@igalia.com>

On Wed, Sep 24, 2025 at 12:26:31PM +0100, Tvrtko Ursulin wrote:
> 
> 
> On 24/09/2025 12:11, Patchwork wrote:
> 
> [snip]
> 
> >         Possible regressions
> > 
> >   *
> > 
> >     igt@kms_async_flips@async-flip-with-page-flip-events-tiled-
> >     atomic@pipe-d-dp-2-4:
> > 
> >       o shard-bmg: NOTRUN -> DMESG-WARN <https://intel-gfx-ci.01.org/
> >         tree/intel-xe/xe-pw-154941v1/shard-bmg-1/
> >         igt@kms_async_flips@async-flip-with-page-flip-events-tiled-
> >         atomic@pipe-d-dp-2-4.html> +8 other tests dmesg-warn
> >   *
> > 
> >     igt@kms_async_flips@async-flip-with-page-flip-events-tiled@pipe-c-
> >     hdmi-a-3-4:
> > 
> >       o shard-bmg: PASS <https://intel-gfx-ci.01.org/tree/intel-xe/
> >         xe-3822-af3cdefd0a1ad2dad29fcde0854ccbce494cc28a/shard-bmg-5/
> >         igt@kms_async_flips@async-flip-with-page-flip-events-tiled@pipe-
> >         c-hdmi-a-3-4.html> -> DMESG-WARN <https://intel-gfx-ci.01.org/
> >         tree/intel-xe/xe-pw-154941v1/shard-bmg-8/
> >         igt@kms_async_flips@async-flip-with-page-flip-events-tiled@pipe-
> >         c-hdmi-a-3-4.html> +18 other tests dmesg-warn
> 
> Mysterious page faults. Looks like the BMG failures due DPT alignment / 
> 512 were real after all. Does it provide any hint to you Ville?

Sounds like the same thing that's happening on MTL.

For ADL I previously forced the DPT alignment in GGTT to 16k
in tgl_plane_min_alignment() to cure async flip faults.

I later tried the same trick on MTL, but I think I had to
bump it to 32k to completely eliminate the faults there.

And the MTL faults didn't seem to be entirely about alignment
because IIRC the other way I could eliminate them was by adding
a few dummy pages around the DPT in GGTT, even while forcing the
start of DPT to be only 4k aligned. And that's when I got pulled
into some other work and didn't have time to finish the
investigation :/

-- 
Ville Syrjälä
Intel

      reply	other threads:[~2025-09-24 11:56 UTC|newest]

Thread overview: 9+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2025-09-24  8:13 [PATCH 1/2] drm/xe: Fix ggtt fb alignment Tvrtko Ursulin
2025-09-24  8:13 ` [PATCH 2/2] drm/xe: Fix DPT frame buffer alignment Tvrtko Ursulin
2025-09-24  8:19 ` ✗ CI.checkpatch: warning for series starting with [1/2] drm/xe: Fix ggtt fb alignment Patchwork
2025-09-24  8:21 ` ✓ CI.KUnit: success " Patchwork
2025-09-24  8:40 ` [PATCH 1/2] " Ville Syrjälä
2025-09-24  9:05 ` ✓ Xe.CI.BAT: success for series starting with [1/2] " Patchwork
2025-09-24 11:11 ` ✗ Xe.CI.Full: failure " Patchwork
2025-09-24 11:26   ` Tvrtko Ursulin
2025-09-24 11:56     ` Ville Syrjälä [this message]

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=aNPcaaAn0-MXCv0U@intel.com \
    --to=ville.syrjala@linux.intel.com \
    --cc=intel-xe@lists.freedesktop.org \
    --cc=tvrtko.ursulin@igalia.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).