Re: [PATCH 1/2] drm/i915/cdclk: Avoid spurious cdclk sanitization on PTL+

[Pierre Pinon <pierre@pinon1.fr>]

On Fri, Jul 17, 2026 at 06:51:06PM +0300, Ville Syrjälä wrote:
> Apparently PTL+ no longer has the cd2x pipe select field in
> CDCLK_CTL. Take that into account during CDCLK sanitization.
>
> This currently triggers a spurious CDCLK sanitization during
> driver load on PTL+ which will causes a visible glitch on all
> active displays.

Tested on a Panther Lake laptop (Intel Graphics [8086:b090], xe driver,
internal eDP panel), Fedora 43.

For what it is worth, on this machine the symptom is more than a glitch:
the internal panel comes up corrupted on cold boot and stays corrupted
until a suspend/resume cycle. dmesg shows:

  xe 0000:00:02.0: [drm] Selective fetch area calculation failed in pipe A
  xe 0000:00:02.0: [drm] *ERROR* CPU pipe A FIFO underrun
  xe 0000:00:02.0: [drm] *ERROR* CPU pipe A FIFO underrun

Counting "CPU pipe A FIFO underrun" in dmesg across boots of the same
machine:

  6.17.1-300.fc43           unpatched   0   (unaffected)
  7.1.12-100.fc43           unpatched   2
  7.1.12 + this patch [*]   patched     0   (4 boots)
  7.1.13-100.fc43           unpatched   2
  7.3.0-rc1 + this patch    patched     0

  [*] 7.1.x predates bxt_cdclk_cd2x_pipe_mask() and needs a slightly
      different form, see below; the 7.3.0-rc1 build carries this patch
      as posted, unmodified.

With the patch the panel comes up clean from cold boot and the underruns
are gone.

Tested-by: Pierre Pinon <pierre@pinon1.fr>

This covers 1/2 only; I have not tested 2/2.

Note the "Selective fetch area calculation failed in pipe A" drm_info_once
is still printed with the patch applied. As far as I can tell that one is
unrelated -- it is the deliberate full-update fallback in
intel_psr2_sel_fetch_update() -- and harmless here.

I realise this is already committed, so the tag may well arrive too late
to be folded in. I am sending it anyway as a data point for the stable
backport, which is the part that matters to me here: neither 7.1.y nor
7.2.y carries the fix yet.

On 7.1.y the change does not apply as posted, since that tree predates
bxt_cdclk_cd2x_pipe_mask(). The form I have been running there guards the
two original lines instead:

	if (DISPLAY_VER(display) < 30) {
		cdctl &= ~bxt_cdclk_cd2x_pipe(display, INVALID_PIPE);
		cdctl |= bxt_cdclk_cd2x_pipe(display, INVALID_PIPE);
	}

That is the variant covered by the four 7.1.12 boots above. The same form
was posted independently by Chang Jiong as
<20260816062344.796615-1-xspeed1989@gmail.com>, which I only noticed
afterwards.