* [PATCH v3 0/4] drm/i915/dp: Attach HDR_OUTPUT_METADATA on eDP
@ 2026-11-21 0:54 Yitian Shao
2026-11-21 1:43 ` Ville Syrjälä
0 siblings, 1 reply; 3+ messages in thread
From: Yitian Shao @ 2026-11-21 0:54 UTC (permalink / raw)
To: intel-gfx
Cc: dri-devel, ville.syrjala, jani.nikula, sebastian.wick,
uma.shankar, suraj.kandpal
This series makes HDR work on pre-ICL eDP panels that decode the metadata
through the Intel AUX HDR interface instead of a gamut metadata DIP.
Patch 1 removes the LSPCON check from intel_dp_has_gamut_metadata_dip(),
which made the helper answer the wrong question.
Patch 2 attaches HDR_OUTPUT_METADATA on eDP when the panel's TCON advertises
PQ decoding, and skips the infoframe when no DIP is available.
Patch 3 is the pre-existing LSPCON issue pointed out during review: refuse
HDR support when the port cannot transmit the gamut metadata DIP.
Patch 4 writes the content luminance over DPCD only when the SDP path is not
in use, since the SDP takes precedence.
v3:
- Drop the LSPCON check from intel_dp_has_gamut_metadata_dip() and move it
to intel_dp_add_properties(), per Ville.
- Fold the checks into intel_dp_hdr_metadata_supported() instead of an
ever growing if-statement.
- Add the LSPCON DIP check as a separate patch, per Ville.
- Invert the content luminance condition: use DPCD when there is no SDP.
- Dropped the fastset concern from v2. The analysis was confirmed correct,
but HDR metadata fastset support is already broken and needs a separate
fix (possibly crtc_state->backlight_changed), so it does not belong here.
v2:
- Do not skip the infoframe on LSPCON (superseded by patch 1 and 2).
- Only attach the property when the Intel AUX HDR backlight interface is in
use, to avoid reading the untagged union.
Testing: CometLake-U (Gen9.5) laptop with a 3840x2160 10bpc eDP panel that
advertises BT2020 + ST2084. With the equivalent change applied the connector
exposes HDR_OUTPUT_METADATA, GNOME 50 offers the HDR switch, and the panel
enters PQ/BT2020 mode. The TCON HDR backlight interface was already in use.
v3 was compile-tested only; I have not re-run the runtime test.
Yitian Shao (4):
drm/i915/dp: Drop LSPCON check from intel_dp_has_gamut_metadata_dip()
drm/i915/dp: Attach HDR_OUTPUT_METADATA on eDP with TCON HDR support
drm/i915/display: Check gamut metadata DIP support in LSPCON HDR
detection
drm/i915/dp: Prefer SDP over DPCD for HDR content luminance
drivers/gpu/drm/i915/display/intel_dp.c | 32 ++++++++++++++++---
.../drm/i915/display/intel_dp_aux_backlight.c | 11 ++++++-
.../drm/i915/display/intel_dp_aux_backlight.h | 3 ++
drivers/gpu/drm/i915/display/intel_lspcon.c | 5 +++
4 files changed, 46 insertions(+), 5 deletions(-)
--
2.53.0
^ permalink raw reply [flat|nested] 3+ messages in thread* Re: [PATCH v3 0/4] drm/i915/dp: Attach HDR_OUTPUT_METADATA on eDP
2026-11-21 0:54 [PATCH v3 0/4] drm/i915/dp: Attach HDR_OUTPUT_METADATA on eDP Yitian Shao
@ 2026-11-21 1:43 ` Ville Syrjälä
0 siblings, 0 replies; 3+ messages in thread
From: Ville Syrjälä @ 2026-11-21 1:43 UTC (permalink / raw)
To: Yitian Shao
Cc: intel-gfx, dri-devel, jani.nikula, sebastian.wick, uma.shankar,
suraj.kandpal
On Sat, Nov 21, 2026 at 08:54:05AM +0800, Yitian Shao wrote:
> This series makes HDR work on pre-ICL eDP panels that decode the metadata
> through the Intel AUX HDR interface instead of a gamut metadata DIP.
>
> Patch 1 removes the LSPCON check from intel_dp_has_gamut_metadata_dip(),
> which made the helper answer the wrong question.
>
> Patch 2 attaches HDR_OUTPUT_METADATA on eDP when the panel's TCON advertises
> PQ decoding, and skips the infoframe when no DIP is available.
>
> Patch 3 is the pre-existing LSPCON issue pointed out during review: refuse
> HDR support when the port cannot transmit the gamut metadata DIP.
>
> Patch 4 writes the content luminance over DPCD only when the SDP path is not
> in use, since the SDP takes precedence.
We still haven't solved the fastset issue, but I think we can mostly
sidestep it for now by always forcing a full modeset when the HDR
metadata changes.
Probably the most straightforward way to do that is by changing
intel_digital_connector_atomic_check() to flag connectors_changed=true
instead of mode_changed=true if the metadata changes on a connector
that uses the AUX HDR interface. A commit with connectors_change==true
will always result in a full modeset.
And I'm thinking that we do want that check to be just the
"are we using the HDR AUX interface" and nothing more. We should
not consider the SDP override stuff at all. That's because the HDR
AUX interface also sets up other stuff such as the 2084 decode and
segmented backlight. I believe this is in fact a pre-existing bug
in the code; If the metadata changes between PQ and SDR but otherwise
remains in place we are currently failing to update the TCON.
Eventually we'll probably want a proper fastset support for this
stuff, but no need to rush it now IMO.
>
> v3:
> - Drop the LSPCON check from intel_dp_has_gamut_metadata_dip() and move it
> to intel_dp_add_properties(), per Ville.
> - Fold the checks into intel_dp_hdr_metadata_supported() instead of an
> ever growing if-statement.
> - Add the LSPCON DIP check as a separate patch, per Ville.
> - Invert the content luminance condition: use DPCD when there is no SDP.
> - Dropped the fastset concern from v2. The analysis was confirmed correct,
> but HDR metadata fastset support is already broken and needs a separate
> fix (possibly crtc_state->backlight_changed), so it does not belong here.
>
> v2:
> - Do not skip the infoframe on LSPCON (superseded by patch 1 and 2).
> - Only attach the property when the Intel AUX HDR backlight interface is in
> use, to avoid reading the untagged union.
>
> Testing: CometLake-U (Gen9.5) laptop with a 3840x2160 10bpc eDP panel that
> advertises BT2020 + ST2084. With the equivalent change applied the connector
> exposes HDR_OUTPUT_METADATA, GNOME 50 offers the HDR switch, and the panel
> enters PQ/BT2020 mode. The TCON HDR backlight interface was already in use.
> v3 was compile-tested only; I have not re-run the runtime test.
>
> Yitian Shao (4):
> drm/i915/dp: Drop LSPCON check from intel_dp_has_gamut_metadata_dip()
> drm/i915/dp: Attach HDR_OUTPUT_METADATA on eDP with TCON HDR support
> drm/i915/display: Check gamut metadata DIP support in LSPCON HDR
> detection
> drm/i915/dp: Prefer SDP over DPCD for HDR content luminance
>
> drivers/gpu/drm/i915/display/intel_dp.c | 32 ++++++++++++++++---
> .../drm/i915/display/intel_dp_aux_backlight.c | 11 ++++++-
> .../drm/i915/display/intel_dp_aux_backlight.h | 3 ++
> drivers/gpu/drm/i915/display/intel_lspcon.c | 5 +++
> 4 files changed, 46 insertions(+), 5 deletions(-)
>
> --
> 2.53.0
--
Ville Syrjälä
Intel
^ permalink raw reply [flat|nested] 3+ messages in thread
[parent not found: <20261121005409.17626-1-fighters390@qq.com>]
* Re: [PATCH v3 0/4] drm/i915/dp: Attach HDR_OUTPUT_METADATA on eDP
[not found] <20261121005409.17626-1-fighters390@qq.com>
@ 2026-11-21 1:17 ` Yitian Shao
0 siblings, 0 replies; 3+ messages in thread
From: Yitian Shao @ 2026-11-21 1:17 UTC (permalink / raw)
To: intel-gfx
Cc: dri-devel, ville.syrjala, jani.nikula, sebastian.wick,
uma.shankar, suraj.kandpal
Runtime-tested v3 on the same CometLake-U machine (kernel 7.0.0-44):
- HDR_OUTPUT_METADATA is attached to eDP-1
- the DP Colorspace property switches to BT2020_RGB with eotf PQ
- a 4K HDR10 HEVC file plays with VA-API and the output stays in BT2020/PQ
for the whole playback
- no state checker warnings in dmesg
As before, Mutter sends all-zero mastering metadata, so the effect of
max_cll/max_fall reaching the TCON cannot be observed visually.
Thanks,
Yitian
^ permalink raw reply [flat|nested] 3+ messages in thread
end of thread, other threads:[~2026-11-21 1:43 UTC | newest]
Thread overview: 3+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2026-11-21 0:54 [PATCH v3 0/4] drm/i915/dp: Attach HDR_OUTPUT_METADATA on eDP Yitian Shao
2026-11-21 1:43 ` Ville Syrjälä
[not found] <20261121005409.17626-1-fighters390@qq.com>
2026-11-21 1:17 ` Yitian Shao
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox