<feed xmlns='http://www.w3.org/2005/Atom'>
<title>linux-toradex.git/drivers/gpu/drm/amd/display/amdgpu_dm, branch master</title>
<subtitle>Linux kernel for Apalis and Colibri modules</subtitle>
<link rel='alternate' type='text/html' href='https://git.toradex.cn/cgit/linux-toradex.git/'/>
<entry>
<title>drm/amd/display: Use FreeSync v3 infopacket for sinks above 255 Hz</title>
<updated>2026-11-18T06:52:09+00:00</updated>
<author>
<name>Cooper Morgan</name>
<email>ext-github@cwooper.dev</email>
</author>
<published>2026-10-31T04:18:26+00:00</published>
<link rel='alternate' type='text/html' href='https://git.toradex.cn/cgit/linux-toradex.git/commit/?id=6a9d5e2f06cd6f5c6aa0acb223a1c0a9dbadfce6'/>
<id>6a9d5e2f06cd6f5c6aa0acb223a1c0a9dbadfce6</id>
<content type='text'>
Native DP, eDP, and HDMI sinks get the v1 FreeSync infopacket unless
they are a whitelisted PCON or support Replay. v1 holds each refresh
rate in one byte, so a sink whose range goes above 255 Hz cannot be
told its real maximum. The v3 packet carries the high bits in
PB11/PB12.

The AMD VSDB version is already stored for every DP, eDP, and HDMI
sink. Send v3 when the sink advertises AMD VSDB v3 and its maximum
refresh rate is above 255 Hz. The choice follows the sink's maximum
rather than the current mode, so such a sink gets v3 at every mode
and the packet version stays fixed per sink. Sinks whose maximum is
255 Hz or less are unchanged.

Tested with this series backported to v7.2.13 on an MSI MAG 271QPX32
(AMD VSDB v3, 48-320 Hz) over DP 1.4 on an RX 9070 XT: 320 Hz shows a
correct picture, and with VRR active the refresh rate tracks frame
caps of 300, 270, and 100 Hz. At 240 Hz, where the sink now gets v3
instead of v1, the picture is correct and VRR tracks a 100 Hz cap.

Link: https://gitlab.freedesktop.org/drm/amd/-/issues/5988
Reviewed-by: Harry Wentland &lt;harry.wentland@amd.com&gt;
Signed-off-by: Cooper Morgan &lt;ext-github@cwooper.dev&gt;
Signed-off-by: Wayne Lin &lt;Wayne.Lin@amd.com&gt;
Signed-off-by: Alex Deucher &lt;alexander.deucher@amd.com&gt;
(cherry picked from commit e1c2682397f2f8f40eada17c817b5d3b300124b6)
Cc: stable@vger.kernel.org
</content>
<content type='xhtml'>
<div xmlns='http://www.w3.org/1999/xhtml'>
<pre>
Native DP, eDP, and HDMI sinks get the v1 FreeSync infopacket unless
they are a whitelisted PCON or support Replay. v1 holds each refresh
rate in one byte, so a sink whose range goes above 255 Hz cannot be
told its real maximum. The v3 packet carries the high bits in
PB11/PB12.

The AMD VSDB version is already stored for every DP, eDP, and HDMI
sink. Send v3 when the sink advertises AMD VSDB v3 and its maximum
refresh rate is above 255 Hz. The choice follows the sink's maximum
rather than the current mode, so such a sink gets v3 at every mode
and the packet version stays fixed per sink. Sinks whose maximum is
255 Hz or less are unchanged.

Tested with this series backported to v7.2.13 on an MSI MAG 271QPX32
(AMD VSDB v3, 48-320 Hz) over DP 1.4 on an RX 9070 XT: 320 Hz shows a
correct picture, and with VRR active the refresh rate tracks frame
caps of 300, 270, and 100 Hz. At 240 Hz, where the sink now gets v3
instead of v1, the picture is correct and VRR tracks a 100 Hz cap.

Link: https://gitlab.freedesktop.org/drm/amd/-/issues/5988
Reviewed-by: Harry Wentland &lt;harry.wentland@amd.com&gt;
Signed-off-by: Cooper Morgan &lt;ext-github@cwooper.dev&gt;
Signed-off-by: Wayne Lin &lt;Wayne.Lin@amd.com&gt;
Signed-off-by: Alex Deucher &lt;alexander.deucher@amd.com&gt;
(cherry picked from commit e1c2682397f2f8f40eada17c817b5d3b300124b6)
Cc: stable@vger.kernel.org
</pre>
</div>
</content>
</entry>
<entry>
<title>drm/amd/display: Saturate refresh rates in v1/v2 FreeSync infopackets</title>
<updated>2026-11-18T06:51:39+00:00</updated>
<author>
<name>Cooper Morgan</name>
<email>ext-github@cwooper.dev</email>
</author>
<published>2026-10-31T04:18:25+00:00</published>
<link rel='alternate' type='text/html' href='https://git.toradex.cn/cgit/linux-toradex.git/commit/?id=76850a457474a629dc0ceffaafb3aed78ff5f469'/>
<id>76850a457474a629dc0ceffaafb3aed78ff5f469</id>
<content type='text'>
The v1 and v2 FreeSync infopackets carry the minimum and maximum
refresh rate in one byte each, and build_vrr_infopacket_data_v1() casts
the rate in Hz to unsigned char. On a sink whose range goes above
255 Hz the value wraps: a 320 Hz mode is advertised as a 64 Hz maximum.

Sinks that treat the field as authoritative misbehave once the wrapped
value lands inside their range. An MSI MAG 271QPX32 (48-320 Hz) over
native DP looks like a ~720p image upscaled to the full screen at 320 Hz
and runs its self-test pattern at 305-315 Hz, while 300 Hz (wrapping to
44 Hz, below its minimum) is fine. Capping the EDID range at 255 Hz, so
the packet says 255, gives a correct picture at 320 Hz.

Saturate the fields at 255 Hz instead of wrapping.

Capping v1 at 255 as a fallback was suggested by oystersuki in the linked
issue.

Fixes: 98e6436d3af5 ("drm/amd/display: Refactor FreeSync module")
Closes: https://gitlab.freedesktop.org/drm/amd/-/issues/5988
Reviewed-by: Harry Wentland &lt;harry.wentland@amd.com&gt;
Signed-off-by: Cooper Morgan &lt;ext-github@cwooper.dev&gt;
Signed-off-by: Wayne Lin &lt;Wayne.Lin@amd.com&gt;
Signed-off-by: Alex Deucher &lt;alexander.deucher@amd.com&gt;
(cherry picked from commit 8f91c13d9a14085216725f60b55546fce5c2dd19)
Cc: stable@vger.kernel.org
</content>
<content type='xhtml'>
<div xmlns='http://www.w3.org/1999/xhtml'>
<pre>
The v1 and v2 FreeSync infopackets carry the minimum and maximum
refresh rate in one byte each, and build_vrr_infopacket_data_v1() casts
the rate in Hz to unsigned char. On a sink whose range goes above
255 Hz the value wraps: a 320 Hz mode is advertised as a 64 Hz maximum.

Sinks that treat the field as authoritative misbehave once the wrapped
value lands inside their range. An MSI MAG 271QPX32 (48-320 Hz) over
native DP looks like a ~720p image upscaled to the full screen at 320 Hz
and runs its self-test pattern at 305-315 Hz, while 300 Hz (wrapping to
44 Hz, below its minimum) is fine. Capping the EDID range at 255 Hz, so
the packet says 255, gives a correct picture at 320 Hz.

Saturate the fields at 255 Hz instead of wrapping.

Capping v1 at 255 as a fallback was suggested by oystersuki in the linked
issue.

Fixes: 98e6436d3af5 ("drm/amd/display: Refactor FreeSync module")
Closes: https://gitlab.freedesktop.org/drm/amd/-/issues/5988
Reviewed-by: Harry Wentland &lt;harry.wentland@amd.com&gt;
Signed-off-by: Cooper Morgan &lt;ext-github@cwooper.dev&gt;
Signed-off-by: Wayne Lin &lt;Wayne.Lin@amd.com&gt;
Signed-off-by: Alex Deucher &lt;alexander.deucher@amd.com&gt;
(cherry picked from commit 8f91c13d9a14085216725f60b55546fce5c2dd19)
Cc: stable@vger.kernel.org
</pre>
</div>
</content>
</entry>
<entry>
<title>drm/amd/display: Hold the VRR vblank reference only on active CRTCs</title>
<updated>2026-11-18T06:51:12+00:00</updated>
<author>
<name>Hugo Sales</name>
<email>hugo@hsal.es</email>
</author>
<published>2026-10-31T14:33:51+00:00</published>
<link rel='alternate' type='text/html' href='https://git.toradex.cn/cgit/linux-toradex.git/commit/?id=ad40065dc8bd5a5e16b13523d9fef6ac94cf2b66'/>
<id>ad40065dc8bd5a5e16b13523d9fef6ac94cf2b66</id>
<content type='text'>
amdgpu_dm_handle_vrr_transition() takes a vblank reference when a CRTC
state turns VRR on and drops it when a later state turns it off. It only
looks at freesync_config.state, and disabling a CRTC keeps that state,
so a suspend or DPMS off with VRR active keeps the reference

On resume, drm_atomic_helper_resume() first calls drm_mode_config_reset(),
which swaps the CRTC state for a zeroed one. The restored state then goes
from VRR_STATE_UNSUPPORTED to active and takes a second reference, and
the first is never dropped. The vblank irq stays on for good, so the CRTC
never idles and PSR is never entered

A VRR change on a disabled CRTC can't take the reference at all, since
vblank is off there and drm_crtc_vblank_get() fails with a WARN

On a Cezanne laptop with VRR always on, each suspend added one to the
eDP CRTC's vblank refcount, and it stayed at 1 rather than 0 after VRR
was turned off again

Treat VRR as active for the reference only while the CRTC is active too.
Disabling the CRTC then drops it and enabling it takes it again, which
balances suspend, resume and DPMS, and the get only happens once
manage_dm_interrupts() has turned vblank on

Fixes: 66b0c973d7f7 ("drm/amd/display: Prevent vblank irq disable while VRR is active. (v3)")
Reviewed-by: Ivan Lipski &lt;ivan.lipski@amd.com&gt;
Signed-off-by: Hugo Sales &lt;hugo@hsal.es&gt;
Signed-off-by: Wayne Lin &lt;Wayne.Lin@amd.com&gt;
Signed-off-by: Alex Deucher &lt;alexander.deucher@amd.com&gt;
(cherry picked from commit 4232ec2333ab3ef16520b399b30d50f17ceb9c63)
Cc: stable@vger.kernel.org
</content>
<content type='xhtml'>
<div xmlns='http://www.w3.org/1999/xhtml'>
<pre>
amdgpu_dm_handle_vrr_transition() takes a vblank reference when a CRTC
state turns VRR on and drops it when a later state turns it off. It only
looks at freesync_config.state, and disabling a CRTC keeps that state,
so a suspend or DPMS off with VRR active keeps the reference

On resume, drm_atomic_helper_resume() first calls drm_mode_config_reset(),
which swaps the CRTC state for a zeroed one. The restored state then goes
from VRR_STATE_UNSUPPORTED to active and takes a second reference, and
the first is never dropped. The vblank irq stays on for good, so the CRTC
never idles and PSR is never entered

A VRR change on a disabled CRTC can't take the reference at all, since
vblank is off there and drm_crtc_vblank_get() fails with a WARN

On a Cezanne laptop with VRR always on, each suspend added one to the
eDP CRTC's vblank refcount, and it stayed at 1 rather than 0 after VRR
was turned off again

Treat VRR as active for the reference only while the CRTC is active too.
Disabling the CRTC then drops it and enabling it takes it again, which
balances suspend, resume and DPMS, and the get only happens once
manage_dm_interrupts() has turned vblank on

Fixes: 66b0c973d7f7 ("drm/amd/display: Prevent vblank irq disable while VRR is active. (v3)")
Reviewed-by: Ivan Lipski &lt;ivan.lipski@amd.com&gt;
Signed-off-by: Hugo Sales &lt;hugo@hsal.es&gt;
Signed-off-by: Wayne Lin &lt;Wayne.Lin@amd.com&gt;
Signed-off-by: Alex Deucher &lt;alexander.deucher@amd.com&gt;
(cherry picked from commit 4232ec2333ab3ef16520b399b30d50f17ceb9c63)
Cc: stable@vger.kernel.org
</pre>
</div>
</content>
</entry>
<entry>
<title>drm/amd/display: Add the AUX backlight minimum once on the custom curve</title>
<updated>2026-11-18T06:50:35+00:00</updated>
<author>
<name>Hugo Sales</name>
<email>hugo@hsal.es</email>
</author>
<published>2026-11-03T13:15:57+00:00</published>
<link rel='alternate' type='text/html' href='https://git.toradex.cn/cgit/linux-toradex.git/commit/?id=988ab04d773ac089588ce17742f491e9ee8b1793'/>
<id>988ab04d773ac089588ce17742f491e9ee8b1793</id>
<content type='text'>
On an AUX backlight, convert_brightness_from_user() maps the user level
through the ATIF custom brightness curve with
convert_custom_brightness(), which rescales the result to
user_min..user_max. convert_brightness_from_user() then rescales it
again, from 0..user_max to user_min..user_max, so the minimum goes in
twice. On the 1..512 nits AUX fallback range, a request for 0 gives
1998 millinits instead of 1000

Rescale the curve's output to 0..user_max and leave the offset to the
caller

Fixes: 8dbd72cb7900 ("drm/amd/display: Export full brightness range to userspace")
Reviewed-by: Ray Wu &lt;ray.wu@amd.com&gt;
Signed-off-by: Hugo Sales &lt;hugo@hsal.es&gt;
Signed-off-by: Wayne Lin &lt;Wayne.Lin@amd.com&gt;
Signed-off-by: Alex Deucher &lt;alexander.deucher@amd.com&gt;
(cherry picked from commit 5f95ef2cfd208ca7efe446abfc9e3144dc5ce4fa)
Cc: stable@vger.kernel.org
</content>
<content type='xhtml'>
<div xmlns='http://www.w3.org/1999/xhtml'>
<pre>
On an AUX backlight, convert_brightness_from_user() maps the user level
through the ATIF custom brightness curve with
convert_custom_brightness(), which rescales the result to
user_min..user_max. convert_brightness_from_user() then rescales it
again, from 0..user_max to user_min..user_max, so the minimum goes in
twice. On the 1..512 nits AUX fallback range, a request for 0 gives
1998 millinits instead of 1000

Rescale the curve's output to 0..user_max and leave the offset to the
caller

Fixes: 8dbd72cb7900 ("drm/amd/display: Export full brightness range to userspace")
Reviewed-by: Ray Wu &lt;ray.wu@amd.com&gt;
Signed-off-by: Hugo Sales &lt;hugo@hsal.es&gt;
Signed-off-by: Wayne Lin &lt;Wayne.Lin@amd.com&gt;
Signed-off-by: Alex Deucher &lt;alexander.deucher@amd.com&gt;
(cherry picked from commit 5f95ef2cfd208ca7efe446abfc9e3144dc5ce4fa)
Cc: stable@vger.kernel.org
</pre>
</div>
</content>
</entry>
<entry>
<title>drm/amd/display: Keep FreeSync flip timing across plane resets</title>
<updated>2026-11-18T06:50:09+00:00</updated>
<author>
<name>Davey Hughes</name>
<email>davidralphhughes@gmail.com</email>
</author>
<published>2026-10-28T22:54:12+00:00</published>
<link rel='alternate' type='text/html' href='https://git.toradex.cn/cgit/linux-toradex.git/commit/?id=fff46a322847a54cf11d5af3d7fc7fa07bc71b29'/>
<id>fff46a322847a54cf11d5af3d7fc7fa07bc71b29</id>
<content type='text'>
dm_update_plane_state() replaces the DC plane with a newly allocated one
whenever the plane is reset. The new plane's flip timing is zero, and
mod_freesync_handle_preflip() runs before DC stores the new plane's
first flip timestamp, so FreeSync takes the 32-bit microsecond timestamp
itself as the last render time: the monotonic clock modulo about 71.6
minutes.

On a panel with below-the-range compensation (BTR), that engages BTR and
pins the frame time at the BTR mid-point (13.447 ms on a 48-165 Hz
panel). A panel without BTR (max refresh below twice the min) goes
through the fixed refresh path instead and locks to its max refresh
after six frames.

It takes a reset on every flip to keep the panel there, and color
management changes reset the plane. DRM sets color_mgmt_changed whenever
a color property is given a different blob, even one with the same
contents, and KWin 6.8.1 with HDR enabled and its color accuracy setting
at "Prefer efficiency" sets a new GAMMA_LUT blob on every commit
(current KWin master no longer does). Plane resets for CRTC color
changes, added by
commit 7316c4ad2996 ("drm/amd/display: Reset planes for color management
changes"), and later for plane color properties and colorops, made this
reachable on every flip.

Copy the flip timing from the old DC plane into the new one when the
plane stays on the same CRTC. Do it in amdgpu_dm_commit_planes() before
the preflip, after drm_atomic_helper_wait_for_dependencies() has waited
for the previous commit, so the copy holds that commit's flip and does
not race with DC writing it. This covers both the BTR and the fixed
refresh path. A plane that is newly enabled or moved to another CRTC
still starts without history.

Fixes: 180db303ff46 ("drm/amd/display: Add below the range support for FreeSync")
Reviewed-by: Harry Wentland &lt;harry.wentland@amd.com&gt;
Signed-off-by: Davey Hughes &lt;davidralphhughes@gmail.com&gt;
Signed-off-by: Wayne Lin &lt;Wayne.Lin@amd.com&gt;
Signed-off-by: Alex Deucher &lt;alexander.deucher@amd.com&gt;
(cherry picked from commit 4acf82d2a2dcca2d7df8649ecafb3da6c1245003)
Cc: stable@vger.kernel.org
</content>
<content type='xhtml'>
<div xmlns='http://www.w3.org/1999/xhtml'>
<pre>
dm_update_plane_state() replaces the DC plane with a newly allocated one
whenever the plane is reset. The new plane's flip timing is zero, and
mod_freesync_handle_preflip() runs before DC stores the new plane's
first flip timestamp, so FreeSync takes the 32-bit microsecond timestamp
itself as the last render time: the monotonic clock modulo about 71.6
minutes.

On a panel with below-the-range compensation (BTR), that engages BTR and
pins the frame time at the BTR mid-point (13.447 ms on a 48-165 Hz
panel). A panel without BTR (max refresh below twice the min) goes
through the fixed refresh path instead and locks to its max refresh
after six frames.

It takes a reset on every flip to keep the panel there, and color
management changes reset the plane. DRM sets color_mgmt_changed whenever
a color property is given a different blob, even one with the same
contents, and KWin 6.8.1 with HDR enabled and its color accuracy setting
at "Prefer efficiency" sets a new GAMMA_LUT blob on every commit
(current KWin master no longer does). Plane resets for CRTC color
changes, added by
commit 7316c4ad2996 ("drm/amd/display: Reset planes for color management
changes"), and later for plane color properties and colorops, made this
reachable on every flip.

Copy the flip timing from the old DC plane into the new one when the
plane stays on the same CRTC. Do it in amdgpu_dm_commit_planes() before
the preflip, after drm_atomic_helper_wait_for_dependencies() has waited
for the previous commit, so the copy holds that commit's flip and does
not race with DC writing it. This covers both the BTR and the fixed
refresh path. A plane that is newly enabled or moved to another CRTC
still starts without history.

Fixes: 180db303ff46 ("drm/amd/display: Add below the range support for FreeSync")
Reviewed-by: Harry Wentland &lt;harry.wentland@amd.com&gt;
Signed-off-by: Davey Hughes &lt;davidralphhughes@gmail.com&gt;
Signed-off-by: Wayne Lin &lt;Wayne.Lin@amd.com&gt;
Signed-off-by: Alex Deucher &lt;alexander.deucher@amd.com&gt;
(cherry picked from commit 4acf82d2a2dcca2d7df8649ecafb3da6c1245003)
Cc: stable@vger.kernel.org
</pre>
</div>
</content>
</entry>
<entry>
<title>drm/amd/display: Don't warn about vendor infoframe for non-infoframe sinks</title>
<updated>2026-11-18T06:49:43+00:00</updated>
<author>
<name>Paul Menzel</name>
<email>pmenzel@molgen.mpg.de</email>
</author>
<published>2026-10-27T20:52:22+00:00</published>
<link rel='alternate' type='text/html' href='https://git.toradex.cn/cgit/linux-toradex.git/commit/?id=2b784f5e78694d4d852053b3eef3de3f0ff1d02e'/>
<id>2b784f5e78694d4d852053b3eef3de3f0ff1d02e</id>
<content type='text'>
With an EIZO EV2781 connected over HDMI to an AMD Ryzen 3 2200G (Raven)
on an MSI B350M MORTAR (MS-7A37), Linux logs at every boot since v6.16:

    amdgpu 0000:30:00.0: [drm] Failed to setup vendor infoframe on connector HDMI-A-1: -22

Commit 6027cbee1900 ("drm/amd/display: Add error check for avi and
vendor infoframe setup function") started to log the return value of
drm_hdmi_vendor_infoframe_from_display_mode(). With `frame` and `mode`
being non-NULL, the helper only fails with -EINVAL, if the sink does
not support HDMI infoframes, that means, if
`connector-&gt;display_info.has_hdmi_infoframe` is false, and its
kernel-doc says, that this error can be safely ignored.
drm_parse_hdmi_vsdb_video() only sets `has_hdmi_infoframe`, if the HDMI
VSDB has HDMI_Video_present set, and drm_parse_hdmi_forum_scds() sets
it for sinks with an HF-SCDB. So an HDMI sink, whose VSDB lacks
HDMI_Video_present, triggers the warning, although nothing failed.

The EIZO EV2781 is such a sink. Its HDMI VSDB has a payload of only
seven bytes (source physical address, Supports_AI, and a maximum TMDS
clock of 250 MHz), so the byte with HDMI_Video_present is absent, and
the EDID has no HF-SCDB:

    67 03 0c 00 10 00 80 32

Only warn, if the sink supports HDMI infoframes. The infoframe setup
itself is unchanged.

Fixes: 6027cbee1900 ("drm/amd/display: Add error check for avi and vendor infoframe setup function")
Closes: https://gitlab.freedesktop.org/drm/amd/-/work_items/5025
Reviewed-by: Wayne Lin &lt;Wayne.Lin@amd.com&gt;
Signed-off-by: Paul Menzel &lt;pmenzel@molgen.mpg.de&gt;
Signed-off-by: Wayne Lin &lt;Wayne.Lin@amd.com&gt;
Signed-off-by: Alex Deucher &lt;alexander.deucher@amd.com&gt;
(cherry picked from commit 2abfc667f622eb7132b139db2e175642713662f8)
Cc: stable@vger.kernel.org
</content>
<content type='xhtml'>
<div xmlns='http://www.w3.org/1999/xhtml'>
<pre>
With an EIZO EV2781 connected over HDMI to an AMD Ryzen 3 2200G (Raven)
on an MSI B350M MORTAR (MS-7A37), Linux logs at every boot since v6.16:

    amdgpu 0000:30:00.0: [drm] Failed to setup vendor infoframe on connector HDMI-A-1: -22

Commit 6027cbee1900 ("drm/amd/display: Add error check for avi and
vendor infoframe setup function") started to log the return value of
drm_hdmi_vendor_infoframe_from_display_mode(). With `frame` and `mode`
being non-NULL, the helper only fails with -EINVAL, if the sink does
not support HDMI infoframes, that means, if
`connector-&gt;display_info.has_hdmi_infoframe` is false, and its
kernel-doc says, that this error can be safely ignored.
drm_parse_hdmi_vsdb_video() only sets `has_hdmi_infoframe`, if the HDMI
VSDB has HDMI_Video_present set, and drm_parse_hdmi_forum_scds() sets
it for sinks with an HF-SCDB. So an HDMI sink, whose VSDB lacks
HDMI_Video_present, triggers the warning, although nothing failed.

The EIZO EV2781 is such a sink. Its HDMI VSDB has a payload of only
seven bytes (source physical address, Supports_AI, and a maximum TMDS
clock of 250 MHz), so the byte with HDMI_Video_present is absent, and
the EDID has no HF-SCDB:

    67 03 0c 00 10 00 80 32

Only warn, if the sink supports HDMI infoframes. The infoframe setup
itself is unchanged.

Fixes: 6027cbee1900 ("drm/amd/display: Add error check for avi and vendor infoframe setup function")
Closes: https://gitlab.freedesktop.org/drm/amd/-/work_items/5025
Reviewed-by: Wayne Lin &lt;Wayne.Lin@amd.com&gt;
Signed-off-by: Paul Menzel &lt;pmenzel@molgen.mpg.de&gt;
Signed-off-by: Wayne Lin &lt;Wayne.Lin@amd.com&gt;
Signed-off-by: Alex Deucher &lt;alexander.deucher@amd.com&gt;
(cherry picked from commit 2abfc667f622eb7132b139db2e175642713662f8)
Cc: stable@vger.kernel.org
</pre>
</div>
</content>
</entry>
<entry>
<title>drm/amd: Select the firmware signing scheme the PSP reports</title>
<updated>2026-11-18T06:49:31+00:00</updated>
<author>
<name>Mario Limonciello</name>
<email>mario.limonciello@amd.com</email>
</author>
<published>2026-11-14T16:09:25+00:00</published>
<link rel='alternate' type='text/html' href='https://git.toradex.cn/cgit/linux-toradex.git/commit/?id=6d8e709912753f175256e4fafb73ba1ea50baf9f'/>
<id>6d8e709912753f175256e4fafb73ba1ea50baf9f</id>
<content type='text'>
Some platforms ship BIOS with either a default or an alternative trust
authority, so firmware for every IP must be signed with the matching
scheme.
The PSP advertises which one is in use in bit 6 of C2PMSG_63, and in bit
28 whether this is a preproduction part, which needs firmware of its own.

Add amdgpu_psp_init_fw_variant() which reads those bits through amdgpu's
own MMIO, and keep the result in adev-&gt;firmware.sign_type next to
load_type, where preproduction is a type of its own. It is initialized
once, after IP discovery has populated the register bases and only when
firmware is loaded through the front door (the PSP). Only MP0 15.0.0 and
15.0.9 report it. For all other ASICs, SR-IOV VFs and an inaccessible
register (all ones) the defaults are kept: AMDGPU_FW_SIGN_DEFAULT. A
preproduction part that also reports the alternative scheme is not
expected. Mark it AMDGPU_FW_SIGN_UNKNOWN, so that firmware requests warn
and fail with -EINVAL rather than load firmware that would be rejected.

Firmware for preproduction parts is named with a "_p" suffix before the
extension and firmware for the alternative scheme with an "_l" suffix
(e.g. "foo.bin" becomes "foo_p.bin" or "foo_l.bin"). Production parts on
the default scheme use the plain names. Do the substitution centrally in
amdgpu_ucode_request() so it covers every IP that loads firmware through
it (PSP, SMU, DMUB, GC, IMU, MES, SDMA, VCN, VPE, ...) without each
loader having to care. There is no fallback to the default file: it would
be rejected anyway. Declare the variants with MODULE_FIRMWARE() so that
initramfs generators include them.

Acked-by: Alex Deucher &lt;alexander.deucher@amd.com&gt;
Link: https://patch.msgid.link/20261114160925.390350-1-superm1@kernel.org
Signed-off-by: Mario Limonciello &lt;mario.limonciello@amd.com&gt;
Signed-off-by: Alex Deucher &lt;alexander.deucher@amd.com&gt;
(cherry picked from commit 2fe41c5fb704e05ccf2bf00634e1b5ad2db4e5b7)
</content>
<content type='xhtml'>
<div xmlns='http://www.w3.org/1999/xhtml'>
<pre>
Some platforms ship BIOS with either a default or an alternative trust
authority, so firmware for every IP must be signed with the matching
scheme.
The PSP advertises which one is in use in bit 6 of C2PMSG_63, and in bit
28 whether this is a preproduction part, which needs firmware of its own.

Add amdgpu_psp_init_fw_variant() which reads those bits through amdgpu's
own MMIO, and keep the result in adev-&gt;firmware.sign_type next to
load_type, where preproduction is a type of its own. It is initialized
once, after IP discovery has populated the register bases and only when
firmware is loaded through the front door (the PSP). Only MP0 15.0.0 and
15.0.9 report it. For all other ASICs, SR-IOV VFs and an inaccessible
register (all ones) the defaults are kept: AMDGPU_FW_SIGN_DEFAULT. A
preproduction part that also reports the alternative scheme is not
expected. Mark it AMDGPU_FW_SIGN_UNKNOWN, so that firmware requests warn
and fail with -EINVAL rather than load firmware that would be rejected.

Firmware for preproduction parts is named with a "_p" suffix before the
extension and firmware for the alternative scheme with an "_l" suffix
(e.g. "foo.bin" becomes "foo_p.bin" or "foo_l.bin"). Production parts on
the default scheme use the plain names. Do the substitution centrally in
amdgpu_ucode_request() so it covers every IP that loads firmware through
it (PSP, SMU, DMUB, GC, IMU, MES, SDMA, VCN, VPE, ...) without each
loader having to care. There is no fallback to the default file: it would
be rejected anyway. Declare the variants with MODULE_FIRMWARE() so that
initramfs generators include them.

Acked-by: Alex Deucher &lt;alexander.deucher@amd.com&gt;
Link: https://patch.msgid.link/20261114160925.390350-1-superm1@kernel.org
Signed-off-by: Mario Limonciello &lt;mario.limonciello@amd.com&gt;
Signed-off-by: Alex Deucher &lt;alexander.deucher@amd.com&gt;
(cherry picked from commit 2fe41c5fb704e05ccf2bf00634e1b5ad2db4e5b7)
</pre>
</div>
</content>
</entry>
<entry>
<title>drm/amd/display: Apply the HDMI RGB range default only to HDMI signals</title>
<updated>2026-11-18T06:48:47+00:00</updated>
<author>
<name>Adrian Betschart</name>
<email>adrian.betschart@cinemaone.ch</email>
</author>
<published>2026-10-29T07:45:58+00:00</published>
<link rel='alternate' type='text/html' href='https://git.toradex.cn/cgit/linux-toradex.git/commit/?id=6fd6e75162065c7ac9a4a7f3cf303574f0505d97'/>
<id>6fd6e75162065c7ac9a4a7f3cf303574f0505d97</id>
<content type='text'>
Since d23f735d0eac, Broadcast RGB "Automatic" selects limited range on
CTA modes when display_info.is_hdmi is set. That flag comes from an HDMI
VSDB in the EDID and says nothing about the link, so a native
DisplayPort monitor whose EDID carries an HDMI VSDB now gets limited
range on CTA modes too. Monitors that ignore the DP range signal show
grey blacks, where 7.2 sent full range.

Require an HDMI signal as well, as the original commit intended. DP++
passive adapters drive an HDMI signal and keep the HDMI behaviour;
native DP, DP MST and active DP-to-HDMI converters get full range again.

Reported and analysed by Andrés Rodríguez in the issue below.

Fixes: d23f735d0eac ("drm/amd/display: Default HDMI RGB output to limited range on CTA modes")
Closes: https://gitlab.freedesktop.org/drm/amd/-/issues/6052
Reviewed-by: Wayne Lin &lt;Wayne.Lin@amd.com&gt;
Signed-off-by: Adrian Betschart &lt;adrian.betschart@cinemaone.ch&gt;
Signed-off-by: Wayne Lin &lt;Wayne.Lin@amd.com&gt;
Assisted-by: Claude Code:claude-opus-5-5
Signed-off-by: Alex Deucher &lt;alexander.deucher@amd.com&gt;
(cherry picked from commit 39f37a2346410a33edf4af17a873ac96bb5244ce)
Cc: stable@vger.kernel.org # 7.3
</content>
<content type='xhtml'>
<div xmlns='http://www.w3.org/1999/xhtml'>
<pre>
Since d23f735d0eac, Broadcast RGB "Automatic" selects limited range on
CTA modes when display_info.is_hdmi is set. That flag comes from an HDMI
VSDB in the EDID and says nothing about the link, so a native
DisplayPort monitor whose EDID carries an HDMI VSDB now gets limited
range on CTA modes too. Monitors that ignore the DP range signal show
grey blacks, where 7.2 sent full range.

Require an HDMI signal as well, as the original commit intended. DP++
passive adapters drive an HDMI signal and keep the HDMI behaviour;
native DP, DP MST and active DP-to-HDMI converters get full range again.

Reported and analysed by Andrés Rodríguez in the issue below.

Fixes: d23f735d0eac ("drm/amd/display: Default HDMI RGB output to limited range on CTA modes")
Closes: https://gitlab.freedesktop.org/drm/amd/-/issues/6052
Reviewed-by: Wayne Lin &lt;Wayne.Lin@amd.com&gt;
Signed-off-by: Adrian Betschart &lt;adrian.betschart@cinemaone.ch&gt;
Signed-off-by: Wayne Lin &lt;Wayne.Lin@amd.com&gt;
Assisted-by: Claude Code:claude-opus-5-5
Signed-off-by: Alex Deucher &lt;alexander.deucher@amd.com&gt;
(cherry picked from commit 39f37a2346410a33edf4af17a873ac96bb5244ce)
Cc: stable@vger.kernel.org # 7.3
</pre>
</div>
</content>
</entry>
<entry>
<title>drm/amd/display: Take every stream out of self-refresh before a modeset</title>
<updated>2026-11-18T06:47:25+00:00</updated>
<author>
<name>Hugo Sales</name>
<email>hugo@hsal.es</email>
</author>
<published>2026-10-29T16:54:20+00:00</published>
<link rel='alternate' type='text/html' href='https://git.toradex.cn/cgit/linux-toradex.git/commit/?id=09b6bf1a04cc6bbcbb8bb29fa1e38fa8a9f8d1dc'/>
<id>09b6bf1a04cc6bbcbb8bb29fa1e38fa8a9f8d1dc</id>
<content type='text'>
Commit 3c108046e1d6 ("drm/amd/display: Add power module on Linux")
dropped amdgpu_dm_psr_disable_all() and amdgpu_dm_replay_disable_all()
from amdgpu_dm_commit_streams(). Since then, a modeset only disables
self-refresh on the CRTCs in the commit, so an eDP on another CRTC stays
in PSR through dc_commit_streams()

Commit 6ee90e88b976 ("drm/amd/display: PSR eDP p-state warning occurs
intermittently after unplug DP") added that disable-all for this very
case. With it gone, turning off or unplugging an HDMI monitor while the
eDP is in PSR hangs DMUB on a Cezanne laptop (DCN 2.1), on the first
try:

  dc_dmub_srv_log_diagnostic_data: DMCUB error
  Error queueing DMUB command: status=2

The eDP then stays frozen until a GPU reset

Do the same through the power module's events. When a CRTC that is or
was active needs a modeset, raise the PSR and Replay pause events on
every stream mod_power tracks and wait for self-refresh to exit, before
amdgpu_dm_mod_power_update_streams() changes the stream set. Clear them
on the commit's streams after amdgpu_dm_mod_power_setup_streams(), since
enabling a stream clears psr_feature_enabled until then. Use
psr_event_pause and replay_event_pause rather than the hw_programming
events, which a fast update on the eDP can clear in
amdgpu_dm_enable_self_refresh() while the modeset is still running

Fixes: 3c108046e1d6 ("drm/amd/display: Add power module on Linux")
Reviewed-by: Ray Wu &lt;ray.wu@amd.com&gt;
Signed-off-by: Hugo Sales &lt;hugo@hsal.es&gt;
Signed-off-by: Wayne Lin &lt;Wayne.Lin@amd.com&gt;
Signed-off-by: Alex Deucher &lt;alexander.deucher@amd.com&gt;
(cherry picked from commit 5d90fdb959505b5f42f2dec699521ac56b4174be)
Cc: stable@vger.kernel.org
</content>
<content type='xhtml'>
<div xmlns='http://www.w3.org/1999/xhtml'>
<pre>
Commit 3c108046e1d6 ("drm/amd/display: Add power module on Linux")
dropped amdgpu_dm_psr_disable_all() and amdgpu_dm_replay_disable_all()
from amdgpu_dm_commit_streams(). Since then, a modeset only disables
self-refresh on the CRTCs in the commit, so an eDP on another CRTC stays
in PSR through dc_commit_streams()

Commit 6ee90e88b976 ("drm/amd/display: PSR eDP p-state warning occurs
intermittently after unplug DP") added that disable-all for this very
case. With it gone, turning off or unplugging an HDMI monitor while the
eDP is in PSR hangs DMUB on a Cezanne laptop (DCN 2.1), on the first
try:

  dc_dmub_srv_log_diagnostic_data: DMCUB error
  Error queueing DMUB command: status=2

The eDP then stays frozen until a GPU reset

Do the same through the power module's events. When a CRTC that is or
was active needs a modeset, raise the PSR and Replay pause events on
every stream mod_power tracks and wait for self-refresh to exit, before
amdgpu_dm_mod_power_update_streams() changes the stream set. Clear them
on the commit's streams after amdgpu_dm_mod_power_setup_streams(), since
enabling a stream clears psr_feature_enabled until then. Use
psr_event_pause and replay_event_pause rather than the hw_programming
events, which a fast update on the eDP can clear in
amdgpu_dm_enable_self_refresh() while the modeset is still running

Fixes: 3c108046e1d6 ("drm/amd/display: Add power module on Linux")
Reviewed-by: Ray Wu &lt;ray.wu@amd.com&gt;
Signed-off-by: Hugo Sales &lt;hugo@hsal.es&gt;
Signed-off-by: Wayne Lin &lt;Wayne.Lin@amd.com&gt;
Signed-off-by: Alex Deucher &lt;alexander.deucher@amd.com&gt;
(cherry picked from commit 5d90fdb959505b5f42f2dec699521ac56b4174be)
Cc: stable@vger.kernel.org
</pre>
</div>
</content>
</entry>
<entry>
<title>drm/amd/display: Stop HDMI FRL status polling on fini</title>
<updated>2026-11-18T06:46:10+00:00</updated>
<author>
<name>Fangzhi Zuo</name>
<email>jerry.zuo@amd.com</email>
</author>
<published>2026-10-24T04:51:40+00:00</published>
<link rel='alternate' type='text/html' href='https://git.toradex.cn/cgit/linux-toradex.git/commit/?id=449141afc84a4434db864abf9051dc8beb3b7c81'/>
<id>449141afc84a4434db864abf9051dc8beb3b7c81</id>
<content type='text'>
[Why]
The HDMI FRL status polling work re-queues itself and its workqueue was not
stopped during amdgpu_dm_fini(). During unbind or unload, the worker could
access freed DC state, and the workqueue was leaked. This fixes the FRL
polling teardown introduced by the HDMI FRL support change.

[How]
Cancel the delayed polling work synchronously and destroy the workqueue in
amdgpu_dm_fini() before the remaining display state is torn down.

Cc: Harry Wentland &lt;harry.wentland@amd.com&gt;
Reported-by: Eren Özen &lt;erenozen@protonmail.com&gt;
Link: https://lore.kernel.org/all/20261110-frl-polling-wq-teardown-v1-0-0ff028d5c35b@protonmail.com/
Reviewed-by: ChiaHsuan Chung &lt;chiahsuan.chung@amd.com&gt;
Signed-off-by: Fangzhi Zuo &lt;jerry.zuo@amd.com&gt;
Signed-off-by: Wayne Lin &lt;wayne.lin@amd.com&gt;
Tested-by: Dan Wheeler &lt;daniel.wheeler@amd.com&gt;
Signed-off-by: Alex Deucher &lt;alexander.deucher@amd.com&gt;
(cherry picked from commit 709b7a5aff3addf2288dedf42bc62c45ae1ab163)
</content>
<content type='xhtml'>
<div xmlns='http://www.w3.org/1999/xhtml'>
<pre>
[Why]
The HDMI FRL status polling work re-queues itself and its workqueue was not
stopped during amdgpu_dm_fini(). During unbind or unload, the worker could
access freed DC state, and the workqueue was leaked. This fixes the FRL
polling teardown introduced by the HDMI FRL support change.

[How]
Cancel the delayed polling work synchronously and destroy the workqueue in
amdgpu_dm_fini() before the remaining display state is torn down.

Cc: Harry Wentland &lt;harry.wentland@amd.com&gt;
Reported-by: Eren Özen &lt;erenozen@protonmail.com&gt;
Link: https://lore.kernel.org/all/20261110-frl-polling-wq-teardown-v1-0-0ff028d5c35b@protonmail.com/
Reviewed-by: ChiaHsuan Chung &lt;chiahsuan.chung@amd.com&gt;
Signed-off-by: Fangzhi Zuo &lt;jerry.zuo@amd.com&gt;
Signed-off-by: Wayne Lin &lt;wayne.lin@amd.com&gt;
Tested-by: Dan Wheeler &lt;daniel.wheeler@amd.com&gt;
Signed-off-by: Alex Deucher &lt;alexander.deucher@amd.com&gt;
(cherry picked from commit 709b7a5aff3addf2288dedf42bc62c45ae1ab163)
</pre>
</div>
</content>
</entry>
</feed>
