[PATCH 1/2] drm/amd/display: Fix preferred link rate for NUTMEG

[PATCH 1/2] drm/amd/display: Fix preferred link rate for NUTMEG

[Timur Kristóf]

When there is a preferred link rate setting, it needs to be
applied to both the current and initial link rate.
This was regressed by a "coding style" fix, which caused
the current link rate to not respect the preferred value.

This commit restores the functionality of NUTMEG,
the DP bridge encoder found on old APUs such as Kaveri.

Fixes: b74322eea36b ("drm/amd/display: Fix coding style issue")
Cc: Chuanyu Tseng <Chuanyu.Tseng@amd.com>
Signed-off-by: Timur Kristóf <timur.kristof@gmail.com>
---
 .../drm/amd/display/dc/link/protocols/link_dp_capability.c    | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)

diff --git a/drivers/gpu/drm/amd/display/dc/link/protocols/link_dp_capability.c b/drivers/gpu/drm/amd/display/dc/link/protocols/link_dp_capability.c
index 817b4010edcbe..f44c13300a59c 100644
--- a/drivers/gpu/drm/amd/display/dc/link/protocols/link_dp_capability.c
+++ b/drivers/gpu/drm/amd/display/dc/link/protocols/link_dp_capability.c
@@ -750,8 +750,10 @@ static bool decide_dp_link_settings(struct dc_link *link, struct dc_link_setting
 	if (req_bw > dp_link_bandwidth_kbps(link, &link->verified_link_cap))
 		return false;
 
-	if (link->preferred_link_setting.link_rate != LINK_RATE_UNKNOWN)
+	if (link->preferred_link_setting.link_rate != LINK_RATE_UNKNOWN) {
 		initial_link_setting.link_rate = link->preferred_link_setting.link_rate;
+		current_link_setting.link_rate = link->preferred_link_setting.link_rate;
+	}
 
 	/* search for the minimum link setting that:
 	 * 1. is supported according to the link training result
-- 
2.53.0

[PATCH 2/2] drm/amd/display: Add dp_skip_rbr flag for NUTMEG

[Timur Kristóf]

No functional changes. Just clean up a conceptual mismatch.

Based on feedback on the NUTMEG code in DC, the
preferred_link_setting is meant to force the DP link to a
specific setting, meaning both the link rate and lane count
should be locked to an exact value. What NUTMEG needs is
a lower bound on the link rate, which is not the same concept.

Implement this as a HW workaround flag instead.

Suggested-by: Wenjing Liu <wenjing.liu@amd.com>
Signed-off-by: Timur Kristóf <timur.kristof@gmail.com>
---
 drivers/gpu/drm/amd/display/dc/dc.h                         | 2 ++
 drivers/gpu/drm/amd/display/dc/link/link_detection.c        | 2 +-
 .../drm/amd/display/dc/link/protocols/link_dp_capability.c  | 6 +++---
 3 files changed, 6 insertions(+), 4 deletions(-)

diff --git a/drivers/gpu/drm/amd/display/dc/dc.h b/drivers/gpu/drm/amd/display/dc/dc.h
index 75f59ca927ffe..3fb97860722df 100644
--- a/drivers/gpu/drm/amd/display/dc/dc.h
+++ b/drivers/gpu/drm/amd/display/dc/dc.h
@@ -1731,6 +1731,8 @@ struct dc_scratch_space {
 		bool dp_skip_DID2;
 		bool dp_skip_reset_segment;
 		bool dp_skip_fs_144hz;
+		/* Some DP bridges don't work with RBR and must use HBR. */
+		bool dp_skip_rbr;
 		bool dp_mot_reset_segment;
 		/* Some USB4 docks do not handle turning off MST DSC once it has been enabled. */
 		bool dpia_mst_dsc_always_on;
diff --git a/drivers/gpu/drm/amd/display/dc/link/link_detection.c b/drivers/gpu/drm/amd/display/dc/link/link_detection.c
index 1ac6a22fecfee..9f4e90493fed5 100644
--- a/drivers/gpu/drm/amd/display/dc/link/link_detection.c
+++ b/drivers/gpu/drm/amd/display/dc/link/link_detection.c
@@ -621,7 +621,7 @@ static bool detect_dp(struct dc_link *link,
 		link->dpcd_caps.sink_count.bits.SINK_COUNT = 1;
 		/* NUTMEG requires that we use HBR, doesn't work with RBR. */
 		if (link->dpcd_caps.branch_dev_id == DP_BRANCH_DEVICE_ID_00001A)
-			link->preferred_link_setting.link_rate = LINK_RATE_HIGH;
+			link->wa_flags.dp_skip_rbr = true;
 	}
 
 	return true;
diff --git a/drivers/gpu/drm/amd/display/dc/link/protocols/link_dp_capability.c b/drivers/gpu/drm/amd/display/dc/link/protocols/link_dp_capability.c
index f44c13300a59c..708dd6f336d8c 100644
--- a/drivers/gpu/drm/amd/display/dc/link/protocols/link_dp_capability.c
+++ b/drivers/gpu/drm/amd/display/dc/link/protocols/link_dp_capability.c
@@ -750,9 +750,9 @@ static bool decide_dp_link_settings(struct dc_link *link, struct dc_link_setting
 	if (req_bw > dp_link_bandwidth_kbps(link, &link->verified_link_cap))
 		return false;
 
-	if (link->preferred_link_setting.link_rate != LINK_RATE_UNKNOWN) {
-		initial_link_setting.link_rate = link->preferred_link_setting.link_rate;
-		current_link_setting.link_rate = link->preferred_link_setting.link_rate;
+	if (link->wa_flags.dp_skip_rbr) {
+		initial_link_setting.link_rate = LINK_RATE_HIGH;
+		current_link_setting.link_rate = LINK_RATE_HIGH;
 	}
 
 	/* search for the minimum link setting that:
-- 
2.53.0

Re: [PATCH 2/2] drm/amd/display: Add dp_skip_rbr flag for NUTMEG

[Alex Deucher]

Applied the series.  Thanks!

Alex

On Fri, May 29, 2026 at 5:34 AM Timur Kristóf <timur.kristof@gmail.com> wrote:
>
> No functional changes. Just clean up a conceptual mismatch.
>
> Based on feedback on the NUTMEG code in DC, the
> preferred_link_setting is meant to force the DP link to a
> specific setting, meaning both the link rate and lane count
> should be locked to an exact value. What NUTMEG needs is
> a lower bound on the link rate, which is not the same concept.
>
> Implement this as a HW workaround flag instead.
>
> Suggested-by: Wenjing Liu <wenjing.liu@amd.com>
> Signed-off-by: Timur Kristóf <timur.kristof@gmail.com>
> ---
>  drivers/gpu/drm/amd/display/dc/dc.h                         | 2 ++
>  drivers/gpu/drm/amd/display/dc/link/link_detection.c        | 2 +-
>  .../drm/amd/display/dc/link/protocols/link_dp_capability.c  | 6 +++---
>  3 files changed, 6 insertions(+), 4 deletions(-)
>
> diff --git a/drivers/gpu/drm/amd/display/dc/dc.h b/drivers/gpu/drm/amd/display/dc/dc.h
> index 75f59ca927ffe..3fb97860722df 100644
> --- a/drivers/gpu/drm/amd/display/dc/dc.h
> +++ b/drivers/gpu/drm/amd/display/dc/dc.h
> @@ -1731,6 +1731,8 @@ struct dc_scratch_space {
>                 bool dp_skip_DID2;
>                 bool dp_skip_reset_segment;
>                 bool dp_skip_fs_144hz;
> +               /* Some DP bridges don't work with RBR and must use HBR. */
> +               bool dp_skip_rbr;
>                 bool dp_mot_reset_segment;
>                 /* Some USB4 docks do not handle turning off MST DSC once it has been enabled. */
>                 bool dpia_mst_dsc_always_on;
> diff --git a/drivers/gpu/drm/amd/display/dc/link/link_detection.c b/drivers/gpu/drm/amd/display/dc/link/link_detection.c
> index 1ac6a22fecfee..9f4e90493fed5 100644
> --- a/drivers/gpu/drm/amd/display/dc/link/link_detection.c
> +++ b/drivers/gpu/drm/amd/display/dc/link/link_detection.c
> @@ -621,7 +621,7 @@ static bool detect_dp(struct dc_link *link,
>                 link->dpcd_caps.sink_count.bits.SINK_COUNT = 1;
>                 /* NUTMEG requires that we use HBR, doesn't work with RBR. */
>                 if (link->dpcd_caps.branch_dev_id == DP_BRANCH_DEVICE_ID_00001A)
> -                       link->preferred_link_setting.link_rate = LINK_RATE_HIGH;
> +                       link->wa_flags.dp_skip_rbr = true;
>         }
>
>         return true;
> diff --git a/drivers/gpu/drm/amd/display/dc/link/protocols/link_dp_capability.c b/drivers/gpu/drm/amd/display/dc/link/protocols/link_dp_capability.c
> index f44c13300a59c..708dd6f336d8c 100644
> --- a/drivers/gpu/drm/amd/display/dc/link/protocols/link_dp_capability.c
> +++ b/drivers/gpu/drm/amd/display/dc/link/protocols/link_dp_capability.c
> @@ -750,9 +750,9 @@ static bool decide_dp_link_settings(struct dc_link *link, struct dc_link_setting
>         if (req_bw > dp_link_bandwidth_kbps(link, &link->verified_link_cap))
>                 return false;
>
> -       if (link->preferred_link_setting.link_rate != LINK_RATE_UNKNOWN) {
> -               initial_link_setting.link_rate = link->preferred_link_setting.link_rate;
> -               current_link_setting.link_rate = link->preferred_link_setting.link_rate;
> +       if (link->wa_flags.dp_skip_rbr) {
> +               initial_link_setting.link_rate = LINK_RATE_HIGH;
> +               current_link_setting.link_rate = LINK_RATE_HIGH;
>         }
>
>         /* search for the minimum link setting that:
> --
> 2.53.0
>

RE: [PATCH 1/2] drm/amd/display: Fix preferred link rate for NUTMEG

[Zuo, Jerry]

AMD General

Reviewed-by: Fangzhi Zuo <Jerry.Zuo@amd.com>

> -----Original Message-----
> From: amd-gfx <amd-gfx-bounces@lists.freedesktop.org> On Behalf Of Timur
> Kristóf
> Sent: Friday, May 29, 2026 05:09
> To: amd-gfx@lists.freedesktop.org; Deucher, Alexander
> <Alexander.Deucher@amd.com>; LIPSKI, IVAN <IVAN.LIPSKI@amd.com>;
> Wentland, Harry <Harry.Wentland@amd.com>; Hung, Alex
> <Alex.Hung@amd.com>; Wu, Ray <Ray.Wu@amd.com>; Liu, Wenjing
> <Wenjing.Liu@amd.com>; Pillai, Aurabindo <Aurabindo.Pillai@amd.com>;
> Chuanyu Tseng <Chuanyu.Tseng@amd.com>; Li, Roman
> <Roman.Li@amd.com>; Wheeler, Daniel <Daniel.Wheeler@amd.com>
> Cc: Timur Kristóf <timur.kristof@gmail.com>
> Subject: [PATCH 1/2] drm/amd/display: Fix preferred link rate for NUTMEG
>
> When there is a preferred link rate setting, it needs to be applied to both the
> current and initial link rate.
> This was regressed by a "coding style" fix, which caused the current link rate
> to not respect the preferred value.
>
> This commit restores the functionality of NUTMEG, the DP bridge encoder
> found on old APUs such as Kaveri.
>
> Fixes: b74322eea36b ("drm/amd/display: Fix coding style issue")
> Cc: Chuanyu Tseng <Chuanyu.Tseng@amd.com>
> Signed-off-by: Timur Kristóf <timur.kristof@gmail.com>
> ---
>  .../drm/amd/display/dc/link/protocols/link_dp_capability.c    | 4 +++-
>  1 file changed, 3 insertions(+), 1 deletion(-)
>
> diff --git
> a/drivers/gpu/drm/amd/display/dc/link/protocols/link_dp_capability.c
> b/drivers/gpu/drm/amd/display/dc/link/protocols/link_dp_capability.c
> index 817b4010edcbe..f44c13300a59c 100644
> --- a/drivers/gpu/drm/amd/display/dc/link/protocols/link_dp_capability.c
> +++ b/drivers/gpu/drm/amd/display/dc/link/protocols/link_dp_capability.c
> @@ -750,8 +750,10 @@ static bool decide_dp_link_settings(struct dc_link
> *link, struct dc_link_setting
>       if (req_bw > dp_link_bandwidth_kbps(link, &link->verified_link_cap))
>               return false;
>
> -     if (link->preferred_link_setting.link_rate != LINK_RATE_UNKNOWN)
> +     if (link->preferred_link_setting.link_rate != LINK_RATE_UNKNOWN) {
>               initial_link_setting.link_rate = link-
> >preferred_link_setting.link_rate;
> +             current_link_setting.link_rate = link-
> >preferred_link_setting.link_rate;
> +     }
>
>       /* search for the minimum link setting that:
>        * 1. is supported according to the link training result
> --
> 2.53.0

Re: [PATCH 1/2] drm/amd/display: Fix preferred link rate for NUTMEG

[John Olender]

Ping.

On 5/29/26 5:09 AM, Timur Kristóf wrote:
> When there is a preferred link rate setting, it needs to be
> applied to both the current and initial link rate.
> This was regressed by a "coding style" fix, which caused
> the current link rate to not respect the preferred value.
> 
> This commit restores the functionality of NUTMEG,
> the DP bridge encoder found on old APUs such as Kaveri.
> 
> Fixes: b74322eea36b ("drm/amd/display: Fix coding style issue")
> Cc: Chuanyu Tseng <Chuanyu.Tseng@amd.com>
> Signed-off-by: Timur Kristóf <timur.kristof@gmail.com>

A user recently reported this issue to the tracker and confirmed
this patch is an effective fix.

Closes: https://gitlab.freedesktop.org/drm/amd/-/work_items/5465

Thanks,
John

> ---
>  .../drm/amd/display/dc/link/protocols/link_dp_capability.c    | 4 +++-
>  1 file changed, 3 insertions(+), 1 deletion(-)
> 
> diff --git a/drivers/gpu/drm/amd/display/dc/link/protocols/link_dp_capability.c b/drivers/gpu/drm/amd/display/dc/link/protocols/link_dp_capability.c
> index 817b4010edcbe..f44c13300a59c 100644
> --- a/drivers/gpu/drm/amd/display/dc/link/protocols/link_dp_capability.c
> +++ b/drivers/gpu/drm/amd/display/dc/link/protocols/link_dp_capability.c
> @@ -750,8 +750,10 @@ static bool decide_dp_link_settings(struct dc_link *link, struct dc_link_setting
>  	if (req_bw > dp_link_bandwidth_kbps(link, &link->verified_link_cap))
>  		return false;
>  
> -	if (link->preferred_link_setting.link_rate != LINK_RATE_UNKNOWN)
> +	if (link->preferred_link_setting.link_rate != LINK_RATE_UNKNOWN) {
>  		initial_link_setting.link_rate = link->preferred_link_setting.link_rate;
> +		current_link_setting.link_rate = link->preferred_link_setting.link_rate;
> +	}
>  
>  	/* search for the minimum link setting that:
>  	 * 1. is supported according to the link training result

Re: [PATCH 1/2] drm/amd/display: Fix preferred link rate for NUTMEG

[Lionso Alejandro Pacheco Vacacela]

Hi,

I independently hit this regression on a desktop Kaveri system and can
confirm this patch fixes it:

  - AMD A10-7850K (Kaveri, DCE 8.1), MSI A68HM-E33, VGA monitor on the
    onboard output (NUTMEG DP-to-VGA bridge)
  - Debian kernel 7.1.3 (amdgpu.cik_support=1): VGA monitor receives no
    signal from KMS takeover on; dmesg shows
    "[drm] enabling link 2 failed: 15" (DC_FAIL_DP_LINK_TRAINING).
  - Applying just this patch (1/2) on top of v7.1.3 restores VGA output
    with DC enabled: the link trains, the CRTC comes up at 1360x768@60,
    and the fix survived three consecutive reboots.

Tested-by: Lionso Alejandro Pacheco Vacacela <lionso@murena.io>

Two notes:

1. The Fixes: tag references b74322eea36b, but that hash does not exist
   in mainline — it looks like an amd-staging-internal id. In Linus'
   tree the offending commit is:

     a62346043a89 ("drm/amd/display: Fix coding style issue")

   It would be good to correct this before the patch is pushed so the
   stable tooling can resolve the dependency.

2. Since the offending commit shipped in v7.1, could this be annotated
   Cc: stable@vger.kernel.org # 7.1.x (or nominated for 7.1.y once it
   lands in mainline)? Every Kaveri/Kabini/Mullins system with a VGA
   monitor loses display output on release kernels.

Thanks,
Lionso

Re: [PATCH 1/2] drm/amd/display: Fix preferred link rate for NUTMEG

[Timur Kristóf]

On 2026. július 11., szombat 21:35:07 közép-európai nyári idő Lionso Alejandro 
Pacheco Vacacela wrote:
> Hi,
> 
> I independently hit this regression on a desktop Kaveri system and can
> confirm this patch fixes it:
> 
>   - AMD A10-7850K (Kaveri, DCE 8.1), MSI A68HM-E33, VGA monitor on the
>     onboard output (NUTMEG DP-to-VGA bridge)
>   - Debian kernel 7.1.3 (amdgpu.cik_support=1): VGA monitor receives no
>     signal from KMS takeover on; dmesg shows
>     "[drm] enabling link 2 failed: 15" (DC_FAIL_DP_LINK_TRAINING).
>   - Applying just this patch (1/2) on top of v7.1.3 restores VGA output
>     with DC enabled: the link trains, the CRTC comes up at 1360x768@60,
>     and the fix survived three consecutive reboots.
> 
> Tested-by: Lionso Alejandro Pacheco Vacacela <lionso@murena.io>

Thank you for the testing!
I would appreciate if you could verify that the second patch also works well.

> 
> Two notes:
> 
> 1. The Fixes: tag references b74322eea36b, but that hash does not exist
>    in mainline — it looks like an amd-staging-internal id. In Linus'
>    tree the offending commit is:
> 
>      a62346043a89 ("drm/amd/display: Fix coding style issue")
> 
>    It would be good to correct this before the patch is pushed so the
>    stable tooling can resolve the dependency.

As far as I remember the regression hasn't made it upstream yet when I made 
the fix, so the hash didn't exist in Linus's tree yet.

> 
> 2. Since the offending commit shipped in v7.1, could this be annotated
>    Cc: stable@vger.kernel.org # 7.1.x (or nominated for 7.1.y once it
>    lands in mainline)? Every Kaveri/Kabini/Mullins system with a VGA
>    monitor loses display output on release kernels.

Agreed. Alex can you please backport this to 7.1?

Thanks & best regards,
Timur

Re: [PATCH 1/2] drm/amd/display: Fix preferred link rate for NUTMEG

[Lionso Alejandro Pacheco Vacacela]

Hi Timur,

On Sunday, 12 July 2026 19:44:31 CEST Timur Kristóf wrote:
> Thank you for the testing!
> I would appreciate if you could verify that the second patch also
> works well.

Happy to! I applied both patches (1/2 and 2/2) on top of v7.1.3,
rebuilt, and re-tested on the same machine (A10-7850K / MSI A68HM-E33,
VGA on the onboard output). With DC enabled the NUTMEG link trains, the
CRTC comes up at 1360x768@60, VGA output is back, and it happily
survived a bunch of reboots. dmesg shows "Display Core v3.2.378
initialized on DCE 8.1" with no "enabling link failed" this time.
Nice to have this old box on DC at last.

So for 2/2 as well:

Tested-by: Lionso Alejandro Pacheco Vacacela <lionso@murena.io>

> As far as I remember the regression hasn't made it upstream yet when
> I made the fix, so the hash didn't exist in Linus's tree yet.

Ah, that explains it — makes sense. It has landed since then, so it's
really just a matter of refreshing the tag. In Linus' tree the commit
is:

  a62346043a89 ("drm/amd/display: Fix coding style issue")

and it's an ancestor of v7.1.3 (it shipped in the v7.1 release), so
pointing the Fixes: line at that hash when the fix goes out should let
the stable tooling pull it into 7.1.y on its own.

Thanks a lot for the quick fix, and thanks Alex for picking up the 7.1
backport — much appreciated.

Best regards,
Lionso