Skip to contentSkip to conversation

drm/tegra: fixup primary/overlay format for tegra210

RFC PATCH In review Nicolas Chauvet 3 participants 6 messages Mar 23 – Mar 25

Subsystems (3)
lore
Share & tools
lore
#0 lore raw

RFC PATCH 1/1 drm/tegra: fixup primary/overlay format for tegra210

The primary_format and overlay_format were picked from earlier tegra114 generation instead of using the previous tegra124 format leading to missing format.

This patch is RFC because while it's unlikely that format availability have skipped to earlier tegra210 soc generation I haven't confirmed any runtime error nor experienced any regression by the lack of availability of theses format.

---
 drivers/gpu/drm/tegra/dc.c | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)
drivers/gpu/drm/tegra/dc.c +4 -4
index 06370b7e0e56..d5896f12f25f 100644
@@ -2942,10 +2942,10 @@ static const struct tegra_dc_soc_info tegra210_dc_soc_info = {
 	.has_powergate = true,
 	.coupled_pm = false,
 	.has_nvdisplay = false,
-	.num_primary_formats = ARRAY_SIZE(tegra114_primary_formats),
-	.primary_formats = tegra114_primary_formats,
-	.num_overlay_formats = ARRAY_SIZE(tegra114_overlay_formats),
-	.overlay_formats = tegra114_overlay_formats,
+	.num_primary_formats = ARRAY_SIZE(tegra124_primary_formats),
+	.primary_formats = tegra124_primary_formats,
+	.num_overlay_formats = ARRAY_SIZE(tegra124_overlay_formats),
+	.overlay_formats = tegra124_overlay_formats,
 	.modifiers = tegra124_modifiers,
 	.has_win_a_without_filters = false,
 	.has_win_b_vfilter_mem_client = false,
Signature · 1 line
2.53.0
#1 lore raw

RFC PATCH 1/1 drm/tegra: fixup primary/overlay format for tegra210

The primary_format and overlay_format were picked from earlier tegra114 generation instead of using the previous tegra124 format leading to missing format.

This patch is RFC because while it's unlikely that format availability have skipped to earlier tegra210 soc generation I haven't confirmed any runtime error nor experienced any regression by the lack of availability of theses format.

---
 drivers/gpu/drm/tegra/dc.c | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)
drivers/gpu/drm/tegra/dc.c +4 -4
index 06370b7e0e56..d5896f12f25f 100644
@@ -2942,10 +2942,10 @@ static const struct tegra_dc_soc_info tegra210_dc_soc_info = {
 	.has_powergate = true,
 	.coupled_pm = false,
 	.has_nvdisplay = false,
-	.num_primary_formats = ARRAY_SIZE(tegra114_primary_formats),
-	.primary_formats = tegra114_primary_formats,
-	.num_overlay_formats = ARRAY_SIZE(tegra114_overlay_formats),
-	.overlay_formats = tegra114_overlay_formats,
+	.num_primary_formats = ARRAY_SIZE(tegra124_primary_formats),
+	.primary_formats = tegra124_primary_formats,
+	.num_overlay_formats = ARRAY_SIZE(tegra124_overlay_formats),
+	.overlay_formats = tegra124_overlay_formats,
 	.modifiers = tegra124_modifiers,
 	.has_win_a_without_filters = false,
 	.has_win_b_vfilter_mem_client = false,
Signature · 1 line
2.53.0
maintainer
#2 lore raw
In reply to Nicolas Chauvet
Quoted 34 lines · Nicolas Chauvet

The primary_format and overlay_format were picked from earlier tegra114 generation instead of using the previous tegra124 format leading to missing format.

This patch is RFC because while it's unlikely that format availability have skipped to earlier tegra210 soc generation I haven't confirmed any runtime error nor experienced any regression by the lack of availability of theses format.

---
 drivers/gpu/drm/tegra/dc.c | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)
drivers/gpu/drm/tegra/dc.c +4 -5
index 06370b7e0e56..d5896f12f25f 100644
@@ -2942,10 +2942,10 @@ static const struct tegra_dc_soc_info
 tegra210_dc_soc_info = { .has_powergate = true,
 	.coupled_pm = false,
 	.has_nvdisplay = false,
-	.num_primary_formats = ARRAY_SIZE(tegra114_primary_formats),
-	.primary_formats = tegra114_primary_formats,
-	.num_overlay_formats = ARRAY_SIZE(tegra114_overlay_formats),
-	.overlay_formats = tegra114_overlay_formats,
+	.num_primary_formats = ARRAY_SIZE(tegra124_primary_formats),
+	.primary_formats = tegra124_primary_formats,
+	.num_overlay_formats = ARRAY_SIZE(tegra124_overlay_formats),
+	.overlay_formats = tegra124_overlay_formats,
 	.modifiers = tegra124_modifiers,
 	.has_win_a_without_filters = false,
 	.has_win_b_vfilter_mem_client = false,
--
Quoted 1 line

2.53.0

Looking at the TRMs, I think indeed Tegra210 also supports the DRM_FORMAT_RGBX8888/DRM_FORMAT_BGRX8888 formats, so this patch should be correct. Functionality-wise, the formats aren't handled currently in plane.c, so I don't think there's any change in functionality.

These formats require enabling byteswap -- so do RGBA8888/BGRA8888, but plane.c doesn't currently do that, so I wonder if these formats are just not working right now.

Thanks!
Mikko

#3 lore raw
Quoted 43 lines · Mikko Perttunen
Quoted 34 lines · Nicolas Chauvet

The primary_format and overlay_format were picked from earlier tegra114 generation instead of using the previous tegra124 format leading to missing format.

This patch is RFC because while it's unlikely that format availability have skipped to earlier tegra210 soc generation I haven't confirmed any runtime error nor experienced any regression by the lack of availability of theses format.

---
 drivers/gpu/drm/tegra/dc.c | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)
drivers/gpu/drm/tegra/dc.c +4 -5
index 06370b7e0e56..d5896f12f25f 100644
@@ -2942,10 +2942,10 @@ static const struct tegra_dc_soc_info
 tegra210_dc_soc_info = { .has_powergate = true,
      .coupled_pm = false,
      .has_nvdisplay = false,
-     .num_primary_formats = ARRAY_SIZE(tegra114_primary_formats),
-     .primary_formats = tegra114_primary_formats,
-     .num_overlay_formats = ARRAY_SIZE(tegra114_overlay_formats),
-     .overlay_formats = tegra114_overlay_formats,
+     .num_primary_formats = ARRAY_SIZE(tegra124_primary_formats),
+     .primary_formats = tegra124_primary_formats,
+     .num_overlay_formats = ARRAY_SIZE(tegra124_overlay_formats),
+     .overlay_formats = tegra124_overlay_formats,
      .modifiers = tegra124_modifiers,
      .has_win_a_without_filters = false,
      .has_win_b_vfilter_mem_client = false,
--
Quoted 1 line

2.53.0

Looking at the TRMs, I think indeed Tegra210 also supports the DRM_FORMAT_RGBX8888/DRM_FORMAT_BGRX8888 formats, so this patch should be correct. Functionality-wise, the formats aren't handled currently in plane.c, so I don't think there's any change in functionality.

Quoted 3 lines

These formats require enabling byteswap -- so do RGBA8888/BGRA8888, but plane.c doesn't currently do that, so I wonder if these formats are just not working right now.

Thanks for the review,

For info, I've spotted this when trying to understand a graphic regression in mesa starting with mesa-25.2
https://gitlab.freedesktop.org/mesa/mesa/-/issues/14640

That was initially an issue with mutter been way less efficient than weston for some reason (with earlier mesa-25.1)
https://gitlab.gnome.org/GNOME/mutter/-/issues/4501

I expect something related to how some formats are selected/advertised...

maintainer
#4 lore raw
In reply to Mikko Perttunen
Quoted 42 lines · Mikko Perttunen
Quoted 34 lines · Nicolas Chauvet

The primary_format and overlay_format were picked from earlier tegra114 generation instead of using the previous tegra124 format leading to missing format.

This patch is RFC because while it's unlikely that format availability have skipped to earlier tegra210 soc generation I haven't confirmed any runtime error nor experienced any regression by the lack of availability of theses format.

---
 drivers/gpu/drm/tegra/dc.c | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)
drivers/gpu/drm/tegra/dc.c +4 -5
index 06370b7e0e56..d5896f12f25f 100644
@@ -2942,10 +2942,10 @@ static const struct tegra_dc_soc_info
 tegra210_dc_soc_info = { .has_powergate = true,
 	.coupled_pm = false,
 	.has_nvdisplay = false,
-	.num_primary_formats = ARRAY_SIZE(tegra114_primary_formats),
-	.primary_formats = tegra114_primary_formats,
-	.num_overlay_formats = ARRAY_SIZE(tegra114_overlay_formats),
-	.overlay_formats = tegra114_overlay_formats,
+	.num_primary_formats = ARRAY_SIZE(tegra124_primary_formats),
+	.primary_formats = tegra124_primary_formats,
+	.num_overlay_formats = ARRAY_SIZE(tegra124_overlay_formats),
+	.overlay_formats = tegra124_overlay_formats,
 	.modifiers = tegra124_modifiers,
 	.has_win_a_without_filters = false,
 	.has_win_b_vfilter_mem_client = false,
--
Quoted 1 line

2.53.0

Looking at the TRMs, I think indeed Tegra210 also supports the DRM_FORMAT_RGBX8888/DRM_FORMAT_BGRX8888 formats, so this patch should be correct. Functionality-wise, the formats aren't handled currently in plane.c, so I don't think there's any change in functionality.

Quoted 3 lines

These formats require enabling byteswap -- so do RGBA8888/BGRA8888, but plane.c doesn't currently do that, so I wonder if these formats are just not working right now.

It's been a really long time that I've looked at this, but if I recall correctly at least on pre-Tegra186 we properly program the byteswap registers (see DC_WIN_BYTE_SWAP in dc.c's tegra_dc_setup_window()). It is possible that I never implemented this for Tegra186 and later, where all the plane programming happens in hub.c with the shared planes.

I suspect that for Tegra210 we could add a bunch more formats, and I have a vague recollection that I had prototypes a few more a long time ago (notable some more YUV variants) but may never have submitted those patches because I couldn't find a good way to test them.

Thierry

maintainer
#5 lore raw
Quoted 58 lines · Thierry Reding
Quoted 47 lines · Mikko Perttunen
Quoted 39 lines · Nicolas Chauvet

The primary_format and overlay_format were picked from earlier tegra114 generation instead of using the previous tegra124 format leading to missing format.

This patch is RFC because while it's unlikely that format availability have skipped to earlier tegra210 soc generation I haven't confirmed any runtime error nor experienced any regression by the lack of availability of theses format.

---

 drivers/gpu/drm/tegra/dc.c | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)
drivers/gpu/drm/tegra/dc.c +4 -4
index 06370b7e0e56..d5896f12f25f 100644
@@ -2942,10 +2942,10 @@ static const struct tegra_dc_soc_info
 tegra210_dc_soc_info = { .has_powergate = true,
 
 	.coupled_pm = false,
 	.has_nvdisplay = false,
 
-	.num_primary_formats = ARRAY_SIZE(tegra114_primary_formats),
-	.primary_formats = tegra114_primary_formats,
-	.num_overlay_formats = ARRAY_SIZE(tegra114_overlay_formats),
-	.overlay_formats = tegra114_overlay_formats,
+	.num_primary_formats = ARRAY_SIZE(tegra124_primary_formats),
+	.primary_formats = tegra124_primary_formats,
+	.num_overlay_formats = ARRAY_SIZE(tegra124_overlay_formats),
+	.overlay_formats = tegra124_overlay_formats,
 
drivers/gpu/drm/tegra/dc.c +0 -1
 	.modifiers = tegra124_modifiers,
 	.has_win_a_without_filters = false,
 	.has_win_b_vfilter_mem_client = false,
 
--
Quoted 1 line

2.53.0

Looking at the TRMs, I think indeed Tegra210 also supports the DRM_FORMAT_RGBX8888/DRM_FORMAT_BGRX8888 formats, so this patch should be correct. Functionality-wise, the formats aren't handled currently in plane.c, so I don't think there's any change in functionality.

Quoted 3 lines

These formats require enabling byteswap -- so do RGBA8888/BGRA8888, but plane.c doesn't currently do that, so I wonder if these formats are just not working right now.

It's been a really long time that I've looked at this, but if I recall correctly at least on pre-Tegra186 we properly program the byteswap registers (see DC_WIN_BYTE_SWAP in dc.c's tegra_dc_setup_window()). It is possible that I never implemented this for Tegra186 and later, where all the plane programming happens in hub.c with the shared planes.

Quoted 6 lines

I suspect that for Tegra210 we could add a bunch more formats, and I have a vague recollection that I had prototypes a few more a long time ago (notable some more YUV variants) but may never have submitted those patches because I couldn't find a good way to test them.

Thierry

TBH I'm probably getting all the endiannesses of these format names mixed up in my head. FWIW, for pre-Tegra186, we are programming the swap register but plane.c always sets it to NOSWAP for the RGB-y formats.

Mikko