* [PATCH v17 13/28] drm/rockchip: vop2: Recognise 10-bit YUV422 as YUV format
From: Nicolas Frattaroli @ 2026-06-09 12:44 UTC (permalink / raw)
To: Harry Wentland, Leo Li, Rodrigo Siqueira, Alex Deucher,
Christian König, David Airlie, Simona Vetter,
Maarten Lankhorst, Maxime Ripard, Thomas Zimmermann,
Andrzej Hajda, Neil Armstrong, Robert Foss, Laurent Pinchart,
Jonas Karlman, Jernej Skrabec, Sandy Huang, Heiko Stübner,
Andy Yan, Jani Nikula, Rodrigo Vivi, Joonas Lahtinen,
Tvrtko Ursulin, Dmitry Baryshkov, Sascha Hauer, Rob Herring,
Jonathan Corbet, Shuah Khan, Daniel Stone
Cc: kernel, amd-gfx, dri-devel, linux-kernel, linux-arm-kernel,
linux-rockchip, intel-gfx, intel-xe, linux-doc, wayland-devel,
Nicolas Frattaroli, Cristian Ciocaltea
In-Reply-To: <20260609-color-format-v17-0-35739b5782cc@collabora.com>
The Rockchip VOP2 video output driver has a "is_yuv_output" function,
which returns true when a given bus format is a YUV format, and false
otherwise.
This switch statement is lacking the bus format used for YUV422 10-bit.
Add the two component orderings of the YUV422 10-bit bus formats to the
switch statement.
Fixes: 604be85547ce ("drm/rockchip: Add VOP2 driver")
Reviewed-by: Cristian Ciocaltea <cristian.ciocaltea@collabora.com>
Reviewed-by: Daniel Stone <daniel@fooishbar.org>
Signed-off-by: Nicolas Frattaroli <nicolas.frattaroli@collabora.com>
---
drivers/gpu/drm/rockchip/rockchip_drm_vop2.c | 2 ++
1 file changed, 2 insertions(+)
diff --git a/drivers/gpu/drm/rockchip/rockchip_drm_vop2.c b/drivers/gpu/drm/rockchip/rockchip_drm_vop2.c
index 6230735bbabc..3fad79fe2d89 100644
--- a/drivers/gpu/drm/rockchip/rockchip_drm_vop2.c
+++ b/drivers/gpu/drm/rockchip/rockchip_drm_vop2.c
@@ -352,6 +352,8 @@ static bool is_yuv_output(u32 bus_format)
switch (bus_format) {
case MEDIA_BUS_FMT_YUV8_1X24:
case MEDIA_BUS_FMT_YUV10_1X30:
+ case MEDIA_BUS_FMT_YUYV10_1X20:
+ case MEDIA_BUS_FMT_UYVY10_1X20:
case MEDIA_BUS_FMT_UYYVYY8_0_5X24:
case MEDIA_BUS_FMT_UYYVYY10_0_5X30:
case MEDIA_BUS_FMT_YUYV8_2X8:
--
2.54.0
^ permalink raw reply related
* [PATCH v17 12/28] drm/rockchip: vop2: Add RK3576 to the RG swap special case
From: Nicolas Frattaroli @ 2026-06-09 12:43 UTC (permalink / raw)
To: Harry Wentland, Leo Li, Rodrigo Siqueira, Alex Deucher,
Christian König, David Airlie, Simona Vetter,
Maarten Lankhorst, Maxime Ripard, Thomas Zimmermann,
Andrzej Hajda, Neil Armstrong, Robert Foss, Laurent Pinchart,
Jonas Karlman, Jernej Skrabec, Sandy Huang, Heiko Stübner,
Andy Yan, Jani Nikula, Rodrigo Vivi, Joonas Lahtinen,
Tvrtko Ursulin, Dmitry Baryshkov, Sascha Hauer, Rob Herring,
Jonathan Corbet, Shuah Khan, Daniel Stone
Cc: kernel, amd-gfx, dri-devel, linux-kernel, linux-arm-kernel,
linux-rockchip, intel-gfx, intel-xe, linux-doc, wayland-devel,
Nicolas Frattaroli, Andy Yan
In-Reply-To: <20260609-color-format-v17-0-35739b5782cc@collabora.com>
Much like RK3588, RK3576 requires an RG swap to be performed for YUV444
8-bit and YUV444 10-bit bus formats.
Add its version to the already existing check for RK3588, so that YUV444
output is correct on this platform.
Fixes: 944757a4cba6 ("drm/rockchip: vop2: Add support for rk3576")
Reviewed-by: Andy Yan <andyshrk@163.com>
Reviewed-by: Daniel Stone <daniel@fooishbar.org>
Signed-off-by: Nicolas Frattaroli <nicolas.frattaroli@collabora.com>
---
drivers/gpu/drm/rockchip/rockchip_drm_vop2.c | 3 ++-
1 file changed, 2 insertions(+), 1 deletion(-)
diff --git a/drivers/gpu/drm/rockchip/rockchip_drm_vop2.c b/drivers/gpu/drm/rockchip/rockchip_drm_vop2.c
index 9b261f0c7eec..6230735bbabc 100644
--- a/drivers/gpu/drm/rockchip/rockchip_drm_vop2.c
+++ b/drivers/gpu/drm/rockchip/rockchip_drm_vop2.c
@@ -337,7 +337,8 @@ static bool vop2_output_uv_swap(u32 bus_format, u32 output_mode)
static bool vop2_output_rg_swap(struct vop2 *vop2, u32 bus_format)
{
- if (vop2->version == VOP_VERSION_RK3588) {
+ if (vop2->version == VOP_VERSION_RK3588 ||
+ vop2->version == VOP_VERSION_RK3576) {
if (bus_format == MEDIA_BUS_FMT_YUV8_1X24 ||
bus_format == MEDIA_BUS_FMT_YUV10_1X30)
return true;
--
2.54.0
^ permalink raw reply related
* [PATCH v17 11/28] drm/rockchip: Add YUV422 output mode constants for VOP2
From: Nicolas Frattaroli @ 2026-06-09 12:43 UTC (permalink / raw)
To: Harry Wentland, Leo Li, Rodrigo Siqueira, Alex Deucher,
Christian König, David Airlie, Simona Vetter,
Maarten Lankhorst, Maxime Ripard, Thomas Zimmermann,
Andrzej Hajda, Neil Armstrong, Robert Foss, Laurent Pinchart,
Jonas Karlman, Jernej Skrabec, Sandy Huang, Heiko Stübner,
Andy Yan, Jani Nikula, Rodrigo Vivi, Joonas Lahtinen,
Tvrtko Ursulin, Dmitry Baryshkov, Sascha Hauer, Rob Herring,
Jonathan Corbet, Shuah Khan, Daniel Stone
Cc: kernel, amd-gfx, dri-devel, linux-kernel, linux-arm-kernel,
linux-rockchip, intel-gfx, intel-xe, linux-doc, wayland-devel,
Nicolas Frattaroli, Andy Yan
In-Reply-To: <20260609-color-format-v17-0-35739b5782cc@collabora.com>
The Rockchip display controller has a general YUV422 output mode, and
some SoC-specific connector-specific output modes for RK3576.
Add them, based on the values in downstream and the TRM (dsp_out_mode in
RK3576 TRM Part 2, register POST*_CTRL_POST_DSP_CTRL).
Reviewed-by: Andy Yan <andyshrk@163.com>
Reviewed-by: Daniel Stone <daniel@fooishbar.org>
Signed-off-by: Nicolas Frattaroli <nicolas.frattaroli@collabora.com>
---
drivers/gpu/drm/rockchip/rockchip_drm_drv.h | 4 ++++
1 file changed, 4 insertions(+)
diff --git a/drivers/gpu/drm/rockchip/rockchip_drm_drv.h b/drivers/gpu/drm/rockchip/rockchip_drm_drv.h
index 2e86ad00979c..4705dc6b8bd7 100644
--- a/drivers/gpu/drm/rockchip/rockchip_drm_drv.h
+++ b/drivers/gpu/drm/rockchip/rockchip_drm_drv.h
@@ -30,10 +30,14 @@
#define ROCKCHIP_OUT_MODE_P565 2
#define ROCKCHIP_OUT_MODE_BT656 5
#define ROCKCHIP_OUT_MODE_S888 8
+#define ROCKCHIP_OUT_MODE_YUV422 9
#define ROCKCHIP_OUT_MODE_S888_DUMMY 12
#define ROCKCHIP_OUT_MODE_YUV420 14
/* for use special outface */
#define ROCKCHIP_OUT_MODE_AAAA 15
+/* SoC specific output modes */
+#define ROCKCHIP_OUT_MODE_YUV422_RK3576_DP 12
+#define ROCKCHIP_OUT_MODE_YUV422_RK3576_HDMI 13
/* output flags */
#define ROCKCHIP_OUTPUT_DSI_DUAL BIT(0)
--
2.54.0
^ permalink raw reply related
* [PATCH v17 10/28] drm/amdgpu: Implement "color format" DRM property
From: Nicolas Frattaroli @ 2026-06-09 12:43 UTC (permalink / raw)
To: Harry Wentland, Leo Li, Rodrigo Siqueira, Alex Deucher,
Christian König, David Airlie, Simona Vetter,
Maarten Lankhorst, Maxime Ripard, Thomas Zimmermann,
Andrzej Hajda, Neil Armstrong, Robert Foss, Laurent Pinchart,
Jonas Karlman, Jernej Skrabec, Sandy Huang, Heiko Stübner,
Andy Yan, Jani Nikula, Rodrigo Vivi, Joonas Lahtinen,
Tvrtko Ursulin, Dmitry Baryshkov, Sascha Hauer, Rob Herring,
Jonathan Corbet, Shuah Khan, Daniel Stone
Cc: kernel, amd-gfx, dri-devel, linux-kernel, linux-arm-kernel,
linux-rockchip, intel-gfx, intel-xe, linux-doc, wayland-devel,
Nicolas Frattaroli, Werner Sembach, Andri Yngvason, Marius Vlad
In-Reply-To: <20260609-color-format-v17-0-35739b5782cc@collabora.com>
The "color format" DRM property allows userspace to explicitly pick a
color format to use. If an unsupported color format is requested,
userspace will be given an error instead of silently having its request
disobeyed.
The default case, which is AUTO, picks YCbCr 4:2:0 if it's a 4:2:0-only
mode, and RGB in all other cases.
Co-developed-by: Werner Sembach <wse@tuxedocomputers.com>
Signed-off-by: Werner Sembach <wse@tuxedocomputers.com>
Co-developed-by: Andri Yngvason <andri@yngvason.is>
Signed-off-by: Andri Yngvason <andri@yngvason.is>
Co-developed-by: Marius Vlad <marius.vlad@collabora.com>
Signed-off-by: Marius Vlad <marius.vlad@collabora.com>
Reviewed-by: Daniel Stone <daniel@fooishbar.org>
Signed-off-by: Nicolas Frattaroli <nicolas.frattaroli@collabora.com>
---
drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c | 90 ++++++++++++++++++++---
1 file changed, 79 insertions(+), 11 deletions(-)
diff --git a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c
index dfe97897127c..ece581609cf7 100644
--- a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c
+++ b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c
@@ -6899,11 +6899,14 @@ static void fill_stream_properties_from_drm_display_mode(
const struct dc_stream_state *old_stream,
int requested_bpc)
{
+ bool is_dp_or_hdmi = dc_is_hdmi_signal(stream->signal) || dc_is_dp_signal(stream->signal);
struct dc_crtc_timing *timing_out = &stream->timing;
const struct drm_display_info *info = &connector->display_info;
struct amdgpu_dm_connector *aconnector = NULL;
struct hdmi_vendor_infoframe hv_frame;
struct hdmi_avi_infoframe avi_frame;
+ bool want_420;
+ bool want_422;
ssize_t err;
if (connector->connector_type != DRM_MODE_CONNECTOR_WRITEBACK)
@@ -6916,20 +6919,41 @@ static void fill_stream_properties_from_drm_display_mode(
timing_out->h_border_right = 0;
timing_out->v_border_top = 0;
timing_out->v_border_bottom = 0;
- /* TODO: un-hardcode */
- if (drm_mode_is_420_only(info, mode_in) ||
- (aconnector && aconnector->force_yuv420_output &&
- drm_mode_is_420_also(info, mode_in)))
+
+ want_420 = (aconnector && aconnector->force_yuv420_output) ||
+ (connector_state->color_format == DRM_CONNECTOR_COLOR_FORMAT_YCBCR420);
+ want_422 = (aconnector && aconnector->force_yuv422_output) ||
+ (connector_state->color_format == DRM_CONNECTOR_COLOR_FORMAT_YCBCR422);
+
+ if (drm_mode_is_420_only(info, mode_in) &&
+ (want_420 || connector_state->color_format == DRM_CONNECTOR_COLOR_FORMAT_AUTO)) {
timing_out->pixel_encoding = PIXEL_ENCODING_YCBCR420;
- else if ((connector->display_info.color_formats & BIT(DRM_OUTPUT_COLOR_FORMAT_YCBCR422))
- && aconnector
- && aconnector->force_yuv422_output)
+ } else if (drm_mode_is_420_also(info, mode_in) && want_420) {
+ timing_out->pixel_encoding = PIXEL_ENCODING_YCBCR420;
+ } else if ((info->color_formats & BIT(DRM_OUTPUT_COLOR_FORMAT_YCBCR422)) &&
+ want_422 && is_dp_or_hdmi) {
timing_out->pixel_encoding = PIXEL_ENCODING_YCBCR422;
- else if ((connector->display_info.color_formats & BIT(DRM_OUTPUT_COLOR_FORMAT_YCBCR444))
- && stream->signal == SIGNAL_TYPE_HDMI_TYPE_A)
+ } else if (connector_state->color_format == DRM_CONNECTOR_COLOR_FORMAT_YCBCR444 &&
+ (info->color_formats & BIT(DRM_OUTPUT_COLOR_FORMAT_YCBCR444)) &&
+ is_dp_or_hdmi) {
timing_out->pixel_encoding = PIXEL_ENCODING_YCBCR444;
- else
+ } else if (connector_state->color_format == DRM_CONNECTOR_COLOR_FORMAT_RGB444 ||
+ connector_state->color_format == DRM_CONNECTOR_COLOR_FORMAT_AUTO) {
timing_out->pixel_encoding = PIXEL_ENCODING_RGB;
+ } else {
+ /*
+ * If a format was explicitly requested but the requested format
+ * can't be satisfied, set it to an invalid value so that an
+ * error bubbles up to userspace. This way, userspace knows it
+ * needs to make a better choice.
+ */
+ if (connector_state->color_format != DRM_CONNECTOR_COLOR_FORMAT_AUTO)
+ timing_out->pixel_encoding = PIXEL_ENCODING_UNDEFINED;
+ else if (drm_mode_is_420_only(info, mode_in))
+ timing_out->pixel_encoding = PIXEL_ENCODING_YCBCR420;
+ else
+ timing_out->pixel_encoding = PIXEL_ENCODING_RGB;
+ }
timing_out->timing_3d_format = TIMING_3D_FORMAT_NONE;
timing_out->display_color_depth = convert_color_depth_from_display_info(
@@ -8279,6 +8303,38 @@ static enum dc_status dm_validate_stream_and_context(struct dc *dc,
return dc_result;
}
+static enum dc_status
+dm_validate_stream_color_format(const struct drm_connector_state *drm_state,
+ const struct dc_stream_state *stream)
+{
+ enum dc_pixel_encoding encoding;
+
+ if (!drm_state->color_format)
+ return DC_OK;
+
+ switch (drm_state->color_format) {
+ case DRM_CONNECTOR_COLOR_FORMAT_AUTO:
+ case DRM_CONNECTOR_COLOR_FORMAT_RGB444:
+ encoding = PIXEL_ENCODING_RGB;
+ break;
+ case DRM_CONNECTOR_COLOR_FORMAT_YCBCR444:
+ encoding = PIXEL_ENCODING_YCBCR444;
+ break;
+ case DRM_CONNECTOR_COLOR_FORMAT_YCBCR422:
+ encoding = PIXEL_ENCODING_YCBCR422;
+ break;
+ case DRM_CONNECTOR_COLOR_FORMAT_YCBCR420:
+ encoding = PIXEL_ENCODING_YCBCR420;
+ break;
+ default:
+ encoding = PIXEL_ENCODING_UNDEFINED;
+ break;
+ }
+
+ return encoding == stream->timing.pixel_encoding ?
+ DC_OK : DC_UNSUPPORTED_VALUE;
+}
+
struct dc_stream_state *
create_validate_stream_for_sink(struct drm_connector *connector,
const struct drm_display_mode *drm_mode,
@@ -8325,6 +8381,9 @@ create_validate_stream_for_sink(struct drm_connector *connector,
if (dc_result == DC_OK)
dc_result = dm_validate_stream_and_context(adev->dm.dc, stream);
+ if (dc_result == DC_OK)
+ dc_result = dm_validate_stream_color_format(drm_state, stream);
+
if (dc_result != DC_OK) {
drm_dbg_kms(connector->dev, "Pruned mode %d x %d (clk %d) %s %s -- %s\n",
drm_mode->hdisplay,
@@ -9154,6 +9213,12 @@ static const u32 supported_colorspaces =
BIT(DRM_MODE_COLORIMETRY_BT2020_RGB) |
BIT(DRM_MODE_COLORIMETRY_BT2020_YCC);
+static const u32 supported_colorformats =
+ BIT(DRM_OUTPUT_COLOR_FORMAT_RGB444) |
+ BIT(DRM_OUTPUT_COLOR_FORMAT_YCBCR444) |
+ BIT(DRM_OUTPUT_COLOR_FORMAT_YCBCR422) |
+ BIT(DRM_OUTPUT_COLOR_FORMAT_YCBCR420);
+
void amdgpu_dm_connector_init_helper(struct amdgpu_display_manager *dm,
struct amdgpu_dm_connector *aconnector,
int connector_type,
@@ -9270,8 +9335,11 @@ void amdgpu_dm_connector_init_helper(struct amdgpu_display_manager *dm,
connector_type == DRM_MODE_CONNECTOR_eDP) {
drm_connector_attach_hdr_output_metadata_property(&aconnector->base);
- if (!aconnector->mst_root)
+ if (!aconnector->mst_root) {
drm_connector_attach_vrr_capable_property(&aconnector->base);
+ drm_connector_attach_color_format_property(&aconnector->base,
+ supported_colorformats);
+ }
if (adev->dm.hdcp_workqueue)
drm_connector_attach_content_protection_property(&aconnector->base, true);
--
2.54.0
^ permalink raw reply related
* [PATCH v17 09/28] drm/display: hdmi-state-helper: Try subsampling in mode_valid
From: Nicolas Frattaroli @ 2026-06-09 12:43 UTC (permalink / raw)
To: Harry Wentland, Leo Li, Rodrigo Siqueira, Alex Deucher,
Christian König, David Airlie, Simona Vetter,
Maarten Lankhorst, Maxime Ripard, Thomas Zimmermann,
Andrzej Hajda, Neil Armstrong, Robert Foss, Laurent Pinchart,
Jonas Karlman, Jernej Skrabec, Sandy Huang, Heiko Stübner,
Andy Yan, Jani Nikula, Rodrigo Vivi, Joonas Lahtinen,
Tvrtko Ursulin, Dmitry Baryshkov, Sascha Hauer, Rob Herring,
Jonathan Corbet, Shuah Khan, Daniel Stone
Cc: kernel, amd-gfx, dri-devel, linux-kernel, linux-arm-kernel,
linux-rockchip, intel-gfx, intel-xe, linux-doc, wayland-devel,
Nicolas Frattaroli, Dmitry Baryshkov
In-Reply-To: <20260609-color-format-v17-0-35739b5782cc@collabora.com>
drm_hdmi_connector_mode_valid assumes modes are only valid if they work
with RGB. The reality is more complex however: YCbCr 4:2:0
chroma-subsampled modes only require half the pixel clock that the same
mode would require in RGB.
This leads to drm_hdmi_connector_mode_valid rejecting perfectly valid
420-only or 420-also modes.
Fix this by checking whether the mode is 420-capable first. If so, then
proceed by checking it with DRM_OUTPUT_COLOR_FORMAT_YCBCR420 so long as
the connector has legalized 420, otherwise error out. If the mode is not
420-capable, check with RGB as was previously always the case.
Fixes: 47368ab437fd ("drm/display: hdmi: add generic mode_valid helper")
Reviewed-by: Maxime Ripard <mripard@kernel.org>
Reviewed-by: Dmitry Baryshkov <dmitry.baryshkov@oss.qualcomm.com>
Reviewed-by: Daniel Stone <daniel@fooishbar.org>
Signed-off-by: Nicolas Frattaroli <nicolas.frattaroli@collabora.com>
---
drivers/gpu/drm/display/drm_hdmi_state_helper.c | 15 ++++++++++++++-
1 file changed, 14 insertions(+), 1 deletion(-)
diff --git a/drivers/gpu/drm/display/drm_hdmi_state_helper.c b/drivers/gpu/drm/display/drm_hdmi_state_helper.c
index 069f9bf2489e..ce17eeefc2da 100644
--- a/drivers/gpu/drm/display/drm_hdmi_state_helper.c
+++ b/drivers/gpu/drm/display/drm_hdmi_state_helper.c
@@ -943,8 +943,21 @@ drm_hdmi_connector_mode_valid(struct drm_connector *connector,
const struct drm_display_mode *mode)
{
unsigned long long clock;
+ enum drm_output_color_format fmt;
+
+ if (drm_mode_is_420_only(&connector->display_info, mode)) {
+ if (connector->ycbcr_420_allowed)
+ fmt = DRM_OUTPUT_COLOR_FORMAT_YCBCR420;
+ else
+ return MODE_NO_420;
+ } else if (drm_mode_is_420_also(&connector->display_info, mode) &&
+ connector->ycbcr_420_allowed) {
+ fmt = DRM_OUTPUT_COLOR_FORMAT_YCBCR420;
+ } else {
+ fmt = DRM_OUTPUT_COLOR_FORMAT_RGB444;
+ }
- clock = drm_hdmi_compute_mode_clock(mode, 8, DRM_OUTPUT_COLOR_FORMAT_RGB444);
+ clock = drm_hdmi_compute_mode_clock(mode, 8, fmt);
if (!clock)
return MODE_ERROR;
--
2.54.0
^ permalink raw reply related
* [PATCH v17 08/28] drm/display: hdmi-state-helper: Act on color format DRM property
From: Nicolas Frattaroli @ 2026-06-09 12:43 UTC (permalink / raw)
To: Harry Wentland, Leo Li, Rodrigo Siqueira, Alex Deucher,
Christian König, David Airlie, Simona Vetter,
Maarten Lankhorst, Maxime Ripard, Thomas Zimmermann,
Andrzej Hajda, Neil Armstrong, Robert Foss, Laurent Pinchart,
Jonas Karlman, Jernej Skrabec, Sandy Huang, Heiko Stübner,
Andy Yan, Jani Nikula, Rodrigo Vivi, Joonas Lahtinen,
Tvrtko Ursulin, Dmitry Baryshkov, Sascha Hauer, Rob Herring,
Jonathan Corbet, Shuah Khan, Daniel Stone
Cc: kernel, amd-gfx, dri-devel, linux-kernel, linux-arm-kernel,
linux-rockchip, intel-gfx, intel-xe, linux-doc, wayland-devel,
Nicolas Frattaroli, Dmitry Baryshkov
In-Reply-To: <20260609-color-format-v17-0-35739b5782cc@collabora.com>
With the introduction of the "color format" DRM property, which allows
userspace to request a specific color format, the HDMI state helper
should implement this.
Implement it by translating the requested drm_connector_color_format to
a drm_output_color_format enum value as per the logic HDMI should use
for this: Auto is translated to RGB, and a fallback to YUV420 is only
performed if the original color format was auto.
Reviewed-by: Dmitry Baryshkov <dmitry.baryshkov@oss.qualcomm.com>
Reviewed-by: Maxime Ripard <mripard@kernel.org>
Reviewed-by: Daniel Stone <daniel@fooishbar.org>
Signed-off-by: Nicolas Frattaroli <nicolas.frattaroli@collabora.com>
---
drivers/gpu/drm/display/drm_hdmi_state_helper.c | 31 +++++++++++++++++++++++++
1 file changed, 31 insertions(+)
diff --git a/drivers/gpu/drm/display/drm_hdmi_state_helper.c b/drivers/gpu/drm/display/drm_hdmi_state_helper.c
index c68b159f0a7f..069f9bf2489e 100644
--- a/drivers/gpu/drm/display/drm_hdmi_state_helper.c
+++ b/drivers/gpu/drm/display/drm_hdmi_state_helper.c
@@ -671,8 +671,39 @@ hdmi_compute_config(const struct drm_connector *connector,
unsigned int max_bpc = clamp_t(unsigned int,
conn_state->max_bpc,
8, connector->max_bpc);
+ enum drm_output_color_format fmt;
int ret;
+ if (conn_state->color_format != DRM_CONNECTOR_COLOR_FORMAT_AUTO) {
+ switch (conn_state->color_format) {
+ case DRM_CONNECTOR_COLOR_FORMAT_AUTO:
+ drm_warn(connector->dev, "AUTO format in non-AUTO path.\n");
+ fallthrough;
+ case DRM_CONNECTOR_COLOR_FORMAT_RGB444:
+ fmt = DRM_OUTPUT_COLOR_FORMAT_RGB444;
+ break;
+ case DRM_CONNECTOR_COLOR_FORMAT_YCBCR444:
+ fmt = DRM_OUTPUT_COLOR_FORMAT_YCBCR444;
+ break;
+ case DRM_CONNECTOR_COLOR_FORMAT_YCBCR422:
+ fmt = DRM_OUTPUT_COLOR_FORMAT_YCBCR422;
+ break;
+ case DRM_CONNECTOR_COLOR_FORMAT_YCBCR420:
+ fmt = DRM_OUTPUT_COLOR_FORMAT_YCBCR420;
+ break;
+ default:
+ drm_dbg_kms(connector->dev, "HDMI does not support color format '%d'.\n",
+ conn_state->color_format);
+ return -EINVAL;
+ }
+
+ return hdmi_compute_format_bpc(connector, conn_state, mode, max_bpc, fmt);
+ }
+
+ /*
+ * For %DRM_CONNECTOR_COLOR_FORMAT_AUTO, try RGB first, and fall back
+ * to the less bandwidth-intensive YCBCR420 if RGB fails.
+ */
ret = hdmi_compute_format_bpc(connector, conn_state, mode, max_bpc,
DRM_OUTPUT_COLOR_FORMAT_RGB444);
if (ret) {
--
2.54.0
^ permalink raw reply related
* [PATCH v17 07/28] drm/atomic-helper: Add HDMI bridge output bus formats helper
From: Nicolas Frattaroli @ 2026-06-09 12:43 UTC (permalink / raw)
To: Harry Wentland, Leo Li, Rodrigo Siqueira, Alex Deucher,
Christian König, David Airlie, Simona Vetter,
Maarten Lankhorst, Maxime Ripard, Thomas Zimmermann,
Andrzej Hajda, Neil Armstrong, Robert Foss, Laurent Pinchart,
Jonas Karlman, Jernej Skrabec, Sandy Huang, Heiko Stübner,
Andy Yan, Jani Nikula, Rodrigo Vivi, Joonas Lahtinen,
Tvrtko Ursulin, Dmitry Baryshkov, Sascha Hauer, Rob Herring,
Jonathan Corbet, Shuah Khan, Daniel Stone
Cc: kernel, amd-gfx, dri-devel, linux-kernel, linux-arm-kernel,
linux-rockchip, intel-gfx, intel-xe, linux-doc, wayland-devel,
Nicolas Frattaroli, Dmitry Baryshkov
In-Reply-To: <20260609-color-format-v17-0-35739b5782cc@collabora.com>
The drm_bridge_funcs atomic_get_output_bus_fmts operation should be the
same for likely every HDMI connector bridge, unless such an HDMI
connector bridge has some special hardware restrictions that I cannot
envision yet.
To avoid code duplication and standardize on a set of media bus formats
that the HDMI output color formats translate to, add a common helper
function that implements this operation to the drm bridge helpers.
The function returns a list of output bus formats based on the HDMI
bridge's current output bits-per-component, and its bitmask of supported
color formats.
To guard against future expansion of DRM_OUTPUT_COLOR_FORMAT outgrowing
the hweight8 call, add a BUILD_BUG_ON statement where it's used that
checks for DRM_OUTPUT_COLOR_FORMAT_COUNT. The justification for not
using hweight32 in all cases is that not all ISAs have a popcount
instruction, and will benefit from a smaller/faster software
implementation that doesn't have to operate across all bits.
The justification for not defining an hweight_color depending on the
value of DRM_OUTPUT_COLOR_FORMAT_COUNT is that this count enum value is
only known at compile time, not at preprocessor time.
Reviewed-by: Dmitry Baryshkov <dmitry.baryshkov@oss.qualcomm.com>
Reviewed-by: Daniel Stone <daniel@fooishbar.org>
Signed-off-by: Nicolas Frattaroli <nicolas.frattaroli@collabora.com>
---
drivers/gpu/drm/drm_atomic_helper.c | 81 +++++++++++++++++++++++++++++++++++++
include/drm/drm_atomic_helper.h | 7 ++++
2 files changed, 88 insertions(+)
diff --git a/drivers/gpu/drm/drm_atomic_helper.c b/drivers/gpu/drm/drm_atomic_helper.c
index 3547f797a850..285aac3554df 100644
--- a/drivers/gpu/drm/drm_atomic_helper.c
+++ b/drivers/gpu/drm/drm_atomic_helper.c
@@ -28,6 +28,7 @@
#include <linux/export.h>
#include <linux/dma-fence.h>
#include <linux/ktime.h>
+#include <linux/media-bus-format.h>
#include <drm/drm_atomic.h>
#include <drm/drm_atomic_helper.h>
@@ -4107,3 +4108,83 @@ drm_atomic_helper_bridge_propagate_bus_fmt(struct drm_bridge *bridge,
return input_fmts;
}
EXPORT_SYMBOL(drm_atomic_helper_bridge_propagate_bus_fmt);
+
+/**
+ * drm_atomic_helper_bridge_get_hdmi_output_bus_fmts - helper implementing
+ * atomic_get_output_bus_fmts for HDMI
+ * @bridge: pointer to &struct drm_bridge
+ * @bridge_state: pointer to the current bridge state
+ * @crtc_state: pointer to the current CRTC state
+ * @conn_state: pointer to the current connector state
+ * @num_output_fmts: pointer to where the number of entries in the returned array
+ * will be stored. Set to 0 if unsuccessful.
+ *
+ * Common implementation for the &drm_bridge_funcs.atomic_get_output_bus_fmts
+ * operation that's applicable to HDMI connectors.
+ *
+ * Returns: a newly allocated array of u32 values of length \*@num_output_fmts,
+ * representing all the MEDIA_BUS_FMTS\_ for the current connector state's
+ * chosen HDMI output bits per compoennt, or %NULL if it fails to allocate one.
+ */
+u32 *
+drm_atomic_helper_bridge_get_hdmi_output_bus_fmts(struct drm_bridge *bridge,
+ struct drm_bridge_state *bridge_state,
+ struct drm_crtc_state *crtc_state,
+ struct drm_connector_state *conn_state,
+ unsigned int *num_output_fmts)
+{
+ unsigned int num_fmts = 0;
+ u32 *out_fmts;
+
+ /*
+ * bridge->supported_formats is a bit field of BIT(enum drm_output_color_format)
+ * values. The smallest hweight that is smaller than or equal to
+ * %DRM_OUTPUT_COLOR_FORMAT_COUNT will do for counting set bits here.
+ */
+ BUILD_BUG_ON(const_true(DRM_OUTPUT_COLOR_FORMAT_COUNT > 8));
+ out_fmts = kmalloc_array(hweight8(bridge->supported_formats),
+ sizeof(u32), GFP_KERNEL);
+ if (!out_fmts) {
+ *num_output_fmts = 0;
+ return NULL;
+ }
+
+ switch (conn_state->hdmi.output_bpc) {
+ case 12:
+ if (bridge->supported_formats & BIT(DRM_OUTPUT_COLOR_FORMAT_RGB444))
+ out_fmts[num_fmts++] = MEDIA_BUS_FMT_RGB121212_1X36;
+ if (bridge->supported_formats & BIT(DRM_OUTPUT_COLOR_FORMAT_YCBCR444))
+ out_fmts[num_fmts++] = MEDIA_BUS_FMT_YUV12_1X36;
+ if (bridge->supported_formats & BIT(DRM_OUTPUT_COLOR_FORMAT_YCBCR422))
+ out_fmts[num_fmts++] = MEDIA_BUS_FMT_UYVY12_1X24;
+ if (bridge->supported_formats & BIT(DRM_OUTPUT_COLOR_FORMAT_YCBCR420))
+ out_fmts[num_fmts++] = MEDIA_BUS_FMT_UYYVYY12_0_5X36;
+ break;
+ case 10:
+ if (bridge->supported_formats & BIT(DRM_OUTPUT_COLOR_FORMAT_RGB444))
+ out_fmts[num_fmts++] = MEDIA_BUS_FMT_RGB101010_1X30;
+ if (bridge->supported_formats & BIT(DRM_OUTPUT_COLOR_FORMAT_YCBCR444))
+ out_fmts[num_fmts++] = MEDIA_BUS_FMT_YUV10_1X30;
+ if (bridge->supported_formats & BIT(DRM_OUTPUT_COLOR_FORMAT_YCBCR422))
+ out_fmts[num_fmts++] = MEDIA_BUS_FMT_UYVY10_1X20;
+ if (bridge->supported_formats & BIT(DRM_OUTPUT_COLOR_FORMAT_YCBCR420))
+ out_fmts[num_fmts++] = MEDIA_BUS_FMT_UYYVYY10_0_5X30;
+ break;
+ default:
+ if (bridge->supported_formats & BIT(DRM_OUTPUT_COLOR_FORMAT_RGB444))
+ out_fmts[num_fmts++] = MEDIA_BUS_FMT_RGB888_1X24;
+ if (bridge->supported_formats & BIT(DRM_OUTPUT_COLOR_FORMAT_YCBCR444))
+ out_fmts[num_fmts++] = MEDIA_BUS_FMT_YUV8_1X24;
+ if (bridge->supported_formats & BIT(DRM_OUTPUT_COLOR_FORMAT_YCBCR422))
+ out_fmts[num_fmts++] = MEDIA_BUS_FMT_UYVY8_1X16;
+ if (bridge->supported_formats & BIT(DRM_OUTPUT_COLOR_FORMAT_YCBCR420))
+ out_fmts[num_fmts++] = MEDIA_BUS_FMT_UYYVYY8_0_5X24;
+ break;
+ }
+
+ *num_output_fmts = num_fmts;
+
+ return out_fmts;
+}
+EXPORT_SYMBOL(drm_atomic_helper_bridge_get_hdmi_output_bus_fmts);
+
diff --git a/include/drm/drm_atomic_helper.h b/include/drm/drm_atomic_helper.h
index b84152810abb..4cfeec70d648 100644
--- a/include/drm/drm_atomic_helper.h
+++ b/include/drm/drm_atomic_helper.h
@@ -295,4 +295,11 @@ drm_atomic_helper_bridge_propagate_bus_fmt(struct drm_bridge *bridge,
u32 output_fmt,
unsigned int *num_input_fmts);
+u32 *
+drm_atomic_helper_bridge_get_hdmi_output_bus_fmts(struct drm_bridge *bridge,
+ struct drm_bridge_state *bridge_state,
+ struct drm_crtc_state *crtc_state,
+ struct drm_connector_state *conn_state,
+ unsigned int *num_output_fmts);
+
#endif /* DRM_ATOMIC_HELPER_H_ */
--
2.54.0
^ permalink raw reply related
* [PATCH v17 06/28] drm/bridge: Act on the DRM color format property
From: Nicolas Frattaroli @ 2026-06-09 12:43 UTC (permalink / raw)
To: Harry Wentland, Leo Li, Rodrigo Siqueira, Alex Deucher,
Christian König, David Airlie, Simona Vetter,
Maarten Lankhorst, Maxime Ripard, Thomas Zimmermann,
Andrzej Hajda, Neil Armstrong, Robert Foss, Laurent Pinchart,
Jonas Karlman, Jernej Skrabec, Sandy Huang, Heiko Stübner,
Andy Yan, Jani Nikula, Rodrigo Vivi, Joonas Lahtinen,
Tvrtko Ursulin, Dmitry Baryshkov, Sascha Hauer, Rob Herring,
Jonathan Corbet, Shuah Khan, Daniel Stone
Cc: kernel, amd-gfx, dri-devel, linux-kernel, linux-arm-kernel,
linux-rockchip, intel-gfx, intel-xe, linux-doc, wayland-devel,
Nicolas Frattaroli, Dmitry Baryshkov
In-Reply-To: <20260609-color-format-v17-0-35739b5782cc@collabora.com>
The new DRM color format property allows userspace to request a specific
color format on a connector. In turn, this fills the connector state's
color_format member to switch color formats.
Make drm_bridges consider the color_format set in the connector state
during the atomic bridge check. Call into the connector function to get
the connector state's connector color format. For bridge connectors
including an HDMI bridge, this will make use of whatever the HDMI
implementation set as output formats, and AUTO will never be part of the
rejection logic.
Reject any output bus formats that do not correspond to the requested
color format. DRM_CONNECTOR_COLOR_FORMAT_AUTO is always accepted as a
matching color format for a bus format, meaning that non-HDMI bridge
chains will end up picking the first bus format choice that works, as
has already been the case previously.
Reviewed-by: Dmitry Baryshkov <dmitry.baryshkov@oss.qualcomm.com>
Reviewed-by: Daniel Stone <daniel@fooishbar.org>
Signed-off-by: Nicolas Frattaroli <nicolas.frattaroli@collabora.com>
---
drivers/gpu/drm/drm_bridge.c | 64 +++++++++++++++++++++++++++++++++++++++++++-
1 file changed, 63 insertions(+), 1 deletion(-)
diff --git a/drivers/gpu/drm/drm_bridge.c b/drivers/gpu/drm/drm_bridge.c
index 9a185032a3bd..9d6fcf986022 100644
--- a/drivers/gpu/drm/drm_bridge.c
+++ b/drivers/gpu/drm/drm_bridge.c
@@ -1151,6 +1151,47 @@ static int select_bus_fmt_recursive(struct drm_bridge *first_bridge,
return ret;
}
+static bool __pure bus_format_is_color_fmt(u32 bus_fmt, enum drm_connector_color_format fmt)
+{
+ if (fmt == DRM_CONNECTOR_COLOR_FORMAT_AUTO)
+ return true;
+
+ switch (bus_fmt) {
+ case MEDIA_BUS_FMT_FIXED:
+ return true;
+ case MEDIA_BUS_FMT_RGB888_1X24:
+ case MEDIA_BUS_FMT_RGB101010_1X30:
+ case MEDIA_BUS_FMT_RGB121212_1X36:
+ case MEDIA_BUS_FMT_RGB161616_1X48:
+ return fmt == DRM_CONNECTOR_COLOR_FORMAT_RGB444;
+ case MEDIA_BUS_FMT_YUV8_1X24:
+ case MEDIA_BUS_FMT_YUV10_1X30:
+ case MEDIA_BUS_FMT_YUV12_1X36:
+ case MEDIA_BUS_FMT_YUV16_1X48:
+ return fmt == DRM_CONNECTOR_COLOR_FORMAT_YCBCR444;
+ case MEDIA_BUS_FMT_UYVY8_1X16:
+ case MEDIA_BUS_FMT_VYUY8_1X16:
+ case MEDIA_BUS_FMT_YUYV8_1X16:
+ case MEDIA_BUS_FMT_YVYU8_1X16:
+ case MEDIA_BUS_FMT_UYVY10_1X20:
+ case MEDIA_BUS_FMT_YUYV10_1X20:
+ case MEDIA_BUS_FMT_VYUY10_1X20:
+ case MEDIA_BUS_FMT_YVYU10_1X20:
+ case MEDIA_BUS_FMT_UYVY12_1X24:
+ case MEDIA_BUS_FMT_VYUY12_1X24:
+ case MEDIA_BUS_FMT_YUYV12_1X24:
+ case MEDIA_BUS_FMT_YVYU12_1X24:
+ return fmt == DRM_CONNECTOR_COLOR_FORMAT_YCBCR422;
+ case MEDIA_BUS_FMT_UYYVYY8_0_5X24:
+ case MEDIA_BUS_FMT_UYYVYY10_0_5X30:
+ case MEDIA_BUS_FMT_UYYVYY12_0_5X36:
+ case MEDIA_BUS_FMT_UYYVYY16_0_5X48:
+ return fmt == DRM_CONNECTOR_COLOR_FORMAT_YCBCR420;
+ default:
+ return false;
+ }
+}
+
/*
* This function is called by &drm_atomic_bridge_chain_check() just before
* calling &drm_bridge_funcs.atomic_check() on all elements of the chain.
@@ -1194,6 +1235,7 @@ drm_atomic_bridge_chain_select_bus_fmts(struct drm_bridge *bridge,
struct drm_encoder *encoder = bridge->encoder;
struct drm_bridge_state *last_bridge_state;
unsigned int i, num_out_bus_fmts = 0;
+ enum drm_connector_color_format fmt;
u32 *out_bus_fmts;
int ret = 0;
@@ -1235,11 +1277,31 @@ drm_atomic_bridge_chain_select_bus_fmts(struct drm_bridge *bridge,
out_bus_fmts[0] = MEDIA_BUS_FMT_FIXED;
}
+ /*
+ * Instead of directly accessing conn_state.color_format, call into a
+ * connector function that allows connector implementations (e.g. for
+ * bridge connectors including HDMI bridges, where the HDMI helpers will
+ * have already chosen an appropriate output format) to override the
+ * selected format.
+ */
+ fmt = drm_connector_get_color_format(conn_state);
+
for (i = 0; i < num_out_bus_fmts; i++) {
+ if (!bus_format_is_color_fmt(out_bus_fmts[i], fmt)) {
+ drm_dbg_kms(last_bridge->dev,
+ "Skipping bus format 0x%04x as it doesn't match format %d\n",
+ out_bus_fmts[i], fmt);
+ ret = -ENOTSUPP;
+ continue;
+ }
ret = select_bus_fmt_recursive(bridge, last_bridge, crtc_state,
conn_state, out_bus_fmts[i]);
- if (ret != -ENOTSUPP)
+ if (ret != -ENOTSUPP) {
+ drm_dbg_kms(last_bridge->dev,
+ "Found bridge chain ending with bus format 0x%04x\n",
+ out_bus_fmts[i]);
break;
+ }
}
kfree(out_bus_fmts);
--
2.54.0
^ permalink raw reply related
* [PATCH v17 05/28] drm/display: bridge_connector: Use HDMI color format for HDMI conns
From: Nicolas Frattaroli @ 2026-06-09 12:43 UTC (permalink / raw)
To: Harry Wentland, Leo Li, Rodrigo Siqueira, Alex Deucher,
Christian König, David Airlie, Simona Vetter,
Maarten Lankhorst, Maxime Ripard, Thomas Zimmermann,
Andrzej Hajda, Neil Armstrong, Robert Foss, Laurent Pinchart,
Jonas Karlman, Jernej Skrabec, Sandy Huang, Heiko Stübner,
Andy Yan, Jani Nikula, Rodrigo Vivi, Joonas Lahtinen,
Tvrtko Ursulin, Dmitry Baryshkov, Sascha Hauer, Rob Herring,
Jonathan Corbet, Shuah Khan, Daniel Stone
Cc: kernel, amd-gfx, dri-devel, linux-kernel, linux-arm-kernel,
linux-rockchip, intel-gfx, intel-xe, linux-doc, wayland-devel,
Nicolas Frattaroli, Dmitry Baryshkov
In-Reply-To: <20260609-color-format-v17-0-35739b5782cc@collabora.com>
For bridge connectors which contain an HDMI bridge at some stage, the
HDMI state helpers' format selection logic should be involved.
Add an implementation for the drm_bridge_funcs color_format function,
which translates from the HDMI state's output format to a connector
format for bridge connectors involving an HDMI bridge, but return the
connector state's color_format member unchanged otherwise.
Reviewed-by: Dmitry Baryshkov <dmitry.baryshkov@oss.qualcomm.com>
Reviewed-by: Daniel Stone <daniel@fooishbar.org>
Signed-off-by: Nicolas Frattaroli <nicolas.frattaroli@collabora.com>
---
drivers/gpu/drm/display/drm_bridge_connector.c | 24 ++++++++++++++++++++++++
1 file changed, 24 insertions(+)
diff --git a/drivers/gpu/drm/display/drm_bridge_connector.c b/drivers/gpu/drm/display/drm_bridge_connector.c
index cafa498c3848..92f8a2d7aab4 100644
--- a/drivers/gpu/drm/display/drm_bridge_connector.c
+++ b/drivers/gpu/drm/display/drm_bridge_connector.c
@@ -283,6 +283,29 @@ drm_bridge_connector_create_state(struct drm_connector *connector)
return conn_state;
}
+static enum drm_connector_color_format
+drm_bridge_connector_color_format(const struct drm_connector_state *conn_state)
+{
+ struct drm_bridge_connector *bridge_connector =
+ to_drm_bridge_connector(conn_state->connector);
+
+ if (bridge_connector->bridge_hdmi) {
+ switch (conn_state->hdmi.output_format) {
+ default:
+ case DRM_OUTPUT_COLOR_FORMAT_RGB444:
+ return DRM_CONNECTOR_COLOR_FORMAT_RGB444;
+ case DRM_OUTPUT_COLOR_FORMAT_YCBCR444:
+ return DRM_CONNECTOR_COLOR_FORMAT_YCBCR444;
+ case DRM_OUTPUT_COLOR_FORMAT_YCBCR422:
+ return DRM_CONNECTOR_COLOR_FORMAT_YCBCR422;
+ case DRM_OUTPUT_COLOR_FORMAT_YCBCR420:
+ return DRM_CONNECTOR_COLOR_FORMAT_YCBCR420;
+ }
+ }
+
+ return conn_state->color_format;
+}
+
static const struct drm_connector_funcs drm_bridge_connector_funcs = {
.detect = drm_bridge_connector_detect,
.force = drm_bridge_connector_force,
@@ -292,6 +315,7 @@ static const struct drm_connector_funcs drm_bridge_connector_funcs = {
.atomic_destroy_state = drm_atomic_helper_connector_destroy_state,
.debugfs_init = drm_bridge_connector_debugfs_init,
.oob_hotplug_event = drm_bridge_connector_oob_hotplug_event,
+ .color_format = drm_bridge_connector_color_format,
};
/* -----------------------------------------------------------------------------
--
2.54.0
^ permalink raw reply related
* [PATCH v17 04/28] drm/connector: Let connectors have a say in their color format
From: Nicolas Frattaroli @ 2026-06-09 12:43 UTC (permalink / raw)
To: Harry Wentland, Leo Li, Rodrigo Siqueira, Alex Deucher,
Christian König, David Airlie, Simona Vetter,
Maarten Lankhorst, Maxime Ripard, Thomas Zimmermann,
Andrzej Hajda, Neil Armstrong, Robert Foss, Laurent Pinchart,
Jonas Karlman, Jernej Skrabec, Sandy Huang, Heiko Stübner,
Andy Yan, Jani Nikula, Rodrigo Vivi, Joonas Lahtinen,
Tvrtko Ursulin, Dmitry Baryshkov, Sascha Hauer, Rob Herring,
Jonathan Corbet, Shuah Khan, Daniel Stone
Cc: kernel, amd-gfx, dri-devel, linux-kernel, linux-arm-kernel,
linux-rockchip, intel-gfx, intel-xe, linux-doc, wayland-devel,
Nicolas Frattaroli, Dmitry Baryshkov
In-Reply-To: <20260609-color-format-v17-0-35739b5782cc@collabora.com>
Add a function to get the connector color format from a connector state,
and a new function pointer in drm_connector_funcs to allow connectors to
override what connector color format it returns.
This is useful for the bridge chain recursive bus format selection code,
which does not wish to implement connector implementation specific
checks like whether it involves HDMI.
Reviewed-by: Dmitry Baryshkov <dmitry.baryshkov@oss.qualcomm.com>
Reviewed-by: Daniel Stone <daniel@fooishbar.org>
Signed-off-by: Nicolas Frattaroli <nicolas.frattaroli@collabora.com>
---
drivers/gpu/drm/drm_connector.c | 16 ++++++++++++++++
include/drm/drm_connector.h | 12 ++++++++++++
2 files changed, 28 insertions(+)
diff --git a/drivers/gpu/drm/drm_connector.c b/drivers/gpu/drm/drm_connector.c
index 296195087a4c..95c65ed18709 100644
--- a/drivers/gpu/drm/drm_connector.c
+++ b/drivers/gpu/drm/drm_connector.c
@@ -3694,6 +3694,22 @@ void drm_connector_oob_hotplug_event(struct fwnode_handle *connector_fwnode,
}
EXPORT_SYMBOL(drm_connector_oob_hotplug_event);
+/**
+ * drm_connector_get_color_format - Return connector color format of @conn_state
+ * @conn_state: pointer to the &struct drm_connector_state to go check
+ *
+ */
+enum drm_connector_color_format
+drm_connector_get_color_format(const struct drm_connector_state *conn_state)
+{
+ struct drm_connector *connector = conn_state->connector;
+
+ if (connector->funcs->color_format)
+ return connector->funcs->color_format(conn_state);
+
+ return conn_state->color_format;
+}
+EXPORT_SYMBOL(drm_connector_get_color_format);
/**
* DOC: Tile group
diff --git a/include/drm/drm_connector.h b/include/drm/drm_connector.h
index e1f6072ce7c9..4317166562cf 100644
--- a/include/drm/drm_connector.h
+++ b/include/drm/drm_connector.h
@@ -1803,6 +1803,16 @@ struct drm_connector_funcs {
* Allows connectors to create connector-specific debugfs files.
*/
void (*debugfs_init)(struct drm_connector *connector, struct dentry *root);
+
+ /**
+ * @color_format:
+ *
+ * Allows connectors to return a connector color format other than
+ * @conn_state.color_format for purposes of e.g. display protocol
+ * specific helper logic having already mapped it to an output format.
+ */
+ enum drm_connector_color_format (*color_format)(
+ const struct drm_connector_state *conn_state);
};
/**
@@ -2619,6 +2629,8 @@ drm_connector_is_unregistered(struct drm_connector *connector)
void drm_connector_oob_hotplug_event(struct fwnode_handle *connector_fwnode,
enum drm_connector_status status);
+enum drm_connector_color_format
+drm_connector_get_color_format(const struct drm_connector_state *conn_state);
const char *drm_get_connector_type_name(unsigned int connector_type);
const char *drm_get_connector_status_name(enum drm_connector_status status);
const char *drm_get_subpixel_order_name(enum subpixel_order order);
--
2.54.0
^ permalink raw reply related
* [PATCH v17 03/28] drm: Add new general DRM property "color format"
From: Nicolas Frattaroli @ 2026-06-09 12:43 UTC (permalink / raw)
To: Harry Wentland, Leo Li, Rodrigo Siqueira, Alex Deucher,
Christian König, David Airlie, Simona Vetter,
Maarten Lankhorst, Maxime Ripard, Thomas Zimmermann,
Andrzej Hajda, Neil Armstrong, Robert Foss, Laurent Pinchart,
Jonas Karlman, Jernej Skrabec, Sandy Huang, Heiko Stübner,
Andy Yan, Jani Nikula, Rodrigo Vivi, Joonas Lahtinen,
Tvrtko Ursulin, Dmitry Baryshkov, Sascha Hauer, Rob Herring,
Jonathan Corbet, Shuah Khan, Daniel Stone
Cc: kernel, amd-gfx, dri-devel, linux-kernel, linux-arm-kernel,
linux-rockchip, intel-gfx, intel-xe, linux-doc, wayland-devel,
Nicolas Frattaroli, Werner Sembach, Andri Yngvason, Marius Vlad
In-Reply-To: <20260609-color-format-v17-0-35739b5782cc@collabora.com>
Add a new general DRM property named "color format" which can be used by
userspace to request the display driver to output a particular color
format.
Possible string values for the new enum property are:
- "AUTO" (setup by default, driver internally picks the color format)
- "RGB"
- "YUV 4:4:4"
- "YUV 4:2:2"
- "YUV 4:2:0"
Drivers should advertise from this list the formats they support in an
optimistic best-case scenario. EDID data from the sink can then be used
in the kernel's atomic check phase to restrict this set of formats, as
well as by userspace to make a correct choice in the first place.
Co-developed-by: Werner Sembach <wse@tuxedocomputers.com>
Signed-off-by: Werner Sembach <wse@tuxedocomputers.com>
Co-developed-by: Andri Yngvason <andri@yngvason.is>
Signed-off-by: Andri Yngvason <andri@yngvason.is>
Signed-off-by: Marius Vlad <marius.vlad@collabora.com>
Reviewed-by: Maxime Ripard <mripard@kernel.org>
Reviewed-by: Daniel Stone <daniel@fooishbar.org>
Signed-off-by: Nicolas Frattaroli <nicolas.frattaroli@collabora.com>
---
Documentation/gpu/drm-kms.rst | 6 ++
drivers/gpu/drm/drm_atomic_helper.c | 5 ++
drivers/gpu/drm/drm_atomic_uapi.c | 4 +
drivers/gpu/drm/drm_connector.c | 155 ++++++++++++++++++++++++++++++++++++
include/drm/drm_connector.h | 84 +++++++++++++++++++
5 files changed, 254 insertions(+)
diff --git a/Documentation/gpu/drm-kms.rst b/Documentation/gpu/drm-kms.rst
index e4f7153543b3..0dd440a14946 100644
--- a/Documentation/gpu/drm-kms.rst
+++ b/Documentation/gpu/drm-kms.rst
@@ -610,6 +610,12 @@ Color Management Properties
.. kernel-doc:: drivers/gpu/drm/drm_color_mgmt.c
:doc: overview
+Color Format Property
+---------------------
+
+.. kernel-doc:: drivers/gpu/drm/drm_connector.c
+ :doc: Color format
+
Tile Group Property
-------------------
diff --git a/drivers/gpu/drm/drm_atomic_helper.c b/drivers/gpu/drm/drm_atomic_helper.c
index 51f39edc31ed..3547f797a850 100644
--- a/drivers/gpu/drm/drm_atomic_helper.c
+++ b/drivers/gpu/drm/drm_atomic_helper.c
@@ -737,6 +737,11 @@ drm_atomic_helper_check_modeset(struct drm_device *dev,
if (old_connector_state->max_requested_bpc !=
new_connector_state->max_requested_bpc)
new_crtc_state->connectors_changed = true;
+
+ if (old_connector_state->color_format !=
+ new_connector_state->color_format)
+ new_crtc_state->connectors_changed = true;
+
}
if (funcs->atomic_check)
diff --git a/drivers/gpu/drm/drm_atomic_uapi.c b/drivers/gpu/drm/drm_atomic_uapi.c
index 6441b55cc274..c7f80d90794c 100644
--- a/drivers/gpu/drm/drm_atomic_uapi.c
+++ b/drivers/gpu/drm/drm_atomic_uapi.c
@@ -935,6 +935,8 @@ static int drm_atomic_connector_set_property(struct drm_connector *connector,
state->privacy_screen_sw_state = val;
} else if (property == connector->broadcast_rgb_property) {
state->hdmi.broadcast_rgb = val;
+ } else if (property == connector->color_format_property) {
+ state->color_format = val;
} else if (connector->funcs->atomic_set_property) {
return connector->funcs->atomic_set_property(connector,
state, property, val);
@@ -1020,6 +1022,8 @@ drm_atomic_connector_get_property(struct drm_connector *connector,
*val = state->privacy_screen_sw_state;
} else if (property == connector->broadcast_rgb_property) {
*val = state->hdmi.broadcast_rgb;
+ } else if (property == connector->color_format_property) {
+ *val = state->color_format;
} else if (connector->funcs->atomic_get_property) {
return connector->funcs->atomic_get_property(connector,
state, property, val);
diff --git a/drivers/gpu/drm/drm_connector.c b/drivers/gpu/drm/drm_connector.c
index a5d13b92b665..296195087a4c 100644
--- a/drivers/gpu/drm/drm_connector.c
+++ b/drivers/gpu/drm/drm_connector.c
@@ -1397,6 +1397,18 @@ static const u32 hdmi_colorspaces =
BIT(DRM_MODE_COLORIMETRY_DCI_P3_RGB_D65) |
BIT(DRM_MODE_COLORIMETRY_DCI_P3_RGB_THEATER);
+static const u32 hdmi_colorformats =
+ BIT(DRM_OUTPUT_COLOR_FORMAT_RGB444) |
+ BIT(DRM_OUTPUT_COLOR_FORMAT_YCBCR444) |
+ BIT(DRM_OUTPUT_COLOR_FORMAT_YCBCR422) |
+ BIT(DRM_OUTPUT_COLOR_FORMAT_YCBCR420);
+
+static const u32 dp_colorformats =
+ BIT(DRM_OUTPUT_COLOR_FORMAT_RGB444) |
+ BIT(DRM_OUTPUT_COLOR_FORMAT_YCBCR444) |
+ BIT(DRM_OUTPUT_COLOR_FORMAT_YCBCR422) |
+ BIT(DRM_OUTPUT_COLOR_FORMAT_YCBCR420);
+
/*
* As per DP 1.4a spec, 2.2.5.7.5 VSC SDP Payload for Pixel Encoding/Colorimetry
* Format Table 2-120
@@ -2944,6 +2956,149 @@ int drm_connector_attach_colorspace_property(struct drm_connector *connector)
}
EXPORT_SYMBOL(drm_connector_attach_colorspace_property);
+/**
+ * DOC: Color format
+ *
+ * The connector "color format" property allows userspace to request a specific
+ * color model on the output of the connector. Not all values listed by the
+ * property are guaranteed to work for every sink; rather, it is an optimistic
+ * listing of color formats that the source could output depending on
+ * circumstances.
+ *
+ * Whether it actually can output a certain color format is determined during
+ * the atomic check phase. Consequently, a userspace application that sets the
+ * color format to a value other than "AUTO" should check whether its atomic
+ * commit succeeded.
+ *
+ * Possible values for "color format":
+ *
+ * "AUTO":
+ * The driver or display protocol helpers should pick a suitable color
+ * format. All implementations of a specific display protocol will behave
+ * the same way with "AUTO", but different display protocols do not
+ * necessarily have the same "AUTO" semantics.
+ *
+ * For HDMI connectors, "AUTO" picks RGB, but falls back to YUV 4:2:0 if
+ * the bandwidth required for full-scale RGB is not available, or the mode
+ * is YUV 4:2:0-only, as long as the mode, source, and sink all support
+ * YUV 4:2:0.
+ * "RGB":
+ * RGB output format. The quantization range (limited/full) depends on the
+ * value of the "Broadcast RGB" property if it is present on the connector.
+ * "YUV 4:4:4":
+ * YUV 4:4:4 (a.k.a. YCbCr 4:4:4) output format. Chroma is not subsampled.
+ * The quantization range defaults to limited.
+ * "YUV 4:2:2":
+ * YUV 4:2:2 (a.k.a. YCbCr 4:2:2) output format. Chroma has half the
+ * horizontal resolution of Luma. The quantization range defaults to
+ * limited.
+ * "YUV 4:2:0":
+ * YUV 4:2:0 (a.k.a. YCbCr 4:2:0) output format. Chroma has half the
+ * horizontal and vertical resolution of Luma. The quantization range
+ * defaults to limited.
+ *
+ * A sink may only support some color formats in specific modes and at specific
+ * bit depths. The atomic modesetting API should be used to set a working
+ * configuration in one go, as an unsupported combination of parameters is
+ * rejected.
+ */
+
+/**
+ * drm_connector_attach_color_format_property - create and attach color format property
+ * @connector: connector to create the color format property on
+ * @supported_color_formats: bitmask of bit-shifted &enum drm_output_color_format
+ * values the connector supports
+ *
+ * Called by a driver to create a color format property. The property is
+ * attached to the connector automatically on success.
+ *
+ * @supported_color_formats should only include color formats the connector
+ * type can actually support.
+ *
+ * Returns:
+ * 0 on success, negative errno on error
+ */
+int drm_connector_attach_color_format_property(struct drm_connector *connector,
+ unsigned long supported_color_formats)
+{
+ struct drm_device *dev = connector->dev;
+ struct drm_prop_enum_list enum_list[DRM_CONNECTOR_COLOR_FORMAT_COUNT];
+ unsigned int i = 0;
+ unsigned long fmt;
+
+ if (connector->color_format_property)
+ return 0;
+
+ if (!supported_color_formats) {
+ drm_err(dev, "No supported color formats provided on [CONNECTOR:%d:%s]\n",
+ connector->base.id, connector->name);
+ return -EINVAL;
+ }
+
+ if (supported_color_formats & ~GENMASK(DRM_OUTPUT_COLOR_FORMAT_COUNT - 1, 0)) {
+ drm_err(dev, "Unknown color formats provided on [CONNECTOR:%d:%s]\n",
+ connector->base.id, connector->name);
+ return -EINVAL;
+ }
+
+ switch (connector->connector_type) {
+ case DRM_MODE_CONNECTOR_HDMIA:
+ case DRM_MODE_CONNECTOR_HDMIB:
+ if (supported_color_formats & ~hdmi_colorformats) {
+ drm_err(dev, "Color formats not allowed for HDMI on [CONNECTOR:%d:%s]\n",
+ connector->base.id, connector->name);
+ return -EINVAL;
+ }
+ break;
+ case DRM_MODE_CONNECTOR_DisplayPort:
+ case DRM_MODE_CONNECTOR_eDP:
+ if (supported_color_formats & ~dp_colorformats) {
+ drm_err(dev, "Color formats not allowed for DP on [CONNECTOR:%d:%s]\n",
+ connector->base.id, connector->name);
+ return -EINVAL;
+ }
+ break;
+ }
+
+ enum_list[0].name = "AUTO";
+ enum_list[0].type = DRM_CONNECTOR_COLOR_FORMAT_AUTO;
+
+ for_each_set_bit(fmt, &supported_color_formats, DRM_OUTPUT_COLOR_FORMAT_COUNT) {
+ switch (fmt) {
+ case DRM_OUTPUT_COLOR_FORMAT_RGB444:
+ enum_list[++i].type = DRM_CONNECTOR_COLOR_FORMAT_RGB444;
+ break;
+ case DRM_OUTPUT_COLOR_FORMAT_YCBCR444:
+ enum_list[++i].type = DRM_CONNECTOR_COLOR_FORMAT_YCBCR444;
+ break;
+ case DRM_OUTPUT_COLOR_FORMAT_YCBCR422:
+ enum_list[++i].type = DRM_CONNECTOR_COLOR_FORMAT_YCBCR422;
+ break;
+ case DRM_OUTPUT_COLOR_FORMAT_YCBCR420:
+ enum_list[++i].type = DRM_CONNECTOR_COLOR_FORMAT_YCBCR420;
+ break;
+ default:
+ drm_warn(dev, "Unknown supported format %ld on [CONNECTOR:%d:%s]\n",
+ fmt, connector->base.id, connector->name);
+ continue;
+ }
+ enum_list[i].name = drm_hdmi_connector_get_output_format_name(fmt);
+ }
+
+ connector->color_format_property =
+ drm_property_create_enum(dev, DRM_MODE_PROP_ENUM, "color format",
+ enum_list, i + 1);
+
+ if (!connector->color_format_property)
+ return -ENOMEM;
+
+ drm_object_attach_property(&connector->base, connector->color_format_property,
+ DRM_CONNECTOR_COLOR_FORMAT_AUTO);
+
+ return 0;
+}
+EXPORT_SYMBOL(drm_connector_attach_color_format_property);
+
/**
* drm_connector_atomic_hdr_metadata_equal - checks if the hdr metadata changed
* @old_state: old connector state to compare
diff --git a/include/drm/drm_connector.h b/include/drm/drm_connector.h
index 529755c2e862..e1f6072ce7c9 100644
--- a/include/drm/drm_connector.h
+++ b/include/drm/drm_connector.h
@@ -571,12 +571,80 @@ enum drm_colorspace {
* YCbCr 4:2:2 output format (ie. with horizontal subsampling)
* @DRM_OUTPUT_COLOR_FORMAT_YCBCR420:
* YCbCr 4:2:0 output format (ie. with horizontal and vertical subsampling)
+ * @DRM_OUTPUT_COLOR_FORMAT_COUNT:
+ * Number of valid output color format values in this enum
*/
enum drm_output_color_format {
DRM_OUTPUT_COLOR_FORMAT_RGB444 = 0,
DRM_OUTPUT_COLOR_FORMAT_YCBCR444,
DRM_OUTPUT_COLOR_FORMAT_YCBCR422,
DRM_OUTPUT_COLOR_FORMAT_YCBCR420,
+ DRM_OUTPUT_COLOR_FORMAT_COUNT,
+};
+
+/**
+ * enum drm_connector_color_format - Connector Color Format Request
+ *
+ * This enum, unlike &enum drm_output_color_format, is used to specify requests
+ * for a specific color format on a connector through the DRM "color format"
+ * property. The difference is that it has an "AUTO" value to specify that
+ * no specific choice has been made.
+ */
+enum drm_connector_color_format {
+ /**
+ * @DRM_CONNECTOR_COLOR_FORMAT_AUTO: The driver or display protocol
+ * helpers should pick a suitable color format. All implementations of a
+ * specific display protocol must behave the same way with "AUTO", but
+ * different display protocols do not necessarily have the same "AUTO"
+ * semantics.
+ *
+ * For HDMI, "AUTO" picks RGB, but falls back to YCbCr 4:2:0 if the
+ * bandwidth required for full-scale RGB is not available, or the mode
+ * is YCbCr 4:2:0-only, as long as the mode and output both support
+ * YCbCr 4:2:0.
+ *
+ * For display protocols other than HDMI, the recursive bridge chain
+ * format selection picks the first chain of bridge formats that works,
+ * as has already been the case before the introduction of the "color
+ * format" property. Non-HDMI bridges should therefore either sort their
+ * bus output formats by preference, or agree on a unified auto format
+ * selection logic that's implemented in a common state helper (like
+ * how HDMI does it).
+ */
+ DRM_CONNECTOR_COLOR_FORMAT_AUTO = 0,
+
+ /**
+ * @DRM_CONNECTOR_COLOR_FORMAT_RGB444: RGB output format. The
+ * quantization range depends on the value of the "Broadcast RGB"
+ * property if it is present on the connector.
+ */
+ DRM_CONNECTOR_COLOR_FORMAT_RGB444,
+
+ /**
+ * @DRM_CONNECTOR_COLOR_FORMAT_YCBCR444: YCbCr 4:4:4 output format (ie.
+ * not subsampled). Quantization range is "Limited" by default.
+ */
+ DRM_CONNECTOR_COLOR_FORMAT_YCBCR444,
+
+ /**
+ * @DRM_CONNECTOR_COLOR_FORMAT_YCBCR422: YCbCr 4:2:2 output format (ie.
+ * with horizontal subsampling). Quantization range is "Limited" by
+ * default.
+ */
+ DRM_CONNECTOR_COLOR_FORMAT_YCBCR422,
+
+ /**
+ * @DRM_CONNECTOR_COLOR_FORMAT_YCBCR420: YCbCr 4:2:0 output format (ie.
+ * with horizontal and vertical subsampling). Quantization range is
+ * "Limited" by default.
+ */
+ DRM_CONNECTOR_COLOR_FORMAT_YCBCR420,
+
+ /**
+ * @DRM_CONNECTOR_COLOR_FORMAT_COUNT: Number of valid connector color
+ * format values in this enum
+ */
+ DRM_CONNECTOR_COLOR_FORMAT_COUNT,
};
const char *
@@ -1167,6 +1235,13 @@ struct drm_connector_state {
*/
enum drm_colorspace colorspace;
+ /**
+ * @color_format: State variable for Connector property to request
+ * color format change on Sink. This is most commonly used to switch
+ * between RGB to YUV and vice-versa.
+ */
+ enum drm_connector_color_format color_format;
+
/**
* @writeback_job: Writeback job for writeback connectors
*
@@ -2181,6 +2256,12 @@ struct drm_connector {
*/
struct drm_property *colorspace_property;
+ /**
+ * @color_format_property: Connector property to set the suitable
+ * color format supported by the sink.
+ */
+ struct drm_property *color_format_property;
+
/**
* @path_blob_ptr:
*
@@ -2664,6 +2745,9 @@ bool drm_connector_has_possible_encoder(struct drm_connector *connector,
struct drm_encoder *encoder);
const char *drm_get_colorspace_name(enum drm_colorspace colorspace);
+int drm_connector_attach_color_format_property(struct drm_connector *connector,
+ unsigned long supported_color_formats);
+
/**
* drm_for_each_connector_iter - connector_list iterator macro
* @connector: &struct drm_connector pointer used as cursor
--
2.54.0
^ permalink raw reply related
* [PATCH v17 01/28] drm/amd/display: Remove unnecessary SIGNAL_TYPE_HDMI_TYPE_A check
From: Nicolas Frattaroli @ 2026-06-09 12:43 UTC (permalink / raw)
To: Harry Wentland, Leo Li, Rodrigo Siqueira, Alex Deucher,
Christian König, David Airlie, Simona Vetter,
Maarten Lankhorst, Maxime Ripard, Thomas Zimmermann,
Andrzej Hajda, Neil Armstrong, Robert Foss, Laurent Pinchart,
Jonas Karlman, Jernej Skrabec, Sandy Huang, Heiko Stübner,
Andy Yan, Jani Nikula, Rodrigo Vivi, Joonas Lahtinen,
Tvrtko Ursulin, Dmitry Baryshkov, Sascha Hauer, Rob Herring,
Jonathan Corbet, Shuah Khan, Daniel Stone
Cc: kernel, amd-gfx, dri-devel, linux-kernel, linux-arm-kernel,
linux-rockchip, intel-gfx, intel-xe, linux-doc, wayland-devel,
Nicolas Frattaroli, Werner Sembach, Andri Yngvason
In-Reply-To: <20260609-color-format-v17-0-35739b5782cc@collabora.com>
From: Werner Sembach <wse@tuxedocomputers.com>
Remove unnecessary SIGNAL_TYPE_HDMI_TYPE_A check that was performed in the
drm_mode_is_420_only() case, but not in the drm_mode_is_420_also() &&
force_yuv420_output case.
Without further knowledge if YCbCr 4:2:0 is supported outside of HDMI,
there is no reason to use RGB when the display
reports drm_mode_is_420_only() even on a non HDMI connection.
This patch also moves both checks in the same if-case. This eliminates an
extra else-if-case.
Signed-off-by: Werner Sembach <wse@tuxedocomputers.com>
Signed-off-by: Andri Yngvason <andri@yngvason.is>
Tested-by: Andri Yngvason <andri@yngvason.is>
Reviewed-by: Daniel Stone <daniel@fooishbar.org>
Signed-off-by: Nicolas Frattaroli <nicolas.frattaroli@collabora.com>
---
drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c | 9 +++------
1 file changed, 3 insertions(+), 6 deletions(-)
diff --git a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c
index ba7f98a87808..dfe97897127c 100644
--- a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c
+++ b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c
@@ -6917,12 +6917,9 @@ static void fill_stream_properties_from_drm_display_mode(
timing_out->v_border_top = 0;
timing_out->v_border_bottom = 0;
/* TODO: un-hardcode */
- if (drm_mode_is_420_only(info, mode_in)
- && stream->signal == SIGNAL_TYPE_HDMI_TYPE_A)
- timing_out->pixel_encoding = PIXEL_ENCODING_YCBCR420;
- else if (drm_mode_is_420_also(info, mode_in)
- && aconnector
- && aconnector->force_yuv420_output)
+ if (drm_mode_is_420_only(info, mode_in) ||
+ (aconnector && aconnector->force_yuv420_output &&
+ drm_mode_is_420_also(info, mode_in)))
timing_out->pixel_encoding = PIXEL_ENCODING_YCBCR420;
else if ((connector->display_info.color_formats & BIT(DRM_OUTPUT_COLOR_FORMAT_YCBCR422))
&& aconnector
--
2.54.0
^ permalink raw reply related
* [PATCH v17 00/28] Add new general DRM property "color format"
From: Nicolas Frattaroli @ 2026-06-09 12:43 UTC (permalink / raw)
To: Harry Wentland, Leo Li, Rodrigo Siqueira, Alex Deucher,
Christian König, David Airlie, Simona Vetter,
Maarten Lankhorst, Maxime Ripard, Thomas Zimmermann,
Andrzej Hajda, Neil Armstrong, Robert Foss, Laurent Pinchart,
Jonas Karlman, Jernej Skrabec, Sandy Huang, Heiko Stübner,
Andy Yan, Jani Nikula, Rodrigo Vivi, Joonas Lahtinen,
Tvrtko Ursulin, Dmitry Baryshkov, Sascha Hauer, Rob Herring,
Jonathan Corbet, Shuah Khan, Daniel Stone
Cc: kernel, amd-gfx, dri-devel, linux-kernel, linux-arm-kernel,
linux-rockchip, intel-gfx, intel-xe, linux-doc, wayland-devel,
Nicolas Frattaroli, Werner Sembach, Andri Yngvason,
Cristian Ciocaltea, Marius Vlad, Dmitry Baryshkov, Andy Yan
Hello,
this is a follow-up to
https://lore.kernel.org/all/20250911130739.4936-1-marius.vlad@collabora.com/
which in of itself is a follow-up to
https://lore.kernel.org/dri-devel/20240115160554.720247-1-andri@yngvason.is/ where
a new DRM connector property has been added allowing users to
force a particular color format.
That in turn was actually also a follow-up from Werner Sembach's posted at
https://lore.kernel.org/dri-devel/20210630151018.330354-1-wse@tuxedocomputers.com/
As the number of cooks have reached critical mass, I'm hoping I'll be
the last person to touch this particular series.
We have an implementation in Weston at
https://gitlab.freedesktop.org/wayland/weston/-/merge_requests/1825 that
adds support for this property. This patch series has been tested
against that MR on i915 (HDMI, DP), amdgpu (HDMI, DP) and on rockchip
(HDMI).
You can also manually test this with modetest like so, but beware that
this is a non-atomic invocation, so testing YUV420 like this will result
in weird outcomes if only some of the modes support YUV420:
$ modetest -s 115:1920x1080-60@NV12 -w 115:'color format':4
where 115 is the connector ID and '4' is the enum value for a particular
color format.
General notes on the approach taken by me: instead of silently switching
to a different format than was explicitly requested, or even worse,
outputting something to the sink the sink doesn't support, bubble up an
error to userspace instead. "color format" is a "I want this" type
property, not a "force this" type property, i.e. the kernel will respect
the limits imposed by the hardware.
Things I've tested:
- HDMI (YCbCr 4:4:4 + YCbCr 4:2:2 (8-bit) + RGB + Auto) on RK3588
- HDMI (YCbCr 4:4:4 + YCbCr 4:2:2 (8-bit) + RGB + Auto) on RK3576
- HDMI (YCbCr 4:4:4, YCbCr 4:2:0, RGB, Auto) + DP (YCbCr 4:4:4, RGB,
Auto) on Intel N97 (i915).
- HDMI (YCbCr 4:4:4, YCbCr 4:2:2, YCbCr 4:2:0, RGB, Auto) + DP (YCbCr
4:4:4, RGB, Auto) on an AMD Radeon RX 550 (amdgpu).
Signed-off-by: Nicolas Frattaroli <nicolas.frattaroli@collabora.com>
---
Changes in v17:
- Rebase onto drm-misc-next
- Reorder i915 patches to end of series
- Link to v16: https://patch.msgid.link/20260523-color-format-v16-0-24340c5e4732@collabora.com
Changes in v16:
- drm/i915/dp: Don't check for dfp conversion to determine ycbcr444
output capability
- drm/i915/dp: Replicate intel_dp_has_hdmi_sink check for YCBCR444 as well
- drm/i915/dp: Reorder formats and checks in sink_format_valid
- drm/i915/hdmi: Fix inverted HAS_GMCH check for YCBCR444 capability
- drm/i915/hdmi: Add has_hdmi_sink check to YCBCR444 sink_format_valid
- drm/i915/hdmi: Reorder formats in sink_format_valid
- drm/i915/hdmi: Drop documentation for internal function
- Link to v15: https://patch.msgid.link/20260522-color-format-v15-0-21fb136c9df2@collabora.com
Changes in v15:
- Drop redundant drm_connector_color_format_valid() function
- dw-hdmi-qp-rockchip: make dw_hdmi_qp_rockchip_get_vop_format()'s
return type int, check bstate for IS_ERR rather than NULL, and pass up
error codes to caller
- rebase onto recent drm-tip, necessitating
s/drm_atomic_state/drm_atomic_commit/ in the KUnit tests
- Link to v14: https://patch.msgid.link/20260423-color-format-v14-0-449a419ccbd4@collabora.com
Changes in v14:
- i915: Check for source support of YCbCr444 in both the sink format
validation and when registering the properites on the connectors.
- i915: Split HDMI and DP implementations into separate patches, with
their own supported color formats masks.
- Link to v13: https://patch.msgid.link/20260413-color-format-v13-0-ab37d4dfba48@collabora.com
Changes in v13:
- Introduce a new drm_connector_funcs member that returns the
connector state's color format with additional bells and whistles.
- Implement this new func in drm_bridge_connector to return whatever the
HDMI implementation did if an HDMI bridge is present.
- Call into the drm_connector.h function wrapping this new func in the
recursive bridge chain bus format selection code.
- Link to v12: https://patch.msgid.link/20260409-color-format-v12-0-ce84e1817a27@collabora.com
Changes in v12:
- Rebase on top of Ville's i915 format selection cleanup series
- i915: Reimplement based on new format selection code. Drop the DP-MST
implementation, as sinks with different sets of supported formats
complicate the matter.
- Adjust documentation on drm_connector_color_format enum values to
mention quantization ranges.
- Add documentation to drm-kms page to document the string values for
the property. Mostly the same documentation as the enum values, but
from a more userspace-centric perspective. This is done in the patch
that introduces the property, but I didn't drop the existing R-b as
it isn't a functional change.
- Update documentation of "colorspace" property to mention that "color
format" property controls the output format.
- amdgpu: Remove property from DP-MST, as it's unclear whether it'd work
properly with sinks that have different sets of supported formats
- Link to v11: https://patch.msgid.link/20260324-color-format-v11-0-605559af4fb4@collabora.com
Changes in v11:
- amdgpu: fix property registration on DP-MST
- i915: fix property registration on DP-MST
- rebase on drm-tip, which includes Maxime's refactor series that was
previously declared a dependency of this series
- Link to v10: https://lore.kernel.org/r/20260305-color-format-v10-0-a58c68a11868@collabora.com
Changes in v10:
- Make DRM_OUTPUT_COLOR_FORMAT_COUNT and
DRM_CONNECTOR_COLOR_FORMAT_COUNT part of the enum definition (thanks
to Maxime)
- Preemptively avoid the warning that would be generated by the
enumification of DRM_OUTPUT_COLOR_FORMAT_COUNT by modifying the
problematic switch statement in drm_hdmi_state_helper's
sink_supports_format_bpc.
- drm/bridge: Change HDMI check from checking for the last bridge having
a DRM_BRIDGE_OP_HDMI in ops to checking if last_bridge->type is HDMIA.
This is not quite the suggestion Dmitry had, but according to the
documentation of the drm_bridge.type member, and the
display-connector.c code, it should be correct.
- Combine drm_mode_create_color_format_property and
drm_connector_attach_color_format_property into one function named the
latter. (thanks to Dmitry Baryshkov)
- Change author of 'drm: Add new general DRM property "color format"'
to myself as it has by now changed quite a bit, and add Andri and
Werner as Co-developed-by, as per Andri's suggestion.
- hdmi-state-helper: Rework hdmi_compute_config to make code flow more
obvious, and drop Dmitry's R-b as a consequence (thanks to Maxime)
- Move dw-hdmi-qp's atomic_get_output_bus_fmts into
drm_bridge_helper.c, along with kernel doc string (thanks to Dmitry)
- Future-proof the aforementioned get_output_bus_fmts use of hweight8 on
the supported_formats bitmask with a BUILD_BUG_ON.
- Add a KUnit test for the HDMI output bus formats helper
- Link to v9: https://lore.kernel.org/r/20260227-color-format-v9-0-658c3b9db7ef@collabora.com
Changes in v9:
- Document what the "AUTO" behaviour is in the color format enum (thanks
to Maxime)
- drm/bridge: dw-hdmi-qp: Fix a rebase oopsie that reintroduced some
functions that were dropped. (thanks to Cristian)
- drm/bridge: Shuffle "1:1" in the bridge fmt selection docs to earlier
in the sentence. (thanks to Randy Dunlap)
- i915: Check chosen output format against requested format for dp-mst
- All color format driver implementations: rebase and rework on top of
Maxime's series
- As part of this rework, rename drm_color_format_enum to
drm_connector_color_format
- drm kunit tests: rework for the new enums. Changes were trivial, so
trailers were kept
- Link to v8: https://lore.kernel.org/r/20260216-color-format-v8-0-5722ce175dd5@collabora.com
Changes in v8:
- Drop "drm/rockchip: vop2: Fix YUV444 output", as the original problem
could not be reproduced anymore, and the justification did not make
sense.
- Remove the 12-bit format from "drm/rockchip: vop2: Recognise 10/12-bit
YUV422 as YUV formats".
- Refactor to keep the original DRM_COLOR_FORMAT bitshifted defines
as-is, but introduce a new drm_color_format_enum enum.
- Adjust conversion functions for the newly refactored enum, ensuring
they only return valid enum values, and only convert in directions
that open up no error value cans of worms.
- Rework the property uapi code for the newly refactored enum, since
it no longer needs to do any bitshifting or ffs().
- Rework all the device drivers for the new enum.
- Rework all the tests for the refactored enum.
- Rework the hdmi state helper for the new enum, and also make it more
explicit about the auto behaviour by not relying on a conversion
function to map AUTO to RGB, but do this in the framework itself.
- rockchip dw_hdmi_qp: Fix the GRF value to check for color >= 0 instead
of color > 0, as the latter broke switching back to RGB.
- Rebase onto a recent drm-tip. This necessitated blindly reworking some
of the i915 dp-mst code.
- Drop the __maybe_unused edid test patch, as I could no longer
reproduce the build warnings I added it for. I blame ghosts.
- drm_bridge tests: remove "destroyed" member from struct
drm_bridge_chain_priv and all associated code, as it was not used in
any test.
- Link to v7: https://lore.kernel.org/r/20260121-color-format-v7-0-ef790dae780c@collabora.com
Changes in v7:
- Fix drm_bridge kunit test build failure caused by rebasing across an
API change.
- Make compilers shut up about unused EDID definitions in the test
suites.
- Empty line checkpatch fixes that b4 prep --check didn't catch.
- Link to v6: https://lore.kernel.org/r/20260121-color-format-v6-0-7b81a771cd0b@collabora.com
Changes in v6:
- Checkpatch fixes
- Add drm_bridge.c kerneldoc fix patch to b4 deps so the kernel docs
required for every contribution to the subsystem can be built
- dw-hdmi-qp core has gained the atomic_get_output_bus_fmts bridge func,
which allows it to participate in the drm_bridge chain recursive format
selection code properly.
- The Rockchip dw-hdmi-qp integration now no longer reimplements the
color format logic (improperly), but reads the bus format of the first
bridge as set by the recursive bridge format selection. If the input
format is FIXED, it'll use the output format. Otherwise, the input
format is used.
- In the synopsys drivers, YUV422 uses the same bus format as the non-qp
hdmi encoder driver. Probably correcter this way. The Rockchip vop2
is_yuv function has been extended to recognise this format as well.
- KUnit tests for drm_bridge chains are now included, which exercise the
chain's recursive bus format selection.
- On HDMI connectors, the drm_bridge bus format selection will try to target
the color format that the HDMI layer came up with. This means the AUTO
logic is not duplicated for HDMI connectors.
- The enum conversion function commit gained a function for converting
from hdmi_colorspace to drm_color_format, and its author changed as no
original code remains anyway. Marius is still included as a
Co-developer.
- Some tests for the HDMI state helper's mode_valid have been written.
They are incomplete as we lack a test EDID for a 420-also mode that
would violate the clock constraints on RGB. I hacked one together with
a hex editor, but it reports a too high of a clock rate, and there's
no EDID editor I could find which supports these extension blocks.
- The color_format KUnit tests have been more heavily parameterised, the
auto case absorbed into other tests, and the comments around them
rewritten.
- Add a few paragraphs of documentation that explain the bridge format
selection, and how to make use of it in a display driver.
- Link to v5: https://lore.kernel.org/r/20251128-color-format-v5-0-63e82f1db1e1@collabora.com
Changes in v5:
- Rebase onto drm-tip
- Drop DRM_MODE_COLOR_FORMAT_* as an enum
- Unify DRM_COLOR_FORMAT_NONE and DRM_COLOR_FORMAT_AUTO, with AUTO being
0. This makes conversion and general logic much easier.
- Adjust the drm_color_format enum to not needlessly renumber the
existing defines, as it doesn't need to correspond to how HDMI numbers
them.
- Make the DRM-to-HDMI conversion function static inline __pure, because
the assembly it generates is tiny, and the function is pure.
- Don't accept nothing as the list of supported color formats for
registration of the property.
- Drop the per-connector variants of the color format registration
function, as it's not needed.
- drm_hdmi_state_helper: Fix mode_valid rejecting 420-only modes.
- drm_hdmi_state_helper: Only fall back to YUV420 with
DRM_COLOR_FORMAT_AUTO.
- drm_hdmi_state_helper: Remove redundant AUTO->RGB condition, as the
conversion already does this.
- Add KUnit tests for hdmi_compute_config.
- drm/bridge: Refactor bus_format_is_color_fmt and add a few more YUV422
formats.
- Register the color format property in drmm_connector_hdmi_init based
on the supported HDMI formats passed to it. This means rockchip
dw_hdmi_qp no longer needs to register it.
- amdgpu: Simplify YUV420 logic
- amdgpu: Don't try to pick YUV444 on YUV420-only modes
- i915: Try to make behaviour more or less the same as that of the drm
hdmi state helper.
- rockchip dw_hdmi_qp: Set supported HDMI formats
- rockchip dw_hdmi_qp: Set the right VO GRF values depending on color
format.
- rockchip dw_hdmi_qp: Act on the color format property in this driver,
rather than in VOP2, by setting the bus_format appropriately.
- rockchip VOP2: Can the BCSH-based implementation. BCSH isn't available
on all video ports of the hardware, and the code was extremely
suspect. Instead, plug into the existing YUV-to-RGB/RGB-to-YUV code,
which can be done now that the HDMI driver sets the bus format.
- A whole bunch of Rockchip VOP2 fixes.
- Link to v4: https://lore.kernel.org/r/20251117-color-format-v4-0-0ded72bd1b00@collabora.com
Changes in v4:
- Rebase onto next-20251117
- Get rid of HDMI_COLORSPACE_AUTO
- Split hdmi_compute_config change into separate patch
- Add missing symbol export for color_format_to_hdmi_colorspace to fix
builds in certain configurations
- Drop "drm: Pass supported color formats straight onto drm_bridge"
- Make dw-hdmi-qp set the platform data's supported color formats as
the bridge's supported HDMI color formats
- drm_hdmi_state_helper: pass requested color format to
hdmi_compute_format_bpc if set.
- drm_bridge: limit the bus formats to those explicitly requested with
the color format property during the atomic bridge check call,
specifically in drm_atomic_bridge_chain_select_bus_fmts.
- i915: Remove INTEL_OUTPUT_FORMAT_AUTO, as automatic format selection
does not need to involve the hardware state
- i915: Deduplicate ntel_output_format_to_drm_color_format code by
moving it as a static inline __pure function into a shared header
- i915: rework logic in HDMI, DP and DP-MST output config functions to
remove redundant locals, simplify execution flow, and return an error
to userspace if an explicit color_format request can't be satisfied.
- i915: assign myself as the author and make the others Co-developers,
so that they don't get the blame for any of my bugs.
- amdgpu: refactor fill_stream_properties_from_drm_display_mode to
improve readability and ensure that impossible color format requests
get bubbled up to userspace as errors
- amdgpu: don't pick YUV444 over RGB.
- amdgpu: assign authorship to myself, with others as Co-developers, as
logic was modified and the blame should fall on me
- dw_hdmi_qp-rockchip: set the supported color formats platform data
member
- rockchip: remove drm property registration for rk3066_hdmi and
inno_hdmi. None of the platforms that use these use vop2 as the
video output processor.
- Link to v3: https://lore.kernel.org/all/20250911130739.4936-1-marius.vlad@collabora.com/
Changes in v3 by mvlad compared to Andri's v2 series:
- renamed the property to just 'color format'
- the property is added dynamically similar to the Colorspace property
- a key point from previous comments was that drivers should advertise
the color formats they support and userspace would query EDID and
perform an intersection from those color formats which users can
further use. With this patch set each driver that adds this property
has such list of hard-coded color formats, but fundamentally the idea
is that driver can query the HW and do that on its own. The
infrastructure is now in place to allow to do that
- by default the 'AUTO' color format is set. With this patch series that
has been introduced as a fallback to RGB. Drivers could further
customize this behavour and could perform additional checks on the sink
to pick another suitable color format they'd like for AUTO
- drm_bridge bridge code has been improved to allow initialization with
the same color formats list as the DRM connector property. Similarly, bpc
pick-up now takes the color format into consideration when deciding
which bpc to choose from
- The new DRM color format re-uses HDMI_COLORPSACE enum and provides an
enum translations between the two to avoid touching all other drivers that
use HDMI_COLORPSACE enum. I believe at this point that this allows the
least amount of disruption and avoids a massive bike shedding around
that part
- a rockchip implementation has been by my colleague Derek Foreman
- YUV444 color format has been added in i915
- address comment about "Remove unnecessary SIGNAL_TYPE_HDMI_TYPE_A
check" where aconnector might be invalid
- Link to v2: https://lore.kernel.org/dri-devel/20240115160554.720247-1-andri@yngvason.is/
---
Nicolas Frattaroli (27):
drm/display: hdmi-state-helper: Use default case for unsupported formats
drm: Add new general DRM property "color format"
drm/connector: Let connectors have a say in their color format
drm/display: bridge_connector: Use HDMI color format for HDMI conns
drm/bridge: Act on the DRM color format property
drm/atomic-helper: Add HDMI bridge output bus formats helper
drm/display: hdmi-state-helper: Act on color format DRM property
drm/display: hdmi-state-helper: Try subsampling in mode_valid
drm/amdgpu: Implement "color format" DRM property
drm/rockchip: Add YUV422 output mode constants for VOP2
drm/rockchip: vop2: Add RK3576 to the RG swap special case
drm/rockchip: vop2: Recognise 10-bit YUV422 as YUV format
drm/rockchip: vop2: Set correct output format for RK3576 YUV422
drm/bridge: dw-hdmi-qp: Use common HDMI output bus fmts helper
drm/rockchip: dw_hdmi_qp: Implement "color format" DRM property
drm/rockchip: dw_hdmi_qp: Set supported_formats platdata
drm/connector: Register color format property on HDMI connectors
drm/tests: hdmi: Add tests for the color_format property
drm/tests: hdmi: Add tests for HDMI helper's mode_valid
drm/tests: bridge: Add KUnit tests for bridge chain format selection
drm/tests: bridge: Add test for HDMI output bus formats helper
drm/bridge: Document bridge chain format selection
drm/connector: Update docs of "colorspace" for color format prop
drm/i915/hdmi: Add YCBCR444 handling for sink formats
drm/i915/dp: Add YCBCR444 handling for sink formats
drm/i915/hdmi: Implement "color format" DRM property
drm/i915/dp: Implement "color format" DRM property
Werner Sembach (1):
drm/amd/display: Remove unnecessary SIGNAL_TYPE_HDMI_TYPE_A check
Documentation/gpu/drm-kms-helpers.rst | 6 +
Documentation/gpu/drm-kms.rst | 6 +
drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c | 91 +-
drivers/gpu/drm/bridge/synopsys/dw-hdmi-qp.c | 1 +
drivers/gpu/drm/display/drm_bridge_connector.c | 24 +
drivers/gpu/drm/display/drm_hdmi_state_helper.c | 53 +-
drivers/gpu/drm/drm_atomic_helper.c | 86 ++
drivers/gpu/drm/drm_atomic_uapi.c | 4 +
drivers/gpu/drm/drm_bridge.c | 104 ++-
drivers/gpu/drm/drm_connector.c | 178 +++-
drivers/gpu/drm/i915/display/intel_dp.c | 81 +-
drivers/gpu/drm/i915/display/intel_hdmi.c | 76 +-
drivers/gpu/drm/rockchip/dw_hdmi_qp-rockchip.c | 111 ++-
drivers/gpu/drm/rockchip/rockchip_drm_drv.h | 4 +
drivers/gpu/drm/rockchip/rockchip_drm_vop2.c | 21 +-
drivers/gpu/drm/tests/drm_bridge_test.c | 971 +++++++++++++++++++++
drivers/gpu/drm/tests/drm_hdmi_state_helper_test.c | 345 ++++++++
include/drm/drm_atomic_helper.h | 7 +
include/drm/drm_connector.h | 96 ++
19 files changed, 2228 insertions(+), 37 deletions(-)
---
base-commit: f2f4e6d45fecc6ead92bf5ea06a04a00c245904e
change-id: 20251028-color-format-49fd202b7183
Best regards,
--
Nicolas Frattaroli <nicolas.frattaroli@collabora.com>
^ permalink raw reply
* [PATCH v17 02/28] drm/display: hdmi-state-helper: Use default case for unsupported formats
From: Nicolas Frattaroli @ 2026-06-09 12:43 UTC (permalink / raw)
To: Harry Wentland, Leo Li, Rodrigo Siqueira, Alex Deucher,
Christian König, David Airlie, Simona Vetter,
Maarten Lankhorst, Maxime Ripard, Thomas Zimmermann,
Andrzej Hajda, Neil Armstrong, Robert Foss, Laurent Pinchart,
Jonas Karlman, Jernej Skrabec, Sandy Huang, Heiko Stübner,
Andy Yan, Jani Nikula, Rodrigo Vivi, Joonas Lahtinen,
Tvrtko Ursulin, Dmitry Baryshkov, Sascha Hauer, Rob Herring,
Jonathan Corbet, Shuah Khan, Daniel Stone
Cc: kernel, amd-gfx, dri-devel, linux-kernel, linux-arm-kernel,
linux-rockchip, intel-gfx, intel-xe, linux-doc, wayland-devel,
Nicolas Frattaroli, Cristian Ciocaltea
In-Reply-To: <20260609-color-format-v17-0-35739b5782cc@collabora.com>
Switch statements that do not handle all possible values of an
enumeration will generate a warning during compilation. In preparation
for adding a COUNT value to the end of the enum, this needs to be dealt
with.
Add a default case to sink_supports_format_bpc's DRM_OUTPUT_COLOR_FORMAT
switch statement, and move the log-and-return unknown pixel format
handling into it.
No functional change.
Reviewed-by: Maxime Ripard <mripard@kernel.org>
Reviewed-by: Cristian Ciocaltea <cristian.ciocaltea@collabora.com>
Reviewed-by: Daniel Stone <daniel@fooishbar.org>
Signed-off-by: Nicolas Frattaroli <nicolas.frattaroli@collabora.com>
---
drivers/gpu/drm/display/drm_hdmi_state_helper.c | 7 ++++---
1 file changed, 4 insertions(+), 3 deletions(-)
diff --git a/drivers/gpu/drm/display/drm_hdmi_state_helper.c b/drivers/gpu/drm/display/drm_hdmi_state_helper.c
index a331ebdd65af..c68b159f0a7f 100644
--- a/drivers/gpu/drm/display/drm_hdmi_state_helper.c
+++ b/drivers/gpu/drm/display/drm_hdmi_state_helper.c
@@ -542,10 +542,11 @@ sink_supports_format_bpc(const struct drm_connector *connector,
drm_dbg_kms(dev, "YUV444 format supported in that configuration.\n");
return true;
- }
- drm_dbg_kms(dev, "Unsupported pixel format.\n");
- return false;
+ default:
+ drm_dbg_kms(dev, "Unsupported pixel format.\n");
+ return false;
+ }
}
static enum drm_mode_status
--
2.54.0
^ permalink raw reply related
* RE: [PATCH net-next v3 07/13] net: ethernet: oa_tc6: Add read_mms/write_mms register access functions
From: Regus, Ciprian @ 2026-06-09 12:13 UTC (permalink / raw)
To: Andrew Lunn
Cc: Parthiban Veerasooran, Andrew Lunn, David S. Miller, Eric Dumazet,
Jakub Kicinski, Paolo Abeni, Simon Horman, Jonathan Corbet,
Shuah Khan, Heiner Kallweit, Russell King, Rob Herring,
Krzysztof Kozlowski, Conor Dooley, netdev@vger.kernel.org,
linux-kernel@vger.kernel.org, linux-doc@vger.kernel.org,
devicetree@vger.kernel.org
In-Reply-To: <905f32c3-9638-46e4-8f52-1bc874f4e8b9@lunn.ch>
> > +/**
> > + * oa_tc6_read_register_mms - function for reading a MAC-PHY register
> in a
> > + * memory map other than 0.
>
> Is MMS 0 somehow special? Does it not work if passed to this function?
For addressing registers in MMS 0, we can just use the already existing oa_tc6_read_register
and oa_tc6_write_register functions. So, I meant to say that for transfers to memory maps
other than 0, this function can be used instead. Passing 0 in the mms parameter would
result in the same behavior as if the user called oa_tc6_read_register() with the upper 16 bits
of the addr masked (so there is nothing special about this case).
I agree the documentation is poorly worded and I'll change it to something like this in v4:
/**
* oa_tc6_read_register_mms - function for reading a MAC-PHY register in a specified
* memory map.
>
> I'm not suggesting all the current calls are replaced with this, just
> that the documentation might be inaccurate?
>
> > + * @tc6: oa_tc6 struct.
> > + * @mms: Memory map selector for the register.
> > + * @address: register address of the MAC-PHY to be read.
> > + * @value: value read from the @address register address of the MAC-
> PHY.
> > + *
> > + * Return: 0 on success otherwise failed.
>
> Isn't the "otherwise" actually an error code?
This function can return the same set of error codes as oa_tc6_read_register/ oa_tc6_write_register.
Those can be -EPROTO, -EINVAL or any error code returned by the spi_sync(). Since the other register
access functions have the same documentation for the return values, I kept it the same here.
I'll change it to "0 on success or a negative error code on failure".
>
> Andrew
^ permalink raw reply
* Re: [PATCH v3 1/3] PM: core: Rename module parameters prefix to "power"
From: Tomasz Figa @ 2026-06-09 11:17 UTC (permalink / raw)
To: Rafael J. Wysocki
Cc: Tzung-Bi Shih, Jonathan Corbet, Greg Kroah-Hartman,
Danilo Krummrich, Shuah Khan, Pavel Machek, Len Brown, linux-doc,
linux-kernel, linux-pm, driver-core, senozhatsky, Randy Dunlap
In-Reply-To: <CAJZ5v0ja1ocM+W4hbVsHoZ=fFdmQ-=z4xzviNwo0eEyWT8ssBA@mail.gmail.com>
On Tue, Jun 9, 2026 at 8:10 PM Rafael J. Wysocki <rafael@kernel.org> wrote:
>
> On Tue, Jun 9, 2026 at 11:03 AM Tzung-Bi Shih <tzungbi@kernel.org> wrote:
> >
> > On Mon, Jun 08, 2026 at 04:11:30PM +0200, Rafael J. Wysocki wrote:
> > > On Mon, Jun 8, 2026 at 4:16 AM Tzung-Bi Shih <tzungbi@kernel.org> wrote:
> > > >
> > > > Currently, the module parameters defined in drivers/base/power/main.c
> > > > use the default prefix "main" (derived from the filename). The prefix
> > > > "main" is too generic and non-descriptive for power management
> > > > parameters.
> > > >
> > > > Redefine MODULE_PARAM_PREFIX to "power." at the beginning of the file
> > > > to group the module parameters under the "power" namespace instead.
> > > > This makes the parameters more descriptive.
> > > >
> > > > Signed-off-by: Tzung-Bi Shih <tzungbi@kernel.org>
> > > > ---
> > > > v3:
> > > > - No changes.
> > > >
> > > > v2: https://lore.kernel.org/all/20260604090756.2884671-2-tzungbi@kernel.org
> > > > - New to the series.
> > > >
> > > > v1: Doesn't exist.
> > > >
> > > > drivers/base/power/main.c | 3 +++
> > > > 1 file changed, 3 insertions(+)
> > > >
> > > > diff --git a/drivers/base/power/main.c b/drivers/base/power/main.c
> > > > index ed48c292f575..cd864f3a2799 100644
> > > > --- a/drivers/base/power/main.c
> > > > +++ b/drivers/base/power/main.c
> > > > @@ -40,6 +40,9 @@
> > > > #include "../base.h"
> > > > #include "power.h"
> > > >
> > > > +#undef MODULE_PARAM_PREFIX
> > > > +#define MODULE_PARAM_PREFIX "power."
> > >
> > > "power" may be confused with the power supply support, so I'd rather
> > > use "pm" or even "pm_sleep" (in which case the "dpm_" prefix could be
> > > dropped from the new module param name in the next patch).
> >
> > Ack, will use "pm_sleep" in the next version.
> >
> > Regarding dropping the "dpm_" prefix, should this also apply to the existing
> > dpm_watchdog_all_cpu_backtrace parameter? Or should we leave it as-is to
> > avoid breaking existing configurations?
>
> Breaking things for someone would be unfortunate.
>
> For consistency, let's retain the "dpm_watchdog" part in this name and
> use it in the new one.
Hmm, doesn't the prefix change already break things?
Best,
Tomasz
^ permalink raw reply
* Re: [PATCH v6 06/12] PCI: liveupdate: Auto-preserve upstream bridges across Live Update
From: Pranjal Shrivastava @ 2026-06-09 11:15 UTC (permalink / raw)
To: David Matlack
Cc: kexec, linux-doc, linux-kernel, linux-mm, linux-pci,
Adithya Jayachandran, Alexander Graf, Alex Williamson,
Bjorn Helgaas, Chris Li, David Rientjes, Jacob Pan,
Jason Gunthorpe, Jonathan Corbet, Josh Hilke, Leon Romanovsky,
Lukas Wunner, Mike Rapoport, Parav Pandit, Pasha Tatashin,
Pratyush Yadav, Saeed Mahameed, Samiullah Khawaja, Shuah Khan,
Vipin Sharma, William Tu, Yi Liu
In-Reply-To: <aic1gdmisIT4sqEN@google.com>
On Mon, Jun 08, 2026 at 09:34:57PM +0000, David Matlack wrote:
> On 2026-06-06 10:15 PM, Pranjal Shrivastava wrote:
> > On Fri, May 22, 2026 at 08:24:04PM +0000, David Matlack wrote:
> > > When a PCI device is preserved across a Live Update, all of its upstream
> > > bridges up to the root port must also be preserved. This enables the PCI
> > > core and any drivers bound to the bridges to manage bridges correctly
> > > across a Live Update.
> > >
> > > Notably, this will be used in subsequent commits to ensure that
> > > preserved devices can continue performing memory transactions without a
> > > disruption or change in routing.
> > >
> > > To preserve bridges, the PCI core tracks the number of downstream
> > > devices preserved under each bridge using a reference count in struct
> > > pci_dev_ser. This allows a bridge to remain preserved until all its
> > > downstream preserved devices are unpreserved or finish their
> > > participation in the Live Update.
> > >
> > > Signed-off-by: David Matlack <dmatlack@google.com>
> > > ---
> > > drivers/pci/liveupdate.c | 136 +++++++++++++++++++++++++++++++-----
> > > include/linux/kho/abi/pci.h | 5 +-
> > > 2 files changed, 122 insertions(+), 19 deletions(-)
> > >
> >
> > [...]
> >
> > > +
> > > +#define for_each_pci_dev_in_path(_d, _start, _end) \
> > > + for ((_d) = (_start); (_d) != (_end); (_d) = (_d)->bus->self)
> > > +
> > > +static void __pci_liveupdate_unpreserve_path(struct pci_ser *ser,
> > > + struct pci_dev *start,
> > > + struct pci_dev *end)
> > > +{
> > > + struct pci_dev *dev;
> > > +
> > > + for_each_pci_dev_in_path(dev, start, end) {
> > > + if (pci_liveupdate_unpreserve_device(ser, dev))
> >
> > I might be reading this wrong but are we leaking some upstream devs if
> > an intermediate node fails?
> >
> > EP0
> > /
> > Assume we have: RC -> B1 -> B2
> > \
> > EP1
> >
> > and EP0 & EP1 were preserved successfully.
> >
> > And then we try unpreserving EP1, we follow:
> >
> > unpreserve EP1 -> unpreserve B2 failed due to a corruption.
> >
> > This aborts the loop, skipping B1 and RC completely?
> > Their refcounts remain elevated, effectively leaking them as preserved
> > state permanently? (i.e. if we unpreserve EP0 after this, B1 & RC will
> > still get preserved).
>
> Yes, but that would only happen if there is some sort of kernel bug or
> silent data corruption. I guess we could proceed with trying to
> unpreserve the bridges upstream. But I opted to log a big warning and
> bail immediately.
>
> pci_liveupdate_finish_path() has the same behavior BTW.
Fair point. I agree we are in a broken state if we hit this.
I was originally thinking of a situation where we'd want to keep the
failure localized. For example: unpreserve EP1 fails -> user sees the
warning -> resets EP1 -> retries preserving it later.
But given the recent discussion/decision that retrieve operations
will no longer be retried, I guess there isn't really a use-case for
retrying anything. It makes sense to just bail here.
>
> >
> > > + return;
> > > + }
> > > +}
> > > +
> > > +static void pci_liveupdate_unpreserve_path(struct pci_ser *ser,
> > > + struct pci_dev *start)
> > > +{
> > > + __pci_liveupdate_unpreserve_path(ser, start, /*end=*/NULL);
> > > +}
> > > +
> > > +static int pci_liveupdate_preserve_path(struct pci_ser *ser,
> > > + struct pci_dev *start)
> > > +{
> > > + struct pci_dev *dev;
> > > + int ret;
> > > +
> > > + for_each_pci_dev_in_path(dev, start, NULL) {
> > > + ret = pci_liveupdate_preserve_device(ser, dev);
> > > + if (ret) {
> > > + __pci_liveupdate_unpreserve_path(ser, start, dev);
> > > + return ret;
> > > + }
> > > + }
> > > +
> > > + return 0;
> > > +}
> > > +
> > > /**
> > > * pci_liveupdate_preserve() - Preserve a PCI device across Live Update
> > > * @dev: The PCI device to preserve.
> > > @@ -321,6 +403,9 @@ static int pci_liveupdate_preserve_device(struct pci_ser *ser, struct pci_dev *d
> > > * pci_liveupdate_preserve() from their struct liveupdate_file_handler
> > > * preserve() callback to ensure the outgoing struct pci_ser is already set up.
> > > *
> > > + * pci_liveupdate_preserve() automatically preserves all bridges upstream of
> > > + * @dev.
> > > + *
> > > * Returns: 0 on success, <0 on failure.
> > > */
> > > int pci_liveupdate_preserve(struct pci_dev *dev)
> > > @@ -336,7 +421,7 @@ int pci_liveupdate_preserve(struct pci_dev *dev)
> > > if (IS_ERR(ser))
> > > return PTR_ERR(ser);
> > >
> > > - return pci_liveupdate_preserve_device(ser, dev);
> > > + return pci_liveupdate_preserve_path(ser, dev);
> >
> > Minor nit: I might be too nitpicky here (and it's NOT a strong opinion)
> > but naming it pci_liveupdate_preserve_path_for_dev() reads better to me.
>
> Noted :). I'll keep the current name for now since that is pretty long,
> but if anyone else votes for it I'm happy to be overridden.
Sounds good.
Thanks,
Praan
^ permalink raw reply
* Re: [PATCH v3 1/3] PM: core: Rename module parameters prefix to "power"
From: Rafael J. Wysocki @ 2026-06-09 11:10 UTC (permalink / raw)
To: Tzung-Bi Shih
Cc: Rafael J. Wysocki, Jonathan Corbet, Greg Kroah-Hartman,
Danilo Krummrich, Shuah Khan, Pavel Machek, Len Brown, linux-doc,
linux-kernel, linux-pm, driver-core, tfiga, senozhatsky,
Randy Dunlap
In-Reply-To: <aifWwEyHuFejW4ac@google.com>
On Tue, Jun 9, 2026 at 11:03 AM Tzung-Bi Shih <tzungbi@kernel.org> wrote:
>
> On Mon, Jun 08, 2026 at 04:11:30PM +0200, Rafael J. Wysocki wrote:
> > On Mon, Jun 8, 2026 at 4:16 AM Tzung-Bi Shih <tzungbi@kernel.org> wrote:
> > >
> > > Currently, the module parameters defined in drivers/base/power/main.c
> > > use the default prefix "main" (derived from the filename). The prefix
> > > "main" is too generic and non-descriptive for power management
> > > parameters.
> > >
> > > Redefine MODULE_PARAM_PREFIX to "power." at the beginning of the file
> > > to group the module parameters under the "power" namespace instead.
> > > This makes the parameters more descriptive.
> > >
> > > Signed-off-by: Tzung-Bi Shih <tzungbi@kernel.org>
> > > ---
> > > v3:
> > > - No changes.
> > >
> > > v2: https://lore.kernel.org/all/20260604090756.2884671-2-tzungbi@kernel.org
> > > - New to the series.
> > >
> > > v1: Doesn't exist.
> > >
> > > drivers/base/power/main.c | 3 +++
> > > 1 file changed, 3 insertions(+)
> > >
> > > diff --git a/drivers/base/power/main.c b/drivers/base/power/main.c
> > > index ed48c292f575..cd864f3a2799 100644
> > > --- a/drivers/base/power/main.c
> > > +++ b/drivers/base/power/main.c
> > > @@ -40,6 +40,9 @@
> > > #include "../base.h"
> > > #include "power.h"
> > >
> > > +#undef MODULE_PARAM_PREFIX
> > > +#define MODULE_PARAM_PREFIX "power."
> >
> > "power" may be confused with the power supply support, so I'd rather
> > use "pm" or even "pm_sleep" (in which case the "dpm_" prefix could be
> > dropped from the new module param name in the next patch).
>
> Ack, will use "pm_sleep" in the next version.
>
> Regarding dropping the "dpm_" prefix, should this also apply to the existing
> dpm_watchdog_all_cpu_backtrace parameter? Or should we leave it as-is to
> avoid breaking existing configurations?
Breaking things for someone would be unfortunate.
For consistency, let's retain the "dpm_watchdog" part in this name and
use it in the new one.
^ permalink raw reply
* Re: [PATCH v3 3/3] PM: dpm_watchdog: Add sysctl interface for DPM watchdog timeouts
From: Rafael J. Wysocki @ 2026-06-09 11:08 UTC (permalink / raw)
To: Tzung-Bi Shih
Cc: Rafael J. Wysocki, Jonathan Corbet, Greg Kroah-Hartman,
Danilo Krummrich, Shuah Khan, Pavel Machek, Len Brown, linux-doc,
linux-kernel, linux-pm, driver-core, tfiga, senozhatsky,
Randy Dunlap
In-Reply-To: <aifWjW93SOQiv_Eh@google.com>
On Tue, Jun 9, 2026 at 11:02 AM Tzung-Bi Shih <tzungbi@kernel.org> wrote:
>
> On Mon, Jun 08, 2026 at 04:22:35PM +0200, Rafael J. Wysocki wrote:
> > On Mon, Jun 8, 2026 at 4:16 AM Tzung-Bi Shih <tzungbi@kernel.org> wrote:
> > >
> > > Introduce sysctl knobs to allow configuring DPM watchdog timeouts at
> > > runtime.
> > >
> > > Currently, these timeouts are fixed at compile time via
> > > CONFIG_DPM_WATCHDOG_TIMEOUT and CONFIG_DPM_WATCHDOG_WARNING_TIMEOUT.
> > > This limits flexibility if the timeouts need to be adjusted for
> > > different testing scenarios or hardware behaviors without rebuilding
> > > the kernel.
> > >
> > > Add the following sysctl files under /proc/sys/kernel/:
> > > - dpm_watchdog_timeout_secs: The total timeout before panic. The
> > > maximum value is capped at CONFIG_DPM_WATCHDOG_TIMEOUT to prevent
> > > unreasonably large timeouts.
> > > - dpm_watchdog_warning_timeout_secs: The warning timeout. The maximum
> > > value is capped at the current dpm_watchdog_timeout_secs.
> > > Both sysctls have a minimum value of 1.
> > >
> > > Signed-off-by: Tzung-Bi Shih <tzungbi@kernel.org>
> > > ---
> > > v3:
> > > - No changes.
> > >
> > > v2: https://lore.kernel.org/all/20260604090756.2884671-4-tzungbi@kernel.org
> > > - New to the series.
> > >
> > > v1: Doesn't exist.
> > >
> > > drivers/base/power/main.c | 61 ++++++++++++++++++++++++++++++++++++---
> > > 1 file changed, 57 insertions(+), 4 deletions(-)
> > >
> > > diff --git a/drivers/base/power/main.c b/drivers/base/power/main.c
> > > index 7822c29b7c8d..c1a4b30fafb2 100644
> > > --- a/drivers/base/power/main.c
> > > +++ b/drivers/base/power/main.c
> > > @@ -28,6 +28,7 @@
> > > #include <linux/interrupt.h>
> > > #include <linux/sched.h>
> > > #include <linux/sched/debug.h>
> > > +#include <linux/sysctl.h>
> > > #include <linux/async.h>
> > > #include <linux/suspend.h>
> > > #include <trace/events/power.h>
> > > @@ -539,6 +540,58 @@ static bool __read_mostly dpm_watchdog_enabled =
> > > module_param(dpm_watchdog_enabled, bool, 0644);
> > > MODULE_PARM_DESC(dpm_watchdog_enabled, "Enable DPM watchdog");
> > >
> > > +static unsigned int __read_mostly dpm_watchdog_timeout = CONFIG_DPM_WATCHDOG_TIMEOUT;
> > > +static unsigned int __read_mostly dpm_watchdog_warning_timeout =
> > > + CONFIG_DPM_WATCHDOG_WARNING_TIMEOUT;
> > > +static const unsigned int dpm_watchdog_timeout_max = CONFIG_DPM_WATCHDOG_TIMEOUT;
> > > +
> > > +static int proc_dodpm_watchdog_timeout_secs(const struct ctl_table *table,
> > > + int write, void *buffer,
> > > + size_t *lenp, loff_t *ppos)
> > > +{
> > > + struct ctl_table ctl = *table;
> > > + unsigned int val = dpm_watchdog_timeout;
> > > + int ret;
> > > +
> > > + ctl.data = &val;
> > > + ret = proc_douintvec_minmax(&ctl, write, buffer, lenp, ppos);
> > > + if (ret || !write)
> > > + return ret;
> > > +
> > > + if (val < dpm_watchdog_warning_timeout)
> > > + dpm_watchdog_warning_timeout = val;
> > > + dpm_watchdog_timeout = val;
> > > +
> > > + return 0;
> > > +}
> > > +
> > > +static const struct ctl_table dpm_watchdog_sysctls[] = {
> > > + {
> > > + .procname = "dpm_watchdog_timeout_secs",
> > > + .maxlen = sizeof(unsigned int),
> > > + .mode = 0644,
> > > + .proc_handler = proc_dodpm_watchdog_timeout_secs,
> > > + .extra1 = SYSCTL_ONE,
> > > + .extra2 = (void *)&dpm_watchdog_timeout_max,
> > > + },
> > > + {
> > > + .procname = "dpm_watchdog_warning_timeout_secs",
> > > + .data = &dpm_watchdog_warning_timeout,
> > > + .maxlen = sizeof(unsigned int),
> > > + .mode = 0644,
> > > + .proc_handler = proc_douintvec_minmax,
> > > + .extra1 = SYSCTL_ONE,
> > > + .extra2 = (void *)&dpm_watchdog_timeout,
> > > + },
> > > +};
> > > +
> > > +static int __init dpm_watchdog_sysctl_init(void)
> > > +{
> > > + register_sysctl_init("kernel", dpm_watchdog_sysctls);
> > > + return 0;
> > > +}
> > > +subsys_initcall(dpm_watchdog_sysctl_init);
> > > +
> > > /**
> > > * dpm_watchdog_handler - Driver suspend / resume watchdog handler.
> > > * @t: The timer that PM watchdog depends on.
> > > @@ -564,9 +617,9 @@ static void dpm_watchdog_handler(struct timer_list *t)
> > > dev_driver_string(wd->dev), dev_name(wd->dev));
> > > }
> > >
> > > - time_left = CONFIG_DPM_WATCHDOG_TIMEOUT - CONFIG_DPM_WATCHDOG_WARNING_TIMEOUT;
> > > + time_left = dpm_watchdog_timeout - dpm_watchdog_warning_timeout;
> > > dev_warn(wd->dev, "**** DPM device timeout after %u seconds; %u seconds until panic ****\n",
> > > - CONFIG_DPM_WATCHDOG_WARNING_TIMEOUT, time_left);
> > > + dpm_watchdog_warning_timeout, time_left);
> > > show_stack(wd->tsk, NULL, KERN_WARNING);
> > >
> > > wd->fatal = true;
> > > @@ -587,11 +640,11 @@ static void dpm_watchdog_set(struct dpm_watchdog *wd, struct device *dev)
> > >
> > > wd->dev = dev;
> > > wd->tsk = current;
> > > - wd->fatal = CONFIG_DPM_WATCHDOG_TIMEOUT == CONFIG_DPM_WATCHDOG_WARNING_TIMEOUT;
> > > + wd->fatal = dpm_watchdog_timeout == dpm_watchdog_warning_timeout;
> > >
> > > timer_setup_on_stack(timer, dpm_watchdog_handler, 0);
> > > /* use same timeout value for both suspend and resume */
> > > - timer->expires = jiffies + HZ * CONFIG_DPM_WATCHDOG_WARNING_TIMEOUT;
> > > + timer->expires = jiffies + HZ * dpm_watchdog_warning_timeout;
> > > add_timer(timer);
> > > }
> > >
> > > --
> >
> > I think that this can be applied without the other two patches in the
> > series, so please let me know if you want me to apply it separately.
>
> Ack. The patch does have adjacent hunks with the preceding patch, which
> might cause minor contextual conflicts if applied independently.
That's not a problem.
> Would you want me to reorder the series in the next version to make this
> the first patch? In case the rest two patches may still take some time
> to review.
I'll pick it up as is, so it can make it into 7.2.
^ permalink raw reply
* Re: [PATCH mm-unstable v19 11/14] mm/khugepaged: Introduce mTHP collapse support
From: Lance Yang @ 2026-06-09 11:01 UTC (permalink / raw)
To: Nico Pache
Cc: David Hildenbrand (Arm), linux-doc, linux-kernel, linux-mm,
linux-trace-kernel, aarcange, akpm, anshuman.khandual, apopple,
baohua, baolin.wang, byungchul, catalin.marinas, cl, corbet,
dave.hansen, dev.jain, gourry, hannes, hughd, jack, jackmanb,
jannh, jglisse, joshua.hahnjy, kas, liam, ljs, mathieu.desnoyers,
matthew.brost, mhiramat, mhocko, peterx, pfalcato, rakie.kim,
raquini, rdunlap, richard.weiyang, rientjes, rostedt, rppt,
ryan.roberts, shivankg, sunnanyong, surenb, thomas.hellstrom,
tiwai, usamaarif642, vbabka, vishal.moola, wangkefeng.wang, will,
willy, yang, ying.huang, ziy, zokeefe
In-Reply-To: <CAA1CXcD7WAiA1b9GTLAuNZ+kHaFx0SzZwpBkqAZ=s+RHsTUaow@mail.gmail.com>
On 2026/6/9 18:50, Nico Pache wrote:
> On Tue, Jun 9, 2026 at 4:37 AM Lance Yang <lance.yang@linux.dev> wrote:
>>
>>
>>
>> On 2026/6/9 17:32, Nico Pache wrote:
>>> On Tue, Jun 9, 2026 at 3:26 AM David Hildenbrand (Arm) <david@kernel.org> wrote:
>>>>
>>>> On 6/9/26 11:06, Nico Pache wrote:
>>>>> On Mon, Jun 8, 2026 at 8:57 AM David Hildenbrand (Arm) <david@kernel.org> wrote:
>>>>>>
>>>>>> On 6/6/26 12:28, Lance Yang wrote:
>>>>>>>
>>>>>>>
>>>>>>> Looks broken for swap PTEs in PMD collapse ...
>>>>>>>
>>>>>>> collapse_scan_pmd() allows them up to max_ptes_swap and record them in
>>>>>>> unmapped, but they don't get a bit in mthp_present_ptes. And then
>>>>>>> mthp_collapse() does the check above:
>>>>>>
>>>>>> Right. I assumed this is implicitly handled by the optimization in collapse_scan_pmd:
>>>>>>
>>>>>> if (enabled_orders != BIT(HPAGE_PMD_ORDER))
>>>>>> max_ptes_none = KHUGEPAGED_MAX_PTES_LIMIT;
>>>>>>
>>>>>> But we perform the check a second time.
>>>>>>
>>>>>>>
>>>>>>> nr_occupied_ptes >= nr_ptes - max_ptes_none
>>>>>>>
>>>>>>> So max_ptes_none=0 + 511 present PTEs + one allowed swap PTE won't even
>>>>>>> call collapse_huge_page() for PMD order.
>>>>>>>
>>>>>>> Shouldn't we account for them in the PMD-order check? Something like:
>>>>>>>
>>>>>>> if (is_pmd_order(order))
>>>>>>> nr_occupied_ptes += unmapped;
>>>>>
>>>>> This solution seems good for a temporary fixup. but longterm we may
>>>>> want something else. I'm still not sure how we plan on supporting
>>>>> swapin without causing creep. So I'd be ok with adding a fix for
>>>>> legacy PMD behavior until we know how to handle mTHP creep correctly.
>>>>>
>>>>>> As an alternative, we could either 1) skip the check there for
>>>>>> pmd order (as the check was already done); or 2) introduce+maintain
>>>>>> a bitmap that tracks non-present PTEs.
>>>>>>
>>>>>> @@ -1475,7 +1477,9 @@ static enum scan_result mthp_collapse(struct mm_struct *mm,
>>>>>> nr_occupied_ptes = bitmap_weight_from(cc->mthp_present_ptes, offset,
>>>>>> offset + nr_ptes);
>>>>>>
>>>>>> - if (nr_occupied_ptes >= nr_ptes - max_ptes_none) {
>>>>>> + /* Check was already done in the caller. */
>>>>>> + if (is_pmd_order(order) ||
>>>>>> + nr_occupied_ptes >= nr_ptes - max_ptes_none) {
>>>>>> enum scan_result ret;
>>>>>>
>>>>>> collapse_address = address + offset * PAGE_SIZE;
>>>>>>
>>>>>> 2) would probably be cleanest long-term.
>>>>>
>>>>> That would be best for future swapin support in mTHP, but I still
>>>>> don't think it solves the creep issue.
>>>>
>>>> It wouldn't, we'd simply maintain the state we collect + rely on in separate
>>>> bitmaps. On swapin, we'd have to update/refresh bitmaps I guess.
>>>
>>> Yeah, I'm saying for the future, it obviously solves this issue here
>>> as well, but if we have positional tracking of the swapout, shared,
>>> and none PTEs, I think we can use this to determine whether the
>>> collapse would lead to creep. If we detect creep would happen it may
>>> be best to automatically collapse to the N+1 (or greater) candidate.
>>> Just thinking outloud here.
>>>
>>>>
>>>>> Perhaps we could combine the
>>>>> two bitmaps to determine if it would make the future collapse eligible
>>>>> again? Not sure but ill start thinking about it.
>>>>>
>>>>> Should I send a fixup for this using Lance's solution? Or does Lance
>>>>> want to send a patch out with the fixes tag?
>>>>
>>>> If Lance could send a fixup, explaining the situation, that would be nice.
>>
>> Sure, happy to send a fixup :P
>>
>> Should I send it as a fixup to be folded into this patch, or as a
>> separate patch with a Fixes: tag?
>
> Id assume a seperate patch so you can keep credit for the discovery :)
Okay :D
> Thank you for all the review you provided on this series, its been
> really helpful!
Appreciate it!
Nice work getting it this far. Nice one, Nico :P
^ permalink raw reply
* [PATCH v9 6/6] selftests/mm: add hwpoison-panic destructive test
From: Breno Leitao @ 2026-06-09 10:57 UTC (permalink / raw)
To: Miaohe Lin, Andrew Morton, David Hildenbrand, Lorenzo Stoakes,
Vlastimil Babka, Mike Rapoport, Suren Baghdasaryan, Michal Hocko,
Shuah Khan, Naoya Horiguchi, Jonathan Corbet, Shuah Khan,
Liam R. Howlett, lance.yang, Steven Rostedt, Masami Hiramatsu,
Mathieu Desnoyers, Liam R. Howlett
Cc: linux-mm, linux-kernel, linux-doc, linux-kselftest, Breno Leitao,
linux-trace-kernel, kernel-team
In-Reply-To: <20260609-ecc_panic-v9-0-432a74002e74@debian.org>
Add a destructive selftest that verifies
vm.panic_on_unrecoverable_memory_failure actually panics when a
hwpoison error hits a kernel-owned page.
Three "kinds" of kernel-owned page can be targeted, selectable via
the script's first positional argument (default: rodata):
rodata - a PG_reserved page in the kernel rodata range, sourced
from the "Kernel rodata" sub-resource of "System RAM" in
/proc/iomem. That entry is reported on every major
architecture and guarantees the chosen PFN is backed by
struct page (an online System RAM range, not a firmware
hole), is PG_reserved, and is read-only -- so even if
the panic fails to fire for some reason, the resulting
PG_hwpoison marker on rodata does not corrupt writable
kernel state.
slab - a slab page found by walking /proc/kpageflags for the
first PFN with KPF_SLAB set (and KPF_HWPOISON / KPF_NOPAGE
/ KPF_COMPOUND_TAIL clear). Exercises the get_any_page()
path on a non PG_reserved kernel-owned page and so
catches regressions where get_any_page() collapses
kernel-owned pages into a transient -EIO instead of
-ENOTRECOVERABLE.
pgtable - same as slab, but the PFN is selected via KPF_PGTABLE.
PageLargeKmalloc, the fourth page type matched by
HWPoisonKernelOwned(), is intentionally not covered: it is a
PAGE_TYPE_OPS flag with no /proc/kpageflags bit, so selecting such
a PFN from userspace is not feasible. The slab and pgtable
variants already exercise the same get_any_page() positive-check
branch.
The script enables the sysctl and writes the selected physical
address to /sys/devices/system/memory/hard_offline_page. A
successful run crashes the kernel with
Memory failure: <pfn>: unrecoverable page
A return from the inject means the panic did not fire and the test
fails. Test outcome is therefore observed externally (serial
console, kdump) rather than from the script's own exit code.
The script is intentionally NOT wired into run_vmtests.sh: every
successful run panics the kernel, which is incompatible with the
sequential "run each category in the same VM" model that
run_vmtests.sh assumes. It is also not registered as a TEST_PROGS /
ksft_* wrapper so a default kselftest run does not opt itself into
a panic. The script is meant to be executed manually inside a
disposable VM (e.g. virtme-ng), one variant per VM boot, and
requires RUN_DESTRUCTIVE=1 in the environment as a safety net.
Signed-off-by: Breno Leitao <leitao@debian.org>
---
tools/testing/selftests/mm/Makefile | 4 +
tools/testing/selftests/mm/hwpoison-panic.sh | 208 +++++++++++++++++++++++++++
2 files changed, 212 insertions(+)
diff --git a/tools/testing/selftests/mm/Makefile b/tools/testing/selftests/mm/Makefile
index e6df968f0971..ed321ae709da 100644
--- a/tools/testing/selftests/mm/Makefile
+++ b/tools/testing/selftests/mm/Makefile
@@ -174,6 +174,10 @@ TEST_PROGS += ksft_userfaultfd.sh
TEST_PROGS += ksft_vma_merge.sh
TEST_PROGS += ksft_vmalloc.sh
+# Destructive: every successful run panics the kernel. Installed and
+# kept executable, but not run from a default kselftest invocation.
+TEST_PROGS_EXTENDED += hwpoison-panic.sh
+
TEST_FILES := test_vmalloc.sh
TEST_FILES += test_hmm.sh
TEST_FILES += va_high_addr_switch.sh
diff --git a/tools/testing/selftests/mm/hwpoison-panic.sh b/tools/testing/selftests/mm/hwpoison-panic.sh
new file mode 100755
index 000000000000..fe58e7638a8b
--- /dev/null
+++ b/tools/testing/selftests/mm/hwpoison-panic.sh
@@ -0,0 +1,208 @@
+#!/bin/bash
+# SPDX-License-Identifier: GPL-2.0
+#
+# Verify vm.panic_on_unrecoverable_memory_failure by injecting a hwpoison
+# error on a kernel-owned page and confirming the kernel panics.
+#
+# Three "kinds" of kernel-owned page can be targeted, selectable via the
+# first positional argument (default: rodata):
+#
+# rodata - a PG_reserved page in the kernel rodata range
+# (sourced from /proc/iomem "Kernel rodata"). Exercises
+# memory_failure() -> get_any_page() on a PageReserved page.
+#
+# slab - a slab page found via /proc/kpageflags (KPF_SLAB).
+# Exercises memory_failure() -> get_any_page() on a non
+# PG_reserved kernel-owned page. This path is what catches
+# regressions where get_any_page() collapses kernel-owned
+# pages into a transient -EIO instead of -ENOTRECOVERABLE.
+#
+# pgtable - a page-table page found via /proc/kpageflags (KPF_PGTABLE).
+# Same path as slab, different page type.
+#
+# This test is DESTRUCTIVE: a successful run crashes the kernel. It is
+# meant to be executed inside a disposable VM (e.g. virtme-ng) with a
+# serial console captured by the harness. It is skipped unless the
+# caller opts in via RUN_DESTRUCTIVE=1.
+#
+# Test passes externally: the kernel must panic with
+# "Memory failure: <pfn>: unrecoverable page"
+# A return from the inject means the panic did not fire and the test
+# fails.
+#
+# Author: Breno Leitao <leitao@debian.org>
+
+set -u
+
+ksft_skip=4
+sysctl_path=/proc/sys/vm/panic_on_unrecoverable_memory_failure
+inject_path=/sys/devices/system/memory/hard_offline_page
+kpageflags_path=/proc/kpageflags
+
+# /proc/kpageflags bit positions (see include/uapi/linux/kernel-page-flags.h)
+KPF_SLAB=7
+KPF_COMPOUND_TAIL=16
+KPF_HWPOISON=19
+KPF_NOPAGE=20
+KPF_PGTABLE=26
+
+kind=${1:-rodata}
+
+ksft_print() { echo "# $*"; }
+ksft_exit_skip() { ksft_print "$*"; exit "$ksft_skip"; }
+ksft_exit_fail() { echo "not ok 1 $*"; exit 1; }
+
+if [ "$(id -u)" -ne 0 ]; then
+ ksft_exit_skip "must run as root"
+fi
+
+if [ ! -w "$sysctl_path" ]; then
+ ksft_exit_skip "$sysctl_path not present (kernel without the sysctl?)"
+fi
+
+if [ ! -w "$inject_path" ]; then
+ ksft_exit_skip "$inject_path not present (no MEMORY_HOTPLUG?)"
+fi
+
+if [ "${RUN_DESTRUCTIVE:-0}" != "1" ]; then
+ ksft_exit_skip "destructive test; re-run with RUN_DESTRUCTIVE=1 inside a disposable VM"
+fi
+
+# Pick a PFN inside the kernel image rodata region of /proc/iomem.
+# This is preferred over a top-level "Reserved" entry because top-level
+# Reserved ranges are often firmware holes that have no backing struct
+# page; pfn_to_online_page() returns NULL on those and memory_failure()
+# bails out with -ENXIO before reaching the panic path.
+#
+# "Kernel rodata" is reported as a sub-resource of "System RAM" on every
+# major architecture, which guarantees:
+# - the PFN is backed by struct page (within an online memory range);
+# - PG_reserved is set on the page (kernel image area);
+# - the memory is read-only, so setting PG_hwpoison on it does not
+# corrupt writable kernel state if the panic somehow does not fire.
+#
+# /proc/iomem entries look like (indented for sub-resources):
+# " 02500000-02ffffff : Kernel rodata"
+pick_rodata_phys_addr() {
+ awk -v pagesize="$(getconf PAGE_SIZE)" '
+ # Convert a hex string to a number without relying on the gawk-only
+ # strtonum(). mawk lacks it and would otherwise spuriously skip
+ # this test on distros that ship mawk as /usr/bin/awk.
+ function hex2num(s, n, i, c, v) {
+ n = 0
+ for (i = 1; i <= length(s); i++) {
+ c = tolower(substr(s, i, 1))
+ v = index("0123456789abcdef", c) - 1
+ if (v < 0)
+ return -1
+ n = n * 16 + v
+ }
+ return n
+ }
+ /: Kernel rodata[[:space:]]*$/ {
+ sub(/^[[:space:]]+/, "")
+ n = split($0, a, /[- ]/)
+ start = hex2num(a[1])
+ end = hex2num(a[2])
+ if (end <= start)
+ next
+ # Page-align upward and emit the first byte of that page.
+ pfn = int((start + pagesize - 1) / pagesize)
+ printf "0x%x\n", pfn * pagesize
+ exit 0
+ }
+ ' /proc/iomem
+}
+
+# Walk /proc/kpageflags and return the phys addr of the first PFN that
+# has bit $1 set, with KPF_HWPOISON, KPF_NOPAGE and KPF_COMPOUND_TAIL
+# all clear (so we attack a real, non-tail, not-already-poisoned page).
+#
+# We skip the first 16 MiB of PFNs to step past low-memory special
+# ranges (BIOS/EFI/ACPI/etc.) that often are PG_reserved and would not
+# exhibit the slab/pgtable type we are looking for.
+pick_kpageflags_phys_addr() {
+ local want_bit=$1
+ local pagesize skip_pfn
+
+ [ -r "$kpageflags_path" ] || return
+
+ pagesize=$(getconf PAGE_SIZE)
+ skip_pfn=$(((16 * 1024 * 1024) / pagesize))
+
+ od -An -tx8 -v -w8 -j "$((skip_pfn * 8))" "$kpageflags_path" 2>/dev/null | \
+ awk -v want_bit="$want_bit" \
+ -v hwp_bit="$KPF_HWPOISON" \
+ -v nopage_bit="$KPF_NOPAGE" \
+ -v tail_bit="$KPF_COMPOUND_TAIL" \
+ -v base_pfn="$skip_pfn" \
+ -v pagesize="$pagesize" '
+ # Test whether bit "b" is set in the 16-hex-digit value "hex".
+ # Done with substring + per-digit lookup so we never rely on awk
+ # bitwise operators (mawk lacks them), 64-bit FP precision or the
+ # gawk-only strtonum().
+ function bit_set(hex, b, di, bi, c, v) {
+ di = int(b / 4)
+ bi = b - di * 4
+ c = substr(hex, length(hex) - di, 1)
+ v = index("0123456789abcdef", tolower(c)) - 1
+ if (bi == 0) return (v % 2) == 1
+ if (bi == 1) return int(v / 2) % 2 == 1
+ if (bi == 2) return int(v / 4) % 2 == 1
+ return int(v / 8) % 2 == 1
+ }
+ {
+ gsub(/^[[:space:]]+/, "")
+ h = $1
+ if (bit_set(h, want_bit) &&
+ !bit_set(h, hwp_bit) &&
+ !bit_set(h, nopage_bit) &&
+ !bit_set(h, tail_bit)) {
+ pfn = base_pfn + NR - 1
+ printf "0x%x\n", pfn * pagesize
+ exit 0
+ }
+ }
+ '
+}
+
+case "$kind" in
+rodata)
+ phys_addr=$(pick_rodata_phys_addr)
+ missing_msg='no "Kernel rodata" entry in /proc/iomem'
+ ;;
+slab)
+ phys_addr=$(pick_kpageflags_phys_addr "$KPF_SLAB")
+ missing_msg="no usable slab PFN found in $kpageflags_path"
+ ;;
+pgtable)
+ phys_addr=$(pick_kpageflags_phys_addr "$KPF_PGTABLE")
+ missing_msg="no usable page-table PFN found in $kpageflags_path"
+ ;;
+*)
+ ksft_exit_fail "unknown kind '$kind' (expected: rodata|slab|pgtable)"
+ ;;
+esac
+
+if [ -z "$phys_addr" ]; then
+ ksft_exit_skip "$missing_msg"
+fi
+
+ksft_print "enabling $sysctl_path"
+prior=$(cat "$sysctl_path")
+echo 1 > "$sysctl_path" || ksft_exit_fail "failed to enable sysctl"
+
+ksft_print "injecting hwpoison at phys 0x$(printf '%x' "$phys_addr") (kind=$kind)"
+ksft_print "expecting kernel panic: 'Memory failure: <pfn>: unrecoverable page'"
+
+# If this returns, the kernel did not panic → test failed. Restore the
+# sysctl before reporting so the system is left as we found it.
+if echo "$phys_addr" > "$inject_path"; then
+ echo "$prior" > "$sysctl_path"
+ ksft_exit_fail "inject returned without panic; sysctl ineffective"
+fi
+
+# Write failed (e.g. -EINVAL on offlining a non-online region): also a
+# failure for this test, since we expected the panic path.
+echo "$prior" > "$sysctl_path"
+ksft_exit_fail "inject failed before reaching the panic path"
--
2.53.0-Meta
^ permalink raw reply related
* [PATCH v9 5/6] Documentation: document panic_on_unrecoverable_memory_failure sysctl
From: Breno Leitao @ 2026-06-09 10:56 UTC (permalink / raw)
To: Miaohe Lin, Andrew Morton, David Hildenbrand, Lorenzo Stoakes,
Vlastimil Babka, Mike Rapoport, Suren Baghdasaryan, Michal Hocko,
Shuah Khan, Naoya Horiguchi, Jonathan Corbet, Shuah Khan,
Liam R. Howlett, lance.yang, Steven Rostedt, Masami Hiramatsu,
Mathieu Desnoyers, Liam R. Howlett
Cc: linux-mm, linux-kernel, linux-doc, linux-kselftest, Breno Leitao,
linux-trace-kernel, kernel-team
In-Reply-To: <20260609-ecc_panic-v9-0-432a74002e74@debian.org>
Add documentation for the new vm.panic_on_unrecoverable_memory_failure
sysctl, describing which failures trigger a panic (kernel-owned pages
the handler cannot recover) and which are intentionally left out
(transient allocator races and unclassified pages).
Signed-off-by: Breno Leitao <leitao@debian.org>
---
Documentation/admin-guide/sysctl/vm.rst | 85 +++++++++++++++++++++++++++++++++
1 file changed, 85 insertions(+)
diff --git a/Documentation/admin-guide/sysctl/vm.rst b/Documentation/admin-guide/sysctl/vm.rst
index 97e12359775c..f71d87039904 100644
--- a/Documentation/admin-guide/sysctl/vm.rst
+++ b/Documentation/admin-guide/sysctl/vm.rst
@@ -67,6 +67,7 @@ Currently, these files are in /proc/sys/vm:
- page-cluster
- page_lock_unfairness
- panic_on_oom
+- panic_on_unrecoverable_memory_failure
- percpu_pagelist_high_fraction
- stat_interval
- stat_refresh
@@ -925,6 +926,90 @@ panic_on_oom=2+kdump gives you very strong tool to investigate
why oom happens. You can get snapshot.
+panic_on_unrecoverable_memory_failure
+======================================
+
+When a hardware memory error (e.g. multi-bit ECC) hits a kernel page
+that cannot be recovered by the memory failure handler, the default
+behaviour is to ignore the error and continue operation. This is
+dangerous because the corrupted data remains accessible to the kernel,
+risking silent data corruption or a delayed crash when the poisoned
+memory is next accessed.
+
+When enabled, this sysctl triggers a panic on memory failure events
+hitting kernel-owned pages that the handler cannot recover:
+``PageReserved`` (firmware reservations, kernel image, vDSO, zero
+page, and similar memblock-reserved regions), ``PageSlab``,
+``PageTable``, and ``PageLargeKmalloc``. These are owned by the
+kernel and the memory failure handler cannot reliably evict their
+contents.
+
+For soft offline (``madvise(MADV_SOFT_OFFLINE)``,
+``/sys/devices/system/memory/soft_offline_page``), pages owned by
+``movable_ops`` are exempted, since soft offline is allowed to
+migrate them even though they are not on the LRU.
+
+Other unrecoverable kernel-owned populations (vmalloc allocations,
+kernel stack pages, ...) are not currently covered because the
+handler has no page-type signal that distinguishes them from a
+userspace folio temporarily off the LRU during migration or
+compaction. Such pages still go through the standard
+MF_MSG_GET_HWPOISON path: ``PG_hwpoison`` is set on them and a
+delayed crash on the next access remains possible. Coverage may
+grow as the handler gains stronger kernel-ownership signals.
+
+Recoverable failure paths are also intentionally left out: in-flight
+buddy allocations and other transient races with the page allocator
+can reach the same diagnostic, and panicking on them would risk
+killing the box for a page destined for userspace where the standard
+SIGBUS recovery path applies. Pages whose state could not be
+classified at all are not covered either, since an unknown state is
+not a sound basis for a panic decision.
+
+For many environments it is preferable to panic immediately with a clean
+crash dump that captures the original error context, rather than to
+continue and face a random crash later whose cause is difficult to
+diagnose.
+
+Use cases
+---------
+
+This option is most useful in environments where unattributed crashes
+are expensive to debug or where data integrity must take precedence
+over availability:
+
+* Large fleets, where multi-bit ECC errors on kernel pages are observed
+ regularly and post-mortem analysis of an unrelated downstream crash
+ (often seconds to minutes after the original error) consumes
+ significant engineering effort.
+
+* Systems configured with kdump, where panicking at the moment of the
+ hardware error produces a vmcore that still contains the faulting
+ address, the affected page state, and the originating MCE/GHES
+ record — context that is typically lost by the time a delayed crash
+ occurs.
+
+* High-availability clusters that rely on fast, deterministic node
+ failure for failover, and prefer an immediate panic over silent data
+ corruption propagating to replicas or persistent storage.
+
+* Kernel and platform developers reproducing hwpoison issues with
+ tools such as ``mce-inject`` or error-injection debugfs interfaces,
+ where panicking on the unrecoverable path makes regressions
+ immediately visible instead of surfacing as later, unrelated
+ failures.
+
+= =====================================================================
+0 Try to continue operation (default).
+1 Panic immediately. If the ``panic`` sysctl is also non-zero then the
+ machine will be rebooted.
+= =====================================================================
+
+Example::
+
+ echo 1 > /proc/sys/vm/panic_on_unrecoverable_memory_failure
+
+
percpu_pagelist_high_fraction
=============================
--
2.53.0-Meta
^ permalink raw reply related
* [PATCH v9 4/6] mm/memory-failure: add panic option for unrecoverable pages
From: Breno Leitao @ 2026-06-09 10:56 UTC (permalink / raw)
To: Miaohe Lin, Andrew Morton, David Hildenbrand, Lorenzo Stoakes,
Vlastimil Babka, Mike Rapoport, Suren Baghdasaryan, Michal Hocko,
Shuah Khan, Naoya Horiguchi, Jonathan Corbet, Shuah Khan,
Liam R. Howlett, lance.yang, Steven Rostedt, Masami Hiramatsu,
Mathieu Desnoyers, Liam R. Howlett
Cc: linux-mm, linux-kernel, linux-doc, linux-kselftest, Breno Leitao,
linux-trace-kernel, kernel-team
In-Reply-To: <20260609-ecc_panic-v9-0-432a74002e74@debian.org>
Add a sysctl panic_on_unrecoverable_memory_failure (disabled by
default) that triggers a kernel panic when memory_failure()
encounters pages that cannot be recovered. This provides a clean
crash with useful debug information rather than allowing silent
data corruption or a delayed crash at an unrelated code path.
Panic eligibility is intentionally narrow: only MF_MSG_KERNEL with
result == MF_IGNORED panics. After the previous patch, MF_MSG_KERNEL
covers PG_reserved pages and the kernel-owned pages promoted from
get_hwpoison_page() via -ENOTRECOVERABLE (slab, page tables,
large-kmalloc).
All other action types are excluded:
- MF_MSG_GET_HWPOISON and MF_MSG_KERNEL_HIGH_ORDER can be reached by
transient refcount races with the page allocator (an in-flight buddy
allocation has refcount 0 and is no longer on the buddy free list,
briefly), and panicking on them would risk killing the box for what
is actually a recoverable userspace page.
- MF_MSG_UNKNOWN means identify_page_state() could not classify the
page; that is precisely the wrong basis for a panic decision.
Signed-off-by: Breno Leitao <leitao@debian.org>
---
mm/memory-failure.c | 23 +++++++++++++++++++++++
1 file changed, 23 insertions(+)
diff --git a/mm/memory-failure.c b/mm/memory-failure.c
index 35f2b5d89fbe..a8b466a48b02 100644
--- a/mm/memory-failure.c
+++ b/mm/memory-failure.c
@@ -74,6 +74,8 @@ static int sysctl_memory_failure_recovery __read_mostly = 1;
static int sysctl_enable_soft_offline __read_mostly = 1;
+static int sysctl_panic_on_unrecoverable_mf __read_mostly;
+
atomic_long_t num_poisoned_pages __read_mostly = ATOMIC_LONG_INIT(0);
static bool hw_memory_failure __read_mostly = false;
@@ -155,6 +157,15 @@ static const struct ctl_table memory_failure_table[] = {
.proc_handler = proc_dointvec_minmax,
.extra1 = SYSCTL_ZERO,
.extra2 = SYSCTL_ONE,
+ },
+ {
+ .procname = "panic_on_unrecoverable_memory_failure",
+ .data = &sysctl_panic_on_unrecoverable_mf,
+ .maxlen = sizeof(sysctl_panic_on_unrecoverable_mf),
+ .mode = 0644,
+ .proc_handler = proc_dointvec_minmax,
+ .extra1 = SYSCTL_ZERO,
+ .extra2 = SYSCTL_ONE,
}
};
@@ -1255,6 +1266,15 @@ static void update_per_node_mf_stats(unsigned long pfn,
++mf_stats->total;
}
+static bool panic_on_unrecoverable_mf(enum mf_action_page_type type,
+ enum mf_result result)
+{
+ if (!sysctl_panic_on_unrecoverable_mf)
+ return false;
+
+ return type == MF_MSG_KERNEL && result == MF_IGNORED;
+}
+
/*
* "Dirty/Clean" indication is not 100% accurate due to the possibility of
* setting PG_dirty outside page lock. See also comment above set_page_dirty().
@@ -1272,6 +1292,9 @@ static int action_result(unsigned long pfn, enum mf_action_page_type type,
pr_err("%#lx: recovery action for %s: %s\n",
pfn, action_page_types[type], action_name[result]);
+ if (panic_on_unrecoverable_mf(type, result))
+ panic("Memory failure: %#lx: unrecoverable page", pfn);
+
return (result == MF_RECOVERED || result == MF_DELAYED) ? 0 : -EBUSY;
}
--
2.53.0-Meta
^ permalink raw reply related
* [PATCH v9 3/6] mm/memory-failure: report MF_MSG_KERNEL for unrecoverable kernel pages
From: Breno Leitao @ 2026-06-09 10:56 UTC (permalink / raw)
To: Miaohe Lin, Andrew Morton, David Hildenbrand, Lorenzo Stoakes,
Vlastimil Babka, Mike Rapoport, Suren Baghdasaryan, Michal Hocko,
Shuah Khan, Naoya Horiguchi, Jonathan Corbet, Shuah Khan,
Liam R. Howlett, lance.yang, Steven Rostedt, Masami Hiramatsu,
Mathieu Desnoyers, Liam R. Howlett
Cc: linux-mm, linux-kernel, linux-doc, linux-kselftest, Breno Leitao,
linux-trace-kernel, kernel-team
In-Reply-To: <20260609-ecc_panic-v9-0-432a74002e74@debian.org>
The previous patch teaches get_any_page() to return -ENOTRECOVERABLE
for stable unhandlable kernel pages (PG_reserved, slab, page tables,
large-kmalloc). memory_failure() still folds every negative return
into MF_MSG_GET_HWPOISON, so callers that want to react to the
unrecoverable cases (a panic option, smarter logging) cannot tell
them apart from transient page-allocator races.
Turn the post-call branch into a switch over the get_hwpoison_page()
return code: map -ENOTRECOVERABLE to MF_MSG_KERNEL and any other
negative return to MF_MSG_GET_HWPOISON. case 0 keeps the existing
free-buddy / kernel-high-order handling and case 1 falls through to
the rest of memory_failure() unchanged.
The MF_MSG_KERNEL label and tracepoint string are kept as
"reserved kernel page" to avoid breaking userspace tools that match
on those literals; the enum value still adequately tags the failure
even though it now also covers slab, page tables and large-kmalloc
pages.
Suggested-by: David Hildenbrand <david@kernel.org>
Acked-by: David Hildenbrand (Arm) <david@kernel.org>
Acked-by: Miaohe Lin <linmiaohe@huawei.com>
Signed-off-by: Breno Leitao <leitao@debian.org>
---
mm/memory-failure.c | 17 +++++++++++++++--
1 file changed, 15 insertions(+), 2 deletions(-)
diff --git a/mm/memory-failure.c b/mm/memory-failure.c
index eed9de387694..35f2b5d89fbe 100644
--- a/mm/memory-failure.c
+++ b/mm/memory-failure.c
@@ -2444,7 +2444,8 @@ int memory_failure(unsigned long pfn, int flags)
* that may make page_ref_freeze()/page_ref_unfreeze() mismatch.
*/
res = get_hwpoison_page(p, flags);
- if (!res) {
+ switch (res) {
+ case 0:
if (is_free_buddy_page(p)) {
if (take_page_off_buddy(p)) {
page_ref_inc(p);
@@ -2463,7 +2464,19 @@ int memory_failure(unsigned long pfn, int flags)
res = action_result(pfn, MF_MSG_KERNEL_HIGH_ORDER, MF_IGNORED);
}
goto unlock_mutex;
- } else if (res < 0) {
+ case 1:
+ /* Got a refcount on a handlable page. */
+ break;
+ case -ENOTRECOVERABLE:
+ /*
+ * Stable unhandlable kernel-owned page (PG_reserved,
+ * slab, page tables, large-kmalloc).
+ * No recovery possible.
+ */
+ res = action_result(pfn, MF_MSG_KERNEL, MF_IGNORED);
+ goto unlock_mutex;
+ default:
+ /* Transient lifecycle race with the page allocator. */
res = action_result(pfn, MF_MSG_GET_HWPOISON, MF_IGNORED);
goto unlock_mutex;
}
--
2.53.0-Meta
^ permalink raw reply related
* [PATCH v9 2/6] mm/memory-failure: surface unhandlable kernel pages as -ENOTRECOVERABLE
From: Breno Leitao @ 2026-06-09 10:56 UTC (permalink / raw)
To: Miaohe Lin, Andrew Morton, David Hildenbrand, Lorenzo Stoakes,
Vlastimil Babka, Mike Rapoport, Suren Baghdasaryan, Michal Hocko,
Shuah Khan, Naoya Horiguchi, Jonathan Corbet, Shuah Khan,
Liam R. Howlett, lance.yang, Steven Rostedt, Masami Hiramatsu,
Mathieu Desnoyers, Liam R. Howlett
Cc: linux-mm, linux-kernel, linux-doc, linux-kselftest, Breno Leitao,
linux-trace-kernel, kernel-team
In-Reply-To: <20260609-ecc_panic-v9-0-432a74002e74@debian.org>
get_any_page() collapses every HWPoisonHandlable() rejection into a
single -EIO via the __get_hwpoison_page() -> -EBUSY -> shake_page()
-> retry path. That is correct for the transient case (a userspace
folio briefly off LRU during migration or compaction, which a later
shake can drag back), but wrong for stable kernel-owned pages: slab,
page-table, large-kmalloc and PG_reserved pages will never become
HWPoisonHandlable(), so the retry loop is wasted work and the final
-EIO loses the "this is structurally unrecoverable" information.
memory_failure() then maps -EIO into MF_MSG_GET_HWPOISON, which the
panic-on-unrecoverable sysctl deliberately does not act on.
Introduce HWPoisonKernelOwned(), a small predicate that positively
identifies pages the hwpoison handler cannot recover from:
HWPoisonKernelOwned(p, flags) :=
!(MF_SOFT_OFFLINE && page_has_movable_ops(p)) &&
(PageReserved(p) ||
PageSlab(head) || PageTable(head) || PageLargeKmalloc(head))
where head = compound_head(p).
PG_reserved is a per-page flag (PF_NO_COMPOUND) and is tested on the
page directly. The slab, page-table and large-kmalloc page-type bits
are only stored on the head page, so those tests resolve the compound
head first, then re-read compound_head(page) afterwards: a concurrent
split or compound free that moves head invalidates the just-read flags
and the loop retries. The lookup still takes no refcount, mirroring
the rest of get_any_page(); the recheck closes the common split race,
and a residual free->alloc->free in the same window can only mis-tag
a genuinely poisoned page, never reclassify a handlable one.
The MF_SOFT_OFFLINE / page_has_movable_ops() opt-out mirrors the
same exception in HWPoisonHandlable(): soft-offline is allowed to
migrate movable_ops pages even though they are not on the LRU, and
we must not pre-empt that with an unrecoverable verdict.
The list is intentionally not exhaustive. vmalloc and kernel-stack
pages, for example, do not carry a page_type bit and would need a
different oracle; they keep going through the existing retry path
unchanged. This is the smallest set we can identify with certainty
by page type.
Wire the helper into the top of get_any_page() to short-circuit
those pages before the retry loop runs. On a hit, drop the caller's
MF_COUNT_INCREASED reference (if any) and return -ENOTRECOVERABLE
straight away. Pages outside the helper's positive list still take
the existing retry path and return -EIO, leaving operator-visible
behaviour for those cases unchanged.
Extend the unhandlable-page pr_err() to fire for either errno and
update the get_hwpoison_page() kerneldoc to document the new return.
memory_failure() still folds every negative return into
MF_MSG_GET_HWPOISON via its existing "else if (res < 0)" branch, so
this patch on its own only changes the errno that soft_offline_page()
can propagate to its callers. A follow-up wires -ENOTRECOVERABLE
through memory_failure() and reports MF_MSG_KERNEL for the
unrecoverable cases, which is what the
panic_on_unrecoverable_memory_failure sysctl observes.
Suggested-by: David Hildenbrand <david@kernel.org>
Suggested-by: Lance Yang <lance.yang@linux.dev>
Signed-off-by: Breno Leitao <leitao@debian.org>
---
mm/memory-failure.c | 60 +++++++++++++++++++++++++++++++++++++++++++++++++++--
1 file changed, 58 insertions(+), 2 deletions(-)
diff --git a/mm/memory-failure.c b/mm/memory-failure.c
index f4d3e6e20e13..eed9de387694 100644
--- a/mm/memory-failure.c
+++ b/mm/memory-failure.c
@@ -1325,6 +1325,46 @@ static inline bool HWPoisonHandlable(struct page *page, unsigned long flags)
return PageLRU(page) || is_free_buddy_page(page);
}
+/*
+ * Positive identification of pages the hwpoison handler cannot recover.
+ * These page types are owned by kernel internals (no userspace mapping
+ * to unmap, no file mapping to invalidate, no migration target), so the
+ * shake_page() / retry loop in get_any_page() can never turn them into
+ * something HWPoisonHandlable() will accept. Short-circuit them to
+ * -ENOTRECOVERABLE so callers can panic on operator request instead of
+ * spinning through retries that exit as a transient-looking -EIO.
+ *
+ * The MF_SOFT_OFFLINE / page_has_movable_ops() opt-out mirrors
+ * HWPoisonHandlable(): soft-offline is allowed to migrate movable_ops
+ * pages even though they are not on the LRU.
+ */
+static inline bool HWPoisonKernelOwned(struct page *page, unsigned long flags)
+{
+ struct page *head;
+
+ if ((flags & MF_SOFT_OFFLINE) && page_has_movable_ops(page))
+ return false;
+
+ /* PG_reserved is a per-page flag, never set on a compound page. */
+ if (PageReserved(page))
+ return true;
+
+ /*
+ * Page-type bits live only on the head page, so resolve any tail
+ * first. The check takes no refcount; recheck the head afterwards
+ * so a concurrent split or compound free cannot leave us trusting
+ * a stale view. A free->alloc->free in the same window is still
+ * possible but closing it would require taking a reference here.
+ */
+retry:
+ head = compound_head(page);
+ if (!(PageSlab(head) || PageTable(head) || PageLargeKmalloc(head)))
+ return false;
+ if (head != compound_head(page))
+ goto retry;
+ return true;
+}
+
static int __get_hwpoison_page(struct page *page, unsigned long flags)
{
struct folio *folio = page_folio(page);
@@ -1371,6 +1411,19 @@ static int get_any_page(struct page *p, unsigned long flags)
if (flags & MF_COUNT_INCREASED)
count_increased = true;
+ /*
+ * Page types we know are kernel-owned and cannot be recovered.
+ * Short-circuit before the shake_page() / retry loop, which
+ * cannot turn any of these into something HWPoisonHandlable().
+ * Drop the caller's reference if MF_COUNT_INCREASED took one.
+ */
+ if (HWPoisonKernelOwned(p, flags)) {
+ if (count_increased)
+ put_page(p);
+ ret = -ENOTRECOVERABLE;
+ goto out;
+ }
+
try_again:
if (!count_increased) {
ret = __get_hwpoison_page(p, flags);
@@ -1418,7 +1471,7 @@ static int get_any_page(struct page *p, unsigned long flags)
ret = -EIO;
}
out:
- if (ret == -EIO)
+ if (ret == -EIO || ret == -ENOTRECOVERABLE)
pr_err("%#lx: unhandlable page.\n", page_to_pfn(p));
return ret;
@@ -1475,7 +1528,10 @@ static int __get_unpoison_page(struct page *page)
* -EIO for pages on which we can not handle memory errors,
* -EBUSY when get_hwpoison_page() has raced with page lifecycle
* operations like allocation and free,
- * -EHWPOISON when the page is hwpoisoned and taken off from buddy.
+ * -EHWPOISON when the page is hwpoisoned and taken off from buddy,
+ * -ENOTRECOVERABLE for kernel-owned pages identified by
+ * HWPoisonKernelOwned() (PG_reserved, slab,
+ * page-table, large-kmalloc) that the handler cannot recover.
*/
static int get_hwpoison_page(struct page *p, unsigned long flags)
{
--
2.53.0-Meta
^ permalink raw reply related
page: next (older) | prev (newer) | latest
- recent:[subjects (threaded)|topics (new)|topics (active)]
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox