On 5/20/2026 9:30 PM, Jani Nikula wrote:
> On Wed, 20 May 2026, Chenyu Chen <[email protected]> wrote:
>> Parse the Display Parameters Data Block (tag 0x21) defined in
>> DisplayID v2.1a Section 4.2.6. Extract the Display Device Technology
>> field from payload byte 27 bits [6:4], which indicates whether the
>> panel uses LCD (001b) or OLED (010b) technology.
>>
>> Add a did_panel_type field to struct drm_display_info and populate it
>> during DisplayID iteration so downstream drivers can use it for
>> panel-type-dependent behavior. Add DRM_MODE_PANEL_TYPE_LCD to the
>> UAPI constants for use as an internal communication value between
>> DRM core and drivers.
>>
>> Assisted-by: Copilot:Claude-Opus-4.6
>> Signed-off-by: Chenyu Chen <[email protected]>
>> Reviewed-by: Mario Limonciello (AMD) <[email protected]>
>> ---
>> drivers/gpu/drm/drm_displayid_internal.h | 25 ++++++++++++++
>> drivers/gpu/drm/drm_edid.c | 44 ++++++++++++++++++++++++
>> include/drm/drm_connector.h | 6 ++++
>> include/uapi/drm/drm_mode.h | 1 +
>> 4 files changed, 76 insertions(+)
>>
>> diff --git a/drivers/gpu/drm/drm_displayid_internal.h
>> b/drivers/gpu/drm/drm_displayid_internal.h
>> index 5b1b32f73516..e0f7c54d2987 100644
>> --- a/drivers/gpu/drm/drm_displayid_internal.h
>> +++ b/drivers/gpu/drm/drm_displayid_internal.h
>> @@ -142,6 +142,31 @@ struct displayid_formula_timing_block {
>> struct displayid_formula_timings_9 timings[];
>> } __packed;
>>
>> +/*
>> + * DisplayID v2.x Display Parameters Data Block (tag 0x21).
>> + *
>> + * Per VESA DisplayID v2.1a, Section 4.2.6, Table 4-14:
>> + * Offset 0x1E (payload byte 27) contains Native Color Depth and
>> + * Display Device Technology fields.
>> + * bits [2:0] = Native Color Depth
>> + * bit [3] = RESERVED
>> + * bits [6:4] = Display Device Technology
>> + * 000b = not specified, 001b = LCD, 010b = OLED, others reserved
>> + * bit [7] = Display Device Theme Preference
>> + */
>
> Please drop the comment...
>
>> +#define DISPLAYID_DISPLAY_PARAMS_DEVICE_TECH GENMASK(6, 4)
>> +
>> +struct displayid_display_params_block {
>> + struct displayid_block base;
>> + u8 payload[27];
>> + u8 device_tech_byte; /* bits [6:4] = Display Device Technology */
>> + u8 reserved;
>
> ...and actually define the information here.
>
> Sure, you're only interested in one thing, but don't leave the rest for
> someone who comes after you, since you clearly have access to the spec,
> and whoever reviews this also needs to have access to the spec.
>
I will replace the block comment and the u8 payload[27] with fully named
fields per spec Table 4-7 (image size, pixel count, features, primary
colors, white point, luminance, color_depth_and_tech, gamma_eotf).
I also add the named defines DISPLAYID_DEVICE_TECH_{UNSPECIFIED,LCD,OLED}
for the technology values.
>> +} __packed;
>> +
>> +#define DISPLAYID_DISPLAY_PARAMS_MIN_LEN \
>> + (sizeof(struct displayid_display_params_block) - \
>> + sizeof(struct displayid_block))
>
> Is this helpful? Do you suggest we should add the define for all of the
> blocks? What about when there's a minimum, and you can optionally have
> more?
>
I think defining a macro for every block doesn’t make sense.
So I will remove the macro and use sizeof(*params) - sizeof(params->base)
inline instead.
>> +
>> #define DISPLAYID_VESA_MSO_OVERLAP GENMASK(3, 0)
>> #define DISPLAYID_VESA_MSO_MODE GENMASK(6, 5)
>>
>> diff --git a/drivers/gpu/drm/drm_edid.c b/drivers/gpu/drm/drm_edid.c
>> index 04878478ab78..d1de1a398677 100644
>> --- a/drivers/gpu/drm/drm_edid.c
>> +++ b/drivers/gpu/drm/drm_edid.c
>> @@ -6713,6 +6713,8 @@ static void drm_reset_display_info(struct
>> drm_connector *connector)
>>
>> info->source_physical_address = CEC_PHYS_ADDR_INVALID;
>> memset(&info->amd_vsdb, 0, sizeof(info->amd_vsdb));
>> +
>> + info->did_panel_type = DRM_MODE_PANEL_TYPE_UNKNOWN;
>> }
>>
>> static void drm_displayid_process_section_header(struct drm_connector
>> *connector,
>> @@ -6731,6 +6733,44 @@ static void
>> drm_displayid_process_section_header(struct drm_connector *connector
>> info->non_desktop = true;
>> }
>>
>> +static void
>> +drm_displayid_parse_display_params(struct drm_connector *connector,
>> + const struct displayid_block *block)
>> +{
>> + struct drm_display_info *info = &connector->display_info;
>> + const struct displayid_display_params_block *params =
>> + (const struct displayid_display_params_block *)block;
>> +
>
> Superfluous blank line.
>
Will fix it.
>> + u8 tech;
>> +
>> + if (block->num_bytes < DISPLAYID_DISPLAY_PARAMS_MIN_LEN) {
>> + drm_dbg_kms(connector->dev,
>> + "[CONNECTOR:%d:%s] DisplayID Display Parameters
>> block too short (%u < %zu)\n",
>> + connector->base.id, connector->name,
>> + block->num_bytes,
>> + DISPLAYID_DISPLAY_PARAMS_MIN_LEN);
>> + return;
>> + }
>> +
>> + tech = FIELD_GET(DISPLAYID_DISPLAY_PARAMS_DEVICE_TECH,
>> + params->device_tech_byte);
>> +
>> + drm_dbg_kms(connector->dev,
>> + "[CONNECTOR:%d:%s] DisplayID Display Parameters: device
>> technology %u\n",
>> + connector->base.id, connector->name, tech);
>
> It would be more useful to print LCD or OLED, not the number.
>
I will change it to print "LCD", "OLED", or "unspecified".
>> +
>> + switch (tech) {
>> + case 1: /* LCD */
>> + info->did_panel_type = DRM_MODE_PANEL_TYPE_LCD;
>> + break;
>> + case 2: /* OLED */
>
> The comments above are useless.
>
I will remove the inline comments and use the named defines instead of
magic numbers in the switch cases.
>> + info->did_panel_type = DRM_MODE_PANEL_TYPE_OLED;
>> + break;
>> + default:
>> + break;
>> + }
>> +}
>> +
>> static void update_displayid_info(struct drm_connector *connector,
>> const struct drm_edid *drm_edid)
>> {
>> @@ -6744,6 +6784,10 @@ static void update_displayid_info(struct
>> drm_connector *connector,
>> drm_displayid_process_section_header(connector, &iter);
>> header_processed = true;
>> }
>> +
>> + if (displayid_version(&iter) == DISPLAY_ID_STRUCTURE_VER_20 &&
>> + block->tag == DATA_BLOCK_2_DISPLAY_PARAMETERS)
>> + drm_displayid_parse_display_params(connector, block);
>> }
>> displayid_iter_end(&iter);
>> }
>> diff --git a/include/drm/drm_connector.h b/include/drm/drm_connector.h
>> index c398dbc68bbc..b95aec34ddb7 100644
>> --- a/include/drm/drm_connector.h
>> +++ b/include/drm/drm_connector.h
>> @@ -899,6 +899,12 @@ struct drm_display_info {
>> * @amd_vsdb: AMD-specific VSDB information.
>> */
>> struct drm_amd_vsdb_info amd_vsdb;
>> +
>> + /**
>> + * @did_panel_type: Panel type from DisplayID Display Parameters
>> + * Data Block (tag 0x21). Uses DRM_MODE_PANEL_TYPE_* constants.
>> + */
>> + u8 did_panel_type;
>
> What does the did_ prefix give us?
>
Good point, it doesn't add much value.
The intent was to indicate that the value is derived from DisplayID,
but I agree that encoding the data source in the field name is not
particularly helpful.
I'll rename it to "panel_type" and treat it as a generic field,
currently populated from DisplayID if available.
>> };
>>
>> int drm_display_info_set_bus_formats(struct drm_display_info *info,
>> diff --git a/include/uapi/drm/drm_mode.h b/include/uapi/drm/drm_mode.h
>> index 3693d82b5279..d7ca1040b92e 100644
>> --- a/include/uapi/drm/drm_mode.h
>> +++ b/include/uapi/drm/drm_mode.h
>> @@ -169,6 +169,7 @@ extern "C" {
>> /* Panel type property */
>> #define DRM_MODE_PANEL_TYPE_UNKNOWN 0
>> #define DRM_MODE_PANEL_TYPE_OLED 1
>> +#define DRM_MODE_PANEL_TYPE_LCD 2
>
> This is a UABI header.
>
> Should we add this to the panel type property too?
>
>
> BR,
> Jani.
>
>
Yes. I think we can expose the panel type too.
Regards,
Chenyu
>>
>> /*
>> * DRM_MODE_ROTATE_<degrees>
>
