Re: [PATCH] drm/edid: Use struct cea_db when parsing HDMI VSDB

["Ville Syrjälä" <ville.syrjala@linux.intel.com>]

On Tue, Jan 20, 2026 at 03:01:14PM -0500, jpeisach@ubuntu.com wrote:
> 
> 
> > On Jan 20, 2026, at 1:09 PM, Ville Syrjälä <ville.syrjala@linux.intel.com> wrote:
> > 
> > On Tue, Jan 20, 2026 at 12:46:59PM -0500, jpeisach@ubuntu.com <mailto:jpeisach@ubuntu.com> wrote:
> >> 
> >> 
> >>> On Jan 20, 2026, at 12:03 PM, Ville Syrjälä <ville.syrjala@linux.intel.com <mailto:ville.syrjala@linux.intel.com>> wrote:
> >>> 
> >>> On Tue, Jan 20, 2026 at 11:37:59AM -0500, jpeisach@ubuntu.com <mailto:jpeisach@ubuntu.com> <mailto:jpeisach@ubuntu.com> wrote:
> >>>> 
> >>>> 
> >>>>> On Jan 20, 2026, at 10:41 AM, Ville Syrjälä <ville.syrjala@linux.intel.com> wrote:
> >>>>> 
> >>>>> On Mon, Jan 19, 2026 at 03:39:12PM +0200, Jani Nikula wrote:
> >>>>>> On Sat, 17 Jan 2026, Joshua Peisach <jpeisach@ubuntu.com> wrote:
> >>>>>>> drm_parse_hdmi_vsdb_video is one of the parsers that still do not use the
> >>>>>>> cea_db struct, and currently passes a u8 pointer.
> >>>>>>> 
> >>>>>>> Set the correct struct type and update references to the data accordingly.
> >>>>>>> This also makes the same change to drm_parse_hdmi_deep_color_info as
> >>>>>>> necessary.
> >>>>>>> 
> >>>>>>> Signed-off-by: Joshua Peisach <jpeisach@ubuntu.com>
> >>>>>>> ---
> >>>>>>> drivers/gpu/drm/drm_edid.c | 26 +++++++++++++-------------
> >>>>>>> 1 file changed, 13 insertions(+), 13 deletions(-)
> >>>>>>> 
> >>>>>>> diff --git a/drivers/gpu/drm/drm_edid.c b/drivers/gpu/drm/drm_edid.c
> >>>>>>> index 26bb7710a..15bd99e65 100644
> >>>>>>> --- a/drivers/gpu/drm/drm_edid.c
> >>>>>>> +++ b/drivers/gpu/drm/drm_edid.c
> >>>>>>> @@ -6290,7 +6290,7 @@ static void drm_parse_hdmi_forum_scds(struct drm_connector *connector,
> >>>>>>> }
> >>>>>>> 
> >>>>>>> static void drm_parse_hdmi_deep_color_info(struct drm_connector *connector,
> >>>>>>> -					   const u8 *hdmi)
> >>>>>>> +					   const struct cea_db *db)
> >>>>>>> {
> >>>>>>> 	struct drm_display_info *info = &connector->display_info;
> >>>>>>> 	unsigned int dc_bpc = 0;
> >>>>>>> @@ -6298,24 +6298,24 @@ static void drm_parse_hdmi_deep_color_info(struct drm_connector *connector,
> >>>>>>> 	/* HDMI supports at least 8 bpc */
> >>>>>>> 	info->bpc = 8;
> >>>>>>> 
> >>>>>>> -	if (cea_db_payload_len(hdmi) < 6)
> >>>>>>> +	if (cea_db_payload_len(db) < 6)
> >>>>>>> 		return;
> >>>>>>> 
> >>>>>>> -	if (hdmi[6] & DRM_EDID_HDMI_DC_30) {
> >>>>>>> +	if (db->data[6] & DRM_EDID_HDMI_DC_30) {
> >>>>>> 
> >>>>>> That's not the same thing, but off-by-one now. Ditto everywhere that
> >>>>>> changes from u8* to db->data[].
> >>>>>> 
> >>>>>> The main problem with the change (even with fixed offsets) is that the
> >>>>>> *specs* typically use indexing from the beginning of the data block, not
> >>>>>> from the beginning of payload data.
> >>>>>> 
> >>>>>> We've discussed this before with Ville (Cc'd) but I'm not sure if we
> >>>>>> reached a conclusion.
> >>>>> 
> >>>>> I guess we could give up on the index matching the spec byte#.
> >>>>> Looks like the HDMI VSDB parsing is the only place where we
> >>>>> actually have the two match, and everwhere else it's
> >>>>> already inconsistent.
> >>>>> 
> >>>>> Also maybe we should add something to also exclude the
> >>>>> extended tag from the payload, for the blocks that use
> >>>>> the extended tag...
> >>>>> 
> >>>>> -- 
> >>>>> Ville Syrjälä
> >>>>> Intel
> >>>> 
> >>>> I personally believe it is important to match the spec for consistency
> >>>> and reference. Unless I am looking at the wrong thing, bit 6 should have
> >>>> the correct index.
> >>>> 
> >>>> Also I think cea_db in the context of the function calls here are
> >>>> just the data blocks. Unless you mean that by having the struct's first
> >>>> member being the tag_length if offsetting everything, but I don't think
> >>>> it is? Let me know if I'm wrong :)
> >>> 
> >>> The tag+length byte is:
> >>> byte# 1 in the CTA spec
> >>> byte# 0 in the HDMI spec
> >>> 
> >>> There's no super nice way to use the byte# as the index for both.
> >>> Also the length checks end up looking somewhat confusing when
> >>> comparing them with the corresponding index.
> >>> 
> >>> -- 
> >>> Ville Syrjälä
> >>> Intel
> >> 
> >> The name of the functions specifically say HDMI, so I think that's the
> >> system to use: there are functions that say CTA in the name, like
> >> parse_cta_y420cmdb - so that is CTA, and these functions follow HDMI.
> > 
> > I'm saying that there is no really sane way to deal with the CTA byte#
> > convention. So I think it's probably best to just go for a single
> > consistent approach for both CTA and HDMI. That way people at least
> > won't get confused by the different convetion between the functions.
> > And the length checks wouldn't look incorrect.
> > 
> 
> I agree. I can't think of anything though, other than to assume CTA.
> 
> The other parsers that already use struct cea_db refer to the data
> using the CTA spec (starting at 1). Maybe just go with that?

Nothing is indexed using the CTA byte#. To do that we'd have to
do something like 'u8 *db = start_of_db - 1'.

I can more or less see these variants:
1. relative to the start of the data block
   (eg. drm_parse_hdr_metadata_block(), drm_parse_vcdb())
2. relative to the start of the payload
   (eg. parse_cta_vdb())
3. relative to the first byte past the extended tag
   (y420* stuff)

And I'm suggesting that perhaps everything should use either
2 or 3 depending on whether the extended tag is present.

-- 
Ville Syrjälä
Intel