Re: [PATCH 1/1] drm/vc4: hdmi: Replace drm_detect_hdmi_monitor() with is_hdmi

[José Expósito]

Hi Maxime,

Thanks for your comments.

On Fri, Apr 08, 2022 at 09:41:10AM +0200, Maxime Ripard wrote:
> Hi Jose,
> 
> On Wed, Apr 06, 2022 at 06:55:14PM +0200, José Expósito wrote:
> > Once EDID is parsed, the monitor HDMI support information is cached in
> > drm_display_info.is_hdmi by drm_parse_hdmi_vsdb_video().
> > 
> > This driver calls drm_detect_hdmi_monitor() to receive the same
> > information and stores its own cached value, which is less efficient.
> > 
> > Avoid calling drm_detect_hdmi_monitor() and use drm_display_info.is_hdmi
> > instead.
> > 
> > drm_detect_hdmi_monitor() is called in vc4_hdmi_connector_detect() and
> > vc4_hdmi_connector_get_modes(). In both cases it is safe to rely on
> > drm_display_info.is_hdmi as shown by ftrace:
> 
> How do you use ftrace to generate that?

I had to add noinline to a couple of relevant functions and run the
usual:

$ sudo trace-cmd record -p function_graph -l "vc4_hdmi_*" [...]

I'll add the command to v2.

 
> > vc4_hdmi_connector_detect:
> > 
> >     vc4_hdmi_connector_detect() {
> >       drm_get_edid() {
> >         drm_connector_update_edid_property() {
> >           drm_add_display_info() {
> >             drm_reset_display_info();
> >             drm_for_each_detailed_block.part.0();
> >             drm_parse_cea_ext() {
> >               drm_find_cea_extension();
> >               cea_db_offsets.part.0();
> >               cea_db_is_hdmi_vsdb.part.0();
> >               drm_parse_hdmi_vsdb_video();
> >               /* drm_display_info.is_hdmi is cached here */
> >             }
> >           }
> >         }
> >       }
> >       /* drm_display_info.is_hdmi is used here */
> >     }
> > 
> > vc4_hdmi_connector_get_modes:
> > 
> >     vc4_hdmi_connector_get_modes() {
> >       drm_get_edid() {
> >         drm_connector_update_edid_property() {
> >           drm_add_display_info() {
> >             drm_reset_display_info();
> >             drm_for_each_detailed_block.part.0();
> >             drm_parse_cea_ext() {
> >               drm_find_cea_extension();
> >               cea_db_offsets.part.0();
> >               cea_db_is_hdmi_vsdb.part.0();
> >               drm_parse_hdmi_vsdb_video();
> >               /* drm_display_info.is_hdmi is cached here */
> >             }
> >           }
> >         }
> >       }
> >       /* drm_display_info.is_hdmi is used here */
> >       drm_connector_update_edid_property();
> >     }
> > 
> > Signed-off-by: José Expósito <jose.exposito89@xxxxxxxxx>
> 
> I think what you're hinting at in the cover letter would be best though:
> we should just remove that caching entirely and use is_hdmi everywhere

Cool, I'll work on a follow up patch to remove vc4_encoder.hdmi_monitor
and add it to v2.

Thanks,
Jose