drm/bridge: it6505: Add caching for EDID

  Add caching when EDID is read, and invalidate the cache until the
bridge detects HPD low or sink count changes on HPD_IRQ.

It takes 1.2s for IT6505 bridge to read a 3-block EDID, and skipping
one EDID read would be a notable difference on user experience.

Signed-off-by: Pin-yen Lin <treapking@chromium.org>

---

 drivers/gpu/drm/bridge/ite-it6505.c | 26 ++++++++++++++++++++------
 1 file changed, 20 insertions(+), 6 deletions(-)
  

On Tue, 15 Nov 2022 at 12:27, Pin-yen Lin <treapking@chromium.org> wrote:
>
> Add caching when EDID is read, and invalidate the cache until the
> bridge detects HPD low or sink count changes on HPD_IRQ.
>
> It takes 1.2s for IT6505 bridge to read a 3-block EDID, and skipping
> one EDID read would be a notable difference on user experience.
>
> Signed-off-by: Pin-yen Lin <treapking@chromium.org>
>
> ---
>
>  drivers/gpu/drm/bridge/ite-it6505.c | 26 ++++++++++++++++++++------
>  1 file changed, 20 insertions(+), 6 deletions(-)
>
> diff --git a/drivers/gpu/drm/bridge/ite-it6505.c b/drivers/gpu/drm/bridge/ite-it6505.c
> index 21a9b8422bda..4b818f31668f 100644
> --- a/drivers/gpu/drm/bridge/ite-it6505.c
> +++ b/drivers/gpu/drm/bridge/ite-it6505.c
> @@ -457,6 +457,8 @@ struct it6505 {
>
>         /* it6505 driver hold option */
>         bool enable_drv_hold;
> +
> +       struct edid *cached_edid;
>  };
>
>  struct it6505_step_train_para {
> @@ -2244,6 +2246,13 @@ static void it6505_plugged_status_to_codec(struct it6505 *it6505)
>                                    status == connector_status_connected);
>  }
>
> +
> +static void it6505_remove_edid(struct it6505 *it6505)
> +{
> +       kfree(it6505->cached_edid);
> +       it6505->cached_edid = NULL;
> +}
> +
>  static int it6505_process_hpd_irq(struct it6505 *it6505)
>  {
>         struct device *dev = &it6505->client->dev;
> @@ -2270,6 +2279,7 @@ static int it6505_process_hpd_irq(struct it6505 *it6505)
>                 it6505_reset_logic(it6505);
>                 it6505_int_mask_enable(it6505);
>                 it6505_init(it6505);
> +               it6505_remove_edid(it6505);
>                 return 0;
>         }
>
> @@ -2353,6 +2363,7 @@ static void it6505_irq_hpd(struct it6505 *it6505)
>                         it6505_video_reset(it6505);
>         } else {
>                 memset(it6505->dpcd, 0, sizeof(it6505->dpcd));
> +               it6505_remove_edid(it6505);
>
>                 if (it6505->hdcp_desired)
>                         it6505_stop_hdcp(it6505);
> @@ -3016,16 +3027,18 @@ static struct edid *it6505_bridge_get_edid(struct drm_bridge *bridge,
>  {
>         struct it6505 *it6505 = bridge_to_it6505(bridge);
>         struct device *dev = &it6505->client->dev;
> -       struct edid *edid;
>
> -       edid = drm_do_get_edid(connector, it6505_get_edid_block, it6505);
> +       if (!it6505->cached_edid) {
> +               it6505->cached_edid = drm_do_get_edid(connector, it6505_get_edid_block,
> +                                                     it6505);
>
> -       if (!edid) {
> -               DRM_DEV_DEBUG_DRIVER(dev, "failed to get edid!");
> -               return NULL;
> +               if (!it6505->cached_edid) {
> +                       DRM_DEV_DEBUG_DRIVER(dev, "failed to get edid!");
> +                       return NULL;
> +               }
>         }
>
> -       return edid;
> +       return drm_edid_duplicate(it6505->cached_edid);
>  }
>
>  static const struct drm_bridge_funcs it6505_bridge_funcs = {
> @@ -3367,6 +3380,7 @@ static void it6505_i2c_remove(struct i2c_client *client)
>         drm_dp_aux_unregister(&it6505->aux);
>         it6505_debugfs_remove(it6505);
>         it6505_poweroff(it6505);
> +       it6505_remove_edid(it6505);
>  }
>
>  static const struct i2c_device_id it6505_id[] = {
> --
> 2.38.1.493.g58b659f92b-goog
>

Reviewed-by: Robert Foss <robert.foss@linaro.org>
  

On Tue, 15 Nov 2022 19:27:20 +0800, Pin-yen Lin wrote:
> Add caching when EDID is read, and invalidate the cache until the
> bridge detects HPD low or sink count changes on HPD_IRQ.
> 
> It takes 1.2s for IT6505 bridge to read a 3-block EDID, and skipping
> one EDID read would be a notable difference on user experience.
> 
> 
> [...]

Applied, thanks!

Repo: https://cgit.freedesktop.org/drm/drm-misc/


[1/1] drm/bridge: it6505: Add caching for EDID
      (no commit info)



rob
  

Hi Robert,

Thanks for the review.

I didn't see this patch on drm-misc-next, but my another patch merged
instead: https://cgit.freedesktop.org/drm/drm-misc/commit/?id=5eb9a4314053bda7642643f70f49a2b415920812

Did something go wrong? Or is it me missing something?

Regards,
Pin-yen

On Thu, Dec 15, 2022 at 12:51 AM Robert Foss <robert.foss@linaro.org> wrote:
>
> On Tue, 15 Nov 2022 19:27:20 +0800, Pin-yen Lin wrote:
> > Add caching when EDID is read, and invalidate the cache until the
> > bridge detects HPD low or sink count changes on HPD_IRQ.
> >
> > It takes 1.2s for IT6505 bridge to read a 3-block EDID, and skipping
> > one EDID read would be a notable difference on user experience.
> >
> >
> > [...]
>
> Applied, thanks!
>
> Repo: https://cgit.freedesktop.org/drm/drm-misc/
>
>
> [1/1] drm/bridge: it6505: Add caching for EDID
>       (no commit info)
>
>
>
> rob
>
  

On Thu, 15 Dec 2022 at 09:36, Pin-yen Lin <treapking@chromium.org> wrote:
>
> Hi Robert,
>
> Thanks for the review.
>
> I didn't see this patch on drm-misc-next, but my another patch merged
> instead: https://cgit.freedesktop.org/drm/drm-misc/commit/?id=5eb9a4314053bda7642643f70f49a2b415920812
>
> Did something go wrong? Or is it me missing something?

This is my mistake. I'm switching to a 'b4 ty' based workflow, and the
wrong thank you message got sent out.

Let me have a look at this patch too.

>
> Regards,
> Pin-yen
>
> On Thu, Dec 15, 2022 at 12:51 AM Robert Foss <robert.foss@linaro.org> wrote:
> >
> > On Tue, 15 Nov 2022 19:27:20 +0800, Pin-yen Lin wrote:
> > > Add caching when EDID is read, and invalidate the cache until the
> > > bridge detects HPD low or sink count changes on HPD_IRQ.
> > >
> > > It takes 1.2s for IT6505 bridge to read a 3-block EDID, and skipping
> > > one EDID read would be a notable difference on user experience.
> > >
> > >
> > > [...]
> >
> > Applied, thanks!
> >
> > Repo: https://cgit.freedesktop.org/drm/drm-misc/
> >
> >
> > [1/1] drm/bridge: it6505: Add caching for EDID
> >       (no commit info)
> >
> >
> >
> > rob
> >
  

On Tue, 15 Nov 2022 19:27:20 +0800, Pin-yen Lin wrote:
> Add caching when EDID is read, and invalidate the cache until the
> bridge detects HPD low or sink count changes on HPD_IRQ.
> 
> It takes 1.2s for IT6505 bridge to read a 3-block EDID, and skipping
> one EDID read would be a notable difference on user experience.
> 
> 
> [...]

Applied, thanks!

Repo: https://cgit.freedesktop.org/drm/drm-misc/


[1/1] drm/bridge: it6505: Add caching for EDID
      commit: 11feaef69d0cb81278294299bbfd86f94c2004e3



Rob