[Intel-gfx] [PATCH v6 3/3] drm/i915: Add support for integrated privacy screens

[Rajat Jain]

Thanks Jani so much for the detailed explanation.

I was able to write the code for this, but I am facing one problem, see below.

On Fri, Mar 6, 2020 at 2:15 AM Jani Nikula <jani.nikula at linux.intel.com> wrote:
>
> On Thu, 05 Mar 2020, Rajat Jain <rajatja at google.com> wrote:
> > OK, will do. In order to do that I may need to introduce driver level
> > hooks that i915 driver can populate and drm core can call (or may be
> > some functions to add privacy screen property that drm core exports
> > and i915 driver will call).
>
> The latter. Look at drm_connector_attach_*() functions in
> drm_connector.c. i915 (or any other driver) can create and attach the
> property as needed. drm_atomic_connector_{get,set}_property in
> drm_atomic_uapi.c need to handle the properties, but *only* to get/set
> the value in drm_connector_state, nothing more. How that value is
> actually used is up to the drivers, but the userspace interface will be
> the same instead of being driver specific.

Understood, done.

>
> >> > @@ -93,15 +97,18 @@ int intel_digital_connector_atomic_set_property(struct drm_connector *connector,
> >> >       struct drm_i915_private *dev_priv = to_i915(dev);
> >> >       struct intel_digital_connector_state *intel_conn_state =
> >> >               to_intel_digital_connector_state(state);
> >> > +     struct intel_connector *intel_connector = to_intel_connector(connector);
> >> >
> >> >       if (property == dev_priv->force_audio_property) {
> >> >               intel_conn_state->force_audio = val;
> >> >               return 0;
> >> > -     }
> >> > -
> >> > -     if (property == dev_priv->broadcast_rgb_property) {
> >> > +     } else if (property == dev_priv->broadcast_rgb_property) {
> >> >               intel_conn_state->broadcast_rgb = val;
> >> >               return 0;
> >> > +     } else if (property == intel_connector->privacy_screen_property) {
> >> > +             intel_privacy_screen_set_val(intel_connector, val);
> >>
> >> I think this part should only change the connector state. The driver
> >> would then do the magic at commit stage according to the property value.
>
> Also, this would be the part that's done in drm core level.
>

Yup.

> > Can you please point me to some code reference as to where in code
> > does the "commit stage" apply the changes?
>
> Look at, say, drm_connector_attach_scaling_mode_property(). In the
> getter/setter code it'll just read/change state->scaling_mode. You can
> use the value in encoder ->enable, ->disable, and ->update_pipe
> hooks. Enable should enable the privacy screen if desired, disable
> should probably unconditionally disable the privacy screen while
> disabling the display, and update should just change the state according
> to the value. Update is called if there isn't a full modeset. (Scaling
> mode is a bit more indirect than that, affecting other things in the
> encoder ->compute_config hook, leading to similar effects.)

For my testing purposes, I'm testing this using the proptest utility
in our distribution (I think from
https://github.com/CPFL/drm/blob/master/tests/proptest/proptest.c). I
notice that when I change the value of the property from userspace,
even though the drm_connector_state->privacy_screen_status gets
updated and reflects the change, the encoder->update_pipe() is not
getting called. Just wanted to ask if this is expected since you seem
to imply this update_pipe() might *not* get called if there *is* a
full modeset? (What is the hook that gets called for a full modeset
where i915 driver should commit this property change to the hardware?)

Thanks & Best Regards,

Rajat


>
> Ville, anything I missed?
>
> BR,
> Jani.
>
>
> --
> Jani Nikula, Intel Open Source Graphics Center