[PATCH v17 4/8] drm: bridge: Cadence: Add MHDP8501 DP/HDMI driver

[Dmitry Baryshkov]

On Mon, Sep 30, 2024 at 10:18:06AM +0200, Maxime Ripard wrote:
> On Sun, Sep 29, 2024 at 02:34:36AM GMT, Sandor Yu wrote:
> > > > +static void cdns_hdmi_sink_config(struct cdns_mhdp8501_device *mhdp)
> > > > +{
> > > > +	struct drm_display_info *display = &mhdp->curr_conn->display_info;
> > > > +	struct drm_connector_state *conn_state = mhdp->curr_conn->state;
> > > 
> > > That looks a bit hackish to me. We should probably provide a helper to get the
> > > connector state the bridge is attached to.
> > 
> > How about code change to followed, is it more clear?
> > 370         struct drm_connector *connector = mhdp->curr_conn;
> > 371         struct drm_connector_state *conn_state = connector->state;
> > 372         struct drm_display_info *display = &connector->display_info;
> > 373         struct drm_scdc *scdc = &display->hdmi.scdc;
> 
> What I meant was that I wish bridges had a way to get their connector
> pointer. It doesn't look like it's possible with drm_bridge_connector,
> and we don't have access to drm_display_info anymore.
> 
> I don't really see a good way to do this yet, so maybe that kind of
> workaround is ok. Eventually, I guess we'll have the scrambler setup in
> the HDMI connector helpers anyway.
> 
> Dmitry, any idea?

Unfortunately nothing significant, I didn't have time to look at the
scrambler yet. The platforms that I'm working with either do not support
HDMI 2.0 or require no additional setup there.

Regarding the drm_connector_state, I had to go via the bridge->encoder,
then drm_atomic_get_new_connector_for_encoder() and finally
drm_atomic_get_new_connector_state(). I don't like the idea of storing
the connector in drm_bridge driver data, it seems like a bad idea to me.

> 
> > > > +static enum drm_mode_status
> > > > +cdns_hdmi_tmds_char_rate_valid(const struct drm_bridge *bridge,
> > > > +			       const struct drm_display_mode *mode,
> > > > +			       unsigned long long tmds_rate) {
> > > > +	struct cdns_mhdp8501_device *mhdp = bridge->driver_private;
> > > > +	union phy_configure_opts phy_cfg;
> > > > +	int ret;
> > > > +
> > > > +	phy_cfg.hdmi.tmds_char_rate = tmds_rate;
> > > > +
> > > > +	ret = phy_validate(mhdp->phy, PHY_MODE_HDMI, 0, &phy_cfg);
> > > > +	if (ret < 0)
> > > > +		return MODE_CLOCK_RANGE;
> > > > +
> > > > +	return MODE_OK;
> > > > +}
> > > > +
> > > > +static enum drm_mode_status
> > > > +cdns_hdmi_bridge_mode_valid(struct drm_bridge *bridge,
> > > > +			    const struct drm_display_info *info,
> > > > +			    const struct drm_display_mode *mode) {
> > > > +	unsigned long long tmds_rate;
> > > > +
> > > > +	/* We don't support double-clocked and Interlaced modes */
> > > > +	if (mode->flags & DRM_MODE_FLAG_DBLCLK ||
> > > > +	    mode->flags & DRM_MODE_FLAG_INTERLACE)
> > > > +		return MODE_BAD;
> > > > +
> > > > +	if (mode->hdisplay > 3840)
> > > > +		return MODE_BAD_HVALUE;
> > > > +
> > > > +	if (mode->vdisplay > 2160)
> > > > +		return MODE_BAD_VVALUE;
> > > > +
> > > > +	tmds_rate = mode->clock * 1000ULL;
> > > > +	return cdns_hdmi_tmds_char_rate_valid(bridge, mode, tmds_rate); }
> > > 
> > > Didn't we agree on creating a mode_valid helper?
> > 
> > In fact, now I'm no idea where should add the mode_valid helper function.
> > 
> > In struct drm_bridge_funcs, it had mode_valid() and hdmi_tmds_char_rate_valid().
> > 
> > If create a new mode_valid helper function in struct drm_connector_hdmi_funcs,
> > Is it appropriate to call another API function(tmds_char_rate_valid)
> > at the same level within this API function?
> 
> I'm not quite sure what you mean, but a reasonable approach to me would
> be to turn drm_hdmi_state_helper.c hdmi_clock_valid into a public
> function, and then call it from drm_bridge_connector mode_valid hook.
> 
> It's a similar discussion to the previous one really: in order to
> implement it properly, we need access to drm_display_info.

I've sent a proposal, [1]. I don't think we should be using
hdmi_clock_valid() directly (at least not before sorting out the EDID /
hotplug handling in the HDMI Connector code) exactly because of the
info->max_tmds_clock. If it gets stale, we might filter modes
incorrectly. Also, I'm not sure if it should be left at 0 by default (or
in drm_parse_hdmi_forum_scds()).

[1] https://lore.kernel.org/dri-devel/20241018-hdmi-mode-valid-v1-0-6e49ae4801f7@linaro.org/

-- 
With best wishes
Dmitry