[Intel-gfx] [PATCH 1/3] drm/i915/dp: Make sure all tiled connectors get added to the state with full modeset

[Ville Syrjälä]

On Mon, Dec 16, 2019 at 11:13:09AM -0800, Manasi Navare wrote:
> On Mon, Dec 16, 2019 at 04:37:38PM +0200, Ville Syrjälä wrote:
> > On Wed, Dec 11, 2019 at 01:14:23PM -0800, Manasi Navare wrote:
> > > In case of tiled displays, all the tiles are linke dto each other
> > > for transcoder port sync. So in intel_atomic_check() we need to make
> > > sure that we add all the tiles to the modeset and if one of the
> > > tiles needs a full modeset then mark all other tiles for a full modeset.
> > > 
> > > Suggested-by: Ville Syrjälä <ville.syrjala at linux.intel.com>
> > > Cc: Ville Syrjälä <ville.syrjala at linux.intel.com>
> > > Cc: José Roberto de Souza <jose.souza at intel.com>
> > > Bugzilla: https://gitlab.freedesktop.org/drm/intel/issues/5
> > > Signed-off-by: Manasi Navare <manasi.d.navare at intel.com>
> > > ---
> > >  drivers/gpu/drm/i915/display/intel_display.c | 78 ++++++++++++++++++++
> > >  1 file changed, 78 insertions(+)
> > > 
> > > diff --git a/drivers/gpu/drm/i915/display/intel_display.c b/drivers/gpu/drm/i915/display/intel_display.c
> > > index 803993a01ca7..7263eaa66cda 100644
> > > --- a/drivers/gpu/drm/i915/display/intel_display.c
> > > +++ b/drivers/gpu/drm/i915/display/intel_display.c
> > > @@ -14066,6 +14066,80 @@ static int intel_atomic_check_crtcs(struct intel_atomic_state *state)
> > >  	return 0;
> > >  }
> > >  
> > > +static int
> > > +intel_dp_modeset_all_tiles(struct drm_i915_private *dev_priv,
> > > +			   struct intel_atomic_state *state, int tile_grp_id)
> > > +{
> > > +	struct drm_connector *conn_iter;
> > > +	struct drm_connector_list_iter conn_list_iter;
> > > +	struct drm_crtc_state *crtc_state;
> > > +
> > > +	drm_connector_list_iter_begin(&dev_priv->drm, &conn_list_iter);
> > > +	drm_for_each_connector_iter(conn_iter, &conn_list_iter) {
> > > +		struct drm_connector_state *conn_iter_state;
> > > +
> > > +		if (!conn_iter->has_tile)
> > > +			continue;
> > > +		conn_iter_state = drm_atomic_get_connector_state(&state->base,
> > > +								 conn_iter);
> > > +		if (IS_ERR(conn_iter_state)) {
> > > +			drm_connector_list_iter_end(&conn_list_iter);
> > > +			return PTR_ERR(conn_iter_state);
> > > +		}
> > > +
> > > +		if (!conn_iter_state->crtc)
> > > +			continue;
> > > +
> > > +		if (conn_iter->tile_group->id != tile_grp_id)
> > > +			continue;
> > > +
> > > +		crtc_state = drm_atomic_get_crtc_state(&state->base, conn_iter_state->crtc);
> > > +		if (IS_ERR(crtc_state)) {
> > > +			drm_connector_list_iter_end(&conn_list_iter);
> > > +			return PTR_ERR(conn_iter_state);
> > > +		}
> > > +		crtc_state->mode_changed = true;
> > > +	}
> > > +	drm_connector_list_iter_end(&conn_list_iter);
> > > +
> > > +	return 0;
> > > +}
> > > +
> > > +static int
> > > +intel_dp_atomic_trans_port_sync_check(struct drm_i915_private *dev_priv,
> > > +				      struct intel_atomic_state *state)
> > > +{
> > > +	struct drm_connector *connector;
> > > +	struct drm_crtc_state *crtc_state;
> > > +	struct drm_connector_state *connector_state;
> > > +	int i, ret, tile_grp_id = 0;
> > > +
> > > +	if (INTEL_GEN(dev_priv) < 11)
> > > +		return 0;
> > > +
> > > +	/* Is tiled, mark all other tiled CRTCs as needing a modeset */
> > > +	for_each_new_connector_in_state(&state->base, connector, connector_state, i) {
> > > +		if (!connector->has_tile)
> > > +			continue;
> > > +		if (connector_state->crtc &&
> > > +		    tile_grp_id != connector->tile_group->id) {
> > > +			crtc_state = drm_atomic_get_new_crtc_state(&state->base,
> > > +								   connector_state->crtc);
> > > +			if (!drm_atomic_crtc_needs_modeset(crtc_state))
> > > +				continue;
> > > +
> > > +			tile_grp_id = connector->tile_group->id;
> > > +		} else
> > > +			continue;
> > > +
> > > +		ret = intel_dp_modeset_all_tiles(dev_priv, state, tile_grp_id);
> > > +		if (ret)
> > > +			return ret;
> > > +	}
> > > +
> > > +	return 0;
> > > +}
> > 
> > BTW after some more pondering I don't think this alone is sufficient.
> > The tile information may have already disppeared so I believe we also
> > need to make sure we mark all currently synced crtcs as needing a
> > modeset if any of them need a modeset. And I guess that's pretty much
> > the same function we'll need to handle fastset correctly.
> >
> 
> The crtcs in the current state get synced and master/slave assignments happen in icl_add_sync_mode_crtcs before compute_config call
> So are you suggesting basically moving this function after the crtcs are synced and then just setting modeset
> to true for all synced crtcs if one of them needs modeset?

I think it should look something like:

modeset_tiled_things();
modeset_synced_things();

for_each() {
	modeset_pipes_config();
	if (can_fastset()) {
		modeset=false;
		fastset=true;
	}
}

modeset_synced_things();

for_each() {
	if (!modeset && fastset)
		copy_state();
}

> 
> And why would the tile information be disappeared?  

It'll get updated whenever someone does a getconnector() or whatever.

Example:
1. sync pipe A and B, pipe A is master
2. swap pipe B display for something else
3. getconnector() -> tile info goes poof
4. do something on pipe A that needs a modeset
   no tile info so we miss that pipe B also needs a modeset

-- 
Ville Syrjälä
Intel