[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 08:40:24AM -0800, Manasi Navare wrote:
> On Mon, Dec 16, 2019 at 02:03:43PM +0200, Ville Syrjälä wrote:
> > On Fri, Dec 13, 2019 at 06:28:40PM -0800, Manasi Navare wrote:
> > > On Fri, Dec 13, 2019 at 10:05:49PM +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;
> > > > 'connector'
> > > > > +	struct drm_connector_list_iter conn_list_iter;
> > > > > +	struct drm_crtc_state *crtc_state;
> > > > 
> > > > crtc_state has needlessly wide scope.
> > > > 
> > > > > +
> > > > > +	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;
> > > > 
> > > > 'conn_state' is the most popular name.
> > > > 
> > > > > +
> > > > > +		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;
> > > > 
> > > > The tile group check should be part of the same if with the has_tile
> > > > check.
> > > > 
> > > > > +
> > > > > +		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,
> > > > 
> > > > Pointless variable. Can be extracted from the atomic state.
> > > > 
> > > > > +				      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;
> > > > 
> > > > tile_grp_id is rather pointless. crtc_state and ret can move into
> > > > tighter scope. And the next suggestion allows you to kill crtc_state
> > > > entirely...
> > > 
> > > Its not clear why tile_grp_id is pointless, I am using tile_grp_id for the first connector with has_tile
> > > and I make sure that I dont enter into the loop to check modeset again for the connector with
> > > same tile_grp_id because we have already set its mode changed to true in intel_dp_modeset_all_tiles()
> > > 
> > > How can I achieve this instead?
> > 
> > 
> > Instead of 
> > foo()
> > {
> > 	tile_grp_id = 0;
> > 
> > 	for_each() {
> > 		tile_grp_id = conn->tile_grp_id;
> > 
> > 		intel_dp_modeset_all_tiles(tile_grp_id);
> > 	}
> > }
> > 
> > you just do
> > foo()
> > {
> > 	for_each() {
> > 		intel_dp_modeset_all_tiles(conn->tile_grp_id);
> > 	}
> > }
> 
> Yes I understand that we can pass the conn->tile_grp_id directly. But currently I am using tile_grp_id in for_each(), to call intel_dp_modeset_all_tiles()
> only if that cnnector is from a differnt tile grp id else continue.
> 
> foo()
> {
> 	for_each() {
> 		if(tile_grp_id != conn->tile_grp_id)
> 			intel_dpmodeset_all_tiles();
> 		else
> 			continue;
> 	}
> }
> 
> calling intel_dp_modeset_all_tiles() for all tiles is kind of redundant. How can I do this with your suggestion, or do you think
> we should call intel_dp_modeset_all_tiles for each connector anyway and it shouldnt matter?

Yes if that connector is tiled.

Eg. if you have something like
connector A: tile group 1
connector B: tile group 2
connector C: tile group 3
connector D: tile group 2

then your single tile_grp_id variable doesn't work anyway. You'd need
track which tile groups you've already handled and skip those. But
much easier to just blindly plow ahead.

-- 
Ville Syrjälä
Intel