[Intel-gfx] [PATCH 1/4] drm/i915/tgl: Select master transcoder for MST stream
Ville Syrjälä
ville.syrjala at linux.intel.com
Mon Dec 9 19:43:35 UTC 2019
On Mon, Dec 09, 2019 at 06:45:43PM +0000, Souza, Jose wrote:
> On Mon, 2019-12-09 at 18:40 +0200, Ville Syrjälä wrote:
> > On Fri, Dec 06, 2019 at 05:18:29PM -0800, José Roberto de Souza
> > wrote:
> > > On TGL the blending of all the streams have moved from DDI to
> > > transcoder, so now every transcoder working over the same MST port
> > > must
> > > send its stream to a master transcoder and master will send to DDI
> > > respecting the time slots.
> > >
> > > So here adding all the CRTCs that shares the same MST stream if
> > > needed and computing their state again, it will pick the first
> > > transcoder among the ones in the same stream to be master.
> > >
> > > Most of the time skl_commit_modeset_enables() enables pipes in a
> > > crescent order but due DDB overlapping it might not happen, this
> > > scenario will be handled in the next patch.
> > >
> > > BSpec: 50493
> > > BSpec: 49190
> > > Cc: Ville Syrjälä <ville.syrjala at linux.intel.com>
> > > Cc: Lucas De Marchi <lucas.demarchi at intel.com>
> > > Signed-off-by: José Roberto de Souza <jose.souza at intel.com>
> > > ---
> > > drivers/gpu/drm/i915/display/intel_ddi.c | 14 +-
> > > drivers/gpu/drm/i915/display/intel_display.c | 15 +-
> > > .../drm/i915/display/intel_display_types.h | 3 +
> > > drivers/gpu/drm/i915/display/intel_dp_mst.c | 163
> > > ++++++++++++++++++
> > > drivers/gpu/drm/i915/display/intel_dp_mst.h | 5 +
> > > 5 files changed, 196 insertions(+), 4 deletions(-)
> > >
> > > diff --git a/drivers/gpu/drm/i915/display/intel_ddi.c
> > > b/drivers/gpu/drm/i915/display/intel_ddi.c
> > > index 3cacb1e279c1..be5bc506d4d3 100644
> > > --- a/drivers/gpu/drm/i915/display/intel_ddi.c
> > > +++ b/drivers/gpu/drm/i915/display/intel_ddi.c
> > > @@ -1903,8 +1903,13 @@ intel_ddi_transcoder_func_reg_val_get(const
> > > struct intel_crtc_state *crtc_state)
> > > temp |= TRANS_DDI_MODE_SELECT_DP_MST;
> > > temp |= DDI_PORT_WIDTH(crtc_state->lane_count);
> > >
> > > - if (INTEL_GEN(dev_priv) >= 12)
> > > - temp |=
> > > TRANS_DDI_MST_TRANSPORT_SELECT(crtc_state->cpu_transcoder);
> > > + if (INTEL_GEN(dev_priv) >= 12) {
> > > + enum transcoder master;
> > > +
> > > + master = crtc_state->mst_master_transcoder;
> > > + WARN_ON(master == INVALID_TRANSCODER);
> >
> > I'd drop the WARN_ON(). If we keep adding these for every little
> > thing
> > we'll drown in them.
> >
> > > + temp |= TRANS_DDI_MST_TRANSPORT_SELECT(master);
> > > + }
> > > } else {
> > > temp |= TRANS_DDI_MODE_SELECT_DP_SST;
> > > temp |= DDI_PORT_WIDTH(crtc_state->lane_count);
> > > @@ -4372,6 +4377,11 @@ void intel_ddi_get_config(struct
> > > intel_encoder *encoder,
> > > pipe_config->output_types |= BIT(INTEL_OUTPUT_DP_MST);
> > > pipe_config->lane_count =
> > > ((temp & DDI_PORT_WIDTH_MASK) >>
> > > DDI_PORT_WIDTH_SHIFT) + 1;
> > > +
> > > + if (INTEL_GEN(dev_priv) >= 12)
> > > + pipe_config->mst_master_transcoder =
> > > + REG_FIELD_GET(TRANS_DDI_MST_TRA
> > > NSPORT_SELECT_MASK, temp);
> > > +
> > > intel_dp_get_m_n(intel_crtc, pipe_config);
> > > break;
> > > default:
> > > diff --git a/drivers/gpu/drm/i915/display/intel_display.c
> > > b/drivers/gpu/drm/i915/display/intel_display.c
> > > index 821ba8053f9d..f89494c849ce 100644
> > > --- a/drivers/gpu/drm/i915/display/intel_display.c
> > > +++ b/drivers/gpu/drm/i915/display/intel_display.c
> > > @@ -46,6 +46,7 @@
> > > #include "display/intel_crt.h"
> > > #include "display/intel_ddi.h"
> > > #include "display/intel_dp.h"
> > > +#include "display/intel_dp_mst.h"
> > > #include "display/intel_dsi.h"
> > > #include "display/intel_dvo.h"
> > > #include "display/intel_gmbus.h"
> > > @@ -12542,6 +12543,11 @@ static void intel_dump_pipe_config(const
> > > struct intel_crtc_state *pipe_config,
> > > pipe_config->csc_mode, pipe_config-
> > > >gamma_mode,
> > > pipe_config->gamma_enable, pipe_config-
> > > >csc_enable);
> > >
> > > + if (INTEL_GEN(dev_priv) >= 12 &&
> > > + intel_crtc_has_type(pipe_config, INTEL_OUTPUT_DP_MST))
> >
> > Could probably print this unconditionally to keep the code less
> > messy.
>
> I'm not setting mst_master_transcoder = INVALID_TRANSCODER in the other
> code paths, so it would print transcoder A for HDMI, DP SST and to DP
> MST in gen < 12.
> That is fine? Should I add mst_master_transcoder = INVALID_TRANSCODER
> to all those code paths? For me the best option is keep this checks.
I think setting to invalid would be nice.
With https://patchwork.freedesktop.org/series/69129/
we could even do it in a nice central place just the once.
>
>
> >
> > > + DRM_DEBUG_KMS("MST master transcoder: %s\n",
> > > + transcoder_name(pipe_config-
> > > >mst_master_transcoder));
> > > +
> > > dump_planes:
> > > if (!state)
> > > return;
> > > @@ -13318,6 +13324,10 @@ intel_pipe_config_compare(const struct
> > > intel_crtc_state *current_config,
> > > PIPE_CONF_CHECK_I(sync_mode_slaves_mask);
> > > PIPE_CONF_CHECK_I(master_transcoder);
> > >
> > > + if (INTEL_GEN(dev_priv) >= 12 &&
> > > + intel_crtc_has_type(pipe_config, INTEL_OUTPUT_DP_MST))
> >
> > These checks are definitely pointless.
>
> Similar with the above, it would not cause mismatch because for non gen
> 12 DP MST it would compare trans A with trans A, will change that
> depending on your answer to the above.
>
>
> >
> > > + PIPE_CONF_CHECK_I(mst_master_transcoder);
> > > +
> > > #undef PIPE_CONF_CHECK_X
> > > #undef PIPE_CONF_CHECK_I
> > > #undef PIPE_CONF_CHECK_BOOL
> > > @@ -14406,7 +14416,7 @@ static void
> > > intel_commit_modeset_disables(struct intel_atomic_state *state)
> > > u32 handled = 0;
> > > int i;
> > >
> > > - /* Only disable port sync slaves */
> > > + /* Only disable port sync and MST slaves */
> > > for_each_oldnew_intel_crtc_in_state(state, crtc,
> > > old_crtc_state,
> > > new_crtc_state, i) {
> > > if (!needs_modeset(new_crtc_state))
> > > @@ -14420,7 +14430,8 @@ static void
> > > intel_commit_modeset_disables(struct intel_atomic_state *state)
> > > * slave CRTCs are disabled first and then master CRTC
> > > since
> > > * Slave vblanks are masked till Master Vblanks.
> > > */
> > > - if (!is_trans_port_sync_slave(old_crtc_state))
> > > + if (!is_trans_port_sync_slave(old_crtc_state) &&
> > > + !intel_dp_mst_is_slave_trans(old_crtc_state))
> > > continue;
> > >
> > > intel_pre_plane_update(state, crtc);
> > > diff --git a/drivers/gpu/drm/i915/display/intel_display_types.h
> > > b/drivers/gpu/drm/i915/display/intel_display_types.h
> > > index 83ea04149b77..630a94892b7b 100644
> > > --- a/drivers/gpu/drm/i915/display/intel_display_types.h
> > > +++ b/drivers/gpu/drm/i915/display/intel_display_types.h
> > > @@ -1054,6 +1054,9 @@ struct intel_crtc_state {
> > >
> > > /* Bitmask to indicate slaves attached */
> > > u8 sync_mode_slaves_mask;
> > > +
> > > + /* Only valid on TGL+ */
> > > + enum transcoder mst_master_transcoder;
> > > };
> > >
> > > struct intel_crtc {
> > > diff --git a/drivers/gpu/drm/i915/display/intel_dp_mst.c
> > > b/drivers/gpu/drm/i915/display/intel_dp_mst.c
> > > index 926e49f449a6..49f1518e3ab9 100644
> > > --- a/drivers/gpu/drm/i915/display/intel_dp_mst.c
> > > +++ b/drivers/gpu/drm/i915/display/intel_dp_mst.c
> > > @@ -87,6 +87,49 @@ static int
> > > intel_dp_mst_compute_link_config(struct intel_encoder *encoder,
> > > return 0;
> > > }
> > >
> > > +/*
> > > + * Iterate over all the CRTCs and set mst_master_transcoder to the
> > > first active
> > > + * transcoder that shares the same MST connector.
> > > + *
> > > + * TODO: maybe this could be optimized to keep the old master
> > > transcoder
> > > + */
> > > +static int
> > > +intel_dp_mst_master_trans_compute(struct intel_encoder *encoder,
> > > + struct intel_crtc_state *pipe_config,
> >
> > s/pipe_config/crtc_state/
>
> Okay
>
> >
> > > + struct drm_connector_state
> > > *conn_state)
> > > +{
> > > + struct intel_atomic_state *state =
> > > to_intel_atomic_state(pipe_config->uapi.state);
> > > + struct intel_connector *conn = to_intel_connector(conn_state-
> > > >connector);
> >
> > ocd: s/conn/connector/ all over.
>
> Okay
>
> >
> > > + struct drm_i915_private *dev_priv = to_i915(encoder->base.dev);
> > > + struct intel_digital_connector_state *iter_conn_state;
> > > + struct intel_connector *iter_conn;
> > > + int i;
> > > +
> > > + if (INTEL_GEN(dev_priv) < 12)
> > > + return 0;
> > > +
> > > + for_each_new_intel_connector_in_state(state, iter_conn,
> > > + iter_conn_state, i) {
> > > + struct intel_crtc_state *iter_crtc_state;
> > > + struct intel_crtc *iter_crtc;
> > > +
> > > + if (conn->mst_port != iter_conn->mst_port ||
> > > + !iter_conn_state->base.crtc)
> > > + continue;
> > > +
> > > + iter_crtc = to_intel_crtc(iter_conn_state->base.crtc);
> > > + iter_crtc_state =
> > > intel_atomic_get_new_crtc_state(state,
> > > + iter_
> > > crtc);
> > > + if (!iter_crtc_state->uapi.active)
> > > + continue;
> > > +
> > > + pipe_config->mst_master_transcoder = iter_crtc_state-
> > > >cpu_transcoder;
> > > + break;
> >
> > So we're going to pick this based on the connector order. How does
> > that
> > fit in with the port sync, or did we not have port sync for MST yet?
> >
> > To keep everything consistent and easy my idea was to pick the first
> > pipe as the master for both MST and port sync.
> >
> > Although I can easily imagine topologies where even that would
> > land us in some interesting mess. For example:
> > pipe A -> MST port B -> ... -> whatever : MST master
> > pipe B -> MST port B -> ... -> tile 0 : port sync master / MST slave
> > pipe C -> MST port C -> ... -> tile 1 : port sync slave / MST master
> > pipe D -> MST port C -> ... -> whatever : MST slave
> >
> > pipe A -> MST port B -> ... -> whatever : MST master
> > pipe B -> MST port B -> ... -> tile 0 : port sync master / MST slave
> > pipe C -> MST port C -> ... -> whatever : MST master
> > pipe D -> MST port C -> ... -> tile 1 : port sync slave / MST slave
> >
> > No idea if the hw can even do something like that. Needs more thought
> > if/when we enable port sync with MST (assuming we're not already
> > doing
> > that).
>
> According to BSpec is possible to have port sync with MST but it is not
> implemented. But sure it can be easily fixed, will loop throgh all the
> connectors updating mst_master_transcoder with the smallest pipe.
I don't think it's quite that easy since the ordering constraints of
MST vs. port sync might conflict as in my examples above. But let's
just ignore that particular can of worms for now.
--
Ville Syrjälä
Intel
More information about the Intel-gfx
mailing list