[Intel-gfx] [PATCH 1/3] drm/i915/icl: Release TC ports when unloading or suspending driver
Imre Deak
imre.deak at intel.com
Thu Nov 29 21:03:15 UTC 2018
On Thu, Nov 29, 2018 at 10:18:40PM +0200, Souza, Jose wrote:
> On Thu, 2018-11-29 at 15:52 +0200, Imre Deak wrote:
> > On Wed, Nov 28, 2018 at 11:54:21PM +0200, Souza, Jose wrote:
> > > On Wed, 2018-11-28 at 13:34 +0200, Imre Deak wrote:
> > > > On Wed, Nov 07, 2018 at 04:05:52PM -0800, José Roberto de Souza
> > > > wrote:
> > > > > When suspending or unloading the driver, it needs to release
> > > > > the
> > > > > TC ports so HW can change it state without wait for driver
> > > > > handshake.
> > > >
> > > > According to
> > > > https://bugs.freedesktop.org/show_bug.cgi?id=108070#c26
> > > >
> > > > this patch should fix the bug reported at
> > > > https://bugs.freedesktop.org/show_bug.cgi?id=108070#c17
> > > >
> > > > but, I can't see how the change here would fix the corresponding
> > > > problem
> > > > described in
> > > > https://bugs.freedesktop.org/show_bug.cgi?id=108070#c18
> > > >
> > > > Would you explain?
> > > >
> > > > I think there are more fundamental problems in TypeC HPD handling
> > > > as
> > > > we
> > > > discussed earlier here on the list, which should be fixed first:
> > > >
> > > > - Switching to/from type C mode from the interrupt handler
> > > > without
> > > > considering if we have an active mode or an ongoing AUX
> > > > transfer.
> > >
> > > Yes that is still missing, I plan to work on that after I deliver a
> > > few
> > > tasks. Also I read the emails in discussion that you started about
> > > that.
> >
> > The point is that imo those fundamental problems should be fixed
> > first.
> >
> > > > - Not having any way for handling the case where we'd do a
> > > > modeset
> > > > after
> > > > an HPD disconnect interrupt, like in the case of
> > > > common-hpd-after-suspend in the bug report.
> > >
> > > For what I understood about this test it will suspend, disconnect
> > > chamelium board from source, wakeup and check if connector state
> > > match,
> > > as we did not marked port as safe before suspending we are left in
> > > a
> > > invalid state where PD firmware do not trigger a interruption after
> > > wakeup.
> >
> > No, it's not about the lack of HPD interrupts. What happens is that
> > the
> > sink disappears during system suspend/resume, but we still have an
> > active mode enabled on the corresponding port. We try to restore this
> > mode during resume but that modeset will fail.
>
> Before suspend, all display output is disabled so there is no active
> modeset.
Yes, there is:
https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_5129/fi-icl-u2/dmesg0.log :
<7>[ 204.806237] [drm:intel_atomic_check [i915]] [CONNECTOR:209:HDMI-A-2] Limiting display bpp to 24 instead of EDID bpp 24, requested bpp 36, max platform bpp 36
...
<7>[ 204.806411] [drm:intel_dump_pipe_config [i915]] [CRTC:80:pipe A][modeset]
...
<7>[ 204.809132] [drm:intel_enable_pipe [i915]] enabling pipe A
...
<7>[ 204.909519] [IGT] kms_chamelium: executing
...
<7>[ 204.947440] [IGT] kms_chamelium: starting subtest common-hpd-after-suspend
...
<6>[ 205.936472] PM: suspend entry (deep)
After which we will try to restore this active mode during resume.
>
> >
> > > > More comments below:
> > > >
> > > > > Cc: Imre Deak <imre.deak at intel.com>
> > > > > Cc: Ville Syrjälä <ville.syrjala at linux.intel.com>
> > > > > Signed-off-by: José Roberto de Souza <jose.souza at intel.com>
> > > > > ---
> > > > > drivers/gpu/drm/i915/i915_drv.c | 1 +
> > > > > drivers/gpu/drm/i915/i915_drv.h | 1 +
> > > > > drivers/gpu/drm/i915/i915_irq.c | 1 +
> > > > > drivers/gpu/drm/i915/intel_dp.c | 19 +++++++++++++++++++
> > > > > drivers/gpu/drm/i915/intel_drv.h | 1 +
> > > > > drivers/gpu/drm/i915/intel_hotplug.c | 9 +++++++++
> > > > > 6 files changed, 32 insertions(+)
> > > > >
> > > > > diff --git a/drivers/gpu/drm/i915/i915_drv.c
> > > > > b/drivers/gpu/drm/i915/i915_drv.c
> > > > > index acb516308262..14331c396278 100644
> > > > > --- a/drivers/gpu/drm/i915/i915_drv.c
> > > > > +++ b/drivers/gpu/drm/i915/i915_drv.c
> > > > > @@ -1911,6 +1911,7 @@ static int i915_drm_suspend(struct
> > > > > drm_device
> > > > > *dev)
> > > > >
> > > > > intel_runtime_pm_disable_interrupts(dev_priv);
> > > > > intel_hpd_cancel_work(dev_priv);
> > > > > + intel_hpd_suspend(dev_priv);
> > > >
> > > > What about runtime suspend? We won't receive HPD interrupts after
> > > > that
> > > > either, so we'd need to disconnect in that case too according to
> > > > the
> > > > spec.
> > >
> > > Oh my bad I was thinking that the regular suspend would handle this
> > > too
> > > but I was wrong.
> > >
> > > So I will add a intel_hpd_suspend() call to intel_runtime_suspend()
> > > leaving the i915_drv.c changes as this.
> > >
> > > diff --git a/drivers/gpu/drm/i915/i915_drv.c
> > > b/drivers/gpu/drm/i915/i915_drv.c
> > > index b1d23c73c147..948914a79c67 100644
> > > --- a/drivers/gpu/drm/i915/i915_drv.c
> > > +++ b/drivers/gpu/drm/i915/i915_drv.c
> > > @@ -1905,6 +1905,7 @@ static int i915_drm_suspend(struct drm_device
> > > *dev)
> > >
> > > intel_runtime_pm_disable_interrupts(dev_priv);
> > > intel_hpd_cancel_work(dev_priv);
> > > + intel_hpd_suspend(dev_priv);
> > >
> > > intel_suspend_encoders(dev_priv);
> > >
> > > @@ -2909,6 +2910,7 @@ static int intel_runtime_suspend(struct
> > > device
> > > *kdev)
> > > intel_uc_suspend(dev_priv);
> > >
> > > intel_runtime_pm_disable_interrupts(dev_priv);
> > > + intel_hpd_suspend(dev_priv);
> > >
> > > intel_uncore_suspend(dev_priv);
> > >
> > > Also I will drop the next patch as we don't call
> > > intel_hpd_cancel_work() in intel_runtime_suspend() path.
> > >
> > >
> > > > And how are we handling resume then where we'll be in a
> > > > disconnected
> > > > state due to the change in this patch and would continue to do a
> > > > modeset (to restore the mode we had before suspend).
> > >
> > > The drm_kms_helper_poll_enable() calls will trigger the call to the
> > > detection function of the connectors that will call
> > > intel_digital_port_connected() thal will call ICL TC functions that
> > > will take care of update everything.
> >
> > No detection will be triggered, since we're not polling this port
> > (and
> > polling can be disabled as a whole). In any case that would be too
> > late,
> > since we're trying to restore the mode before we call
> > drm_kms_helper_poll_enable(). Also doing a detection is not enough
> > here, since if detection fails (as it would in our case) we still
> > have a
> > mode that we need to restore. The corresponding modeset would fail,
> > but
> > we don't have a way to fail it gracefully as I wrote above.
>
>
> For the regular suspend it will run the connector detection functions
> for sure when trying to restore the previous modeset.
Not sure what you mean.
>From drm_kms_helper_poll_enable():
if (!dev->mode_config.poll_enabled || !drm_kms_helper_poll)
return;
so we won't do anything in this function if polling is globally disabled.
if (connector->polled & (DRM_CONNECTOR_POLL_CONNECT |
DRM_CONNECTOR_POLL_DISCONNECT))
poll = true;
so we won't do anything if we haven't enabled polling for the connector.
Generally we don't enable polling for connectors which provide HPD
interrupts.
There is actually a detection work scheduled from intel_hpd_init(), but
again that will run only after we tried to restore the pre-suspend mode.
Also note what I wrote about detection being not enough for us. The sink
can be gone by the time we resume and we will still try to restore the
mode.
> About the runtime resume I need to check if with the polling disabled
> it will still update the connector state.
>
>
> >
> > > > >
> > > > > intel_suspend_encoders(dev_priv);
> > > > >
> > > > > diff --git a/drivers/gpu/drm/i915/i915_drv.h
> > > > > b/drivers/gpu/drm/i915/i915_drv.h
> > > > > index 0c8438de3c1b..96d5ddc36f4e 100644
> > > > > --- a/drivers/gpu/drm/i915/i915_drv.h
> > > > > +++ b/drivers/gpu/drm/i915/i915_drv.h
> > > > > @@ -2792,6 +2792,7 @@ enum hpd_pin intel_hpd_pin_default(struct
> > > > > drm_i915_private *dev_priv,
> > > > > enum port port);
> > > > > bool intel_hpd_disable(struct drm_i915_private *dev_priv, enum
> > > > > hpd_pin pin);
> > > > > void intel_hpd_enable(struct drm_i915_private *dev_priv, enum
> > > > > hpd_pin pin);
> > > > > +void intel_hpd_suspend(struct drm_i915_private *dev_priv);
> > > > >
> > > > > /* i915_irq.c */
> > > > > static inline void i915_queue_hangcheck(struct
> > > > > drm_i915_private
> > > > > *dev_priv)
> > > > > diff --git a/drivers/gpu/drm/i915/i915_irq.c
> > > > > b/drivers/gpu/drm/i915/i915_irq.c
> > > > > index d7e47d6082de..23084d227e2a 100644
> > > > > --- a/drivers/gpu/drm/i915/i915_irq.c
> > > > > +++ b/drivers/gpu/drm/i915/i915_irq.c
> > > > > @@ -4996,6 +4996,7 @@ void intel_irq_uninstall(struct
> > > > > drm_i915_private *dev_priv)
> > > > > {
> > > > > drm_irq_uninstall(&dev_priv->drm);
> > > > > intel_hpd_cancel_work(dev_priv);
> > > > > + intel_hpd_suspend(dev_priv);
> > > > > dev_priv->runtime_pm.irqs_enabled = false;
> > > > > }
> > > > >
> > > > > diff --git a/drivers/gpu/drm/i915/intel_dp.c
> > > > > b/drivers/gpu/drm/i915/intel_dp.c
> > > > > index 5258c9d654f4..27c6163426d6 100644
> > > > > --- a/drivers/gpu/drm/i915/intel_dp.c
> > > > > +++ b/drivers/gpu/drm/i915/intel_dp.c
> > > > > @@ -5043,6 +5043,25 @@ bool intel_digital_port_connected(struct
> > > > > intel_encoder *encoder)
> > > > > return false;
> > > > > }
> > > > >
> > > > > +/*
> > > > > + * intel_digital_port_force_disconnection - Used to force port
> > > > > disconnection or
> > > > > + * release any control from display driver before going to a
> > > > > state
> > > > > that driver
> > > > > + * will not handle interruptions.
> > > > > + */
> > > > > +void intel_digital_port_force_disconnection(struct
> > > > > intel_encoder
> > > > > *encoder)
> > > > > +{
> > > > > + struct drm_i915_private *dev_priv = to_i915(encoder-
> > > > > >base.dev);
> > > > > + struct intel_digital_port *dig_port;
> > > > > +
> > > > > + dig_port = enc_to_dig_port(&encoder->base);
> > > > > + /* Skip fake MST ports */
> > > > > + if (!dig_port)
> > > > > + return;
> > > > > +
> > > > > + if (INTEL_GEN(dev_priv) >= 11)
> > > > > + icl_tc_phy_disconnect(dev_priv, dig_port);
> > > > > +}
> > > > > +
> > > > > static struct edid *
> > > > > intel_dp_get_edid(struct intel_dp *intel_dp)
> > > > > {
> > > > > diff --git a/drivers/gpu/drm/i915/intel_drv.h
> > > > > b/drivers/gpu/drm/i915/intel_drv.h
> > > > > index 6772e9974751..bc30f107b430 100644
> > > > > --- a/drivers/gpu/drm/i915/intel_drv.h
> > > > > +++ b/drivers/gpu/drm/i915/intel_drv.h
> > > > > @@ -1851,6 +1851,7 @@ bool intel_dp_read_dpcd(struct intel_dp
> > > > > *intel_dp);
> > > > > int intel_dp_link_required(int pixel_clock, int bpp);
> > > > > int intel_dp_max_data_rate(int max_link_clock, int max_lanes);
> > > > > bool intel_digital_port_connected(struct intel_encoder
> > > > > *encoder);
> > > > > +void intel_digital_port_force_disconnection(struct
> > > > > intel_encoder
> > > > > *encoder);
> > > > >
> > > > > /* intel_dp_aux_backlight.c */
> > > > > int intel_dp_aux_init_backlight_funcs(struct intel_connector
> > > > > *intel_connector);
> > > > > diff --git a/drivers/gpu/drm/i915/intel_hotplug.c
> > > > > b/drivers/gpu/drm/i915/intel_hotplug.c
> > > > > index 42e61e10f517..d6745b7b79d5 100644
> > > > > --- a/drivers/gpu/drm/i915/intel_hotplug.c
> > > > > +++ b/drivers/gpu/drm/i915/intel_hotplug.c
> > > > > @@ -655,3 +655,12 @@ void intel_hpd_enable(struct
> > > > > drm_i915_private
> > > > > *dev_priv, enum hpd_pin pin)
> > > > > dev_priv->hotplug.stats[pin].state = HPD_ENABLED;
> > > > > spin_unlock_irq(&dev_priv->irq_lock);
> > > > > }
> > > > > +
> > > > > +void intel_hpd_suspend(struct drm_i915_private *dev_priv)
> > > > > +{
> > > > > + struct drm_device *dev = &dev_priv->drm;
> > > > > + struct intel_encoder *encoder;
> > > > > +
> > > > > + for_each_intel_encoder(dev, encoder)
> > > > > + intel_digital_port_force_disconnection(encoder)
> > > > > ;
> > > > > +}
> > > > > --
> > > > > 2.19.1
> > > > >
> >
> >
More information about the Intel-gfx
mailing list