| Message ID | 1472304805-4409-1-git-send-email-matthew.auld@intel.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
On Sat, 2016-08-27 at 14:33 +0100, Matthew Auld wrote: > Currently it's entirely possible to go through the link training step > without first determining the lane_count, which is silly since we end > up > doing a bunch of aux transfers of size = 0, as highlighted by > WARN_ON(!msg->buffer != !msg->size), and can only ever result in a > 'failed to update link training' message. This can be observed during > intel_dp_long_pulse where we can do the link training step, but > before > we have had a chance to set the link params. To avoid this we add an > extra check for the lane_count in intel_dp_check_link_status, which > should prevent us from doing the link training step prematurely. > > References: https://bugs.freedesktop.org/show_bug.cgi?id=97344 > Cc: Jani Nikula <jani.nikula@linux.intel.com> > Signed-off-by: Matthew Auld <matthew.auld@intel.com> > --- > drivers/gpu/drm/i915/intel_dp.c | 3 +++ > 1 file changed, 3 insertions(+) > > diff --git a/drivers/gpu/drm/i915/intel_dp.c > b/drivers/gpu/drm/i915/intel_dp.c > index a3c7dd8..0dbb672 100644 > --- a/drivers/gpu/drm/i915/intel_dp.c > +++ b/drivers/gpu/drm/i915/intel_dp.c > @@ -3927,6 +3927,9 @@ intel_dp_check_link_status(struct intel_dp > *intel_dp) > if (!to_intel_crtc(intel_encoder->base.crtc)->active) > return; > > + if (!intel_dp->lane_count) > + return; > + > /* if link training is requested we should perform it always > */ > if ((intel_dp->compliance_test_type == > DP_TEST_LINK_TRAINING) || > (!drm_dp_channel_eq_ok(link_status, intel_dp- > >lane_count))) { Should we place this check as part drm_dp_helper()'s drm_dp_channel_eq_ok() routine as this may happen with other than our i915 driver as well?
On Sat, Aug 27, 2016 at 02:33:25PM +0100, Matthew Auld wrote: > Currently it's entirely possible to go through the link training step > without first determining the lane_count, which is silly since we end up > doing a bunch of aux transfers of size = 0, as highlighted by > WARN_ON(!msg->buffer != !msg->size), and can only ever result in a > 'failed to update link training' message. This can be observed during > intel_dp_long_pulse where we can do the link training step, but before > we have had a chance to set the link params. To avoid this we add an > extra check for the lane_count in intel_dp_check_link_status, which > should prevent us from doing the link training step prematurely. > > References: https://bugs.freedesktop.org/show_bug.cgi?id=97344 > Cc: Jani Nikula <jani.nikula@linux.intel.com> > Signed-off-by: Matthew Auld <matthew.auld@intel.com> > --- > drivers/gpu/drm/i915/intel_dp.c | 3 +++ > 1 file changed, 3 insertions(+) > > diff --git a/drivers/gpu/drm/i915/intel_dp.c b/drivers/gpu/drm/i915/intel_dp.c > index a3c7dd8..0dbb672 100644 > --- a/drivers/gpu/drm/i915/intel_dp.c > +++ b/drivers/gpu/drm/i915/intel_dp.c > @@ -3927,6 +3927,9 @@ intel_dp_check_link_status(struct intel_dp *intel_dp) > if (!to_intel_crtc(intel_encoder->base.crtc)->active) > return; > > + if (!intel_dp->lane_count) > + return; This pretty much amounts to the same thing that I did with 'active_streams' in my series [1]. The problem is that we need to sync up the state on readout to make it really work. Daniel wasn't too happy with my .sync_state() hook, so maybe we need something fancier. [1] https://patchwork.freedesktop.org/series/10354/ > + > /* if link training is requested we should perform it always */ > if ((intel_dp->compliance_test_type == DP_TEST_LINK_TRAINING) || > (!drm_dp_channel_eq_ok(link_status, intel_dp->lane_count))) { > -- > 2.7.4 > > _______________________________________________ > Intel-gfx mailing list > Intel-gfx@lists.freedesktop.org > https://lists.freedesktop.org/mailman/listinfo/intel-gfx
On Tue, Sep 06, 2016 at 05:43:44PM +0300, Ville Syrjälä wrote: > On Sat, Aug 27, 2016 at 02:33:25PM +0100, Matthew Auld wrote: > > Currently it's entirely possible to go through the link training step > > without first determining the lane_count, which is silly since we end up > > doing a bunch of aux transfers of size = 0, as highlighted by > > WARN_ON(!msg->buffer != !msg->size), and can only ever result in a > > 'failed to update link training' message. This can be observed during > > intel_dp_long_pulse where we can do the link training step, but before > > we have had a chance to set the link params. To avoid this we add an > > extra check for the lane_count in intel_dp_check_link_status, which > > should prevent us from doing the link training step prematurely. > > > > References: https://bugs.freedesktop.org/show_bug.cgi?id=97344 > > Cc: Jani Nikula <jani.nikula@linux.intel.com> > > Signed-off-by: Matthew Auld <matthew.auld@intel.com> > > --- > > drivers/gpu/drm/i915/intel_dp.c | 3 +++ > > 1 file changed, 3 insertions(+) > > > > diff --git a/drivers/gpu/drm/i915/intel_dp.c b/drivers/gpu/drm/i915/intel_dp.c > > index a3c7dd8..0dbb672 100644 > > --- a/drivers/gpu/drm/i915/intel_dp.c > > +++ b/drivers/gpu/drm/i915/intel_dp.c > > @@ -3927,6 +3927,9 @@ intel_dp_check_link_status(struct intel_dp *intel_dp) > > if (!to_intel_crtc(intel_encoder->base.crtc)->active) > > return; > > > > + if (!intel_dp->lane_count) > > + return; > > This pretty much amounts to the same thing that I did with 'active_streams' in > my series [1]. The problem is that we need to sync up the state on readout to > make it really work. Daniel wasn't too happy with my .sync_state() hook, so > maybe we need something fancier. > > [1] https://patchwork.freedesktop.org/series/10354/ So, looking at some recent logs I see we may hit this WARN spew a *lot*. So we probably want this patch as a temporary measure. Maybe with WARN_ONCE() + a comment, eg. /* * FIXME need to synchronize this sort of * stuff properly with hardware readout. */ ? > > > + > > /* if link training is requested we should perform it always */ > > if ((intel_dp->compliance_test_type == DP_TEST_LINK_TRAINING) || > > (!drm_dp_channel_eq_ok(link_status, intel_dp->lane_count))) { > > -- > > 2.7.4 > > > > _______________________________________________ > > Intel-gfx mailing list > > Intel-gfx@lists.freedesktop.org > > https://lists.freedesktop.org/mailman/listinfo/intel-gfx > > -- > Ville Syrjälä > Intel OTC > _______________________________________________ > Intel-gfx mailing list > Intel-gfx@lists.freedesktop.org > https://lists.freedesktop.org/mailman/listinfo/intel-gfx
diff --git a/drivers/gpu/drm/i915/intel_dp.c b/drivers/gpu/drm/i915/intel_dp.c index a3c7dd8..0dbb672 100644 --- a/drivers/gpu/drm/i915/intel_dp.c +++ b/drivers/gpu/drm/i915/intel_dp.c @@ -3927,6 +3927,9 @@ intel_dp_check_link_status(struct intel_dp *intel_dp) if (!to_intel_crtc(intel_encoder->base.crtc)->active) return; + if (!intel_dp->lane_count) + return; + /* if link training is requested we should perform it always */ if ((intel_dp->compliance_test_type == DP_TEST_LINK_TRAINING) || (!drm_dp_channel_eq_ok(link_status, intel_dp->lane_count))) {
Currently it's entirely possible to go through the link training step without first determining the lane_count, which is silly since we end up doing a bunch of aux transfers of size = 0, as highlighted by WARN_ON(!msg->buffer != !msg->size), and can only ever result in a 'failed to update link training' message. This can be observed during intel_dp_long_pulse where we can do the link training step, but before we have had a chance to set the link params. To avoid this we add an extra check for the lane_count in intel_dp_check_link_status, which should prevent us from doing the link training step prematurely. References: https://bugs.freedesktop.org/show_bug.cgi?id=97344 Cc: Jani Nikula <jani.nikula@linux.intel.com> Signed-off-by: Matthew Auld <matthew.auld@intel.com> --- drivers/gpu/drm/i915/intel_dp.c | 3 +++ 1 file changed, 3 insertions(+)