On Tue, Feb 11, 2020 at 07:13:31PM +0200, Ville Syrjälä wrote:
> On Tue, Feb 11, 2020 at 06:02:33PM +0100, Daniel Vetter wrote:
> > On Tue, Feb 11, 2020 at 06:22:06PM +0200, Ville Syrjala wrote:
> > > From: Ville Syrjälä <ville.syrjala@xxxxxxxxxxxxxxx>
> > >
> > > Many drivers are populating encoder->possible_clones wrong. Let's
> > > persuade them to get it right by adding some loud WARNs.
> > >
> > > We'll cross check the bits between any two encoders. So either
> > > both encoders can clone with the other, or neither can.
> > >
> > > We'll also complain about effectively empty possible_clones, and
> > > possible_clones containing bits for encoders that don't exist.
> > >
> > > v2: encoder->possible_clones now includes the encoder itelf
> > > v3: Move to drm_mode_config_validate() (Daniel)
> > > Document that you get a WARN when this is wrong (Daniel)
> > > Extract full_encoder_mask()
> > >
> > > Acked-by: Thomas Zimmermann <tzimmermann@xxxxxxx>
> > > Cc: Daniel Vetter <daniel@xxxxxxxx>
> > > Signed-off-by: Ville Syrjälä <ville.syrjala@xxxxxxxxxxxxxxx>
> >
> > I wonder whether we should start to have some unit tests for stuff like
> > this, like set up broken driver, make sure we have a WARN in dmesg. But
> > ideally we'd do that with the mocking stuff Kunit hopefully has soon.
> >
> > </idle musings>
> >
> >
> > > ---
> > > drivers/gpu/drm/drm_mode_config.c | 40 +++++++++++++++++++++++++++++++
> > > include/drm/drm_encoder.h | 2 ++
> > > 2 files changed, 42 insertions(+)
> > >
> > > diff --git a/drivers/gpu/drm/drm_mode_config.c b/drivers/gpu/drm/drm_mode_config.c
> > > index 75e357c7e84d..afc91447293a 100644
> > > --- a/drivers/gpu/drm/drm_mode_config.c
> > > +++ b/drivers/gpu/drm/drm_mode_config.c
> > > @@ -533,6 +533,17 @@ void drm_mode_config_cleanup(struct drm_device *dev)
> > > }
> > > EXPORT_SYMBOL(drm_mode_config_cleanup);
> > >
> > > +static u32 full_encoder_mask(struct drm_device *dev)
> > > +{
> > > + struct drm_encoder *encoder;
> > > + u32 encoder_mask = 0;
> > > +
> > > + drm_for_each_encoder(encoder, dev)
> > > + encoder_mask |= drm_encoder_mask(encoder);
> > > +
> > > + return encoder_mask;
> > > +}
> > > +
> > > /*
> > > * For some reason we want the encoder itself included in
> > > * possible_clones. Make life easy for drivers by allowing them
> > > @@ -544,10 +555,39 @@ static void fixup_encoder_possible_clones(struct drm_encoder *encoder)
> > > encoder->possible_clones = drm_encoder_mask(encoder);
> > > }
> > >
> > > +static void validate_encoder_possible_clones(struct drm_encoder *encoder)
> > > +{
> > > + struct drm_device *dev = encoder->dev;
> > > + u32 encoder_mask = full_encoder_mask(dev);
> > > + struct drm_encoder *other;
> > > +
> > > + drm_for_each_encoder(other, dev) {
> > > + WARN(!(encoder->possible_clones & drm_encoder_mask(other)) !=
> > > + !(other->possible_clones & drm_encoder_mask(encoder)),
> >
> > Bikeshed: !! as canonical "make this a bool value" might be slightly
> > clearer, but whatever.
>
> Can repaint.
>
> >
> > > + "possible_clones mismatch: "
> > > + "[ENCODER:%d:%s] mask=0x%x possible_clones=0x%x vs. "
> > > + "[ENCODER:%d:%s] mask=0x%x possible_clones=0x%x\n",
> > > + encoder->base.id, encoder->name,
> > > + drm_encoder_mask(encoder), encoder->possible_clones,
> > > + other->base.id, other->name,
> > > + drm_encoder_mask(other), other->possible_clones);
> > > + }
> > > +
> > > + WARN((encoder->possible_clones & drm_encoder_mask(encoder)) == 0 ||
> > > + (encoder->possible_clones & ~encoder_mask) != 0,
> > > + "Bogus possible_clones: "
> > > + "[ENCODER:%d:%s] possible_clones=0x%x (full encoder mask=0x%x)\n",
> > > + encoder->base.id, encoder->name,
> > > + encoder->possible_clones, encoder_mask);
> > > +}
> >
> > Since it's next to each another double-checking that the fixup did add the
> > self-clone is probably too much :-)
>
> I changed the fixup to be just
> if (possible_clones == 0)
> possible_clones = drm_encoder_mask();
>
> So if the driver tries to set it up but fails and forgets the
> encoder itself this WARN will still trip.
Duh, I was just blind, everything is looking good.
-Daniel
>
> >
> > > +
> > > void drm_mode_config_validate(struct drm_device *dev)
> > > {
> > > struct drm_encoder *encoder;
> > >
> > > drm_for_each_encoder(encoder, dev)
> > > fixup_encoder_possible_clones(encoder);
> > > +
> > > + drm_for_each_encoder(encoder, dev)
> > > + validate_encoder_possible_clones(encoder);
> >
> > > }
> > > diff --git a/include/drm/drm_encoder.h b/include/drm/drm_encoder.h
> > > index 22d6cdf729f1..3741963b9587 100644
> > > --- a/include/drm/drm_encoder.h
> > > +++ b/include/drm/drm_encoder.h
> > > @@ -163,6 +163,8 @@ struct drm_encoder {
> > > * any cloning it can leave @possible_clones set to 0. The core will
> > > * automagically fix this up by setting the bit for the encoder itself.
> > > *
> > > + * You will get a WARN if you get this wrong in the driver.
> >
> > Nice.
> >
> > Reviewed-by: Daniel Vetter <daniel.vetter@xxxxxxxx>
> >
> > > + *
> > > * Note that since encoder objects can't be hotplugged the assigned indices
> > > * are stable and hence known before registering all objects.
> > > */
> > > --
> > > 2.24.1
> > >
> >
> > --
> > Daniel Vetter
> > Software Engineer, Intel Corporation
> > http://blog.ffwll.ch
>
> --
> Ville Syrjälä
> Intel
--
Daniel Vetter
Software Engineer, Intel Corporation
http://blog.ffwll.ch
_______________________________________________
Intel-gfx mailing list
Intel-gfx@xxxxxxxxxxxxxxxxxxxxx
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
