Re: [PATCH 2/5] drm: Use new mode_valid() helpers in connector probe helper

[Daniel Vetter <daniel@ffwll.ch>]

On Mon, May 08, 2017 at 11:13:37AM +0100, Jose Abreu wrote:
> Hi Daniel,
> 
> 
> On 08-05-2017 08:50, Daniel Vetter wrote:
> > On Thu, May 04, 2017 at 03:11:39PM +0100, Jose Abreu wrote:
> >> This changes the connector probe helper function to use the new
> >> encoder->mode_valid() and crtc->mode_valid() helper callbacks to
> >> validate the modes.
> >>
> >> The new callbacks are optional so the behaviour remains the same
> >> if they are not implemented. If they are, then the code loops
> >> through all the connector's encodersXcrtcs and calls the
> >> callback.
> >>
> >> If at least a valid encoderXcrtc combination is found which
> >> accepts the mode then the function returns MODE_OK.
> >>
> >> Signed-off-by: Jose Abreu <joabreu@synopsys.com>
> >> Cc: Carlos Palminha <palminha@synopsys.com>
> >> Cc: Alexey Brodkin <abrodkin@synopsys.com>
> >> Cc: Ville Syrjälä <ville.syrjala@linux.intel.com>
> >> Cc: Daniel Vetter <daniel.vetter@ffwll.ch>
> >> Cc: Dave Airlie <airlied@linux.ie>
> >> Cc: Andrzej Hajda <a.hajda@samsung.com>
> >> Cc: Archit Taneja <architt@codeaurora.org>
> > A few comments below.
> > -Daniel
> 
> Thanks for the review!
> 
> >
> >> ---
> >>  drivers/gpu/drm/drm_probe_helper.c | 71 +++++++++++++++++++++++++++++++++++---
> >>  1 file changed, 67 insertions(+), 4 deletions(-)
> >>
> >> diff --git a/drivers/gpu/drm/drm_probe_helper.c b/drivers/gpu/drm/drm_probe_helper.c
> >> index 1b0c14a..bf19f82 100644
> >> --- a/drivers/gpu/drm/drm_probe_helper.c
> >> +++ b/drivers/gpu/drm/drm_probe_helper.c
> >> @@ -80,6 +80,67 @@
> >>  	return MODE_OK;
> >>  }
> >>  
> >> +static enum drm_mode_status
> >> +drm_mode_validate_connector(struct drm_connector *connector,
> >> +			    struct drm_display_mode *mode)
> >> +{
> >> +	const struct drm_connector_helper_funcs *connector_funcs =
> >> +		connector->helper_private;
> >> +	struct drm_device *dev = connector->dev;
> >> +	uint32_t *ids = connector->encoder_ids;
> >> +	enum drm_mode_status ret = MODE_OK;
> >> +	unsigned int i;
> >> +
> >> +	/* Step 1: Validate against connector */
> >> +	if (connector_funcs && connector_funcs->mode_valid)
> >> +		ret = connector_funcs->mode_valid(connector, mode);
> >> +
> >> +	if (ret != MODE_OK)
> >> +		return ret;
> >> +
> >> +	/* Step 2: Validate against encoders and crtcs */
> >> +	for (i = 0; i < DRM_CONNECTOR_MAX_ENCODER; i++) {
> >> +		struct drm_encoder *encoder = drm_encoder_find(dev, ids[i]);
> >> +		const struct drm_encoder_helper_funcs *encoder_funcs;
> >> +		struct drm_crtc *crtc;
> >> +
> >> +		if (!encoder)
> >> +			continue;
> >> +
> >> +		encoder_funcs = encoder->helper_private;
> >> +		if (encoder_funcs && encoder_funcs->mode_valid)
> >> +			ret = encoder_funcs->mode_valid(encoder, mode);
> >> +		else
> >> +			ret = MODE_OK; /* encoder accepts everything */
> >> +
> >> +		/* No point in continuing for crtc check as this encoder will
> >> +		 * not accept the mode anyway. If all encoders reject the mode
> >> +		 * then, at exit, ret will not be MODE_OK. */
> >> +		if (ret != MODE_OK)
> >> +			continue;
> > I think we should validate encoders against connector->possible_ids here.
> > Might be that not many drivers fill this out correctly, and in case fixing
> > that is much work, perhaps as a follow-up. But would be good to at least
> > help look down that part of uapi a bit more.
> 
> I'm sorry but I'm not following you here (I think I need more
> coffee :)).
> 
> What do you mean by connector->possible_ids ? Is this something
> new ? Because I didn't find anything in my tree and I'm based on
> today's drm-next from Dave.

Oops, I guess I was on an old tree or whatever by accident. I meant
drm_connector->encoder_ids, that limits the encoders you can use on a
crtc. For many drivers there's only one.

> > This isn't checked within atomic and legacy helpers since we assume that
> > ->best_encoder respectively ->atomic_best_encoder gives us a valid encoder
> > ...
> >
> >> +
> >> +		drm_for_each_crtc(crtc, dev) {
> >> +			const struct drm_crtc_helper_funcs *crtc_funcs;
> >> +
> >> +			if (!drm_encoder_crtc_ok(encoder, crtc))
> >> +				continue;
> > We check this one here already in both atomic and legacy helpers, so all
> > drivers should get this right.
> 
> But drm_for_each_crtc() iterates over all crtc from a device and
> some crtcs may only be used by specific encoders (by setting
> encoder->possible_crtcs), right ? We need to iterate only over
> the encoder's crtc we want to validate.

This was a comment about ->encoder_ids, since we don't validate that in
the atomic helpers (but instead rely on ->(atomic_)best_encoder to give us
the right encoder) validating here in this function might be a problem and
uncover driver bugs. But for drm_encoder_crtc_ok there's no such risk, so
this is safe.

I was simply dumping my thoughts while reviewing, the code is all good :-)

Cheers, Daniel
-- 
Daniel Vetter
Software Engineer, Intel Corporation
http://blog.ffwll.ch