[PATCH v3 1/6] drm: bridge: Propagate the bus flags from bridge->timings

[Laurent Pinchart]

Hi Tomi,

On Mon, Nov 30, 2020 at 11:46:31AM +0200, Tomi Valkeinen wrote:
> On 30/11/2020 11:36, Laurent Pinchart wrote:
> > On Thu, Nov 19, 2020 at 09:31:29PM +0530, Nikhil Devshatwar wrote:
> >> bus_flags can be specified by a bridge in the timings.
> >> If the bridge provides it, Override the bus_flags when propagating
> >> from next bridge.
> >>
> >> Signed-off-by: Nikhil Devshatwar <nikhil.nd at ti.com>
> >> Reviewed-by: Tomi Valkeinen <tomi.valkeinen at ti.com>
> >> ---
> >>
> >> Notes:
> >>     changes from v2:
> >>     * update comment
> >>     changes from v1:
> >>     * Check for timings
> >>     * Prioritize timings flags over next bridge's flags
> >>
> >>  drivers/gpu/drm/drm_bridge.c | 8 ++++++++
> >>  1 file changed, 8 insertions(+)
> >>
> >> diff --git a/drivers/gpu/drm/drm_bridge.c b/drivers/gpu/drm/drm_bridge.c
> >> index 64f0effb52ac..13b67fc0dad3 100644
> >> --- a/drivers/gpu/drm/drm_bridge.c
> >> +++ b/drivers/gpu/drm/drm_bridge.c
> >> @@ -975,6 +975,14 @@ drm_atomic_bridge_propagate_bus_flags(struct drm_bridge *bridge,
> >>  	 * duplicate the "dummy propagation" logic.
> >>  	 */
> >>  	bridge_state->input_bus_cfg.flags = output_flags;
> >> +
> >> +	/*
> >> +	 * If legacy bus flags are provided in bridge->timings, use those as
> >> +	 * input flags instead of propagating the output flags.
> >> +	 */
> >> +	if (bridge->timings && bridge->timings->input_bus_flags)
> >> +		bridge_state->input_bus_cfg.flags =
> >> +			bridge->timings->input_bus_flags;
> > 
> > Hasn't Boris commented in his review of v1 that bus flags should be set
> > in atomic_check, even when they're static ? We're moving towards
> > removing timings->input_bus_flags, so this patch goes in the wrong
> > direction :-S
> 
> We have atomic_check only if the bridge has implemented atomic funcs. And even if there's
> atomic_check, not all bridges set the bus_flags there. So we need to either 1) fix the issue for now
> as in this patch, or 2) convert all bridges to use atomic funcs and fix all the bridges to set the
> bus_flags.

The second option is what we'd like to achieve. Wouldn't it be best to
already start going in that direction ? We don't need to convert all
bridge drivers in one go here, just the ones that are used by tidss.

-- 
Regards,

Laurent Pinchart