Re: [PATCH] drm/omap: dss: fix handling deferrals

[Andreas Kemnade <andreas@kemnade.info>]

On Wed, 11 Nov 2026 16:46:51 +0100
"Luca Ceresoli" <luca.ceresoli@bootlin.com> wrote:

> Hello Andreas,
> 
> On Wed Nov 11, 2026 at 3:56 PM CET, Tomi Valkeinen wrote:
> > Hi,
> >
> > On 11/11/2026 14:22, Andreas Kemnade wrote:  
> >> Forward the error given by of_drm_get_bridge_by_endpoint() if it is
> >> EPROBE_DEFER to ensure output gets initialized later when available.
> >>
> >> On GTA04/OMAP3 no display was found, but it appeared after doing
> >> rmmod omapdrm ; modprobe omapdrm
> >>
> >> Fixes: b7e9b93fc8ecf ("drm/omap: dss: don't create a panel_bridge")  
> 
> Ouch, sorry about breaking this driver. To my excuse the old code removed
> by b7e9b93fc8ecf was (ans still is) pretty obscure.
> 
> >> Signed-off-by: Andreas Kemnade <andreas@kemnade.info>
> >> ---
> >>   drivers/gpu/drm/omapdrm/dss/output.c | 5 +++++
> >>   1 file changed, 5 insertions(+)
> >>
> >> diff --git a/drivers/gpu/drm/omapdrm/dss/output.c b/drivers/gpu/drm/omapdrm/dss/output.c
> >> index 36d37deac8026..220ed7e003080 100644
> >> --- a/drivers/gpu/drm/omapdrm/dss/output.c
> >> +++ b/drivers/gpu/drm/omapdrm/dss/output.c
> >> @@ -24,6 +24,11 @@ int omapdss_device_init_output(struct omap_dss_device *out,
> >>
> >>   	bridge = of_drm_get_bridge_by_endpoint(out->dev->of_node, out->of_port, 0);
> >>   	if (IS_ERR(bridge)) {
> >> +		if (PTR_ERR(bridge) == -EPROBE_DEFER) {
> >> +			ret = PTR_ERR(bridge);
> >> +			goto error;
> >> +		}
> >> +  
> > Shouldn't we just return PTR_ERR(bridge) here? The error label will call
> > omapdss_device_cleanup_output(), which cleans up things set up later in
> > this function.  
> 
> Indeed.
> 
> 
> Additionally, by looking at the omapdss_device_init_output() code right
> now, the goto-based error management appears:
> 
>  1. unnecessarily complicated: only one goto for only one line of rollback
>     code
> 
>  2. potentially wrong, or at least doing too much, because
>     omapdss_device_cleanup_output() does two things:
> 
>       a. drm_bridge_put(out->next_bridge): but when goto is invoked the
>          out->next_bridge pointer has not been touched yet
>          (I assume we should not touch the refcount in this case)
> 
>       b. drm_bridge_put(out->bridge): unneeded because the goto is inside
>          an 'if (!out->bridge)', thus there is no reference at all to put
> 
> Do you agree with these findings?
> 
of_drm_get_bridge_by_endpoint never returns NULL (indicated by Sashiko).
So we have dead code there anyways. So the patch is dead code elimination
and not fixing bugs.

Regards,
Andreas