Re: [PATCH] drm/vc4: hdmi: Fix pointer dereference before check

[Maxime Ripard <maxime@cerno.tech>]

On Wed, Nov 02, 2022 at 12:10:03PM +0100, José Expósito wrote:
> Hi Maxime,
> 
> Thanks a lot for looking into the patch.
> 
> On Wed, Nov 02, 2022 at 10:01:53AM +0100, Maxime Ripard wrote:
> > Hi,
> > 
> > On Sat, Oct 29, 2022 at 11:34:13AM +0200, José Expósito wrote:
> > > Commit 6bed2ea3cb38 ("drm/vc4: hdmi: Reset link on hotplug") introduced
> > > the vc4_hdmi_reset_link() function. This function dereferences the
> > > "connector" pointer before checking whether it is NULL or not.
> > > 
> > > Rework variable assignment to avoid this issue.
> > > 
> > > Fixes: 6bed2ea3cb38 ("drm/vc4: hdmi: Reset link on hotplug")
> > > Signed-off-by: José Expósito <jose.exposito89@gmail.com>
> > > ---
> > >  drivers/gpu/drm/vc4/vc4_hdmi.c | 10 +++++++---
> > >  1 file changed, 7 insertions(+), 3 deletions(-)
> > > 
> > > diff --git a/drivers/gpu/drm/vc4/vc4_hdmi.c b/drivers/gpu/drm/vc4/vc4_hdmi.c
> > > index 4a73fafca51b..07d058b6afb7 100644
> > > --- a/drivers/gpu/drm/vc4/vc4_hdmi.c
> > > +++ b/drivers/gpu/drm/vc4/vc4_hdmi.c
> > > @@ -319,9 +319,9 @@ static int reset_pipe(struct drm_crtc *crtc,
> > >  static int vc4_hdmi_reset_link(struct drm_connector *connector,
> > >  			       struct drm_modeset_acquire_ctx *ctx)
> > >  {
> > > -	struct drm_device *drm = connector->dev;
> > > -	struct vc4_hdmi *vc4_hdmi = connector_to_vc4_hdmi(connector);
> > > -	struct drm_encoder *encoder = &vc4_hdmi->encoder.base;
> > > +	struct drm_device *drm;
> > > +	struct vc4_hdmi *vc4_hdmi;
> > > +	struct drm_encoder *encoder;
> > >  	struct drm_connector_state *conn_state;
> > >  	struct drm_crtc_state *crtc_state;
> > >  	struct drm_crtc *crtc;
> > > @@ -332,6 +332,10 @@ static int vc4_hdmi_reset_link(struct drm_connector *connector,
> > >  	if (!connector)
> > >  		return 0;
> > >  
> > > +	drm = connector->dev;
> > > +	vc4_hdmi = connector_to_vc4_hdmi(connector);
> > > +	encoder = &vc4_hdmi->encoder.base;
> > > +
> > 
> > I don't think that's right. Connector shouldn't be NULL to begin with,
> > how did you notice this?
> > 
> > Maxime
> 
> This issue was reported by Coverity. At the moment this function is not
> invoked with a NULL connector by any code path. However, since the NULL
> check is present, in my opinion, it makes sense to either remove it or
> make it usefull just in case the preconditions change in the future.

Yeah, it makes sense

I'd ask for a small cosmetic change then, could you add the assignments
where they are actually needed instead of at the top of the function?

Something like

 if (!connector)
 	return 0;

+drm = connector->dev;
 ret = drm_modeset_lock(&drm->mode_config.connection_mutex, ctx);

...

+vc4_hdmi = connector_to_vc4_hdmi(connector);
 if (!vc4_hdmi_supports_scrambling(vc4_hdmi))

...

Changing the prototype of vc4_hdmi_supports_scrambling to take a struct
vc4_hdmi pointer would also help, it's much more convenient.

Maxime