drivers/media/platform/video-mux.c | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-)
The current order of locking between the driver mutex and the v4l2
subdev state lock causes a circuluar locking dependency when trying to
set up a link. Fix this.
Signed-off-by: Paul Elder <paul.elder@ideasonboard.com>
---
drivers/media/platform/video-mux.c | 6 +++---
1 file changed, 3 insertions(+), 3 deletions(-)
diff --git a/drivers/media/platform/video-mux.c b/drivers/media/platform/video-mux.c
index 31e9e92e723eb..f43849db56800 100644
--- a/drivers/media/platform/video-mux.c
+++ b/drivers/media/platform/video-mux.c
@@ -52,6 +52,7 @@ static int video_mux_link_setup(struct media_entity *entity,
const struct media_pad *remote, u32 flags)
{
struct v4l2_subdev *sd = media_entity_to_v4l2_subdev(entity);
+ struct v4l2_subdev_state *sd_state;
struct video_mux *vmux = v4l2_subdev_to_video_mux(sd);
u16 source_pad = entity->num_pads - 1;
int ret = 0;
@@ -67,10 +68,10 @@ static int video_mux_link_setup(struct media_entity *entity,
remote->entity->name, remote->index, local->entity->name,
local->index, flags & MEDIA_LNK_FL_ENABLED);
+ sd_state = v4l2_subdev_lock_and_get_active_state(sd);
mutex_lock(&vmux->lock);
if (flags & MEDIA_LNK_FL_ENABLED) {
- struct v4l2_subdev_state *sd_state;
struct v4l2_mbus_framefmt *source_mbusformat;
if (vmux->active == local->index)
@@ -88,12 +89,10 @@ static int video_mux_link_setup(struct media_entity *entity,
vmux->active = local->index;
/* Propagate the active format to the source */
- sd_state = v4l2_subdev_lock_and_get_active_state(sd);
source_mbusformat = v4l2_subdev_state_get_format(sd_state,
source_pad);
*source_mbusformat = *v4l2_subdev_state_get_format(sd_state,
vmux->active);
- v4l2_subdev_unlock_state(sd_state);
} else {
if (vmux->active != local->index)
goto out;
@@ -105,6 +104,7 @@ static int video_mux_link_setup(struct media_entity *entity,
out:
mutex_unlock(&vmux->lock);
+ v4l2_subdev_unlock_state(sd_state);
return ret;
}
--
2.39.2
Hi Paul, Thank you for the patch. On Mon, Sep 09, 2024 at 05:48:28PM +0200, Paul Elder wrote: > The current order of locking between the driver mutex and the v4l2 > subdev state lock causes a circuluar locking dependency when trying to > set up a link. Fix this. > > Signed-off-by: Paul Elder <paul.elder@ideasonboard.com> Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com> This being said, I think we should deprecate the video-mux driver and implement a driver that uses the V4L2 subdev internal routing API instead of basing the routing configuration on link setup. Any opinion ? > --- > drivers/media/platform/video-mux.c | 6 +++--- > 1 file changed, 3 insertions(+), 3 deletions(-) > > diff --git a/drivers/media/platform/video-mux.c b/drivers/media/platform/video-mux.c > index 31e9e92e723eb..f43849db56800 100644 > --- a/drivers/media/platform/video-mux.c > +++ b/drivers/media/platform/video-mux.c > @@ -52,6 +52,7 @@ static int video_mux_link_setup(struct media_entity *entity, > const struct media_pad *remote, u32 flags) > { > struct v4l2_subdev *sd = media_entity_to_v4l2_subdev(entity); > + struct v4l2_subdev_state *sd_state; > struct video_mux *vmux = v4l2_subdev_to_video_mux(sd); > u16 source_pad = entity->num_pads - 1; > int ret = 0; > @@ -67,10 +68,10 @@ static int video_mux_link_setup(struct media_entity *entity, > remote->entity->name, remote->index, local->entity->name, > local->index, flags & MEDIA_LNK_FL_ENABLED); > > + sd_state = v4l2_subdev_lock_and_get_active_state(sd); > mutex_lock(&vmux->lock); > > if (flags & MEDIA_LNK_FL_ENABLED) { > - struct v4l2_subdev_state *sd_state; > struct v4l2_mbus_framefmt *source_mbusformat; > > if (vmux->active == local->index) > @@ -88,12 +89,10 @@ static int video_mux_link_setup(struct media_entity *entity, > vmux->active = local->index; > > /* Propagate the active format to the source */ > - sd_state = v4l2_subdev_lock_and_get_active_state(sd); > source_mbusformat = v4l2_subdev_state_get_format(sd_state, > source_pad); > *source_mbusformat = *v4l2_subdev_state_get_format(sd_state, > vmux->active); > - v4l2_subdev_unlock_state(sd_state); > } else { > if (vmux->active != local->index) > goto out; > @@ -105,6 +104,7 @@ static int video_mux_link_setup(struct media_entity *entity, > > out: > mutex_unlock(&vmux->lock); > + v4l2_subdev_unlock_state(sd_state); > return ret; > } > -- Regards, Laurent Pinchart
Hi Laurent, On Mon, Sep 09, 2024 at 06:56:03PM +0300, Laurent Pinchart wrote: > Hi Paul, > > Thank you for the patch. > > On Mon, Sep 09, 2024 at 05:48:28PM +0200, Paul Elder wrote: > > The current order of locking between the driver mutex and the v4l2 > > subdev state lock causes a circuluar locking dependency when trying to > > set up a link. Fix this. > > > > Signed-off-by: Paul Elder <paul.elder@ideasonboard.com> > > Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com> > > This being said, I think we should deprecate the video-mux driver and > implement a driver that uses the V4L2 subdev internal routing API > instead of basing the routing configuration on link setup. Any opinion ? Sounds good to me. But do we need a new driver? The subdev client streams capability flag should help here, just as it does on other drivers. I applied this one, with spelling fixed in the commit message. -- Kind regards, Sakari Ailus
On Thu, Oct 10, 2024 at 08:45:04AM +0000, Sakari Ailus wrote: > On Mon, Sep 09, 2024 at 06:56:03PM +0300, Laurent Pinchart wrote: > > On Mon, Sep 09, 2024 at 05:48:28PM +0200, Paul Elder wrote: > > > The current order of locking between the driver mutex and the v4l2 > > > subdev state lock causes a circuluar locking dependency when trying to > > > set up a link. Fix this. > > > > > > Signed-off-by: Paul Elder <paul.elder@ideasonboard.com> > > > > Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com> > > > > This being said, I think we should deprecate the video-mux driver and > > implement a driver that uses the V4L2 subdev internal routing API > > instead of basing the routing configuration on link setup. Any opinion ? > > Sounds good to me. But do we need a new driver? The subdev client streams > capability flag should help here, just as it does on other drivers. Good point, we could switch the driver behaviour depending on client capabilities. I like this ! > I applied this one, with spelling fixed in the commit message. -- Regards, Laurent Pinchart
© 2016 - 2024 Red Hat, Inc.