Message ID | 1353804080-25492-1-git-send-email-laurent.pinchart@ideasonboard.com (mailing list archive) |
---|---|
State | New, archived |
Headers | show |
Hi Laurent, On 11/25/2012 01:41 AM, Laurent Pinchart wrote: > When subdev registration fails the subdev v4l2_dev field is left to a > non-NULL value. Later calls to v4l2_device_unregister_subdev() will > consider the subdev as registered and will module_put() the subdev > module without any matching module_get(). > > Fix this by setting the subdev v4l2_dev field to NULL in > v4l2_device_register_subdev() when the function fails. > > Signed-off-by: Laurent Pinchart<laurent.pinchart@ideasonboard.com> Acked-by: Sylwester Nawrocki <s.nawrocki@samsung.com> I'm just wondering whether including this patch in stable kernel releases could potentially break anything. > --- > drivers/media/v4l2-core/v4l2-device.c | 30 ++++++++++++++---------------- > 1 files changed, 14 insertions(+), 16 deletions(-) > > diff --git a/drivers/media/v4l2-core/v4l2-device.c b/drivers/media/v4l2-core/v4l2-device.c > index 513969f..98a7f5e 100644 > --- a/drivers/media/v4l2-core/v4l2-device.c > +++ b/drivers/media/v4l2-core/v4l2-device.c > @@ -159,31 +159,21 @@ int v4l2_device_register_subdev(struct v4l2_device *v4l2_dev, > sd->v4l2_dev = v4l2_dev; > if (sd->internal_ops&& sd->internal_ops->registered) { > err = sd->internal_ops->registered(sd); > - if (err) { > - module_put(sd->owner); > - return err; > - } > + if (err) > + goto error_module; > } > > /* This just returns 0 if either of the two args is NULL */ > err = v4l2_ctrl_add_handler(v4l2_dev->ctrl_handler, sd->ctrl_handler, NULL); > - if (err) { > - if (sd->internal_ops&& sd->internal_ops->unregistered) > - sd->internal_ops->unregistered(sd); > - module_put(sd->owner); > - return err; > - } > + if (err) > + goto error_unregister; > > #if defined(CONFIG_MEDIA_CONTROLLER) > /* Register the entity. */ > if (v4l2_dev->mdev) { > err = media_device_register_entity(v4l2_dev->mdev, entity); > - if (err< 0) { > - if (sd->internal_ops&& sd->internal_ops->unregistered) > - sd->internal_ops->unregistered(sd); > - module_put(sd->owner); > - return err; > - } > + if (err< 0) > + goto error_unregister; > } > #endif > > @@ -192,6 +182,14 @@ int v4l2_device_register_subdev(struct v4l2_device *v4l2_dev, > spin_unlock(&v4l2_dev->lock); > > return 0; > + > +error_unregister: > + if (sd->internal_ops&& sd->internal_ops->unregistered) > + sd->internal_ops->unregistered(sd); > +error_module: > + module_put(sd->owner); > + sd->v4l2_dev = NULL; > + return err; > } > EXPORT_SYMBOL_GPL(v4l2_device_register_subdev); -- To unsubscribe from this list: send the line "unsubscribe linux-media" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
Hi Sylwester, On Tuesday 27 November 2012 23:54:40 Sylwester Nawrocki wrote: > On 11/25/2012 01:41 AM, Laurent Pinchart wrote: > > When subdev registration fails the subdev v4l2_dev field is left to a > > non-NULL value. Later calls to v4l2_device_unregister_subdev() will > > consider the subdev as registered and will module_put() the subdev > > module without any matching module_get(). > > > > Fix this by setting the subdev v4l2_dev field to NULL in > > v4l2_device_register_subdev() when the function fails. > > > > Signed-off-by: Laurent Pinchart<laurent.pinchart@ideasonboard.com> > > Acked-by: Sylwester Nawrocki <s.nawrocki@samsung.com> Thank you. > I'm just wondering whether including this patch in stable kernel releases > could potentially break anything. I don't think it would, the patch only touches error paths, and clearly fixes a bug. So Cc: stable@vger.kernel.org looks like a good idea.
diff --git a/drivers/media/v4l2-core/v4l2-device.c b/drivers/media/v4l2-core/v4l2-device.c index 513969f..98a7f5e 100644 --- a/drivers/media/v4l2-core/v4l2-device.c +++ b/drivers/media/v4l2-core/v4l2-device.c @@ -159,31 +159,21 @@ int v4l2_device_register_subdev(struct v4l2_device *v4l2_dev, sd->v4l2_dev = v4l2_dev; if (sd->internal_ops && sd->internal_ops->registered) { err = sd->internal_ops->registered(sd); - if (err) { - module_put(sd->owner); - return err; - } + if (err) + goto error_module; } /* This just returns 0 if either of the two args is NULL */ err = v4l2_ctrl_add_handler(v4l2_dev->ctrl_handler, sd->ctrl_handler, NULL); - if (err) { - if (sd->internal_ops && sd->internal_ops->unregistered) - sd->internal_ops->unregistered(sd); - module_put(sd->owner); - return err; - } + if (err) + goto error_unregister; #if defined(CONFIG_MEDIA_CONTROLLER) /* Register the entity. */ if (v4l2_dev->mdev) { err = media_device_register_entity(v4l2_dev->mdev, entity); - if (err < 0) { - if (sd->internal_ops && sd->internal_ops->unregistered) - sd->internal_ops->unregistered(sd); - module_put(sd->owner); - return err; - } + if (err < 0) + goto error_unregister; } #endif @@ -192,6 +182,14 @@ int v4l2_device_register_subdev(struct v4l2_device *v4l2_dev, spin_unlock(&v4l2_dev->lock); return 0; + +error_unregister: + if (sd->internal_ops && sd->internal_ops->unregistered) + sd->internal_ops->unregistered(sd); +error_module: + module_put(sd->owner); + sd->v4l2_dev = NULL; + return err; } EXPORT_SYMBOL_GPL(v4l2_device_register_subdev);
When subdev registration fails the subdev v4l2_dev field is left to a non-NULL value. Later calls to v4l2_device_unregister_subdev() will consider the subdev as registered and will module_put() the subdev module without any matching module_get(). Fix this by setting the subdev v4l2_dev field to NULL in v4l2_device_register_subdev() when the function fails. Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com> --- drivers/media/v4l2-core/v4l2-device.c | 30 ++++++++++++++---------------- 1 files changed, 14 insertions(+), 16 deletions(-)