Skip to contentSkip to conversation

media: usbtv: fix use-after-free and lockdep panic on unbind

PATCH New Svyatoslav Nikolenko 1 participant 1 message Sep 30 Media (V4L/DVB)

lore
Share & tools
lore
#0 lore raw

PATCH v2 1/1 media: usbtv: fix use-after-free and lockdep panic on unbind

When a usbtv device is unplugged or unbound, a lockdep panic and use-after-free occurs in the driver core.

The sequence of events leading to the crash is:
1. device_release_driver_internal() acquires dev->mutex.
2. usbtv_disconnect() is called, which in turn calls usbtv_video_free().
3. usbtv_video_free() prematurely calls v4l2_device_disconnect() and v4l2_device_put().
4. This drops the final reference on the parent USB device (intf->dev), triggering synchronous device deletion while the USB core is still holding dev->mutex.
5. When the core attempts to unlock dev->mutex, it operates on freed memory, causing a lockdep panic.

Furthermore, usbtv_disconnect() continues to access usbtv->udev after usbtv_video_free() returns, leading to a direct use-after-free.

Note that commit cb705f75d03e ("media: usbtv: fix null-pointer dereference on disconnect") recently added a mutex to usbtv_video_free() to resolve a separate issue, but left the premature v4l2_device calls intact. Because the final reference is still dropped too early, the lockdep panic still occurs.

Fix this by removing the manual v4l2_device_disconnect() and v4l2_device_put() calls from usbtv_video_free(). This allows usbtv_video_free() to safely unregister the video node without destroying the underlying structures. The final v4l2_device_put() at the end of usbtv_disconnect() will now correctly trigger usbtv_release() asynchronously, safely dropping the parent device reference only after all system locks are released.

As a result of this cleanup, the dummy v4l2_device_get() previously used to prevent premature deletion in the usbtv_probe() error path is no longer needed and has been removed.

Fixes a3550ea665ac ("[media] usbtv: split core and video implementation")
Reported-by [email protected]

---
Changes in v2:
- Rebased onto media-committers/next.
- Updated commit message to clarify why commit cb705f75d03e does not resolve this specific lockdep panic.

 drivers/media/usb/usbtv/usbtv-core.c  | 3 ---
 drivers/media/usb/usbtv/usbtv-video.c | 3 ---
 2 files changed, 6 deletions(-)
drivers/media/usb/usbtv/usbtv-core.c +0 -3
index 89c3424e7685..2cd982ca49bf 100644
@@ -112,9 +112,6 @@ static int usbtv_probe(struct usb_interface *intf,
 	return 0;
 
 usbtv_audio_fail:
-	/* we must not free at this point */
-	v4l2_device_get(&usbtv->v4l2_dev);
-	/* this will undo the v4l2_device_get() */
 	usbtv_video_free(usbtv);
 
 usbtv_video_fail:
drivers/media/usb/usbtv/usbtv-video.c +0 -3
index 2a91af8d1e30..999bce621c04 100644
@@ -964,11 +964,8 @@ int usbtv_video_init(struct usbtv *usbtv)
 void usbtv_video_free(struct usbtv *usbtv)
 {
 	vb2_video_unregister_device(&usbtv->vdev);
-	v4l2_device_disconnect(&usbtv->v4l2_dev);
 
 	mutex_lock(&usbtv->v4l2_lock);
 	usbtv->udev = NULL;
 	mutex_unlock(&usbtv->v4l2_lock);
-
-	v4l2_device_put(&usbtv->v4l2_dev);
 }
Signature · 1 line
2.47.3