On 05/15/2018 05:46 PM, Oliver Neukum wrote: > Am Dienstag, den 15.05.2018, 16:28 +0200 schrieb Hans Verkuil: >> On 05/15/18 15:07, Oliver Neukum wrote: >>> The premature free in the error path is blocked by V4L >>> refcounting, not USB refcounting. Thanks to >>> Ben Hutchings for review. >>> >>> [v2] corrected attributions >>> >>> Signed-off-by: Oliver Neukum <oneukum@xxxxxxxx> >>> Fixes: 50e704453553 ("media: usbtv: prevent double free in error case") >>> CC: stable@xxxxxxxxxxxxxxx >>> Reported-by: Ben Hutchings <ben.hutchings@xxxxxxxxxxxxxxx> >>> --- >>> drivers/media/usb/usbtv/usbtv-core.c | 3 ++- >>> 1 file changed, 2 insertions(+), 1 deletion(-) >>> >>> diff --git a/drivers/media/usb/usbtv/usbtv-core.c b/drivers/media/usb/usbtv/usbtv-core.c >>> index 5095c380b2c1..4a03c4d66314 100644 >>> --- a/drivers/media/usb/usbtv/usbtv-core.c >>> +++ b/drivers/media/usb/usbtv/usbtv-core.c >>> @@ -113,7 +113,8 @@ static int usbtv_probe(struct usb_interface *intf, >>> >>> usbtv_audio_fail: >>> /* we must not free at this point */ >>> - usb_get_dev(usbtv->udev); >>> + v4l2_device_get(&usbtv->v4l2_dev); >> >> This is very confusing. I think it is much better to move the > > Yes. It confused me. > >> v4l2_device_register() call from usbtv_video_init to this probe function. > > Yes, but it is called here. So you want to do it after registering the > audio? No, before. It's a global data structure, so this can be done before the call to usbtv_video_init() as part of the top-level initialization of the driver. Regards, Hans