David Ellingsworth wrote:
>> Hi Mariusz,
>>
>> You have the oops thanks to poison and it is not enabled in my kernel.
>>
>> I found the real bug: the device structure was part of the gspca device
>> and it was freed on close after webcam unplug while streaming.
>
> Jean-Francois,
>
> I reviewed your patch and in my opinion it is the wrong thing to do.
> With the recent modifications to v4l2 it is very safe and practical to
> embed the video_device struct within a driver struct. The containing
> structure however should not be freed until the release callback in
> the video_device structure is called. This callback is called after
> all open handles have been closed, it is no longer called immediately
> after video_unregister_device is called.
>
Correct, the v4l2 currently does exactly the same ref counting as we are doing
in the gspca core now this can be removed.
I'm working on a patch which uses the v4l2 core ref counting instead of our own.
Regards,
Hans
_______________________________________________
v4l-dvb-maintainer mailing list
v4l-dvb-maintainer@xxxxxxxxxxx
http://www.linuxtv.org/cgi-bin/mailman/listinfo/v4l-dvb-maintainer
[Linux Media]
[Older V4L]
[Linux DVB]
[Video Disk Recorder]
[Asterisk]
[Photo]
[DCCP]
[Netdev]
[Xorg]
[Util Linux NG]
[Xfree86]
[Free Photo Albums]
[Fedora Users]
[Fedora Women]
[ALSA Users]
[ALSA Devel]
[SSH]
[Linux USB]
 |
 |
-->