Em 13-12-2010 13:59, iceberg escreveu:
> KERNEL_VERSION: 2.6.36
> SUBJECT: double mutex_lock in drivers/media/video/tlg2300/pd-video.c
> in function vidioc_s_fmt
> SUBSCRIBE:
> First mutex_unlock in function pd_vidioc_s_fmt in line 767:
>
> 764 ret |= send_set_req(pd, VIDEO_ROSOLU_SEL,
> 765 vid_resol, &cmd_status);
> 766 if (ret || cmd_status) {
> 767 mutex_unlock(&pd->lock);
> 768 return -EBUSY;
> 769 }
>
> Second mutex_unlock in function vidioc_s_fmt in line 806:
>
> 805 pd_vidioc_s_fmt(pd, &f->fmt.pix);
> 806 mutex_unlock(&pd->lock);
>
> Found by Linux Device Drivers Verification Project
>
> ?hecks the return code of pd_vidioc_s_fm before mutex_unlocking.
>
> Signed-off-by: Alexander Strakh <strakh@xxxxxxxxx>
>
> ---
> --
> To unsubscribe from this list: send the line "unsubscribe linux-media" in
> the body of a message to majordomo@xxxxxxxxxxxxxxx
> More majordomo info at http://vger.kernel.org/majordomo-info.html
>
> diff --git a/drivers/media/video/tlg2300/pd-video.c
> b/drivers/media/video/tlg2300/pd-video.c
> index a1ffe18..fe6bd2b 100644
> --- a/drivers/media/video/tlg2300/pd-video.c
> +++ b/drivers/media/video/tlg2300/pd-video.c
> @@ -802,8 +802,8 @@ static int vidioc_s_fmt(struct file *file, void *fh,
> struct v4l2_format *f)
> return -EINVAL;
> }
>
> - pd_vidioc_s_fmt(pd, &f->fmt.pix);
> - mutex_unlock(&pd->lock);
> + if (!pd_vidioc_s_fmt(pd, &f->fmt.pix))
> + mutex_unlock(&pd->lock);
> return 0;
> }
>
Thanks for the patch, but the better is to keep the calls to mutex_lock/mutex_unlock
at the same function. So, instead of adding an "if" at vidioc_s_fmt, please remove
the mutex_unlock() from pd_vidioc_s_fmt().
Cheers,
Mauro
--
To unsubscribe from this list: send the line "unsubscribe linux-media" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
