Hi Ricardo,
Thank you for the patch.
On Mon, Dec 21, 2020 at 05:48:08PM +0100, Ricardo Ribalda wrote:
> The control and its data needs to be copied to the workqueue at the same
> time to avoid half-updates of the events.
> This is, events reported to userspace were the control id does not match
> its value.
Actually, after discussing this with you on IRC, I'm not sure there's a
problem. The URB is resubmitted by uvc_ctrl_status_event_work(), so the
data shouldn't be overwritten before it is processed.
> Signed-off-by: Ricardo Ribalda <riba...@chromium.org>
> ---
> drivers/media/usb/uvc/uvc_ctrl.c | 2 +-
> drivers/media/usb/uvc/uvcvideo.h | 2 +-
> 2 files changed, 2 insertions(+), 2 deletions(-)
>
> diff --git a/drivers/media/usb/uvc/uvc_ctrl.c
> b/drivers/media/usb/uvc/uvc_ctrl.c
> index 011e69427b7c..aa18dcdf8165 100644
> --- a/drivers/media/usb/uvc/uvc_ctrl.c
> +++ b/drivers/media/usb/uvc/uvc_ctrl.c
> @@ -1332,7 +1332,7 @@ bool uvc_ctrl_status_event(struct urb *urb, struct
> uvc_video_chain *chain,
> return false;
> }
>
> - w->data = data;
> + memcpy(w->data, data, ctrl->info.size);
> w->urb = urb;
> w->chain = chain;
> w->ctrl = ctrl;
> diff --git a/drivers/media/usb/uvc/uvcvideo.h
> b/drivers/media/usb/uvc/uvcvideo.h
> index a3dfacf069c4..0db6c2e0bd98 100644
> --- a/drivers/media/usb/uvc/uvcvideo.h
> +++ b/drivers/media/usb/uvc/uvcvideo.h
> @@ -678,7 +678,7 @@ struct uvc_device {
> struct urb *urb;
> struct uvc_video_chain *chain;
> struct uvc_control *ctrl;
> - const void *data;
> + u8 data[UVC_MAX_STATUS_SIZE];
> } async_ctrl;
> };
>
--
Regards,
Laurent Pinchart
