by Ricardo Ribalda

[permalink] [raw]

Subject: Re: [PATCH v9 16/22] media: uvcvideo: Return -EACCES to inactive controls

Hi Laurent

Thanks for your review!

On Thu, 10 Jun 2021 at 19:28, Laurent Pinchart
<[email protected]> wrote:
>
> Hi Ricardo,
>
> Thank you for the patch.
>
> On Fri, Mar 26, 2021 at 10:58:34AM +0100, Ricardo Ribalda wrote:
> > If a control is inactive return -EACCES to let the userspace know that
> > the value will not be applied automatically when the control is active
> > again.
>
> Isn't the device supposed to stall the control set in that case, with
> the bRequestErrorCode set to "Wrong state", which uvc_query_ctrl()
> translates to -EACCES already ? Could you elaborate on why this patch is
> needed ?

The problem is that the hardware was not returning the equivalent of
EACCES, so we had to capture it manually.
I will try to revert the patch and capture an error.

>
> > Suggested-by: Hans Verkuil <[email protected]>
> > Signed-off-by: Ricardo Ribalda <[email protected]>
> > ---
> > drivers/media/usb/uvc/uvc_ctrl.c | 71 +++++++++++++++++++++-----------
> > 1 file changed, 48 insertions(+), 23 deletions(-)
> >
> > diff --git a/drivers/media/usb/uvc/uvc_ctrl.c b/drivers/media/usb/uvc/uvc_ctrl.c
> > index bcebf9d1a46f..d9d4add1e813 100644
> > --- a/drivers/media/usb/uvc/uvc_ctrl.c
> > +++ b/drivers/media/usb/uvc/uvc_ctrl.c
> > @@ -1082,13 +1082,36 @@ static const char *uvc_map_get_name(const struct uvc_control_mapping *map)
> > return "Unknown Control";
> > }
> >
> > +static bool uvc_ctrl_is_inactive(struct uvc_video_chain *chain,
> > + struct uvc_control *ctrl,
> > + struct uvc_control_mapping *mapping)
> > +{
> > + struct uvc_control_mapping *master_map = NULL;
> > + struct uvc_control *master_ctrl = NULL;
> > + s32 val;
> > + int ret;
> > +
> > + if (!mapping->master_id)
> > + return false;
> > +
> > + __uvc_find_control(ctrl->entity, mapping->master_id, &master_map,
> > + &master_ctrl, 0);
> > +
> > + if (!master_ctrl || !(master_ctrl->info.flags & UVC_CTRL_FLAG_GET_CUR))
> > + return false;
> > +
> > + ret = __uvc_ctrl_get(chain, master_ctrl, master_map, &val);
> > + if (ret < 0 || val == mapping->master_manual)
> > + return false;
> > +
> > + return true;
> > +}
> > +
> > static int __uvc_query_v4l2_ctrl(struct uvc_video_chain *chain,
> > struct uvc_control *ctrl,
> > struct uvc_control_mapping *mapping,
> > struct v4l2_queryctrl *v4l2_ctrl)
> > {
> > - struct uvc_control_mapping *master_map = NULL;
> > - struct uvc_control *master_ctrl = NULL;
> > const struct uvc_menu_info *menu;
> > unsigned int i;
> >
> > @@ -1104,18 +1127,8 @@ static int __uvc_query_v4l2_ctrl(struct uvc_video_chain *chain,
> > if (!(ctrl->info.flags & UVC_CTRL_FLAG_SET_CUR))
> > v4l2_ctrl->flags |= V4L2_CTRL_FLAG_READ_ONLY;
> >
> > - if (mapping->master_id)
> > - __uvc_find_control(ctrl->entity, mapping->master_id,
> > - &master_map, &master_ctrl, 0);
> > - if (master_ctrl && (master_ctrl->info.flags & UVC_CTRL_FLAG_GET_CUR)) {
> > - s32 val;
> > - int ret = __uvc_ctrl_get(chain, master_ctrl, master_map, &val);
> > - if (ret < 0)
> > - return ret;
>
> There's a small change in behaviour here, the driver used to return an
> error, now it will ignore it. Is it intentional ?

AFAIK The error did not follow the v4l2 spec. You shall always be able
to query a ctrl.

I will add it to the commit message to make it clear.

>
> > -
> > - if (val != mapping->master_manual)
> > - v4l2_ctrl->flags |= V4L2_CTRL_FLAG_INACTIVE;
> > - }
> > + if (uvc_ctrl_is_inactive(chain, ctrl, mapping))
> > + v4l2_ctrl->flags |= V4L2_CTRL_FLAG_INACTIVE;
> >
> > if (!ctrl->cached) {
> > int ret = uvc_ctrl_populate_cache(chain, ctrl);
> > @@ -1638,25 +1651,37 @@ static int uvc_ctrl_commit_entity(struct uvc_device *dev,
> > return 0;
> > }
> >
> > -static int uvc_ctrl_find_ctrlidx(struct uvc_entity *entity,
> > +static int uvc_ctrl_commit_error(struct uvc_video_chain *chain,
> > + struct uvc_entity *entity,
> > struct v4l2_ext_controls *ctrls,
> > - struct uvc_control *uvc_control)
> > + struct uvc_control *err_control,
> > + int ret)
> > {
> > struct uvc_control_mapping *mapping;
> > struct uvc_control *ctrl_found;
> > unsigned int i;
> >
> > - if (!entity)
> > - return ctrls->count;
> > + if (!entity) {
> > + ctrls->error_idx = ctrls->count;
> > + return ret;
> > + }
> >
> > for (i = 0; i < ctrls->count; i++) {
> > __uvc_find_control(entity, ctrls->controls[i].id, &mapping,
> > &ctrl_found, 0);
> > - if (uvc_control == ctrl_found)
> > - return i;
> > + if (err_control == ctrl_found)
> > + break;
> > }
> > + ctrls->error_idx = i;
>
> I think this line should be moved after the next check.

Not really, if we cannot find a control, we cannot blame it on control 0 ;)

>
> > +
> > + /* We could not find the control that failed. */
> > + if (i == ctrls->count)
> > + return ret;
> >
> > - return ctrls->count;
> > + if (uvc_ctrl_is_inactive(chain, err_control, mapping))
> > + return -EACCES;
> > +
> > + return ret;
> > }
> >
> > int __uvc_ctrl_commit(struct uvc_fh *handle, int rollback,
> > @@ -1679,8 +1704,8 @@ int __uvc_ctrl_commit(struct uvc_fh *handle, int rollback,
> > uvc_ctrl_send_events(handle, ctrls->controls, ctrls->count);
> > done:
> > if (ret < 0 && ctrls)
> > - ctrls->error_idx = uvc_ctrl_find_ctrlidx(entity, ctrls,
> > - err_ctrl);
> > + ret = uvc_ctrl_commit_error(chain, entity, ctrls, err_ctrl,
> > + ret);
> > mutex_unlock(&chain->ctrl_mutex);
> > return ret;
> > }
>
> --
> Regards,
>
> Laurent Pinchart

--
Ricardo Ribalda

2021-06-10 19:51:39

by Ricardo Ribalda Delgado

[permalink] [raw]

Subject: Re: [PATCH v9 16/22] media: uvcvideo: Return -EACCES to inactive controls

Hi Laurent

On Thu, Jun 10, 2021 at 8:43 PM Ricardo Ribalda <[email protected]> wrote:
>
> Hi Laurent
>
> Thanks for your review!
>
> On Thu, 10 Jun 2021 at 19:28, Laurent Pinchart
> <[email protected]> wrote:
> >
> > Hi Ricardo,
> >
> > Thank you for the patch.
> >
> > On Fri, Mar 26, 2021 at 10:58:34AM +0100, Ricardo Ribalda wrote:
> > > If a control is inactive return -EACCES to let the userspace know that
> > > the value will not be applied automatically when the control is active
> > > again.
> >
> > Isn't the device supposed to stall the control set in that case, with
> > the bRequestErrorCode set to "Wrong state", which uvc_query_ctrl()
> > translates to -EACCES already ? Could you elaborate on why this patch is
> > needed ?
>
> The problem is that the hardware was not returning the equivalent of
> EACCES, so we had to capture it manually.
> I will try to revert the patch and capture an error.

In my hardware after reverting the patch I am still passing v4l2-compliance.

The benefits of this patch are:

1) We return -EACCESS even if the hardware mistakenly returns a different error
2) query_ctrl returns a value for hardware errors.

Hans can you comment on this? Shall we remove the patch ?

Thanks!

>
> >
> > > Suggested-by: Hans Verkuil <[email protected]>
> > > Signed-off-by: Ricardo Ribalda <[email protected]>
> > > ---
> > > drivers/media/usb/uvc/uvc_ctrl.c | 71 +++++++++++++++++++++-----------
> > > 1 file changed, 48 insertions(+), 23 deletions(-)
> > >
> > > diff --git a/drivers/media/usb/uvc/uvc_ctrl.c b/drivers/media/usb/uvc/uvc_ctrl.c
> > > index bcebf9d1a46f..d9d4add1e813 100644
> > > --- a/drivers/media/usb/uvc/uvc_ctrl.c
> > > +++ b/drivers/media/usb/uvc/uvc_ctrl.c
> > > @@ -1082,13 +1082,36 @@ static const char *uvc_map_get_name(const struct uvc_control_mapping *map)
> > > return "Unknown Control";
> > > }
> > >
> > > +static bool uvc_ctrl_is_inactive(struct uvc_video_chain *chain,
> > > + struct uvc_control *ctrl,
> > > + struct uvc_control_mapping *mapping)
> > > +{
> > > + struct uvc_control_mapping *master_map = NULL;
> > > + struct uvc_control *master_ctrl = NULL;
> > > + s32 val;
> > > + int ret;
> > > +
> > > + if (!mapping->master_id)
> > > + return false;
> > > +
> > > + __uvc_find_control(ctrl->entity, mapping->master_id, &master_map,
> > > + &master_ctrl, 0);
> > > +
> > > + if (!master_ctrl || !(master_ctrl->info.flags & UVC_CTRL_FLAG_GET_CUR))
> > > + return false;
> > > +
> > > + ret = __uvc_ctrl_get(chain, master_ctrl, master_map, &val);
> > > + if (ret < 0 || val == mapping->master_manual)
> > > + return false;
> > > +
> > > + return true;
> > > +}
> > > +
> > > static int __uvc_query_v4l2_ctrl(struct uvc_video_chain *chain,
> > > struct uvc_control *ctrl,
> > > struct uvc_control_mapping *mapping,
> > > struct v4l2_queryctrl *v4l2_ctrl)
> > > {
> > > - struct uvc_control_mapping *master_map = NULL;
> > > - struct uvc_control *master_ctrl = NULL;
> > > const struct uvc_menu_info *menu;
> > > unsigned int i;
> > >
> > > @@ -1104,18 +1127,8 @@ static int __uvc_query_v4l2_ctrl(struct uvc_video_chain *chain,
> > > if (!(ctrl->info.flags & UVC_CTRL_FLAG_SET_CUR))
> > > v4l2_ctrl->flags |= V4L2_CTRL_FLAG_READ_ONLY;
> > >
> > > - if (mapping->master_id)
> > > - __uvc_find_control(ctrl->entity, mapping->master_id,
> > > - &master_map, &master_ctrl, 0);
> > > - if (master_ctrl && (master_ctrl->info.flags & UVC_CTRL_FLAG_GET_CUR)) {
> > > - s32 val;
> > > - int ret = __uvc_ctrl_get(chain, master_ctrl, master_map, &val);
> > > - if (ret < 0)
> > > - return ret;
> >
> > There's a small change in behaviour here, the driver used to return an
> > error, now it will ignore it. Is it intentional ?
>
> AFAIK The error did not follow the v4l2 spec. You shall always be able
> to query a ctrl.
>
> I will add it to the commit message to make it clear.
>
> >
> > > -
> > > - if (val != mapping->master_manual)
> > > - v4l2_ctrl->flags |= V4L2_CTRL_FLAG_INACTIVE;
> > > - }
> > > + if (uvc_ctrl_is_inactive(chain, ctrl, mapping))
> > > + v4l2_ctrl->flags |= V4L2_CTRL_FLAG_INACTIVE;
> > >
> > > if (!ctrl->cached) {
> > > int ret = uvc_ctrl_populate_cache(chain, ctrl);
> > > @@ -1638,25 +1651,37 @@ static int uvc_ctrl_commit_entity(struct uvc_device *dev,
> > > return 0;
> > > }
> > >
> > > -static int uvc_ctrl_find_ctrlidx(struct uvc_entity *entity,
> > > +static int uvc_ctrl_commit_error(struct uvc_video_chain *chain,
> > > + struct uvc_entity *entity,
> > > struct v4l2_ext_controls *ctrls,
> > > - struct uvc_control *uvc_control)
> > > + struct uvc_control *err_control,
> > > + int ret)
> > > {
> > > struct uvc_control_mapping *mapping;
> > > struct uvc_control *ctrl_found;
> > > unsigned int i;
> > >
> > > - if (!entity)
> > > - return ctrls->count;
> > > + if (!entity) {
> > > + ctrls->error_idx = ctrls->count;
> > > + return ret;
> > > + }
> > >
> > > for (i = 0; i < ctrls->count; i++) {
> > > __uvc_find_control(entity, ctrls->controls[i].id, &mapping,
> > > &ctrl_found, 0);
> > > - if (uvc_control == ctrl_found)
> > > - return i;
> > > + if (err_control == ctrl_found)
> > > + break;
> > > }
> > > + ctrls->error_idx = i;
> >
> > I think this line should be moved after the next check.
>
> Not really, if we cannot find a control, we cannot blame it on control 0 ;)
>
> >
> > > +
> > > + /* We could not find the control that failed. */
> > > + if (i == ctrls->count)
> > > + return ret;
> > >
> > > - return ctrls->count;
> > > + if (uvc_ctrl_is_inactive(chain, err_control, mapping))
> > > + return -EACCES;
> > > +
> > > + return ret;
> > > }
> > >
> > > int __uvc_ctrl_commit(struct uvc_fh *handle, int rollback,
> > > @@ -1679,8 +1704,8 @@ int __uvc_ctrl_commit(struct uvc_fh *handle, int rollback,
> > > uvc_ctrl_send_events(handle, ctrls->controls, ctrls->count);
> > > done:
> > > if (ret < 0 && ctrls)
> > > - ctrls->error_idx = uvc_ctrl_find_ctrlidx(entity, ctrls,
> > > - err_ctrl);
> > > + ret = uvc_ctrl_commit_error(chain, entity, ctrls, err_ctrl,
> > > + ret);
> > > mutex_unlock(&chain->ctrl_mutex);
> > > return ret;
> > > }
> >
> > --
> > Regards,
> >
> > Laurent Pinchart
>
>
>
> --
> Ricardo Ribalda

--
Ricardo Ribalda