Hi Edgar,
Thank you for the patch respin.
I'm still a bit concerned about that -ENOMEM though:
On 06/10/17 11:34, Edgar Thier wrote:
>
> Use flags the device exposes for UVC controls.
> This allows the device to define which property flags are set.
>
> Since some cameras offer auto-adjustments for properties (e.g. auto-gain),
> the values of other properties (e.g. gain) can change in the camera.
> Examining the flags ensures that the driver is aware of such properties.
>
> Signed-off-by: Edgar Thier <i...@edgarthier.net>
> ---
> drivers/media/usb/uvc/uvc_ctrl.c | 56
> +++++++++++++++++++++++++++-------------
> 1 file changed, 38 insertions(+), 18 deletions(-)
>
> diff --git a/drivers/media/usb/uvc/uvc_ctrl.c
> b/drivers/media/usb/uvc/uvc_ctrl.c
> index 20397ab..5091086 100644
> --- a/drivers/media/usb/uvc/uvc_ctrl.c
> +++ b/drivers/media/usb/uvc/uvc_ctrl.c
> @@ -1630,6 +1630,41 @@ static void uvc_ctrl_fixup_xu_info(struct uvc_device
> *dev,
> }
>
> /*
> + * Retrieve flags for a given control
> + */
> +static int uvc_ctrl_get_flags(struct uvc_device *dev, const struct
> uvc_control *ctrl,
> + const struct uvc_control_info *info)
> +{
> + u8 *data;
> + int ret = 0;
> + int flags = 0;
> +
> + data = kmalloc(2, GFP_KERNEL);
> + if (data == NULL)
> + return -ENOMEM;
> +
> + ret = uvc_query_ctrl(dev, UVC_GET_INFO, ctrl->entity->id, dev->intfnum,
> + info->selector, data, 1);
> + if (ret < 0) {
> + uvc_trace(UVC_TRACE_CONTROL,
> + "GET_INFO failed on control %pUl/%u (%d).\n",
> + info->entity, info->selector, ret);
> + } else {
> + flags = UVC_CTRL_FLAG_GET_MIN | UVC_CTRL_FLAG_GET_MAX
> + | UVC_CTRL_FLAG_GET_RES | UVC_CTRL_FLAG_GET_DEF
> + | (data[0] & UVC_CONTROL_CAP_GET ?
> + UVC_CTRL_FLAG_GET_CUR : 0)
> + | (data[0] & UVC_CONTROL_CAP_SET ?
> + UVC_CTRL_FLAG_SET_CUR : 0)
> + | (data[0] & UVC_CONTROL_CAP_AUTOUPDATE ?
> + UVC_CTRL_FLAG_AUTO_UPDATE : 0);
> + }
> + kfree(data);
> + return flags;
> +}
> +
> +
(Minor nit) There's a double blank line there ...
> +/*
> * Query control information (size and flags) for XU controls.
> */
> static int uvc_ctrl_fill_xu_info(struct uvc_device *dev,
> @@ -1659,24 +1694,7 @@ static int uvc_ctrl_fill_xu_info(struct uvc_device
> *dev,
>
> info->size = le16_to_cpup((__le16 *)data);
>
> - /* Query the control information (GET_INFO) */
> - ret = uvc_query_ctrl(dev, UVC_GET_INFO, ctrl->entity->id, dev->intfnum,
> - info->selector, data, 1);
> - if (ret < 0) {
> - uvc_trace(UVC_TRACE_CONTROL,
> - "GET_INFO failed on control %pUl/%u (%d).\n",
> - info->entity, info->selector, ret);
> - goto done;
> - }
> -
> - info->flags = UVC_CTRL_FLAG_GET_MIN | UVC_CTRL_FLAG_GET_MAX
> - | UVC_CTRL_FLAG_GET_RES | UVC_CTRL_FLAG_GET_DEF
> - | (data[0] & UVC_CONTROL_CAP_GET ?
> - UVC_CTRL_FLAG_GET_CUR : 0)
> - | (data[0] & UVC_CONTROL_CAP_SET ?
> - UVC_CTRL_FLAG_SET_CUR : 0)
> - | (data[0] & UVC_CONTROL_CAP_AUTOUPDATE ?
> - UVC_CTRL_FLAG_AUTO_UPDATE : 0);
In this original code : info->flags is set based on the flags.
If the higher kmalloc failed, we wouldn't get here - and an error would have
already been returned to the caller.
> + info->flags = uvc_ctrl_get_flags(dev, ctrl, info);
But now ... info->flags could be set to an error return code if
uvc_ctrl_get_flags fails.
That error code is not checked anywhere and could be mis-interpreted as 'real
(invalid) flags' in other code....
And at no point - have we presented an error message to state that these flags
are now set ... but completely invalid.
> uvc_ctrl_fixup_xu_info(dev, ctrl, info);
>
> @@ -1902,6 +1920,8 @@ static int uvc_ctrl_add_info(struct uvc_device *dev,
> struct uvc_control *ctrl,
> goto done;
> }
>
> + ctrl->info.flags = uvc_ctrl_get_flags(dev, ctrl, info);
> +
And again here of course - now ctrl->info.flags could be set to -ENOMEM, or
rather -12 ... 0x(FFFFFFFF)FFFFFFF4
> ctrl->initialized = 1;
>
> uvc_trace(UVC_TRACE_CONTROL, "Added control %pUl/%u to device %s "
Regards
Kieran
