Re: [PATCHv2] v4l2-ioctl.c: fix incorrect error path

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



Hi Hans,

Thank you for the patch.

On Mon, Mar 21, 2022 at 12:03:30PM +0100, Hans Verkuil wrote:
> If allocating array_buf fails, or copying data from userspace into that
> buffer fails, then just free memory and return the error. Don't attempt
> to call video_put_user() since there is no point.
> 
> If writing the array back to userspace fails, then don't go to out_array_args,
> instead just continue with the regular code that just returns the error
> unless 'always_copy' is set. That flag was just completely ignored when
> writing back the array data.
> 
> Update the VIDIOC_G/S/TRY_EXT_CTRLS ioctls to set the ALWAYS_COPY flag
> since they now need it.
> 
> Signed-off-by: Hans Verkuil <hverkuil-cisco@xxxxxxxxx>
> ---
> Changes since v1:
> - drop the now unused out_array_args label
> - set the INFO_FL_ALWAYS_COPY for the VIDIOC_G/S/TRY_EXT_CTRLS ioctls
> ---
>  drivers/media/v4l2-core/v4l2-ioctl.c | 12 +++++-------
>  1 file changed, 5 insertions(+), 7 deletions(-)
> 
> diff --git a/drivers/media/v4l2-core/v4l2-ioctl.c b/drivers/media/v4l2-core/v4l2-ioctl.c
> index 96e307fe3aab..3eb8f5f0958d 100644
> --- a/drivers/media/v4l2-core/v4l2-ioctl.c
> +++ b/drivers/media/v4l2-core/v4l2-ioctl.c
> @@ -2816,9 +2816,9 @@ static const struct v4l2_ioctl_info v4l2_ioctls[] = {
>  	IOCTL_INFO(VIDIOC_S_PRIORITY, v4l_s_priority, v4l_print_u32, INFO_FL_PRIO),
>  	IOCTL_INFO(VIDIOC_G_SLICED_VBI_CAP, v4l_g_sliced_vbi_cap, v4l_print_sliced_vbi_cap, INFO_FL_CLEAR(v4l2_sliced_vbi_cap, type)),
>  	IOCTL_INFO(VIDIOC_LOG_STATUS, v4l_log_status, v4l_print_newline, 0),
> -	IOCTL_INFO(VIDIOC_G_EXT_CTRLS, v4l_g_ext_ctrls, v4l_print_ext_controls, INFO_FL_CTRL),
> -	IOCTL_INFO(VIDIOC_S_EXT_CTRLS, v4l_s_ext_ctrls, v4l_print_ext_controls, INFO_FL_PRIO | INFO_FL_CTRL),
> -	IOCTL_INFO(VIDIOC_TRY_EXT_CTRLS, v4l_try_ext_ctrls, v4l_print_ext_controls, INFO_FL_CTRL),
> +	IOCTL_INFO(VIDIOC_G_EXT_CTRLS, v4l_g_ext_ctrls, v4l_print_ext_controls, INFO_FL_CTRL | INFO_FL_ALWAYS_COPY),
> +	IOCTL_INFO(VIDIOC_S_EXT_CTRLS, v4l_s_ext_ctrls, v4l_print_ext_controls, INFO_FL_PRIO | INFO_FL_CTRL | INFO_FL_ALWAYS_COPY),
> +	IOCTL_INFO(VIDIOC_TRY_EXT_CTRLS, v4l_try_ext_ctrls, v4l_print_ext_controls, INFO_FL_CTRL | INFO_FL_ALWAYS_COPY),
>  	IOCTL_INFO(VIDIOC_ENUM_FRAMESIZES, v4l_stub_enum_framesizes, v4l_print_frmsizeenum, INFO_FL_CLEAR(v4l2_frmsizeenum, pixel_format)),
>  	IOCTL_INFO(VIDIOC_ENUM_FRAMEINTERVALS, v4l_stub_enum_frameintervals, v4l_print_frmivalenum, INFO_FL_CLEAR(v4l2_frmivalenum, height)),
>  	IOCTL_INFO(VIDIOC_G_ENC_INDEX, v4l_stub_g_enc_index, v4l_print_enc_idx, 0),
> @@ -3311,7 +3311,7 @@ video_usercopy(struct file *file, unsigned int orig_cmd, unsigned long arg,
>  		array_buf = kvmalloc(array_size, GFP_KERNEL);
>  		err = -ENOMEM;
>  		if (array_buf == NULL)
> -			goto out_array_args;
> +			goto out;
>  		err = -EFAULT;
>  		if (in_compat_syscall())
>  			err = v4l2_compat_get_array_args(file, array_buf,
> @@ -3321,7 +3321,7 @@ video_usercopy(struct file *file, unsigned int orig_cmd, unsigned long arg,
>  			err = copy_from_user(array_buf, user_ptr, array_size) ?
>  								-EFAULT : 0;
>  		if (err)
> -			goto out_array_args;
> +			goto out;
>  		*kernel_ptr = array_buf;
>  	}
> 
> @@ -3353,7 +3353,6 @@ video_usercopy(struct file *file, unsigned int orig_cmd, unsigned long arg,
>  		} else if (copy_to_user(user_ptr, array_buf, array_size)) {
>  			err = -EFAULT;
>  		}
> -		goto out_array_args;
>  	}
>  	/*
>  	 * Some ioctls can return an error, but still have valid
> @@ -3362,7 +3361,6 @@ video_usercopy(struct file *file, unsigned int orig_cmd, unsigned long arg,
>  	if (err < 0 && !always_copy)
>  		goto out;
> 
> -out_array_args:
>  	if (video_put_user((void __user *)arg, parg, cmd, orig_cmd))
>  		err = -EFAULT;

Could we turn this into

  	if (err >= 0 || always_copy) {
  		if (video_put_user((void __user *)arg, parg, cmd, orig_cmd))
  			err = -EFAULT;
	}

to avoid the goto ?

With this,

Reviewed-by: Laurent Pinchart <laurent.pinchart@xxxxxxxxxxxxxxxx>

I also wonder if we should skip the copy to user for the array args when
an error occurs (unless always_copy is set). What do you think ? It's
not a candidate for this patch of course.

>  out:

-- 
Regards,

Laurent Pinchart



[Index of Archives]     [Linux Input]     [Video for Linux]     [Gstreamer Embedded]     [Mplayer Users]     [Linux USB Devel]     [Linux Audio Users]     [Linux Kernel]     [Linux SCSI]     [Yosemite Backpacking]

  Powered by Linux