Re: [PATCH 1/1] v4l: ioctl: Validate num_planes before using it

[Hans Verkuil <hverkuil@xs4all.nl>]

On 2/15/19 4:52 PM, Ezequiel Garcia wrote:
> On Thu, 2019-01-10 at 14:43 +0200, Sakari Ailus wrote:
>> The for loop to reset the memory of the plane reserved fields runs over
>> num_planes provided by the user without validating it. Ensure num_planes
>> is no more than VIDEO_MAX_PLANES before the loop.
>>
>> Fixes: 4e1e0eb0e074 ("media: v4l2-ioctl: Zero v4l2_plane_pix_format reserved fields")
>> Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
>> ---
>> Hi folks,
>>
>> This patch goes on top of Thierry's patch "media: v4l2-ioctl: Clear only
>> per-plane reserved fields".
>>
>>  drivers/media/v4l2-core/v4l2-ioctl.c | 8 ++++++++
>>  1 file changed, 8 insertions(+)
>>
>> diff --git a/drivers/media/v4l2-core/v4l2-ioctl.c b/drivers/media/v4l2-core/v4l2-ioctl.c
>> index 392f1228af7b5..9e68a608ac6d3 100644
>> --- a/drivers/media/v4l2-core/v4l2-ioctl.c
>> +++ b/drivers/media/v4l2-core/v4l2-ioctl.c
>> @@ -1551,6 +1551,8 @@ static int v4l_s_fmt(const struct v4l2_ioctl_ops *ops,
>>  		if (unlikely(!ops->vidioc_s_fmt_vid_cap_mplane))
>>  			break;
>>  		CLEAR_AFTER_FIELD(p, fmt.pix_mp.xfer_func);
>> +		if (p->fmt.pix_mp.num_planes > VIDEO_MAX_PLANES)
>> +			break;
>>  		for (i = 0; i < p->fmt.pix_mp.num_planes; i++)
>>  			CLEAR_AFTER_FIELD(&p->fmt.pix_mp.plane_fmt[i], bytesperline);
>>  		return ops->vidioc_s_fmt_vid_cap_mplane(file, fh, arg);
>> @@ -1581,6 +1583,8 @@ static int v4l_s_fmt(const struct v4l2_ioctl_ops *ops,
>>  		if (unlikely(!ops->vidioc_s_fmt_vid_out_mplane))
>>  			break;
>>  		CLEAR_AFTER_FIELD(p, fmt.pix_mp.xfer_func);
>> +		if (p->fmt.pix_mp.num_planes > VIDEO_MAX_PLANES)
>> +			break;
>>  		for (i = 0; i < p->fmt.pix_mp.num_planes; i++)
>>  			CLEAR_AFTER_FIELD(&p->fmt.pix_mp.plane_fmt[i], bytesperline);
>>  		return ops->vidioc_s_fmt_vid_out_mplane(file, fh, arg);
>> @@ -1648,6 +1652,8 @@ static int v4l_try_fmt(const struct v4l2_ioctl_ops *ops,
>>  		if (unlikely(!ops->vidioc_try_fmt_vid_cap_mplane))
>>  			break;
>>  		CLEAR_AFTER_FIELD(p, fmt.pix_mp.xfer_func);
>> +		if (p->fmt.pix_mp.num_planes > VIDEO_MAX_PLANES)
>> +			break;
>>  		for (i = 0; i < p->fmt.pix_mp.num_planes; i++)
>>  			CLEAR_AFTER_FIELD(&p->fmt.pix_mp.plane_fmt[i], bytesperline);
>>  		return ops->vidioc_try_fmt_vid_cap_mplane(file, fh, arg);
>> @@ -1678,6 +1684,8 @@ static int v4l_try_fmt(const struct v4l2_ioctl_ops *ops,
>>  		if (unlikely(!ops->vidioc_try_fmt_vid_out_mplane))
>>  			break;
>>  		CLEAR_AFTER_FIELD(p, fmt.pix_mp.xfer_func);
>> +		if (p->fmt.pix_mp.num_planes > VIDEO_MAX_PLANES)
>> +			break;
>>  		for (i = 0; i < p->fmt.pix_mp.num_planes; i++)
>>  			CLEAR_AFTER_FIELD(&p->fmt.pix_mp.plane_fmt[i], bytesperline);
>>  		return ops->vidioc_try_fmt_vid_out_mplane(file, fh, arg);
> 
> I'm under the impression this commit is causing the following v4l2-compliance warning:
> 
> TRY_FMT cannot handle an invalid pixelformat.
> This may or may not be a problem. For more information see:
> http://www.mail-archive.com/linux-media@vger.kernel.org/msg56550.html
> 
> This applies to try_fmt and s_fmt.
> 
> After checking the v4l2 spec, it seems we can only error on a bad type field [1]:
> 
> "Drivers should not return an error code unless the type field is invalid".
> 
> Perhaps we can clamp num_planes to VIDEO_MAX_PLANES, instead of erroring out?

Yes, that would be the right approach.

Ezequiel, can you clamp num_planes and test the patch? And post it if it fixes the
warning?

This really should have been caught by testing with e.g. vivid and v4l2-compliance.

Regards,

	Hans

> 
> [1] https://www.kernel.org/doc/html/latest/media/uapi/v4l/vidioc-g-fmt.html
> 
> Regards,
> Eze
>