On 5/29/19 5:36 PM, Ezequiel Garcia wrote:
> On Wed, 2019-05-29 at 16:41 +0200, Hans Verkuil wrote:
>> On 5/3/19 1:42 PM, Boris Brezillon wrote:
>>> Make sure the default value at least passes the std_validate() tests.
>>>
>>> Signed-off-by: Boris Brezillon <[email protected]>
>>> ---
>>> drivers/media/v4l2-core/v4l2-ctrls.c | 20 +++++++++++++++++++-
>>> 1 file changed, 19 insertions(+), 1 deletion(-)
>>>
>>> diff --git a/drivers/media/v4l2-core/v4l2-ctrls.c
>>> b/drivers/media/v4l2-core/v4l2-ctrls.c
>>> index b1ae2e555c68..19d40cc6e565 100644
>>> --- a/drivers/media/v4l2-core/v4l2-ctrls.c
>>> +++ b/drivers/media/v4l2-core/v4l2-ctrls.c
>>> @@ -1461,7 +1461,14 @@ static bool std_equal(const struct v4l2_ctrl *ctrl,
>>> u32 idx,
>>> static void std_init(const struct v4l2_ctrl *ctrl, u32 idx,
>>> union v4l2_ctrl_ptr ptr)
>>> {
>>> - switch (ctrl->type) {
>>> + struct v4l2_ctrl_mpeg2_slice_params *p_mpeg2_slice_params;
>>> +
>>> + /*
>>> + * The cast is needed to get rid of a gcc warning complaining that
>>> + * V4L2_CTRL_TYPE_MPEG2_SLICE_PARAMS is not part of the
>>> + * v4l2_ctrl_type enum.
>>> + */
>>> + switch ((u32)ctrl->type) {
>>> case V4L2_CTRL_TYPE_STRING:
>>> idx *= ctrl->elem_size;
>>> memset(ptr.p_char + idx, ' ', ctrl->minimum);
>>> @@ -1486,6 +1493,17 @@ static void std_init(const struct v4l2_ctrl *ctrl,
>>> u32 idx,
>>> case V4L2_CTRL_TYPE_U32:
>>> ptr.p_u32[idx] = ctrl->default_value;
>>> break;
>>> + case V4L2_CTRL_TYPE_MPEG2_SLICE_PARAMS:
>>> + p_mpeg2_slice_params = ptr.p;
>>> + /* 4:2:0 */
>>> + p_mpeg2_slice_params->sequence.chroma_format = 1;
>>> + /* 8 bits */
>>> + p_mpeg2_slice_params->picture.intra_dc_precision = 0;
>>> + /* interlaced top field */
>>> + p_mpeg2_slice_params->picture.picture_structure = 1;
>>> + p_mpeg2_slice_params->picture.picture_coding_type =
>>> + V4L2_MPEG2_PICTURE_CODING_TYPE_I;
>>
>> Oops, this isn't complete. It should still zero the p_mpeg2_slice_params
>> struct first. Right now any fields not explicitly set just have whatever
>> was in memory.
>>
>> Can you post a patch fixing this?
>>
>>
>
>
> I was wondering if we want to zero all the cases, and not just
> the struct types ones.
The others either overwrite the data with the default_value, or memset
the whole control (default case). It's only for these compound controls
that something special needs to be done.
The code can be restructured, though: instead of break do return in all
the simple type cases.
Then call memset followed by a new switch for the compound types where you
need to set some fields to non-zero.
>
> Or a different approach would be to define statically-allocated default
> structs, and init-assign to them. Although this take more space,
> and has no benefits, except maybe code clarity.
>
> OTOH, I can push a quick fix if you want, just zeroing this struct.
>
A quick fix is fine, but as we add new compound types here we need something
along the lines of what I described above. A statically-allocated default
struct isn't a bad idea either, but a bit overkill for the simple
mpeg2_slice_params.
Regards,
Hans
