On 11/16/2014 05:40 PM, Sakari Ailus wrote:
> Hi Hans,
>
> Thank you for the review.
>
> On Fri, Nov 14, 2014 at 09:46:54AM +0100, Hans Verkuil wrote:
>> On 11/09/2014 12:04 AM, Sakari Ailus wrote:
>>> The V4L2_SEL_TGT_NATIVE_SIZE target is used to denote e.g. the size of a
>>> sensor's pixel array.
>>>
>>> Signed-off-by: Sakari Ailus <[email protected]>
>>> ---
>>> Documentation/DocBook/media/v4l/selections-common.xml | 8 ++++++++
>>> include/uapi/linux/v4l2-common.h | 2 ++
>>> 2 files changed, 10 insertions(+)
>>>
>>> diff --git a/Documentation/DocBook/media/v4l/selections-common.xml
>>> b/Documentation/DocBook/media/v4l/selections-common.xml
>>> index 7502f78..5fc833a 100644
>>> --- a/Documentation/DocBook/media/v4l/selections-common.xml
>>> +++ b/Documentation/DocBook/media/v4l/selections-common.xml
>>> @@ -63,6 +63,14 @@
>>> <entry>Yes</entry>
>>> </row>
>>> <row>
>>> + <entry><constant>V4L2_SEL_TGT_NATIVE_SIZE</constant></entry>
>>> + <entry>0x0003</entry>
>>> + <entry>The native size of the device, e.g. a sensor's
>>> + pixel array.</entry>
>>
>> You might want to state that top and left are always 0.
>
> Fixed. I also added a patch to fix this in the smiapp driver --- the values
> were uninitialised. :-P
>
>>> + <entry>Yes</entry>
>>> + <entry>Yes</entry>
>>> + </row>
>>> + <row>
>>> <entry><constant>V4L2_SEL_TGT_COMPOSE</constant></entry>
>>> <entry>0x0100</entry>
>>> <entry>Compose rectangle. Used to configure scaling
>>> diff --git a/include/uapi/linux/v4l2-common.h
>>> b/include/uapi/linux/v4l2-common.h
>>> index 2f6f8ca..1527398 100644
>>> --- a/include/uapi/linux/v4l2-common.h
>>> +++ b/include/uapi/linux/v4l2-common.h
>>> @@ -43,6 +43,8 @@
>>> #define V4L2_SEL_TGT_CROP_DEFAULT 0x0001
>>> /* Cropping bounds */
>>> #define V4L2_SEL_TGT_CROP_BOUNDS 0x0002
>>> +/* Native frame size */
>>> +#define V4L2_SEL_TGT_NATIVE_SIZE 0x0003
>>> /* Current composing area */
>>> #define V4L2_SEL_TGT_COMPOSE 0x0100
>>> /* Default composing area */
>>>
>>
>> I like this. This would also make it possible to set the 'canvas' size of an
>> mem2mem device. Currently calling S_FMT for a mem2mem device cannot setup any
>> scaler since there is no native size. Instead S_FMT effectively *sets* the
>> native
>> size. The same is true for webcams with a scaler, which is why you added
>> this in
>> the first place. Obviously for sensors this target is read-only, but for a
>> mem2mem
>> device it can be writable as well.
>>
>> However, to make full use of this you also need to add input and output
>> capabilities if the native size can be set:
>>
>> V4L2_IN_CAP_NATIVE_SIZE
>> V4L2_OUT_CAP_NATIVE_SIZE
>
> Do you think this would require a capability flag, rather than just
> returning an error if the target is unsettable, as we otherwise already do
> if a selection target isn't supported? For the compound controls it's even
> easier, you just don't have a read-only flag set in the equivalent control.
No, I really want a capability flag here. Otherwise applications would have to
call ENUMINPUT *and* call G_SELECTION followed by S_SELECTION just to test if
it can be set. Besides, this is also a per-input capability, so you want to get
hold of the capabilities without having to do a S_INPUT first. I.e. you don't
want to have to do this:
// pseudo code just to give the idea
for (i = 0; i < max_input; i++) {
struct v4l2_selection sel = { NATIVE_SIZE };
ioctl(S_INPUT, i);
ioctl(G_SELECTION, &sel);
if (ioctl(S_SELECTION, &sel))
// does not support NATIVE_SIZE
}
>
>> (see ENUMINPUT/ENUMOUTPUT)
>>
>> This would nicely fill in a hole in the V4L2 Spec.
>
Regards,
Hans
--
To unsubscribe from this list: send the line "unsubscribe linux-media" in
the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html
