On 12/12/18 9:43 AM, Hans Verkuil wrote:
> On 12/12/18 9:18 AM, Mauro Carvalho Chehab wrote:
>>> /* ioctls */
>>> @@ -368,6 +446,8 @@ struct media_v2_topology {
>>> #define MEDIA_IOC_ENUM_ENTITIES _IOWR('|', 0x01, struct media_entity_desc)
>>> #define MEDIA_IOC_ENUM_LINKS _IOWR('|', 0x02, struct media_links_enum)
>>> #define MEDIA_IOC_SETUP_LINK _IOWR('|', 0x03, struct media_link_desc)
>>> +/* Old MEDIA_IOC_G_TOPOLOGY ioctl without props support */
>>> +#define MEDIA_IOC_G_TOPOLOGY_OLD 0xc0487c04
>>
>> I would avoid calling it "_OLD". we may have a V3, V4, V5, ... of this
>> ioctl.
>>
>> I would, instead, define an _IOWR_COMPAT macro that would take an extra
>> parameter, in order to get the size of part of the struct, e. g. something
>> like:
>>
>> #define MEDIA_IOC_G_TOPOLOGY_V1 _IOWR_COMPAT('|', 0x04, struct media_v2_topology, num_props)
>
> That's not a bad idea, actually.
>
>>
>> Also, I don't see any good reason why to keep this at uAPI (except for a
>> mc-compliance tool that would test both versions - but this could be
>> defined directly there).
>
> Makes sense.
Ah, this means that applications like gstreamer cannot fall back to the older
G_TOPOLOGY ioctl. They don't know which kernel they are running against, so
they will probably want to try the new one first before falling back to the
old version.
I wonder how drm does this. Does anyone know?
Regards,
Hans
>
>>
>>> #define MEDIA_IOC_G_TOPOLOGY _IOWR('|', 0x04, struct media_v2_topology)
>>> #define MEDIA_IOC_REQUEST_ALLOC _IOR ('|', 0x05, int)
>>>
>>
>> Thanks,
>> Mauro
>>
>
> Regards,
>
> Hans
>
