> through ioctl()? It's not as immediate and safe as controlling the device
quoted text
> registers through /sysfs (not /proc). However, the sysfs interface in those
> drivers appeared before V4L2 had its own ioctls and we agreed to keep and
> export the interface to the only users selecting CONFIG_VIDEO_ADV_DEBUG (ok,
> this is actually valid for the sn9c102, I'll submit a patch for the et61x251
> in the future).

Direct register access for debug ok, but not this is not ok for 
normal usage.
quoted text
>
>
>> From my POV, a driver that is creating its own userspace API is not
>> fully compliant with V4L2 API.
>
> Isn't Video4Linux2 ioctl() supersedeing sysfs in this case?

It should be. However, things like direct register access (for non-debug 
mode) may allow some controls that weren't visible via ioctl. That's why 
the sysfs usage may be evil: a driver may have some parts accessible only 
via sysfs interface, on a non-standard way, without offering the official 
API support. So, some device functionalities may be hidden from userspace 
apps that are compliant with V4L2.

Summarizing:

Linus patch seems to be the better solution to solve the V4L1_COMPAT bug.

I would also convert the other container_of stuff to to_video_device (to 
have code uniformity).

et61x251 direct register interfaces should be available only if 
CONFIG_VIDEO_ADV_DEBUG is selected to avoid its miss-usage. If there are 
other device configurations that needs specific register settings not yet 
provided, this should be provided via V4L2 standard ioctls.

This way, et61x251 will be compliant with V4L2.

I still think that we should work at the remaining sysfs classes to make 
them coherent on all V4L devices.

Cheers,
Mauro.
-
unsubscribe notice
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/