Hello Hans,
Thank you very much for the quick response.
On 15.01.2021 13:11, Hans Verkuil wrote:
On 15/01/2021 12:53, Sebastian Fricke wrote:
Hello,
I believe you are the creator of v4l2-compliance, right? My hope is that you might be able to lead me in the right direction.
I am currently finishing up a sensor driver that I ported from a downstream kernel. And I try to clear up the last compliance error:
```
info: checking control event 'User Controls' (0x00980001)
VIDIOC_SUBSCRIBE_EVENT returned -1 (Inappropriate ioctl for device)
This returns ENOTTY, which is weird...
fail: v4l2-test-controls.cpp(818): subscribe event for control 'User Controls' failed
test VIDIOC_(UN)SUBSCRIBE_EVENT/DQEVENT: FAIL
```
I notice that it calls a normal v4l2 ioctl on my subdevice. I implemented the event handling just like I was able to find in other devices like:
imx219, imx355, ov2640 by adding:
```
#include <media/v4l2-event.h>
...
sd->flags |= V4L2_SUBDEV_FL_HAS_DEVNODE |
V4L2_SUBDEV_FL_HAS_EVENTS;
...
static const struct v4l2_subdev_core_ops ov13850_core_ops = {
...
.subscribe_event = v4l2_ctrl_subdev_subscribe_event,
.unsubscribe_event = v4l2_event_subdev_unsubscribe
...
```
...since this looks exactly as it should be.
Am I supposed to correct that error, or asked in another way is it actually an error when a subdevice doesn't implement VIDIOC_SUBSCRIBE_EVENT?
Yes, it is an error if subscribe_event isn't implemented, but you DO have controls in your driver.
Are you compiling v4l2-compliance from the git repo? You shouldn't use the version packages by distros since that is typically much too old.
Yes I have compiled it directly from the git repo `git clone https://git.linuxtv.org/v4l-utils.git/`, to be more specific I have an ansible playbook that shows exactly how I compiled it:
https://github.com/initBasti/NanoPC-T4_armbian_configuration/blob/main/ansible/setup_nanopct4.yml#L95
The build I currently use is 3 days old.
I can't give support either if you are using an old version.
Additionally, I noticed that the compliance check doesn't look at my VIDIOC_SUBDEV_G_SELECTION implementation but instead searches for any implementation of the VIDIOC_G_SELECTION ioctl.
It shouldn't, at least v4l2-test-subdevs.cpp only uses VIDIOC_SUBDEV_G_SELECTION.
That is exactly what confuses me, I cannot find a *single* SUBDEV ioctl
within the output of the tests.
Here is the complete output: https://paste.debian.net/1181280/
Regards,
Hans
Thanks a lot
Sebastian
I invoked the compliance test with:
v4l2-compliance -u /dev/v4l-subdev3 -v -T
My current code can be located here: https://github.com/initBasti/Linux_kernel_media_tree_fork/blob/fix_subscribe_event_user_controls/drivers/media/i2c/ov13850.c
Greetings and thanks in advance,
Sebastian
