camera control interface

camera control interface

[Vinod]

Hi folks,

I am writing a driver for camera control inteface which is an i2c
controller. So looking up the code I think it can be a v4l subdev,
right? Can it be an independent i2c master and not v4l subdev?

Second the control sports GPIOs. It can support  a set of
synchronization primitives so it's possible to drive I2C clients and
GPIOs with hardware controlled timing to allow for sync control of
sensors hooked and also for fancy strobe. How would we represent these
gpios in v4l2 and allow the control, any ideas on that.

Thanks
-- 
~Vinod

Re: camera control interface

[Laurent Pinchart]

Hi Vinod,

On Tuesday, 29 May 2018 11:29:32 EEST Vinod wrote:
> Hi folks,
> 
> I am writing a driver for camera control inteface which is an i2c
> controller. So looking up the code I think it can be a v4l subdev,
> right? Can it be an independent i2c master and not v4l subdev?

What do you mean by "camera control interface" here ? A hardware device 
handling communication with camera sensors ? I assume the communication bus is 
I2C ? Is that "camera control interface" plain I2C or does it have additional 
features ?

If we're talking about an I2C controller a V4L2 subdev is not only unneeded, 
but it wouldn't help. You need an I2C master.

> Second the control sports GPIOs. It can support  a set of
> synchronization primitives so it's possible to drive I2C clients and
> GPIOs with hardware controlled timing to allow for sync control of
> sensors hooked and also for fancy strobe. How would we represent these
> gpios in v4l2 and allow the control, any ideas on that.

Even if your main use case it related to camera, synchronization of I2C and 
GPIO doesn't seem to be a V4L2 feature to me. It sounds that you need to 
implement that int he I2C and GPIO subsystems.

-- 
Regards,

Laurent Pinchart

Re: camera control interface

[Vinod]

Hey Laurent,

On 30-05-18, 10:04, Laurent Pinchart wrote:
> > 
> > I am writing a driver for camera control inteface which is an i2c
> > controller. So looking up the code I think it can be a v4l subdev,
> > right? Can it be an independent i2c master and not v4l subdev?
> 
> What do you mean by "camera control interface" here ? A hardware device 
> handling communication with camera sensors ? I assume the communication bus is 
> I2C ? Is that "camera control interface" plain I2C or does it have additional 
> features ?
> 
> If we're talking about an I2C controller a V4L2 subdev is not only unneeded, 
> but it wouldn't help. You need an I2C master.

Sorry if I wasn't quite right in description, the control interface is
indeed i2c master and gpio. The camera sensors are i2c slaves connected to
this i2c master and gpio for sensors are connected to this as well.

> > Second the control sports GPIOs. It can support  a set of
> > synchronization primitives so it's possible to drive I2C clients and
> > GPIOs with hardware controlled timing to allow for sync control of
> > sensors hooked and also for fancy strobe. How would we represent these
> > gpios in v4l2 and allow the control, any ideas on that.
> 
> Even if your main use case it related to camera, synchronization of I2C and 
> GPIO doesn't seem to be a V4L2 feature to me. It sounds that you need to 
> implement that int he I2C and GPIO subsystems.

Well if a user wants to capture multiple cameras and synchronise,
wouldn't that need sync of i2c and gpio. I understand it may not be
supported but the question is would it be a nice feature for v4l, if so
how to go about it?

Thanks
-- 
~Vinod

Re: camera control interface

[Laurent Pinchart]

Hi Vinod,

On Wednesday, 30 May 2018 10:28:58 EEST Vinod wrote:
> On 30-05-18, 10:04, Laurent Pinchart wrote:
> >> I am writing a driver for camera control inteface which is an i2c
> >> controller. So looking up the code I think it can be a v4l subdev,
> >> right? Can it be an independent i2c master and not v4l subdev?
> > 
> > What do you mean by "camera control interface" here ? A hardware device
> > handling communication with camera sensors ? I assume the communication
> > bus is I2C ? Is that "camera control interface" plain I2C or does it have
> > additional features ?
> > 
> > If we're talking about an I2C controller a V4L2 subdev is not only
> > unneeded, but it wouldn't help. You need an I2C master.
> 
> Sorry if I wasn't quite right in description, the control interface is
> indeed i2c master and gpio. The camera sensors are i2c slaves connected to
> this i2c master and gpio for sensors are connected to this as well.

No worries. This clarifies the context.

> >> Second the control sports GPIOs. It can support  a set of
> >> synchronization primitives so it's possible to drive I2C clients and
> >> GPIOs with hardware controlled timing to allow for sync control of
> >> sensors hooked and also for fancy strobe. How would we represent these
> >> gpios in v4l2 and allow the control, any ideas on that.
> > 
> > Even if your main use case it related to camera, synchronization of I2C
> > and GPIO doesn't seem to be a V4L2 feature to me. It sounds that you need
> > to implement that int he I2C and GPIO subsystems.
> 
> Well if a user wants to capture multiple cameras and synchronise,
> wouldn't that need sync of i2c and gpio. I understand it may not be
> supported but the question is would it be a nice feature for v4l, if so
> how to go about it?

You need two parts to implement this in my opinion. First, you need a 
synchronized I2C + GPIO mechanism on which to base the implementation. That's 
not V4L2-specific and should I believe be handled in the I2C and GPIO 
subsystems. Then, you need to expose the concept of camera synchronization in 
V4L2. That part will likely require API extensions, both inside the kernel and 
towards userspace.

-- 
Regards,

Laurent Pinchart

Re: camera control interface

[Vinod]

Hi Laurent,

On 30-05-18, 10:31, Laurent Pinchart wrote:
> On Wednesday, 30 May 2018 10:28:58 EEST Vinod wrote:
> > On 30-05-18, 10:04, Laurent Pinchart wrote:
> > >> I am writing a driver for camera control inteface which is an i2c
> > >> controller. So looking up the code I think it can be a v4l subdev,
> > >> right? Can it be an independent i2c master and not v4l subdev?
> > > 
> > > What do you mean by "camera control interface" here ? A hardware device
> > > handling communication with camera sensors ? I assume the communication
> > > bus is I2C ? Is that "camera control interface" plain I2C or does it have
> > > additional features ?
> > > 
> > > If we're talking about an I2C controller a V4L2 subdev is not only
> > > unneeded, but it wouldn't help. You need an I2C master.
> > 
> > Sorry if I wasn't quite right in description, the control interface is
> > indeed i2c master and gpio. The camera sensors are i2c slaves connected to
> > this i2c master and gpio for sensors are connected to this as well.
> 
> No worries. This clarifies the context.
> 
> > >> Second the control sports GPIOs. It can support  a set of
> > >> synchronization primitives so it's possible to drive I2C clients and
> > >> GPIOs with hardware controlled timing to allow for sync control of
> > >> sensors hooked and also for fancy strobe. How would we represent these
> > >> gpios in v4l2 and allow the control, any ideas on that.
> > > 
> > > Even if your main use case it related to camera, synchronization of I2C
> > > and GPIO doesn't seem to be a V4L2 feature to me. It sounds that you need
> > > to implement that int he I2C and GPIO subsystems.
> > 
> > Well if a user wants to capture multiple cameras and synchronise,
> > wouldn't that need sync of i2c and gpio. I understand it may not be
> > supported but the question is would it be a nice feature for v4l, if so
> > how to go about it?
> 
> You need two parts to implement this in my opinion. First, you need a 
> synchronized I2C + GPIO mechanism on which to base the implementation. That's 
> not V4L2-specific and should I believe be handled in the I2C and GPIO 
> subsystems. Then, you need to expose the concept of camera synchronization in 
> V4L2. That part will likely require API extensions, both inside the kernel and 
> towards userspace.

Okay thanks. So should I do a v4l subdev for i2c master or leave it in
i2c subsystem. Does subdev make sense for gpio too.. How does one go
about 'linking' the two?

Btw where is v4l userspace located?

-- 
~Vinod

Re: camera control interface

[Laurent Pinchart]

Hi Vinod,

On Wednesday, 30 May 2018 10:40:56 EEST Vinod wrote:
> On 30-05-18, 10:31, Laurent Pinchart wrote:
> > On Wednesday, 30 May 2018 10:28:58 EEST Vinod wrote:
> >> On 30-05-18, 10:04, Laurent Pinchart wrote:
> >>>> I am writing a driver for camera control inteface which is an i2c
> >>>> controller. So looking up the code I think it can be a v4l subdev,
> >>>> right? Can it be an independent i2c master and not v4l subdev?
> >>> 
> >>> What do you mean by "camera control interface" here ? A hardware
> >> device handling communication with camera sensors ? I assume the
> >>> communication bus is I2C ? Is that "camera control interface" plain
> >>> I2C or does it have additional features ?
> >>> 
> >>> If we're talking about an I2C controller a V4L2 subdev is not only
> >>> unneeded, but it wouldn't help. You need an I2C master.
> >> 
> >> Sorry if I wasn't quite right in description, the control interface is
> >> indeed i2c master and gpio. The camera sensors are i2c slaves connected
> >> to this i2c master and gpio for sensors are connected to this as well.
> > 
> > No worries. This clarifies the context.
> > 
> >>>> Second the control sports GPIOs. It can support  a set of
> >>>> synchronization primitives so it's possible to drive I2C clients and
> >>>> GPIOs with hardware controlled timing to allow for sync control of
> >>>> sensors hooked and also for fancy strobe. How would we represent
> >>>> these gpios in v4l2 and allow the control, any ideas on that.
> >>> 
> >>> Even if your main use case it related to camera, synchronization of
> >>> I2C and GPIO doesn't seem to be a V4L2 feature to me. It sounds that
> >>> you need to implement that int he I2C and GPIO subsystems.
> >> 
> >> Well if a user wants to capture multiple cameras and synchronise,
> >> wouldn't that need sync of i2c and gpio. I understand it may not be
> >> supported but the question is would it be a nice feature for v4l, if so
> >> how to go about it?
> > 
> > You need two parts to implement this in my opinion. First, you need a
> > synchronized I2C + GPIO mechanism on which to base the implementation.
> > That's not V4L2-specific and should I believe be handled in the I2C and
> > GPIO subsystems. Then, you need to expose the concept of camera
> > synchronization in V4L2. That part will likely require API extensions,
> > both inside the kernel and towards userspace.
> 
> Okay thanks. So should I do a v4l subdev for i2c master or leave it in
> i2c subsystem. Does subdev make sense for gpio too..

A V4L2 subdev doesn't make sense for any of these. Roughly speaking, a subdev 
represent an entity in a video data pipeline (with the notable exception of 
lens and flash controllers). I2C, GPIO and other resources (such as clocks) 
are not part of the video data pipeline (unless video data itself transits on 
I2C, but that's another story).

> How does one go about 'linking' the two?

That's what you have to invent :-)

> Btw where is v4l userspace located?

There's no official "V4L2 userspace". V4L2 is a kernel to userspace API. Lots 
of applications then use that API, either directly, or through a framework as 
GStreamer.

-- 
Regards,

Laurent Pinchart