Hi Sakari,
Just saw the v2 and remembered I didn't follow-up
on your questions here. Sorry about that.
On Wed, 2021-03-10 at 00:21 +0200, Sakari Ailus wrote:
> Hi Ezequiel,
>
> On Tue, Mar 09, 2021 at 09:22:16AM -0300, Ezequiel Garcia wrote:
> > On Tue, 2021-03-09 at 12:03 +0200, Sakari Ailus wrote:
> > > Hi Ezequiel,
> > >
> > > On Mon, Mar 08, 2021 at 04:01:58PM -0300, Ezequiel Garcia wrote:
> > > > Hi Sakari,
> > > >
> > > > Thanks a lot for the patch. I like where this is going.
> > >
> > > Thanks for the review, too!
> > >
> > > >
> > > > On Mon, 2021-03-08 at 12:02 +0200, Sakari Ailus wrote:
> > > > > The V4L2 async framework is generally used with the V4L2 fwnode, which
> > > > > also depends on the former. There is only one exception, the CAFE_CCIC
> > > > > driver, which uses V4L2 async but does not need V4L2 fwnode.
> > > > >
> > > > > At the same time there is a vast number of systems that need videodev
> > > > > module, but have no use for v4l2-async that's now part of videodev.
> > > > >
> > > > > In order to improve, build v4l2-async and v4l2-fwnode as a single module
> > > > > called v4l2-async (the v4l2-async.c file is renamed as v4l2-async-core.c).
> > > > > Also the menu item V4L2_FWNODE is renamed as V4L2_ASYNC.
> > > > >
> > > > > This also moves the initialisation of the debufs entries for async subdevs
> > > > > to loading of the v4l2-async module. The directory is named as
> > > > > "v4l2-async".
> > > > >
> > > > > Signed-off-by: Sakari Ailus <sakari.ailus@xxxxxxxxxxxxxxx>
> > > > > ---
> > > > > This goes on top of my recent async patches.
> > > > >
> > > > > drivers/media/i2c/Kconfig | 80 +++++++++----------
> > > > > drivers/media/i2c/ccs/Kconfig | 2 +-
> > > > > drivers/media/i2c/et8ek8/Kconfig | 2 +-
> > > > > drivers/media/pci/intel/ipu3/Kconfig | 2 +-
> > > > > drivers/media/platform/Kconfig | 16 ++--
> > > > > drivers/media/platform/am437x/Kconfig | 2 +-
> > > > > drivers/media/platform/atmel/Kconfig | 4 +-
> > > > > drivers/media/platform/cadence/Kconfig | 4 +-
> > > > > drivers/media/platform/davinci/Kconfig | 2 +-
> > > > > drivers/media/platform/exynos4-is/Kconfig | 4 +-
> > > > > drivers/media/platform/marvell-ccic/Kconfig | 1 +
> > > > > drivers/media/platform/rcar-vin/Kconfig | 4 +-
> > > > > .../media/platform/sunxi/sun4i-csi/Kconfig | 2 +-
> > > > > .../media/platform/sunxi/sun6i-csi/Kconfig | 2 +-
> > > > > drivers/media/platform/xilinx/Kconfig | 2 +-
> > > > > drivers/media/v4l2-core/Kconfig | 3 +-
> > > > > drivers/media/v4l2-core/Makefile | 6 +-
> > > > > .../{v4l2-async.c => v4l2-async-core.c} | 23 +++++-
> > > > > drivers/media/v4l2-core/v4l2-dev.c | 5 --
> > > > > drivers/staging/media/imx/Kconfig | 2 +-
> > > > > drivers/staging/media/tegra-video/Kconfig | 2 +-
> > > > > 21 files changed, 94 insertions(+), 76 deletions(-)
> > > > > rename drivers/media/v4l2-core/{v4l2-async.c => v4l2-async-core.c} (96%)
> > > > >
> > > > [..]
> > > > > diff --git a/drivers/media/v4l2-core/Kconfig b/drivers/media/v4l2-core/Kconfig
> > > > > index bf49f83cb86f..26e12db0a4e8 100644
> > > > > --- a/drivers/media/v4l2-core/Kconfig
> > > > > +++ b/drivers/media/v4l2-core/Kconfig
> > > > > @@ -62,13 +62,14 @@ config V4L2_FLASH_LED_CLASS
> > > > > tristate "V4L2 flash API for LED flash class devices"
> > > > > depends on VIDEO_V4L2 && VIDEO_V4L2_SUBDEV_API
> > > > > depends on LEDS_CLASS_FLASH
> > > > > + select V4L2_ASYNC
> > > > > help
> > > > > Say Y here to enable V4L2 flash API support for LED flash
> > > > > class drivers.
> > > > >
> > > > > When in doubt, say N.
> > > > >
> > > > > -config V4L2_FWNODE
> > > > > +config V4L2_ASYNC
> > > > > tristate
> > > > >
> > > >
> > > > So we don't expect (or want) out-of-tree drivers to be able to use this.
> > > > Is that correct?
> > >
> > > Hmm. I somehow manage to miss how this is related to out-of-tree drivers.
> > >
> >
> > As it is, the v4l2-async module is not user-selectable, because
> > V4L2_ASYNC has no prompt (the symbol is non-visible).
> >
> > For the user to be able to select the v4l2-async module,
> > even if no driver is selecting it, it needs a prompt:
> >
> > config V4L2_ASYNC
> > tristate "V4L2 asynchronous subdevice support"
>
> Well, yes, you could do this in principle.
>
> But this patch does not change this in any way, does it?
>
I can be wrong, but it does change it. Before the patch,
v4l2-async kernel API is in the videodev module, which
is built with CONFIG_VIDEO_V4L2, which is enabled
when the user selects CONFIG_MEDIA_SUPPORT.
If the kernel was built with CONFIG_VIDEO_V4L2,
out-of-tree drivers gets support for v4l2-async.
After this patch, the v4l2-async kernel API is now selected
by drivers that need it, and can't be enabled. If there are
no such drivers, out-of-tree drivers won't get it.
I guess this is a bit subtle, and perhaps not worth
considering.
> >
> > This is usually useful for out-of-tree drivers, AFAIK.
>
> If someone needs this out-of-tree, they could also enable one of the
> drivers that need v4l2-async in the kernel.
>
True :)
Thanks,
Ezequiel
