Hi Ezequiel,
On Wed, Feb 26, 2020 at 06:08:40PM -0300, Ezequiel Garcia wrote:
> On Wed, 2020-02-26 at 22:38 +0200, Sakari Ailus wrote:
> > Hi Ezequiel,
> >
> > On Wed, Feb 26, 2020 at 02:28:08PM -0300, Ezequiel Garcia wrote:
> > > Hello Sakari,
> > >
> > > Thanks a lot for your comments.
> > >
> > > On Wed, 2020-02-26 at 17:53 +0200, Sakari Ailus wrote:
> > > > Hi Ezequiel,
> > > >
> > > > On Fri, Jan 24, 2020 at 05:35:43PM -0300, Ezequiel Garcia wrote:
> > > > > Currently, v4l2_pipeline_pm_use() prototype is:
> > > > >
> > > > > int v4l2_pipeline_pm_use(struct media_entity *entity, int use)
> > > > >
> > > > > Where the 'use' argument shall only be set to '1' for enable/power-on,
> > > > > or to '0' for disable/power-off. The integer return is specified
> > > > > as only meaningful when 'use' is set to '1'.
> > > > >
> > > > > Let's enforce this semantic by splitting the function in two:
> > > > > v4l2_pipeline_pm_get and v4l2_pipeline_pm_put. This is done
> > > > > for several reasons.
> > > > >
> > > > > It makes the API easier to use (or harder to misuse).
> > > > > It removes the constraint on the values the 'use' argument
> > > > > shall take. Also, it removes the need to constraint
> > > > > the return value, by making v4l2_pipeline_pm_put void return.
> > > > >
> > > > > And last, it's more consistent with other kernel APIs, such
> > > > > as the runtime pm APIs, which makes the code more symmetric.
> > > >
> > > > Indeed. These functions only exist because not all sensor etc. drivers have
> > > > been converted to runtime PM yet. New drivers no longer implement s_power
> > > > callbacks.
> > > >
> > > > I don't object the patch as such, but I think you could also add a note
> > > > that relying on the s_power callback is deprecated. This probably should be
> > > > a separate patch.
> > > >
> > >
> > > Hans picked this patch, sending a pull request yesterday which includes it.
> > >
> > > Since you know this API better than me, I thikn it would be best
> > > if you take care of sending a patch for it.
> > >
> > > In particular, I'd like to know as reference, if any changes are needed
> > > RKISP1 and sensors such as IMX219 in order to avoid relying in the deprecated
> > > API.
> >
> > I do look for the s_power callback when reviewing the driver. :-)
> >
> > ISP drivers may, I think, omit calling s_power if they don't need to work
> > with sensor drivers that require it. In that case, one could as well fix
> > the sensor driver.
> >
> > > Moreover, is there any way we can add some build time or run time warning,
> > > to avoid developers from using an API that is deprecated?
> >
> > Getting rid of s_power is a long project, so a warning every time it's used
> > would be quite a nuisance. I think documentation is the way to go.
> >
> > I can send a patch.
> >
>
> Hey Sakari,
>
> I know everyone should always read headers, comments and documentation,
> but since reality might be different how about:
>
> diff --git a/drivers/media/v4l2-core/v4l2-subdev.c b/drivers/media/v4l2-core/v4l2-subdev.c
> index a376b351135f..eca341c3cb17 100644
> --- a/drivers/media/v4l2-core/v4l2-subdev.c
> +++ b/drivers/media/v4l2-core/v4l2-subdev.c
> @@ -802,6 +802,8 @@ void v4l2_subdev_init(struct v4l2_subdev *sd, const struct v4l2_subdev_ops *ops)
> {
> INIT_LIST_HEAD(&sd->list);
> BUG_ON(!ops);
> + if (ops->core && ops->core->s_power)
> + pr_warn_once("Warning: s_power is deprecated. Please use foo and bar instead\n");
> sd->ops = ops;
> sd->v4l2_dev = NULL;
> sd->flags = 0;
>
> Do you think that's too noisy?
There are probably quite a few similar matters one could complain about. So
what else should be similarly flagged...?
>From the message alone it's also unclear which driver that gets loaded
causes the line to be printed.
Perhaps a Kconfig option to flag all deprecated stuff, so you'd get such
messages only if you enabled that? Might be overkill...
I wonder what others think.
--
Kind regards,
Sakari Ailus
