Re: [PATCH v2 09/20] media: i2c: ov4689: Use runtime PM autosuspend

[Sakari Ailus]

Hi Mikhail,

On Mon, Jan 08, 2024 at 06:06:52PM +0300, Mikhail Rudenko wrote:
> Hi Sakari,
> 
> Thanks for the review!
> 
> On 2024-01-08 at 11:18 GMT, Sakari Ailus <sakari.ailus@xxxxxx> wrote:
> 
> > Hi Mikhail,
> >
> > On Mon, Dec 18, 2023 at 08:40:30PM +0300, Mikhail Rudenko wrote:
> >> Use runtime PM autosuspend to avoid powering off the sensor during
> >> fast stop-reconfigure-restart cycles.
> >>
> >> Signed-off-by: Mikhail Rudenko <mike.rudenko@xxxxxxxxx>
> >> ---
> >>  drivers/media/i2c/ov4689.c | 22 +++++++++++++++-------
> >>  1 file changed, 15 insertions(+), 7 deletions(-)
> >>
> >> diff --git a/drivers/media/i2c/ov4689.c b/drivers/media/i2c/ov4689.c
> >> index 5300e621ff90..64cc6d9e48cc 100644
> >> --- a/drivers/media/i2c/ov4689.c
> >> +++ b/drivers/media/i2c/ov4689.c
> >> @@ -407,26 +407,27 @@ static int ov4689_s_stream(struct v4l2_subdev *sd, int on)
> >>  					  ov4689->cur_mode->num_regs,
> >>  					  NULL);
> >>  		if (ret) {
> >> -			pm_runtime_put(dev);
> >> +			pm_runtime_put_sync(dev);
> >
> > Why are you switching to pm_runtime_put_sync() here? That isn't covered by
> > the commit message (nor I think should be done).
> 
> PM autosuspend conversion was suggested earlier by Laurent in his review
> of this series [1], and he adviced looking at how it was done for the
> imx290 driver. I followed along the lines of the corresponding patch
> [2].

Ah, I suppose all of these are error cases. I suppose it won't do any harm
in this case but it's not really useful either.

You can get more benefits from autosuspend if you can avoid writing
registers that already have the same values you're writing to them. Thay
may be better left outside this set as it's already fairly big.

> 
> >>  			goto unlock_and_return;
> >>  		}
> >>
> >>  		ret = __v4l2_ctrl_handler_setup(&ov4689->ctrl_handler);
> >>  		if (ret) {
> >> -			pm_runtime_put(dev);
> >> +			pm_runtime_put_sync(dev);
> >>  			goto unlock_and_return;
> >>  		}
> >>
> >>  		ret = cci_write(ov4689->regmap, OV4689_REG_CTRL_MODE,
> >>  				OV4689_MODE_STREAMING, NULL);
> >>  		if (ret) {
> >> -			pm_runtime_put(dev);
> >> +			pm_runtime_put_sync(dev);
> >>  			goto unlock_and_return;
> >>  		}
> >>  	} else {
> >>  		cci_write(ov4689->regmap, OV4689_REG_CTRL_MODE,
> >>  			  OV4689_MODE_SW_STANDBY, NULL);
> >> -		pm_runtime_put(dev);
> >> +		pm_runtime_mark_last_busy(dev);
> >> +		pm_runtime_put_autosuspend(dev);
> >>  	}
> >>
> >>  unlock_and_return:
> >> @@ -606,7 +607,9 @@ static int ov4689_set_ctrl(struct v4l2_ctrl *ctrl)
> >>  		break;
> >>  	}
> >>
> >> -	pm_runtime_put(dev);
> >> +	pm_runtime_mark_last_busy(dev);
> >> +	pm_runtime_put_autosuspend(dev);
> >
> > Also note that with runtime PM autosuspend,  you have to use
> > pm_runtime_get_if_active() instead of pm_runtime_get_if_in_use().
> 
> Noted, will do so in v3.
> 
> >> +
> >>  	return ret;
> >>  }
> >>
> >> @@ -877,8 +880,10 @@ static int ov4689_probe(struct i2c_client *client)
> >>  	}
> >>
> >>  	pm_runtime_set_active(dev);
> >> +	pm_runtime_get_noresume(dev);
> >>  	pm_runtime_enable(dev);
> >> -	pm_runtime_idle(dev);
> >> +	pm_runtime_set_autosuspend_delay(dev, 1000);
> >> +	pm_runtime_use_autosuspend(dev);
> >>
> >>  	ret = v4l2_async_register_subdev_sensor(sd);
> >>  	if (ret) {
> >> @@ -886,11 +891,14 @@ static int ov4689_probe(struct i2c_client *client)
> >>  		goto err_clean_subdev_pm;
> >>  	}
> >>
> >> +	pm_runtime_mark_last_busy(dev);
> >> +	pm_runtime_put_autosuspend(dev);
> >> +
> >>  	return 0;
> >>
> >>  err_clean_subdev_pm:
> >>  	pm_runtime_disable(dev);
> >> -	pm_runtime_set_suspended(dev);
> >> +	pm_runtime_put_noidle(dev);
> >>  	v4l2_subdev_cleanup(sd);
> >>  err_clean_entity:
> >>  	media_entity_cleanup(&sd->entity);
> 
> [1] https://lore.kernel.org/all/20231211181935.GG27535@xxxxxxxxxxxxxxxxxxxxxxxxxx/
> [2] https://lore.kernel.org/all/20230116144454.1012-14-laurent.pinchart@xxxxxxxxxxxxxxxx/
> 

-- 
Regards,

Sakari Ailus