Re: [PATCH 1/4] media: i2c: ov5695: use regulator_bulk_enable/regulator_bulk disable instead of for loop

From: Tommaso Merciai
Date: Thu Jun 09 2022 - 06:12:10 EST


Hi All,

On Thu, Jun 02, 2022 at 07:57:02AM +0200, Michael Nazzareno Trimarchi wrote:
> Hi Mark
>
> Add Dongchun Zhu, for the patch of regulator changes and mark brown to
> clarify the API for bulk regulator.
>
> The commit f1a64f56663e9d03e509439016dcbddd0166b2da states that the
> regulator bulk api does not guarantee the order.
> Can you help me with this?

Just a gentle ping on this point.
Thanks

Regards,
Tommaso

>
> On Wed, Jun 1, 2022 at 10:39 AM Michael Nazzareno Trimarchi
> <michael@xxxxxxxxxxxxxxxxxxxx> wrote:
> >
> > Hi
> >
> > On Wed, Jun 1, 2022 at 10:11 AM Jacopo Mondi <jacopo@xxxxxxxxxx> wrote:
> > >
> > > Hi Micheal,
> > >
> > > On Tue, May 31, 2022 at 05:50:51PM +0200, Michael Nazzareno Trimarchi wrote:
> > > > Hi
> > > >
> > > > On Tue, May 31, 2022 at 5:40 PM Tommaso Merciai
> > > > <tommaso.merciai@xxxxxxxxxxxxxxxxxxxx> wrote:
> > > > >
> > > > > Hi Jacopo,
> > > > > On Tue, May 31, 2022 at 03:14:09PM +0200, Jacopo Mondi wrote:
> > > > > > Hi Tommaso,
> > > > > >
> > > > > > On Thu, May 19, 2022 at 09:51:14AM +0200, Tommaso Merciai wrote:
> > > > > > > Enable regulator using regulator_bulk_enable/regulatore_bulk_disable
> > > > > > > function in __ov5695_power_on/__ov5695_power_off function instead of for loop.
> > > > > > > This reduce code size and make things more clear
> > > > > > >
> > > > > > > Signed-off-by: Tommaso Merciai <tommaso.merciai@xxxxxxxxxxxxxxxxxxxx>
> > > > > > > Co-Developed-by: Michael Trimarchi <michael@xxxxxxxxxxxxxxxxxxxx>
> > > > > > > ---
> > > > > > > drivers/media/i2c/ov5695.c | 25 +++++++------------------
> > > > > > > 1 file changed, 7 insertions(+), 18 deletions(-)
> > > > > > >
> > > > > > > diff --git a/drivers/media/i2c/ov5695.c b/drivers/media/i2c/ov5695.c
> > > > > > > index 439385938a51..880b586e55fe 100644
> > > > > > > --- a/drivers/media/i2c/ov5695.c
> > > > > > > +++ b/drivers/media/i2c/ov5695.c
> > > > > > > @@ -972,7 +972,7 @@ static int ov5695_s_stream(struct v4l2_subdev *sd, int on)
> > > > > > >
> > > > > > > static int __ov5695_power_on(struct ov5695 *ov5695)
> > > > > > > {
> > > > > > > - int i, ret;
> > > > > > > + int ret;
> > > > > > > struct device *dev = &ov5695->client->dev;
> > > > > > >
> > > > > > > ret = clk_prepare_enable(ov5695->xvclk);
> > > > > > > @@ -987,13 +987,10 @@ static int __ov5695_power_on(struct ov5695 *ov5695)
> > > > > > > * The hardware requires the regulators to be powered on in order,
> > > > > > > * so enable them one by one.
> > > > > > > */
> > > > > >
> > > > > > The comment says that the hardware requires regulators to be enabled
> > > > > > in precise order
> > > > > >
> > > >
> > > > They are enabled on the array order.
> > > >
> > > > > > > - for (i = 0; i < OV5695_NUM_SUPPLIES; i++) {
> > > > > > > - ret = regulator_enable(ov5695->supplies[i].consumer);
> > > > > > > - if (ret) {
> > > > > > > - dev_err(dev, "Failed to enable %s: %d\n",
> > > > > > > - ov5695->supplies[i].supply, ret);
> > > > > > > - goto disable_reg_clk;
> > > > > > > - }
> > > > > > > + ret = regulator_bulk_enable(ARRAY_SIZE(ov5695->supplies), ov5695->supplies);
> > > > > >
> > > > > > bulk_enable() uses the async API (async_schedule_domain() in
> > > > > > particular) which by the name makes me think such ordering guarantee
> > > > > > cannot be respected.
> > > >
> > > > I don't think so. Will make no sense because if it fails, revert them.
> > > > Even the bulk disable disable them
> > > > in reverse order
> > > >
> > >
> > > I understand your points, but even the commit message in the patch
> > > linked by Tommaso [1] (which I see in mainline as
> > > f1a64f56663e ("media: i2c: ov5695: Fix power on and off sequences"))
> > > reports:
> > >
> > > "Given the bulk API does not give any guarantee about the order of
> > > regulators, change the driver to use regulator_disable() instead."
> > >
> > > However I would have expected the core regulator API to clearly document
> > > this behaviour.
> > >
> >
> > Yes, I agree. I see two points:
> > - patch f1a64f56663e is not fully consistent
> > - a patch is needed to the regulator api documentation
> >
> > I think that we need better documentation of the api but:
> > Work-queues are SMP-safe and guarantee serialization of actual work performed.
> >
> > Michael
> >
> >
> >
> > >
> > > > > >
> > > > > > However most sensors require some kind of ordering when enabling
> > > > > > regulators, and most of the use the bulk API anyhow. The fact this
> > > > > > driver uses the bulk API to get an release the regulators but not for
> > > > > > enabling them and the above comment, makes me think it has been done
> > > > > > on purpose ? Could you check with the driver author maybe ?
> > > > >
> > > > > Thanks for suggestion, good question.
> > > > > I see also ov5693 driver use bulk_enable/bulk_disable
> > > > > on ov5693_sensor_powerdown and ov5693_sensor_powerup functions, I take
> > > > > this as reference (and I'm wrong)
> > > > >
> > > > > In a functional test on PX30_Mini_evb_v11_20190507, after this series
> > > > > I'm able to see the correct chip id during probe and do some capture.
> > > > >
> > > > > I think you are right Jacopo, we can drop off this [PATCH 1/4]
> > > > > On the following link I found the issue that you describe: [1]
> > > > >
> > > >
> > > > WHy drop?
> > >
> > > As this is a partial revert of [1].
> > >
> > > I think in practice this won't make any actual difference, but if not
> > > 100% sure, better leave it the way it is as the authors of [1] might
> > > have actually been experiencing issues. Even more as this patch is
> > > not a bugfix but a nice-to-have. Up to you :)
> > >
> > >
> > > >
> > > > Michael
> > > >
> > > > > >
> > > > > > > + if (ret) {
> > > > > > > + dev_err(dev, "Failed to enable regulators %d\n", ret);
> > > > > > > + goto disable_reg_clk;
> > > > > > > }
> > > > > > >
> > > > > > > gpiod_set_value_cansleep(ov5695->reset_gpio, 0);
> > > > > > > @@ -1003,8 +1000,7 @@ static int __ov5695_power_on(struct ov5695 *ov5695)
> > > > > > > return 0;
> > > > > > >
> > > > > > > disable_reg_clk:
> > > > > > > - for (--i; i >= 0; i--)
> > > > > > > - regulator_disable(ov5695->supplies[i].consumer);
> > > > > > > + regulator_bulk_disable(ARRAY_SIZE(ov5695->supplies), ov5695->supplies);
> > > > > >
> > > > > > FYI the bulk API does this for you if enabling any of the regulators fails.
> > > > > > Hence this should not be necessary.
> > > > >
> > > > > Thanks for sharing! This is new to me.
> > > > > I'll update the series on v2 removing this patch.
> > > > >
> > > > > Regards,
> > > > > Tommaso
> > > > >
> > > > > [1]: https://mailweb.openeuler.org/hyperkitty/list/kernel@xxxxxxxxxxxxx/message/4X54QYJDRRE4K5BW4FTDZUGRAL4GRQWY/
> > > > >
> > > > > > Thanks
> > > > > > j
> > > > > >
> > > > > > > clk_disable_unprepare(ov5695->xvclk);
> > > > > > >
> > > > > > > return ret;
> > > > > > > @@ -1012,8 +1008,6 @@ static int __ov5695_power_on(struct ov5695 *ov5695)
> > > > > > >
> > > > > > > static void __ov5695_power_off(struct ov5695 *ov5695)
> > > > > > > {
> > > > > > > - struct device *dev = &ov5695->client->dev;
> > > > > > > - int i, ret;
> > > > > > >
> > > > > > > clk_disable_unprepare(ov5695->xvclk);
> > > > > > > gpiod_set_value_cansleep(ov5695->reset_gpio, 1);
> > > > > > > @@ -1022,12 +1016,7 @@ static void __ov5695_power_off(struct ov5695 *ov5695)
> > > > > > > * The hardware requires the regulators to be powered off in order,
> > > > > > > * so disable them one by one.
> > > > > > > */
> > > > > > > - for (i = OV5695_NUM_SUPPLIES - 1; i >= 0; i--) {
> > > > > > > - ret = regulator_disable(ov5695->supplies[i].consumer);
> > > > > > > - if (ret)
> > > > > > > - dev_err(dev, "Failed to disable %s: %d\n",
> > > > > > > - ov5695->supplies[i].supply, ret);
> > > > > > > - }
> > > > > > > + regulator_bulk_disable(ARRAY_SIZE(ov5695->supplies), ov5695->supplies);
> > > > > > > }
> > > > > > >
> > > > > > > static int __maybe_unused ov5695_runtime_resume(struct device *dev)
> > > > > > > --
> > > > > > > 2.25.1
> > > > > > >
> > > > >
> > > > > --
> > > > > Tommaso Merciai
> > > > > Embedded Linux Engineer
> > > > > tommaso.merciai@xxxxxxxxxxxxxxxxxxxx
> > > > > __________________________________
> > > > >
> > > > > Amarula Solutions SRL
> > > > > Via Le Canevare 30, 31100 Treviso, Veneto, IT
> > > > > T. +39 042 243 5310
> > > > > info@xxxxxxxxxxxxxxxxxxxx
> > > > > www.amarulasolutions.com
> > > >
> > > >
> > > >
> > > > --
> > > > Michael Nazzareno Trimarchi
> > > > Co-Founder & Chief Executive Officer
> > > > M. +39 347 913 2170
> > > > michael@xxxxxxxxxxxxxxxxxxxx
> > > > __________________________________
> > > >
> > > > Amarula Solutions BV
> > > > Joop Geesinkweg 125, 1114 AB, Amsterdam, NL
> > > > T. +31 (0)85 111 9172
> > > > info@xxxxxxxxxxxxxxxxxxxx
> > > > www.amarulasolutions.com
> >
> >
> >
> > --
> > Michael Nazzareno Trimarchi
> > Co-Founder & Chief Executive Officer
> > M. +39 347 913 2170
> > michael@xxxxxxxxxxxxxxxxxxxx
> > __________________________________
> >
> > Amarula Solutions BV
> > Joop Geesinkweg 125, 1114 AB, Amsterdam, NL
> > T. +31 (0)85 111 9172
> > info@xxxxxxxxxxxxxxxxxxxx
> > www.amarulasolutions.com
>
>
>
> --
> Michael Nazzareno Trimarchi
> Co-Founder & Chief Executive Officer
> M. +39 347 913 2170
> michael@xxxxxxxxxxxxxxxxxxxx
> __________________________________
>
> Amarula Solutions BV
> Joop Geesinkweg 125, 1114 AB, Amsterdam, NL
> T. +31 (0)85 111 9172
> info@xxxxxxxxxxxxxxxxxxxx
> www.amarulasolutions.com

--
Tommaso Merciai
Embedded Linux Engineer
tommaso.merciai@xxxxxxxxxxxxxxxxxxxx
__________________________________

Amarula Solutions SRL
Via Le Canevare 30, 31100 Treviso, Veneto, IT
T. +39 042 243 5310
info@xxxxxxxxxxxxxxxxxxxx
www.amarulasolutions.com