Re: [PATCH V3 3/3] media: i2c: gc5035: Add GC5035 image sensor driver

From: Tomasz Figa
Date: Wed Sep 02 2020 - 19:00:01 EST


Hi Sakari,

On Mon, Aug 31, 2020 at 7:41 PM Sakari Ailus
<sakari.ailus@xxxxxxxxxxxxxxx> wrote:
>
> Hi Xingyu,
>
> Thanks for the update. I've got a few more comments below.
>
> Do you happen to have some insight on what the OTP data contains and what
> does the driver do with it?
>
> At least in principle the OTP data may be programmed for the customer so
> the same sensor could contain something else what the driver expects to
> find there.
>

Thanks for the review. For anything without my reply, assume fixed. :)

As far as I can see, the data is being read from an area that is
supposed to be reserved for Galaxycore, so I'd assume it doesn't
depend on the customer.

[snip]
> > diff --git a/drivers/media/i2c/Kconfig b/drivers/media/i2c/Kconfig
> > index da11036..aeaf594 100644
> > --- a/drivers/media/i2c/Kconfig
> > +++ b/drivers/media/i2c/Kconfig
> > @@ -712,6 +712,18 @@ config VIDEO_APTINA_PLL
> > config VIDEO_SMIAPP_PLL
> > tristate
> >
> > +config VIDEO_GC5035
> > + tristate "Galaxycore GC5035 sensor support"
> > + depends on I2C && VIDEO_V4L2
> > + select MEDIA_CONTROLLER
> > + select VIDEO_V4L2_SUBDEV_API
>
> Add:
>
> V4L2_FWNODE
> OF

This driver doesn't depend on OF. It uses the firmware-independent
property access API. (v4 I sent actually uses device_property_*()).

[snip]
> > +static int __gc5035_power_on(struct gc5035 *gc5035)
> > +{
> > + struct device *dev = &gc5035->client->dev;
> > + int i, ret;
> > +
> > + ret = clk_prepare_enable(gc5035->xvclk);
> > + if (ret < 0) {
> > + dev_err(dev, "Failed to enable xvclk\n");
> > + return ret;
> > + }
> > +
> > + gpiod_set_value_cansleep(gc5035->reset_gpio, 1);
> > +
> > + for (i = 0; i < GC5035_NUM_SUPPLIES; i++) {
> > + ret = regulator_enable(gc5035->supplies[i].consumer);
> > + if (ret) {
> > + dev_err(dev, "Failed to enable %s: %d\n",
> > + gc5035->supplies[i].supply, ret);
> > + goto disable_reg_clk;
>
> Please use regulator_bulk_enable() here, and regulator_bulk_disable()
> below.
>

This actually needs to have one of the regulators (iovdd) enabled
before the other ones, but regulator_bulk_enable() is async. In v4 I
used regulator_enable() for iovdd and regulator_bulk_enable() for the
other two for optimal sequencing.

Best regards,
Tomasz