Re: [PATCH v12 04/13] mfd: Add Ingenic TCU driver

[Lee Jones <lee.jones@linaro.org>]

On Thu, 27 Jun 2019, Paul Cercueil wrote:
> Le jeu. 27 juin 2019 à 8:58, Lee Jones <lee.jones@linaro.org> a écrit :
> > On Wed, 26 Jun 2019, Paul Cercueil wrote:
> > >  Le mer. 26 juin 2019 à 15:18, Lee Jones <lee.jones@linaro.org> a
> > > écrit :
> > >  > On Tue, 21 May 2019, Paul Cercueil wrote:
> > >  >
> > >  > >  This driver will provide a regmap that can be retrieved very
> > > early
> > >  > > in
> > >  > >  the boot process through the API function
> > > ingenic_tcu_get_regmap().
> > >  > >
> > >  > >  Additionally, it will call devm_of_platform_populate() so that
> > > all
> > >  > > the
> > >  > >  children devices will be probed.
> > >  > >
> > >  > >  Signed-off-by: Paul Cercueil <paul@crapouillou.net>
> > >  > >  ---
> > >  > >
> > >  > >  Notes:
> > >  > >      v12: New patch
> > >  > >
> > >  > >   drivers/mfd/Kconfig             |   8 +++
> > >  > >   drivers/mfd/Makefile            |   1 +
> > >  > >   drivers/mfd/ingenic-tcu.c       | 113
> > >  > > ++++++++++++++++++++++++++++++++
> > >  > >   include/linux/mfd/ingenic-tcu.h |   8 +++
> > >  > >   4 files changed, 130 insertions(+)
> > >  > >   create mode 100644 drivers/mfd/ingenic-tcu.c
> > 
> > [...]
> > 
> > >  > >  +static struct regmap * __init ingenic_tcu_create_regmap(struct
> > >  > > device_node *np)
> > >  > >  +{
> > >  > >  +	struct resource res;
> > >  > >  +	void __iomem *base;
> > >  > >  +	struct regmap *map;
> > >  > >  +
> > >  > >  +	if (!of_match_node(ingenic_tcu_of_match, np))
> > >  > >  +		return ERR_PTR(-EINVAL);
> > 
> > Drop this check.
> > 
> > >  > >  +	base = of_io_request_and_map(np, 0, "TCU");
> > >  > >  +	if (IS_ERR(base))
> > >  > >  +		return ERR_PTR(PTR_ERR(base));
> > >  > >  +
> > >  > >  +	map = regmap_init_mmio(NULL, base,
> > > &ingenic_tcu_regmap_config);
> > >  > >  +	if (IS_ERR(map))
> > >  > >  +		goto err_iounmap;
> > 
> > Place this inside probe().
> > 
> > >  > >  +	return map;
> > >  > >  +
> > >  > >  +err_iounmap:
> > >  > >  +	iounmap(base);
> > >  > >  +	of_address_to_resource(np, 0, &res);
> > >  > >  +	release_mem_region(res.start, resource_size(&res));
> > >  > >  +
> > >  > >  +	return map;
> > >  > >  +}
> > >  >
> > >  > Why does this need to be set-up earlier than probe()?
> > > 
> > >  See the explanation below.
> > 
> > I think the answer is, it doesn't.
> > 
> > >  > >  +static int __init ingenic_tcu_probe(struct platform_device
> > > *pdev)
> > >  > >  +{
> > >  > >  +	struct regmap *map =
> > > ingenic_tcu_get_regmap(pdev->dev.of_node);
> > >  > >  +
> > >  > >  +	platform_set_drvdata(pdev, map);
> > >  > >  +
> > >  > >  +	regmap_attach_dev(&pdev->dev, map,
> > > &ingenic_tcu_regmap_config);
> > >  > >  +
> > >  > >  +	return devm_of_platform_populate(&pdev->dev);
> > >  > >  +}
> > >  > >  +
> > >  > >  +static struct platform_driver ingenic_tcu_driver = {
> > >  > >  +	.driver = {
> > >  > >  +		.name = "ingenic-tcu",
> > >  > >  +		.of_match_table = ingenic_tcu_of_match,
> > >  > >  +	},
> > >  > >  +};
> > >  > >  +
> > >  > >  +static int __init ingenic_tcu_platform_init(void)
> > >  > >  +{
> > >  > >  +	return platform_driver_probe(&ingenic_tcu_driver,
> > >  > >  +				     ingenic_tcu_probe);
> > >  >
> > >  > What?  Why?
> > > 
> > >  The device driver probed here will populate the children devices,
> > >  which will be able to retrieve the pointer to the regmap through
> > >  device_get_regmap(dev->parent).
> > 
> > I've never heard of this call.  Where is it?
> 
> dev_get_regmap, in <linux/regmap.h>.
> 
> > >  The children devices are normal platform drivers that can be probed
> > >  the normal way. These are the PWM driver, the watchdog driver, and
> > > the
> > >  OST (OS Timer) clocksource driver, all part of the same hardware
> > > block
> > >  (the Timer/Counter Unit or TCU).
> > 
> > If they are normal devices, then there is no need to roll your own
> > regmap-getter implementation like this.
> > 
> > >  > >  +}
> > >  > >  +subsys_initcall(ingenic_tcu_platform_init);
> > >  > >  +
> > >  > >  +struct regmap * __init ingenic_tcu_get_regmap(struct
> > > device_node
> > >  > > *np)
> > >  > >  +{
> > >  > >  +	if (!tcu_regmap)
> > >  > >  +		tcu_regmap = ingenic_tcu_create_regmap(np);
> > >  > >  +
> > >  > >  +	return tcu_regmap;
> > >  > >  +}
> > >  >
> > >  > This makes me pretty uncomfortable.
> > >  >
> > >  > What calls it?
> > > 
> > >  The TCU IRQ driver (patch [06/13]), clocks driver (patch [05/13]),
> > > and the
> > >  non-OST clocksource driver (patch [07/13]) all probe very early in
> > > the boot
> > >  process, and share the same devicetree node. They call this
> > > function to get
> > >  a pointer to the regmap.
> > 
> > Horrible!
> > 
> > Instead, you should send it through platform_set_drvdata() and collect
> > it in the child drivers with platform_get_drvdata(dev->parent).
> 
> The IRQ, clocks and clocksource driver do NOT have a "struct device" to
> begin with. They are not platform drivers, and cannot be platform drivers,
> as they must register so early in the boot process, before "struct device"
> is even a thing.
> 
> All they get is a pointer to the same devicetree node. Since all of these
> have to use the same registers, they need to use a shared regmap, which
> they obtain by calling ingenic_tcu_get_regmap() below.
> 
> Then, when this driver's probe gets called, the regmap is retrieved and
> attached to the struct device, and then the children devices will be
> probed: the watchdog device, the PWM device, the OST device. These three
> will retrieve the regmap by calling dev_get_regmap(dev->parent, NULL).

That makes sense.

This explanation certainly belongs in the commit log.

Can you send your v14, as you intended.  I will re-review it with new
eyes when you do.

> > >  > >  +bool ingenic_tcu_pwm_can_use_chn(struct device *dev, unsigned
> > > int
> > >  > > channel)
> > >  > >  +{
> > >  > >  +	const struct ingenic_soc_info *soc =
> > >  > > device_get_match_data(dev->parent);
> > >  > >  +
> > >  > >  +	/* Enable all TCU channels for PWM use by default except
> > > channels
> > >  > > 0/1 */
> > >  > >  +	u32 pwm_channels_mask = GENMASK(soc->num_channels - 1, 2);
> > >  > >  +
> > >  > >  +	device_property_read_u32(dev->parent,
> > > "ingenic,pwm-channels-mask",
> > >  > >  +				 &pwm_channels_mask);
> > 
> > Doesn't this call overwrite the previous assignment above?
> 
> Yes, that's intended. You have a default value, that can be overriden
> by a device property.

You should provide a comment here to make your intentions clear.

> > >  > >  +	return !!(pwm_channels_mask & BIT(channel));
> > >  > >  +}
> > >  > >  +EXPORT_SYMBOL_GPL(ingenic_tcu_pwm_can_use_chn);
> > 
> > Where is this called from?
> 
> This is called from the PWM driver.

Why can't it live in the PWM driver?

-- 
Lee Jones [李琼斯]
Linaro Services Technical Lead
Linaro.org │ Open source software for ARM SoCs
Follow Linaro: Facebook | Twitter | Blog