Re: [PATCH 5/6] gpiolib: consolidate GPIO lookups

[Dmitry Torokhov]

Hi Andy,

On Fri, Nov 04, 2022 at 07:17:27PM +0200, Andy Shevchenko wrote:
> On Thu, Nov 03, 2022 at 11:10:15PM -0700, Dmitry Torokhov wrote:
> > Ensure that all paths to obtain/look up GPIOD from generic
> > consumer-visible APIs go through the new gpiod_find_and_request()
> > helper, so that we can easily extend it with support for new firmware
> > mechanisms.
> 
> ...
> 
> > +static struct gpio_desc *gpiod_find_by_fwnode(struct fwnode_handle *fwnode,
> > +					      struct device *consumer,
> > +					      const char *con_id,
> > +					      unsigned int idx,
> > +					      enum gpiod_flags *flags,
> > +					      unsigned long *lookupflags)
> >  {
> 
> > +	struct gpio_desc *desc = ERR_PTR(-ENOENT);
> 
> No need, just return directly.
> 
> > +	dev_dbg(consumer, "GPIO lookup for consumer %s in node '%s'\n",
> > +		con_id, fwnode_get_name(fwnode));
> 
> %pfwP ?

OK. Although, I think I like %pfw (without 'P') better as it gives
results like:

	/soc/i2c@11007000/edp-bridge@8

or

	\_SB.PCI0.I2C1.D010

which should help identifying the exact node.

> 
> > +
> > +	/* Using device tree? */
> >  	if (is_of_node(fwnode)) {
> > +		dev_dbg(consumer, "using device tree for GPIO lookup\n");
> > +		desc = of_find_gpio(to_of_node(fwnode),
> > +				    con_id, idx, lookupflags);
> >  	} else if (is_acpi_node(fwnode)) {
> 
> With direct return, no need for 'else' here.

When we have several branches of equal weight I prefer not to have
early/inline returns, but I can add:

	} else {
		desc = ERR_PTR(-ENOENT);
	}

at the end, what do you think?

> 
> > +		dev_dbg(consumer, "using ACPI for GPIO lookup\n");
> > +		desc = acpi_find_gpio(fwnode, con_id, idx, flags, lookupflags);
> >  	}
> >  
> > +	return desc;
> > +}
> 
> ...
> 
> > +static struct gpio_desc *gpiod_find_and_request(struct device *consumer,
> > +						struct fwnode_handle *fwnode,
> > +						const char *con_id,
> > +						unsigned int idx,
> > +						enum gpiod_flags flags,
> > +						const char *label,
> > +						bool platform_lookup_allowed)
> > +{
> 
> > +	struct gpio_desc *desc = ERR_PTR(-ENOENT);
> 
> We can get rid of the assignment, see below.
> 
> 
> > +	unsigned long lookupflags;
> > +	int ret;
> 
> > +	if (fwnode)
> 
> Do we need this check?

Yes, I would prefer to have it as it clearly informs the reader that we
are only doing lookup by node if we actually have a node.

gpiod_find_and_request() expects that it gets a valid node and in the
followup change it will be dereferencing fwnode without checking for
NULL-ness.

> 
> Debug message above (when %pfw is used) would be even useful when
> fwnode == NULL.
> 
> > +		desc = gpiod_find_by_fwnode(fwnode, consumer, con_id, idx,
> > +					    &flags, &lookupflags);
> 
> > +
> 
> The blank line can be removed after above comments being addressed.
> 
> > +	if (gpiod_not_found(desc) && platform_lookup_allowed) {
> > +		/*
> > +		 * Either we are not using DT or ACPI, or their lookup did not
> > +		 * return a result. In that case, use platform lookup as a
> > +		 * fallback.
> > +		 */
> > +		dev_dbg(consumer, "using lookup tables for GPIO lookup\n");
> > +		desc = gpiod_find(consumer, con_id, idx, &lookupflags);
> > +	}
> > +
> > +	if (IS_ERR(desc)) {
> > +		dev_dbg(consumer, "No GPIO consumer %s found\n", con_id);
> > +		return desc;
> > +	}
> > +
> > +	/*
> > +	 * If a connection label was passed use that, else attempt to use
> > +	 * the device name as label
> > +	 */
> >  	ret = gpiod_request(desc, label);
> > +	if (ret) {
> > +		if (!(ret == -EBUSY && flags & GPIOD_FLAGS_BIT_NONEXCLUSIVE))
> > +			return ERR_PTR(ret);
> > +
> > +		/*
> > +		 * This happens when there are several consumers for
> > +		 * the same GPIO line: we just return here without
> > +		 * further initialization. It is a bit of a hack.
> > +		 * This is necessary to support fixed regulators.
> > +		 *
> > +		 * FIXME: Make this more sane and safe.
> > +		 */
> 
> > +		dev_info(consumer,
> > +			 "nonexclusive access to GPIO for %s\n", con_id);
> 
> Cam be one line.

I still have not embraced the new 100 columns limit. Linus, Bart, are
you OK with moving to 100 or do you want to stay with 80 for a while?

> 
> > +		return desc;
> > +	}
> >  
> > +	ret = gpiod_configure_flags(desc, con_id, lookupflags, flags);
> >  	if (ret < 0) {
> > +		dev_dbg(consumer, "setup of GPIO %s failed\n", con_id);
> >  		gpiod_put(desc);
> >  		return ERR_PTR(ret);
> >  	}
> 
> ...
> 
> >  struct gpio_desc *fwnode_gpiod_get_index(struct fwnode_handle *fwnode,
> > +					 const char *con_id,
> > +					 int index,
> >  					 enum gpiod_flags flags,
> >  					 const char *label)
> >  {
> >  
> 
> Unnecessary blank line?

Indeed, I'll fix it.

> 
> > +	return gpiod_find_and_request(NULL, fwnode, con_id, index, flags, label,
> > +				      false);
> 
> Can be one line.

Yep, depending on 80/100 column answer.

Thanks for the review!


-- 
Dmitry