Re: [PATCH v1 3/3] gpiolib: append SFI helpers for GPIO API

[Andy Shevchenko <andriy.shevchenko@linux.intel.com>]

On Fri, 2013-12-06 at 10:52 +0900, Alex Courbot wrote:
> On 12/06/2013 01:36 AM, Andy Shevchenko wrote:
> > To support some (legacy) firmwares and platforms let's make life easier for
> > their customers.
> >
> > This patch provides a function which converts sfi_gpio_table_entry to
> > gpio_desc. The use of it is integrated into GPIO library to enable generic
> > access to the SFI GPIO resources.
> >
> > Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
> > ---
> >   drivers/gpio/Kconfig       |  4 ++++
> >   drivers/gpio/Makefile      |  1 +
> >   drivers/gpio/gpiolib-sfi.c | 28 ++++++++++++++++++++++++++++
> >   drivers/gpio/gpiolib.c     |  3 +++
> >   drivers/gpio/gpiolib.h     | 13 +++++++++++++
> >   5 files changed, 49 insertions(+)
> >   create mode 100644 drivers/gpio/gpiolib-sfi.c
> >
> > diff --git a/drivers/gpio/Kconfig b/drivers/gpio/Kconfig
> > index ae3682d..a12752a 100644
> > --- a/drivers/gpio/Kconfig
> > +++ b/drivers/gpio/Kconfig
> > @@ -51,6 +51,10 @@ config OF_GPIO
> >   	def_bool y
> >   	depends on OF
> >
> > +config GPIO_SFI
> > +	def_bool y
> > +	depends on SFI
> > +
> >   config GPIO_ACPI
> >   	def_bool y
> >   	depends on ACPI
> > diff --git a/drivers/gpio/Makefile b/drivers/gpio/Makefile
> > index ee95154..5373e3a 100644
> > --- a/drivers/gpio/Makefile
> > +++ b/drivers/gpio/Makefile
> > @@ -5,6 +5,7 @@ ccflags-$(CONFIG_DEBUG_GPIO)	+= -DDEBUG
> >   obj-$(CONFIG_GPIO_DEVRES)	+= devres.o
> >   obj-$(CONFIG_GPIOLIB)		+= gpiolib.o
> >   obj-$(CONFIG_OF_GPIO)		+= gpiolib-of.o
> > +obj-$(CONFIG_GPIO_SFI)		+= gpiolib-sfi.o
> >   obj-$(CONFIG_GPIO_ACPI)		+= gpiolib-acpi.o
> >
> >   # Device drivers. Generally keep list sorted alphabetically
> > diff --git a/drivers/gpio/gpiolib-sfi.c b/drivers/gpio/gpiolib-sfi.c
> > new file mode 100644
> > index 0000000..c804314
> > --- /dev/null
> > +++ b/drivers/gpio/gpiolib-sfi.c
> > @@ -0,0 +1,28 @@
> > +/*
> > + * Simple Firmware Interface (SFI) helpers for GPIO API
> > + *
> > + * Copyright (C) 2013, Intel Corporation
> > + * Author: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
> > + *
> > + * This program is free software; you can redistribute it and/or modify
> > + * it under the terms of the GNU General Public License version 2 as
> > + * published by the Free Software Foundation.
> > + */
> > +
> > +#include <linux/gpio/consumer.h>
> > +#include <linux/sfi.h>
> > +#include <linux/errno.h>
> > +#include <linux/err.h>
> > +
> > +#include "gpiolib.h"
> > +
> > +struct gpio_desc *sfi_get_gpiod_by_name(const char *name)
> > +{
> > +	struct sfi_gpio_table_entry *pentry;
> > +
> > +	pentry = sfi_gpio_get_entry_by_name(name);
> > +	if (!pentry)
> > +		return ERR_PTR(-ENODEV);
> > +
> > +	return gpio_to_desc(pentry->pin_no);
> > +}
> > diff --git a/drivers/gpio/gpiolib.c b/drivers/gpio/gpiolib.c
> > index bad400c..789ae1c 100644
> > --- a/drivers/gpio/gpiolib.c
> > +++ b/drivers/gpio/gpiolib.c
> > @@ -2451,6 +2451,9 @@ struct gpio_desc *__must_check gpiod_get_index(struct device *dev,
> >   	} else if (IS_ENABLED(CONFIG_ACPI) && dev && ACPI_HANDLE(dev)) {
> >   		dev_dbg(dev, "using ACPI for GPIO lookup\n");
> >   		desc = acpi_find_gpio(dev, con_id, idx, &flags);
> > +	} else if (IS_ENABLED(CONFIG_SFI)) {
> > +		dev_dbg(dev, "using SFI for GPIO lookup\n");
> > +		desc = sfi_get_gpiod_by_name(con_id);
> 
> Your lookup function is ignoring the dev argument. Are SFI GPIOs always 
> supposed to be system-global?

It's not clear. It could be device related, though SFI itself has
probably wrong design. I rather prefer to avoid a dev parameter check at
all.

>  In this case, your if condition should 
> likely be
> 
> 	} else if (IS_ENABLED(CONFIG_SFI) && !dev) {
> 
> So that a global SFI GPIO does not get mistakenly assigned to a device 
> that has, say, a more suited platform mapping on the same con_id.

So, for example in the driver that could be enumerated from SFI, DT, and
via platform data you suggest to have something like


desc = gpiod_get(dev, "con_id_device_tree");
...

if (IS_ERR(desc))
 desc = gpiod_get(NULL, "con_id_sfi");

if (IS_ERR(desc))
 desc = gpiod_get(???, "con_id_from_platdata");

Correct?

>From my point of view, the SFI, platform data, and DT cases should have
the same con_id name. And I don't see a problem here, if we use dev
parameter != NULL for SFI case. Since you proposed to distinguish no
entity case vs. others, it would be easy to return it from SFI gpio
helper.

-- 
Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Intel Finland Oy