diff mbox series

[v4,03/11] gpiolib: provide gpio_device_find()

Message ID 20230927142931.19798-4-brgl@bgdev.pl
State New
Headers show
Series gpiolib: work towards removing gpiochip_find() | expand

Commit Message

Bartosz Golaszewski Sept. 27, 2023, 2:29 p.m. UTC
From: Bartosz Golaszewski <bartosz.golaszewski@linaro.org>

gpiochip_find() is wrong and its kernel doc is misleading as the
function doesn't return a reference to the gpio_chip but just a raw
pointer. The chip itself is not guaranteed to stay alive, in fact it can
be deleted at any point. Also: other than GPIO drivers themselves,
nobody else has any business accessing gpio_chip structs.

Provide a new gpio_device_find() function that returns a real reference
to the opaque gpio_device structure that is guaranteed to stay alive for
as long as there are active users of it.

Signed-off-by: Bartosz Golaszewski <bartosz.golaszewski@linaro.org>
Reviewed-by: Linus Walleij <linus.walleij@linaro.org>
---
 drivers/gpio/gpiolib.c      | 71 +++++++++++++++++++++++++++----------
 include/linux/gpio/driver.h |  3 ++
 2 files changed, 56 insertions(+), 18 deletions(-)

Comments

Andy Shevchenko Oct. 2, 2023, 9:42 a.m. UTC | #1
On Wed, Sep 27, 2023 at 04:29:23PM +0200, Bartosz Golaszewski wrote:
> From: Bartosz Golaszewski <bartosz.golaszewski@linaro.org>
> 
> gpiochip_find() is wrong and its kernel doc is misleading as the
> function doesn't return a reference to the gpio_chip but just a raw
> pointer. The chip itself is not guaranteed to stay alive, in fact it can
> be deleted at any point. Also: other than GPIO drivers themselves,
> nobody else has any business accessing gpio_chip structs.
> 
> Provide a new gpio_device_find() function that returns a real reference
> to the opaque gpio_device structure that is guaranteed to stay alive for
> as long as there are active users of it.

...

>  struct gpio_chip *gpiochip_find(void *data,
>  				int (*match)(struct gpio_chip *gc,

> +struct gpio_device *gpio_device_find(void *data,
> +				     int (*match)(struct gpio_chip *gc,
> +						  void *data))

Why not

typedef int (*gpio_chip_match_fn)(struct gpio_chip *gc, void *data);

?
Bartosz Golaszewski Oct. 2, 2023, 9:52 a.m. UTC | #2
On Mon, Oct 2, 2023 at 11:42 AM Andy Shevchenko
<andriy.shevchenko@linux.intel.com> wrote:
>
> On Wed, Sep 27, 2023 at 04:29:23PM +0200, Bartosz Golaszewski wrote:
> > From: Bartosz Golaszewski <bartosz.golaszewski@linaro.org>
> >
> > gpiochip_find() is wrong and its kernel doc is misleading as the
> > function doesn't return a reference to the gpio_chip but just a raw
> > pointer. The chip itself is not guaranteed to stay alive, in fact it can
> > be deleted at any point. Also: other than GPIO drivers themselves,
> > nobody else has any business accessing gpio_chip structs.
> >
> > Provide a new gpio_device_find() function that returns a real reference
> > to the opaque gpio_device structure that is guaranteed to stay alive for
> > as long as there are active users of it.
>
> ...
>
> >  struct gpio_chip *gpiochip_find(void *data,
> >                               int (*match)(struct gpio_chip *gc,
>
> > +struct gpio_device *gpio_device_find(void *data,
> > +                                  int (*match)(struct gpio_chip *gc,
> > +                                               void *data))
>
> Why not
>
> typedef int (*gpio_chip_match_fn)(struct gpio_chip *gc, void *data);
>

Because gpiochip_find() will go away as soon as we convert all users.

Bart

> ?
>
> --
> With Best Regards,
> Andy Shevchenko
>
>
Andy Shevchenko Oct. 2, 2023, 9:57 a.m. UTC | #3
On Mon, Oct 02, 2023 at 11:52:52AM +0200, Bartosz Golaszewski wrote:
> On Mon, Oct 2, 2023 at 11:42 AM Andy Shevchenko
> <andriy.shevchenko@linux.intel.com> wrote:
> > On Wed, Sep 27, 2023 at 04:29:23PM +0200, Bartosz Golaszewski wrote:
> > > From: Bartosz Golaszewski <bartosz.golaszewski@linaro.org>

...

> > >  struct gpio_chip *gpiochip_find(void *data,
> > >                               int (*match)(struct gpio_chip *gc,
> >
> > > +struct gpio_device *gpio_device_find(void *data,
> > > +                                  int (*match)(struct gpio_chip *gc,
> > > +                                               void *data))
> >
> > Why not
> >
> > typedef int (*gpio_chip_match_fn)(struct gpio_chip *gc, void *data);
> 
> Because gpiochip_find() will go away as soon as we convert all users.

And gpio_device_find() does not. So, I didn't get this argument.
Bartosz Golaszewski Oct. 2, 2023, 9:59 a.m. UTC | #4
On Mon, Oct 2, 2023 at 11:57 AM Andy Shevchenko
<andriy.shevchenko@linux.intel.com> wrote:
>
> On Mon, Oct 02, 2023 at 11:52:52AM +0200, Bartosz Golaszewski wrote:
> > On Mon, Oct 2, 2023 at 11:42 AM Andy Shevchenko
> > <andriy.shevchenko@linux.intel.com> wrote:
> > > On Wed, Sep 27, 2023 at 04:29:23PM +0200, Bartosz Golaszewski wrote:
> > > > From: Bartosz Golaszewski <bartosz.golaszewski@linaro.org>
>
> ...
>
> > > >  struct gpio_chip *gpiochip_find(void *data,
> > > >                               int (*match)(struct gpio_chip *gc,
> > >
> > > > +struct gpio_device *gpio_device_find(void *data,
> > > > +                                  int (*match)(struct gpio_chip *gc,
> > > > +                                               void *data))
> > >
> > > Why not
> > >
> > > typedef int (*gpio_chip_match_fn)(struct gpio_chip *gc, void *data);
> >
> > Because gpiochip_find() will go away as soon as we convert all users.
>
> And gpio_device_find() does not. So, I didn't get this argument.
>

This symbol would only be used in a single place. But whatever, I can
do it as there will be another respin anyway.

Bart
diff mbox series

Patch

diff --git a/drivers/gpio/gpiolib.c b/drivers/gpio/gpiolib.c
index f84ad54d8dbd..0371d23f0a46 100644
--- a/drivers/gpio/gpiolib.c
+++ b/drivers/gpio/gpiolib.c
@@ -1014,16 +1014,10 @@  void gpiochip_remove(struct gpio_chip *gc)
 }
 EXPORT_SYMBOL_GPL(gpiochip_remove);
 
-/**
- * gpiochip_find() - iterator for locating a specific gpio_chip
- * @data: data to pass to match function
- * @match: Callback function to check gpio_chip
+/*
+ * FIXME: This will be removed soon.
  *
- * Similar to bus_find_device.  It returns a reference to a gpio_chip as
- * determined by a user supplied @match callback.  The callback should return
- * 0 if the device doesn't match and non-zero if it does.  If the callback is
- * non-zero, this function will return to the caller and not iterate over any
- * more gpio_chips.
+ * This function is depracated, don't use.
  */
 struct gpio_chip *gpiochip_find(void *data,
 				int (*match)(struct gpio_chip *gc,
@@ -1031,21 +1025,62 @@  struct gpio_chip *gpiochip_find(void *data,
 {
 	struct gpio_device *gdev;
 	struct gpio_chip *gc = NULL;
-	unsigned long flags;
 
-	spin_lock_irqsave(&gpio_lock, flags);
-	list_for_each_entry(gdev, &gpio_devices, list)
-		if (gdev->chip && match(gdev->chip, data)) {
-			gc = gdev->chip;
-			break;
-		}
-
-	spin_unlock_irqrestore(&gpio_lock, flags);
+	gdev = gpio_device_find(data, match);
+	if (gdev) {
+		gc = gdev->chip;
+		gpio_device_put(gdev);
+	}
 
 	return gc;
 }
 EXPORT_SYMBOL_GPL(gpiochip_find);
 
+/**
+ * gpio_device_find() - find a specific GPIO device
+ * @data: data to pass to match function
+ * @match: Callback function to check gpio_chip
+ *
+ * Returns:
+ * New reference to struct gpio_device.
+ *
+ * Similar to bus_find_device(). It returns a reference to a gpio_device as
+ * determined by a user supplied @match callback. The callback should return
+ * 0 if the device doesn't match and non-zero if it does. If the callback
+ * returns non-zero, this function will return to the caller and not iterate
+ * over any more gpio_devices.
+ *
+ * The callback takes the GPIO chip structure as argument. During the execution
+ * of the callback function the chip is protected from being freed. TODO: This
+ * actually has yet to be implemented.
+ *
+ * If the function returns non-NULL, the returned reference must be freed by
+ * the caller using gpio_device_put().
+ */
+struct gpio_device *gpio_device_find(void *data,
+				     int (*match)(struct gpio_chip *gc,
+						  void *data))
+{
+	struct gpio_device *gdev;
+
+	/*
+	 * Not yet but in the future the spinlock below will become a mutex.
+	 * Annotate this function before anyone tries to use it in interrupt
+	 * context like it happened with gpiochip_find().
+	 */
+	might_sleep();
+
+	guard(spinlock_irqsave)(&gpio_lock);
+
+	list_for_each_entry(gdev, &gpio_devices, list) {
+		if (gdev->chip && match(gdev->chip, data))
+			return gpio_device_get(gdev);
+	}
+
+	return NULL;
+}
+EXPORT_SYMBOL_GPL(gpio_device_find);
+
 static int gpiochip_match_name(struct gpio_chip *gc, void *data)
 {
 	const char *name = data;
diff --git a/include/linux/gpio/driver.h b/include/linux/gpio/driver.h
index 1cedbc3d3200..6ad1f1a8ef2e 100644
--- a/include/linux/gpio/driver.h
+++ b/include/linux/gpio/driver.h
@@ -608,6 +608,9 @@  int devm_gpiochip_add_data_with_key(struct device *dev, struct gpio_chip *gc,
 struct gpio_chip *gpiochip_find(void *data,
 				int (*match)(struct gpio_chip *gc, void *data));
 
+struct gpio_device *gpio_device_find(void *data,
+				int (*match)(struct gpio_chip *gc, void *data));
+
 struct gpio_device *gpio_device_get(struct gpio_device *gdev);
 void gpio_device_put(struct gpio_device *gdev);