Skip to contentSkip to conversation

gpiolib: protect the list of GPIO devices with a mutex

PATCH Superseded Bartosz Golaszewski 2 participants 9 messages Dec 8, 2023 – Dec 14, 2023 GPIO

v2 of 3 Superseded Latest is v3 →

lore
Share & tools
lore
#0 lore raw

v2 cover gpiolib: protect the list of GPIO devices with a mutex

From: Bartosz Golaszewski <[email protected]>

I figured that - since the descriptor locking is going to take some more time - we should at least start the conversion and protect the GPIO device list with a mutex.

Patch 2/2 here is actually a v2 of the original submission.

v1 -> v2:
- add a patch renaming two functions
- protect the list in gpio_device_find() too
- coding style tweaks

Bartosz Golaszewski (2):
gpiolib: rename static functions that are called with the lock taken
gpiolib: use a mutex to protect the list of GPIO devices

 drivers/gpio/gpiolib-sysfs.c |  26 +-----
 drivers/gpio/gpiolib-sysfs.h |   6 ++
 drivers/gpio/gpiolib.c       | 166 ++++++++++++++++++-----------------
 drivers/gpio/gpiolib.h       |   1 -
 4 files changed, 94 insertions(+), 105 deletions(-)
Signature · 1 line
2.40.1
#1 lore raw

PATCH v2 1/2 gpiolib: rename static functions that are called with the lock taken

From: Bartosz Golaszewski <[email protected]>

Rename two functions that read or modify the global GPIO device list but don't take the lock themselves (and need to be called with it already acquired). Use the _unlocked() suffix which seems to be used quite consistently across the kernel despite there also existing the _locked() suffix for the same purpose.

---
 drivers/gpio/gpiolib.c | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)
drivers/gpio/gpiolib.c +4 -4
index 4e190be75dc2..779f8b21bf05 100644
@@ -290,7 +290,7 @@ struct gpio_chip *gpio_device_get_chip(struct gpio_device *gdev)
 EXPORT_SYMBOL_GPL(gpio_device_get_chip);
 
 /* dynamic allocation of GPIOs, e.g. on a hotplugged device */
-static int gpiochip_find_base(int ngpio)
+static int gpiochip_find_base_unlocked(int ngpio)
 {
 	struct gpio_device *gdev;
 	int base = GPIO_DYNAMIC_BASE;
@@ -363,7 +363,7 @@ EXPORT_SYMBOL_GPL(gpiod_get_direction);
  * Return -EBUSY if the new chip overlaps with some other chip's integer
  * space.
  */
-static int gpiodev_add_to_list(struct gpio_device *gdev)
+static int gpiodev_add_to_list_unlocked(struct gpio_device *gdev)
 {
 	struct gpio_device *prev, *next;
 
@@ -907,7 +907,7 @@ int gpiochip_add_data_with_key(struct gpio_chip *gc, void *data,
 	 */
 	base = gc->base;
 	if (base < 0) {
-		base = gpiochip_find_base(gc->ngpio);
+		base = gpiochip_find_base_unlocked(gc->ngpio);
 		if (base < 0) {
 			spin_unlock_irqrestore(&gpio_lock, flags);
 			ret = base;
@@ -927,7 +927,7 @@ int gpiochip_add_data_with_key(struct gpio_chip *gc, void *data,
 	}
 	gdev->base = base;
 
-	ret = gpiodev_add_to_list(gdev);
+	ret = gpiodev_add_to_list_unlocked(gdev);
 	if (ret) {
 		spin_unlock_irqrestore(&gpio_lock, flags);
 		chip_err(gc, "GPIO integer space overlap, cannot add chip\n");
Signature · 1 line
2.40.1
Reviewed-by on v2 1/2
#2 lore raw

Re: [PATCH v2 1/2] gpiolib: rename static functions that are called with the lock taken

Quoted 7 lines · Bartosz Golaszewski

From: Bartosz Golaszewski <[email protected]>

Rename two functions that read or modify the global GPIO device list but don't take the lock themselves (and need to be called with it already acquired). Use the _unlocked() suffix which seems to be used quite consistently across the kernel despite there also existing the _locked() suffix for the same purpose.

Okay,

Signature · 2 lines
With Best Regards,
Andy Shevchenko
#3 lore raw

PATCH v2 2/2 gpiolib: use a mutex to protect the list of GPIO devices

In reply to Bartosz Golaszewski

From: Bartosz Golaszewski <[email protected]>

The global list of GPIO devices is never modified or accessed from atomic context so it's fine to protect it using a mutex. Add a new global lock dedicated to the gpio_devices list and use it whenever accessing or modifying it.

While at it: fold the sysfs registering of existing devices into gpiolib.c and make gpio_devices static within its compilation unit.

---
 drivers/gpio/gpiolib-sysfs.c |  26 +-----
 drivers/gpio/gpiolib-sysfs.h |   6 ++
 drivers/gpio/gpiolib.c       | 162 ++++++++++++++++++-----------------
 drivers/gpio/gpiolib.h       |   1 -
 4 files changed, 92 insertions(+), 103 deletions(-)
drivers/gpio/gpiolib-sysfs.c +2 -24
index 6f309a3b2d9a..c538568604e8 100644
@@ -790,9 +790,7 @@ void gpiochip_sysfs_unregister(struct gpio_device *gdev)
 
 static int __init gpiolib_sysfs_init(void)
 {
-	int		status;
-	unsigned long	flags;
-	struct gpio_device *gdev;
+	int status;
 
 	status = class_register(&gpio_class);
 	if (status < 0)
@@ -804,26 +802,6 @@ static int __init gpiolib_sysfs_init(void)
 	 * We run before arch_initcall() so chip->dev nodes can have
 	 * registered, and so arch_initcall() can always gpiod_export().
 	 */
-	spin_lock_irqsave(&gpio_lock, flags);
-	list_for_each_entry(gdev, &gpio_devices, list) {
-		if (gdev->mockdev)
-			continue;
-
-		/*
-		 * TODO we yield gpio_lock here because
-		 * gpiochip_sysfs_register() acquires a mutex. This is unsafe
-		 * and needs to be fixed.
-		 *
-		 * Also it would be nice to use gpio_device_find() here so we
-		 * can keep gpio_chips local to gpiolib.c, but the yield of
-		 * gpio_lock prevents us from doing this.
-		 */
-		spin_unlock_irqrestore(&gpio_lock, flags);
-		status = gpiochip_sysfs_register(gdev);
-		spin_lock_irqsave(&gpio_lock, flags);
-	}
-	spin_unlock_irqrestore(&gpio_lock, flags);
-
-	return status;
+	return gpiochip_sysfs_register_all();
 }
 postcore_initcall(gpiolib_sysfs_init);
drivers/gpio/gpiolib-sysfs.h +6 -0
index b794b396d6a5..ab157cec0b4b 100644
@@ -8,6 +8,7 @@ struct gpio_device;
 #ifdef CONFIG_GPIO_SYSFS
 
 int gpiochip_sysfs_register(struct gpio_device *gdev);
+int gpiochip_sysfs_register_all(void);
 void gpiochip_sysfs_unregister(struct gpio_device *gdev);
 
 #else
@@ -17,6 +18,11 @@ static inline int gpiochip_sysfs_register(struct gpio_device *gdev)
 	return 0;
 }
 
+static inline int gpiochip_sysfs_register_all(void)
+{
+	return 0;
+}
+
 static inline void gpiochip_sysfs_unregister(struct gpio_device *gdev)
 {
 }
drivers/gpio/gpiolib.c +84 -78

272 lines not on this page · raw message

drivers/gpio/gpiolib.h +0 -1
index 3ccacf3c1288..9278796db079 100644
@@ -135,7 +135,6 @@ int gpiod_set_array_value_complex(bool raw, bool can_sleep,
 int gpiod_set_transitory(struct gpio_desc *desc, bool transitory);
 
 extern spinlock_t gpio_lock;
-extern struct list_head gpio_devices;
 
 void gpiod_line_state_notify(struct gpio_desc *desc, unsigned long action);
 
Signature · 1 line
2.40.1
#4 lore raw

Re: [PATCH v2 2/2] gpiolib: use a mutex to protect the list of GPIO devices

Quoted 8 lines · Bartosz Golaszewski

From: Bartosz Golaszewski <[email protected]>

The global list of GPIO devices is never modified or accessed from atomic context so it's fine to protect it using a mutex. Add a new global lock dedicated to the gpio_devices list and use it whenever accessing or modifying it.

Quoted 4 lines

While at it: fold the sysfs registering of existing devices into gpiolib.c and make gpio_devices static within its compilation unit.

If there are no objections, I'll apply it tomorrow.

Bart

#5 lore raw

Re: [PATCH v2 2/2] gpiolib: use a mutex to protect the list of GPIO devices

In reply to Bartosz Golaszewski
Quoted 6 lines · Bartosz Golaszewski

From: Bartosz Golaszewski <[email protected]>

The global list of GPIO devices is never modified or accessed from atomic context so it's fine to protect it using a mutex. Add a new global lock dedicated to the gpio_devices list and use it whenever accessing or modifying it.

...

Quoted 2 lines

While at it: fold the sysfs registering of existing devices into gpiolib.c and make gpio_devices static within its compilation unit.

TBH I do not like injecting sysfs (legacy!) code into gpiolib. It makes things at very least confusing.

That _ugly_ ifdeffery and sysfs in the function name are not okay.

If you want do that, please create a separate change and explain the rationale behind with answering to the Q "Why do we need all that and why is it better than any alternatives?".

Signature · 2 lines
With Best Regards,
Andy Shevchenko
#6 lore raw

Re: [PATCH v2 2/2] gpiolib: use a mutex to protect the list of GPIO devices

Quoted 14 lines · Andy Shevchenko
Quoted 6 lines · Bartosz Golaszewski

From: Bartosz Golaszewski <[email protected]>

The global list of GPIO devices is never modified or accessed from atomic context so it's fine to protect it using a mutex. Add a new global lock dedicated to the gpio_devices list and use it whenever accessing or modifying it.

...

Quoted 2 lines

While at it: fold the sysfs registering of existing devices into gpiolib.c and make gpio_devices static within its compilation unit.

Quoted 9 lines

TBH I do not like injecting sysfs (legacy!) code into gpiolib. It makes things at very least confusing.

That _ugly_ ifdeffery and sysfs in the function name are not okay.

If you want do that, please create a separate change and explain the rationale behind with answering to the Q "Why do we need all that and why is it better than any alternatives?".

I can move it back to gpiolib-sysfs.c but this way we'll have to keep the GPIO device mutex public in gpiolib.h.

Bart

Quoted 5 lines

--
With Best Regards,
Andy Shevchenko

#7 lore raw

Re: [PATCH v2 2/2] gpiolib: use a mutex to protect the list of GPIO devices

Quoted 10 lines · Bartosz Golaszewski
Quoted 8 lines · Andy Shevchenko
Quoted 6 lines · Bartosz Golaszewski

From: Bartosz Golaszewski <[email protected]>

The global list of GPIO devices is never modified or accessed from atomic context so it's fine to protect it using a mutex. Add a new global lock dedicated to the gpio_devices list and use it whenever accessing or modifying it.

...

Quoted 6 lines
Quoted 6 lines
Quoted 2 lines

While at it: fold the sysfs registering of existing devices into gpiolib.c and make gpio_devices static within its compilation unit.

TBH I do not like injecting sysfs (legacy!) code into gpiolib. It makes things at very least confusing.

Quoted 8 lines
Quoted 5 lines

That _ugly_ ifdeffery and sysfs in the function name are not okay.

If you want do that, please create a separate change and explain the rationale behind with answering to the Q "Why do we need all that and why is it better than any alternatives?".

I can move it back to gpiolib-sysfs.c but this way we'll have to keep the GPIO device mutex public in gpiolib.h.

And I'm fine with that. Again, we can discuss this in a separate change that will do that (make that mutex local with the explanation why).

Signature · 2 lines
With Best Regards,
Andy Shevchenko
#8 lore raw

Re: [PATCH v2 2/2] gpiolib: use a mutex to protect the list of GPIO devices

Quoted 27 lines · Andy Shevchenko
Quoted 10 lines · Bartosz Golaszewski
Quoted 8 lines · Andy Shevchenko
Quoted 6 lines · Bartosz Golaszewski

From: Bartosz Golaszewski <[email protected]>

The global list of GPIO devices is never modified or accessed from atomic context so it's fine to protect it using a mutex. Add a new global lock dedicated to the gpio_devices list and use it whenever accessing or modifying it.

...

Quoted 12 lines
Quoted 11 lines
Quoted 2 lines

While at it: fold the sysfs registering of existing devices into gpiolib.c and make gpio_devices static within its compilation unit.

TBH I do not like injecting sysfs (legacy!) code into gpiolib. It makes things at very least confusing.

That _ugly_ ifdeffery and sysfs in the function name are not okay.

If you want do that, please create a separate change and explain the rationale behind with answering to the Q "Why do we need all that and why is it better than any alternatives?".

Quoted 6 lines
Quoted 2 lines

I can move it back to gpiolib-sysfs.c but this way we'll have to keep the GPIO device mutex public in gpiolib.h.

And I'm fine with that. Again, we can discuss this in a separate change that will do that (make that mutex local with the explanation why).

No, I won't be sending one. I'll send another iteration of this with sysfs stuff contained to gpiolib-sysfs.c.

Bart

Quoted 5 lines

--
With Best Regards,
Andy Shevchenko