2023-12-12 10:56:07

by Bartosz Golaszewski

Subject: [PATCH] gpiolib: allocate memory atomically with a spinlock held

From: Bartosz Golaszewski <[email protected]>

We will eventually switch to protecting the GPIO descriptors with a mutex
but until then, we need to allocate memory for the label copy atomically
while we're holding the global spinlock.

Reported-by: Dan Carpenter <[email protected]>
Closes: https://lore.kernel.org/linux-gpio/[email protected]/T/#u
Fixes: f8d05e276b45 ("gpiolib: remove gpiochip_is_requested()")
Signed-off-by: Bartosz Golaszewski <[email protected]>
---
drivers/gpio/gpiolib.c | 10 +++++++++-
1 file changed, 9 insertions(+), 1 deletion(-)

diff --git a/drivers/gpio/gpiolib.c b/drivers/gpio/gpiolib.c
index 4e190be75dc2..6efe44570333 100644
--- a/drivers/gpio/gpiolib.c
+++ b/drivers/gpio/gpiolib.c
@@ -2399,7 +2399,15 @@ char *gpiochip_dup_line_label(struct gpio_chip *gc, unsigned int offset)
if (!test_bit(FLAG_REQUESTED, &desc->flags))
return NULL;

- label = kstrdup(desc->label, GFP_KERNEL);
+ /*
+ * FIXME: Once we mark gpiod_direction_input/output() and
+ * gpiod_get_direction() with might_sleep(), we'll be able to protect
+ * the GPIO descriptors with mutex (while value setting operations will
+ * become lockless).
+ *
+ * Until this happens, this allocation needs to be atomic.
+ */
+ label = kstrdup(desc->label, GFP_ATOMIC);
if (!label)
return ERR_PTR(-ENOMEM);

--
2.40.1

2023-12-12 12:33:08

by Linus Walleij

Subject: Re: [PATCH] gpiolib: allocate memory atomically with a spinlock held

On Tue, Dec 12, 2023 at 11:55 AM Bartosz Golaszewski <[email protected]> wrote:

> From: Bartosz Golaszewski <[email protected]>
>
> We will eventually switch to protecting the GPIO descriptors with a mutex
> but until then, we need to allocate memory for the label copy atomically
> while we're holding the global spinlock.
>
> Reported-by: Dan Carpenter <[email protected]>
> Closes: https://lore.kernel.org/linux-gpio/[email protected]/T/#u
> Fixes: f8d05e276b45 ("gpiolib: remove gpiochip_is_requested()")
> Signed-off-by: Bartosz Golaszewski <[email protected]>

Makes sense.
Reviewed-by: Linus Walleij <[email protected]>

Yours,
Linus Walleij

2023-12-15 15:44:11

by Bartosz Golaszewski

Subject: Re: [PATCH] gpiolib: allocate memory atomically with a spinlock held

On Tue, Dec 12, 2023 at 11:55 AM Bartosz Golaszewski <[email protected]> wrote:
>
> From: Bartosz Golaszewski <[email protected]>
>
> We will eventually switch to protecting the GPIO descriptors with a mutex
> but until then, we need to allocate memory for the label copy atomically
> while we're holding the global spinlock.
>
> Reported-by: Dan Carpenter <[email protected]>
> Closes: https://lore.kernel.org/linux-gpio/[email protected]/T/#u
> Fixes: f8d05e276b45 ("gpiolib: remove gpiochip_is_requested()")
> Signed-off-by: Bartosz Golaszewski <[email protected]>
> ---
> drivers/gpio/gpiolib.c | 10 +++++++++-
> 1 file changed, 9 insertions(+), 1 deletion(-)
>
> diff --git a/drivers/gpio/gpiolib.c b/drivers/gpio/gpiolib.c
> index 4e190be75dc2..6efe44570333 100644
> --- a/drivers/gpio/gpiolib.c
> +++ b/drivers/gpio/gpiolib.c
> @@ -2399,7 +2399,15 @@ char *gpiochip_dup_line_label(struct gpio_chip *gc, unsigned int offset)
> if (!test_bit(FLAG_REQUESTED, &desc->flags))
> return NULL;
>
> - label = kstrdup(desc->label, GFP_KERNEL);
> + /*
> + * FIXME: Once we mark gpiod_direction_input/output() and
> + * gpiod_get_direction() with might_sleep(), we'll be able to protect
> + * the GPIO descriptors with mutex (while value setting operations will
> + * become lockless).
> + *
> + * Until this happens, this allocation needs to be atomic.
> + */
> + label = kstrdup(desc->label, GFP_ATOMIC);
> if (!label)
> return ERR_PTR(-ENOMEM);
>
> --
> 2.40.1
>

Patch applied.

Bart