[PATCH v2 4/6] gpio: cdev: put emitting the line state events on a workqueue

Bartosz Golaszewski posted 6 patches 1 month, 2 weeks ago
There is a newer version of this series
[PATCH v2 4/6] gpio: cdev: put emitting the line state events on a workqueue
Posted by Bartosz Golaszewski 1 month, 2 weeks ago
From: Bartosz Golaszewski <bartosz.golaszewski@linaro.org>

In order to allow line state notifications to be emitted from atomic
context (for instance: from gpiod_direction_input/output()), we must
stop calling any sleeping functions in lineinfo_changed_notify(). To
that end let's use the new workqueue.

Let's atomically allocate small structures containing the required data
and fill it with information immediately upon being notified about the
change except for the pinctrl state which will be retrieved later from
process context. We can pretty reliably do this as pin functions are
typically set once per boot.

Let's make sure to bump the reference count of GPIO device and the GPIO
character device file descriptor to keep both alive until the event was
queued.

Signed-off-by: Bartosz Golaszewski <bartosz.golaszewski@linaro.org>
---
 drivers/gpio/gpiolib-cdev.c | 72 +++++++++++++++++++++++++++++++++++++--------
 1 file changed, 60 insertions(+), 12 deletions(-)

diff --git a/drivers/gpio/gpiolib-cdev.c b/drivers/gpio/gpiolib-cdev.c
index dec6c9a6005f..2677134b52cd 100644
--- a/drivers/gpio/gpiolib-cdev.c
+++ b/drivers/gpio/gpiolib-cdev.c
@@ -2451,6 +2451,7 @@ struct gpio_chardev_data {
 #ifdef CONFIG_GPIO_CDEV_V1
 	atomic_t watch_abi_version;
 #endif
+	struct file *fp;
 };
 
 static int chipinfo_get(struct gpio_chardev_data *cdev, void __user *ip)
@@ -2621,29 +2622,75 @@ static long gpio_ioctl_compat(struct file *file, unsigned int cmd,
 }
 #endif
 
+struct lineinfo_changed_ctx {
+	struct work_struct work;
+	struct gpio_v2_line_info_changed chg;
+	struct gpio_device *gdev;
+	struct gpio_chardev_data *cdev;
+};
+
+static void lineinfo_changed_func(struct work_struct *work)
+{
+	struct lineinfo_changed_ctx *ctx =
+			container_of(work, struct lineinfo_changed_ctx, work);
+	struct gpio_chip *gc;
+	int ret;
+
+	scoped_guard(srcu, &ctx->gdev->srcu) {
+		gc = srcu_dereference(ctx->gdev->chip, &ctx->gdev->srcu);
+		if (!gc)
+			return;
+
+		/*
+		 * We're doing this late because it's a sleeping function. Pin
+		 * functions are in general much more static and while it's not
+		 * 100% bullet-proof, it's good enough for most cases.
+		 */
+		if (!pinctrl_gpio_can_use_line(gc, ctx->chg.info.offset))
+			ctx->chg.info.flags |= GPIO_V2_LINE_FLAG_USED;
+	}
+
+	ret = kfifo_in_spinlocked(&ctx->cdev->events, &ctx->chg, 1,
+				  &ctx->cdev->wait.lock);
+	if (ret)
+		wake_up_poll(&ctx->cdev->wait, EPOLLIN);
+	else
+		pr_debug_ratelimited("lineinfo event FIFO is full - event dropped\n");
+
+	gpio_device_put(ctx->gdev);
+	fput(ctx->cdev->fp);
+	kfree(ctx);
+}
+
 static int lineinfo_changed_notify(struct notifier_block *nb,
 				   unsigned long action, void *data)
 {
 	struct gpio_chardev_data *cdev =
 		container_of(nb, struct gpio_chardev_data, lineinfo_changed_nb);
-	struct gpio_v2_line_info_changed chg;
+	struct lineinfo_changed_ctx *ctx;
 	struct gpio_desc *desc = data;
-	int ret;
 
 	if (!test_bit(gpio_chip_hwgpio(desc), cdev->watched_lines))
 		return NOTIFY_DONE;
 
-	memset(&chg, 0, sizeof(chg));
-	chg.event_type = action;
-	chg.timestamp_ns = ktime_get_ns();
-	gpio_desc_to_lineinfo(desc, &chg.info, false);
-	supinfo_to_lineinfo(desc, &chg.info);
+	ctx = kzalloc(sizeof(*ctx), GFP_ATOMIC);
+	if (!ctx) {
+		pr_err("Failed to allocate memory for line info notification\n");
+		return NOTIFY_DONE;
+	}
 
-	ret = kfifo_in_spinlocked(&cdev->events, &chg, 1, &cdev->wait.lock);
-	if (ret)
-		wake_up_poll(&cdev->wait, EPOLLIN);
-	else
-		pr_debug_ratelimited("lineinfo event FIFO is full - event dropped\n");
+	ctx->chg.event_type = action;
+	ctx->chg.timestamp_ns = ktime_get_ns();
+	gpio_desc_to_lineinfo(desc, &ctx->chg.info, true);
+	supinfo_to_lineinfo(desc, &ctx->chg.info);
+	/* Keep the GPIO device alive until we emit the event. */
+	ctx->gdev = gpio_device_get(desc->gdev);
+	ctx->cdev = cdev;
+	/* Keep the file descriptor alive too. */
+	get_file(ctx->cdev->fp);
+
+	INIT_WORK(&ctx->work, lineinfo_changed_func);
+	queue_work(ctx->gdev->line_state_wq, &ctx->work);
 
 	return NOTIFY_OK;
 }
@@ -2803,6 +2850,7 @@ static int gpio_chrdev_open(struct inode *inode, struct file *file)
 		goto out_unregister_line_notifier;
 
 	file->private_data = cdev;
+	cdev->fp = file;
 
 	ret = nonseekable_open(inode, file);
 	if (ret)

-- 
2.43.0
Re: [PATCH v2 4/6] gpio: cdev: put emitting the line state events on a workqueue
Posted by Kent Gibson 1 month, 2 weeks ago
On Thu, Oct 10, 2024 at 11:10:25AM +0200, Bartosz Golaszewski wrote:
> From: Bartosz Golaszewski <bartosz.golaszewski@linaro.org>
>
> In order to allow line state notifications to be emitted from atomic
> context (for instance: from gpiod_direction_input/output()), we must
> stop calling any sleeping functions in lineinfo_changed_notify(). To
> that end let's use the new workqueue.
>
> Let's atomically allocate small structures containing the required data
> and fill it with information immediately upon being notified about the
> change except for the pinctrl state which will be retrieved later from
> process context. We can pretty reliably do this as pin functions are
> typically set once per boot.
>
> Let's make sure to bump the reference count of GPIO device and the GPIO
> character device file descriptor to keep both alive until the event was
> queued.
>
> Signed-off-by: Bartosz Golaszewski <bartosz.golaszewski@linaro.org>
> ---
>  drivers/gpio/gpiolib-cdev.c | 72 +++++++++++++++++++++++++++++++++++++--------
>  1 file changed, 60 insertions(+), 12 deletions(-)
>
> diff --git a/drivers/gpio/gpiolib-cdev.c b/drivers/gpio/gpiolib-cdev.c
> index dec6c9a6005f..2677134b52cd 100644
> --- a/drivers/gpio/gpiolib-cdev.c
> +++ b/drivers/gpio/gpiolib-cdev.c
> @@ -2451,6 +2451,7 @@ struct gpio_chardev_data {
>  #ifdef CONFIG_GPIO_CDEV_V1
>  	atomic_t watch_abi_version;
>  #endif
> +	struct file *fp;
>  };
>
>  static int chipinfo_get(struct gpio_chardev_data *cdev, void __user *ip)
> @@ -2621,29 +2622,75 @@ static long gpio_ioctl_compat(struct file *file, unsigned int cmd,
>  }
>  #endif
>
> +struct lineinfo_changed_ctx {
> +	struct work_struct work;
> +	struct gpio_v2_line_info_changed chg;
> +	struct gpio_device *gdev;
> +	struct gpio_chardev_data *cdev;
> +};
> +
> +static void lineinfo_changed_func(struct work_struct *work)
> +{
> +	struct lineinfo_changed_ctx *ctx =
> +			container_of(work, struct lineinfo_changed_ctx, work);
> +	struct gpio_chip *gc;
> +	int ret;
> +
> +	scoped_guard(srcu, &ctx->gdev->srcu) {
> +		gc = srcu_dereference(ctx->gdev->chip, &ctx->gdev->srcu);
> +		if (!gc)
> +			return;
> +
> +		/*
> +		 * We're doing this late because it's a sleeping function. Pin
> +		 * functions are in general much more static and while it's not
> +		 * 100% bullet-proof, it's good enough for most cases.
> +		 */
> +		if (!pinctrl_gpio_can_use_line(gc, ctx->chg.info.offset))
> +			ctx->chg.info.flags |= GPIO_V2_LINE_FLAG_USED;
> +	}
> +

This block should be conditional on GPIO_V2_LINE_FLAG_USED not already
being set - most of the time it will be and then this is pointless work.

Cheers,
Kent.
Re: [PATCH v2 4/6] gpio: cdev: put emitting the line state events on a workqueue
Posted by Bartosz Golaszewski 1 month, 2 weeks ago
On Mon, Oct 14, 2024 at 4:09 AM Kent Gibson <warthog618@gmail.com> wrote:
>
> > +
> > +     scoped_guard(srcu, &ctx->gdev->srcu) {
> > +             gc = srcu_dereference(ctx->gdev->chip, &ctx->gdev->srcu);
> > +             if (!gc)
> > +                     return;
> > +
> > +             /*
> > +              * We're doing this late because it's a sleeping function. Pin
> > +              * functions are in general much more static and while it's not
> > +              * 100% bullet-proof, it's good enough for most cases.
> > +              */
> > +             if (!pinctrl_gpio_can_use_line(gc, ctx->chg.info.offset))
> > +                     ctx->chg.info.flags |= GPIO_V2_LINE_FLAG_USED;
> > +     }
> > +
>
> This block should be conditional on GPIO_V2_LINE_FLAG_USED not already
> being set - most of the time it will be and then this is pointless work.
>

Good point, thanks!

Bart