Re: [PATCH v8 09/20] gpiolib: cdev: support edge detection for uAPI v2

From: Andy Shevchenko
Date: Tue Sep 15 2020 - 06:40:27 EST


On Wed, Sep 9, 2020 at 1:33 PM Kent Gibson <warthog618@xxxxxxxxx> wrote:
>
> Add support for edge detection to lines requested using
> GPIO_V2_GET_LINE_IOCTL.
>
> The edge_detector implementation is based on the v1 lineevent
> implementation.

...

> + * @timestamp: cache for the timestamp storing it between hardirq and IRQ
> + * thread, used to bring the timestamp close to the actual event

_ns ?

> + * @req_seqno: the seqno for the current edge event in the sequence of
> + * events for the corresponding line request. This is drawn from the @req.
> + * @line_seqno: the seqno for the current edge event in the sequence of
> + * events for this line.
> */
> struct line {
> struct gpio_desc *desc;
> + /*
> + * -- edge detector specific fields --
> + */
> + struct linereq *req;
> + unsigned int irq;
> + u64 eflags;

> + /*
> + * timestamp and req_seqno are shared by edge_irq_handler() and
> + * edge_irq_thread() which are themselves mutually exclusive.
> + */

I didn't quite get this comment. Does it mean we should define them as
union here? Or is it about how functions consider this data and no
lock is needed? Or what?

> + u64 timestamp;
> + u32 req_seqno;
> + u32 line_seqno;
> };
>
> /**
> @@ -395,12 +418,22 @@ struct line {
> * @gdev: the GPIO device the line request pertains to
> * @label: consumer label used to tag GPIO descriptors
> * @num_lines: the number of lines in the lines array
> + * @wait: wait queue that handles blocking reads of events
> + * @event_buffer_size: the number of elements allocated in @events
> + * @events: KFIFO for the GPIO events
> + * @seqno: the sequence number for edge events generated on all lines in
> + * this line request. Note that this is not used when @num_lines is 1, as
> + * the line_seqno is then the same and is cheaper to calculate.
> * @lines: the lines held by this line request, with @num_lines elements.
> */
> struct linereq {
> struct gpio_device *gdev;
> const char *label;
> u32 num_lines;
> + wait_queue_head_t wait;
> + u32 event_buffer_size;
> + DECLARE_KFIFO_PTR(events, struct gpio_v2_line_event);
> + atomic_t seqno;
> struct line lines[];
> };
>
> @@ -417,12 +450,138 @@ struct linereq {
> (GPIO_V2_LINE_FLAG_OPEN_DRAIN | \
> GPIO_V2_LINE_FLAG_OPEN_SOURCE)
>
> +#define GPIO_V2_LINE_EDGE_FLAGS \
> + (GPIO_V2_LINE_FLAG_EDGE_RISING | \
> + GPIO_V2_LINE_FLAG_EDGE_FALLING)
> +
> #define GPIO_V2_LINE_VALID_FLAGS \
> (GPIO_V2_LINE_FLAG_ACTIVE_LOW | \
> GPIO_V2_LINE_DIRECTION_FLAGS | \
> GPIO_V2_LINE_DRIVE_FLAGS | \
> + GPIO_V2_LINE_EDGE_FLAGS | \
> GPIO_V2_LINE_BIAS_FLAGS)
>
> +static irqreturn_t edge_irq_thread(int irq, void *p)
> +{
> + struct line *line = p;
> + struct linereq *lr = line->req;
> + struct gpio_v2_line_event le;
> + int ret;
> +
> + /* Do not leak kernel stack to userspace */
> + memset(&le, 0, sizeof(le));
> +
> + /*
> + * We may be running from a nested threaded interrupt in which case
> + * we didn't get the timestamp from edge_irq_handler().
> + */

> + if (!line->timestamp) {

Can it be positive conditional?

> + le.timestamp = ktime_get_ns();
> + if (lr->num_lines != 1)
> + line->req_seqno = atomic_inc_return(&lr->seqno);
> + } else {
> + le.timestamp = line->timestamp;
> + }
> + line->timestamp = 0;
> +
> + if (line->eflags == (GPIO_V2_LINE_FLAG_EDGE_RISING |
> + GPIO_V2_LINE_FLAG_EDGE_FALLING)) {
> + int level = gpiod_get_value_cansleep(line->desc);
> +
> + if (level)
> + /* Emit low-to-high event */
> + le.id = GPIO_V2_LINE_EVENT_RISING_EDGE;
> + else
> + /* Emit high-to-low event */
> + le.id = GPIO_V2_LINE_EVENT_FALLING_EDGE;
> + } else if (line->eflags == GPIO_V2_LINE_FLAG_EDGE_RISING) {
> + /* Emit low-to-high event */
> + le.id = GPIO_V2_LINE_EVENT_RISING_EDGE;
> + } else if (line->eflags == GPIO_V2_LINE_FLAG_EDGE_FALLING) {
> + /* Emit high-to-low event */
> + le.id = GPIO_V2_LINE_EVENT_FALLING_EDGE;
> + } else {
> + return IRQ_NONE;
> + }
> + line->line_seqno++;
> + le.line_seqno = line->line_seqno;
> + le.seqno = (lr->num_lines == 1) ? le.line_seqno : line->req_seqno;
> + le.offset = gpio_chip_hwgpio(line->desc);
> +
> + ret = kfifo_in_spinlocked_noirqsave(&lr->events, &le,
> + 1, &lr->wait.lock);
> + if (ret)
> + wake_up_poll(&lr->wait, EPOLLIN);
> + else

> + pr_debug_ratelimited("event FIFO is full - event dropped\n");

Oh, can we really avoid printf() in IRQ context?

> + return IRQ_HANDLED;
> +}
> +
> +static irqreturn_t edge_irq_handler(int irq, void *p)
> +{
> + struct line *line = p;
> + struct linereq *lr = line->req;
> +
> + /*
> + * Just store the timestamp in hardirq context so we get it as
> + * close in time as possible to the actual event.
> + */
> + line->timestamp = ktime_get_ns();
> +
> + if (lr->num_lines != 1)
> + line->req_seqno = atomic_inc_return(&lr->seqno);
> +
> + return IRQ_WAKE_THREAD;
> +}
> +
> +static void edge_detector_stop(struct line *line)
> +{

> + if (line->irq) {
> + free_irq(line->irq, line);
> + line->irq = 0;
> + }

Perhaps

if (!line->irq)
return;

?

> +}
> +
> +static int edge_detector_setup(struct line *line,
> + u64 eflags)
> +{
> + unsigned long irqflags = 0;
> + int irq, ret;
> +
> + if (eflags && !kfifo_initialized(&line->req->events)) {
> + ret = kfifo_alloc(&line->req->events,
> + line->req->event_buffer_size, GFP_KERNEL);
> + if (ret)
> + return ret;
> + }
> + line->eflags = eflags;
> +
> + if (!eflags)
> + return 0;
> +
> + irq = gpiod_to_irq(line->desc);
> + if (irq <= 0)

> + return -ENODEV;

Why shadowing actual error code?

> + if (eflags & GPIO_V2_LINE_FLAG_EDGE_RISING)
> + irqflags |= test_bit(FLAG_ACTIVE_LOW, &line->desc->flags) ?
> + IRQF_TRIGGER_FALLING : IRQF_TRIGGER_RISING;
> + if (eflags & GPIO_V2_LINE_FLAG_EDGE_FALLING)
> + irqflags |= test_bit(FLAG_ACTIVE_LOW, &line->desc->flags) ?
> + IRQF_TRIGGER_RISING : IRQF_TRIGGER_FALLING;
> + irqflags |= IRQF_ONESHOT;
> +
> + /* Request a thread to read the events */
> + ret = request_threaded_irq(irq, edge_irq_handler, edge_irq_thread,
> + irqflags, line->req->label, line);
> + if (ret)
> + return ret;
> +
> + line->irq = irq;
> + return 0;
> +}
> +
> static u64 gpio_v2_line_config_flags(struct gpio_v2_line_config *lc,
> unsigned int line_idx)
> {
> @@ -465,6 +624,11 @@ static int gpio_v2_line_flags_validate(u64 flags)
> (flags & GPIO_V2_LINE_FLAG_OUTPUT))
> return -EINVAL;
>
> + /* Edge detection requires explicit input. */
> + if ((flags & GPIO_V2_LINE_EDGE_FLAGS) &&
> + !(flags & GPIO_V2_LINE_FLAG_INPUT))
> + return -EINVAL;
> +
> /*
> * Do not allow OPEN_SOURCE & OPEN_DRAIN flags in a single request. If
> * the hardware actually supports enabling both at the same time the
> @@ -526,6 +690,10 @@ static void gpio_v2_line_config_flags_to_desc_flags(u64 flags,
> set_bit(FLAG_IS_OUT, flagsp);
> else if (flags & GPIO_V2_LINE_FLAG_INPUT)
> clear_bit(FLAG_IS_OUT, flagsp);
> + assign_bit(FLAG_EDGE_RISING, flagsp,
> + flags & GPIO_V2_LINE_FLAG_EDGE_RISING);
> + assign_bit(FLAG_EDGE_FALLING, flagsp,
> + flags & GPIO_V2_LINE_FLAG_EDGE_FALLING);
> assign_bit(FLAG_OPEN_DRAIN, flagsp,
> flags & GPIO_V2_LINE_FLAG_OPEN_DRAIN);
> assign_bit(FLAG_OPEN_SOURCE, flagsp,
> @@ -614,14 +782,85 @@ static long linereq_ioctl_compat(struct file *file, unsigned int cmd,
> }
> #endif
>
> +static __poll_t linereq_poll(struct file *file,
> + struct poll_table_struct *wait)
> +{
> + struct linereq *lr = file->private_data;
> + __poll_t events = 0;
> +
> + poll_wait(file, &lr->wait, wait);
> +
> + if (!kfifo_is_empty_spinlocked_noirqsave(&lr->events,
> + &lr->wait.lock))
> + events = EPOLLIN | EPOLLRDNORM;
> +
> + return events;
> +}
> +
> +static ssize_t linereq_read(struct file *file,
> + char __user *buf,
> + size_t count,
> + loff_t *f_ps)
> +{
> + struct linereq *lr = file->private_data;
> + struct gpio_v2_line_event le;
> + ssize_t bytes_read = 0;
> + int ret;
> +
> + if (count < sizeof(le))
> + return -EINVAL;
> +
> + do {
> + spin_lock(&lr->wait.lock);
> + if (kfifo_is_empty(&lr->events)) {
> + if (bytes_read) {
> + spin_unlock(&lr->wait.lock);
> + return bytes_read;
> + }
> +
> + if (file->f_flags & O_NONBLOCK) {
> + spin_unlock(&lr->wait.lock);
> + return -EAGAIN;
> + }
> +
> + ret = wait_event_interruptible_locked(lr->wait,
> + !kfifo_is_empty(&lr->events));
> + if (ret) {
> + spin_unlock(&lr->wait.lock);
> + return ret;
> + }
> + }
> +
> + ret = kfifo_out(&lr->events, &le, 1);
> + spin_unlock(&lr->wait.lock);
> + if (ret != 1) {
> + /*
> + * This should never happen - we were holding the
> + * lock from the moment we learned the fifo is no
> + * longer empty until now.
> + */
> + ret = -EIO;
> + break;
> + }
> +
> + if (copy_to_user(buf + bytes_read, &le, sizeof(le)))
> + return -EFAULT;
> + bytes_read += sizeof(le);
> + } while (count >= bytes_read + sizeof(le));
> +
> + return bytes_read;
> +}
> +
> static void linereq_free(struct linereq *lr)
> {
> unsigned int i;
>
> for (i = 0; i < lr->num_lines; i++) {
> + edge_detector_stop(&lr->lines[i]);
> if (lr->lines[i].desc)
> gpiod_free(lr->lines[i].desc);
> }
> + kfifo_free(&lr->events);
> kfree(lr->label);
> put_device(&lr->gdev->dev);
> kfree(lr);
> @@ -637,6 +876,8 @@ static int linereq_release(struct inode *inode, struct file *file)
>
> static const struct file_operations line_fileops = {
> .release = linereq_release,
> + .read = linereq_read,
> + .poll = linereq_poll,
> .owner = THIS_MODULE,
> .llseek = noop_llseek,
> .unlocked_ioctl = linereq_ioctl,
> @@ -676,6 +917,9 @@ static int linereq_create(struct gpio_device *gdev, void __user *ip)
> lr->gdev = gdev;
> get_device(&gdev->dev);
>

> + for (i = 0; i < ulr.num_lines; i++)
> + lr->lines[i].req = lr;

> /* Make sure this is terminated */
> ulr.consumer[sizeof(ulr.consumer)-1] = '\0';
> if (strlen(ulr.consumer)) {
> @@ -686,6 +930,14 @@ static int linereq_create(struct gpio_device *gdev, void __user *ip)
> }
> }
>
> + init_waitqueue_head(&lr->wait);
> + lr->event_buffer_size = ulr.event_buffer_size;
> + if (lr->event_buffer_size == 0)

> + lr->event_buffer_size = ulr.num_lines*16;
> + else if (lr->event_buffer_size > GPIO_V2_LINES_MAX*16)
> + lr->event_buffer_size = GPIO_V2_LINES_MAX*16;

*16 -> ' * 16'

> +
> + atomic_set(&lr->seqno, 0);
> lr->num_lines = ulr.num_lines;
>
> /* Request each GPIO */
> @@ -724,6 +976,11 @@ static int linereq_create(struct gpio_device *gdev, void __user *ip)
> ret = gpiod_direction_input(desc);
> if (ret)
> goto out_free_linereq;
> +
> + ret = edge_detector_setup(&lr->lines[i],
> + flags & GPIO_V2_LINE_EDGE_FLAGS);
> + if (ret)
> + goto out_free_linereq;
> }
>
> blocking_notifier_call_chain(&desc->gdev->notifier,
> @@ -1245,6 +1502,11 @@ static void gpio_desc_to_lineinfo(struct gpio_desc *desc,
> if (test_bit(FLAG_PULL_UP, &desc->flags))
> info->flags |= GPIO_V2_LINE_FLAG_BIAS_PULL_UP;
>
> + if (test_bit(FLAG_EDGE_RISING, &desc->flags))
> + info->flags |= GPIO_V2_LINE_FLAG_EDGE_RISING;
> + if (test_bit(FLAG_EDGE_FALLING, &desc->flags))
> + info->flags |= GPIO_V2_LINE_FLAG_EDGE_FALLING;
> +
> spin_unlock_irqrestore(&gpio_lock, flags);
> }
>
> diff --git a/drivers/gpio/gpiolib.c b/drivers/gpio/gpiolib.c
> index 80137c1b3cdc..e4c81dca7f8b 100644
> --- a/drivers/gpio/gpiolib.c
> +++ b/drivers/gpio/gpiolib.c
> @@ -2041,6 +2041,8 @@ static bool gpiod_free_commit(struct gpio_desc *desc)
> clear_bit(FLAG_PULL_UP, &desc->flags);
> clear_bit(FLAG_PULL_DOWN, &desc->flags);
> clear_bit(FLAG_BIAS_DISABLE, &desc->flags);
> + clear_bit(FLAG_EDGE_RISING, &desc->flags);
> + clear_bit(FLAG_EDGE_FALLING, &desc->flags);
> clear_bit(FLAG_IS_HOGGED, &desc->flags);
> #ifdef CONFIG_OF_DYNAMIC
> desc->hog = NULL;
> diff --git a/drivers/gpio/gpiolib.h b/drivers/gpio/gpiolib.h
> index 6709f79c02dd..39b356160937 100644
> --- a/drivers/gpio/gpiolib.h
> +++ b/drivers/gpio/gpiolib.h
> @@ -114,6 +114,8 @@ struct gpio_desc {
> #define FLAG_PULL_UP 13 /* GPIO has pull up enabled */
> #define FLAG_PULL_DOWN 14 /* GPIO has pull down enabled */
> #define FLAG_BIAS_DISABLE 15 /* GPIO has pull disabled */
> +#define FLAG_EDGE_RISING 16 /* GPIO CDEV detects rising edge events */
> +#define FLAG_EDGE_FALLING 17 /* GPIO CDEV detects falling edge events */
>
> /* Connection label */
> const char *label;
> --
> 2.28.0
>


--
With Best Regards,
Andy Shevchenko