When an external device generated a level based interrupt then the
interrupt controller could miss the interrupt. The reason is that the
interrupt controller can detect only link changes.
In the following example, if there is a PHY that generates an interrupt
then the following would happen. The GPIO detected that the interrupt
line changed, and then the 'ocelot_irq_handler' will be called. Here it
detects which GPIO line seen the change and for that will call the
following:
1. irq_mask
2. phy interrupt routine
3. irq_eoi
4. irq_unmask
And this works fine for simple cases, but if the PHY generates many
interrupts, for example when doing PTP timestamping, then the following
could happen. Again the function 'ocelot_irq_handler' will be called
and then from here the following could happen:
1. irq_mask
2. phy interrupt routine
3. irq_eoi
4. irq_unmask
Right before step 3(irq_eoi), the PHY will generate another interrupt.
Now the interrupt controller will acknowledge the change in the
interrupt line. So we miss the interrupt.
A solution will be to use 'handle_level_irq' instead of
'handle_fasteoi_irq', because for this will change routine order of
handling the interrupt.
1. irq_mask
2. irq_ack
3. phy interrupt routine
4. irq_unmask
And now if the PHY will generate a new interrupt before irq_unmask, the
interrupt controller will detect this because it already acknowledge the
change in interrupt line at step 2(irq_ack).
But this is not the full solution because there is another issue. In
case there are 2 PHYs that share the interrupt line. For example phy1
generates an interrupt, then the following can happen:
1.irq_mask
2.irq_ack
3.phy0 interrupt routine
4.phy1 interrupt routine
5.irq_unmask
In case phy0 will generate an interrupt while clearing the interrupt
source in phy1, then the interrupt line will be kept down by phy0. So
the interrupt controller will not see any changes in the interrupt line.
The solution here is to update 'irq_unmask' such that it can detect if
the interrupt line is still active or not. And if it is active then call
again the procedure to clear the interrupts. But we don't want to do it
every time, only if we know that the interrupt controller have not seen
already that the interrupt line has changed.
While at this, add support also for IRQ_TYPE_LEVEL_LOW.
Fixes: be36abb71d878f ("pinctrl: ocelot: add support for interrupt controller")
Signed-off-by: Horatiu Vultur <[email protected]>
---
v1->v2:
- fix grammer mistakes
- remove redundant checks for type in ocelot_irq_set_type
- split assignment and declaration of work variable
- use irqd_get_trigger_type instead of getting trigger from action
---
drivers/pinctrl/pinctrl-ocelot.c | 93 +++++++++++++++++++++++++++-----
1 file changed, 80 insertions(+), 13 deletions(-)
diff --git a/drivers/pinctrl/pinctrl-ocelot.c b/drivers/pinctrl/pinctrl-ocelot.c
index c5fd154990c8..29887d55c93b 100644
--- a/drivers/pinctrl/pinctrl-ocelot.c
+++ b/drivers/pinctrl/pinctrl-ocelot.c
@@ -338,6 +338,11 @@ struct ocelot_match_data {
struct ocelot_pincfg_data pincfg_data;
};
+struct ocelot_irq_work {
+ struct work_struct irq_work;
+ struct irq_desc *irq_desc;
+};
+
#define LUTON_P(p, f0, f1) \
static struct ocelot_pin_caps luton_pin_##p = { \
.pin = p, \
@@ -1813,6 +1818,74 @@ static void ocelot_irq_mask(struct irq_data *data)
gpiochip_disable_irq(chip, gpio);
}
+static void ocelot_irq_work(struct work_struct *work)
+{
+ struct ocelot_irq_work *w = container_of(work, struct ocelot_irq_work, irq_work);
+ struct irq_chip *parent_chip = irq_desc_get_chip(w->irq_desc);
+ struct gpio_chip *chip = irq_desc_get_chip_data(w->irq_desc);
+ struct irq_data *data = irq_desc_get_irq_data(w->irq_desc);
+ unsigned int gpio = irqd_to_hwirq(data);
+
+ local_irq_disable();
+ chained_irq_enter(parent_chip, w->irq_desc);
+ generic_handle_domain_irq(chip->irq.domain, gpio);
+ chained_irq_exit(parent_chip, w->irq_desc);
+ local_irq_enable();
+
+ kfree(w);
+}
+
+static void ocelot_irq_unmask_level(struct irq_data *data)
+{
+ struct gpio_chip *chip = irq_data_get_irq_chip_data(data);
+ struct ocelot_pinctrl *info = gpiochip_get_data(chip);
+ struct irq_desc *desc = irq_data_to_desc(data);
+ unsigned int gpio = irqd_to_hwirq(data);
+ bool ack = false, active = false;
+ u8 trigger_level;
+ int val;
+
+ trigger_level = irqd_get_trigger_type(data);
+
+ /* Check if the interrupt line is still active. */
+ regmap_read(info->map, REG(OCELOT_GPIO_IN, info, gpio), &val);
+ if ((!(val & BIT(gpio % 32)) && trigger_level == IRQ_TYPE_LEVEL_LOW) ||
+ (val & BIT(gpio % 32) && trigger_level == IRQ_TYPE_LEVEL_HIGH))
+ active = true;
+
+ /*
+ * Check if the interrupt controller has seen any changes in the
+ * interrupt line.
+ */
+ regmap_read(info->map, REG(OCELOT_GPIO_INTR, info, gpio), &val);
+ if (val & BIT(gpio % 32))
+ ack = true;
+
+ /* Enable the interrupt now */
+ gpiochip_enable_irq(chip, gpio);
+ regmap_update_bits(info->map, REG(OCELOT_GPIO_INTR_ENA, info, gpio),
+ BIT(gpio % 32), BIT(gpio % 32));
+
+ /*
+ * In case the interrupt line is still active and the interrupt
+ * controller has not seen any changes in the interrupt line, then it
+ * means that there happen another interrupt while the line was active.
+ * So we missed that one, so we need to kick again the interrupt
+ * handler.
+ */
+ if (active && !ack) {
+ struct ocelot_irq_work *work;
+
+ work = kmalloc(sizeof(*work), GFP_ATOMIC);
+ if (!work)
+ return;
+
+ work->irq_desc = desc;
+ INIT_WORK(&work->irq_work, ocelot_irq_work);
+ queue_work(system_wq, &work->irq_work);
+ }
+}
+
static void ocelot_irq_unmask(struct irq_data *data)
{
struct gpio_chip *chip = irq_data_get_irq_chip_data(data);
@@ -1836,13 +1909,12 @@ static void ocelot_irq_ack(struct irq_data *data)
static int ocelot_irq_set_type(struct irq_data *data, unsigned int type);
-static struct irq_chip ocelot_eoi_irqchip = {
+static struct irq_chip ocelot_level_irqchip = {
.name = "gpio",
.irq_mask = ocelot_irq_mask,
- .irq_eoi = ocelot_irq_ack,
- .irq_unmask = ocelot_irq_unmask,
- .flags = IRQCHIP_EOI_THREADED | IRQCHIP_EOI_IF_HANDLED |
- IRQCHIP_IMMUTABLE,
+ .irq_ack = ocelot_irq_ack,
+ .irq_unmask = ocelot_irq_unmask_level,
+ .flags = IRQCHIP_IMMUTABLE,
.irq_set_type = ocelot_irq_set_type,
GPIOCHIP_IRQ_RESOURCE_HELPERS
};
@@ -1859,14 +1931,9 @@ static struct irq_chip ocelot_irqchip = {
static int ocelot_irq_set_type(struct irq_data *data, unsigned int type)
{
- type &= IRQ_TYPE_SENSE_MASK;
-
- if (!(type & (IRQ_TYPE_EDGE_BOTH | IRQ_TYPE_LEVEL_HIGH)))
- return -EINVAL;
-
- if (type & IRQ_TYPE_LEVEL_HIGH)
- irq_set_chip_handler_name_locked(data, &ocelot_eoi_irqchip,
- handle_fasteoi_irq, NULL);
+ if (type & (IRQ_TYPE_LEVEL_HIGH | IRQ_TYPE_LEVEL_LOW))
+ irq_set_chip_handler_name_locked(data, &ocelot_level_irqchip,
+ handle_level_irq, NULL);
if (type & IRQ_TYPE_EDGE_BOTH)
irq_set_chip_handler_name_locked(data, &ocelot_irqchip,
handle_edge_irq, NULL);
--
2.33.0
On Wed, Sep 7, 2022 at 10:59 AM Horatiu Vultur
<[email protected]> wrote:
>
> When an external device generated a level based interrupt then the
> interrupt controller could miss the interrupt. The reason is that the
> interrupt controller can detect only link changes.
>
> In the following example, if there is a PHY that generates an interrupt
> then the following would happen. The GPIO detected that the interrupt
> line changed, and then the 'ocelot_irq_handler' will be called. Here it
was called
> detects which GPIO line seen the change and for that will call the
saw
> following:
> 1. irq_mask
> 2. phy interrupt routine
> 3. irq_eoi
> 4. irq_unmask
>
> And this works fine for simple cases, but if the PHY generates many
> interrupts, for example when doing PTP timestamping, then the following
> could happen. Again the function 'ocelot_irq_handler' will be called
> and then from here the following could happen:
> 1. irq_mask
> 2. phy interrupt routine
> 3. irq_eoi
> 4. irq_unmask
>
> Right before step 3(irq_eoi), the PHY will generate another interrupt.
> Now the interrupt controller will acknowledge the change in the
> interrupt line. So we miss the interrupt.
>
> A solution will be to use 'handle_level_irq' instead of
> 'handle_fasteoi_irq', because for this will change routine order of
> handling the interrupt.
> 1. irq_mask
> 2. irq_ack
> 3. phy interrupt routine
> 4. irq_unmask
>
> And now if the PHY will generate a new interrupt before irq_unmask, the
> interrupt controller will detect this because it already acknowledge the
> change in interrupt line at step 2(irq_ack).
>
> But this is not the full solution because there is another issue. In
> case there are 2 PHYs that share the interrupt line. For example phy1
> generates an interrupt, then the following can happen:
> 1.irq_mask
> 2.irq_ack
> 3.phy0 interrupt routine
> 4.phy1 interrupt routine
> 5.irq_unmask
>
> In case phy0 will generate an interrupt while clearing the interrupt
> source in phy1, then the interrupt line will be kept down by phy0. So
> the interrupt controller will not see any changes in the interrupt line.
> The solution here is to update 'irq_unmask' such that it can detect if
> the interrupt line is still active or not. And if it is active then call
> again the procedure to clear the interrupts. But we don't want to do it
> every time, only if we know that the interrupt controller have not seen
has not seen
> already that the interrupt line has changed.
>
> While at this, add support also for IRQ_TYPE_LEVEL_LOW.
...
> + regmap_read(info->map, REG(OCELOT_GPIO_IN, info, gpio), &val);
> + if ((!(val & BIT(gpio % 32)) && trigger_level == IRQ_TYPE_LEVEL_LOW) ||
> + (val & BIT(gpio % 32) && trigger_level == IRQ_TYPE_LEVEL_HIGH))
> + active = true;
You can use temporary variable for the bit, like
unsigned int bit = BIT(gpio % 32);
...
> + /*
> + * In case the interrupt line is still active and the interrupt
> + * controller has not seen any changes in the interrupt line, then it
> + * means that there happen another interrupt while the line was active.
> + * So we missed that one, so we need to kick again the interrupt
the interrupt again
> + * handler.
> + */
> + if (active && !ack) {
> + struct ocelot_irq_work *work;
> +
> + work = kmalloc(sizeof(*work), GFP_ATOMIC);
> + if (!work)
> + return;
> +
> + work->irq_desc = desc;
> + INIT_WORK(&work->irq_work, ocelot_irq_work);
> + queue_work(system_wq, &work->irq_work);
> + }
Here I see potential issues with the object lifetime. 1) The memory is
allocated here and what does guarantee its freeing? 2) What does
guarantee that work will be not scheduled if the driver or its parts
are gone?
--
With Best Regards,
Andy Shevchenko
The 09/07/2022 11:36, Andy Shevchenko wrote:
> > + * handler.
> > + */
> > + if (active && !ack) {
> > + struct ocelot_irq_work *work;
> > +
> > + work = kmalloc(sizeof(*work), GFP_ATOMIC);
> > + if (!work)
> > + return;
> > +
> > + work->irq_desc = desc;
> > + INIT_WORK(&work->irq_work, ocelot_irq_work);
> > + queue_work(system_wq, &work->irq_work);
> > + }
>
> Here I see potential issues with the object lifetime. 1) The memory is
> allocated here and what does guarantee its freeing? 2) What does
> guarantee that work will be not scheduled if the driver or its parts
> are gone?
You are right, if the driver is removed once the work is queued, then
that object will not be freed or even worst get NULL pointers.
A solution to this would be not to use system_wq but allocate own workqueue
and once the driver is removed, make sure to destroy it.
In this way we make sure that all the work is done.
>
> --
> With Best Regards,
> Andy Shevchenko
--
/Horatiu
On Wed, Sep 7, 2022 at 9:59 AM Horatiu Vultur
<[email protected]> wrote:
> When an external device generated a level based interrupt then the
> interrupt controller could miss the interrupt. The reason is that the
> interrupt controller can detect only link changes.
I see there are some further comments, I expect to just merge v3 into
fixes as this needs to go into the -rc:s right?
Yours,
Linus Walleij
The 09/08/2022 10:46, Linus Walleij wrote:
>
> On Wed, Sep 7, 2022 at 9:59 AM Horatiu Vultur
> <[email protected]> wrote:
>
> > When an external device generated a level based interrupt then the
> > interrupt controller could miss the interrupt. The reason is that the
> > interrupt controller can detect only link changes.
>
> I see there are some further comments, I expect to just merge v3 into
> fixes as this needs to go into the -rc:s right?
That is correct, it should go into fixes.
>
> Yours,
> Linus Walleij
--
/Horatiu