zephyr: Implement machine.Pin.irq() for setting callbacks on pin change. #6146
Conversation
dpgeorge
force-pushed
the
zephyr-pin-irq-v2
branch
from
bc989d3
to
e8f35b2
Compare
There was a problem hiding this comment.
Thanks for this @dpgeorge . I need to come up to speed on uevent and uasyncio
I tested this on mimxrt1050_evk and frdm_k64f. Hard interrupts work on both, but soft interrupts don't seem to work on either. Is that expected?
ports/zephyr/machine_pin.c
Outdated
| } | ||
|
|
||
| STATIC void gpio_callback_handler(struct device *port, struct gpio_callback *cb, gpio_port_pins_t pins) { | ||
| machine_pin_irq_obj_t *irq = (void *)((uintptr_t)cb - offsetof(machine_pin_irq_obj_t, callback)); |
There was a problem hiding this comment.
Zephyr has a useful CONTAINER_OF macro:
machine_pin_irq_obj_t *irq = CONTAINER_OF(cb, machine_pin_irq_obj_t, callback);| const mp_obj_base_t machine_pin_obj_template = {&machine_pin_type}; | ||
|
|
||
| void machine_pin_deinit(void) { | ||
| for (machine_pin_irq_obj_t *irq = MP_STATE_PORT(machine_pin_irq_list); irq != NULL; irq = irq->next) { |
There was a problem hiding this comment.
I'd consider using a Zephyr slist
There was a problem hiding this comment.
I had look at slist and I don't think it'll work: for the garbage collector to trace the links of the list (which are allocated on the uPy heap) the link node pointers must point to the start of the allocated struct, but if it used slist then the next pointers would point to the interior of the struct. Note also that mp_irq_obj_t base; must go at the start of the struct so it's a proper uPy object, so sys_snode_t would need to go in the interior of the struct.
ports/zephyr/machine_pin.c
Outdated
|
|
||
| if (n_args > 1 || kw_args->used != 0) { | ||
| // configure irq | ||
| self->irq->callback.pin_mask &= ~BIT(self->pin); |
There was a problem hiding this comment.
We shouldn't be modifying the callback structure internals; pin_mask should get set by gpio_init_callback.
There was a problem hiding this comment.
Ok. I was a bit uneasy about modifying it directly but the comment in gpio.h did say "Such pin_mask can be modified whenever necessary by the owner".
Anyway, I changed it now and made it a bit simpler so gpio_init_callback() is only called once with BIT(self->pin), and gpio_pin_interrupt_configure() is used exclusively to disable/enable the IRQ.
|
Thanks for the review!
uevent is still WIP, and understanding how polling on zephyr works could help to refine how uevent behaves. Ultimately we want to be able to sleep efficiently while waiting for a set of events, including a pin edge/irq. But you don't need to know about these modules to use pin irq callbacks.
They should work if the VM is busy, eg: import time, machine
usr = machine.Pin(('GPIO_2', 6))
usr.irq(lambda p: print('irq', p)) # soft by default
# busy loop so that soft callbacks are processed
for i in range(5000):
time.sleep_ms(1) |
So it can be unconditionally included in a port's build even if certain configurations in that port do not use its features, to simplify the Makefile. Signed-off-by: Damien George <damien@micropython.org>
Supports hard and soft interrupts. In the current implementation, soft interrupt callbacks will only be called when the VM is executing, ie they will not be called during a blocking kernel call like k_msleep. And the behaviour of hard interrupt callbacks will depend on the underlying device, as well as the amount of ISR stack space. Soft and hard interrupts tested on frdm_k64f and nucleo_f767zi boards. Signed-off-by: Damien George <damien@micropython.org>
Signed-off-by: Damien George <damien@micropython.org>
dpgeorge
force-pushed
the
zephyr-pin-irq-v2
branch
from
4fd44e3
to
41b7734
Compare
|
Ok, this was merged. It possible to make follow-up changes if needed. |
This PR adds support for hard and soft pin interrupts in the zephyr port. In the current implementation, soft interrupt callbacks will only be called when the VM is executing, ie they will not be called during a blocking kernel call like k_msleep. And the behaviour of hard interrupt callbacks will depend on the underlying device, as well as the amount of ISR stack space.
Soft and hard interrupts tested on frdm_k64f and nucleo_f767zi boards.
This will eventually help with testing #6110 and #6125 (uevent and uasyncio)
To test on a frdm_k64f board do:
@MaureenHelm FYI