Peripherals that can generate interrupts can be divided in two types:
- External peripherals, within the ESP32 but outside the Xtensa cores themselves. Most ESP32 peripherals are of this type.
- Internal peripherals, part of the Xtensa CPU cores themselves.
Interrupt handling differs slightly between these two types of peripherals.
Internal peripheral interrupts
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
Each Xtensa CPU core has its own set of six internal peripherals:
- Three timer comparators
- A performance monitor
- Two software interrupts.
Internal interrupt sources are defined in esp_intr_alloc.h as ``ETS_INTERNAL_*_INTR_SOURCE``.
These peripherals can only be configured from the core they are associated with. When generating an interrupt,
the interrupt they generate is hard-wired to their associated core; it's not possible to have e.g. an internal
timer comparator of one core generate an interrupt on another core. That is why these sources can only be managed
using a task running on that specific core. Internal interrupt sources are still allocatable using esp_intr_alloc
as normal, but they cannot be shared and will always have a fixed interrupt level (namely, the one associated in
hardware with the peripheral).
External Peripheral Interrupts
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
The remaining interrupt sources are from external peripherals. These are defined in soc/soc.h as ``ETS_*_INTR_SOURCE``.
Non-internal interrupt slots in both CPU cores are wired to an interrupt multiplexer, which can be used to
route any external interrupt source to any of these interrupt slots.
- Allocating an external interrupt will always allocate it on the core that does the allocation.
- Freeing an external interrupt must always happen on the same core it was allocated on.
- Disabling and enabling external interrupts from another core is allowed.
- Multiple external interrupt sources can share an interrupt slot by passing ``ESP_INTR_FLAG_SHARED`` as a flag to esp_intr_alloc().
Care should be taken when calling esp_intr_alloc() from a task which is not pinned to a core. During task switching, these tasks can migrate between cores. Therefore it is impossible to tell which CPU the interrupt is allocated on, which makes it difficult to free the interrupt handle and may also cause debugging difficulties. It is advised to use xTaskCreatePinnedToCore() with a specific CoreID argument to create tasks that will allocate interrupts. In the case of internal interrupt sources, this is required.
IRAM-Safe Interrupt Handlers
----------------------------
The ``ESP_INTR_FLAG_IRAM`` flag registers an interrupt handler that always runs from IRAM (and reads all its data from DRAM), and therefore does not need to be disabled during flash erase and write operations.
This is useful for interrupts which need a guaranteed minimum execution latency, as flash write and erase operations can be slow (erases can take tens or hundreds of milliseconds to complete).
It can also be useful to keep an interrupt handler in IRAM if it is called very frequently, to avoid flash cache misses.
Refer to the :ref:`SPI flash API documentation <iram-safe-interrupt-handlers>` for more details.
Several handlers can be assigned to a same source, given that all handlers are allocated using the ``ESP_INTR_FLAG_SHARED`` flag.
They'll be all allocated to the interrupt, which the source is attached to, and called sequentially when the source is active.
The handlers can be disabled and freed individually. The source is attached to the interrupt (enabled), if one or more handlers are enabled, otherwise detached.
A handler will never be called when disabled, while **its source may still be triggered** if any one of its handler enabled.
Sources attached to non-shared interrupt do not support this feature.
Though the framework support this feature, you have to use it *very carefully*. There usually exist 2 ways to stop a interrupt from being triggered: *disable the sourse* or *mask peripheral interrupt status*.
IDF only handles the enabling and disabling of the source itself, leaving status and mask bits to be handled by users. **Status bits should always be masked before the handler responsible for it is disabled,
or the status should be handled in other enabled interrupt properly**. You may leave some status bits unhandled if you just disable one of all the handlers without mask the status bits, which causes the interrupt being triggered infinitely,
