Inter-Processor Call¶

Overview¶

Due to the dual core nature of the ESP32, there are instances where a certain function must be run in the context of a particular core (e.g. allocating ISR to an interrupt source of a particular core). The IPC (Inter-Processor Call) feature allows for the execution of functions on a particular CPU.

A given function can be executed on a particular core by calling esp_ipc_call() or esp_ipc_call_blocking(). IPC is implemented via two high priority FreeRTOS tasks pinned to each CPU known as the IPC Tasks. The two IPC Tasks remain inactive (blocked) until esp_ipc_call() or esp_ipc_call_blocking() is called. When an IPC Task of a particular core is unblocked, it will preempt the current running task on that core and execute a given function.

Usage¶

esp_ipc_call() unblocks the IPC task on a particular core to execute a given function. The task that calls esp_ipc_call() will be blocked until the IPC Task begins execution of the given function. esp_ipc_call_blocking() is similar but will block the calling task until the IPC Task has completed execution of the given function.

Functions executed by IPCs must be functions of type void func(void *arg). To run more complex functions which require a larger stack, the IPC tasks’ stack size can be configured by modifying CONFIG_ESP_IPC_TASK_STACK_SIZE in menuconfig. The IPC API is protected by a mutex hence simultaneous IPC calls are not possible.

Care should taken to avoid deadlock when writing functions to be executed by IPC, especially when attempting to take a mutex within the function.

Hi-priority IPC¶

In some cases, we need to quickly get the state of the other CPU (for example, in the core dump, in the GDB stub, in some unit tests, in the dport workaround, etc.), and usually it can be implemented as a fairly simple function and can be written in assembler. For these purposes, the IPC API has special functions esp_ipc_isr_asm_call() and esp_ipc_isr_asm_call_blocking(). The esp_ipc_isr_asm…(asm_func, arg) functions trigger the High-priority interrupt (4 level) on the other CPU. The given assembler function (asm_func) will run on the other CPU in the context of the interrupt. This assembler function will be called by the CALLX0 command therefore it must be written in assembler.

esp_ipc_isr_asm_call() - triggers the Hi-priority interrupt on a particular core to execute a given function. The CPU which called this function will be blocked until the other CPU begins execution of the given function.

esp_ipc_isr_asm_call_blocking() is similar but will block the calling CPU until the other CPU has completed the execution of the given function.

esp_ipc_isr_stall_other_cpu() stalls the other CPU and the calling CPU disables interrupts with level 3 and lower. To finish stalling the other CPU call esp_ipc_isr_release_other_cpu(). The stalled CPU disables interrupts with level 5 and lower.

Functions executed by Hi-priority IPC must be functions of type void func(void *arg). Examples of a assembler function see in components/esp_ipc/src/esp_ipc_isr/esp_ipc_isr_routines.S, components/esp_ipc/test/test_ipc_isr.S and below. In the asm function, you can use only a few registers as they were saved in the interrupt handler before calling this function, their use is safe. The registers:a2 as void *arg, a3 and a4 are free for use.

Some feature:

The asm function should be placed in the IRAM memory and aligned on a memory address multiple of 4.
As the asm function is run in the context of the High-priority interrupt, a C function can no be called because the windows spill is disabled.
Use only a2, a3 and a4 registers in the asm function (to workaround it see esp_complex_asm_func).
A CPU, that called these APIs esp_ipc_isr_asm…(asm_func, arg), disables interrupts with level 3 and lower.
A CPU, where the asm function is executed, disables interrupts with level 5 and lower.
You do not need to take care about handling of the Hi-priority interrupt.

/* esp_test_ipc_isr_get_cycle_count_other_cpu(void *arg) */
// this function puts CCOUNT of the other CPU in the arg.
// use only a2, a3 and a4 regs here.
.section    .iram1, "ax"
.align      4
.global     esp_test_ipc_isr_get_cycle_count_other_cpu
.type       esp_test_ipc_isr_get_cycle_count_other_cpu, @function
// Args:
// a2 - void* arg
esp_test_ipc_isr_get_cycle_count_other_cpu:
rsr.ccount a3
s32i    a3, a2, 0
ret

This number of registers available for use is sufficient for most simple cases. But if you need to run a more complex asm function, you can pass as an argument a pointer to a structure that can accept additional registers to a buffer to make them free to use. Remember to restore them before returning See the system/ipc/ipc_isr.

API Reference¶

Header File¶

components/esp_ipc/include/esp_ipc.h

Functions¶

esp_err_t esp_ipc_call(uint32_t cpu_id, esp_ipc_func_t func, void *arg)¶

Execute a function on the given CPU.

Run a given function on a particular CPU. The given function must accept a void* argument and return void. The given function is run in the context of the IPC task of the CPU specified by the cpu_id parameter. The calling task will be blocked until the IPC task begins executing the given function. If another IPC call is ongoing, the calling task will block until the other IPC call completes. The stack size allocated for the IPC task can be configured in the “Inter-Processor Call (IPC) task stack size” setting in menuconfig. Increase this setting if the given function requires more stack than default.

Note

In single-core mode, returns ESP_ERR_INVALID_ARG for cpu_id 1.

Return

ESP_ERR_INVALID_ARG if cpu_id is invalid
ESP_ERR_INVALID_STATE if the FreeRTOS scheduler is not running
ESP_OK otherwise

Parameters

[in] cpu_id: CPU where the given function should be executed (0 or 1)
[in] func: Pointer to a function of type void func(void* arg) to be executed
[in] arg: Arbitrary argument of type void* to be passed into the function

esp_err_t esp_ipc_call_blocking(uint32_t cpu_id, esp_ipc_func_t func, void *arg)¶

Execute a function on the given CPU and blocks until it completes.

Run a given function on a particular CPU. The given function must accept a void* argument and return void. The given function is run in the context of the IPC task of the CPU specified by the cpu_id parameter. The calling task will be blocked until the IPC task completes execution of the given function. If another IPC call is ongoing, the calling task will block until the other IPC call completes. The stack size allocated for the IPC task can be configured in the “Inter-Processor Call (IPC) task stack size” setting in menuconfig. Increase this setting if the given function requires more stack than default.

Note

In single-core mode, returns ESP_ERR_INVALID_ARG for cpu_id 1.

Return

ESP_ERR_INVALID_ARG if cpu_id is invalid
ESP_ERR_INVALID_STATE if the FreeRTOS scheduler is not running
ESP_OK otherwise

Parameters

[in] cpu_id: CPU where the given function should be executed (0 or 1)
[in] func: Pointer to a function of type void func(void* arg) to be executed
[in] arg: Arbitrary argument of type void* to be passed into the function

Provide feedback about this document