Application Startup Flow

This note explains various steps which happen before app_main function of an ESP-IDF application is called.

The high level view of startup process is as follows:

First stage bootloader in ROM loads second-stage bootloader image to RAM (IRAM & DRAM) from flash offset 0x0.
Second stage bootloader loads partition table and main app image from flash. Main app incorporates both RAM segments and read-only segments mapped via flash cache.
Application startup executes. At this point the second CPU and RTOS scheduler are started.

This process is explained in detail in the following sections.

First stage bootloader

After SoC reset, PRO CPU will start running immediately, executing reset vector code, while APP CPU will be held in reset. During startup process, PRO CPU does all the initialization. APP CPU reset is de-asserted in the call_start_cpu0 function of application startup code. Reset vector code is located in the mask ROM of the ESP32-S3 chip and cannot be modified.

Startup code called from the reset vector determines the boot mode by checking GPIO_STRAP_REG register for bootstrap pin states. Depending on the reset reason, the following takes place:

Reset from deep sleep: if the value in RTC_CNTL_STORE6_REG is non-zero, and CRC value of RTC memory in RTC_CNTL_STORE7_REG is valid, use RTC_CNTL_STORE6_REG as an entry point address and jump immediately to it. If RTC_CNTL_STORE6_REG is zero, or RTC_CNTL_STORE7_REG contains invalid CRC, or once the code called via RTC_CNTL_STORE6_REG returns, proceed with boot as if it was a power-on reset. Note: to run customized code at this point, a deep sleep stub mechanism is provided. Please see deep sleep documentation for this.
For power-on reset, software SOC reset, and watchdog SOC reset: check the GPIO_STRAP_REG register if a custom boot mode (such as UART Download Mode) is requested. If this is the case, this custom loader mode is executed from ROM. Otherwise, proceed with boot as if it was due to software CPU reset. Consult ESP32-S3 datasheet for a description of SoC boot modes and how to execute them.
For software CPU reset and watchdog CPU reset: configure SPI flash based on EFUSE values, and attempt to load the code from flash. This step is described in more detail in the next paragraphs.

Note

During normal boot modes the RTC watchdog is enabled when this happens, so if the process is interrupted or stalled then the watchdog will reset the SOC automatically and repeat the boot process. This may cause the SoC to strap into a new boot mode, if the strapping GPIOs have changed.

Second stage bootloader binary image is loaded from the start of flash at offset 0x0.

Second stage bootloader

In ESP-IDF, the binary image which resides at offset 0x0 in flash is the second stage bootloader. Second stage bootloader source code is available in components/bootloader directory of ESP-IDF. Second stage bootloader is used in ESP-IDF to add flexibility to flash layout (using partition tables), and allow for various flows associated with flash encryption, secure boot, and over-the-air updates (OTA) to take place.

When the first stage bootloader is finished checking and loading the second stage bootloader, it jumps to the second stage bootloader entry point found in the binary image header.

Second stage bootloader reads the partition table found by default at offset 0x8000 (configurable value). See partition tables documentation for more information. The bootloader finds factory and OTA app partitions. If OTA app partitions are found in the partition table, the bootloader consults the otadata partition to determine which one should be booted. See Over The Air Updates (OTA) for more information.

For a full description of the configuration options available for the ESP-IDF bootloader, see Bootloader.

For the selected partition, second stage bootloader reads the binary image from flash one segment at a time:

For segments with load addresses in internal IRAM (Instruction RAM) or DRAM (Data RAM), the contents are copied from flash to the load address.
For segments which have load addresses in DROM (data stored in Flash) or IROM (code executed from Flash) regions, the flash MMU is configured to provide the correct mapping from the flash to the load address.

Note that the second stage bootloader configures flash MMU for both PRO and APP CPUs, but it only enables flash MMU for PRO CPU. Reason for this is that second stage bootloader code is loaded into the memory region used by APP CPU cache. The duty of enabling cache for APP CPU is passed on to the application.

Once all segments are processed - meaning code is loaded and flash MMU is set up, second stage bootloader verifies the integrity of the application and then jumps to the application entry point found in the binary image header.

Application startup

Application startup covers everything that happens after the app starts executing and before the app_main function starts running inside the main task. This is split into three stages:

Port initialization of hardware and basic C runtime environment.
System initialization of software services and FreeRTOS.
Running the main task and calling app_main.

Note

Understanding all stages of ESP-IDF app initialization is often not necessary. To understand initialization from the application developer’s perspective only, skip forward to Running the main task.

Port Initialization

ESP-IDF application entry point is call_start_cpu0 function found in components/esp_system/port/cpu_start.c. This function is executed by the second stage bootloader, and never returns.

This port-layer initialization function initializes the basic C Runtime Environment (“CRT”) and performs initial configuration of the SoC’s internal hardware:

Reconfigure CPU exceptions for the app (allowing app interrupt handlers to run, and causing Fatal Errors to be handled using the options configured for the app rather than the simpler error handler provided by ROM).
If the option CONFIG_BOOTLOADER_WDT_ENABLE is not set then the RTC watchdog timer is disabled.
Initialize internal memory (data & bss).
Finish configuring the MMU cache.
Enable PSRAM if configured.
Set the CPU clocks to the frequencies configured for the project.
If the app is configured to run on multiple cores, start the other core and wait for it to initialize as well (inside the similar “port layer” initialization function call_start_cpu1).

Once call_start_cpu0 completes running, it calls the “system layer” initialization function start_cpu0 found in components/esp_system/startup.c. Other cores will also complete port-layer initialization and call start_other_cores found in the same file.

System Initialization

The main system initialization function is start_cpu0. By default, this function is weak-linked to the function start_cpu0_default. This means that it’s possible to override this function to add some additional initialization steps.

The primary system initialization stage includes:

Log information about this application (project name, App version, etc.) if default log level enables this.
Initialize the heap allocator (before this point all allocations must be static or on the stack).
Initialize newlib component syscalls and time functions.
Configure the brownout detector.
Setup libc stdin, stdout, and stderr according to the serial console configuration.
Perform any security-related checks, including burning efuses that should be burned for this configuration (including permanently limiting ROM download modes).
Initialize SPI flash API support.
Call global C++ constructors and any C functions marked with __attribute__((constructor)).

Secondary system initialization allows individual components to be initialized. If a component has an initialization function annotated with the ESP_SYSTEM_INIT_FN macro, it will be called as part of secondary initialization.

Running the main task

After all other components are initialized, the main task is created and the FreeRTOS scheduler starts running.

After doing some more initialization tasks (that require the scheduler to have started), the main task runs the application-provided function app_main in the firmware.

The main task that runs app_main has a fixed RTOS priority (one higher than the minimum) and a configurable stack size.

The main task core affinity is also configurable: CONFIG_ESP_MAIN_TASK_AFFINITY.

Unlike normal FreeRTOS tasks (or embedded C main functions), the app_main task is allowed to return. If this happens, The task is cleaned up and the system will continue running with other RTOS tasks scheduled normally. Therefore, it is possible to implement app_main as either a function that creates other application tasks and then returns, or as a main application task itself.

Second core startup

A similar but simpler startup process happens on the APP CPU:

When running system initialization, the code on PRO CPU sets the entry point for APP CPU, de-asserts APP CPU reset, and waits for a global flag to be set by the code running on APP CPU, indicating that it has started. Once this is done, APP CPU jumps to call_start_cpu1 function in components/esp_system/port/cpu_start.c.

While PRO CPU does initialization in start_cpu0 function, APP CPU runs start_cpu_other_cores function. Similar to start_cpu0, this function is weak-linked and defaults to the start_cpu_other_cores_default function but can be replaced with a different function by the application.

The start_cpu_other_cores_default function does some core-specific system initialization and then waits for the PRO CPU to start the FreeRTOS scheduler, at which point it executes esp_startup_start_app_other_cores which is another weak-linked function defaulting to esp_startup_start_app_other_cores_default.

By default esp_startup_start_app_other_cores_default does nothing but spin in a busy-waiting loop until the scheduler of the PRO CPU triggers an interrupt to start the RTOS scheduler on the APP CPU.

Provide feedback about this document