f34fa1c269
The bootloader does not use the heap allocator or TLSF, so the ROM patch files esp_rom_tlsf.c and esp_rom_multi_heap.c are not needed in the bootloader build. In the cmakev1 build system this was never an issue because cmakev1 uses an early expansion phase where only Kconfig files for components listed in BUILD_COMPONENTS are processed. Since the heap component is not part of the bootloader's component list, its Kconfig options (CONFIG_HEAP_TLSF_USE_ROM_IMPL, etc.) were never defined and the conditional compilation of these source files was effectively skipped. In the cmakev2 build system, Kconfig options from all discovered components are visible regardless of whether the component is part of the build. Because the bootloader reuses the main project's sdkconfig (where CONFIG_HEAP_TLSF_USE_ROM_IMPL defaults to y on targets with ROM TLSF support like esp32c2), the TLSF patch sources were being compiled into the bootloader's esp_rom. This caused a build failure because esp_rom_tlsf.c includes tlsf_block_functions.h from the heap component, which is not a dependency of esp_rom and is not part of the bootloader build. Guard the TLSF and multi_heap ROM patch sources with NOT BOOTLOADER_BUILD to prevent them from being compiled in the bootloader context. This fix is compatible with both cmakev1 and cmakev2. Signed-off-by: Frantisek Hrbata <frantisek.hrbata@espressif.com>
esp_rom Component
Function Description
esp_rom component contains each chip's ROM functions, which are used in the ROM bootloader, 2nd bootloader, esp_tool flash stub and some driver code (e.g. GPIO matrix). ROM functions as not treated as public APIs, attentions are required when you use ROM functions:
- ROM functions are not thread-safe in RTOS, extra locks are needed to be around the ROM functions.
- Names/signatures/behaviors of ROM function may be different between chips.
- ROM functions are not guaranteed to exist across all chips.
When using ROM functions in esp-idf, the including convention is <target>/rom/<header_file>.h. This can prevent you from using a nonexistent ROM function for a specific <target>. Thus ROM functions are recommended for use in a target-specific source file. For example, bootloader_esp32.c can include esp32/rom/<header_file>.h without any violations. However, this is not the case when it comes to a common source file that also wants to use some of the ROM functions. The include list would be quite extensive:
#if CONFIG_IDF_TARGET_ESP32
#include "esp32/rom/uart.h"
#elif CONFIG_IDF_TARGET_ESP32C3
#include "esp32c3/rom/uart.h"
#elif CONFIG_IDF_TARGET_ESP32S3
#include "esp32s3/rom/uart.h"
...
So, we added a wrapper for those commonly used ROM functions. They're declared in esp_rom/include/esp_rom_xxx.h. Unlike the original ROM functions, these extracted ones are expected to exist across all chips. If some of them are missed in the new chips, we will implement them again in esp_rom/patches. These ROM APIs are always prefixed with the name esp_rom (e.g. esp_rom_printf), so that it's obvious to know whether a function is linked to ROM.
Most of the time, the ROM wrapper APIs are just alias to the original ROM functions by linker script esp_rom/<target>/ld/<target>.rom.api.ld. For example, esp_rom_printf is alias to ets_printf in the following way:
PROVIDE ( esp_rom_printf = ets_printf );
If some original ROM functions have changed the behavior or have bugs, we should override them in the wrapper layer. A common example is the esp_rom_install_uart_printf(), on ESP32 and ESP32S2, it's just alias to ets_install_uart_printf, but on other chips, it's re-implemented in the esp_rom/patches/esp_rom_uart.c. To some extent, the ROM wrapper layer works like an anti-corrosion layer between esp-rom project and esp-idf project.
As ROM functions are unique to each target, features are as well. For example, ESP32 has the tjpgd library built into the ROM, but ESP32S2 hasn't. We have a header file esp_rom/<target>/esp_rom_caps.h declaring the features that are supported by each target. Based on the macros defined there, we can decide whether a function should be patched or whether a feature should be re-implemented.
Directory Structure
.
├── CMakeLists.txt
├── <target/chip_name>
│ ├── esp_rom_caps.h
│ └── ld
│ ├── <target>.rom.api.ld
│ ├── <target>.rom.ld
│ ├── <target>.rom.libgcc.ld
│ ├── <target>.rom.newlib.ld
│ ├── <target>.rom.newlib-nano.ld
│ ├── <target>.rom.version.ld
│ └── ... // other ROM linker scripts, added when bring up new chip
├── include
│ ├── <target/chip_name>
│ │ └── rom
│ │ ├── cache.h
│ │ ├── efuse.h
│ │ ├── esp_flash.h
│ │ ├── ets_sys.h
│ │ ├── gpio.h
│ │ ├── uart.h
│ │ └── ... // other original ROM header files, added when bring up new chip
│ ├── esp_rom_gpio.h
│ ├── esp_rom_md5.h
│ ├── esp_rom_sys.h
│ ├── esp_rom_serial_output.h
│ └── ... // other ROM wrapper api files
├── Kconfig.projbuild
├── linker.lf
├── patches
│ ├── esp_rom_sys.c
│ ├── esp_rom_uart.c
│ └── ... // other patched source files
├── README.md
└── test
├── CMakeLists.txt
├── test_miniz.c
└── ... // other ROM function unit tests
Adding New Chip Support For Debugging
- Dump .rodata strings via
riscv32-esp-elf-readelf -p .rodata esp32c5_rev0_rom.elf - Locate the build date in the output (e.g.
[ 23c4] Mar 29 2024). - Get start address of .rodata section via
riscv32-esp-elf-readelf -S esp32c5_rev0_rom.elf | grep rodata - Compute the absolute address by adding the offset from step 2 to the base address from step 3.
- Verify with GDB:
riscv32-esp-elf-gdb esp32c5_rev0_rom.elf -ex "p (char*) 0x4004b3c4" --batch $1 = 0x4004b3c4 "Mar 29 2024" - Update roms.json:
"esp32c5": [ { "rev": 0, "build_date_str_addr": "0x4004b3c4", "build_date_str": "Mar 29 2024" } ],