make_airoha_atf

A standalone project that re-packages BL2 (and compresses BL31) for Airoha SoCs whose vendor BL2 is delivered as separate bl21 / bl22 / bl23 fragments. The project bundles the SPI-NAND flash table generator, the required host tools (lzma, trx), and a thin Makefile pipeline.

When to use this project

Only SoCs whose unopen_img/<ic>/ directory ships the split layout

bl21.bin
bl22.lzma
bl23.lzma
bl31.bin

need this project. Platforms that already provide a finished bl2.bin (such as AN7581, AN7583, EN7523 in the upstream tree) are pre-assembled and can be flashed as-is — running this project on them would be redundant.

The current reference platform that does require packaging is an7563.

Repository layout

make_airoha_atf/
├── Makefile               # Top-level pipeline (bl2 + bl31).
├── config.mk              # Cross-platform defaults and TCSUPPORT_DEFS expansion.
├── configs/               # One file per SoC, holding only its IC overrides.
│   └── an7563.mk
├── .gitignore
├── README.md
├── src/
│   ├── flash_table/
│   │   ├── Makefile                  # Builds the flash_table_gen host tool.
│   │   └── spi_nand_flash_table.c    # SPI-NAND table source.
│   └── include/
│       └── spi/
│           ├── spi_nand_flash.h
│           └── nand_flash_otp.h
├── tools/
│   ├── lzma                          # Prebuilt host binary.
│   └── trx/
│       ├── Makefile                  # Optional rebuild via `make trx_build`.
│       ├── trx                       # Prebuilt host binary.
│       ├── trx.c
│       ├── trx.h
│       └── trx_config
├── unopen_img/
│   └── an7563/                       # Inputs: bl1.bin, bl21.bin, bl22.lzma,
│                                     #         bl23.lzma, bl31.bin
└── build/                            # Generated outputs (git-ignored).
    └── <PLATFORM>/
        ├── bl2.bin
        └── bl31.lzma

Toolchain

• flash_table_gen is a host tool: it runs on the build machine to emit flash_table.bin, so it always uses HOSTCC (defaults to gcc).

• ARM 32-bit cross-compile variables are exposed for any future ARM target that may be added later:

  CROSS_COMPILE       ?= arm-linux-gnueabihf-
  ARM_ARCH            ?= aarch32
  ARM_INSTRUCTION_SET ?= A32

Quick start

# Default: build bl2.bin and bl31.lzma for PLATFORM=an7563.
make

# Choose a platform explicitly.
make PLATFORM=an7563

# Build a single artifact.
make bl2
make bl31

# Verbose mode prints every command.
make V=1

# Inspect the resolved configuration.
make help

# Cleanup.
make clean         # removes build/ and host tool object files.
make distclean     # also cleans rebuilt trx intermediates.

Outputs land in:

build/<PLATFORM>/bl2.bin
build/<PLATFORM>/bl31.lzma

Pipeline overview

Stage	Command equivalent	Output
1	Build flash_table_gen (host tool)	src/flash_table/flash_table_gen
2	Run flash_table_gen inside build/<PLATFORM>/	flash_table.bin (intermediate)
3	lzma e flash_table.bin flash_table.lzma	flash_table.lzma (intermediate)
4	trx -z bl21.bin bl22.lzma bl23.lzma flash_table.lzma bl2.bin	build/<PLATFORM>/bl2.bin
5	cp bl2.bin bl2.tmp + trx -x bl2.tmp bl2_crc.bin + mv bl2_crc.bin bl2.bin	build/<PLATFORM>/bl2.bin (CRC)
6	lzma e bl31.bin bl31.lzma	build/<PLATFORM>/bl31.lzma

Configuration model

Layered resolution

top Makefile
└── include config.mk
    ├── PLATFORM defaults to an7563 (override on command line if needed)
    ├── -include configs/$(PLATFORM).mk    # IC-specific overrides (":=")
    └── default values for every TCSUPPORT_* / toolchain variable ("?=")

configs/<PLATFORM>.mk uses immediate assignments (:=), so the ?= defaults that follow in config.mk cannot overwrite them. Anything not set by the IC config simply inherits the default.

IC config = intersection of profile and config.mk

A configs/<ic>.mk file lists only the flags that are present in both:

1. The IC's vendor profile (e.g. UNION_<...>.mak).
2. config.mk's TCSUPPORT_FLAGS master list.

Any flag missing from either side stays out of the IC config. As a worked example, configs/an7563.mk contains four entries because those are the only flags that the AN7563 profile and config.mk agree on:

config.mk flag	In AN7563 profile?	configs/an7563.mk
TCSUPPORT_BL2_OPTIMIZATION	yes	:= 1
TCSUPPORT_CPU_ARMV8	yes	:= 1
TCSUPPORT_OPENWRT	yes	:= 1
TCSUPPORT_BOOTROM_LARGE_SIZE	yes	:= 1
TCSUPPORT_NAND_FLASH_OTP	no	(omitted)
TCSUPPORT_AUTOBENCH	no	(omitted)
TCSUPPORT_PARALLEL_NAND	no	(omitted)
TCSUPPORT_SECURE_BOOT_FLASH_OTP	no	(omitted)
CONFIG_ECNT_UBOOT	no	(omitted)
TCSUPPORT_LITTLE_ENDIAN	no	(omitted, default 1)
TCSUPPORT_BB_256KB	no	(omitted)
BOOTROM_LARGE_SIZE	no	(omitted)
TCSUPPORT_CMCCV2	no	(omitted)
TCSUPPORT_SECURE_BOOT	no	(omitted)
TCSUPPORT_SECURE_BOOT_V1	no	(omitted)
TCSUPPORT_SECURE_BOOT_V2	no	(omitted)

Adding a new IC

1. Stage the inputs under unopen_img/<ic>/ (must contain at least bl21.bin, bl22.lzma, bl23.lzma, bl31.bin).
2. Locate the IC's vendor profile.
3. Create configs/<ic>.mk containing only the intersection described above; remember to use :=.
4. Run make PLATFORM=<ic>.

TCSUPPORT_* reference

flash_table host tool

Flag	Default	Effect
TCSUPPORT_BL2_OPTIMIZATION	1	Selects the size-optimised BL2 flash table layout.
TCSUPPORT_NAND_FLASH_OTP		Includes OTP page metadata in each table entry.
TCSUPPORT_AUTOBENCH		Excludes NAND entries that are not relevant for autobench builds.
TCSUPPORT_PARALLEL_NAND		Enables parallel NAND support.
TCSUPPORT_SECURE_BOOT_FLASH_OTP		Enables the NAND OTP RSA public-key interface.
TCSUPPORT_CPU_ARMV8		Targets the ARMv8 SRAM address map.
CONFIG_ECNT_UBOOT		U-Boot build context.

FLASH_TABLE_OPEN is always injected by the sub-Makefile and is required for the host build.

trx host tool

Flag	Default	Effect
TCSUPPORT_OPENWRT	1	Uses the OpenWrt build flow.
TCSUPPORT_LITTLE_ENDIAN	1	Generates little-endian output for ARM 32-bit.
TCSUPPORT_BB_256KB		Uses a 256 KB bootbase layout.
BOOTROM_LARGE_SIZE		Uses the 128 KB bootbase layout.
TCSUPPORT_CMCCV2		CMCC v2 customisation.
TCSUPPORT_SECURE_BOOT		Enables secure boot.
TCSUPPORT_SECURE_BOOT_V1 / V2		Selects the secure boot sub-version.
TCSUPPORT_BOOTROM_LARGE_SIZE		Enables the larger bootrom configuration.

TCSUPPORT_DEFS automatically expands every flag set to 1 into -D<flag>=1; sub-Makefiles can drop $(TCSUPPORT_DEFS) straight into their CFLAGS.

.gitignore

Ignored:

• build/, out/ — every generated artifact (bl2.bin, bl31.lzma, intermediates).
• src/flash_table/flash_table_gen and its *.o objects.
• tools/trx/*.o, *.a from optional rebuilds.
• Editor / IDE noise (.vscode/, .idea/, *.swp, *~, .DS_Store, __pycache__/).

Tracked:

• All sources (src/).
• All configuration (config.mk, configs/).
• Prebuilt host tools (tools/lzma, tools/trx/trx).
• Per-platform inputs (unopen_img/<ic>/).

Notes and limitations

• The BL2 packaging path is only meaningful for SoCs that ship the split bl21 / bl22 / bl23 layout. The check-bl2-inputs target stops the build early if any required input is missing.
• The BL31 path simply lzma-compresses unopen_img/<ic>/bl31.bin; nothing else is required.
• flash_table_gen writes its output in the current working directory, so the Makefile cd's into build/<PLATFORM>/ before invoking it to keep the output co-located with the rest of the build artifacts.
• The shipped lzma and trx binaries are 64-bit Linux ELF executables. If they are not compatible with the build host, rebuild trx via make trx_build (a separate lzma source must be obtained for lzma).