[Tutorial] Build, customize and use MediaTek open-source U-Boot and ATF

Banana Pi Router design
92

Hi all. I am trying to replicate OpenWRT’s UBI partition layout for the spim-nand on the BPI-R4. I am using the uboot and mtk-atf from https://github.com/frank-w/u-boot/tree/{2024-04-bpi,mtk-atf}, with some relevant patches from openwrt/package/boot/uboot-mediatek/patches at main · openwrt/openwrt · GitHub.

Based on OP’s instructions I tried:

make PLAT="mt7988" BOOT_DEVICE="spim-nand" BL33=../u-boot/u-boot.bin.xz USE_MKIMAGE= 1 DRAM_USE_COMB=1 UBI=1 all fip

Unfortunately though BL2 cannot find the fip image. Appreciate any help.

NAND boot:

F0: 102B 0000
FA: 1042 0000
FA: 1042 0000 [0200]
F9: 0000 0000
V0: 0000 0000 [0001]
00: 0000 0000
BP: 0600 0041 [0000]
G0: 1190 0000
EC: 0000 0000 [1000]
MK: 0000 0000 [0000]
T0: 0000 019D [0101]
Jump to BL

NOTICE:  BL2: v2.9(release):4e064f88b83 spim-nand
NOTICE:  BL2: Built : 17:50:18, Jun  6 2024
NOTICE:  WDT: Cold boot
NOTICE:  WDT: disabled
NOTICE:  CPU: MT7988
NOTICE:  EMI: Using DDR unknown settings
NOTICE:  EMI: Detected DRAM size: 4096 MB
NOTICE:  EMI: complex R/W mem test passed
NOTICE:  SPI_NAND parses attributes from parameter page.
NOTICE:  SPI_NAND Detected ID 0xef
NOTICE:  Page size 2048, Block size 131072, size 134217728
ERROR:   BL2: Failed to load image id 3 (-2)

MTD layout, SD boot:

spi-nand: spi_nand spi_nand@0: Winbond SPI NAND was found.
spi-nand: spi_nand spi_nand@0: 128 MiB, block size: 128 KiB, page size: 2048, OOB size: 64
List of MTD devices:
* spi-nand0
  - device: spi_nand@0
  - parent: spi@1100a000
  - driver: spi_nand
  - path: /spi@1100a000/spi_nand@0
  - type: NAND flash
  - block size: 0x20000 bytes
  - min I/O: 0x800 bytes
  - OOB size: 64 bytes
  - OOB available: 24 bytes
  - 0x000000000000-0x000008000000 : "spi-nand0"
          - 0x000000000000-0x000000200000 : "bl2"
          - 0x000000200000-0x000008000000 : "ubi"

UBI layout, SD boot:

BPI-R4>  ubi part ubi
ubi0: attaching mtd2
ubi0: scanning is finished
ubi0: attached mtd2 (name "ubi", size 126 MiB)
ubi0: PEB size: 131072 bytes (128 KiB), LEB size: 126976 bytes
ubi0: min./max. I/O unit sizes: 2048/2048, sub-page size 2048
ubi0: VID header offset: 2048 (aligned 2048), data offset: 4096
ubi0: good PEBs: 1008, bad PEBs: 0, corrupted PEBs: 0
ubi0: user volume: 3, internal volumes: 1, max. volumes count: 128
ubi0: max/mean erase counter: 2/1, WL threshold: 4096, image sequence number: 0
ubi0: available PEBs: 949, total reserved PEBs: 59, PEBs reserved for bad PEB handling: 20
BPI-R4> ubi list
0: fip
1: ubootenv
2: ubootenv2
93

This looks like TF-A bl2 is built without support for loading bl3 from UBI, despite you correctly passing it the UBI=1 build argument. Which TF-A tree are you building from? Excuse my ignorance, you wrote it clearly. And there is a clear answer: you have to use TF-A from https://github.com/mtk-openwrt/arm-trusted-firmware

94

My mtk-atf branch is also based from mtk atf repo,but older state where the ubi patches are not in as i have not used this before, so it uses fixed offset for fip. If you flash fip there it should boot uboot correctly

95

I have done some more testing and I can confirm that both mtk-atf repos ( frank-w’s and mtk-openwrt’s) work if UBI=0, by placing the fip at the default value of 0x580000.

@dangowrt The new repo from mtk-openwrt does not work for me when compiling with UBI=1. It seems to be building an invalid bl2 (see below). Do I need to change anything else to get a valid bl2 when compiling with UBI=1? Perhaps in the uBoot config, or the ATF make commandline?

F0: 102B 0000
FA: 1042 0000
FA: 1042 0000 [0200]
F9: 0000 0000
V0: 7027 6006 [0001]
00: 1017 0000
F9: 0000 0000
V0: 7027 6006 [0001]
01: 102A 0001
02: 1017 0000
BP: 0200 00C0 [0001]
EC: 0000 0000 [1000]
MK: 0000 0000 [0000]
T0: 0000 029E [0101]
System halt!
96

I don’t know what you are missing, but I can confirm that ATF bl2 built from mtk-openwrt tree definitely works fine, you can download the resulting bl2 here to give it a shot: https://downloads.openwrt.org/snapshots/targets/mediatek/filogic/openwrt-mediatek-filogic-bananapi_bpi-r4-snand-preloader.bin

97

Thanks hackpascal!

I was able to build latest version of U-boot with latest version for AT-F with patch for PCIe 3.0 from here: https://forum.banana-pi.org/t/bpi-r4-failed-attempt-add-nvme-support-on-uboot/ and I’m able to see the NVME in U-boot. Now I’m working to a script to automate the images build for different scenarios:

  • MMC/eMMC - Firmware + U-boot + Booting + Root FS on MMC/eMMC
  • MMC/eMMC + NVME - Firmware + U-boot on MMC/eMMC and Booting + Root FS on MMC/eMMC on NVME
  • FLASH + NVME - Firmware + U-boot on SPI Flash and Booting + Root FS on MMC/eMMC on NVME

But I have a question to all of you. The original SPI Flash U-Boot has these commands:

mtkautoboot- Display MediaTek bootmenu
mtkboardboot- Boot MTK firmware
mtkload   - MTK image loading utility
mtkupgrade- MTK firmware/bootloader upgrading utility

Where are these commands present? There is a custom U-boot that has these commands? I have no clue where I can add these commands. Anybody has any idea?

98

These are commands from sdk,not upstreamed in u-boot…why do you need these commands? There is a bootmenu,so you do not need mtk specific implementation.load is maybe combination of load+bootm/go. mtkboardboot sounds similar and upgrade will only work with sdk image…openwrt has different structure and has its own upgrade mechanism.

So why do you need these commands? If you really miss any feature from it it should be added to mainline commands in a generic way.

99

Thanks for clarifications.

I’ve checked the U-boot OpenWRT source and I found all these commands in boards/mediatek/common.

Anyway there are not portable to newer versions of U-boot mainly because ubi.c moved a lot of functions to static, functions used by one of the helpers.

As you says and also as I’ve noticed I don’t need them.

100

@hackpascal In the case of MT7981B/MT7986A, when there is no switch, the three storage media SPI-NOR, SPI-NAND, and SD have systems, and can the default access to the system be from the SD card? When there is no SD card or the SD card system is damaged, can the system be accessed from SPI-NOR or SPI-NAND?

101

Booting from SD is a fallback if booting from SPI-NAND fails. For other boot devices, you must change the bootstrap pins to the corresponding combinations.

102

Here’s updated procedures for flashing OpenWRT v23.05.0 & later on BPi R3:

103

Does mt7981b support nand booting the system, and then mounting emmc or SD card as storage space?

104

Yes, that should work. Since I can see the NAND when booting from eMMC, it should work vice versa.

105

does anyone know How to configure mt7981, uboot, openwrt, dts, default spi nand to boot into the system, and mount SD card as an external simple storage device? Thank you

106

Why asking again? This thread is only uboot+atf,no kernel/userspace.

When you want openwrt please ask there or in a openwrtone thread (maybe create one).

107

Ok, sorry, I’ll try to ask on openwrt one, thank you

108

Hi @hackpascal, it looks like current atf code misses some files (wdt,mmc,spi,efuse)

Or has build been changed? I see menuconfig support compared to my old tree.

after i got the missing efuse error also on my default atf-branch i had to delete ./build/mt7988/release/ manually (make clean does not clean it up) to be able to compile my old branch again, but i got a strange error

  LD      build/mt7988/release/bl2/bl2.elf
/bin/sh: 1: -g: not found

when compiling the new tree, it also seems like the MAKE_BUILD_STRINGS was dropped from make_helpers/build_macros.mk which i had used to print the bootdevice in bl2/fip

looks like the error above was caused by my comit readding the MAKE_BUILD_STRINGS, to where was this moved?

uploaded my full tree to GitHub - frank-w/u-boot at mtk-atf-2025

finally got the 8GB variant working with the new atf-code…

NOTICE:  BL2: v2.12.0(release):v2.11.0-1599-gc6dad9faceb5
NOTICE:  BL2: Built : 23:34:17, Apr  7 2025
NOTICE:  WDT: [40000000] Software reset (reboot)
NOTICE:  CPU: MT7988
NOTICE:  EMI: DDR4 4BG mode
NOTICE:  EMI: Using DDR unknown settings
NOTICE:  EMI: Detected DRAM size: 8192 MB
NOTICE:  EMI: complex R/W mem test passed
NOTICE:  LVTS: Enable thermal HW reset
NOTICE:  BL2: Booting BL31
NOTICE:  BL31: v2.12.0(release):v2.11.0-1599-gc6dad9faceb5
NOTICE:  BL31: Built : 23:34:20, Apr  7 2025

only can’t get the build-version string working

linux also sees full 8G of ram

root@bpi-r4-8G:~# free
               total        used        free      shared  buff/cache   available
Mem:         8140504      202848     7922844        1592       89244     7937656
Swap:              0           0           0

but noticed that powerdown is not supported by atf-code (was working with old atf)

[   82.829046] reboot: Power down
ERROR:   Power-down unsupported.
PA�
F0: 102B 0000
109

power down doesn’t really supported (was implemented as WFI). Only soft reset is supported.

110

But my older tree does not print this eror and does not reboot…then atf imgo should hang if it cannot powerdown the pmic.

111

If I remember correctly, I was also struggling, it has to do with the tags in the git repo. Also pull the tags when cloning.