How to manually update bootloaders

This message will disappear after all relevant tasks have been resolved.

Semantic MediaWiki

There are 1 incomplete or pending task to finish installation of Semantic MediaWiki. An administrator or user with sufficient rights can complete it. This should be done before adding new data to avoid inconsistencies.

Applicable for

STM32MP13x lines, STM32MP15x lines

This page explains how to manually update the bootloaders binary (including U-Boot and TF-A) on an SD card or on the e•MMC.

1 To update an SD card with the Linux dd command[edit source]

When a Linux console has access to the SD card device partitions:

on a Linux PC
- with a card reader of this PC
- through a USB connection to the target and the ums command executed on a U-Boot console
on target, with Linux console.

The 3 first GPT partitions on the SD card are:

FSBL1
FSBL2
FIP

See Boot_chain_overview for the bootloader definitions.

You can use the Linux dd command to copy the FSBL and SSBL directly to the correct partition:

   dd if=<file> of=/dev/<dev> conv=fdatasync

<dev> is:

mmcblk<X>p<n>: PC-embedded card reader case or target Linux console
sd<X><n>: USB-connected SD card reader case

where <X> is the ID of the device, and <n> the ID of the partition.

Note: the dd option conv=fdatasync is used to force synchronous copying.

1.1 U-Boot environment[edit source]

The U-Boot environment is saved at the end of the U-Boot partition, named "fip": ID = 3.

To clear this environment, erase the U-Boot partition before any update; for example, by writing 0 to this partition:

   dd if=/dev/zero of=/dev/mmcblk<X>p3 conv=fdatasync

   dd if=/dev/zero of=sd<X>3 conv=fdatasync

1.2 SD card update example[edit source]

The internal card reader is /dev/mmcblk0 or for a target Linux console, GPT partition <n> is /dev/mmcblk0p<n>:

  dd if=tf-a-sdcard.stm32 of=/dev/mmcblk0p1 conv=fdatasync
  dd if=tf-a-sdcard.stm32 of=/dev/mmcblk0p2 conv=fdatasync
  dd if=/dev/zero of=/dev/mmcblk0p3 conv=fdatasync
  dd if=fip.bin of=/dev/mmcblk0p3 conv=fdatasync

Alternatively, with U-Boot console, dev = 0 (SD card device on ST Microelectronics board), GPT partition <n> is /dev/sda<n>:

  mmc dev 0 
  ums 0 mmc 0
 
  dd if=tf-a-sdcard.stm32 of=/dev/sda1 conv=fdatasync
  dd if=tf-a-sdcard.stm32 of=/dev/sda2 conv=fdatasync
  dd if=/dev/zero of=/dev/sda3 conv=fdatasync
  dd if=fip.bin of=/dev/sda3 conv=fdatasync

1.3 SD card update example with SPL as FSBL[edit source]

The USB card reader is /dev/sdb, GPT partition <n> is /dev/sdb<n>:

  dd if=u-boot-spl.stm32 of=/dev/sdb1 conv=fdatasync
  dd if=u-boot-spl.stm32 of=/dev/sdb2 conv=fdatasync
  dd if=/dev/zero of=/dev/sdb3 conv=fdatasync
  dd if=u-boot.img of=/dev/sdb3 conv=fdatasync

2 Update of e•MMC with the Linux dd command[edit source]

The same command, dd, can be used to update e•MMC memory mapping:

SSBL U-Boot is the first GPT partition in the e•MMC user area
FSBL = TF-A (or SPL) is saved at the beginning of the e•MMC boot partition

The user needs to select the e•MMC hardware partition to update: user data, boot1, or boot2.

2.1 On a Linux console[edit source]

If dev = mmcblk1 for e•MMC device (default on ST Microelectronics board)

The boot partitions are available in /dev/mmcblk1boot0 and /dev/mmcblk1boot1 ^[1].

The user perhaps needs to allow access, for example with:

   echo 0 > /sys/class/block/mmcblk1boot0/force_ro

The mmc tools allow the boot partition to be selected ^[2].

The ROM code requires:

<send_ack> =1
the eMMC boot configuration is: 1 wire configuration and 25 MHz, it is done with the command:

   mmc bootbus set  single_backward x1 x1 dev/mmcblk1

To update TF-A in boot1 and select this boot partition:

  dd if=tf-a-emmc.stm32 of=/dev/mmcblk1boot0 conv=fdatasync
  mmc bootpart enable 1 1 /dev/mmcblk1

To update TF-A in boot2 and select this boot partition:

  dd if=tf-a-emmc.stm32 of=/dev/mmcblk1boot1 conv=fdatasync
  mmc bootpart enable 2 1 /dev/mmcblk1

To update U-Boot in the first GPT partition:

  dd if=/dev/zero of=/dev/mmcblk1p1 conv=fdatasync
  dd if=fip.bin of=/dev/mmcblk1p1 conv=fdatasync

2.2 On a U-Boot console[edit source]

Before the first boot, select the e•MMC correct boot configuration (1 wire, 25 MHz) with the command:

  mmc bootbus 1 0 0 0

2.2.1 mmc partconf[edit source]

The targeted e•MMC HW partition is selected in U-Boot by the command mmc partconf:

   help mmc 
 ...
 mmc bootbus dev boot_bus_width reset_boot_bus_width boot_mode
- Set the BOOT_BUS_WIDTH field of the specified device
 mmc bootpart-resize <dev> <boot part size MB> <RPMB part size MB>
- Change sizes of boot and RPMB partitions of specified device
 mmc partconf dev [boot_ack boot_partition partition_access]
- Show or change the bits of the PARTITION_CONFIG field of the specified device

For example:

* dev = 1 (e•MMC device on ST Microelectronics board)
* boot_ack=1 (Boot Acknowledge is needed by ROM code) 
* boot_partition = 1 (Boot partition 1 enabled for boot)
* partition_access =  0 (user data partition - default)

The command to select the boot partition used by ROM code on e•MMC at next reboot is:

for boot1:

  mmc partconf 1 1 1 0

for boot2:

  mmc partconf 1 1 2 0

0: user data partition (default)
1: boot partition 1
2: boot partition 2

Don't change the last parameter partition_access value if you don't manually access to the boot partition content with mmc read or mmc write:

2.2.2 ums[edit source]

The e•MMC update is done with the ums command, with the U-Boot's partition syntax^[4]:

 <command> <interface> [devnum][.hwpartnum][:partnum|#partname]

The exported e•MMC is mounted on the Linux PC as a block device named sd<X> and the GPT partitions are exported by sd<X><n> where <X> is the ID of the device starting at a and <n> the ID of the partition. In the next example we assume <X> = a, so the block device e•MMC available in /dev/sda but this value need to be verified on your PC. boot partitions To update FSBL=TF-A in the boot1 HW partition.

  ums 0 mmc 1.1
 UMS: LUN 0, dev 1, hwpart 1, sector 0x0, count 0x1000

On boot partitions, without GPT partitions, just copy TF-A binary on the block device /dev/sda exported by U-Boot.

  dd if=tf-a-emmc.stm32 of=/dev/sda  conv=fdatasync

For boot2, change the mmc device to 1.2. user data partition To export all the e•MMC in /dev/sda, including all GPT partitions in /dev/sdaN:

  ums 0 mmc 1

  dd if=fip.bin of=/dev/sda1 conv=fdatasync

And copy the FIP binary in the first partition of this block device:

You can also export only the first GPT partition ("fip" partition index is 1) in /dev/sda:

  ums 0 mmc 1:1 
  dd if=fip.bin of=/dev/sda conv=fdatasync

3 References[edit source]

Please refer to the following links for additional information:

[1] ttps://www.kernel.org/doc/Documentation/mmc/mmc-dev-parts.txt

[2] ttps://manpages.debian.org/buster/mmc-utils/mmc.1.en.html

[3] ttps://www.emcraft.com/som/stm32mp1/booting-linux-from-emmc

[4] ttps://u-boot.readthedocs.io/en/latest/usage/partitions.html

[1]

[2]

[3]

[4]