https://github.com/jameswalmsley/imx6-spi-sd-boot-trampoline

A pico bootloader for imx6 to reset from SPI mode into booting from an SD card.
https://github.com/jameswalmsley/imx6-spi-sd-boot-trampoline

Last synced: about 1 month ago
JSON representation

A pico bootloader for imx6 to reset from SPI mode into booting from an SD card.

• Host: GitHub
• URL: https://github.com/jameswalmsley/imx6-spi-sd-boot-trampoline
• Owner: jameswalmsley
• Created: 2015-08-07T14:02:59.000Z (almost 10 years ago)
• Default Branch: master
• Last Pushed: 2015-08-07T14:09:10.000Z (almost 10 years ago)
• Last Synced: 2025-04-02T04:23:26.376Z (about 2 months ago)
• Language: Assembly
• Size: 97.7 KB
• Stars: 8
• Watchers: 2
• Forks: 2
• Open Issues: 0
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

        
# IMX6 Boot Trampoline

The code in boot.S is a tiny nano boot-loader that causes the iMX.6

processor to reset itself in SD-Card boot mode, so that the main 

boot-loader can be loaded from the another SD-card / uSD card.

## Boot Process overview.

Many imx6 boards are shipped with an spi flash pre-programmed with

something like u-boot.

In order to make the imx6 boot from the spi flash on POR (power-on-reset)

the manufacturer as blown some fuses (the eFUSEs - see HRM (hardware ref manual).

This means that the device always boots from SPI flash, and this can be

inconvenient during development, and even production.

This simple boot-loader can be stored into the SPI flash to cause the imx6 to reboot

but load the main boot-loader from an SD-Card instead.

To do this, it sets the registers at:

    0x020d8040 <= (SD 0x3840) (uSD 0x3040)                ; // The bit-pattern seems to be undocumented. (Typical Sabrelite board)

    0x020d8040 <= (SD 0x3850) (uSD 0x2850) (eMMC 0x5262)  ; // (Congatex QMX-6 board).

    0x020d8044 <= 0x10000000                              ; // Signals a persistent boot.

It then causes a system warm reset, i.e. not POR, which causes the boot rom

to read the values in these registers as arguments to the boot flow.

So far I didn't manage to find any real documentation on the meaning of these

bits. Maybe I should read more carefully.

## Building the boot image

    make

## Flash the SPI flash using u-boot:

Copy the boot image onto an sd-card or other medium usable by u-boot:

    dd if=boot.bin of=/dev/sdc

Note this operation overwrites the MBR of the sd-card, ensure you backup anything important.

Now insert the SD-Card into a slot.

    # Replace 0 with the correct ID for your sd-slot.

    mmc dev 0

    mmc read 0x10800000 0 200

Flash the spi:

    sf probe 1

    # Note the next command erases any current boot-loader.

    sf erase 0 0x40000

    sf write 0x10800000 0 0x40000

## Make a bootable SD-Card.

You should find out if your u-boot image is padded:

    arm-none-eabi-objdump -b binary -D u-boot.bin -marm | more

You'll see output like:

    u-boot.bin:     file format binary

    

    

    Disassembly of section .data:

    

    00000000 <.data>:

           0:       ea0001be        b       0x700

            ...

         400:       402000d1        ldrdmi  r0, [r0], -r1   ; 

         404:       27800700        strcs   r0, [r0, r0, lsl #14]

         408:       00000000        andeq   r0, r0, r0

         40c:       2780042c        strcs   r0, [r0, ip, lsr #8]

         410:       27800420        strcs   r0, [r0, r0, lsr #8]

         414:       27800400        strcs   r0, [r0, r0, lsl #8]

    

If the value 0x402000d1 (as seen at address 0x400) is at 0x0 then your image is not padded.

    dd if=u-boot.bin of=/dev/sdc bs=512 skip=2 seek=2     # Padded image

    dd if=u-boot.bin of=/dev/sdc bs=512 skip=0 seek=2     # Non-padded image