blob: 6532878932c0c73dccf012318f173c9707df1c3f [file] [log] [blame]
.. SPDX-License-Identifier: GPL-2.0-or-later
.. sectionauthor:: Igor Opaniuk <igor.opaniuk@toradex.com>
Colibri iMX7 Modules
====================
- SoM: https://www.toradex.com/computer-on-modules/colibri-arm-family/nxp-freescale-imx7
- Carrier board: https://www.toradex.com/products/carrier-board/colibri-evaluation-board
Quick Start
-----------
- Build U-Boot
- NAND IMX image adjustments before flashing
- Flashing manually U-Boot to eMMC
- Flashing manually U-Boot to NAND
- Using ``update_uboot`` script
Build U-Boot
------------
.. code-block:: bash
$ export CROSS_COMPILE=arm-linux-gnueabi-
$ make colibri_imx7_emmc_defconfig # For NAND: colibri_imx7_defconfig
$ make
After the build succeeds, you will obtain the final ``u-boot-dtb.imx`` IMX
specific image, ready for flashing (but check next section for additional
adjustments).
The final IMX program image includes (section ``6.6.7`` from `IMX7DRM
<https://www.nxp.com/webapp/Download?colCode=IMX7DRM>`_):
* **Image vector table** (IVT) for BootROM
* **Boot data** -indicates the program image location, program image size
in bytes, and the plugin flag.
* **Device configuration data**
* **User image**: U-Boot image (``u-boot-dtb.bin``)
IMX image adjustments prior to flashing
---------------------------------------
1. U-Boot for both Colibri iMX7 NAND and eMMC versions
is built with HABv4 support (`AN4581.pdf
<https://www.nxp.com/docs/en/application-note/AN4581.pdf>`_)
enabled by default, which requires generating a proper
Command Sequence File (CSF) by srktool from NXP (not included in the
U-Boot tree, check additional details in introduction_habv4.txt)
and concatenate it to the final ``u-boot-dtb.imx``.
2. In case you don't want to generate a proper ``CSF`` (for any reason),
you still need to pad the IMX image so it has the same size as specified in
the **Boot Data** section of the IMX image.
To obtain this value, run:
.. code-block:: bash
$ od -X -N 0x30 u-boot-dtb.imx
0000000 402000d1 87800000 00000000 877ff42c
0000020 877ff420 877ff400 878a5000 00000000
^^^^^^^^
0000040 877ff000 000a8060 00000000 40b401d2
^^^^^^^^ ^^^^^^^^
Where:
* ``877ff400`` - IVT self address
* ``877ff000`` - Program image address
* ``000a8060`` - Program image size
To calculate the padding:
* IVT offset = ``0x877ff400`` - ``0x877ff000`` = ``0x400``
* Program image size = ``0xa8060`` - ``0x400`` = ``0xa7c60``
and then pad the image:
.. code-block:: bash
$ objcopy -I binary -O binary --pad-to 0xa7c60 --gap-fill=0x00 \
u-boot-dtb.imx u-boot-dtb.imx.zero-padded
3. Also, according to the requirement from ``6.6.7.1``, the final image
should have ``0x400`` offset for the initial IVT table.
For eMMC setup we handle this by flashing it to ``0x400``, however
for NAND setup we adjust the image prior to flashing, adding padding at the
beginning of the image.
.. code-block:: bash
$ dd if=u-boot-dtb.imx.zero-padded of=u-boot-dtb.imx.ready bs=1024 seek=1
Flash U-Boot IMX image to eMMC
------------------------------
Flash the ``u-boot-dtb.imx.zero-padded`` binary to the primary eMMC hardware
boot area partition:
.. code-block:: bash
=> load mmc 1:1 $loadaddr u-boot-dtb.imx.zero-padded
=> setexpr blkcnt ${filesize} + 0x1ff && setexpr blkcnt ${blkcnt} / 0x200
=> mmc dev 0 1
=> mmc write ${loadaddr} 0x2 ${blkcnt}
Flash U-Boot IMX image to NAND
------------------------------
.. code-block:: bash
=> load mmc 1:1 $loadaddr u-boot-dtb.imx.ready
=> nand erase.part u-boot1
=> nand write ${loadaddr} u-boot1 ${filesize}
=> nand erase.part u-boot2
=> nand write ${loadaddr} u-boot2 ${filesize}
Using update_uboot script
-------------------------
You can also use U-Boot env update_uboot script,
which wraps all eMMC/NAND specific command invocations:
.. code-block:: bash
=> load mmc 1:1 $loadaddr u-boot-dtb.imx.ready
=> run update_uboot