85 lines
3.2 KiB
Plaintext
85 lines
3.2 KiB
Plaintext
Compilation of TF-A (Trusted Firmware-A):
|
|
1. Pre-requisite
|
|
2. Initialise cross-compilation via SDK
|
|
3. Prepare tf-a source code
|
|
4. Management of tf-a source code
|
|
5. Compile tf-a source code
|
|
6. Update software on board
|
|
|
|
1. Pre-requisite:
|
|
-----------------
|
|
OpenSTLinux SDK must be installed.
|
|
|
|
For tf-a build you need to install:
|
|
- git:
|
|
Ubuntu: sudo apt-get install git-core gitk
|
|
Fedora: sudo yum install git
|
|
|
|
If you have never configured you git configuration:
|
|
$ git config --global user.name "your_name"
|
|
$ git config --global user.email "your_email@example.com"
|
|
|
|
2. Initialise cross-compilation via SDK:
|
|
---------------------------------------
|
|
Source SDK environment:
|
|
$ source <path to SDK>/environment-setup-cortexa7t2hf-neon-vfpv4-ostl-linux-gnueabi
|
|
|
|
To verify if your cross-compilation environment have put in place:
|
|
$ set | grep CROSS
|
|
CROSS_COMPILE=arm-ostl-linux-gnueabi-
|
|
|
|
Warning: the environment are valid only on the shell session where you have
|
|
sourced the sdk environment.
|
|
|
|
3. Prepare tf-a source:
|
|
------------------------
|
|
If you have the tarball and the list of patch then you must extract the tarball
|
|
and apply the patch.
|
|
$> tar xfz ##BP##-##PR##.tar.gz
|
|
A new directory containing tf-a standard source code will be created, go into it:
|
|
$> cd ##BP##
|
|
|
|
NB: if there is no git management on source code and you would like to have a
|
|
git management on the code see section 4 [Management of tf-a source code]
|
|
if there is some patch, please apply it on source code
|
|
$> for p in `ls -1 ../*.patch`; do patch -p1 < $p; done
|
|
|
|
4. Management of tf-a source code:
|
|
-----------------------------------
|
|
If you like to have a better management of change made on tf-a source, you
|
|
can use git:
|
|
$ cd <directory to tf-a source code>
|
|
$ test -d .git || git init . && git add . && git commit -m "tf-a source code" && git gc
|
|
$ git checkout -b WORKING
|
|
$ for p in `ls -1 <path to patch>/*.patch`; do git am $p; done
|
|
|
|
NB: you can use directly the source from the community:
|
|
URL: git://github.com/ARM-software/arm-trusted-firmware.git
|
|
Branch: ##GIT_BRANCH##
|
|
Revision: ##GIT_SRCREV##
|
|
|
|
$ git clone git://github.com/ARM-software/arm-trusted-firmware.git -b ##GIT_BRANCH##
|
|
$ cd <directory to tf-a source code>
|
|
$ git checkout -b WORKING ##GIT_SRCREV##
|
|
$ for p in `ls -1 <path to patch>/*.patch`; do git am $p; done
|
|
|
|
5. Build tf-a source code:
|
|
--------------------------------
|
|
Since OpenSTLinux has activated FIP by default, so the FIP_artifacts should be specified before launching compilation
|
|
- In case of using SOURCES-xxxx.tar.gz of Developer package the FIP_DEPLOYDIR_ROOT should be set as below:
|
|
$> export FIP_DEPLOYDIR_ROOT=$PWD/../../FIP_artifacts
|
|
|
|
To compile tf-a source code
|
|
$> make -f $PWD/../Makefile.sdk all
|
|
or for a specific config :
|
|
$ make -f $PWD/../Makefile.sdk TFA_DEVICETREE=stm32mp157c-ev1 TF_A_CONFIG=trusted ELF_DEBUG_ENABLE='1' all
|
|
|
|
NB: TFA_DEVICETREE flag must be set to switch to correct board configuration.
|
|
|
|
The generated FIP images are available in $FIP_DEPLOYDIR_ROOT/fip
|
|
|
|
6. Update software on board:
|
|
----------------------------
|
|
Please use STM32CubeProgrammer then only tick the boot partitions means patitions 0x1 to 0x6 (more informations on the wiki website http://wiki.st.com/stm32mpu)
|
|
|