Fix imx-boot and u-boot-imx generation/instalation for when UBOOT_CONFIG is empty #1707
Conversation
angolini
force-pushed
the
empty_uboot_config
branch
from
3d07918 to
11470f3
Compare
| } | ||
|
|
||
| compile_mx8() { | ||
| bbnote 8QM boot binary build | ||
| cp ${DEPLOY_DIR_IMAGE}/${BOOT_TOOLS}/${SC_FIRMWARE_NAME} ${BOOT_STAGING}/scfw_tcm.bin | ||
| cp ${DEPLOY_DIR_IMAGE}/${ATF_MACHINE_NAME} ${BOOT_STAGING}/bl31.bin |
There was a problem hiding this comment.
Hi @angolini, I happened to be looking at taking the UBOOT_CONFIG changes into our local layer, but I'm confused how it even works. Since this function can be called multiple times, it will fail on the second call for any files which are not UBOOT_CONFIG specific (like the ATF file here) since in that case it will be copying the same file to the same location.
I feel there is a structural problem with do_deploy as well since it calls deploy_${SOC_FAMILY} without setting the _EXTRA variables at all.
There was a problem hiding this comment.
I agree that we have to think of a better way to reorg this recipe. It's becoming more and more disorganized and crazy. But maybe it would be better if we only switch to bootloader container instead?
If you have any suggestions on how to fix/enhance the PR, please let me know.
There was a problem hiding this comment.
Yes, I believe moving to the bootloader container would be better, but in the meantime, I guess the recipe should at least be fixed.
A refactoring of the functions is needed so common logic occurs once and uboot-config-specific logic occurs once per config.
There was a problem hiding this comment.
I changed where the spl is copied.
@thochstein , if you know a better way to test this (without the machines, as I don't have them) please, let me know
There was a problem hiding this comment.
Can you just put two values in UBOOT_CONFIG? I'm pretty sure that will just fail to build.
There was a problem hiding this comment.
./tmp/sysroots-components/imx8mm_lpddr4_evk/imx-boot/boot/imx-boot-imx8mm-lpddr4-evk-sd.bin-flash_evk_flexspi
./tmp/sysroots-components/imx8mm_lpddr4_evk/imx-boot/boot/imx-boot-imx8mm-lpddr4-evk-fspi.bin-flash_evk_flexspi
I've been testing this as well.
But if you think in anything else, please, let me know
(this PR does not touch the multiple UBOOT_CONFIG, it only deals when UBOOT_CONFIG is empty)
The patchset to enable multiple u-boot targets for the same machine is based only on UBOOT_CONFIG variable. For the case where a machine is not using UBOOT_MACHINE instead (and UBOOT_CONFIG is empty) the previous code does not work, this is the reason for this fix. Signed-off-by: Daiane Angolini <[email protected]>
The patchset to enable multiple u-boot targets (and as a consequence, multimple imx-boot binaries) for the same machine is based only on UBOOT_CONFIG variable. For the case where a machine is not using UBOOT_MACHINE instead (and UBOOT_CONFIG is empty) the previous code does not work, this is the reason for this fix. Signed-off-by: Daiane Angolini <[email protected]>
angolini
force-pushed
the
empty_uboot_config
branch
from
11470f3 to
4fd7444
Compare
What
It adds a path to install and compile the files for those cases when a machine is configured with
UBOOT_MACHINEinstead ofUBOOT_CONFIG.tests
Tested with the following changes in
conf/local.conf, it means I only tested forMACHINE = 'imx8mm-lpddr4-evk'andMACHINE = 'imx93-11x11-lpddr4x-evk'.points of attention
I don't know why I cannot set the variablefixedUBOOT_NAME_EXTRAoutside thecompile_${SOC_FAMILY}. If you see why, let me know.u-boot-spl.binandu-boot-nodtb.binare copied to the directory used to create the imx-boot: