Hello Marcel!
-----Original Message----- From: Marcel Ziswiler marcel.ziswiler@toradex.com Sent: Monday, August 23, 2021 8:57 AM To: u-boot@lists.denx.de; ZHIZHIKIN Andrey <andrey.zhizhikin@leica- geosystems.com> Cc: festevam@gmail.com; peng.fan@nxp.com; uboot-imx@nxp.com; sbabic@denx.de; frieder.schrempf@kontron.de; heiko.thiery@gmail.com Subject: Re: [PATCH v1 1/7] imx: mkimage_fit_atf: fix legacy image generation
Hi Andrey
Long time no see (;-p).
Indeed! :D
On Sun, 2021-08-22 at 10:07 +0000, ZHIZHIKIN Andrey wrote:
Hello Marcel,
-----Original Message----- From: U-Boot u-boot-bounces@lists.denx.de On Behalf Of Marcel Ziswiler Sent: Friday, August 20, 2021 10:52 PM To: u-boot@lists.denx.de Cc: Heiko Thiery heiko.thiery@gmail.com; Stefano Babic sbabic@denx.de; Fabio Estevam festevam@gmail.com; Frieder Schrempf frieder.schrempf@kontron.de; Marcel Ziswiler marcel.ziswiler@toradex.com; NXP i.MX U-Boot Team <uboot- imx@nxp.com>; Peng Fan peng.fan@nxp.com Subject: [PATCH v1 1/7] imx: mkimage_fit_atf: fix legacy image generation
From: Marcel Ziswiler marcel.ziswiler@toradex.com
While most boards meanwhile migrated to using binman a few like the verdin- imx8mm are still using the legacy image generation. Unfortunately, the legacy image generation is currently broken which is especially bad for any kind of bisection attempts. Anyway, this fixes it even though we will also migrate to using binman shortly.
This change has been already proposed in [1],
Well, what I do not get is how one can move forward and leave all kinds of stuff just broken. Fact is, that the legacy image creation has been and still is plain simply broken!
but the discussion went into the direction of monolithic "flash.bin" rather than a
migration to use binman.
Well, those two do actually not rule each other out. Remember, later in this patch set I am migrating to using binman which I instruct to generate a monolithic "flash.bin" again.
I guess if this change is really needed due to the fact that the migration of some boards is really difficult
- the original patch can be taken.
What I found extremely problematic is, as mentioned initially, stuff is currently broken which makes e.g. bisecting other issues extremely cumbersome. But in theory, as I propose now to migrate anyway, we could just not care and leave it broken for anybody else. I just feel this is not really too nice of a gesture!
Totally agree, keeping the broken implementation is not a nice thing as people unaware of this might base their work on it and wonder why it does not build on the first place...
Actually, I wanted to propose to remove this script from the tree and let derivatives that rely on it to fail to being forcibly converted to binman. But I was humble enough to do it through...
There are still those boards that use FIT generator script: configs/cgtqmx8_defconfig:CONFIG_SPL_FIT_GENERATOR="arch/arm/mach-imx/mkimage_fit_atf.sh" configs/imx8mm-icore-mx8mm-ctouch2_defconfig:CONFIG_SPL_FIT_GENERATOR="arch/arm/mach-imx/mkimage_fit_atf.sh" configs/imx8mm-icore-mx8mm-edimm2.2_defconfig:CONFIG_SPL_FIT_GENERATOR="arch/arm/mach-imx/mkimage_fit_atf.sh" configs/imx8mm_beacon_defconfig:CONFIG_SPL_FIT_GENERATOR="arch/arm/mach-imx/mkimage_fit_atf.sh" configs/imx8mm_venice_defconfig:CONFIG_SPL_FIT_GENERATOR="arch/arm/mach-imx/mkimage_fit_atf.sh" configs/imx8mn_beacon_2g_defconfig:CONFIG_SPL_FIT_GENERATOR="arch/arm/mach-imx/mkimage_fit_atf.sh" configs/imx8mn_beacon_defconfig:CONFIG_SPL_FIT_GENERATOR="arch/arm/mach-imx/mkimage_fit_atf.sh" configs/imx8mq_evk_defconfig:CONFIG_SPL_FIT_GENERATOR="arch/arm/mach-imx/mkimage_fit_atf.sh" configs/imx8mq_phanbell_defconfig:CONFIG_SPL_FIT_GENERATOR="arch/arm/mach-imx/mkimage_fit_atf.sh" configs/imx8qm_rom7720_a1_4G_defconfig:CONFIG_SPL_FIT_GENERATOR="arch/arm/mach-imx/mkimage_fit_atf.sh" configs/phycore-imx8mm_defconfig:CONFIG_SPL_FIT_GENERATOR="arch/arm/mach-imx/mkimage_fit_atf.sh" configs/pico-imx8mq_defconfig:CONFIG_SPL_FIT_GENERATOR="arch/arm/mach-imx/mkimage_fit_atf.sh" configs/verdin-imx8mm_defconfig:CONFIG_SPL_FIT_GENERATOR="arch/arm/mach-imx/mkimage_fit_atf.sh"
Verdin gets converted with your other patch, while for all the others I'm not sure what the status is.
There is however one board which has been introduced quite recent (Cc'ing Oliver here) and didn't went for binman implementation from the start. This is a clear indication that people are still relying on old FIT generator, even though there is a clear warning about that it is not suppose to be used.
I guess at this stage, I'd rather leave this question for maintainers to decide - enforce the conversion to binman via deleting the FIT generator script, or fix the generator script with the patch from Oliver and let board maintainers to either use it or migrate.
However, I've commented out in that thread that there is a warning regarding the usage of scripts and migration notice, so maybe it does make
sense to spend extra effort to migrate away from this script at all?
Yes, of course, it is the goal to migrate. I just don't get how in IT new stuff gets introduced all the time with leaving past things broken. Just a little bit annoying...
Fixes: commit cb9faa6f98ae ("tools: Use a single target-independent config to enable OpenSSL")
...
Cheers
Marcel
Cheers, Andrey