[U-Boot] Falcon boot breaks on DRA7 because of commit b9c8ccab "env_mmc.c: Allow environment to be used within SPL"

[Jean-Jacques Hiblot]

On 24/01/2017 16:17, Tom Rini wrote:
> On Tue, Jan 24, 2017 at 10:26:38AM +0100, Jean-Jacques Hiblot wrote:
>
>> Hi Tom,
>>
>> I'm using a TI DRA7 platform and the falcon boot from MMC is broken
>> with v2017. The reason is that the standard "boot_os" is used to
>> tell whether the falcon mode should be used or not, but we can't
>> access it. The root cause is that the environment is stored in a
>> eMMC which is dev 1 not dev 0 on those platforms.
>>
>> What is the purpose of commit b9c8ccab. Is it because we want to
>> initialize only one MMC device in the SPL to reduce the boot time ?
> Please note that b9c8ccaba77b has been in since April 2014, so this is
> not some new behavior.
I had noticed that it's quite old indeed. I didn't mean that it's a 
regression. I'm just puzzled by the commit. what is its purpose ? why is 
SPL not using  CONFIG_SYS_MMC_ENV_DEV ?
I can trace this as the reason why I can't access the environment in 
eMMC (MMC2) when booting from SD (MMC1) and the fix is simply use dev = 
CONFIG_SYS_MMC_ENV_DEV instead of dev = 0.
But I guess that there is a reason for enforcing dev = 0. If the reason 
is still valid, I'll have to find another way of dealing with this. 
Otherwise we could revert to dev = CONFIG_SYS_MMC_ENV_DEV or
define a CONFIG_SYS_MMC_ENV_SPL_EV with a default value of 0.
> That said, we have CONFIG_SYS_MMC_ENV_DEV set
> which is what the commit in question uses to know where the env is.  I
> think you need to run a git bisect between v2016.11 (which I assume is
> when you last tested this)
Honestly I don't know when it broke. I still have to figure this. It may 
be that the test case I'm using hasn't work in a long time. From what I 
understand falcon boot on DRA7 is primarily used from QSPI or eMMC in 
raw mode. And those modes are not impacted by this problem with the 
environment.
>   and v2017.01 to see what commit changed
> things.  It's possible something has changed and broken this case
> requiring another tweak somewhere or another, thanks!
>