[Automated-testing] minnowboard boot & control

[William Mills]

On 02/04/2014 05:06 PM, Otavio Salvador wrote:
> On Tue, Feb 4, 2014 at 8:04 PM, William Mills <wmills at ti.com> wrote:
>> On 02/04/2014 04:51 PM, Otavio Salvador wrote:
>>> Hello,
>>>
>>> On Tue, Feb 4, 2014 at 7:45 PM, William Mills <wmills at ti.com> wrote:
>>> ...
>>>> On BeagleBone Black:
>>>> The Si Boot ROM can boot from TFTP.
>>>> The Si Boot ROM can boot from eMMC or SD.
>>>> With relay controlled pull-ups/downs I can control the boot device.
>>>>
>>>> With this control I can recover a device regardless of what a previous
>>>> test
>>>> has done to eMMC or SD Card.
>>>> (My work is documented at:
>>>> http://processors.wiki.ti.com/index.php/BeagleBone_in_a_Board_Farm
>>>> )
>>>
>>> Except U-Boot tests or am I wrong?
>>>
>> I think your asking if I can test experimental (broken) u-boots and
>> still recover.
>>
>> Yes, with this setup I can.
>> With the relays, I set the boot mode to TFTP at the ROM level.
>> The full boot sequence is:
>> ROM -> TFTP SPL -> TFTP Uboot -> whatever U-boot decides
>>
>> With an unpatched u-boot you can use the serial console to interrupt
>> and get control of u-boot and then have it boot whatever you want.
>>
>> (Back in July I had u-boot hacks to make it get uEnv.txt from TFTP also and
>> then the files on the TFTP were in complete control.
>> Resurrecting that work and cleaning it up is on my TODO list :)
> But will the SPL be also run from the TFTP? Otherwise you're changing
> 'part' of it.
>

OK, this was not the point of this thread.  Just pointing to prior art.  
But ...

You can put your experimental stuff on TFTF and just test it that way
  -- or (what I would do) ---
keep know good SPL and U-boot on TFTP and use them to write the SPL and 
U-boot
to be tested to eMMC or uSD.  After you have done the updates you let go 
of the
boot mode relays and power cycle.
You are now testing the experimental SPL/U-boot in its natural 
environment just
like the real users will.