[PING][PATCH] Optionally: Set the serial# environment variable on the i.MX7.

[Mark G]

On Mon, 11 May 2020, Stefano Babic wrote:

> patch was hidden in the flood of other patches and I am unsure if this
> belongs to i.MX:

No worries. To clarify the patch applies to all i.MX7 boards.

> On 19.02.20 22:01, Mark G wrote:
>> +      Linux kernel will use this as the serial number of the machine.
>> +
>
> I do not see the need of *another* CONFIG_ option. To make this
> available, CONFIG_SERIAL_TAG should also be on, else you cannot call
> get_board_serial() to get the serial from OCOTP. It think SET_SERIAL_ENV
> is quite superflous.

The intent here is to pass in the serial number via the environemnt 
(i.e. when not using ATAGS). My patch changes the preprocessor machinery 
so that get_board_serial() is callable if this option is set.

This then lets fdt_root() put it in the FDT passed to the kernel. It also 
means that it's easily changed for testing by being in the environment. 
The origins of this patch were for a customer switching to the Meerkat96 
board that utilized the CPU serial number in Linux for provisioning and 
testing.

> If this should be done global, local solution should be moved too. I
> mean the warp board doing exactly this in board code.

Ah. To be fair I never looked at the port for the Warp7 board as I do not 
have one. This really should be done globally; there isn't anything board 
specific for this behavior. It should apply to all i.MX7 boards, at a 
minimum.

Well - to clarify in theory there is nothing board specific. Could there 
be a board out there with an EEPROM that contains a different serial 
number that a customer would want to pass to the kernel? Possibly but 
unlikely.

> So which is the idea of this ? To move warp code and making global, or
> what ?

I would say this behavior should be global but the Warp7 port seems to add 
a board specific prefix to the CPU serial number. So I am hesitant to 
remove it. If you like I can #ifdef around it.

Additionally the Warp7 code isn't quite the same as it relies on ATAGS 
being enabled where as my patch does not.

Sincerely,
Mark G.