[U-Boot] [PATCH v2] mx6: Read silicon revision from register

[Eric Nelson <eric.nelson@boundarydevices.com>]

On 03/13/2012 08:49 AM, Stefano Babic wrote:

Hi Stefano,

> On 13/03/2012 16:40, Eric Nelson wrote:
>> On 03/13/2012 07:18 AM, Fabio Estevam wrote:
>>> Instead of hardcoding the mx6 silicon revision, read it from the
>>> proper register.
>>>
>>> Signed-off-by: Fabio Estevam<fabio.estevam@freescale.com>
>>> ---
>>> Changes since v1:
>>> - Fix typo on Subject
>>>    arch/arm/cpu/armv7/mx6/soc.c |    3 ++-
>>>    1 files changed, 2 insertions(+), 1 deletions(-)
>>>
>>> diff --git a/arch/arm/cpu/armv7/mx6/soc.c b/arch/arm/cpu/armv7/mx6/soc.c
>>> index 2ac74b5..639bf30 100644
>>> --- a/arch/arm/cpu/armv7/mx6/soc.c
>>> +++ b/arch/arm/cpu/armv7/mx6/soc.c
>>> @@ -32,7 +32,8 @@
>>>
>>>    u32 get_cpu_rev(void)
>>>    {
>>> -    int system_rev = 0x61000 | CHIP_REV_1_0;
>>> +    int reg = readl(ANATOP_BASE_ADDR + 0x260)&   0xFF;
>>> +    int system_rev = 0x61000 | 0x10 + reg;
>>>
>>>        return system_rev;
>>>    }
>>
>> Nice!
>>      MX6Q SABRELITE U-Boot>  md 0x20C8260 1
>>      020c8260: 00630000    ..c.
>>
>> This beats the heck out of my hack:
>>      http://lists.denx.de/pipermail/u-boot/2012-March/120102.html
>>
>> Now the question is how get_cpu_rev() relates to get_board_rev().
>
> Really they should be unrelated.....I know, we have several cases where
> the board revision contains the cpu, but in u-boot they are unrelated.
> get_board_rev() should only have the revision of the board, if any.
>
>>
>> We had a discussion a while back about reserving some space in OTP
>> to allow per-board revision information.
>>
>> Fabio, do you have any guidance about how/whether the system_rev
>> should be updated to include a board revision?
>
> Do you mean which is the version to be put to make the imxlib happy ?
>

That's part of the question. imxlib seems to expect bits 12..31 to
contain the CPU:

	#define mxc_cpu()               (system_rev >> 12)

and bits 0-7 to represent the tapeout of the CPU:
	#define mxc_cpu_rev()           (system_rev & 0xFF)

That's a lot of bits for each, but tells me where the userspace
wants an **actual** board revision (bits 8-11).

That's confirmed here.
	http://opensource.freescale.com/git?p=imx/uboot-imx.git;a=blob;f=include/asm-arm/arch-mx6/mx6.h;h=ae52fa2fed12d468ef4f920afece7478fcb5c479;hb=imx_v2009.08_11.11.01#l617

The other part of the question is where a board revision should be
stored in OTP, and Fabio answered that under separate cover.

The board id and revision should be read from OTP longword #0x26,
bits 8-12, and the mapping won't be 1:1 with system_rev.

I suppose it's time for us to get some more structure in place
for OTP.

Regards,


Eric