[U-Boot] [PATCH 14/25] x86: Support Intel FSP initialization path in start.S

[Bin Meng]

Hi Simon,

On Fri, Dec 5, 2014 at 7:40 AM, Simon Glass <sjg at chromium.org> wrote:
> Hi Bin,
>
> On 4 December 2014 at 08:02, Bin Meng <bmeng.cn at gmail.com> wrote:
>> Per Intel FSP architecture specification, FSP provides 3 routines
>> for bootloader to call. The first one is the TempRamInit (aka
>> Cache-As-Ram initialization) and the second one is the FspInit
>> which does the memory bring up (like MRC for other x86 targets)
>> and chipset initialization. Those two routines have to be called
>> before U-Boot jumping to board_init_f in start.S.
>>
>> The FspInit() will return several memory blocks called Hand Off
>> Blocks (HOBs) whose format is described in Platform Initialization
>> (PI) specification (part of the UEFI specication) to the bootloader.
>> Save this HOB address to the U-Boot global data for later use.
>>
>> Signed-off-by: Bin Meng <bmeng.cn at gmail.com>
>> ---
>>  arch/x86/cpu/start.S               | 37 +++++++++++++++++++++++++++++++++++++
>>  arch/x86/include/asm/global_data.h |  3 +++
>>  arch/x86/lib/asm-offsets.c         |  3 +++
>>  3 files changed, 43 insertions(+)
>>
>> diff --git a/arch/x86/cpu/start.S b/arch/x86/cpu/start.S
>> index f9662fb..b4a69d2 100644
>> --- a/arch/x86/cpu/start.S
>> +++ b/arch/x86/cpu/start.S
>> @@ -75,6 +75,35 @@ early_board_init_ret:
>>         jmp     car_init
>>  .globl car_init_ret
>>  car_init_ret:
>> +#ifdef CONFIG_HAVE_FSP
>
> Can we put this code into car_init?

Yes, there codes can be put into car_init. I actually considered
putting them to car_init before. In the FSP case, the car_init will
not return to car_init_ret. Instead it will jump directly to reserving
space on stack for global data. This may look weird why car_init does
not return to car_init_ret. But it is not a big problem. I will move
it to car_init in v2, and put a comment in start.S to explain.

>> +       /*
>> +        * The FSP TempRamInit initializes the ecx and edx registers to
>> +        * point to a temporary but writable memory range (Cache-As-RAM).
>> +        * ecx: the start of this temporary memory range,
>> +        * edx: the end of this range.
>> +        */
>> +
>> +       /* stack grows down from top of CAR */
>> +       movl    %edx, %esp
>> +
>> +       movl    $CONFIG_FSP_TEMP_RAM_ADDR, %eax
>> +       xorl    %edx, %edx
>> +       xorl    %ecx, %ecx
>> +       call    FspInitWrapper
>
> What does this return? Can you add a a comment?

Sure.

>> +
>> +.global fsp_init_done
>> +fsp_init_done:
>> +       /*
>> +        * We come here from FspInit with eax pointing to the HOB list.
>> +        * Save eax to esi temporarily.
>> +        */
>> +       movl    %eax, %esi
>> +       /*
>> +        * Re-initialize the ebp (BIST) to zero, as we already reach here
>> +        * which means we passed BIST testing before.
>
> How do you know we already reached here?

In the car_init, the BIST testing will be done before any call to FSP
to enable the CAR. If that fails, it will never get here.

[snip]

Regards,
Bin