Re: [PATCH] ArmPlatformPkg/PrePeiCore: seed temporary stack before entering PEI core

[Ard Biesheuvel <ard.biesheuvel@linaro.org>]

On 23 October 2017 at 15:18, Gao, Liming <liming.gao@intel.com> wrote:
> This is for debug purpose. I think UINT32 is OK if AARCH64 can operate 32bit value.
>

Both my ARM and AARCH64 implementations write 16 bytes at a time, and
so whether the source value is UINT32 or UINT64 or UINT16 does not
make any difference at all.

>> -----Original Message-----
>> From: Laszlo Ersek [mailto:lersek@redhat.com]
>> Sent: Saturday, October 21, 2017 1:19 AM
>> To: Ard Biesheuvel <ard.biesheuvel@linaro.org>
>> Cc: Gao, Liming <liming.gao@intel.com>; edk2-devel@lists.01.org; Leif Lindholm <leif.lindholm@linaro.org>
>> Subject: Re: [edk2] [PATCH] ArmPlatformPkg/PrePeiCore: seed temporary stack before entering PEI core
>>
>> On 10/20/17 18:52, Ard Biesheuvel wrote:
>> > On 20 October 2017 at 17:51, Laszlo Ersek <lersek@redhat.com> wrote:
>> >> On 10/20/17 18:39, Ard Biesheuvel wrote:
>> >>> On 20 October 2017 at 17:37, Gao, Liming <liming.gao@intel.com> wrote:
>> >>>> Ard:
>> >>>>   This case is to share the same value between PeiCore and SecCore. I also think it will be better to define one fixed PCD in
>> MdeModulePkg.dec for this value. Could you submit bugzillar to catch this issue first?
>> >>>>
>> >>>
>> >>> Certainly!
>> >>
>> >> Would it be possible to define the PCD as UINT32, and task 64-bit SEC
>> >> (and PEI_CORE) code to first construct the wider value manually (in a
>> >> register or otherwise)?
>> >>
>> >> Just thinking out loud.
>> >>
>> >
>> > Could you think the reasoning behind that out loud as well?
>>
>> Haha, good stab :) Sure.
>>
>> In your patch you have:
>>
>> +#define INIT_CAR_VALUE      0x5AA55AA55AA55AA5
>>
>> for 64-bit, and
>>
>> +#define INIT_CAR_VALUE      0x5AA55AA5
>>
>> for 32-bit.
>>
>> Both 64-bit assembly code in SEC, and 64-bit C-code in the PEI_CORE, can
>> easily compose the large value from the small value, starting from
>> FixedPcdGet32(). The alternatives are:
>>
>> - asking the 32-bit assembly code to truncate the 64-bit constant -- it
>> won't compile,
>>
>> - defining *two* FixedAtBuild PCDs, one for 32-bit, another for 64-bit
>> SEC -- an idea probably not universally liked.
>>
>> ... When I originally started writing my previous email, I even thought
>> about introducing the PCD as UINT16 :) But then I realized, if any
>> platform lacks *some* 32-bit mode when it sets up the stack in assembly,
>> for C-language entry, then the platform won't be supported by edk2.
>>
>> Thanks
>> Laszlo