Re: [PATCH 2/2] ArmPlatformPkg/PrePi: avoid global variable write to mSystemMemoryEnd

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

On 24 October 2016 at 13:05, Ryan Harkin <ryan.harkin@linaro.org> wrote:
> On 24 October 2016 at 09:57, Ard Biesheuvel <ard.biesheuvel@linaro.org> wrote:
>> The global variable mSystemMemoryEnd is initialized by PrePi only if
>> it has not been initialized by ArmPlatformPeiBootAction(). This allows
>> platforms executing under, e.g., ARM Trusted Firmware to dynamically
>> reserve a window at the top of memory that will be used by the secure
>> firmware.
>>
>> However, PrePi is a SEC module, and writing to a global variable
>> violates the SEC constraints, since SEC and PEI may execute from NOR
>> flash.
>>
>> So instead, initialize mSystemMemoryEnd statically. This will ensure
>> it holds the correct value for all implementations where the value
>> is not overridden, but still allows it to be overridden during the
>> call to ArmPlatformPeiBootAction().
>>
>> Note that this patch also fixes a latent bug on 32-bit platforms where
>> a value of mSystemMemoryEnd exceeding 4 GB would be truncated to 32-bits
>> rather than limited to (4 GB - 1)
>>
>> Contributed-under: TianoCore Contribution Agreement 1.0
>> Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
>
> I tested this successfully on FVP Foundation & AEMv8 models, Juno
> R0/1/2 and TC2.
>
> Tested-by: Ryan Harkin <ryan.harkin@linaro.org>
>

Many thanks!