Re: [PATCH v2 3/8] MdeModulePkg/DxeCorePerformanceLib:Track FPDT record in DXE phase

["Zeng, Star" <star.zeng@intel.com>]

I could not recall the whole background.

Let us separate the problem into two parts.

1. For this special case in DxeCorePerformanceLib, we at least have approach (although just for short term).
2. For the general problem about whether introducing new interface (PCD and API, seemingly it needs to be in MdePkg), it depends on whether we agree to add new interface and seems need more time to discuss. If yes, then library/driver could be updated to consume it.

If we agree to add new interface, as current instances of MemoryAllocationLib are very nature by phases (PEI, DXE (Uefi/DxeCore), SMM, etc), but the new interface is just needed in DXE (Uefi/DxeCore) instances, how about adding it in UefiLib?


Thanks,
Star
-----Original Message-----
From: Laszlo Ersek [mailto:lersek@redhat.com] 
Sent: Tuesday, May 22, 2018 6:00 PM
To: Zeng, Star <star.zeng@intel.com>; Ard Biesheuvel <ard.biesheuvel@linaro.org>
Cc: Bi, Dandan <dandan.bi@intel.com>; Andrew Fish <afish@apple.com>; Kinney, Michael D <michael.d.kinney@intel.com>; Leif Lindholm <leif.lindholm@linaro.org>; edk2-devel@lists.01.org; Gao, Liming <liming.gao@intel.com>; Matt Sealey <neko@bakuhatsu.net>
Subject: Re: [edk2] [PATCH v2 3/8] MdeModulePkg/DxeCorePerformanceLib:Track FPDT record in DXE phase

Hi Star,

On 05/22/18 11:38, Zeng, Star wrote:
> It comes to an old question: how to know PEI is 32bit/64bit? But not DXE is 32bit/64bit.

we've discussed this question in the past (as you remember :) ) and people suggested both HOBs and PCDs to expose the "PEI phase max address" to DXE. All of those ideas were rejected -- I'm unsure why, to this day --, but assuming people who opposed those ideas are going to oppose them again, I think Ard's idea is good, about introducing dedicated MemoryAllocationLib interfaces, for PEI-targeted memory allocation.

Once drivers adopt these new interfaces, platforms can at least fork the MemoryAllocationLib instance(s) they need, and customize the PEI phase max address one way or another.

Thanks,
Laszlo

> -----Original Message-----
> From: Ard Biesheuvel [mailto:ard.biesheuvel@linaro.org]
> Sent: Tuesday, May 22, 2018 5:03 PM
> To: Zeng, Star <star.zeng@intel.com>
> Cc: Bi, Dandan <dandan.bi@intel.com>; Laszlo Ersek 
> <lersek@redhat.com>; Andrew Fish <afish@apple.com>; Kinney, Michael D 
> <michael.d.kinney@intel.com>; Leif Lindholm 
> <leif.lindholm@linaro.org>; edk2-devel@lists.01.org; Gao, Liming 
> <liming.gao@intel.com>; Matt Sealey <neko@bakuhatsu.net>
> Subject: Re: [edk2] [PATCH v2 3/8] 
> MdeModulePkg/DxeCorePerformanceLib:Track FPDT record in DXE phase
> 
> pd On 22 May 2018 at 10:58, Zeng, Star <star.zeng@intel.com> wrote:
>> How about the code to try allocating <4G memory, then try allocating any memory if failed to allocate <4G memory?
>>
>> For example, the code snippet in SmbiosDxe.c.
>>
>>     PhysicalAddress = 0xffffffff;
>>     Status = gBS->AllocatePages (
>>                     AllocateMaxAddress,
>>                     EfiRuntimeServicesData,
>>                     EFI_SIZE_TO_PAGES (sizeof (SMBIOS_TABLE_ENTRY_POINT)),
>>                     &PhysicalAddress
>>                     );
>>     if (EFI_ERROR (Status)) {
>>       DEBUG ((EFI_D_ERROR, "SmbiosCreateTable () could not allocate EntryPointStructure < 4GB\n"));
>>       Status = gBS->AllocatePages (
>>                       AllocateAnyPages,
>>                       EfiRuntimeServicesData,
>>                       EFI_SIZE_TO_PAGES (sizeof (SMBIOS_TABLE_ENTRY_POINT)),
>>                       &PhysicalAddress
>>                       );
>>      if (EFI_ERROR (Status)) {
>>         return EFI_OUT_OF_RESOURCES;
>>       }
>>     }
>>
> 
> Yes, that should do the trick.
> 
> But perhaps we should add something like
> 
> AllocatePeiAccessiblePages()
> 
> to MemoryAllocationLib, which just uses AllocateMaxAddress or 
> AllocateAnyPages depending on whether MDE_CPU_X64 is #defined.
>