Re: [edk2-devel] [edk2-platforms:PATCH V1] MinPlatformPkg/SaveMemoryConfig: Support NVS Data compression.

["Isaac Oram" <isaac.w.oram@intel.com>]

Rereading that thread and this request, other opinions, etc, I am convinced that the simpler interface and board responsibility is the correct short/medium term answer.  I don't have an opinion on Mike Kinney's question on encapsulated services.  I generally like that design, though I am sensitive to Michael Kubacki's feedback that variable services are too complex as it is.  I guess it probably depends a lot on the specifics of the proposal.  That said, it is pretty easy to migrate from a board specific solution to a more base layer solution in the future, so adopting this now doesn't seem harmful to me.

Raghava, can you look at Michael's fix?  It looks nearly identical to yours and general convention is to accept the earlier one in case of collision I believe.  I like his PCD naming a little better, but both are fine to me.

Regards,
Isaac

-----Original Message-----
From: devel@edk2.groups.io <devel@edk2.groups.io> On Behalf Of Michael Kubacki
Sent: Tuesday, May 9, 2023 8:26 AM
To: devel@edk2.groups.io; Kinney, Michael D <michael.d.kinney@intel.com>; Chiu, Chasel <chasel.chiu@intel.com>; Oram, Isaac W <isaac.w.oram@intel.com>; Gudla, Raghava <raghava.gudla@intel.com>
Cc: Desimone, Nathaniel L <nathaniel.l.desimone@intel.com>; Kubacki, Michael <michael.kubacki@microsoft.com>; Chaganty, Rangasai V <rangasai.v.chaganty@intel.com>
Subject: Re: [edk2-devel] [edk2-platforms:PATCH V1] MinPlatformPkg/SaveMemoryConfig: Support NVS Data compression.

At the surface, this looks similar to the following patch I sent a while
back:

https://edk2.groups.io/g/devel/message/92644

That triggered a thread where we had a similar discussion about LargeVariableLib responsibilities, etc.

We still have a fork of SaveMemoryConfig that uses the PCD I sent in the
patch:

https://github.com/microsoft/mu_common_intel_min_platform/blob/release/202208/MinPlatformPkg/FspWrapper/SaveMemoryConfig/SaveMemoryConfig.inf

I believe a challenge that led to adding compression (in our fork code) to SaveMemoryConfig was the fact that the data was produced in pre-mem PEI and compression is expensive in CAR.

In general though, I still believe that it is simpler to separate data mutation from service APIs. Platforms can manipulate data to achieve their goals whether size, security, and so on and the service APIs provide a simple interface to store and retrieve that data blob.

I'd also like to see FSP reduce in size and eliminate operations that are not essential to its role and can be consolidated/reused in the wrapper.

Thanks,
Michael

On 5/8/2023 5:48 PM, Michael D Kinney wrote:
> When reviewing the Variable feature that adds integrity and 
> confidentiality, I suggested that the interface between the Variable 
> services and the NVStorage could provide an abstraction to 
> encode/decode the stored data that would support encryption, 
> compression, or both.  Could also support a platform policy for which variables the encode/decode operation is applied.
> 
> Wouldn't that be a better abstraction than piecemeal adding these 
> features?
> 
> Doesn't mean that this can't go in as-is.  But would be an opportunity 
> to consolidate in the future.
> 
> Mike
> 
>> -----Original Message-----
>> From: devel@edk2.groups.io <devel@edk2.groups.io> On Behalf Of Chiu, 
>> Chasel
>> Sent: Monday, May 8, 2023 12:37 PM
>> To: Oram, Isaac W <isaac.w.oram@intel.com>; Gudla, Raghava 
>> <raghava.gudla@intel.com>; devel@edk2.groups.io
>> Cc: Desimone, Nathaniel L <nathaniel.l.desimone@intel.com>; Kubacki, 
>> Michael <michael.kubacki@microsoft.com>; Chaganty, Rangasai V 
>> <rangasai.v.chaganty@intel.com>; Chiu, Chasel <chasel.chiu@intel.com>
>> Subject: Re: [edk2-devel] [edk2-platforms:PATCH V1]
>> MinPlatformPkg/SaveMemoryConfig: Support NVS Data compression.
>>
>>
>> Hi Isaac,
>>
>> Just my thoughts, I would vote for platform/bootloader to decide 
>> compressing the variable data before saving to NVRAM or not.
>> It should be optional and when platform having big SPI flash they 
>> might not want to do this.
>>
>> Thanks,
>> Chasel
>>
>>
>>> -----Original Message-----
>>> From: Oram, Isaac W <isaac.w.oram@intel.com>
>>> Sent: Monday, May 8, 2023 12:15 PM
>>> To: Gudla, Raghava <raghava.gudla@intel.com>; devel@edk2.groups.io
>>> Cc: Chiu, Chasel <chasel.chiu@intel.com>; Desimone, Nathaniel L 
>>> <nathaniel.l.desimone@intel.com>; Kubacki, Michael 
>>> <michael.kubacki@microsoft.com>; Chaganty, Rangasai V 
>>> <rangasai.v.chaganty@intel.com>
>>> Subject: RE: [edk2-platforms:PATCH V1] MinPlatformPkg/SaveMemoryConfig:
>>> Support NVS Data compression.
>>>
>>> The proposed implementation is fine and I will reviewed-by and push 
>>> if that
>> is
>>> the desired direction.
>>>
>>> My question is if we generally like the design of doing compression 
>>> in
>> common
>>> MinPlatform code, decompression in board specific FSP wrapper code.  
>>> The alternative design is to do compression and decompression inside the FSP.
>> This
>>> seems like a slightly simpler separation of responsibilities.
>>> The board code is responsible for save/restore, the FSP code is 
>>> responsible
>> for
>>> data blob creation and use.  Data integrity, authentication, 
>>> compression, etc
>> all
>>> can be done based on more detailed knowledge of the silicon
>> implementation
>>> requirements.
>>>
>>> I can see another argument though that doing it inside FSP 
>>> effectively forces both bootloader and FSP to carry 
>>> compression/decompression.  The compression/decompression aren't 
>>> likely to need to be silicon specific.  And bootloader may have more 
>>> requirements to balance than just the silicon requirements.
>>>
>>> Can I get some votes on preferred answer for 
>>> compressing/decompressing
>> FSP
>>> non-volatile data in bootloader or FSP?
>>>
>>> Thanks,
>>> Isaac
>>>
>>> -----Original Message-----
>>> From: Gudla, Raghava <raghava.gudla@intel.com>
>>> Sent: Friday, May 5, 2023 5:03 PM
>>> To: devel@edk2.groups.io
>>> Cc: Gudla, Raghava <raghava.gudla@intel.com>; Chiu, Chasel 
>>> <chasel.chiu@intel.com>; Desimone, Nathaniel L 
>>> <nathaniel.l.desimone@intel.com>; Oram, Isaac W
>> <isaac.w.oram@intel.com>
>>> Subject: [edk2-platforms:PATCH V1] MinPlatformPkg/SaveMemoryConfig:
>>> Support NVS Data compression.
>>>
>>> Around 50KB "FspNonVolatileStorageHob" data can be compressed to 
>>> approximately 3 KB which can save NVRAM space and enhance life of 
>>> the SPI part by decreasing the number of reclaim cycles needed.
>>>
>>> This patch added support to compress "FspNonVolatileStorageHob" data
>> before
>>> saving to NVRAM.
>>>
>>> A PcdEnableCompressFspNvsHob is introduced to enable/disable this
>> feature per
>>> platform usage.
>>>
>>> Cc: Chasel Chiu <chasel.chiu@intel.com>
>>> Cc: Nate DeSimone <nathaniel.l.desimone@intel.com>
>>> Cc: Isaac Oram <isaac.w.oram@intel.com>
>>>
>>> Signed-off-by: Raghava Gudla <raghava.gudla@intel.com>
>>> ---
>>>   .../SaveMemoryConfig/SaveMemoryConfig.c       | 34 +++++++++++++++++++
>>>   .../SaveMemoryConfig/SaveMemoryConfig.inf     |  6 +++-
>>>   .../Include/Dsc/CoreCommonLib.dsc             |  1 +
>>>   .../Intel/MinPlatformPkg/MinPlatformPkg.dec   |  5 +++
>>>   4 files changed, 45 insertions(+), 1 deletion(-)
>>>
>>> diff --git
>>>
>> a/Platform/Intel/MinPlatformPkg/FspWrapper/SaveMemoryConfig/SaveMemo
>> r
>>> yConfig.c
>>>
>> b/Platform/Intel/MinPlatformPkg/FspWrapper/SaveMemoryConfig/SaveMem
>> or
>>> yConfig.c
>>> index 0215e8eed..8aa935b54 100644
>>> ---
>>>
>> a/Platform/Intel/MinPlatformPkg/FspWrapper/SaveMemoryConfig/SaveMemo
>> r
>>> yConfig.c
>>> +++
>>>
>> b/Platform/Intel/MinPlatformPkg/FspWrapper/SaveMemoryConfig/SaveMem
>> o
>>> +++ ryConfig.c
>>> @@ -21,6 +21,8 @@ SPDX-License-Identifier: BSD-2-Clause-Patent
>>>   #include <Library/LargeVariableWriteLib.h> #include 
>>> <Library/VariableWriteLib.h> #include 
>>> <Guid/FspNonVolatileStorageHob2.h>+#include <Library/PcdLib.h>+#include
>>> <Library/CompressLib.h>  /**   This is the standard EFI driver point that
>> detects
>>> whether there is a@@ -45,6 +47,9 @@ SaveMemoryConfigEntryPoint (
>>>     UINTN             DataSize;   UINTN             BufferSize;   BOOLEAN
>>> DataIsIdentical;+  VOID              *CompressedData;+  UINT64
>>> CompressedSize;+  UINTN             CompressedAllocationPages;    DataSize
>> = 0;
>>> BufferSize      = 0;@@ -73,6 +78,35 @@ SaveMemoryConfigEntryPoint (
>>>       }   } +  if (PcdGetBool (PcdEnableCompressFspNvsHob) == 1) {+
>>> CompressedData            = NULL;+    CompressedSize            = 0;+
>>> CompressedAllocationPages = 0;++    DEBUG ((DEBUG_INFO, "compressing
>> mem
>>> config nvs variable\n"));+    if (DataSize > 0) {+
>> CompressedAllocationPages =
>>> EFI_SIZE_TO_PAGES (DataSize);+      CompressedData = AllocatePages
>>> (CompressedAllocationPages);+      if (CompressedData == NULL) {+
>> DEBUG
>>> ((DEBUG_ERROR, "[%a] - Failed to allocate compressed data buffer.\n",
>>> __func__));+        ASSERT_EFI_ERROR (EFI_OUT_OF_RESOURCES);+        return
>>> EFI_OUT_OF_RESOURCES;+      }++      CompressedSize = EFI_PAGES_TO_SIZE
>>> (CompressedAllocationPages);+      Status = Compress (HobData, DataSize,
>>> CompressedData, &CompressedSize);+      if (EFI_ERROR (Status)) {+
>> DEBUG
>>> ((DEBUG_ERROR, "[%a] - failed to compress data. Status = %r\n", __func__,
>>> Status));+        ASSERT_EFI_ERROR (Status);+
>> FreePool(CompressedData);+
>>> return Status;+      } else {+        HobData  = CompressedData;+        DataSize
>> =
>>> (UINTN) CompressedSize;+      }+    }+  }+   if (HobData != NULL) {     DEBUG
>>> ((DEBUG_INFO, "FspNvsHob.NvsDataLength:%d\n", DataSize));     DEBUG
>>> ((DEBUG_INFO, "FspNvsHob.NvsDataPtr   : 0x%x\n", HobData));diff --git
>>>
>> a/Platform/Intel/MinPlatformPkg/FspWrapper/SaveMemoryConfig/SaveMemo
>> r
>>> yConfig.inf
>>>
>> b/Platform/Intel/MinPlatformPkg/FspWrapper/SaveMemoryConfig/SaveMem
>> or
>>> yConfig.inf
>>> index 61e85a658..77920d031 100644
>>> ---
>>>
>> a/Platform/Intel/MinPlatformPkg/FspWrapper/SaveMemoryConfig/SaveMemo
>> r
>>> yConfig.inf
>>> +++
>>>
>> b/Platform/Intel/MinPlatformPkg/FspWrapper/SaveMemoryConfig/SaveMem
>> o
>>> +++ ryConfig.inf
>>> @@ -26,6 +26,7 @@
>>>     LargeVariableReadLib   LargeVariableWriteLib   BaseLib+  CompressLib
>>> [Packages]   MdePkg/MdePkg.dec@@ -45,6 +46,9 @@
>>>     gFspNonVolatileStorageHob2Guid                ## CONSUMES
>>> gFspNvsBufferVariableGuid                     ## PRODUCES +[Pcd]+
>>> gMinPlatformPkgTokenSpaceGuid.PcdEnableCompressFspNvsHob+ [Depex]
>>> gEfiVariableArchProtocolGuid        AND-  gEfiVariableWriteArchProtocolGuid
>>> \ No newline at end of file
>>> +  gEfiVariableWriteArchProtocolGuiddiff --git 
>>> + a/Platform/Intel/MinPlatformPkg/Include/Dsc/CoreCommonLib.dsc
>>> + b/Platform/Intel/MinPlatformPkg/Include/Dsc/CoreCommonLib.dsc
>>> index 5ce21cf31..dfe7d836d 100644
>>> --- a/Platform/Intel/MinPlatformPkg/Include/Dsc/CoreCommonLib.dsc
>>> +++ b/Platform/Intel/MinPlatformPkg/Include/Dsc/CoreCommonLib.dsc
>>> @@ -147,6 +147,7 @@
>>>
>>>
>> BmpSupportLib|MdeModulePkg/Library/BaseBmpSupportLib/BaseBmpSuppor
>> tLi
>>> b.inf
>>>
>> LargeVariableReadLib|MinPlatformPkg/Library/BaseLargeVariableLib/Base
>> LargeVariableReadLib|Larg
>>> eVariableReadLib.inf
>>>
>> LargeVariableWriteLib|MinPlatformPkg/Library/BaseLargeVariableLib/Bas
>> LargeVariableWriteLib|eLarg
>>> eVariableWriteLib.inf+
>>> CompressLib|MinPlatformPkg/Library/CompressLib/CompressLib.inf    #   #
>>> CryptLibdiff --git 
>>> a/Platform/Intel/MinPlatformPkg/MinPlatformPkg.dec
>>> b/Platform/Intel/MinPlatformPkg/MinPlatformPkg.dec
>>> index 784abb828..e21d55fb3 100644
>>> --- a/Platform/Intel/MinPlatformPkg/MinPlatformPkg.dec
>>> +++ b/Platform/Intel/MinPlatformPkg/MinPlatformPkg.dec
>>> @@ -348,6 +348,11 @@
>>>
>>>
>> gMinPlatformPkgTokenSpaceGuid.PcdFadtFlags|0x000086A5|UINT32|0x90000
>>> 027
>>>
>> gMinPlatformPkgTokenSpaceGuid.PcdFadtMajorVersion|0x06|UINT8|0x90000
>> 0
>>> 30
>>>
>> gMinPlatformPkgTokenSpaceGuid.PcdFadtMinorVersion|0x03|UINT8|0x90000
>> 0
>>> 31+## Controls whether the Memory Config UEFI Variable is saved as
>>> compressed data.+# Data compression can significantly reduce 
>>> variable
>> storage
>>> usage for FSP NVS buffer data.+# Platforms that choose to compress 
>>> the
>> data
>>> will need to decompress the variable data upon+# extraction.+
>>>
>> gMinPlatformPkgTokenSpaceGuid.PcdEnableCompressFspNvsHob|FALSE|BOO
>> L
>>> EAN|0x90000032  [PcdsFixedAtBuild] --
>>> 2.19.1.windows.1
>>
>>
>>
>>
>>
> 
> 
> 
> 
>