Re: [Patch v3 2/3] UefiCpuPkg/S3Resume2Pei: Send S3 resume finished event to SmmCore.

["Dong, Eric" <eric.dong@intel.com>]

Hi Laszlo,

Sorry for this patch break OVMF. Platform without this Ppi is not what we expected before. So we just add ASSERT here. 

I will send a patch ASAP to fix this ASSERT case.   Also we will provide another patch later to handle the platform without this Ppi case.

Thanks,
Eric

-----Original Message-----
From: edk2-devel [mailto:edk2-devel-bounces@lists.01.org] On Behalf Of Laszlo Ersek
Sent: Thursday, October 12, 2017 4:01 PM
To: Dong, Eric <eric.dong@intel.com>
Cc: Ni, Ruiyu <ruiyu.ni@intel.com>; edk2-devel@lists.01.org; Yao, Jiewen <jiewen.yao@intel.com>
Subject: Re: [edk2] [Patch v3 2/3] UefiCpuPkg/S3Resume2Pei: Send S3 resume finished event to SmmCore.

Hi,

I'm sorry that I couldn't comment on this patch in time.

However, I must observe that version 1 of the patch,

http://mid.mail-archive.com/1507688554-10264-3-git-send-email-eric.dong@intel.com

was sent out at 04:22 AM on 11 Oct, in my time zone; and version 3 of the patch was committed at 04:17 AM on 12 Oct in my time zone.

That's not a lot of time for reviewers to comment -- I was very busy with other stuff yesterday. I even set up an auto-reply about that, asking for a bit of patience.

Of course the auto-reply didn't help, because I wasn't CC'd on the patch in the first place; despite my earlier request to be CC'd on all SMM- and/or S3-related patches.

So, after this introduction:

On 10/11/17 10:22, Eric Dong wrote:
> Driver will send S3 resume finished event to SmmCore through 
> communicate buffer after it signals EndOfPei event.
> 
> V2 Changes:
> 1. Change structures name to avoid they start with EFI_.
> 2. Base on DXE phase bits to provide communication buffer, current 
> implement check both PEI and DXE phase.
> 
> V3 Changes:
> 1. Change structure name for better understanding.
> 2. Enhance communication buffer calculate logic to more accurate.
> 
> Cc: Ruiyu Ni <ruiyu.ni@intel.com>
> Cc: Jiewen Yao <jiewen.yao@intel.com>
> Contributed-under: TianoCore Contribution Agreement 1.1
> Signed-off-by: Eric Dong <eric.dong@intel.com>
> ---
>  UefiCpuPkg/Universal/Acpi/S3Resume2Pei/S3Resume.c  | 87 ++++++++++++++++++++++
>  .../Universal/Acpi/S3Resume2Pei/S3Resume2Pei.inf   |  4 +
>  2 files changed, 91 insertions(+)
> 
> diff --git a/UefiCpuPkg/Universal/Acpi/S3Resume2Pei/S3Resume.c 
> b/UefiCpuPkg/Universal/Acpi/S3Resume2Pei/S3Resume.c
> index e53ed21..c2171cb 100644
> --- a/UefiCpuPkg/Universal/Acpi/S3Resume2Pei/S3Resume.c
> +++ b/UefiCpuPkg/Universal/Acpi/S3Resume2Pei/S3Resume.c
> @@ -28,6 +28,9 @@
>  #include <Ppi/SmmAccess.h>
>  #include <Ppi/PostBootScriptTable.h>
>  #include <Ppi/EndOfPeiPhase.h>
> +#include <Ppi/SmmCommunication.h>
> +
> +#include <Protocol/SmmEndOfS3Resume.h>
>  
>  #include <Library/DebugLib.h>
>  #include <Library/BaseLib.h>
> @@ -151,6 +154,22 @@ typedef union {
>    UINT64    Uint64;
>  } PAGE_TABLE_1G_ENTRY;
>  
> +//
> +// Define two type of smm communicate headers.
> +// One for 32 bits  PEI + 64 bits DXE, the other for 32 bits PEI + 32 bits DXE case.
> +//
> +typedef struct {
> +  EFI_GUID  HeaderGuid;
> +  UINT32    MessageLength;
> +  UINT8     Data[1];
> +} SMM_COMMUNICATE_HEADER_32;
> +
> +typedef struct {
> +  EFI_GUID  HeaderGuid;
> +  UINT64    MessageLength;
> +  UINT8     Data[1];
> +} SMM_COMMUNICATE_HEADER_64;
> +
>  #pragma pack()
>  
>  //
> @@ -430,6 +449,68 @@ IsLongModeWakingVector (  }
>  
>  /**
> +  Send EndOfS3Resume event to SmmCore through communication buffer way.
> +
> +  @retval  EFI_SUCCESS                 Return send the event success.
> +**/
> +EFI_STATUS
> +SignalEndOfS3Resume (
> +  VOID
> +  )
> +{
> +  EFI_STATUS                         Status;
> +  EFI_PEI_SMM_COMMUNICATION_PPI      *SmmCommunicationPpi;
> +  UINTN                              CommSize;
> +  SMM_COMMUNICATE_HEADER_32          Header32;
> +  SMM_COMMUNICATE_HEADER_64          Header64;
> +  VOID                               *CommBuffer;
> +
> +  DEBUG ((EFI_D_INFO, "SignalEndOfS3Resume - Enter\n"));
> +
> +  //
> +  // This buffer consumed in DXE phase, so base on DXE mode to prepare communicate buffer.
> +  // Detect whether DXE is 64 bits mode.
> +  // if (sizeof(UINTN) == sizeof(UINT64), PEI already 64 bits, assume DXE also 64 bits.
> +  // or (FeaturePcdGet (PcdDxeIplSwitchToLongMode)), Dxe will switch to 64 bits.
> +  //
> +  if ((sizeof(UINTN) == sizeof(UINT64)) || (FeaturePcdGet (PcdDxeIplSwitchToLongMode))) {
> +    CommBuffer = &Header64;
> +    Header64.MessageLength = 0;
> +    CommSize = OFFSET_OF (SMM_COMMUNICATE_HEADER_64, Data);  } else {
> +    CommBuffer = &Header32;
> +    Header32.MessageLength = 0;
> +    CommSize = OFFSET_OF (SMM_COMMUNICATE_HEADER_32, Data);  }  
> + CopyGuid (CommBuffer, &gEdkiiSmmEndOfS3ResumeProtocolGuid);
> +
> +  //
> +  // Get needed resource
> +  //
> +  Status = PeiServicesLocatePpi (
> +             &gEfiPeiSmmCommunicationPpiGuid,
> +             0,
> +             NULL,
> +             (VOID **)&SmmCommunicationPpi
> +             );
> +  ASSERT_EFI_ERROR (Status);

This patch breaks S3 resume with OVMF+SMM. OVMF does not have or install an EFI_PEI_SMM_COMMUNICATION_PPI. We discussed this ~2 years ago, when SMM was being added to KVM, QEMU, and OVMF.

OVMF would have needed EFI_PEI_SMM_COMMUNICATION_PPI only for SmmLockBoxPeiLib. But we agreed that SmmLockBoxPeiLib would work without this PPI; exactly the same as if the PPI existed, but Communicate() returned EFI_NOT_STARTED.

Namely, if RestoreLockBox() failed to locate EFI_PEI_SMM_COMMUNICATION_PPI, then InternalRestoreLockBoxFromSmram() would be called instead, which only needed PEI_SMM_ACCESS_PPI. And OVMF would provide that, so everything would work fine.

Please see the following edk2 commits:

  https://github.com/tianocore/edk2/commit/bd3afeb1d62c
  https://github.com/tianocore/edk2/commit/9d560947f6d3

Eric, can you please modify SignalEndOfS3Resume() as follows:

- locate EFI_PEI_SMM_COMMUNICATION_PPI as first action,
- if it is not found, return EFI_NOT_FOUND gracefully,
- in S3ResumeBootOs(), simply log the return value, and proceed.

Currently, during S3 resume (with SMM), I get

> SignalEndOfS3Resume - Enter

> ASSERT_EFI_ERROR (Status = Not Found)
> ASSERT UefiCpuPkg/Universal/Acpi/S3Resume2Pei/S3Resume.c(496):
!EFI_ERROR (Status)

Thanks,
Laszlo

> +
> +  //
> +  // Send command
> +  //
> +  Status = SmmCommunicationPpi->Communicate (
> +                                  SmmCommunicationPpi,
> +                                  (VOID *)CommBuffer,
> +                                  &CommSize
> +                                  );
> +  ASSERT_EFI_ERROR (Status);
> +
> +  DEBUG ((EFI_D_INFO, "SignalEndOfS3Resume - Exit (%r)\n", Status));
> +
> +  return Status;
> +}
> +
> +/**
>    Jump to OS waking vector.
>    The function will install boot script done PPI, report S3 resume status code, and then jump to OS waking vector.
>  
> @@ -504,6 +585,12 @@ S3ResumeBootOs (
>    ASSERT_EFI_ERROR (Status);
>  
>    //
> +  // Signal EndOfS3Resume event.
> +  //
> +  Status = SignalEndOfS3Resume ();
> +  ASSERT_EFI_ERROR (Status);
> +
> +  //
>    // report status code on S3 resume
>    //
>    REPORT_STATUS_CODE (EFI_PROGRESS_CODE, EFI_SOFTWARE_PEI_MODULE | 
> EFI_SW_PEI_PC_OS_WAKE); diff --git 
> a/UefiCpuPkg/Universal/Acpi/S3Resume2Pei/S3Resume2Pei.inf 
> b/UefiCpuPkg/Universal/Acpi/S3Resume2Pei/S3Resume2Pei.inf
> index d514523..943f114 100644
> --- a/UefiCpuPkg/Universal/Acpi/S3Resume2Pei/S3Resume2Pei.inf
> +++ b/UefiCpuPkg/Universal/Acpi/S3Resume2Pei/S3Resume2Pei.inf
> @@ -85,6 +85,10 @@
>    gPeiSmmAccessPpiGuid                          ## SOMETIMES_CONSUMES
>    gPeiPostScriptTablePpiGuid                    ## SOMETIMES_PRODUCES
>    gEfiEndOfPeiSignalPpiGuid                     ## SOMETIMES_PRODUCES
> +  gEfiPeiSmmCommunicationPpiGuid                ## SOMETIMES_CONSUMES
> +
> +[Protocols]
> +  gEdkiiSmmEndOfS3ResumeProtocolGuid            ## SOMETIMES_CONSUMES
>  
>  [FeaturePcd]
>    gEfiMdeModulePkgTokenSpaceGuid.PcdDxeIplSwitchToLongMode         ## CONSUMES
> 

_______________________________________________
edk2-devel mailing list
edk2-devel@lists.01.org
https://lists.01.org/mailman/listinfo/edk2-devel