Re: [PATCH v6 6/6] OvmfPkg/AmdSevDxe: Add support for SEV live migration.

["Ashish Kalra" <ashish.kalra@amd.com>]

Hello Tom,

On Mon, Aug 09, 2021 at 09:29:29AM -0500, Tom Lendacky wrote:
> On 8/2/21 7:33 AM, Ashish Kalra wrote:
> > From: Ashish Kalra <ashish.kalra@amd.com>
> > 
> > Check for SEV live migration feature support, if detected
> > setup a new UEFI enviroment variable to indicate OVMF
> > support for SEV live migration.
> > 
> > The new runtime UEFI environment variable is set via the
> > notification function registered for the
> > EFI_END_OF_DXE_EVENT_GROUP_GUID event in AmdSevDxe driver.
> > 
> > AmdSevDxe module is an apriori driver so it gets loaded between PEI
> > and DXE phases and the SetVariable call will fail at the driver's
> > entry point as the Variable DXE module is still not loaded yet.
> > So we need to wait for an event notification which is signaled
> > after the Variable DXE module is loaded, hence, using the
> > EndOfDxe event notification to make this call.
> > 
> > Signed-off-by: Ashish Kalra <ashish.kalra@amd.com>
> > ---
> >  OvmfPkg/AmdSevDxe/AmdSevDxe.c              | 64 ++++++++++++++++++++
> >  OvmfPkg/AmdSevDxe/AmdSevDxe.inf            |  4 ++
> >  OvmfPkg/Include/Guid/AmdSevMemEncryptLib.h | 20 ++++++
> >  OvmfPkg/OvmfPkg.dec                        |  1 +
> >  4 files changed, 89 insertions(+)
> > 
> > diff --git a/OvmfPkg/AmdSevDxe/AmdSevDxe.c b/OvmfPkg/AmdSevDxe/AmdSevDxe.c
> > index c66c4e9b92..bfad71b9c6 100644
> > --- a/OvmfPkg/AmdSevDxe/AmdSevDxe.c
> > +++ b/OvmfPkg/AmdSevDxe/AmdSevDxe.c
> > @@ -15,10 +15,47 @@
> >  #include <Library/BaseMemoryLib.h>
> >  #include <Library/DebugLib.h>
> >  #include <Library/DxeServicesTableLib.h>
> > +#include <Library/UefiRuntimeServicesTableLib.h>
> > +#include <Library/UefiBootServicesTableLib.h>
> >  #include <Library/MemEncryptSevLib.h>
> >  #include <Library/MemoryAllocationLib.h>
> > +#include <Guid/AmdSevMemEncryptLib.h>
> > +#include <Guid/EventGroup.h>
> >  #include <Library/PcdLib.h>
> >  
> > +STATIC
> > +VOID
> > +EFIAPI
> > +AmdSevDxeOnEndOfDxe (
> > +  IN EFI_EVENT Event,
> > +  IN VOID      *EventToSignal
> > +  )
> > +{
> > +  EFI_STATUS Status;
> > +  BOOLEAN SevLiveMigrationEnabled;
> > +
> > +  SevLiveMigrationEnabled = MemEncryptSevLiveMigrationIsEnabled();
> > +
> > +  if (SevLiveMigrationEnabled) {
> > +    Status = gRT->SetVariable (
> > +               L"SevLiveMigrationEnabled",
> > +               &gAmdSevMemEncryptGuid,
> > +               EFI_VARIABLE_NON_VOLATILE |
> > +               EFI_VARIABLE_BOOTSERVICE_ACCESS |
> > +               EFI_VARIABLE_RUNTIME_ACCESS,
> > +               sizeof SevLiveMigrationEnabled,
> > +               &SevLiveMigrationEnabled
> > +               );
> > +
> > +    DEBUG ((
> > +      DEBUG_INFO,
> > +      "%a: Setting SevLiveMigrationEnabled variable, status = %lx\n",
> > +      __FUNCTION__,
> > +      Status
> > +      ));
> > +  }
> > +}
> > +
> >  EFI_STATUS
> >  EFIAPI
> >  AmdSevDxeEntryPoint (
> > @@ -30,6 +67,7 @@ AmdSevDxeEntryPoint (
> >    EFI_GCD_MEMORY_SPACE_DESCRIPTOR  *AllDescMap;
> >    UINTN                            NumEntries;
> >    UINTN                            Index;
> > +  EFI_EVENT                        Event;
> >  
> >    //
> >    // Do nothing when SEV is not enabled
> > @@ -130,5 +168,31 @@ AmdSevDxeEntryPoint (
> >      }
> >    }
> >  
> > +  //
> > +  // AmdSevDxe module is an apriori driver so it gets loaded between PEI
> > +  // and DXE phases and the SetVariable call will fail at the driver's
> > +  // entry point as the Variable DXE module is still not loaded yet.
> > +  // So we need to wait for an event notification which is signaled
> > +  // after the Variable DXE module is loaded, hence, using the
> > +  // EndOfDxe event notification to make this call.
> > +  //
> > +  // Register EFI_END_OF_DXE_EVENT_GROUP_GUID event.
> > +  // The notification function sets the runtime variable indicating OVMF
> > +  // support for SEV live migration.
> > +  //
> > +  Status = gBS->CreateEventEx (
> > +                  EVT_NOTIFY_SIGNAL,
> > +                  TPL_CALLBACK,
> > +                  AmdSevDxeOnEndOfDxe,
> > +                  NULL,
> > +                  &gEfiEndOfDxeEventGroupGuid,
> > +                  &Event
> > +                  );
> > +
> > +  if (EFI_ERROR (Status)) {
> > +      DEBUG ((DEBUG_INFO, "%a: CreateEventEx(): %r\n",
> 
> DEBUG_ERROR?
> 
> > +        __FUNCTION__, Status));
> 
> Should there be an "ASSERT_EFI_ERROR (Status)" after the DEBUG call?
> 

I don't think we should do an assert here and abort booting, failure
here will simply disable live migration support but i don't think that
it is such a fatal error that we should abort booting because of that.

OTOH, if there is a failure when doing page encryption status hypercalls
then aborting boot makes sense as guest page encryption status tracking
may be critical for multiple guest operations and not only live
migration.

Thanks,
Ashish