On Sun, 19 Dec 2021 at 04:36, Nhi Pham <n...@os.amperecomputing.com> wrote:
>
> Hi Rebecca,
>
> Leif is merging the rest of Altra port to the edk2-platforms which has
> SRAT ACPI table consuming the CPU Core Info table. Therefore, we will
> need to fix the SRAT too. I would defer the fix until the Altra port is
> fully merged.
>

This seems to be stalled so to make progress, I am going to merge this.

> On 17/12/2021 05:07, Rebecca Cran wrote:
> > The ARM_CORE_INFO struct has been updated so the MPIDR is now a single
> > field instead of separate cluster/core fields. Update ArmPlatformLib.
> >
> > Signed-off-by: Rebecca Cran <rebe...@nuviainc.com>
> > ---
> >   Silicon/Ampere/AmpereAltraPkg/Library/ArmPlatformLib/ArmPlatformLib.c | 5 
> > ++---
> >   1 file changed, 2 insertions(+), 3 deletions(-)
> >
> > diff --git 
> > a/Silicon/Ampere/AmpereAltraPkg/Library/ArmPlatformLib/ArmPlatformLib.c 
> > b/Silicon/Ampere/AmpereAltraPkg/Library/ArmPlatformLib/ArmPlatformLib.c
> > index 5b4be0e55516..f2ec923d6f8d 100644
> > --- a/Silicon/Ampere/AmpereAltraPkg/Library/ArmPlatformLib/ArmPlatformLib.c
> > +++ b/Silicon/Ampere/AmpereAltraPkg/Library/ArmPlatformLib/ArmPlatformLib.c
> > @@ -108,9 +108,8 @@ PrePeiCoreGetMpCoreInfo (
> >       }
> >       SocketId = SOCKET_ID (Index);
> >       ClusterId = CLUSTER_ID (Index);
> > -    mArmPlatformMpCoreInfoTable[mArmPlatformCoreCount].ClusterId = 
> > SocketId;
> > -    mArmPlatformMpCoreInfoTable[mArmPlatformCoreCount].CoreId =
> > -      (ClusterId << 8) | (Index % PLATFORM_CPU_NUM_CORES_PER_CPM);
> > +    mArmPlatformMpCoreInfoTable[mArmPlatformCoreCount].Mpidr = GET_MPID (
> > +      SocketId, (ClusterId << 8) | (Index % 
> > PLATFORM_CPU_NUM_CORES_PER_CPM));
>
> For Ampere Altra, the correct MPIDR encoding is SocketId << 32 |
> ClusterId << 16 | (Index % PLATFORM_CPU_NUM_CORES_PER_CPM) << 8
>
> It would be the same what
> Platform/Ampere/JadePkg/Drivers/AcpiPlatformDxe/AcpiMadt.c (not
> available yet - being merged in) is describing.
>

Feel free to follow up with a patch that changes this into the correct
representation, but this patch does not make it less correct than it
already is; it just stores the socket ID in the cluster ID field in a
different way.


-=-=-=-=-=-=-=-=-=-=-=-
Groups.io Links: You receive all messages sent to this group.
View/Reply Online (#86225): https://edk2.groups.io/g/devel/message/86225
Mute This Topic: https://groups.io/mt/87777839/21656
Group Owner: devel+ow...@edk2.groups.io
Unsubscribe: https://edk2.groups.io/g/devel/unsub [arch...@mail-archive.com]
-=-=-=-=-=-=-=-=-=-=-=-


Reply via email to