Andrew Fish

On Aug 23, 2022, at 2:28 PM, Rohit Mathew <rohit.mathew@arm.com> wrote:

Hi Andrew,

From: Andrew Fish <afish@apple.com>
Sent: 23 August 2022 21:11
To: devel@edk2.groups.io; Rohit Mathew <Rohit.Mathew@arm.com>
Cc: username@nvidia.com; Sami Mujawar <Sami.Mujawar@arm.com>; Alexei Fedorov <Alexei.Fedorov@arm.com>; Mike Kinney <michael.d.kinney@intel.com>; gaoliming@byosoft.com.cn; zhiguang.liu@intel.com; Swatisri Kantamsetti <swatisrik@nvidia.com>; Thomas Abraham <thomas.abraham@arm.com>; Thanu Rangarajan <Thanu.Rangarajan@arm.com>; nd <nd@arm.com>
Subject: Re: [edk2-devel] [PATCH 1/2] Mde Pkg: Support for MPAM ACPI Table

On Aug 19, 2022, at 1:26 AM, Rohit Mathew <rohit.mathew@arm.com> wrote:

Hi Swatisri,

Thanks for the patch. Please find my comments inline marked [Rohit] -

-----Original Message-----
From: devel@edk2.groups.io <devel@edk2.groups.io> On Behalf Of Name
via groups.io
Sent: 16 August 2022 21:18
To: devel@edk2.groups.io; Sami Mujawar <Sami.Mujawar@arm.com>;
Alexei Fedorov <Alexei.Fedorov@arm.com>; michael.d.kinney@intel.com;
gaoliming@byosoft.com.cn; zhiguang.liu@intel.com
Cc: Swatisri Kantamsetti <swatisrik@nvidia.com>
Subject: [edk2-devel] [PATCH 1/2] Mde Pkg: Support for MPAM ACPI Table

From: Swatisri Kantamsetti <swatisrik@nvidia.com>

Added MPAM table header, MSC and Resource Node info structures

Signed-off-by: Swatisri Kantamsetti <swatisrik@nvidia.com>
---
MdePkg/Include/IndustryStandard/Acpi64.h | 5 ++
MdePkg/Include/IndustryStandard/Mpam.h   | 69
++++++++++++++++++++++++
2 files changed, 74 insertions(+)
create mode 100644 MdePkg/Include/IndustryStandard/Mpam.h

diff --git a/MdePkg/Include/IndustryStandard/Acpi64.h
b/MdePkg/Include/IndustryStandard/Acpi64.h
index fe5ebfac2b..e54f631186 100644
--- a/MdePkg/Include/IndustryStandard/Acpi64.h
+++ b/MdePkg/Include/IndustryStandard/Acpi64.h
@@ -2952,6 +2952,11 @@ typedef struct {
///
#define
EFI_ACPI_6_4_PROCESSOR_PROPERTIES_TOPOLOGY_TABLE_STRUCTURE_SI
GNATURE SIGNATURE_32('P', 'P', 'T', 'T')

+///
+/// "MPAM" Memory System Resource Partitioning And Monitoring Table
///
+#define
+EFI_ACPI_6_4_MEMORY_SYSTEM_RESOURCE_PARTITIONING_MONITORI
NG_TABLE_STRUC
+TURE_SIGNATURE SIGNATURE_32('M', 'P', 'A', 'M')
+
///
/// "PSDT" Persistent System Description Table /// diff --git
a/MdePkg/Include/IndustryStandard/Mpam.h
b/MdePkg/Include/IndustryStandard/Mpam.h
new file mode 100644
index 0000000000..e0f75f0114
--- /dev/null
+++ b/MdePkg/Include/IndustryStandard/Mpam.h
@@ -0,0 +1,69 @@
+/** @file
+ ACPI Memory System Resource Partitioning And Monitoring (MPAM)
+ as specified in ARM spec DEN0065
+
+ Copyright (c) 2022, NVIDIA CORPORATION. All rights reserved.
+ Copyright (c) 2022, ARM Limited. All rights reserved.
+ SPDX-License-Identifier: BSD-2-Clause-Patent **/
+
+#ifndef _MPAM_H_
+#define _MPAM_H_
+
+#pragma pack(1)
+
+///
+/// Memory System Resource Partitioning and Monitoring Table (MPAM)
///
+typedef struct {
+ EFI_ACPI_DESCRIPTION_HEADER    Header;
+ UINT32                         NumNodes;
+ UINT32                         NodeOffset;
+ UINT32                         Reserved;
+}
+EFI_ACPI_6_4_MEMORY_SYSTEM_RESOURCE_PARTITIONING_MONITORI
NG_TABLE_HEADE
+R;

[Rohit] Shouldn't the header be followed by MSC node body type as defined in MPAM ACPI 1.0, section 2, table 3 - The MPAM table and section 2.1, table 4 - MSC Node body?

+
+///
+/// MPAM Revision (as defined in ACPI 6.4 spec.) /// #define
+EFI_ACPI_6_4_MEMORY_SYSTEM_RESOURCE_PARTITIONING_MONITORI
NG_TABLE_REVIS
+ION 0x01
+
+///
+/// Memory System Controller Node Structure ///
+
+typedef struct {
+ UINT16    Length;
+ UINT16    Reserved;
+ UINT32    Identifier;
+ UINT64    BaseAddress;
+ UINT32    MmioSize;
+ UINT32    OverflowInterrupt;
+ UINT32    OverflowInterruptFlags;

[Rohit] Would it be better to have a type (possibly bitfield struct) instead of a plain UINT32 for Flags? (MPAM ACPI 1.0, section 2.1.1, table 5 - Interrupt flags)

Probably better NOT to use bitfields in APIs that are produced and consumed by different worlds. While the the UEFI does speak to the bit order of or a bitfield the rules around packing of bitfields is compiler defined.

I just saw a bug last week with bitfield compatibility that was introduced by clang fixing its -mms-bitfields implementation. The GCC rules for packing bitfields is different than VC++ so that is why the compiler flag -mms-bitfields exists in the 1st place . A clang -mms-bitfields bug got fixed and it broke the code as the extra padding required by VC++ got added to the bitfield.

[Rohit] Thanks for bringing this point. I think, this type could be left untouched in that case.

Thanks,

Andrew Fish

+ UINT32    Reserved1;
+ UINT32    OverflowInterruptAff;
+ UINT32    ErrorInterrupt;
+ UINT32    ErrorInterruptFlags;

[Rohit ] Same comment as before above.

+ UINT32    Reserved2;
+ UINT32    ErrorInterruptAff;
+ UINT32    MaxNRdyUsec;
+ UINT64    LinkedDeviceHwId;
+ UINT32    LinkedDeviceInstanceHwId;
+ UINT32    NumResourceNodes;
+} EFI_ACPI_6_4_MPAM_MSC_NODE;
+
+///
+/// Resource Node Structure
+///
+
+typedef struct {
+ UINT32    Identifier;
+ UINT8     RisIndex;
+ UINT16    Reserved1;
+ UINT8     LocatorType;
+ UINT64    Locator;

[Rohit ] Shouldn't " Locator " field be 12 bytes in size, possibly a separate type? (MPAM ACPI 1.0, section 2.2, table 7 - Resource node and section 2.3.2 table 10 - locator descriptor)

+ UINT32    NumFuncDep;
+} EFI_ACPI_6_4_MPAM_RESOURCE_NODE;

[Rohit] Since "NumFuncDep" field is part of EFI_ACPI_6_4_MPAM_RESOURCE_NODE type and this could be non-zero, should we also need the type for functional dependency descriptors? (MPAM ACPI 1.0, section 2.2.1, table 8 - Functional dependency descriptor)

[Rohit] Also, could some of the commonly used macros be added to this header, please? (location types, MPAM interrupt mode, interrupt types, affinity type, etc)

+
+#pragma pack()
+
+#endif
--
2.17.1

Regards,
Rohit