Hi Leif,

Please find my response inline marked [SAMI].

Regards,

Sami Mujawar

On 25/09/2023 10:40 am, Leif Lindholm wrote:
On Fri, Sep 22, 2023 at 15:40:38 -0600, Swatisri Kantamsetti via groups.io 
wrote:
The IO Remapping Table, Platform Design Document, Revision E.e,
Sept 2022 (https://developer.arm.com/documentation/den0049/ee)
added flags in SMMUv3 node for validity of ID mappings for MSIs
related to control interrupts.
It makes one small addition to SMMUv3 nodes to
describe MSI support independently of wired GSIV support

Therefore, update the IORT generator to:
- increment IORT table revision to 6
- increment SMMUV3 node revision to 5
- for SMMUV3 node revision >=5 check the DeviceID mapping index
   valid flag to populate DeviceIdMappingIndex field

Signed-off-by: Swatisri Kantamsetti <swatis...@nvidia.com>
---
  .../Acpi/Arm/AcpiIortLibArm/IortGenerator.c   | 35 +++++++++++++++----
  1 file changed, 28 insertions(+), 7 deletions(-)

diff --git a/DynamicTablesPkg/Library/Acpi/Arm/AcpiIortLibArm/IortGenerator.c 
b/DynamicTablesPkg/Library/Acpi/Arm/AcpiIortLibArm/IortGenerator.c
index f28973c1a8..a6e4b49cb1 100644
--- a/DynamicTablesPkg/Library/Acpi/Arm/AcpiIortLibArm/IortGenerator.c
+++ b/DynamicTablesPkg/Library/Acpi/Arm/AcpiIortLibArm/IortGenerator.c
@@ -5,7 +5,7 @@
    SPDX-License-Identifier: BSD-2-Clause-Patent
@par Reference(s):
-  - IO Remapping Table, Platform Design Document, Revision E.d, Feb 2022
+  - IO Remapping Table, Platform Design Document, Revision E.e, Sept 2022
      (https://developer.arm.com/documentation/den0049/)
**/
@@ -1554,9 +1554,14 @@ AddSmmuV3Nodes (
      {
        SmmuV3Node->Node.Revision   = 2;
        SmmuV3Node->Node.Identifier = EFI_ACPI_RESERVED_DWORD;
-    } else {
+    } else if (AcpiTableInfo->AcpiTableRevision <
+               EFI_ACPI_IO_REMAPPING_TABLE_REVISION_06)
+    {
        SmmuV3Node->Node.Revision   = 4;
        SmmuV3Node->Node.Identifier = NodeList->Identifier;
+    } else {
+      SmmuV3Node->Node.Revision   = 5;
+      SmmuV3Node->Node.Identifier = NodeList->Identifier;
This is borderline, but I think it would be worth breaking out into a
helper function. It feels like this may have further cases added in
future.

      }
// SMMUv3 specific data
@@ -1577,11 +1582,27 @@ AddSmmuV3Nodes (
        SmmuV3Node->ProximityDomain = 0;
      }
- if ((SmmuV3Node->Event != 0) && (SmmuV3Node->Pri != 0) &&
-        (SmmuV3Node->Gerr != 0) && (SmmuV3Node->Sync != 0))
+    /* For older SMMUV3 nodes rev. < 5.
+       If all the SMMU control interrupts are GSIV based,
+       the DeviceID mapping index field is ignored.
+       DeviceID mapping valid flag was introduced in IORT rev E.e
+       for SMMUV3 nodes rev. > 5.
+       If the DeviceID mapping index valid flag is set to 0,
+       DeviceID mapping index field must be ignored.
+       Where the SMMU uses message signaled interrupts for
+       its control interrupts, DeviceId Mapping Index contains an
+       index into the array of ID mapping.
+     */
+    if (((SmmuV3Node->Node.Revision < 5) &&
+         (SmmuV3Node->Event != 0)        &&
+         (SmmuV3Node->Pri != 0)          &&
+         (SmmuV3Node->Gerr != 0)         &&
+         (SmmuV3Node->Sync != 0)
+         ) ||
+        ((SmmuV3Node->Node.Revision >= 5) &&
+         ((SmmuV3Node->Flags & EFI_ACPI_IORT_SMMUv3_FLAG_DEVICEID_VALID) == 0))
+        )
This is not borderline. I'm sure all of this is correct, but it is no
longer human readable. This type of information overload is better
kept out of a top-level function.
This also means we can give it an informative name.

Best Regards,

Leif

p.s.
Apologies Sami, but since I spotted this before it had been merged by
CI, I dropped the "push" tag since I wanted to make a comment.
You are the maintainer, and the above are my opinions only. If you
disagree, feel free to add it back and push as is.
(And feel free to do the same to push requests initiated by me in
future.)

[SAMI] I agree, this is becoming complex to read. I think it would be good to have helper functions to make the code clearer.

I have closed the pull request and look forward to a new patch from Swatisri.

[/SAMI]

      {
-      // If all the SMMU control interrupts are GSIV based,
-      // the DeviceID mapping index field is ignored.
        SmmuV3Node->DeviceIdMappingIndex = 0;
      } else {
        SmmuV3Node->DeviceIdMappingIndex = NodeList->DeviceIdMappingIndex;
@@ -2819,7 +2840,7 @@ ACPI_IORT_GENERATOR  IortGenerator = {
      // ACPI Table Signature
      EFI_ACPI_6_4_IO_REMAPPING_TABLE_SIGNATURE,
      // ACPI Table Revision supported by this Generator
-    EFI_ACPI_IO_REMAPPING_TABLE_REVISION_05,
+    EFI_ACPI_IO_REMAPPING_TABLE_REVISION_06,
      // Minimum supported ACPI Table Revision
      EFI_ACPI_IO_REMAPPING_TABLE_REVISION_00,
      // Creator ID
--
2.17.1








-=-=-=-=-=-=-=-=-=-=-=-
Groups.io Links: You receive all messages sent to this group.
View/Reply Online (#109040): https://edk2.groups.io/g/devel/message/109040
Mute This Topic: https://groups.io/mt/101535844/21656
Group Owner: devel+ow...@edk2.groups.io
Unsubscribe: https://edk2.groups.io/g/devel/unsub [arch...@mail-archive.com]
-=-=-=-=-=-=-=-=-=-=-=-


Reply via email to