shikra: Add support for GPU and Adreno_SMMU DT

[asherawa]

Add Shikra device tree support for the Adreno A704 GPU by adding the GPU, GMU wrapper, and Adreno SMMU nodes with required resources such as clocks, interconnects, IOMMU, OPP table, and zap-shader region.

Enable GPU support on CQM-EVK, CQS-EVK, and IQS-EVK boards, and add GPU thermal cooling to throttle GPU frequency at 115°C for thermal protection.

CRs-Fixed: 4569898

qli-2.0 GA Critical Fix
exception - https://jira-dc.qualcomm.com/jira/browse/QLIJIRA-130

[quic-ksanadhy]

shikra: Enable GPU (adreno smmu node + thermal + dt support) -> This title doesn't look good.
Make it something like Add support for GPU and Adreno_SMMU DT.

[qlijarvis]

PR #696 — validate-patch

PR: #696

Verdict	Issues	Detailed Report
⚠️	3	Full report

Final Summary

1. Lore link present: No — None of the three commits contain Link: tags to lore.kernel.org
2. Lore link matches PR commits: N/A — No lore links to compare against
3. Upstream patch status: Unknown — No upstream submission found via lore.kernel.org search for authors/subjects; likely vendor-only or not yet posted
4. PR present in qcom-next: Not checked — Would require access to qcom-next branch

Verdict: ⚠️ — click to expand

🔍 Patch Validation

PR: #696 - Add Shikra GPU support (3 commits)
Upstream commit: N/A (not found on lore.kernel.org)
Verdict: ⚠️ PARTIAL

---

Commit 1/3: arm64: dts: qcom: shikra: Add A704 GPU support

Commit Message

Check	Status	Note
Subject matches upstream	N/A	No upstream reference found
Body preserves rationale	✅	Clear description of what is added
Fixes tag present/correct	N/A	Not a fix
Authorship preserved	⚠️	Uses @qti.qualcomm.com (internal) instead of @oss.qualcomm.com
Backport note (if applicable)	N/A	Not a backport

Diff

File	Status	Notes
arch/arm64/boot/dts/qcom/shikra-cqm-evk.dts	✅	Enables GPU and sets zap-shader firmware
arch/arm64/boot/dts/qcom/shikra-cqs-evk.dts	✅	Enables GPU and sets zap-shader firmware
arch/arm64/boot/dts/qcom/shikra-iqs-evk.dts	✅	Enables GPU and sets zap-shader firmware
arch/arm64/boot/dts/qcom/shikra.dtsi	✅	Adds GPU and GMU wrapper nodes with proper bindings

Issues

• Author email uses internal domain (@qti.qualcomm.com) instead of upstream-friendly @oss.qualcomm.com
• No Link: tag to upstream submission
• Missing Signed-off-by chain if this was co-developed

---

Commit 2/3: arm64: dts: qcom: shikra: Add GPU cooling

Commit Message

Check	Status	Note
Subject matches upstream	N/A	No upstream reference found
Body preserves rationale	✅	Clear explanation of why GPU cooling is needed
Fixes tag present/correct	N/A	Not a fix
Authorship preserved	✅	Uses @oss.qualcomm.com
Backport note (if applicable)	N/A	Not a backport

Diff

File	Status	Notes
arch/arm64/boot/dts/qcom/shikra.dtsi	✅	Adds thermal throttling at 115°C with proper cooling-maps

Issues

• No Link: tag to upstream submission
• Depends on commit 1/3 (GPU node must exist first)

---

Commit 3/3: arm64: dts: qcom: shikra: Add Adreno SMMU node

Commit Message

Check	Status	Note
Subject matches upstream	N/A	No upstream reference found
Body preserves rationale	⚠️	Very brief; could explain SMMU purpose
Fixes tag present/correct	N/A	Not a fix
Authorship preserved	✅	Uses @oss.qualcomm.com; proper co-author chain
Backport note (if applicable)	N/A	Not a backport

Diff

File	Status	Notes
arch/arm64/boot/dts/qcom/shikra.dtsi	✅	Adds adreno_smmu node with proper compatible strings and clocks

Issues

• No Link: tag to upstream submission
• Commit message is minimal (one sentence body)
• Multiple Signed-off-by tags suggest co-development but no explanation in commit message

---

Overall Issues

1. Missing Link: tags - None of the three commits contain Link: tags pointing to lore.kernel.org or any upstream submission
2. Inconsistent email domains - Commit 1/3 uses @qti.qualcomm.com (internal) while 2/3 and 3/3 use @oss.qualcomm.com (upstream-friendly)
3. No upstream traceability - Cannot verify if these patches were posted upstream or if they match any upstream submission
4. Commit ordering concern - Commit 3/3 adds the SMMU node that commit 1/3 references (iommus = <&adreno_smmu 0 1>), but commit 3/3 comes after commit 1/3, which could cause bisection issues

---

Verdict

⚠️ PARTIAL PASS - Patches are technically correct and follow DT binding conventions, but lack upstream traceability and have authorship inconsistencies. The commit ordering may cause bisection issues (SMMU node added after GPU node that references it).

Recommended actions:

1. Reorder commits: SMMU node (3/3) should come before GPU node (1/3)
2. Add Link: tags if these were posted upstream
3. Use consistent email domain (@oss.qualcomm.com) for all authors
4. Expand commit message for 3/3 to explain SMMU purpose

---

Final Summary

1. Lore link present: No — None of the three commits contain Link: tags to lore.kernel.org
2. Lore link matches PR commits: N/A — No lore links to compare against
3. Upstream patch status: Unknown — No upstream submission found via lore.kernel.org search for authors/subjects; likely vendor-only or not yet posted
4. PR present in qcom-next: Not checked — Would require access to qcom-next branch

[qlijarvis]

PR #696 — checker-log-analyzer

PR: #696
Checker run: https://github.com/qualcomm-linux/kernel-config/actions/runs/27343479315

Checker	Result	Summary
Checker	Result	Summary
checkpatch	⚠️	1 warning: undocumented DT compatible string
dt-binding-check	⏭️	Skipped (no binding changes)
dtb-check	❌	GPU interrupt property has wrong cell count
sparse-check	⏭️	Skipped (no C code changes)
check-uapi-headers	⏭️	Skipped (no UAPI changes)
check-patch-compliance	❌	All 3 commits missing required subject prefix
tag-check	N/A	Not applicable for this branch
qcom-next-check	N/A	Not applicable for this branch

Detailed report: Full report

Checker analysis — click to expand

🤖 CI Checker Analysis (checker-log-analyzer)

PR: #696 - arm64: dts: qcom: shikra: Add A704 GPU support (3 commits)
Source: https://github.com/qualcomm-linux/kernel-config/actions/runs/27343479315

Checker	Result	Summary
checkpatch	⚠️	1 warning: undocumented DT compatible string
dt-binding-check	⏭️	Skipped (no binding changes)
dtb-check	❌	GPU interrupt property has wrong cell count
sparse-check	⏭️	Skipped (no C code changes)
check-uapi-headers	⏭️	Skipped (no UAPI changes)
check-patch-compliance	❌	All 3 commits missing required subject prefix
tag-check	N/A	Not applicable for this branch
qcom-next-check	N/A	Not applicable for this branch

---

❌ dtb-check

Root cause: GPU node interrupt property uses 3 cells per interrupt instead of required 4 cells

Failure details:

../arch/arm64/boot/dts/qcom/shikra.dtsi:2532.4-27: Warning (interrupts_property): 
/soc@0/gpu@5900000:#interrupt-cells: size is (12), expected multiple of 16

The GPU node at arch/arm64/boot/dts/qcom/shikra.dtsi line ~2236 defines:

interrupts = <GIC_SPI 177 IRQ_TYPE_LEVEL_HIGH>;

This uses only 3 cells (GIC_SPI, IRQ number, trigger type), but the GIC interrupt controller on this platform requires 4 cells per interrupt. The 4th cell is typically used for interrupt affinity/routing and should be set to 0 for default routing.

Fix: Add the 4th cell (set to 0) to the GPU interrupt property:

-			interrupts = <GIC_SPI 177 IRQ_TYPE_LEVEL_HIGH>;
+			interrupts = <GIC_SPI 177 IRQ_TYPE_LEVEL_HIGH 0>;

Reproduce locally:

make ARCH=arm64 CROSS_COMPILE=aarch64-linux-gnu- dtbs_check DT_SCHEMA_FILES=Documentation/devicetree/bindings/
# Or just build the specific DTB:
make ARCH=arm64 CROSS_COMPILE=aarch64-linux-gnu- arch/arm64/boot/dts/qcom/shikra-cqm-evk.dtb

---

❌ check-patch-compliance

Root cause: All 3 commits lack required subject prefix (UPSTREAM:/FROMLIST:/BACKPORT:/etc.)

Failure details:

Checking commit: arm64: dts: qcom: shikra: Add A704 GPU support
Commit summary does not start with a required prefix

Checking commit: arm64: dts: qcom: shikra: Add GPU cooling
Commit summary does not start with a required prefix

Checking commit: arm64: dts: qcom: shikra: Add Adreno SMMU node
Commit summary does not start with a required prefix

The Qualcomm Linux kernel requires commit subjects to be prefixed with tags indicating their origin:

• UPSTREAM: - Already merged in mainline
• FROMLIST: - Posted to upstream mailing list, pending acceptance
• BACKPORT: - Backported from a newer kernel version
• FROMGIT: - Cherry-picked from an upstream maintainer tree

Fix: Reword all 3 commit subjects to include the appropriate prefix. Since these appear to be new downstream patches for Shikra (not yet upstream), use FROMLIST: if they will be/have been posted upstream, or discuss with maintainers if a different tag is appropriate.

Example:

FROMLIST: arm64: dts: qcom: shikra: Add A704 GPU support
FROMLIST: arm64: dts: qcom: shikra: Add GPU cooling
FROMLIST: arm64: dts: qcom: shikra: Add Adreno SMMU node

Reproduce locally:

cd /path/to/kernel
git clone https://github.com/qualcomm-linux/kernel-checkers.git
bash kernel-checkers/check-patch-compliance.sh --kernel-src . --base <base-sha> --head <head-sha>

---

⚠️ checkpatch (non-blocking warning)

Root cause: DT compatible string "qcom,adreno-07000400" is not documented in bindings

Failure details:

WARNING: DT compatible string "qcom,adreno-07000400" appears un-documented 
-- check ./Documentation/devicetree/bindings/
#85: FILE: arch/arm64/boot/dts/qcom/shikra.dtsi:2236:
+			compatible = "qcom,adreno-07000400", "qcom,adreno";

Fix: This is typically a non-blocking warning if:

1. The binding documentation exists but checkpatch didn't find it (false positive), OR
2. The binding will be added in a separate patch/series

Verify that Documentation/devicetree/bindings/gpu/qcom,adreno.yaml (or similar) documents this compatible string. If not, add it to the binding documentation before or alongside this DTS change.

Reproduce locally:

./scripts/checkpatch.pl --strict --git <base-sha>..<head-sha>

---

Verdict

2 blockers must be fixed before merge:

1. dtb-check failure: Add 4th cell (0) to GPU interrupt property
2. check-patch-compliance failure: Add required subject prefix to all 3 commits (e.g., FROMLIST:)

The checkpatch warning about undocumented compatible string should be verified but is typically non-blocking if the binding exists or will be added separately.

[qswat-orbit-external]

Merge Check Failed: No CR Numbers Found

Error: No Change Request numbers were found.

Please add Change Request numbers to your pull request description in the format CRs-Fixed: 12345 or link GitHub issues that are associated with Change Requests.

[shashim-quic]

arm64: dts: qcom: shikra: Add A704 GPU support

add appropriate tag and Link: tag. Downstream patches need exception approval.

[qswat-orbit-external]

Merge Check Failed: CR Not Eligible for Merge

CR 4569898 is not eligible for merge.

The parent software image for kernel.qli.2.0 is not development complete.

Entity: kernel.qli.2.0
CR: 4569898
Reason: CR_CANNOT_MERGE

Please ensure the CR passes both CCT (ComponentChangeTasks) and ICT (Integration Change Tasks) validations.

[qswat-orbit-external]

Merge Check Failed: CR Not Eligible for Merge

CR 4569898 is not eligible for merge.

The parent software image for kernel.qli.2.0 is not development complete.

Entity: kernel.qli.2.0
CR: 4569898
Reason: CR_CANNOT_MERGE

Please ensure the CR passes both CCT (ComponentChangeTasks) and ICT (Integration Change Tasks) validations.

[qswat-orbit-external]

Merge Check Failed: CR Not Eligible for Merge

CR 4569898 is not eligible for merge.

The parent software image for kernel.qli.2.0 is not development complete.

Entity: kernel.qli.2.0
CR: 4569898
Reason: CR_CANNOT_MERGE

Please ensure the CR passes both CCT (ComponentChangeTasks) and ICT (Integration Change Tasks) validations.

[qlijarvis]

PR #696 — validate-patch

PR: #696

Verdict	Issues	Detailed Report
⚠️	2	Full report

Final Summary

1. Lore link present: No — PENDING: prefix indicates patches not yet posted upstream; no lore link expected or required at this stage

2. Lore link matches PR commits: N/A — no lore link to compare against; these are pre-submission PENDING patches

3. Upstream patch status: Not upstream — all three patches carry PENDING: prefix indicating work-in-progress not yet submitted to mailing lists

4. PR present in qcom-next: Not checked — PENDING patches are typically staged in topic branches before qcom-next integration

Recommendation: Fix the patch ordering issue (adreno_smmu definition before use) before merging. Once reordered, the series is suitable for internal integration and subsequent upstream submission.

Verdict: ⚠️ — click to expand

🔍 Patch Validation

PR: #696 - Add Shikra GPU support (3 patches)
Upstream commit: N/A (PENDING patches)
Verdict: ⚠️ PARTIAL

Patch 1/3: PENDING: arm64: dts: qcom: shikra: Add A704 GPU support

Commit Message

Check	Status	Note
Subject matches upstream	N/A	PENDING prefix - not yet upstream
Body preserves rationale	✅	Clear description of GPU/GMU nodes and board enablement
Fixes tag present/correct	N/A	New feature, no Fixes tag needed
Authorship preserved	✅	Aditya Sherawat asherawa@qti.qualcomm.com
Backport note (if applicable)	N/A	PENDING patch, not a backport

Diff

File	Status	Notes
arch/arm64/boot/dts/qcom/shikra-cqm-evk.dts:28	✅	Enables GPU and sets zap-shader firmware
arch/arm64/boot/dts/qcom/shikra-cqs-evk.dts:29	✅	Enables GPU and sets zap-shader firmware
arch/arm64/boot/dts/qcom/shikra-iqs-evk.dts:40	✅	Enables GPU and sets zap-shader firmware
arch/arm64/boot/dts/qcom/shikra.dtsi:2282	✅	Adds complete GPU node with OPP table and GMU wrapper

Issues

• No Link: tag to upstream submission (expected for PENDING patches)
• Missing Reviewed-by/Acked-by tags from maintainers

---

Patch 2/3: PENDING: arm64: dts: qcom: shikra: Add GPU cooling

Commit Message

Check	Status	Note
Subject matches upstream	N/A	PENDING prefix - not yet upstream
Body preserves rationale	✅	Explains GPU thermal throttling at 115°C
Fixes tag present/correct	N/A	New feature, no Fixes tag needed
Authorship preserved	✅	Aastha Pandey aastha.pandey@oss.qualcomm.com
Backport note (if applicable)	N/A	PENDING patch, not a backport

Diff

File	Status	Notes
arch/arm64/boot/dts/qcom/shikra.dtsi:2843	✅	Adds GPU thermal zone with passive cooling

Issues

• No Link: tag to upstream submission (expected for PENDING patches)
• Missing Reviewed-by/Acked-by tags from maintainers

---

Patch 3/3: PENDING: arm64: dts: qcom: shikra: Add Adreno SMMU node

Commit Message

Check	Status	Note
Subject matches upstream	N/A	PENDING prefix - not yet upstream
Body preserves rationale	⚠️	Very brief - could explain SMMU purpose/configuration
Fixes tag present/correct	N/A	New feature, no Fixes tag needed
Authorship preserved	✅	Komal Bajaj (author) + 2 co-authors with Signed-off-by
Backport note (if applicable)	N/A	PENDING patch, not a backport

Diff

File	Status	Notes
arch/arm64/boot/dts/qcom/shikra.dtsi:2393	✅	Adds Adreno SMMU with clocks and power domains

Issues

• Commit message is minimal - could benefit from more context about SMMU configuration
• No Link: tag to upstream submission (expected for PENDING patches)
• Missing Reviewed-by/Acked-by tags from maintainers

---

Overall Series Issues

Commit Message Quality:

• All three patches use PENDING: prefix correctly to indicate work-in-progress
• Patch 3/3 has minimal commit body (one sentence) - could be more descriptive
• No Link: tags pointing to lore.kernel.org (expected for PENDING patches)
• No upstream review tags (Reviewed-by, Acked-by, Tested-by)

Authorship & Tags:

• All patches have proper Signed-off-by tags
• Patch 3/3 correctly preserves co-author chain (Bibek → Imran → Komal)
• Email domains mix @qti.qualcomm.com and @oss.qualcomm.com (acceptable)

Technical Content:

• Patches form a logical series: GPU node → cooling → SMMU
• DTS syntax appears correct (compatible strings, reg, clocks, etc.)
• Patch 1 references &adreno_smmu before Patch 3 defines it (ordering issue)

Critical Ordering Issue:

• ❌ Patch 1/3 uses iommus = <&adreno_smmu 0 1>; at line 2312
• ❌ Patch 3/3 defines adreno_smmu: node
• ❌ This creates a forward reference that will break bisectability
• ✅ Fix: Patch 3 should come before Patch 1, or Patch 1 should not reference adreno_smmu until it's defined

---

Verdict

⚠️ PARTIAL PASS with ordering issue

The patches are technically sound but have a critical bisectability problem: Patch 1 references &adreno_smmu before Patch 3 defines it. This will cause DTC (device tree compiler) errors when building after Patch 1 alone.

Required fix: Reorder patches so adreno_smmu is defined before it's referenced, or remove the iommus property from Patch 1 and add it in a follow-up after Patch 3.

Minor improvements:

• Expand Patch 3/3 commit message with more context
• Add Link: tags when patches are posted upstream
• Obtain maintainer review tags before merging

---

Final Summary

1. Lore link present: No — PENDING: prefix indicates patches not yet posted upstream; no lore link expected or required at this stage

2. Lore link matches PR commits: N/A — no lore link to compare against; these are pre-submission PENDING patches

3. Upstream patch status: Not upstream — all three patches carry PENDING: prefix indicating work-in-progress not yet submitted to mailing lists

4. PR present in qcom-next: Not checked — PENDING patches are typically staged in topic branches before qcom-next integration

Recommendation: Fix the patch ordering issue (adreno_smmu definition before use) before merging. Once reordered, the series is suitable for internal integration and subsequent upstream submission.

[qlijarvis]

PR #696 — checker-log-analyzer

PR: #696
Checker run: https://github.com/qualcomm-linux/kernel-config/actions/runs/27437573961

Checker	Result	Summary
Checker	Result	Summary
checkpatch	⚠️	1 warning: undocumented DT compatible string
dt-binding-check	⏭️	Skipped - no binding changes
dtb-check	✅	Passed
sparse-check	⏭️	Skipped - no C code changes
check-uapi-headers	⏭️	Skipped - no UAPI changes
check-patch-compliance	❌	BLOCKER: All 3 commits use "PENDING:" prefix
tag-check	❌	BLOCKER: Subject prefix required (FROMLIST:/UPSTREAM:/BACKPORT:)
qcom-next-check	⚠️	Applies only to qcom-next branches

Detailed report: Full report

Checker analysis — click to expand

🤖 CI Checker Analysis (checker-log-analyzer)

PR: #696 - PENDING: arm64: dts: qcom: shikra: Add A704 GPU support (3 commits)
Source: https://github.com/qualcomm-linux/kernel-config/actions/runs/27437573961

Checker	Result	Summary
checkpatch	⚠️	1 warning: undocumented DT compatible string
dt-binding-check	⏭️	Skipped - no binding changes
dtb-check	✅	Passed
sparse-check	⏭️	Skipped - no C code changes
check-uapi-headers	⏭️	Skipped - no UAPI changes
check-patch-compliance	❌	BLOCKER: All 3 commits use "PENDING:" prefix
tag-check	❌	BLOCKER: Subject prefix required (FROMLIST:/UPSTREAM:/BACKPORT:)
qcom-next-check	⚠️	Applies only to qcom-next branches

---

❌ check-patch-compliance

Root cause: All three commits use the "PENDING:" prefix which is not a valid upstream-compliant prefix.

Failure details:

Checking commit: PENDING: arm64: dts: qcom: shikra: Add A704 GPU support
Commit summary does not start with a required prefix

Checking commit: PENDING: arm64: dts: qcom: shikra: Add GPU cooling
Commit summary does not start with a required prefix

Checking commit: PENDING: arm64: dts: qcom: shikra: Add Adreno SMMU node
Commit summary does not start with a required prefix

Fix: Replace "PENDING:" with one of the required prefixes:

• FROMLIST: if the patch has been posted to an upstream mailing list but not yet merged
• UPSTREAM: if the patch exists in Linus's tree or a maintainer tree
• BACKPORT: if backporting from a newer kernel version
• DOWNSTREAM: for Qualcomm-specific changes (use sparingly)

For these GPU devicetree additions, if they haven't been posted upstream yet, you should:

1. Post them to the linux-arm-msm mailing list first
2. Then use FROMLIST: prefix with a link to the lore.kernel.org posting

Example fix for commit 1:

FROMLIST: arm64: dts: qcom: shikra: Add A704 GPU support

Reproduce locally:

bash ../kernel-checkers/check-patch-compliance.sh --kernel-src . \
  --base 6ec3dd75be7c4b70733693d6b04a3e94b966f30b --head 369b3ad76136f9206fd49a2e3da1bcda07129240

---

⚠️ checkpatch

Root cause: The compatible string "qcom,adreno-07000400" is not documented in devicetree bindings.

Failure details:

WARNING: DT compatible string "qcom,adreno-07000400" appears un-documented -- check ./Documentation/devicetree/bindings/
#85: FILE: arch/arm64/boot/dts/qcom/shikra.dtsi:2286:
+			compatible = "qcom,adreno-07000400", "qcom,adreno";

Fix: This is a warning, not a blocker. However, for upstream submission you should:

1. Add the compatible string to Documentation/devicetree/bindings/display/msm/gpu.yaml (or the appropriate binding file)
2. Submit the binding update as a separate patch before the DTS changes

Note: This warning can be ignored for internal/downstream branches if the binding documentation will be added later, but it's required for upstream submission.

Reproduce locally:

./scripts/checkpatch.pl --strict --summary-file --ignore FILE_PATH_CHANGES \
  --git 6ec3dd75be7c4b70733693d6b04a3e94b966f30b..369b3ad76136f9206fd49a2e3da1bcda07129240

---

Verdict

2 blockers must be fixed before merge:

1. Critical: Remove "PENDING:" prefix from all 3 commits and replace with appropriate upstream-compliant prefix (FROMLIST:/UPSTREAM:/BACKPORT:/DOWNSTREAM:)
2. Recommended: Document the "qcom,adreno-07000400" compatible string in devicetree bindings (required for upstream, optional for internal branches)

Action items:

• Reword all commit subjects to use valid prefixes
• Consider posting patches upstream first, then use FROMLIST: prefix
• Add DT binding documentation for the new Adreno compatible string

[qswat-orbit-external]

Merge Check Failed: CR Not Eligible for Merge

CR 4569898 is not eligible for merge.

The parent software image for kernel.qli.2.0 is not development complete.

Entity: kernel.qli.2.0
CR: 4569898
Reason: CR_CANNOT_MERGE

Please ensure the CR passes both CCT (ComponentChangeTasks) and ICT (Integration Change Tasks) validations.