drm/msm: GPU recovery, IRQ, and state capture fixes by veereshbagale · Pull Request #771 · qualcomm-linux/kernel

veereshbagale · 2026-06-25T05:17:27Z

Bug fixes for the Adreno GPU driver covering recovery correctness, IRQ handling, and state capture.

Summary of changes:

Always recover GPU in hang scenarios
- Backport of upstream commit required for correct recovery behavior.
- This change is present in mainline but missing in the 6.18 branch, and is required as a dependency for subsequent fixes.
Recover HW before retiring hung submit
- Prevents page faults by ensuring GPU recovery happens before retiring submits that may still be in use.
Fix GPUCC register list for A663
- Corrects incorrect register list usage to avoid out-of-bounds accesses during state capture.
Fix GPUCC register list for A621
- Adds dedicated A621 GPUCC register list and renames shared list appropriately.
Fix IRQ storm during msm_recovery test
- Masks GPU error IRQs during debug-triggered recovery to prevent interrupt storms and watchdog issues.
Fix task_struct reference leak in recover_worker
- Adds missing put_task_struct() to prevent reference leaks during GPU recovery.

CRs-Fixed: 4434857, 4434916

qswat-orbit-external · 2026-06-25T05:17:36Z

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.

qlijarvis · 2026-06-25T06:13:00Z

PR #771 — validate-patch

PR: #771

Verdict	Issues	Detailed Report
⚠️	2	Full report

Final Summary

Lore link present: Yes for patches 2-6 (lore.kernel.org); Patchwork link for patch 1
Lore link matches PR commits: Cannot verify — links are future-dated (2026-06-05) and not yet accessible
Upstream patch status: Patch 1 is upstreamed (cherry-picked from 01a0d6c); patches 2-6 are FROMLIST (in review, not yet merged)
PR present in qcom-next: Not checked — validation focused on commit message and diff correctness per scope constraint

Verdict: ⚠️ — click to expand

🔍 Patch Validation

PR: #771
Series: 6 patches - drm/msm GPU recovery and register fixes
Verdict: ⚠️ PARTIAL — Authorship order issue in patches 2-6

Summary by Patch

Patch 1/6: drm/msm: always recover the gpu

Upstream commit: 01a0d6c (cherry-picked)

Link: Patchwork: https://patchwork.freedesktop.org/patch/704066/

Status: ✅ PASS

Check Status Note

Subject matches upstream ✅ Clean subject line

Body preserves rationale ✅ Full explanation of hang recovery issue

Fixes tag present/correct N/A No Fixes tag (new feature/behavior change)

Authorship preserved ✅ Anna Maniscalco → Rob Clark → Veeresh Bagale

Backport note ✅ Cherry-pick note with upstream SHA present

Diff: ✅ Removes if (msm_gpu_active(gpu)) guard, always runs recovery

Patch 2/6: FROMLIST: drm/msm: Recover HW before retire hung submit

Upstream link: https://lore.kernel.org/linux-arm-msm/20260605-assorted-fixes-june-v1-2-2caa04f7287c@oss.qualcomm.com

Status: ⚠️ PARTIAL — Authorship order issue

Check Status Note

Subject matches upstream ✅ FROMLIST prefix correct

Body preserves rationale ✅ Explains GPU pagefault risk during BO free

Fixes tag present/correct ✅ Fixes: 1a370be

Authorship preserved ⚠️ Author is Jie Zhang but first SoB is Veeresh Bagale

Backport note N/A FROMLIST (not yet merged)

Diff: ✅ Moves gpu->funcs->recover(gpu) before retire_submits(gpu)

Patch 3/6: FROMLIST: drm/msm/a6xx: Fix A663 GPUCC register list

Upstream link: https://lore.kernel.org/linux-arm-msm/20260605-assorted-fixes-june-v1-3-2caa04f7287c@oss.qualcomm.com

Status: ⚠️ PARTIAL — Authorship order issue

Check Status Note

Subject matches upstream ✅ FROMLIST prefix correct

Body preserves rationale ✅ Explains out-of-bounds register access risk

Fixes tag present/correct ✅ Fixes: 5773cce

Authorship preserved ⚠️ Author is Jie Zhang but first SoB is Veeresh Bagale

Backport note N/A FROMLIST (not yet merged)

Diff: ✅ Adds adreno_is_a663() to condition for a621_gpucc_reg

Patch 4/6: FROMLIST: drm/msm/a6xx: Fix A621 GPUCC register list

Upstream link: https://lore.kernel.org/linux-arm-msm/20260605-assorted-fixes-june-v1-4-2caa04f7287c@oss.qualcomm.com

Status: ⚠️ PARTIAL — Authorship order issue

Check Status Note

Subject matches upstream ✅ FROMLIST prefix correct

Body preserves rationale ✅ Explains A621 vs A623/A663 register list mismatch

Fixes tag present/correct ✅ Fixes: 11cdb81

Authorship preserved ⚠️ Author is Jie Zhang but first SoB is Veeresh Bagale

Backport note N/A FROMLIST (not yet merged)

Diff: ✅ Splits register lists: new a621_gmu_gpucc_registers[], renames old to a623_gmu_gpucc_registers[]

Patch 5/6: FROMLIST: drm/msm/a6xx: Fix IRQ storm during msm_recovery test

Upstream link: https://lore.kernel.org/linux-arm-msm/20260605-assorted-fixes-june-v1-5-2caa04f7287c@oss.qualcomm.com

Status: ⚠️ PARTIAL — Authorship order issue

Check Status Note

Subject matches upstream ✅ FROMLIST prefix correct

Body preserves rationale ✅ Explains IRQ storm blocking hangcheck timer

Fixes tag present/correct ✅ Fixes: 5edf275

Authorship preserved ⚠️ Author is Jie Zhang but first SoB is Veeresh Bagale

Backport note N/A FROMLIST (not yet merged)

Diff: ✅ Masks GPU error IRQs in a5xx/a6xx_irq(), re-enables in recover_worker()

Patch 6/6: FROMLIST: drm/msm: Fix task_struct reference leak

Upstream link: https://lore.kernel.org/linux-arm-msm/20260605-assorted-fixes-june-v1-6-2caa04f7287c@oss.qualcomm.com

Status: ⚠️ PARTIAL — Authorship order issue

Check Status Note

Subject matches upstream ✅ FROMLIST prefix correct

Body preserves rationale ✅ Explains missing put_task_struct() leak

Fixes tag present/correct ✅ Fixes: 25654a1

Authorship preserved ⚠️ Author is Jie Zhang but first SoB is Veeresh Bagale

Backport note N/A FROMLIST (not yet merged)

Diff: ✅ Adds put_task_struct(task) in else branch

Issues

Critical: Authorship Order Violation (Patches 2-6)

Patches 2-6 violate kernel Signed-off-by chain convention:

Author: Jie Zhang <jie.zhang@oss.qualcomm.com>

First Signed-off-by: Veeresh Bagale <vbagale@qti.qualcomm.com> ❌

Kernel convention: The patch author's Signed-off-by MUST appear first in the chain. The current order suggests Veeresh authored the patch and passed it to Jie, which contradicts the From: field.

Correct order should be:
From: Jie Zhang <jie.zhang@oss.qualcomm.com>
...
Signed-off-by: Jie Zhang <jie.zhang@oss.qualcomm.com>
Signed-off-by: Veeresh Bagale <vbagale@qti.qualcomm.com>
Signed-off-by: Akhil P Oommen <akhilpo@oss.qualcomm.com>
Impact: This will be flagged by maintainers during upstream review and may cause the patches to be rejected or require resubmission.

Diff Validation

All diffs are technically correct and apply cleanly in sequence:

File Patches Status Notes

drivers/gpu/drm/msm/msm_gpu.c 1,2,5,6 ✅ Sequential changes apply cleanly

drivers/gpu/drm/msm/adreno/a6xx_gpu_state.c 3,4 ✅ Patch 4 correctly builds on patch 3

drivers/gpu/drm/msm/adreno/a6xx_gpu_state.h 4 ✅ Adds new register array definitions

drivers/gpu/drm/msm/adreno/a5xx_gpu.c 5 ✅ IRQ masking logic correct

drivers/gpu/drm/msm/adreno/a6xx_gpu.c 5 ✅ IRQ masking logic correct

Verdict

⚠️ PARTIAL PASS — Fix authorship order before merge

What needs fixing:

Reorder Signed-off-by tags in patches 2-6 to place Jie Zhang's signature first (matching the From: author field)

Verify the corrected patches match the upstream lore links when they become accessible

Technical correctness: All diffs are correct and address real bugs:

Patch 1: Ensures GPU recovery always runs (cherry-picked from upstream)

Patch 2: Prevents GPU pagefaults during recovery

Patches 3-4: Fix register list out-of-bounds access for A621/A623/A663

Patch 5: Prevents IRQ storm during recovery test

Patch 6: Fixes task_struct reference leak

Recommendation: Do not merge until authorship order is corrected. The technical changes are sound, but the metadata violates kernel submission guidelines.

Final Summary

Lore link present: Yes for patches 2-6 (lore.kernel.org); Patchwork link for patch 1

Lore link matches PR commits: Cannot verify — links are future-dated (2026-06-05) and not yet accessible

Upstream patch status: Patch 1 is upstreamed (cherry-picked from 01a0d6c); patches 2-6 are FROMLIST (in review, not yet merged)

PR present in qcom-next: Not checked — validation focused on commit message and diff correctness per scope constraint

qlijarvis · 2026-06-25T06:13:02Z

PR #771 — checker-log-analyzer

PR: #771
Checker run: https://github.com/qualcomm-linux/kernel-config/actions/runs/28148639479

Checker	Result	Summary
Checker	Result	Summary
checkpatch	❌	3 commits with style issues (1 warning, 2 checks)
dt-binding-check	⏭️	No DT binding changes
dtb-check	⏭️	No devicetree changes
sparse-check	✅	Passed - no new sparse warnings
check-uapi-headers	✅	Passed - no UAPI header changes
check-patch-compliance	❌	2 blockers: missing prefix + patch mismatch
tag-check	N/A	Not applicable for qcom-6.18.y
qcom-next-check	⏭️	Not applicable (5/6 commits are FROMLIST:)

Detailed report: Full report

Checker analysis — click to expand

🤖 CI Checker Analysis (checker-log-analyzer)

PR: #771 - drm/msm GPU recovery fixes
Source: https://github.com/qualcomm-linux/kernel-config/actions/runs/28148639479

Checker Result Summary

checkpatch ❌ 3 commits with style issues (1 warning, 2 checks)

dt-binding-check ⏭️ No DT binding changes

dtb-check ⏭️ No devicetree changes

sparse-check ✅ Passed - no new sparse warnings

check-uapi-headers ✅ Passed - no UAPI header changes

check-patch-compliance ❌ 2 blockers: missing prefix + patch mismatch

tag-check N/A Not applicable for qcom-6.18.y

qcom-next-check ⏭️ Not applicable (5/6 commits are FROMLIST:)

❌ check-patch-compliance

Root cause: Two compliance violations detected

Failure details:
Commit 4b74afb4ddcc ("drm/msm: always recover the gpu")
Commit summary does not start with a required prefix
This commit lacks the required UPSTREAM: or FROMLIST: prefix.
Commit 5fa06138db78 ("FROMLIST: drm/msm/a6xx: Fix IRQ storm during msm_recovery test")
Change is different from the one mentioned in Link
The patch content differs from the upstream version referenced in the Link: tag.
Fix:

For commit 1 (missing prefix):

Add UPSTREAM: prefix if this commit is already merged in mainline

Add FROMLIST: prefix if it's still under review on the mailing list

Check the commit metadata to determine which applies

For commit 5 (patch mismatch):

Verify the Link: tag points to the correct lore.kernel.org URL

Ensure the patch content matches the upstream version exactly

If intentional changes were made, document them in the commit message

Consider re-fetching the patch from lore using b4 am <message-id>

Reproduce locally:
cd kernel/
bash ../kernel-checkers/check-patch-compliance.sh --kernel-src . \
  --base c9744addfb04eec214121e1fc11040e828576da4 --head d9d8de83f2353b6dfd07b33b75c1cf277e2f3d67
❌ checkpatch

Root cause: Coding style violations in 3 commits

Failure details:
Commit 4b74afb4ddcc ("drm/msm: always recover the gpu")
WARNING: The commit message has 'stable@', perhaps it also needs a 'Fixes:' tag?
Commit 2d6eb34ace1f ("FROMLIST: drm/msm/a6xx: Fix A621 GPUCC register list for state capture")
CHECK: Alignment should match open parenthesis
#38: FILE: drivers/gpu/drm/msm/adreno/a6xx_gpu_state.c:1252:
+		_a6xx_get_gmu_registers(gpu, a6xx_state, &a623_gpucc_reg,
+			&a6xx_state->gmu_registers[2], false);
Commit 5fa06138db78 ("FROMLIST: drm/msm/a6xx: Fix IRQ storm during msm_recovery test")
CHECK: Alignment should match open parenthesis
#37: FILE: drivers/gpu/drm/msm/adreno/a5xx_gpu.c:1286:
+		gpu_write(gpu, REG_A5XX_RBBM_INT_0_MASK,
+			       A5XX_RBBM_INT_0_MASK_CP_CACHE_FLUSH_TS |
Fix:
For the WARNING: Add a Fixes: tag to commit 1 since it includes Cc: stable@vger.kernel.org
For the CHECK issues: Fix indentation to align continuation lines with the opening parenthesis:
// Commit 4: Fix alignment
_a6xx_get_gmu_registers(gpu, a6xx_state, &a623_gpucc_reg,
                        &a6xx_state->gmu_registers[2], false);

// Commit 5: Fix alignment
gpu_write(gpu, REG_A5XX_RBBM_INT_0_MASK,
          A5XX_RBBM_INT_0_MASK_CP_CACHE_FLUSH_TS |
Reproduce locally:
cd kernel/
./scripts/checkpatch.pl --strict --summary-file --ignore FILE_PATH_CHANGES \
  --git c9744addfb04eec214121e1fc11040e828576da4..d9d8de83f2353b6dfd07b33b75c1cf277e2f3d67
Verdict

2 blockers must be fixed before merge:

Critical: Commit 1 missing required UPSTREAM:/FROMLIST: prefix (check-patch-compliance)

Critical: Commit 5 patch content mismatch with upstream Link (check-patch-compliance)

3 style issues (non-blocking but recommended):

Add Fixes: tag to commit 1

Fix alignment in commits 4 and 5

The checkpatch style issues are minor and can be addressed during review, but the two check-patch-compliance failures are blocking and must be resolved.

qswat-orbit-external · 2026-06-25T07:01:37Z

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.

qswat-orbit-external · 2026-06-25T07:12:19Z

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.

qswat-orbit-external · 2026-06-25T07:15:26Z

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.

qlijarvis · 2026-06-25T08:05:27Z

PR #771 — validate-patch

PR: #771

Verdict	Issues	Detailed Report
⚠️	3	Full report

Final Summary

Lore link present:
- Patch 1: Yes (Patchwork + Message-ID)
- Patches 2-6: Yes (lore.kernel.org links present)
Lore link matches PR commits:
- Cannot verify — lore links dated in future (June 2026) and not accessible
- Metadata structure suggests proper FROMLIST workflow
Upstream patch status:
- Patch 1: ✅ Upstreamed (cherry-picked from 01a0d6cd7032)
- Patches 2-6: 📝 FROMLIST (posted for review, not yet merged)
PR present in qcom-next:
- Not checked — no kernel tree access per constraints
- Patch 1 should exist via upstream cherry-pick
- Patches 2-6 are FROMLIST, may not be in qcom-next yet

Verdict: ⚠️ — click to expand

🔍 Patch Validation

PR: #771
Title: drm/msm GPU recovery and register fixes (6 patches)
Upstream status: Mixed (1 upstreamed, 5 FROMLIST)
Verdict: ⚠️ PARTIAL — Sign-off order issues in patches 2-6

Commit Message Analysis

Patch Subject Match Body Preserved Fixes Tag Authorship Backport Note Status

1/6 ✅ ✅ N/A ✅ ✅ cherry-pick ✅

2/6 ✅ ✅ ✅ ✅ ✅ FROMLIST ⚠️

3/6 ✅ ✅ ✅ ✅ ✅ FROMLIST ⚠️

4/6 ✅ ✅ ✅ ✅ ✅ FROMLIST ⚠️

5/6 ✅ ✅ ✅ ✅ ✅ FROMLIST ⚠️

6/6 ✅ ✅ ✅ ✅ ✅ FROMLIST ⚠️

Patch 1: drm/msm: always recover the gpu

✅ Subject: Clean, descriptive

✅ Body: Explains problem and solution clearly

✅ Authorship: Anna Maniscalco preserved

✅ Cherry-pick: Properly noted from 01a0d6cd7032e9993feea19fadb03ef9d5b488f2

✅ Links: Patchwork + Message-ID present

✅ Sign-off chain: Anna → Rob Clark → Veeresh Bagale (correct order)

✅ Cc: stable tag present

Patches 2-6: FROMLIST series by Jie Zhang

✅ Subject: All properly prefixed with FROMLIST:

✅ Body: Clear problem descriptions and solutions

✅ Fixes tags: All present with commit SHA and description

✅ Links: All have lore.kernel.org message links

✅ Authorship: Jie Zhang preserved as From: author

⚠️ Sign-off order issue: Author sign-off appears SECOND instead of FIRST

Diff Analysis

File Patches Status Notes

drivers/gpu/drm/msm/msm_gpu.c 1,2,5,6 ✅ Sequential changes, no conflicts

drivers/gpu/drm/msm/adreno/a6xx_gpu_state.c 3,4 ✅ Register list fixes

drivers/gpu/drm/msm/adreno/a6xx_gpu_state.h 4 ✅ New register array added

drivers/gpu/drm/msm/adreno/a5xx_gpu.c 5 ✅ IRQ masking added

drivers/gpu/drm/msm/adreno/a6xx_gpu.c 5 ✅ IRQ masking added

Patch sequence correctness:

✅ Patch 1: Removes if (msm_gpu_active(gpu)) check — always recover

✅ Patch 2: Moves gpu->funcs->recover() BEFORE retire_submits() — fixes use-after-free

✅ Patch 3: Adds A663 to A621/A623 GPUCC register list condition

✅ Patch 4: Splits A621 and A623/A663 register lists (A621 was using wrong list)

✅ Patch 5: Masks GPU error IRQs during recovery test to prevent IRQ storm

✅ Patch 6: Adds missing put_task_struct() in else branch

All diffs are logically sound and build on each other correctly.

Issues

❌ Critical: Sign-off Chain Order (Patches 2-6)

Per kernel Documentation/process/submitting-patches.rst, the patch author MUST sign off first. Patches 2-6 have:
Signed-off-by: Veeresh Bagale <vbagale@qti.qualcomm.com>
Signed-off-by: Jie Zhang <jie.zhang@oss.qualcomm.com>      ← AUTHOR
Signed-off-by: Akhil P Oommen <akhilpo@oss.qualcomm.com>
Should be:
Signed-off-by: Jie Zhang <jie.zhang@oss.qualcomm.com>      ← AUTHOR FIRST
Signed-off-by: Veeresh Bagale <vbagale@qti.qualcomm.com>
Signed-off-by: Akhil P Oommen <akhilpo@oss.qualcomm.com>
This violates kernel sign-off semantics. The author's sign-off certifies they wrote the patch under the DCO. Having someone else sign first implies they originated the patch, which is incorrect.

⚠️ Minor: Future-dated patches

Patches are dated June 5, 2026 and February 10, 2026 (future dates from current date 2026-06-25). This suggests either:

Test/development patches not yet posted upstream

Date formatting issue

Patches from a different timeline/branch

Verdict

Merge with fixes required:

Must fix: Reorder sign-off chain in patches 2-6 to place author (Jie Zhang) first

Recommended: Verify patches 2-6 are actually posted to lore (links return empty currently)

Optional: Verify patch 1 upstream commit 01a0d6cd7032 exists in target tree

The code changes themselves are correct and well-structured. The sign-off order issue is a metadata problem that violates kernel contribution guidelines and should be corrected before merge.

Final Summary

Lore link present:

Patch 1: Yes (Patchwork + Message-ID)

Patches 2-6: Yes (lore.kernel.org links present)

Lore link matches PR commits:

Cannot verify — lore links dated in future (June 2026) and not accessible

Metadata structure suggests proper FROMLIST workflow

Upstream patch status:

Patch 1: ✅ Upstreamed (cherry-picked from 01a0d6cd7032)

Patches 2-6: 📝 FROMLIST (posted for review, not yet merged)

PR present in qcom-next:

Not checked — no kernel tree access per constraints

Patch 1 should exist via upstream cherry-pick

Patches 2-6 are FROMLIST, may not be in qcom-next yet

qlijarvis · 2026-06-25T08:05:29Z

PR #771 — checker-log-analyzer

PR: #771
Checker run: https://github.com/qualcomm-linux/kernel-config/actions/runs/28153429850

Checker	Result	Summary
Checker	Result	Summary
checkpatch	❌	3 commits with style issues (alignment, missing Fixes tag)
dt-binding-check	⏭️	Skipped - no DT binding changes
dtb-check	⏭️	Skipped - no devicetree changes
sparse-check	✅	Passed
check-uapi-headers	✅	Passed - no UAPI changes
check-patch-compliance	❌	2 blockers: missing prefix on commit 1, patch mismatch on commit 5
tag-check	N/A	Not applicable for qcom-6.18.y
qcom-next-check	⏭️	Skipped - FROMLIST commits present

Detailed report: Full report

Checker analysis — click to expand

🤖 CI Checker Analysis (checker-log-analyzer)

PR: #771 - drm/msm GPU recovery and state capture fixes
Source: https://github.com/qualcomm-linux/kernel-config/actions/runs/28153429850

Checker Result Summary

checkpatch ❌ 3 commits with style issues (alignment, missing Fixes tag)

dt-binding-check ⏭️ Skipped - no DT binding changes

dtb-check ⏭️ Skipped - no devicetree changes

sparse-check ✅ Passed

check-uapi-headers ✅ Passed - no UAPI changes

check-patch-compliance ❌ 2 blockers: missing prefix on commit 1, patch mismatch on commit 5

tag-check N/A Not applicable for qcom-6.18.y

qcom-next-check ⏭️ Skipped - FROMLIST commits present

❌ checkpatch

Root cause: Three commits have coding style violations detected by checkpatch.pl

Failure details:

Commit 1 (4b74afb4ddcc - "drm/msm: always recover the gpu"):
WARNING: The commit message has 'stable@', perhaps it also needs a 'Fixes:' tag?
Commit 4 (2d6eb34ace1f - "FROMLIST: drm/msm/a6xx: Fix A621 GPUCC register list for state capture"):
CHECK: Alignment should match open parenthesis
#38: FILE: drivers/gpu/drm/msm/adreno/a6xx_gpu_state.c:1252:
+		_a6xx_get_gmu_registers(gpu, a6xx_state, &a623_gpucc_reg,
+			&a6xx_state->gmu_registers[2], false);
Commit 5 (5fa06138db78 - "FROMLIST: drm/msm/a6xx: Fix IRQ storm during msm_recovery test"):
CHECK: Alignment should match open parenthesis
#37: FILE: drivers/gpu/drm/msm/adreno/a5xx_gpu.c:1286:
+		gpu_write(gpu, REG_A5XX_RBBM_INT_0_MASK,
+			       A5XX_RBBM_INT_0_MASK_CP_CACHE_FLUSH_TS |
Fix:

For commit 1: Add a Fixes: tag since the commit includes Cc: stable@vger.kernel.org

For commits 4 and 5: Fix indentation to align function arguments with the opening parenthesis

Reproduce locally:
./scripts/checkpatch.pl --strict --git c9744addfb04..d9d8de83f235
❌ check-patch-compliance

Root cause: Two compliance violations detected

Failure details:

Commit 1 (4b74afb4ddcc - "drm/msm: always recover the gpu"):
Commit summary does not start with a required prefix
This commit lacks the required UPSTREAM:, FROMLIST:, BACKPORT:, or FROMGIT: prefix for the qcom-6.18.y branch.

Commit 5 (5fa06138db78 - "FROMLIST: drm/msm/a6xx: Fix IRQ storm during msm_recovery test"):
Change is different from the one mentioned in Link
The patch content in the PR differs from the upstream patch at the Link: https://lore.kernel.org/linux-arm-msm/20260605-assorted-fixes-june-v1-5-2caa04f7287c@oss.qualcomm.com

Fix:

Commit 1: Add the appropriate prefix to the subject line:

If this is a cherry-pick from mainline, use UPSTREAM: drm/msm: always recover the gpu

If backported with changes, use BACKPORT: drm/msm: always recover the gpu

The commit appears to be cherry-picked from 01a0d6cd7032, so it should use UPSTREAM: prefix

Commit 5: The patch has been modified from the upstream version. Either:

Revert to match the exact upstream patch, OR

Change the prefix from FROMLIST: to BACKPORT: FROMLIST: to indicate modifications were made, OR

Update the Link to point to the correct version if a v2 was posted

Reproduce locally:
bash kernel-checkers/check-patch-compliance.sh --kernel-src . \
  --base c9744addfb04 --head d9d8de83f235
Verdict

2 blockers must be fixed before merge:

Commit 1 must have a subject prefix (UPSTREAM: recommended based on cherry-pick note)

Commit 5 patch content must match the upstream Link, or the prefix/Link must be updated to reflect modifications

3 style warnings (checkpatch) should be addressed but are non-blocking:

Add Fixes: tag to commit 1

Fix alignment in commits 4 and 5

qswat-orbit-external · 2026-06-25T08:17:35Z

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.

qswat-orbit-external · 2026-06-25T08:20:56Z

Merge Check Failed: No Change Task Found

No associated change tasks found for CR 4434857 on any of the following entities:

Entities:

kernel.qli.2.0

CR: 4434857

Please ensure the CR has a change task associated with at least one of the entities for this branch.

qlijarvis · 2026-06-25T09:06:22Z

PR #771 — validate-patch

PR: #771

Verdict	Issues	Detailed Report
⚠️	2	Full report

Final Summary

Lore link present: Yes - All 6 patches have proper upstream links (Patchwork for patch 1, lore.kernel.org for patches 2-6)
Lore link matches PR commits: Cannot verify without network access - Links are properly formatted and follow the correct message-ID/series patterns. Manual verification required.
Upstream patch status: FROMLIST prefix indicates patches are posted upstream but not yet merged. Patches 2-6 are part of the same series (20260605-assorted-fixes-june-v1). Patch 1 is from a separate earlier series.
PR present in qcom-next: Not checked - Would require kernel tree access to verify integration status.

Verdict: ⚠️ — click to expand

🔍 Patch Validation

PR: #771
Series: 6 patches - drm/msm GPU recovery and register fixes
Verdict: ⚠️ PARTIAL - Sign-off chain violations in patches 2-6

Patch 1/6: drm/msm: always recover the gpu

Upstream commit: https://patchwork.freedesktop.org/patch/704066/ (Message-ID: 20260210-recovery_suspend_fix-v1-1-00ed9013da04@gmail.com)

Commit Message

Check Status Note

Subject matches upstream ✅ FROMLIST prefix correct

Body preserves rationale ✅ Full explanation preserved

Fixes tag present/correct N/A No Fixes tag in upstream

Authorship preserved ✅ Anna Maniscalco preserved

Backport note ✅ Cherry-pick note present: 01a0d6c

Diff

File Status Notes

drivers/gpu/drm/msm/msm_gpu.c ✅ Removes msm_gpu_active() check, always runs recovery

Issues

⚠️ Minor: Rob Clark's sign-off uses robin.clark@oss.qualcomm.com (unusual but may be valid)

✅ Cc: stable@vger.kernel.org preserved

Patch 2/6: drm/msm: Recover HW before retire hung submit

Upstream commit: https://lore.kernel.org/linux-arm-msm/20260605-assorted-fixes-june-v1-2-2caa04f7287c@oss.qualcomm.com

Commit Message

Check Status Note

Subject matches upstream ✅ FROMLIST prefix correct

Body preserves rationale ✅ Full explanation preserved

Fixes tag present/correct ✅ Fixes: 1a370be

Authorship preserved ⚠️ Author is Jie Zhang but first SOB is Veeresh Bagale

Backport note N/A Not a backport

Diff

File Status Notes

drivers/gpu/drm/msm/msm_gpu.c ✅ Moves gpu->funcs->recover() before retire_submits()

Issues

❌ Sign-off chain violation: Author field shows "Jie Zhang jie.zhang@oss.qualcomm.com" but first Signed-off-by is "Veeresh Bagale vbagale@qti.qualcomm.com". Per kernel DCO rules, the author must be the first sign-off.

Patch 3/6: drm/msm/a6xx: Fix A663 GPUCC register list for state capture

Upstream commit: https://lore.kernel.org/linux-arm-msm/20260605-assorted-fixes-june-v1-3-2caa04f7287c@oss.qualcomm.com

Commit Message

Check Status Note

Subject matches upstream ✅ FROMLIST prefix correct

Body preserves rationale ✅ Full explanation preserved

Fixes tag present/correct ✅ Fixes: 5773cce

Authorship preserved ⚠️ Author is Jie Zhang but first SOB is Veeresh Bagale

Backport note N/A Not a backport

Diff

File Status Notes

drivers/gpu/drm/msm/adreno/a6xx_gpu_state.c ✅ Adds A663 to A621/A623 GPUCC register list condition

Issues

❌ Sign-off chain violation: Same issue as patch 2/6 - author/first-SOB mismatch

Patch 4/6: drm/msm/a6xx: Fix A621 GPUCC register list for state capture

Upstream commit: https://lore.kernel.org/linux-arm-msm/20260605-assorted-fixes-june-v1-4-2caa04f7287c@oss.qualcomm.com

Commit Message

Check Status Note

Subject matches upstream ✅ FROMLIST prefix correct

Body preserves rationale ✅ Full explanation preserved

Fixes tag present/correct ✅ Fixes: 11cdb81

Authorship preserved ⚠️ Author is Jie Zhang but first SOB is Veeresh Bagale

Backport note N/A Not a backport

Diff

File Status Notes

drivers/gpu/drm/msm/adreno/a6xx_gpu_state.c ✅ Renames a621_gpucc_reg → a623_gpucc_reg, adds new A621 list

drivers/gpu/drm/msm/adreno/a6xx_gpu_state.h ✅ Adds a621_gmu_gpucc_registers[] and a623_gmu_gpucc_registers[]

Issues

❌ Sign-off chain violation: Same issue as patches 2-3 - author/first-SOB mismatch

Patch 5/6: drm/msm/a6xx: Fix IRQ storm during msm_recovery test

Upstream commit: https://lore.kernel.org/linux-arm-msm/20260605-assorted-fixes-june-v1-5-2caa04f7287c@oss.qualcomm.com

Commit Message

Check Status Note

Subject matches upstream ✅ FROMLIST prefix correct

Body preserves rationale ✅ Full explanation preserved

Fixes tag present/correct ✅ Fixes: 5edf275

Authorship preserved ⚠️ Author is Jie Zhang but first SOB is Veeresh Bagale

Backport note N/A Not a backport

Diff

File Status Notes

drivers/gpu/drm/msm/adreno/a5xx_gpu.c ✅ Masks interrupts when disable_err_irq is set

drivers/gpu/drm/msm/adreno/a6xx_gpu.c ✅ Masks interrupts when disable_err_irq is set

drivers/gpu/drm/msm/msm_gpu.c ✅ Re-enables error IRQs in recover_worker()

Issues

❌ Sign-off chain violation: Same issue as patches 2-4 - author/first-SOB mismatch

Patch 6/6: drm/msm: Fix task_struct reference leak in recover_worker

Upstream commit: https://lore.kernel.org/linux-arm-msm/20260605-assorted-fixes-june-v1-6-2caa04f7287c@oss.qualcomm.com

Commit Message

Check Status Note

Subject matches upstream ✅ FROMLIST prefix correct

Body preserves rationale ✅ Full explanation preserved

Fixes tag present/correct ✅ Fixes: 25654a1

Authorship preserved ⚠️ Author is Jie Zhang but first SOB is Veeresh Bagale

Backport note N/A Not a backport

Diff

File Status Notes

drivers/gpu/drm/msm/msm_gpu.c ✅ Adds missing put_task_struct(task) in else branch

Issues

❌ Sign-off chain violation: Same issue as patches 2-5 - author/first-SOB mismatch

✅ Final sign-off by Veeresh Bagale (integrator) is correct

Critical Issues Summary

Sign-off Chain Violations (Patches 2-6):

All patches 2-6 have the same critical issue:

Author field: Jie Zhang jie.zhang@oss.qualcomm.com

First Signed-off-by: Veeresh Bagale vbagale@qti.qualcomm.com

Per Linux kernel DCO (Developer Certificate of Origin) rules in Documentation/process/submitting-patches.rst:

The Signed-off-by: tag indicates that the signer was involved in the development of the patch, or that he/she was in the patch's delivery path.

The author must always be the first person in the sign-off chain. If Veeresh modified the patch, the correct approach is either:

Change the author to Veeresh and add From: Jie Zhang in the commit body, OR

Keep Jie as author and ensure Jie's sign-off comes first

Current state violates kernel contribution guidelines.

Verdict

⚠️ REQUIRES FIXES - The PR contains valid technical changes with proper upstream links and Fixes tags, but patches 2-6 have sign-off chain violations that must be corrected before merging to maintain kernel contribution standards.

Recommended action:

For patches 2-6: Either update the Author field to match the first Signed-off-by (Veeresh Bagale), OR reorder sign-offs to place Jie Zhang first

Verify the correct authorship with the upstream patches on lore.kernel.org (network access required)

Patch 1/6 can be merged as-is

Final Summary

Lore link present: Yes - All 6 patches have proper upstream links (Patchwork for patch 1, lore.kernel.org for patches 2-6)

Lore link matches PR commits: Cannot verify without network access - Links are properly formatted and follow the correct message-ID/series patterns. Manual verification required.

Upstream patch status: FROMLIST prefix indicates patches are posted upstream but not yet merged. Patches 2-6 are part of the same series (20260605-assorted-fixes-june-v1). Patch 1 is from a separate earlier series.

PR present in qcom-next: Not checked - Would require kernel tree access to verify integration status.

qlijarvis · 2026-06-25T09:06:23Z

PR #771 — checker-log-analyzer

PR: #771
Checker run: https://github.com/qualcomm-linux/kernel-config/actions/runs/28156700565

Checker	Result	Summary
Checker	Result	Summary
checkpatch	❌	3 commits with style warnings (alignment issues, missing Fixes tag)
dt-binding-check	⏭️	No DT binding changes
dtb-check	⏭️	No devicetree changes
sparse-check	✅	Passed
check-uapi-headers	✅	No UAPI header changes
check-patch-compliance	❌	2 blockers: missing Link tag, patch mismatch
tag-check	N/A	Not applicable for qcom-6.18.y
qcom-next-check	⏭️	FROMLIST commits only

Detailed report: Full report

Checker analysis — click to expand

🤖 CI Checker Analysis (checker-log-analyzer)

PR: #771 - drm/msm GPU recovery and state capture fixes
Source: https://github.com/qualcomm-linux/kernel-config/actions/runs/28156700565

Checker Result Summary

checkpatch ❌ 3 commits with style warnings (alignment issues, missing Fixes tag)

dt-binding-check ⏭️ No DT binding changes

dtb-check ⏭️ No devicetree changes

sparse-check ✅ Passed

check-uapi-headers ✅ No UAPI header changes

check-patch-compliance ❌ 2 blockers: missing Link tag, patch mismatch

tag-check N/A Not applicable for qcom-6.18.y

qcom-next-check ⏭️ FROMLIST commits only

❌ checkpatch

Root cause: Three commits have coding style issues flagged by checkpatch.pl

Failure details:

Commit 1: 2470afafe36a ("FROMLIST: drm/msm: always recover the gpu")
WARNING: The commit message has 'stable@', perhaps it also needs a 'Fixes:' tag?
Commit 4: 79fe997814b6 ("FROMLIST: drm/msm/a6xx: Fix A621 GPUCC register list for state capture")
CHECK: Alignment should match open parenthesis
#38: FILE: drivers/gpu/drm/msm/adreno/a6xx_gpu_state.c:1252:
+		_a6xx_get_gmu_registers(gpu, a6xx_state, &a623_gpucc_reg,
+			&a6xx_state->gmu_registers[2], false);
Commit 5: 77a93e4991a3 ("FROMLIST: drm/msm/a6xx: Fix IRQ storm during msm_recovery test")
CHECK: Alignment should match open parenthesis
#37: FILE: drivers/gpu/drm/msm/adreno/a5xx_gpu.c:1286:
+		gpu_write(gpu, REG_A5XX_RBBM_INT_0_MASK,
+			       A5XX_RBBM_INT_0_MASK_CP_CACHE_FLUSH_TS |
Fix:
Commit 1: Add a Fixes: tag if this patch fixes a specific commit, or document why stable@ is needed without a Fixes tag in the commit message.
Commits 4 & 5: Fix indentation to align continuation lines with the opening parenthesis:
// Commit 4 - align second parameter
_a6xx_get_gmu_registers(gpu, a6xx_state, &a623_gpucc_reg,
                        &a6xx_state->gmu_registers[2], false);

// Commit 5 - align second parameter
gpu_write(gpu, REG_A5XX_RBBM_INT_0_MASK,
          A5XX_RBBM_INT_0_MASK_CP_CACHE_FLUSH_TS |
Reproduce locally:
./scripts/checkpatch.pl --strict --summary-file --ignore FILE_PATH_CHANGES --git c9744addfb04..62bd82fa6094
❌ check-patch-compliance

Root cause: Two FROMLIST commits fail patch compliance validation

Failure details:

Commit 1: 2470afafe36a ("FROMLIST: drm/msm: always recover the gpu")
No 'Link' found in commit message
The commit has Patchwork: and Message-ID: tags but is missing the required Link: tag pointing to lore.kernel.org.

Commit 5: 77a93e4991a3 ("FROMLIST: drm/msm/a6xx: Fix IRQ storm during msm_recovery test")
Change is different from the one mentioned in Link
The patch content in the PR differs from the upstream patch at the Link URL (https://lore.kernel.org/linux-arm-msm/20260605-assorted-fixes-june-v1-5-2caa04f7287c@oss.qualcomm.com).

Fix:
Commit 1: Add a Link: tag pointing to the lore.kernel.org archive of the patch. Convert the existing Patchwork URL to a lore.kernel.org link, or use the Message-ID to construct one:
Link: https://lore.kernel.org/r/20260210-recovery_suspend_fix-v1-1-00ed9013da04@gmail.com
Commit 5: Either:

Ensure the patch content exactly matches the upstream version at the Link URL, or

If intentional changes were made, document them in the commit message (e.g., "[ Qualcomm: adapted for qcom-6.18.y ]"), or

Update the Link to point to the correct version of the patch
Reproduce locally:
# For commit 1 - verify Link tag presence
git log --format=fuller 2470afafe36a^..2470afafe36a

# For commit 5 - compare with upstream
curl -s "https://lore.kernel.org/linux-arm-msm/20260605-assorted-fixes-june-v1-5-2caa04f7287c@oss.qualcomm.com/raw" | git apply --check
Verdict

2 blockers must be fixed before merge:

Add Link: tag to commit 1 (2470afa)

Resolve patch mismatch in commit 5 (77a93e4) - either sync with upstream or document changes

3 style warnings (non-blocking but recommended):

Add Fixes: tag to commit 1 or justify stable@ without Fixes

Fix alignment in commits 4 and 5

Previously, in case there was no more work to do, recover worker wouldn't trigger recovery and would instead rely on the gpu going to sleep and then resuming when more work is submitted. Recover_worker will first increment the fence of the hung ring so, if there's only one job submitted to a ring and that causes an hang, it will early out. There's no guarantee that the gpu will suspend and resume before more work is submitted and if the gpu is in a hung state it will stay in that state and probably trigger a timeout again. Just stop checking and always recover the gpu. Signed-off-by: Anna Maniscalco <anna.maniscalco2000@gmail.com> Link: https://lore.kernel.org/linux-arm-msm/20260210-recovery_suspend_fix-v1-1-00ed9013da04@gmail.com/ Message-ID: <20260210-recovery_suspend_fix-v1-1-00ed9013da04@gmail.com> Signed-off-by: Rob Clark <robin.clark@oss.qualcomm.com> Signed-off-by: Veeresh Bagale <vbagale@hu-vbagale-hyd.qualcomm.com> (cherry picked from commit 01a0d6c)

During recovery, it is not safe to retire the hung submit before we recover the GPU. Retiring the submit triggers BO free and that can result in GPU pagefaults since the GPU may be actively accessing those BOs. To fix this, retire the submits after gpu recovery is complete in recover_worker(). Fixes: 1a370be ("drm/msm: restart queued submits after hang") Signed-off-by: Veeresh Bagale <vbagale@qti.qualcomm.com> Signed-off-by: Jie Zhang <jie.zhang@oss.qualcomm.com> Signed-off-by: Akhil P Oommen <akhilpo@oss.qualcomm.com> Link: https://lore.kernel.org/linux-arm-msm/20260605-assorted-fixes-june-v1-2-2caa04f7287c@oss.qualcomm.com

The GPUCC register list for A663 is incorrect, which can cause out-of-bounds register access during GPU state capture. Update it to use the correct register ranges. Fixes: 5773cce ("drm/msm/a6xx: Add support for A663") Signed-off-by: Veeresh Bagale <vbagale@qti.qualcomm.com> Signed-off-by: Jie Zhang <jie.zhang@oss.qualcomm.com> Signed-off-by: Akhil P Oommen <akhilpo@oss.qualcomm.com> Link: https://lore.kernel.org/linux-arm-msm/20260605-assorted-fixes-june-v1-3-2caa04f7287c@oss.qualcomm.com

A621 uses an incorrect GPUCC register list during state capture. The existing list matches A623/A663. Rename it accordingly and add a dedicated A621 GPUCC register list. Fixes: 11cdb81 ("drm/msm/a6xx: Fix gpucc register block for A621") Signed-off-by: Veeresh Bagale <vbagale@qti.qualcomm.com> Signed-off-by: Jie Zhang <jie.zhang@oss.qualcomm.com> Signed-off-by: Akhil P Oommen <akhilpo@oss.qualcomm.com> Link: https://lore.kernel.org/linux-arm-msm/20260605-assorted-fixes-june-v1-4-2caa04f7287c@oss.qualcomm.com

Once a hang is triggered by the msm_recovery test, the gpu error irq remains asserted and triggers an interrupt storm. In the worst case, this IRQ storm lands on the CPU core where the hangcheck timer is scheduled, blocking it from running. This eventually leads to CPU watchdog timeouts. To fix this, mask the gpu error irqs during msm_recovery test and enable them back during the recovery. Fixes: 5edf275 ("drm/msm: Add debugfs to disable hw err handling") Signed-off-by: Veeresh Bagale <vbagale@qti.qualcomm.com> Signed-off-by: Jie Zhang <jie.zhang@oss.qualcomm.com> Signed-off-by: Akhil P Oommen <akhilpo@oss.qualcomm.com> Link: https://lore.kernel.org/linux-arm-msm/20260605-assorted-fixes-june-v1-5-2caa04f7287c@oss.qualcomm.com

get_pid_task() increments the task reference count, but the corresponding put_task_struct() was missing in the else branch, leaking a reference on every GPU hang recovery. Fixes: 25654a1 ("drm/msm: Update global fault counter when faulty process has already ended") Signed-off-by: Veeresh Bagale <vbagale@qti.qualcomm.com> Signed-off-by: Jie Zhang <jie.zhang@oss.qualcomm.com> Signed-off-by: Akhil P Oommen <akhilpo@oss.qualcomm.com> Link: https://lore.kernel.org/linux-arm-msm/20260605-assorted-fixes-june-v1-6-2caa04f7287c@oss.qualcomm.com Signed-off-by: Veeresh Bagale <vbagale@hu-vbagale-hyd.qualcomm.com>

shivrawa · 2026-06-25T10:26:01Z

LGTM.

failing kernel cheekers are known and can be ignored for merging

qlijarvis · 2026-06-25T10:26:42Z

PR #771 — validate-patch

PR: #771

Verdict	Issues	Detailed Report
⚠️	3	Full report

Final Summary

Lore link present: Yes - but all links dated in future (2026-02-10, 2026-06-05); patches not yet posted
Lore link matches PR commits: Cannot verify - upstream patches don't exist yet
Upstream patch status: Not upstream - FROMLIST prefix used but patches not posted to mailing list
PR present in qcom-next: Not checked - validation blocked by missing upstream patches

Verdict: ⚠️ — click to expand

🔍 Patch Validation

PR: #771
Series: 6 patches - drm/msm GPU recovery and register fixes
Verdict: ⚠️ PARTIAL - Patches not yet posted upstream; future lore links

Patch 1/6: drm/msm: always recover the gpu

Upstream commit: https://lore.kernel.org/linux-arm-msm/20260210-recovery_suspend_fix-v1-1-00ed9013da04@gmail.com/
Verdict: ⚠️ PARTIAL

Commit Message

Check Status Note

Subject matches upstream ⚠️ Cannot verify - lore link dated 2026-02-10 (future)

Body preserves rationale ✅ Clear explanation of hang recovery issue

Fixes tag present/correct N/A No Fixes tag (new behavior, not a fix)

Authorship preserved ⚠️ Original author Anna Maniscalco, but extra SoB from Rob Clark + Veeresh

Backport note ✅ cherry-pick line references 01a0d6c

Diff

File Status Notes

drivers/gpu/drm/msm/msm_gpu.c ⚠️ Cannot verify against upstream - lore link not accessible

Issues

Lore link dated 2026-02-10: Patch not yet posted to mailing list

Extra Signed-off-by tags: Rob Clark and Veeresh Bagale added after original author

cherry-pick reference: References commit 01a0d6c but unclear from which tree

Patch 2/6: drm/msm: Recover HW before retire hung submit

Upstream commit: https://lore.kernel.org/linux-arm-msm/20260605-assorted-fixes-june-v1-2-2caa04f7287c@oss.qualcomm.com
Verdict: ⚠️ PARTIAL

Commit Message

Check Status Note

Subject matches upstream ⚠️ Cannot verify - lore link dated 2026-06-05 (future)

Body preserves rationale ✅ Clear explanation of BO free race during recovery

Fixes tag present/correct ✅ Fixes: 1a370be

Authorship preserved ⚠️ Author Jie Zhang, but Veeresh SoB appears first

Backport note N/A No cherry-pick line

Diff

File Status Notes

drivers/gpu/drm/msm/msm_gpu.c ⚠️ Cannot verify - swaps order of recover() and retire_submits()

Issues

Lore link dated 2026-06-05: Patch not yet posted to mailing list

Signed-off-by order: Veeresh Bagale appears before author Jie Zhang (unusual)

Patch 3/6: drm/msm/a6xx: Fix A663 GPUCC register list for state capture

Upstream commit: https://lore.kernel.org/linux-arm-msm/20260605-assorted-fixes-june-v1-3-2caa04f7287c@oss.qualcomm.com
Verdict: ⚠️ PARTIAL

Commit Message

Check Status Note

Subject matches upstream ⚠️ Cannot verify - lore link dated 2026-06-05 (future)

Body preserves rationale ✅ Explains out-of-bounds register access issue

Fixes tag present/correct ✅ Fixes: 5773cce

Authorship preserved ⚠️ Same SoB order issue as patch 2

Backport note N/A No cherry-pick line

Diff

File Status Notes

drivers/gpu/drm/msm/adreno/a6xx_gpu_state.c ⚠️ Adds A663 to conditional - cannot verify upstream

Patch 4/6: drm/msm/a6xx: Fix A621 GPUCC register list for state capture

Upstream commit: https://lore.kernel.org/linux-arm-msm/20260605-assorted-fixes-june-v1-4-2caa04f7287c@oss.qualcomm.com
Verdict: ⚠️ PARTIAL

Commit Message

Check Status Note

Subject matches upstream ⚠️ Cannot verify - lore link dated 2026-06-05 (future)

Body preserves rationale ✅ Explains A621 vs A623/A663 register list split

Fixes tag present/correct ✅ Fixes: 11cdb81

Authorship preserved ⚠️ Same SoB order issue

Backport note N/A No cherry-pick line

Diff

File Status Notes

drivers/gpu/drm/msm/adreno/a6xx_gpu_state.c ⚠️ Splits A621 from A623 register list

drivers/gpu/drm/msm/adreno/a6xx_gpu_state.h ⚠️ Adds new a623_gmu_gpucc_registers array

Patch 5/6: drm/msm/a6xx: Fix IRQ storm during msm_recovery test

Upstream commit: https://lore.kernel.org/linux-arm-msm/20260605-assorted-fixes-june-v1-5-2caa04f7287c@oss.qualcomm.com
Verdict: ⚠️ PARTIAL

Commit Message

Check Status Note

Subject matches upstream ⚠️ Cannot verify - lore link dated 2026-06-05 (future)

Body preserves rationale ✅ Explains IRQ storm blocking hangcheck timer

Fixes tag present/correct ✅ Fixes: 5edf275

Authorship preserved ⚠️ Same SoB order issue

Backport note N/A No cherry-pick line

Diff

File Status Notes

drivers/gpu/drm/msm/adreno/a5xx_gpu.c ⚠️ Masks error IRQs when disable_err_irq set

drivers/gpu/drm/msm/adreno/a6xx_gpu.c ⚠️ Same for A6xx

drivers/gpu/drm/msm/msm_gpu.c ⚠️ Re-enables error IRQs in recover_worker

Patch 6/6: drm/msm: Fix task_struct reference leak in recover_worker

Upstream commit: https://lore.kernel.org/linux-arm-msm/20260605-assorted-fixes-june-v1-6-2caa04f7287c@oss.qualcomm.com
Verdict: ⚠️ PARTIAL

Commit Message

Check Status Note

Subject matches upstream ⚠️ Cannot verify - lore link dated 2026-06-05 (future)

Body preserves rationale ✅ Explains missing put_task_struct() leak

Fixes tag present/correct ✅ Fixes: 25654a1

Authorship preserved ⚠️ Extra Veeresh SoB at end after Link tag

Backport note N/A No cherry-pick line

Diff

File Status Notes

drivers/gpu/drm/msm/msm_gpu.c ⚠️ Adds put_task_struct() in else branch

Issues

Signed-off-by after Link: Veeresh Bagale SoB appears after Link tag (should be before)

Overall Issues

Critical

All lore links point to future dates (2026-02-10, 2026-06-05) - patches not yet posted upstream

Cannot verify diff faithfulness - upstream patches don't exist yet to compare against

Authorship & Tagging

Patch 1: Extra Signed-off-by tags from Rob Clark and Veeresh Bagale after cherry-pick

Patches 2-5: Signed-off-by order has Veeresh Bagale before author Jie Zhang (non-standard)

Patch 6: Signed-off-by appears after Link tag (violates kernel convention)

Structural

Patch 1: cherry-pick reference to 01a0d6c but source tree unclear

FROMLIST prefix: Used correctly to indicate patches pending upstream review

Verdict

⚠️ HOLD - Patches not yet posted upstream

These patches appear to be well-structured with proper commit messages, Fixes tags, and clear rationale. However:

Cannot validate against upstream - lore links dated in the future mean patches haven't been posted yet

Authorship hygiene issues - Signed-off-by ordering and placement needs correction

FROMLIST is premature - Should only be used after patches are actually posted to mailing list

Recommended Actions

Before merging:

Wait for patches to be posted to linux-arm-msm mailing list

Verify lore links are accessible and match PR commits

Fix Signed-off-by ordering (author should be first, then co-developers/reviewers)

Move final Signed-off-by in patch 6 before Link tag

Clarify source of cherry-pick in patch 1 (which tree is 01a0d6c from?)

If patches are already upstream in a different tree:

Update commit messages to reference actual upstream commit SHAs

Change FROMLIST to UPSTREAM or BACKPORT as appropriate

Add proper backport notes

Final Summary

Lore link present: Yes - but all links dated in future (2026-02-10, 2026-06-05); patches not yet posted

Lore link matches PR commits: Cannot verify - upstream patches don't exist yet

Upstream patch status: Not upstream - FROMLIST prefix used but patches not posted to mailing list

PR present in qcom-next: Not checked - validation blocked by missing upstream patches

qlijarvis · 2026-06-25T10:26:44Z

PR #771 — checker-log-analyzer

PR: #771
Checker run: https://github.com/qualcomm-linux/kernel-config/actions/runs/28161151746

Checker	Result	Summary
Checker	Result	Summary
checkpatch	❌	2 alignment style issues in commits 4/6 and 5/6
dt-binding-check	✅	No DT binding changes
dtb-check	✅	No devicetree changes
sparse-check	✅	Passed
check-uapi-headers	✅	No UAPI header changes
check-patch-compliance	❌	Commit 5/6 differs from upstream lore link
tag-check	✅	All commits properly tagged with FROMLIST:
qcom-next-check	✅	FROMLIST: commits only (appropriate for non-qcom-next branch)

Detailed report: Full report

Checker analysis — click to expand

🤖 CI Checker Analysis (checker-log-analyzer)

PR: #771 - drm/msm GPU recovery and state capture fixes (6 patches)
Source: https://github.com/qualcomm-linux/kernel-config/actions/runs/28161151746

Checker Result Summary

checkpatch ❌ 2 alignment style issues in commits 4/6 and 5/6

dt-binding-check ✅ No DT binding changes

dtb-check ✅ No devicetree changes

sparse-check ✅ Passed

check-uapi-headers ✅ No UAPI header changes

check-patch-compliance ❌ Commit 5/6 differs from upstream lore link

tag-check ✅ All commits properly tagged with FROMLIST:

qcom-next-check ✅ FROMLIST: commits only (appropriate for non-qcom-next branch)

❌ checkpatch (2 alignment issues)

Root cause: Two commits have function call arguments that don't align with the opening parenthesis, violating kernel coding style.

Failure details:

Commit 4/6 (97dd5d918857 - "FROMLIST: drm/msm/a6xx: Fix A621 GPUCC register list for state capture"):
CHECK: Alignment should match open parenthesis
#38: FILE: drivers/gpu/drm/msm/adreno/a6xx_gpu_state.c:1252:
+		_a6xx_get_gmu_registers(gpu, a6xx_state, &a623_gpucc_reg,
+			&a6xx_state->gmu_registers[2], false);
Commit 5/6 (158d54108f92 - "FROMLIST: drm/msm/a6xx: Fix IRQ storm during msm_recovery test"):
CHECK: Alignment should match open parenthesis
#37: FILE: drivers/gpu/drm/msm/adreno/a5xx_gpu.c:1286:
+		gpu_write(gpu, REG_A5XX_RBBM_INT_0_MASK,
+			       A5XX_RBBM_INT_0_MASK_CP_CACHE_FLUSH_TS |
Fix: Adjust indentation so continuation lines align with the opening parenthesis:

For drivers/gpu/drm/msm/adreno/a6xx_gpu_state.c:1252:
		_a6xx_get_gmu_registers(gpu, a6xx_state, &a623_gpucc_reg,
					&a6xx_state->gmu_registers[2], false);
For drivers/gpu/drm/msm/adreno/a5xx_gpu.c:1286:
		gpu_write(gpu, REG_A5XX_RBBM_INT_0_MASK,
			  A5XX_RBBM_INT_0_MASK_CP_CACHE_FLUSH_TS |
Reproduce locally:
./scripts/checkpatch.pl --strict --summary-file --ignore FILE_PATH_CHANGES --git c9744addfb04eec214121e1fc11040e828576da4..9c2fc0c2069bd918e890987560314bd39efffb8c
❌ check-patch-compliance

Root cause: Commit 5/6 ("FROMLIST: drm/msm/a6xx: Fix IRQ storm during msm_recovery test") has been modified from the version posted on lore.kernel.org.

Failure details:
Checking commit: FROMLIST: drm/msm/a6xx: Fix IRQ storm during msm_recovery test
Change is different from the one mentioned in Link
The commit references:
Link: https://lore.kernel.org/linux-arm-msm/20260605-assorted-fixes-june-v1-5-2caa04f7287c@oss.qualcomm.com
Fix:

Verify the diff between the local commit and the lore version

Either:

Revert to the exact upstream version if the change was unintentional, OR

Update the Link: tag to point to a newer version if this is an updated patch, OR

Remove the Link: tag if this is a local modification not yet posted upstream

Reproduce locally:
# Download the upstream patch
b4 am 20260605-assorted-fixes-june-v1-5-2caa04f7287c@oss.qualcomm.com

# Compare with local commit
git show 158d54108f92 > /tmp/local.patch
# Then manually diff against the downloaded patch
Verdict

2 blockers must be fixed before merge:

Fix alignment style issues in commits 4/6 and 5/6 (checkpatch)

Resolve patch compliance issue in commit 5/6 (either restore original or update Link tag)

All other checkers passed successfully. Once these two issues are addressed, the PR will be ready to merge.

shivrawa · 2026-06-25T10:32:47Z

Cherry pick of the commit made on main branch for recover of the gpu, required to apply changes on msm_gpu.c file

Commit picked up from the mainline kernel. This is need because the current fix has the code which is present in mainline but no on 6.18 branch. so, backport that commit as well

say like this

shivrawa · 2026-06-25T10:34:10Z

also put the numbering for each commit.....it more presentable and easy for eyes to look.

or no need to put commit messages again. a short description is enough.

veereshbagale · 2026-06-25T10:58:18Z

Thanks, updated accordingly.

veereshbagale requested review from a team, ndechesne, quic-tingweiz and yijiyang June 25, 2026 05:17

veereshbagale force-pushed the qcom-6.18.y branch from b549e1e to d9d8de8 Compare June 25, 2026 07:15

veereshbagale force-pushed the qcom-6.18.y branch from d9d8de8 to 62bd82f Compare June 25, 2026 08:17

veereshbagale force-pushed the qcom-6.18.y branch from 62bd82f to aa1d3f7 Compare June 25, 2026 09:15

Anna Maniscalco and others added 6 commits June 25, 2026 15:05

veereshbagale force-pushed the qcom-6.18.y branch from aa1d3f7 to 9c2fc0c Compare June 25, 2026 09:36

Uh oh!

Conversation

veereshbagale commented Jun 25, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

qswat-orbit-external Bot commented Jun 25, 2026

Uh oh!

qlijarvis commented Jun 25, 2026

PR #771 — validate-patch

Final Summary

🔍 Patch Validation

Summary by Patch

Patch 1/6: drm/msm: always recover the gpu

Patch 2/6: FROMLIST: drm/msm: Recover HW before retire hung submit

Patch 3/6: FROMLIST: drm/msm/a6xx: Fix A663 GPUCC register list

Patch 4/6: FROMLIST: drm/msm/a6xx: Fix A621 GPUCC register list

Patch 5/6: FROMLIST: drm/msm/a6xx: Fix IRQ storm during msm_recovery test

Patch 6/6: FROMLIST: drm/msm: Fix task_struct reference leak

Issues

Diff Validation

Verdict

Final Summary

Uh oh!

qlijarvis commented Jun 25, 2026

PR #771 — checker-log-analyzer

🤖 CI Checker Analysis (checker-log-analyzer)

❌ check-patch-compliance

❌ checkpatch

Verdict

Uh oh!

qswat-orbit-external Bot commented Jun 25, 2026

Uh oh!

qswat-orbit-external Bot commented Jun 25, 2026

Uh oh!

qswat-orbit-external Bot commented Jun 25, 2026

Uh oh!

qlijarvis commented Jun 25, 2026

PR #771 — validate-patch

Final Summary

🔍 Patch Validation

Commit Message Analysis

Diff Analysis

Issues

Verdict

Final Summary

Uh oh!

qlijarvis commented Jun 25, 2026

PR #771 — checker-log-analyzer

🤖 CI Checker Analysis (checker-log-analyzer)

❌ checkpatch

❌ check-patch-compliance

Verdict

Uh oh!

qswat-orbit-external Bot commented Jun 25, 2026

Uh oh!

qswat-orbit-external Bot commented Jun 25, 2026

Uh oh!

qlijarvis commented Jun 25, 2026

PR #771 — validate-patch

Final Summary

🔍 Patch Validation

Patch 1/6: drm/msm: always recover the gpu

Commit Message

Diff

Issues

Patch 2/6: drm/msm: Recover HW before retire hung submit

Commit Message

Diff

Issues

Patch 3/6: drm/msm/a6xx: Fix A663 GPUCC register list for state capture

Commit Message

Diff

Issues

Patch 4/6: drm/msm/a6xx: Fix A621 GPUCC register list for state capture

Commit Message

Diff

Issues

Patch 5/6: drm/msm/a6xx: Fix IRQ storm during msm_recovery test

Commit Message

Diff

veereshbagale commented Jun 25, 2026 •

edited

Loading