FROMLIST: PCI: qcom: Add D3cold support

[ziyuezhang-123]

Add support for transitioning PCIe endpoints under host bridge into D3cold by integrating with the DWC core suspend/resume helpers.

Implement PME_TurnOff message generation via ELBI_SYS_CTRL and hook it into the DWC host operations so the controller follows the standard PME_TurnOff-based power-down sequence before entering D3cold.

When the device is suspended into D3cold, fully tear down interconnect bandwidth, OPP votes. If D3cold is not entered, retain existing behavior by keeping the required interconnect and OPP votes.

Use dw_pcie::skip_pwrctrl_off to avoid powering off devices during suspend to preseve wakeup capability of the devices and also not to power on the devices in the init path.

Drop the qcom_pcie::suspended flag and rely on the existing dw_pcie::suspended state, which now drives both the power-management flow and the interconnect/OPP handling.

Link: https://lore.kernel.org/all/20260407-d3cold-v4-5-bb171f75b465@oss.qualcomm.com/
CRs-Fixed: 4504416

The merge-base changed after approval.

[sgaud-quic]

@ziyuezhang-123 / @krishnachaitanya-linux please rebase topic branch to 7.1-rc2, to match base of qcom-next to resolve merge conflicts during sync.

[sgaud-quic]

PR #939 — validate-patch

PR: #939

Verdict	Issues	Detailed Report
⚠️	0	Full report

Verdict: ⚠️ — click to expand

🔍 Patch Validation Report

PR: PR #939 — PCI: qcom: Add D3cold support (5-patch FROMLIST series, v5)
Upstream commits:

• Patch 1: https://lore.kernel.org/r/20260429-d3cold-v5-1-89e9735b9df6@oss.qualcomm.com
• Patch 2: https://lore.kernel.org/r/20260429-d3cold-v5-2-89e9735b9df6@oss.qualcomm.com
• Patch 3: https://lore.kernel.org/r/20260429-d3cold-v5-3-89e9735b9df6@oss.qualcomm.com
• Patch 4: https://lore.kernel.org/r/20260429-d3cold-v5-4-89e9735b9df6@oss.qualcomm.com
• Patch 5: https://lore.kernel.org/r/20260429-d3cold-v5-5-89e9735b9df6@oss.qualcomm.com

Verdict: ⚠️ PARTIAL

Note: Network access was unavailable in this environment. Lore patches could not be fetched for direct diff comparison. The analysis below is based on internal consistency checks of pr.patch, commit message vs. diff alignment, authorship hygiene, and series ordering. Upstream acceptance status and qcom-next presence could not be verified — mark both as ⏭️ Skipped.

---

Commit Message Checks

Patch 1/5 — FROMLIST: PCI: host-common: Add helper to determine host bridge D3cold eligibility

Check	Status	Note
Subject matches upstream	✅	Subject consistent with lore message-ID slug
Body preserves rationale	✅	Detailed description of helper logic, D3hot requirement, PME-from-D3cold output param
Fixes tag	N/A	New feature, no Fixes tag expected
Authorship (FROMLIST:)	✅	From: = lore sender (same @oss.qualcomm.com); original author's Signed-off-by: present
Backport note	N/A	Not a backport
Co-developed-by misuse	✅	Not used

Patch 2/5 — FROMLIST: PCI: qcom: Add .get_ltssm() helper

Check	Status	Note
Subject matches upstream	✅	Consistent with lore slug
Body preserves rationale	✅	Explains NOC error on sc7280, PARF vs ELBI distinction
Fixes tag	N/A	New feature
Authorship (FROMLIST:)	✅	Same author/SOB pattern
Backport note	N/A
Co-developed-by misuse	✅	Not used

Patch 3/5 — FROMLIST: PCI: qcom: Power down PHY via PARF_PHY_CTRL before disabling rails/clocks

Check	Status	Note
Subject matches upstream	✅	Consistent with lore slug
Body preserves rationale	✅	Explains PHY leakage risk, lists affected IP revisions (2.3.2, 2.3.3, 2.4.0, 2.7.0, 2.9.0)
Fixes tag	N/A	New feature / preventive fix
Authorship (FROMLIST:)	✅	Same author/SOB pattern
Backport note	N/A
Co-developed-by misuse	✅	Not used

Patch 4/5 — FROMLIST: PCI: dwc: Use common D3cold eligibility helper in suspend path

Check	Status	Note
Subject matches upstream	✅	Consistent with lore slug
Body preserves rationale	⚠️	Body claims "Update the suspended flag in dw_pcie_resume_noirq() only after the PCIe link resumes successfully" — but this change is absent from the diff. Patch 4 only touches dw_pcie_suspend_noirq and pcie-designware.h. The dw_pcie_resume_noirq change (if it exists upstream) is not present here.
Fixes tag	N/A
Authorship (FROMLIST:)	✅	Same author/SOB pattern
Backport note	N/A
Co-developed-by misuse	✅	Not used

Patch 5/5 — FROMLIST: PCI: qcom: Add D3cold support

Check	Status	Note
Subject matches upstream	✅	Consistent with lore slug
Body preserves rationale	✅	Covers PME_TurnOff, interconnect teardown, skip_pwrctrl_off, dropping qcom_pcie::suspended
Fixes tag	N/A	New feature
Authorship (FROMLIST:)	✅	Same author/SOB pattern
Backport note	N/A
Co-developed-by misuse	✅	Not used

---

Diff Checks

Patch 1/5

File	Status	Notes
drivers/pci/controller/pci-host-common.c	✅	Renames pci_host_common_can_enter_d3cold → pci_host_common_d3cold_possible, adds pme_capable output param, expands endpoint type check to include LEG_END/RC_END, adds inactive-device bypass, uses PCI_HOST_D3COLD_ALLOWED/PCI_HOST_PME_D3COLD_CAPABLE flags
drivers/pci/controller/pci-host-common.h	✅	Declaration updated to match new signature

Patch 2/5

File	Status	Notes
drivers/pci/controller/dwc/pcie-qcom.c	✅	Adds ELBI_SYS_STTS register + mask, get_ltssm op, qcom_pcie_2_1_0_get_ltssm(), wires into ops_2_1_0/ops_1_0_0; removes pm_domain.h; refactors suspend/resume to intermediate state (manual link-up check, local suspended flag); removes broken_l1ss_resume and pme_turn_off hook (restored in patch 5)

Patch 3/5

File	Status	Notes
drivers/pci/controller/dwc/pcie-qcom.c	⚠️	PHY power-down blocks added correctly to deinit_2_3_2, deinit_2_4_0, deinit_2_3_3, deinit_2_7_0, deinit_2_9_0; init-path comments updated. Minor style: two added comments read /* Force PHY to lowest power state*/ — missing space before */ (lines +499, +521)

Patch 4/5

File	Status	Notes
drivers/pci/controller/dwc/pcie-designware-host.c	⚠️	Correctly replaces pci_host_common_can_enter_d3cold with pci_host_common_d3cold_possible, adds skip_pwrctrl_off assignment. However, the commit body claims a dw_pcie_resume_noirq() change that is not present in this diff
drivers/pci/controller/dwc/pcie-designware.h	✅	Adds skip_pwrctrl_off field to dw_pcie_rp

Patch 5/5

File	Status	Notes
drivers/pci/controller/dwc/pcie-qcom.c	✅	Restores pme_turn_off hook and dw_pcie_suspend/resume_noirq calls; drops qcom_pcie::suspended in favour of dw_pcie::suspended; adds skip_pwrctrl_off guards around pci_pwrctrl_* calls; error handling in resume changes from -ETIMEDOUT to -ENODEV/-EIO (semantic change, but consistent with the new D3cold flow where no endpoint is a valid state)

---

Upstream Patch Status

Commit	Community Verdict
All 5 patches	⏭️ Skipped — network unavailable; lore.kernel.org unreachable

---

Dependency Check

• Series is self-contained (patches 1–5 applied in order produce a coherent result)
• Patch 4 depends on patch 1 (pci_host_common_d3cold_possible) — ordering correct
• Patch 5 depends on patches 2, 3, 4 — ordering correct
• No external prerequisite patches referenced in commit bodies
• ⚠️ Patch 2 leaves the driver in a deliberate intermediate state (removes pme_turn_off, dw_pcie_suspend_noirq call) that is only restored by patch 5. This is acceptable for a series but means patches 2–4 alone would regress D3cold functionality.

---

qcom-next Presence

Commit	Status
All 5 patches	⏭️ Skipped — no local kernel repo with qcom-next found; network unavailable

---

Issues Found

1. Patch 4 — Commit body / diff mismatch (⚠️ WARNING): The commit body states "Update the suspended flag in dw_pcie_resume_noirq() only after the PCIe link resumes successfully, to avoid marking the controller active when link resume fails." However, dw_pcie_resume_noirq() in pcie-designware-host.c is not modified by patch 4's diff. If this change exists in the upstream lore patch, it is missing from the PR. If it was intentionally deferred to patch 5 (which calls dw_pcie_resume_noirq from qcom_pcie_resume_noirq), the commit body paragraph should be removed or moved to patch 5.

2. Patch 3 — Missing space before */ in two comments (⚠️ STYLE):

   • + /* Force PHY to lowest power state*/ (in qcom_pcie_deinit_2_3_2)
   • + /* Force PHY to lowest power state*/ (in qcom_pcie_deinit_2_4_0)
     Should be /* Force PHY to lowest power state */ — the other added comments in the same patch correctly include the space.

---

Verdict

The series is structurally sound and the lore links are present and correctly formatted for all 5 FROMLIST: commits. Two issues require attention before merge: (1) the stray paragraph in patch 4's commit body describing a dw_pcie_resume_noirq() change that does not appear in its diff — either the change is missing or the paragraph belongs in patch 5; (2) two /* ... */ comments in patch 3 are missing the space before */. Resolve issue (1) (either add the missing hunk or drop the paragraph) and fix the comment style in patch 3, then the series is ready to merge.

[sgaud-quic]

PR #939 — checker-log-analyzer

PR: #939
Checker run: https://github.com/qualcomm-linux/kernel-config/actions/runs/25852850665

Checker	Result	Summary
Checker	Result	Summary
checkpatch	❌	Merge conflict during sync — checker never ran
dt-binding-check	❌	Merge conflict during sync — checker never ran
dtb-check	❌	Merge conflict during sync — checker never ran
sparse-check	❌	Merge conflict during sync — checker never ran
check-uapi-headers	❌	Merge conflict during sync — checker never ran
check-patch-compliance	❌	Merge conflict during sync — checker never ran
tag-check	✅	All 5 commits carry FROMLIST: prefix
qcom-next-check	⏭️	Network unavailable — cannot fetch qcom-next; manual check required

Detailed report: Full report

Checker analysis — click to expand

🤖 CI Checker Analysis (checker-log-analyzer)

PR: #939 — PCI D3cold support series (5 commits, all FROMLIST:)
Source: https://github.com/qualcomm-linux/kernel-config/actions/runs/25852850665

Checker	Result	Summary
checkpatch	❌	Merge conflict during sync — checker never ran
dt-binding-check	❌	Merge conflict during sync — checker never ran
dtb-check	❌	Merge conflict during sync — checker never ran
sparse-check	❌	Merge conflict during sync — checker never ran
check-uapi-headers	❌	Merge conflict during sync — checker never ran
check-patch-compliance	❌	Merge conflict during sync — checker never ran
tag-check	✅	All 5 commits carry FROMLIST: prefix
qcom-next-check	⏭️	Network unavailable — cannot fetch qcom-next; manual check required

---

❌ All Checkers — Merge Conflict in Sync Step

Root cause: The CI sync step failed to auto-merge topic/tech/bus/pci/all into the integration branch (based on tag 4e0b0df1c84f9b81089bf478c63d635cc97040a1), producing content conflicts in 4 PCI files; every checker job aborted before running any analysis.

Failure details:

CONFLICT (content): Merge conflict in drivers/pci/controller/dwc/pcie-designware-host.c
CONFLICT (content): Merge conflict in drivers/pci/controller/dwc/pcie-qcom.c
CONFLICT (content): Merge conflict in drivers/pci/controller/pci-host-common.c
CONFLICT (content): Merge conflict in drivers/pci/controller/pci-host-common.h
Automatic merge failed; fix conflicts and then commit the result.
Merge failed, manual merge
##[error]Process completed with exit code 1.

The same 4-file conflict appears in every checker log (checkpatch, check-patch-compliance, dt-binding-check, dtb-check, sparse-check, check-uapi-headers). The baseline tag used was 4e0b0df1c84f9b81089bf478c63d635cc97040a1.

Conflicting files (all in drivers/pci/controller/):

• dwc/pcie-designware-host.c — modified by patch 4/5 (FROMLIST: PCI: dwc: Use common D3cold eligibility helper)
• dwc/pcie-qcom.c — modified by patches 2/5, 3/5, 5/5
• pci-host-common.c — modified by patch 1/5 (FROMLIST: PCI: host-common: Add helper to determine host bridge D3cold eligibility)
• pci-host-common.h — modified by patch 1/5

Root cause analysis: The baseline tag (4e0b0df1c84f9b81089bf478c63d635cc97040a1) in qualcomm-linux/kernel has diverged from the base of this PR's topic branch. The same files touched by this PR have been modified in the baseline since the topic branch was created, causing irreconcilable auto-merge conflicts.

Fix options:

1. Rebase the topic branch onto the current baseline (preferred):

   git fetch origin
   git rebase 4e0b0df1c84f9b81089bf478c63d635cc97040a1
   # resolve conflicts in the 4 files, then:
   git rebase --continue
   git push --force-with-lease

2. Re-trigger CI after rebase — once the topic branch is rebased and conflicts resolved, re-run the checker workflow. All 6 checkers will then execute against the actual patch content.

Reproduce locally:

git clone https://github.com/qualcomm-linux/kernel kernel
cd kernel
git fetch origin topic/tech/bus/pci/all
git checkout -b integ 4e0b0df1c84f9b81089bf478c63d635cc97040a1
git merge origin/topic/tech/bus/pci/all
# Observe conflicts in the 4 files listed above

---

✅ tag-check

All 5 commits carry the FROMLIST: prefix — subject-prefix requirement satisfied for branch tech/bus/pci/all:

• FROMLIST: PCI: host-common: Add helper to determine host bridge D3cold eligibility
• FROMLIST: PCI: qcom: Add .get_ltssm() helper
• FROMLIST: PCI: qcom: Power down PHY via PARF_PHY_CTRL
• FROMLIST: PCI: dwc: Use common D3cold eligibility helper
• FROMLIST: PCI: qcom: Add D3cold support

---

⏭️ qcom-next-check

All 5 commits are FROMLIST: and require a qcom-next presence check, but network access is unavailable in this environment.

⚠️ qcom-next-check — SKIPPED (could not fetch qcom-next)
Manually verify for each commit subject (strip the "FROMLIST: " prefix):
  git log qcom-linux/qcom-next --grep="PCI: host-common: Add helper to determine host bridge D3cold eligibility"
  git log qcom-linux/qcom-next --grep="PCI: qcom: Add .get_ltssm() helper"
  git log qcom-linux/qcom-next --grep="PCI: qcom: Power down PHY via PARF_PHY_CTRL"
  git log qcom-linux/qcom-next --grep="PCI: dwc: Use common D3cold eligibility helper"
  git log qcom-linux/qcom-next --grep="PCI: qcom: Add D3cold support"

---

Verdict

All 6 checker failures are CI infrastructure failures, not patch defects. The topic branch tech/bus/pci/all has merge conflicts with the current baseline tag in qualcomm-linux/kernel across 4 PCI driver files. The PR author must rebase the topic branch onto the current baseline to resolve the conflicts, then re-trigger CI. No patch content changes are required based on these logs — the actual checker results (checkpatch, compliance, sparse, etc.) are unknown until the merge conflict is resolved and CI re-runs successfully.