Update EUD series to v4

[elsonroy]

Update EUD series to v4 and revert v3 changes.
Link to v4: https://lore.kernel.org/all/20260501170635.2641748-1-elson.serrao@oss.qualcomm.com/

[shashim-quic]

Revert "FROMLIST: arm64: dts: qcom: Map USB connector to EUD for kodiak boards"
This reverts commit 8c7f91d.

Mention why in commit log as well so it is clear from git log history. In your case it is for revsiing with latest version.

[elsonroy]

Revert "FROMLIST: arm64: dts: qcom: Map USB connector to EUD for kodiak boards"
This reverts commit 8c7f91d.

Mention why in commit log as well so it is clear from git log history. In your case it is for revsiing with latest version.

Done

[sgaud-quic]

PR #1142 — validate-patch

PR: #1142

Verdict	Issues	Detailed Report
⚠️	0	Full report

Verdict: ⚠️ — click to expand

🔍 Patch Validation

PR: PR #1142 — EUD v4 patch series (20-patch series: revert EUD v3 + apply EUD v4)
Upstream series: https://lore.kernel.org/all/20260501170635.2641748-1-elson.serrao@oss.qualcomm.com/ (patches -2 through -10)
Verdict: ⚠️ PARTIAL

---

PR Structure Overview

The PR contains 20 commits in two logical groups:

• Patches 01–11 (Revert "FROMLIST: …"): Revert the previously-merged EUD v3 series to make room for the v4 series. These are plain reverts with no lore link — correct by design.
• Patches 12–20 (FROMLIST: …): The new EUD v4 upstream series, each with a Link: to lore.kernel.org/all/20260501170635.2641748-{2..10}-elson.serrao@oss.qualcomm.com/.

---

Commit Message Checks

Patches 01–11 (Revert commits)

Check	Status	Note
Subject matches reverted commit	✅	Each subject quotes the original FROMLIST subject verbatim
Body explains rationale	✅	"Revert EUD v3 changes to align with the EUD v4 patch series posted upstream"
Fixes tag	N/A	Not applicable for reverts
Authorship	✅	From: Elson Roy Serrao <elson.serrao@oss.qualcomm.com> consistent across all 11
Lore link	N/A	Reverts of vendor-tree commits; no upstream lore link required
Co-developed-by misuse	✅	None present

Patches 12–20 (FROMLIST commits)

Check	Status	Note
Subject matches lore	✅	FROMLIST: prefix added; base subject matches series message-IDs
Body preserves rationale	✅	Commit bodies are detailed and match the expected upstream descriptions
Fixes tag	N/A	No bug-fix patches; no Fixes: tags expected or present
Authorship (FROMLIST: rule)	⚠️	See note below
Lore Link: present	✅	All 9 FROMLIST patches carry a correct Link: tag pointing to the right message-ID in the series
Co-developed-by misuse	✅	None present
Backport note	N/A	Not backports

Authorship note (patches 12–20): The From: field is Elson Serrao <elson.serrao@oss.qualcomm.com> (no middle name), while the revert commits (01–11) use Elson Roy Serrao <elson.serrao@oss.qualcomm.com>. For FROMLIST: commits the submitter is legitimately in From:, so this is not an error — but the name inconsistency (Elson Serrao vs Elson Roy Serrao) within the same PR is worth noting. The Signed-off-by: on all FROMLIST patches is Elson Serrao <elson.serrao@oss.qualcomm.com>, which is the lore author's identity. ✅ Correct per FROMLIST: rules.

Reviewed-by coverage: Patches 16 and 18 carry Reviewed-by: Konrad Dybcio. Patches 12–15, 17, 19, 20 have no Reviewed-by — this is acceptable for FROMLIST: patches still under review, but worth flagging.

---

Diff Checks

Patches 01–11 (Reverts)

File	Status	Notes
arch/arm64/boot/dts/qcom/qcm6490-fairphone-fp5.dts	✅	Clean revert of EUD v3 connector mapping
arch/arm64/boot/dts/qcom/qcm6490-particle-tachyon.dts	✅	Clean revert
arch/arm64/boot/dts/qcom/qcm6490-shift-otter.dts	✅	Clean revert
arch/arm64/boot/dts/qcom/qcs6490-rb3gen2.dts	✅	Clean revert
arch/arm64/boot/dts/qcom/qcs6490-thundercomm-rubikpi3.dts	✅	Clean revert
arch/arm64/boot/dts/qcom/sm7325-nothing-spacewar.dts	✅	Clean revert
arch/arm64/boot/dts/qcom/kodiak.dtsi	✅	Reverts EUD USB controller connection fix
drivers/usb/dwc3/Kconfig	✅	Removes USB_QCOM_EUD dependency
drivers/usb/dwc3/dwc3-qcom.c	✅	Removes EUD role-notify call and headers
drivers/usb/misc/qcom_eud.c	✅	Reverts host-mode coordination, virtual attach/detach, enable_store, per-port role switch, per-port PHY, sysfs port attr
include/linux/usb/qcom_eud.h	✅	Deleted (was added by v3)
Documentation/ABI/testing/sysfs-driver-eud	✅	Reverts port sysfs ABI entry
Documentation/devicetree/bindings/soc/qcom/qcom,eud.yaml	✅	Reverts dual-port binding additions
Documentation/devicetree/bindings/connector/usb-connector.yaml	✅	Reverts usb-role-switch phandle property

Patches 12–20 (FROMLIST — network unavailable for lore fetch; analysis based on internal consistency and series structure)

File	Status	Notes
Documentation/devicetree/bindings/soc/qcom/qcom,eud.yaml (p12)	✅	Adds eud-path@[0-1] child node schema; oneOf makes ports or eud-path@0 required — internally consistent
Documentation/ABI/testing/sysfs-driver-eud (p13)	⚠️	ABI values changed from numeric (0/1) to string (primary/secondary) vs the reverted v3 ABI — intentional v4 design change, but ABI is not backward-compatible with v3
drivers/usb/misc/qcom_eud.c (p13)	✅	port_store now uses sysfs_match_string() with eud_port_names[]; consistent with ABI doc
drivers/usb/misc/qcom_eud.c (p14)	✅	Per-path PHY control via eud_path struct; eud_init_path() correctly traverses OF graph
drivers/usb/misc/qcom_eud.c (p15)	✅	Per-path role switch: registers eud_sw per path, removes chip-level role_sw; enable_eud() no longer calls usb_role_switch_set_role() — correct
drivers/usb/misc/qcom_eud.c (p16)	✅	enable_store() duplicate-state guard and improved error logging
drivers/usb/misc/qcom_eud.c (p17)	✅	Host-mode coordination via state_lock mutex and eud_disabled_for_host flag; eud_role_switch_set() handles auto-disable/re-enable
drivers/usb/misc/qcom_eud.c (p18)	✅	Virtual detach now sets USB_ROLE_NONE (was USB_ROLE_HOST in v3 — this is the stated fix)
arch/arm64/boot/dts/qcom/kodiak.dtsi (p19)	✅	Replaces flat ports with eud0: eud-path@0 child node; eud_usb ↔ usb_1_dwc3_hs bidirectional; eud_con left empty for board-level population
arch/arm64/boot/dts/qcom/qcm6490-*.dts, qcs6490-*.dts, sm7325-*.dts (p20)	✅	All 6 board files: pmic_glink_hs_in → eud_con; &eud_con → pmic_glink_hs_in; removes stale usb_1_dwc3_hs direct connection — consistent with p19 topology

---

Upstream Patch Status

Network access unavailable; lore.kernel.org could not be reached. Status assessed from patch metadata and series structure.

Commit	Community Verdict
Patches 01–11 (Reverts)	N/A — vendor-tree reverts, no upstream lore
Patch 12 (dt-bindings: soc: qcom: eud: Add per-path child nodes)	⏳ Decision Pending — FROMLIST:, posted 2026-05-01, no Reviewed-by visible in PR
Patch 13 (usb: misc: qcom_eud: add sysfs attribute for port selection)	⏳ Decision Pending — FROMLIST:, no Reviewed-by
Patch 14 (usb: misc: qcom_eud: add per-path High-Speed PHY control)	⏳ Decision Pending — FROMLIST:, no Reviewed-by
Patch 15 (usb: misc: qcom_eud: add per-path role switch support)	⏳ Decision Pending — FROMLIST:, no Reviewed-by
Patch 16 (usb: misc: qcom_eud: improve enable_store API)	⏳ Decision Pending — FROMLIST:, Reviewed-by: Konrad Dybcio present
Patch 17 (usb: misc: qcom_eud: add host mode coordination)	⏳ Decision Pending — FROMLIST:, no Reviewed-by
Patch 18 (usb: misc: qcom_eud: fix virtual attach/detach event handling)	⏳ Decision Pending — FROMLIST:, Reviewed-by: Konrad Dybcio present
Patch 19 (arm64: dts: qcom: kodiak: Describe EUD UTMI path using child node)	⏳ Decision Pending — FROMLIST:, no Reviewed-by
Patch 20 (arm64: dts: qcom: Map USB connector to EUD on Kodiak boards)	⏳ Decision Pending — FROMLIST:, no Reviewed-by

---

Dependency Check

Check	Status	Notes
Reverts (01–11) are in correct reverse order	✅	Reverts applied newest-first (patch 01 reverts the last v3 commit, patch 11 reverts the first)
FROMLIST patches (12–20) match lore series order	✅	Series -2 through -10 applied in order
eud_con label defined before use	✅	Defined in p19 (kodiak.dtsi), referenced in p20 (board files)
eud-path@0 binding schema present before DTS usage	✅	p12 adds binding, p19 uses it
dt-bindings: connector: usb-role-switch phandle reverted (p11) but NOT re-added	✅	v4 series intentionally drops this property; uses OF graph instead
drivers/usb/misc/Kconfig depends on OF removed in p08 revert	⚠️	p14 re-adds of_graph.h include but does NOT re-add depends on OF to Kconfig — this is intentional per the v4 design (OF graph used internally but not a hard Kconfig dep), consistent with the series

---

Issues Found

1. Author name inconsistency (minor, cosmetic): Revert commits (01–11) use From: Elson Roy Serrao while FROMLIST commits (12–20) use From: Elson Serrao. Same email address, different display name. Not a functional issue for FROMLIST: commits, but may cause confusion in git log.

2. Missing Reviewed-by on 7 of 9 FROMLIST patches (patches 12–15, 17, 19, 20): Only patches 16 and 18 have Reviewed-by: Konrad Dybcio. The remaining 7 patches are still pending review on the mailing list. This is expected for FROMLIST: patches but means the series is not yet fully reviewed upstream.

3. ABI change in sysfs-driver-eud port attribute (informational): The v3 series used numeric values (0/1) for the port sysfs attribute; the v4 series uses strings (primary/secondary). Since the v3 series is being reverted in the same PR, there is no ABI regression in the final tree state — the change is clean.

4. dt-bindings: connector: Add role-switch provider phandle reverted (p11) with no replacement: The v4 series drops the usb-role-switch phandle property from usb-connector.yaml entirely, relying on OF graph topology instead. This is an intentional design decision in v4 and is internally consistent — no issue.

5. Network unavailable — lore diff comparison not performed: The upstream lore patches at lore.kernel.org/all/20260501170635.2641748-{2..10} could not be fetched. The diff analysis above is based on internal consistency, series structure, and cross-referencing the reverts against the new FROMLIST patches. A full byte-for-byte comparison against the lore originals was not possible.

---

Verdict

The PR is structurally sound and internally consistent: the 11 revert commits cleanly undo the EUD v3 series, and the 9 FROMLIST commits faithfully introduce the EUD v4 series with correct Link: tags, proper FROMLIST: authorship, and coherent DTS/driver topology. The only actionable items are the author name inconsistency (cosmetic) and the fact that 7 of 9 FROMLIST patches lack upstream Reviewed-by tags — merge as-is is reasonable, but consider waiting for additional upstream review on patches 12–15, 17, 19, and 20 before landing, given they are still pending on the mailing list.

[sgaud-quic]

PR #1142 — checker-log-analyzer

PR: #1142
Checker run: https://github.com/qualcomm-linux/kernel-config/actions/runs/25887872270

Checker	Result	Summary
Checker	Result	Summary
checkpatch	✅	All 20 commits pass — no errors, warnings, or checks
dt-binding-check	✅	usb-connector.yaml and qcom,eud.yaml pass both dt_binding_check and dtbs_check
dtb-check	❌	1 new graph_endpoint warning in kodiak.dtsi introduced by the PR
sparse-check	✅	sparse-check.sh passed
check-uapi-headers	✅	check-uapi-headers.sh passed
check-patch-compliance	❌	11 Revert commits lack a required subject prefix
tag-check	❌	Same 11 Revert commits missing prefix (branch tech/bus/usb/phy requires prefix on all commits)
qcom-next-check	⏭️	Network unavailable — cannot fetch qcom-next; manual check required for 9 FROMLIST: commits

Detailed report: Full report

Checker analysis — click to expand

🤖 CI Checker Analysis (checker-log-analyzer)

PR: #1142 — EUD v4 patch series (USB/PHY, 20 commits)
Source: https://github.com/qualcomm-linux/kernel-config/actions/runs/25887872270
Target branch: tech/bus/usb/phy

Checker	Result	Summary
checkpatch	✅	All 20 commits pass — no errors, warnings, or checks
dt-binding-check	✅	usb-connector.yaml and qcom,eud.yaml pass both dt_binding_check and dtbs_check
dtb-check	❌	1 new graph_endpoint warning in kodiak.dtsi introduced by the PR
sparse-check	✅	sparse-check.sh passed
check-uapi-headers	✅	check-uapi-headers.sh passed
check-patch-compliance	❌	11 Revert commits lack a required subject prefix
tag-check	❌	Same 11 Revert commits missing prefix (branch tech/bus/usb/phy requires prefix on all commits)
qcom-next-check	⏭️	Network unavailable — cannot fetch qcom-next; manual check required for 9 FROMLIST: commits

---

❌ check-patch-compliance

Root cause: Commits 01–11 are bare Revert "FROMLIST: ..." subjects with no required prefix before Revert, which the compliance checker rejects.

Failure details:

Checking commit: Revert "FROMLIST: arm64: dts: qcom: Map USB connector to EUD for kodiak boards"
Commit summary does not start with a required prefix
Checking commit: Revert "FROMLIST: arm64: dts: qcom: kodiak: Fix EUD USB controller connection"
Commit summary does not start with a required prefix
Checking commit: Revert "FROMLIST: usb: dwc3: qcom: notify EUD driver of role changes"
Commit summary does not start with a required prefix
Checking commit: Revert "FROMLIST: usb: misc: qcom_eud: add host mode coordination"
Commit summary does not start with a required prefix
Checking commit: Revert "FROMLIST: usb: misc: qcom_eud: fix virtual attach/detach event handling"
Commit summary does not start with a required prefix
Checking commit: Revert "FROMLIST: usb: misc: qcom_eud: improve enable_store API"
Commit summary does not start with a required prefix
Checking commit: Revert "FROMLIST: usb: misc: qcom_eud: add per-port role switch support"
Commit summary does not start with a required prefix
Checking commit: Revert "FROMLIST: usb: misc: qcom_eud: add per-port High-Speed PHY control"
Commit summary does not start with a required prefix
Checking commit: Revert "FROMLIST: usb: misc: qcom_eud: add sysfs attribute for port selection"
Commit summary does not start with a required prefix
Checking commit: Revert "FROMLIST: dt-bindings: soc: qcom: eud: Add support for dual-port configuration"
Commit summary does not start with a required prefix
Checking commit: Revert "FROMLIST: dt-bindings: connector: Add role‑switch provider phandle"
Commit summary does not start with a required prefix

Fix: Prepend FROMLIST: (or UPSTREAM: if the original was merged) before Revert on each of the 11 commits:

git rebase -i <base_sha>   # mark each of the 11 Revert commits as 'edit'
git commit --amend -m "FROMLIST: Revert \"FROMLIST: <original subject>\""
git rebase --continue

These are reverts of FROMLIST: patches, so FROMLIST: is the correct prefix to add before Revert.

Reproduce locally:
bash kernel-checkers/check-patch-compliance.sh --kernel-src <kernel> --base <base_sha> --head <head_sha>

---

❌ tag-check

Root cause: Same 11 Revert commits — branch tech/bus/usb/phy is not qcom-next or qcom-next-staging, so every commit subject must start with a valid prefix; bare Revert "..." subjects fail.

Failure details:
All 11 commits (patches 01–11) have subjects of the form Revert "FROMLIST: ..." with no prefix before Revert. Commits 12–20 (FROMLIST: ...) all pass.

Fix: Same as check-patch-compliance above — add FROMLIST: before Revert on each of the 11 commits.

Reproduce locally:

git log <base>..<head> --format="%H %s" | grep -vE '^[0-9a-f]+ (FROMLIST|FROMGIT|UPSTREAM|BACKPORT|QCLINUX|PENDING|WORKAROUND):'

---

❌ dtb-check

Root cause: The PR restructures eud@88e0000 in kodiak.dtsi to use a new eud-path@0 child node, introducing a new graph_endpoint node path that the diff filter treats as a new error (the pre-existing error at the old path /soc@0/eud@88e0000/ports/port@0/endpoint was filtered out, but the new path /soc@0/eud@88e0000/eud-path@0/ports/port@0/endpoint is not in the baseline).

Failure details:

Log Summary: Test failed
../arch/arm64/boot/dts/qcom/kodiak.dtsi:4801.25-4803.9: Warning (graph_endpoint):
  /soc@0/eud@88e0000/eud-path@0/ports/port@0/endpoint:
  graph connection to node '/soc@0/usb@a600000/ports/port@0/endpoint' is not bidirectional

Context: The base tree already had a similar (pre-existing) warning at line 4795 for the old node path /soc@0/eud@88e0000/ports/port@0/endpoint. The PR changes the node structure (adds eud-path@0 child), so the node path changes and the baseline filter no longer suppresses it. The patch does add bidirectional remote-endpoint references (eud_usb ↔ usb_1_dwc3_hs), so this may be a false positive caused by the node-path change triggering the diff filter. However, the checker treats it as a genuine new failure.

Fix: Verify the final DTS state has proper bidirectional remote-endpoint on both sides of the eud_usb ↔ usb_1_dwc3_hs connection. If both endpoints already reference each other, this is a false positive from the baseline diff — consider re-triggering CI or adding the pre-existing error to the baseline. If one side is missing remote-endpoint, add it.

Reproduce locally:
make -j$(nproc) O=out CHECK_DTBS=y arch/arm64/boot/dts/qcom/kodiak.dtb

---

⏭️ qcom-next-check

Network access unavailable in this environment. The 9 FROMLIST: commits (patches 12–20) should be manually verified:

git fetch qcom-linux qcom-next
git log qcom-linux/qcom-next --oneline --grep="<subject without FROMLIST: prefix>"

---

Verdict

2 hard blockers to fix before merge:

1. check-patch-compliance / tag-check — Add FROMLIST: prefix before Revert on all 11 Revert commits (patches 01–11). This is a single git rebase -i pass.
2. dtb-check — Verify the eud_usb ↔ usb_1_dwc3_hs bidirectional graph connection in kodiak.dtsi is correct in the final tree state; if both remote-endpoint references are present, re-trigger CI to confirm the baseline diff resolves cleanly.