kernel: disable broken EEE on the MT7530 PHY by work423 · Pull Request #25058 · openwrt/openwrt

work423 · 2026-09-06T18:47:41Z

The MT7530 internal GE PHY has broken EEE: on a 2-pair cable where both ends advertise gigabit, the port loops instead of falling back to 100 Mbps, so no link and no DHCP.

Pending-6.18 patch, code identical to the netdev submission (Reviewed-by: Andrew Lunn, ETHERNET PHY LIBRARY maintainer):
https://lore.kernel.org/netdev/[email protected]/

Supersedes #22647 (per netdev review, the fix belongs in the PHY driver, not DSA). Ref: #24752.

Tested on ASUS RT-AX53U (MT7621): clean 100 Mbps WAN link + DHCP over a 2-pair cable; unpatched loops.

openwrt · 2026-09-06T18:47:45Z

Formality Check: Suggestions Available

We checked this pull request against the contribution guidelines. Here is what needs your attention:

⚠️ STYLISTIC WARNINGS & SUGGESTIONS

Commit c2ec0b9:

⚠️ Commit author email '[email protected]' is not linked to any registered GitHub account. Please add and verify this email in your GitHub profile settings.

_{Something broken? Consider reporting an issue.}
_{Running version 059e3de deployed on 2026-09-09 11:53:18 CEST}

raenye · 2026-09-06T22:14:23Z

Why in target/linux/generic/pending-6.18? IMHO this patch should be in target/linux/ramips/patches-6.18.

work423 · 2026-09-06T23:07:11Z

Why in target/linux/generic/pending-6.18? IMHO this patch should be in target/linux/ramips/patches-6.18.

I put it in generic because the upstream fix is not ramips specific. It patches drivers/net/phy/mediatek/mtk-ge.c, the driver for the MT7530 internal GE PHY which is used both by the MT7621 integrated switch and by boards with the dedicated MT7530 IC ( Banana Pi R2 on the mediatek target) same silicon same defect. The netdev review moved the fix from the DSA side to the PHY driver precisely so it covers every user of this PHY.

openwrt-ai

Reviewed 1 new commit.

Generated by Claude Code

openwrt-ai · 2026-09-07T04:42:10Z

+diff --git a/drivers/net/phy/mediatek/mtk-ge.c b/drivers/net/phy/mediatek/mtk-ge.c
+index 73d9b72..35600c7 100644


This is raw git format-patch output, not a quilt-refreshed patch: none of the 1050 patches under target/linux/generic/ carry a diff --git header, an index line, or the trailing --/2.43.0 git footer (lines 85-86 here). Run make target/linux/refresh (with a MT7530-using target selected) and commit the result — that also regenerates hack-6.18/766-net-phy-mediatek-ge-add-LED-configuration-interface.patch,`` which patches the same file and, after this one, applies with offset 21.

Generated by Claude Code

Djfe · 2026-09-07T09:26:40Z

Could this commit be reverted then and it would still be fixed for affected devices?
b7fa9d9

work423 · 2026-09-07T14:16:48Z

Could this commit be reverted then and it would still be fixed for affected devices? b7fa9d9

Functionally yes on main/6.18 with this patch the EEE advertisement of the MT7530 PHY is cleared in probe() and EEE is locked off in the driver, which is earlier and stronger than the eee-broken-100tx/1000t device tree markers from b7fa9d9, so the symptoms that commit addressed are gone either way.
I'd still not revert it now the DT markers are the only protection on the release branches whose kernels do not have the driver fix (24.10, 25.12) they cost nothing and document the broken silicon.

Once the upstream fix has landed and the OpenWrt kernel bumps inherit it, the markers and this pending patch can be dropped together that looks like a cleanup for later rather than something to fold into this PR.
https://lore.kernel.org/netdev/[email protected]/

openwrt-ai

Commit checks

b2a8f2f "kernel: backport MT7530 broken-EEE fix from upstream" — "bound for v7.0" doesn't hold for a commit dated 8 Sep 2026; v7.0-tagged backports in this tree are January 2026 material. Name the release it actually lands in.

Generated by Claude Code

openwrt-ai · 2026-09-10T15:28:14Z

+diff --git a/drivers/net/phy/mediatek/mtk-ge.c b/drivers/net/phy/mediatek/mtk-ge.c
+index 73d9b72f9d9e2..96d8ac5154e5e 100644


Raw git format-patch output: none of the 165 patches in target/linux/generic/backport-6.18/ carry a diff --git or index line, quilt-refreshed patches go straight from the diffstat to --- a/.... Run make target/linux/refresh with an MT7530 target selected and commit the result; that also refreshes hack-6.18/766-net-phy-mediatek-ge-add-LED-configuration-interface.patch,`` which patches the same file and, with this patch applied first, lands its hunks 2 and 3 at offset 28.

Generated by Claude Code

openwrt-ai · 2026-09-10T15:28:20Z

@@ -0,0 +1,101 @@
+From ccbe7540e4aad0d1c3acc249697350b93ccb8025 Mon Sep 17 00:00:00 2001
+From: Vladislav Karmanov <[email protected]>
+Date: Tue, 8 Sep 2026 17:52:13 +0300


The v7.0- prefix in the filename doesn't fit this date: the v7.0- backports in this directory are January 2026 material, and the newest prefix in tree is v7.4 on 786-01-v7.4-net-phy-add-phy_detach_internal-helper.patch`` dated 31 Aug 2026. Rename the file to the release this commit actually lands in.

Generated by Claude Code

fixed, thanks

Generated by Claude Code

work423 · 2026-09-11T23:28:39Z

Update: the fix is in Linus's tree. It was applied to netdev/net.git as ccbe7540e4aa ("net: phy: mediatek-ge: disable EEE on the MT7530 PHY") and has since been pulled into mainline - it will be part of v7.3:

https://git.kernel.org/torvalds/linux/c/ccbe7540e4aa
netdev thread: https://lore.kernel.org/netdev/[email protected]/

This PR now carries that upstream commit verbatim, refiled from pending-6.18 to backport-6.18 as @raenye suggested.

@raenye: I kept it under generic rather than ramips/patches-6.18 because the upstream commit covers every user of this PHY driver, not just ramips/mt7621 - the MT7530 internal GE PHY is also used by boards with the standalone MT7530 IC on the mediatek target (Banana Pi R2), and the netdev review moved the fix to the PHY driver precisely to cover those too.

openwrt-ai

Reviewed 1 new commit.

Generated by Claude Code

openwrt-ai · 2026-09-12T03:24:54Z

@@ -0,0 +1,101 @@
+From ccbe7540e4aad0d1c3acc249697350b93ccb8025 Mon Sep 17 00:00:00 2001


The 800- number puts this in the "Other drivers" range, but PATCHES.md assigns 7xx to Network / PHY and 8xx to other drivers — this patches drivers/net/phy/mediatek/mtk-ge.c. Renumber into the free 7xx slots next to the other PHY backports (781-783, 786-787 are unused in this directory).

Generated by Claude Code

fixed, thanks

Generated by Claude Code

AuthorReflex · 2026-09-12T12:10:52Z

Could this commit be reverted then and it would still be fixed for affected devices? b7fa9d9

Yes
Tested on Wifire s1500 with two-pair cable and 100 mbit linked device

The upstream fix disabling broken EEE on the MT7530 internal GE PHY was applied to netdev/net.git (ccbe7540e4aa) and is bound for v7.3. Carry the upstream commit verbatim as a backport patch. Signed-off-by: Vladislav Karmanov <[email protected]>

Bruno-HiFi · 2026-09-12T14:42:42Z

Hi,
what about a backport for Linux 6.12.103?

The problem was originally observed on this kernel, and it still affects OpenWrt 25.12 on MT7621/MT7530 devices.

work423 · 2026-09-12T15:06:57Z

Hi, what about a backport for Linux 6.12.103?

The problem was originally observed on this kernel, and it still affects OpenWrt 25.12 on MT7621/MT7530 devices.

Right - 25.12 (kernel 6.12) is affected too. Since the upstream commit (ccbe7540e4aa) carries a Fixes: tag for af3b4b0e59de, which is in 6.12.y, the stable process will backport it into 6.12.x once v7.3 is released, and the 25.12 branch then inherits it with the next kernel bump no extra patch needed, but that chain takes a while.

So users don't have to wait, I can also open a PR against the openwrt-25.12 branch carrying the fix adapted for 6.12 (6.12 predates phy_disable_eee(), so the port clears MDIO_AN_EEE_ADV in probe()/config_init() and sets the equivalent eee_cfg/eee_broken_modes state by hand).

@raenye @dangowrt: would you prefer that as a separate PR now, or the usual route - this one merged first, then a backport PR to the release branch?

openwrt-ai

Reviewed 1 new commit; no new issues found.

Generated by Claude Code

raenye · 2026-09-12T16:15:50Z

Usually I'd say first merge snapshot, then stable, but if the fix is different enough (i.e., more than just refreshing a patch) perhaps it's better to prepare the other PR beforehand.

Bartel-C8 · 2026-09-15T11:09:56Z

@work423 I am interested in the backport PR to test out locally!

work423 · 2026-09-15T21:00:59Z

@work423 I am interested in the backport PR to test out locally!

Here it is: #25206

Reports on other MT7621 or MT7530 boards are very welcome - they help the review.

openwrt Bot added kernel pull request/issue with Linux kernel related changes not following guidelines Pull request does not follow formatting guidelines labels Sep 6, 2026

work423 force-pushed the eee-mt7530-pending branch 2 times, most recently from 43e13db to d5043f6 Compare September 6, 2026 19:15

openwrt Bot removed the not following guidelines Pull request does not follow formatting guidelines label Sep 6, 2026

work423 mentioned this pull request Sep 6, 2026

[RFT] ramips: mt7621: disable EEE on the driver side #22647

Open

openwrt-ai reviewed Sep 7, 2026

View reviewed changes

work423 force-pushed the eee-mt7530-pending branch from d5043f6 to b2a8f2f Compare September 10, 2026 13:26

openwrt-ai reviewed Sep 10, 2026

View reviewed changes

work423 force-pushed the eee-mt7530-pending branch from b2a8f2f to b8fb444 Compare September 10, 2026 15:45

work423 mentioned this pull request Sep 11, 2026

ramips/mt7621: Ethernet auto-negotiation fails/loops on 4-wire cable (1000baseT vs 100baseT) #24752

Open

1 task

openwrt-ai reviewed Sep 12, 2026

View reviewed changes

work423 force-pushed the eee-mt7530-pending branch from b8fb444 to c2ec0b9 Compare September 12, 2026 13:35

openwrt-ai reviewed Sep 12, 2026

View reviewed changes

work423 mentioned this pull request Sep 15, 2026

kernel: backport MT7530 broken-EEE fix from upstream (6.12-adapted) #25206

Open

		diff --git a/drivers/net/phy/mediatek/mtk-ge.c b/drivers/net/phy/mediatek/mtk-ge.c
		index 73d9b72..35600c7 100644

		diff --git a/drivers/net/phy/mediatek/mtk-ge.c b/drivers/net/phy/mediatek/mtk-ge.c
		index 73d9b72f9d9e2..96d8ac5154e5e 100644

		@@ -0,0 +1,101 @@
		From ccbe7540e4aad0d1c3acc249697350b93ccb8025 Mon Sep 17 00:00:00 2001

Uh oh!

Uh oh!

Conversation

work423 commented Sep 6, 2026

Uh oh!

openwrt Bot commented Sep 6, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Formality Check: Suggestions Available

⚠️ STYLISTIC WARNINGS & SUGGESTIONS

Uh oh!

raenye commented Sep 6, 2026

Uh oh!

work423 commented Sep 6, 2026

Uh oh!

openwrt-ai left a comment

Choose a reason for hiding this comment

Uh oh!

openwrt-ai Sep 7, 2026

Choose a reason for hiding this comment

Uh oh!

Djfe commented Sep 7, 2026

Uh oh!

work423 commented Sep 7, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

openwrt-ai left a comment

Choose a reason for hiding this comment

Commit checks

Uh oh!

openwrt-ai Sep 10, 2026

Choose a reason for hiding this comment

Uh oh!

openwrt-ai Sep 10, 2026

Choose a reason for hiding this comment

Uh oh!

openwrt-ai Sep 12, 2026

Choose a reason for hiding this comment

Uh oh!

work423 commented Sep 11, 2026

Uh oh!

openwrt-ai left a comment

Choose a reason for hiding this comment

Uh oh!

openwrt-ai Sep 12, 2026

Choose a reason for hiding this comment

Uh oh!

openwrt-ai Sep 12, 2026

Choose a reason for hiding this comment

Uh oh!

AuthorReflex commented Sep 12, 2026

Uh oh!

Bruno-HiFi commented Sep 12, 2026

Uh oh!

work423 commented Sep 12, 2026

Uh oh!

openwrt-ai left a comment

Choose a reason for hiding this comment

Uh oh!

raenye commented Sep 12, 2026

Uh oh!

Bartel-C8 commented Sep 15, 2026

Uh oh!

work423 commented Sep 15, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

7 participants

openwrt Bot commented Sep 6, 2026 •

edited

Loading

work423 commented Sep 7, 2026 •

edited

Loading