"eth0: TX stall detected on queue 0" "re-kicking TSTART" errors on 6.18.37-v8+ on multiple Raspberry Pi 5 devices

[satmandu]

Describe the bug

I'm seeing this on multiple Raspberry PI 5 devices on different networks:

[61929.702886] macb: eth0: TX stall detected on queue 0 (tail=35473558 head=35473559); re-kicking TSTART
[64329.950222] macb: eth0: TX stall detected on queue 0 (tail=35620587 head=35620588); re-kicking TSTART
[66655.445267] macb: eth0: TX stall detected on queue 0 (tail=35621413 head=35621414); re-kicking TSTART
[67457.234075] macb: eth0: TX stall detected on queue 0 (tail=35628477 head=35628478); re-kicking TSTART
[67518.673810] macb: eth0: TX stall detected on queue 0 (tail=35628499 head=35628500); re-kicking TSTART
[68589.774859] macb: eth0: TX stall detected on queue 0 (tail=35673274 head=35673275); re-kicking TSTART
[69153.996905] macb: eth0: TX stall detected on queue 0 (tail=35681215 head=35681216); re-kicking TSTART
[69186.764808] macb: eth0: TX stall detected on queue 0 (tail=35681268 head=35681269); re-kicking TSTART

I'm running current trixie based Raspberry Pi OS on these machines, just with the kernel from the rpi-firmware repo, and the latest eeprom.

sudo rpi-eeprom-update
BOOTLOADER: up to date
   CURRENT: Mon Jun 29 12:07:01 AM UTC 2026 (1782691621)
    LATEST: Mon Jun 29 12:07:01 AM UTC 2026 (1782691621)
   RELEASE: latest (/usr/lib/firmware/raspberrypi/bootloader-2712/latest)
            Use raspi-config to change the release.

sudo vcgencmd version
2026/06/29 00:07:01
Copyright (c) 2012 Broadcom
version f68405bc (release) (embedded)

Steps to reproduce the behaviour

Boot with the current firmware/kernel from the rpi-firmware repo.

Device (s)

Raspberry Pi 5

System

https://pastebin.com/iDyp0jim

Logs

No response

Additional context

This is new behavior. IIRC the previous kernel did not create these error messages.

I'm not seeing this on my Raspberry PI 4B devices with the same kernel running.

[pelwell]

Mmm, yes - I can see them too. This must be as the result of a fairly recent change. Either the stall detection is getting false positives or real stalls have started happening (I've not been affected by them up to now).

[nbuchwitz]

I'm already looking into it. Unfortunately the patches upstream and downstream targeting the stalls seem to have diverting ...

[pelwell]

That's good to know. At least I can reproduce this issue, whatever the cause, so I may have a look as well.

[nbuchwitz]

For further testing (mainline patch + downstream patches reverted): #7472