Additional data point from GitHub issue cilium/cilium#43198:

**Environment (update since my initial report):**
- Ubuntu 25.10, kernel 6.17.0-1006-raspi (upgraded from 6.17.0-1004)
- Raspberry Pi 5 Model B Rev 1.0
- Cilium v1.18.5 (kube-proxy replacement, VXLAN, Gateway API)
- K3s cluster expanded from 2 to 3 nodes (added 1 worker)

**Observations:**

1. **Not eBPF-specific:** Combined with Chris Boot's report (Calico +
nftables backend), we now have confirmed cases on both eBPF (Cilium
kube-proxy replacement) and nftables (Calico) networking stacks. This
suggests the issue is in the macb/RP1 driver layer, not in the packet
processing backend.

2. **Frequency:** All 3 of my nodes crash approximately every 3 days.
Subjectively, failures seem to correlate with higher network/CPU load,
though I don't have hard metrics to confirm this.

3. **Physical layer appears alive but non-functional:** When a node
dies, the link LED stays lit and the ACT LED continues blinking.
However, the router shows the port in a strange state.
Unplugging/replugging the cable has no effect. This suggests the PHY
believes it has a valid link, but something in the macb/RP1 stack is
broken.

4. **UART unresponsive:** I attempted UART connection via Flipper Zero
and via another node's serial passthrough. No data received from the
dead node's TX pin, despite journald logs showing the system was still
writing locally at that time. This suggests the CPU core is alive but
RP1 (which handles both ethernet and UART) may be completely
unresponsive.

5. **Kernel updates ineffective:** Problem persists across 6.17.0-1004 →
6.17.0-1006 updates.

6. **Diagnostics setup:** I've now deployed periodic ethtool stats
collection (every 60s) and link state monitoring (every 30s) on all
nodes. Will provide detailed pre-failure metrics when the next crash
occurs.

-- 
You received this bug notification because you are a member of Ubuntu
Bugs, which is subscribed to Ubuntu.
https://bugs.launchpad.net/bugs/2133877

Title:
  Complete network hang on Raspberry Pi 5 with kernel 6.17 under load -
  possibly related to CPU frequency scaling

To manage notifications about this bug go to:
https://bugs.launchpad.net/ubuntu/+source/linux-raspi/+bug/2133877/+subscriptions


-- 
ubuntu-bugs mailing list
[email protected]
https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs

Reply via email to