I can confirm that linux-raspi 7.0.0-1023.23 fixes this for me. Setup: Raspberry Pi 5 used as a router, Ubuntu 26.04, onboard eth0 (macb/RP1) as WAN, 1 Gbit/s.
Before (7.0.0-1020.20): 5 silent TX stalls in 11 days (Oct 29 - Nov 8). Link up, no errors in ethtool -S, TX counter frozen, qdisc backlog full (1000p), RX still working. It never recovered by itself; only "ip link set eth0 down/up" or a reboot helped. After (7.0.0-1023.23, running since Nov 10, macb_tx_timeout present in /proc/kallsyms): in 7 days the stall still occurred 7 times, but each time the kernel recovered on its own within 5-6 seconds: macb 1f00100000.ethernet eth0: NETDEV WATCHDOG: CPU: 3: transmit queue 0 timed out 5414 ms (same line 7 times, 5116-5975 ms) No link flap, no DHCP renewal, and my external watchdog (pings the gateway via eth0 once a minute) never saw a single failed check. TSO/SG are left enabled. A second Pi 5 with the same kernel had no stalls during the same week. So the stall itself is still there, but e438ec3e9e95 turns a permanent outage into a ~5 s hiccup. -- 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
