** Description changed: [ Impact ] In environments with high network interface churn, the `lldpd` daemon can crash when interfaces are rapidly added and removed. This occurs due to a race condition and memory handling fault when `lldpd` processes Netlink events concurrently with its internal polling loop. The fix backports the upstream resolution for issue #744 (included in upstream release 1.0.20), which safely handles concurrent interface additions/removals in the Netlink event processor without crashing. [ Test Plan ] The upstream bug report has a simple script that can be used to reproduce the issue: ``` #!/bin/env python3 import subprocess import time # add dummy interface ethx subprocess.run(['ip', 'link', 'add', 'ethx', 'type', 'dummy'], check=True) subprocess.run(['ip', 'link', 'set', 'ethx', 'up'], check=True) time.sleep(2) # remove dummy interface ethx（in fact this won't remove nic) subprocess.run(['ip', 'link', 'del', 'ethx'], check=True) # immediately add dummy nic ethx subprocess.run(['ip', 'link', 'add', 'ethx', 'type', 'dummy'], check=True) subprocess.run(['ip', 'link', 'set', 'ethx', 'up'], check=True) time.sleep(1) # delay 1s，now lldpd will update and reinit ethx # now remove ethx, as this will result in bind raw socket failure and trigger "lldpd_hardware_cleanup(cfg, hardware)" subprocess.run(['ip', 'link', 'del', 'ethx'], check=True) ``` I've successfully reproduced this in a VM with 4 vCPUs and 8GB memory. Just run: sudo tail -f /var/log/syslog & sudo python3 repro.py After a few executions attempt the following error is seen: 2026-09-23T00:52:29.271773+00:00 cases-lldpd-noble kernel: lldpd[2126]: segfault at 616dff74f ip 0000616dfea8ced0 sp 00007ffcc5cf3220 error 4 in lldpd[616dfea78000+22000] likely on CPU 3 (core 0, socket 3) 2026-09-23T00:52:29.271786+00:00 cases-lldpd-noble kernel: Code: 48 8b b5 b0 fe ff ff 4c 89 e2 48 89 df e8 18 86 00 00 4c 89 e6 48 89 df e8 5d 84 00 00 4c 8b a3 30 01 00 00 4d 85 e4 74 29 90 <41> 8b 44 24 44 85 c0 74 16 4c 89 e6 48 89 df e8 1c f5 ff ff 4c 89 2026-09-23T00:52:30.075982+00:00 cases-lldpd-noble systemd[1]: lldpd.service: Main process exited, code=dumped, status=11/SEGV Issue is reproduceable in Noble and Resolute. + [ Where problems could occur ] + + The change affects interfaces that fail initialization, particularly + when an interface is being converted from another type. If cleanup is + skipped, a new interface whose initialization fails would leave its' + allocated resources behind. A regression could appear as increasing + memory use after a herd of such failures. However, such scenario seems + uncommon because it means the interface disappears between the netlink + scan and socket setup or creating socket fails. + [ Other Info ] Upstream Issue: https://github.com/lldpd/lldpd/issues/744 Upstream Fix: https://github.com/lldpd/lldpd/commit/705087fe48483f14ae63295bfe67044738eea1d6 Affected Ubuntu series: Noble (24.04), Resolute (26.04)
-- You received this bug notification because you are a member of Ubuntu Bugs, which is subscribed to Ubuntu. https://bugs.launchpad.net/bugs/2167992 Title: SRU: Fix crash on rapid addition/removal of interfaces (Upstream #744) To manage notifications about this bug go to: https://bugs.launchpad.net/ubuntu/+source/lldpd/+bug/2167992/+subscriptions -- ubuntu-bugs mailing list [email protected] https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs
