Public bug reported:
SUMMARY
System was stable for 95+ days on systemd 245.4-4ubuntu3.24+esm3.
Approximately 6 hours after an unattended upgrade to +esm4 on
2026-09-01, systemd (PID 1) began segfaulting/crashing repeatedly.
Three distinct crashes have now occurred in 4 days, with the
interval between crashes shrinking each time:
2026-09-01 ~13:35 - PID 1 crash (also a separate --user instance
crash same day at 13:18)
2026-09-03 18:29 - PID 1 crash (~2 days later)
2026-09-04 12:45 - PID 1 crash (~4h44m later)
No amdgpu/GPU-related errors, no hardware errors (ECC, disk I/O),
no thermal issues. Kernel was also updated during this window
(5.15.0-190 -> 5.15.0-191) with no apparent effect on frequency.
SYSTEM INFO
Ubuntu 20.04.6 LTS (Focal), Ubuntu Pro/ESM active
Kernel: 5.15.0-191-generic
systemd: 245.4-4ubuntu3.24+esm4 (previously +esm3, stable)
Hardware: Dell Precision workstation, Intel + AMD Navi 14 hybrid
graphics, X11 session (not Wayland)
TIMELINE OF THE UPDATE
Per /var/log/apt/history.log:
Start-Date: 2026-09-01 07:36:30
Upgrade: systemd:amd64 (245.4-4ubuntu3.24+esm3,
245.4-4ubuntu3.24+esm4) [+ libsystemd0, udev,
libpam-systemd, systemd-sysv, libnss-systemd,
systemd-container, systemd-timesyncd]
End-Date: 2026-09-01 07:37:11
First crash occurred at 13:18-13:35 that same day.
CRASH DETAILS
Crash 1 (2026-09-01, PID 1, /var/crash/_usr_lib_systemd_systemd.0.crash):
Signal 11 (SIGSEGV)
Backtrace: kill() -> signal handler -> internal_hashmap_remove_value()
-> [systemd internal] -> sd_event_dispatch() -> sd_event_run()
-> __libc_start_main()
(from /lib/systemd/libsystemd-shared-245.so)
Crash 1b (2026-09-01, separate --user instance, uid 1001,
/var/crash/_usr_lib_systemd_systemd.1001.crash):
Signal 11 (SIGSEGV), identical backtrace shape, same
internal_hashmap_remove_value() call site.
Occurred within the same second that a
gnome-launched-google-chrome.desktop systemd scope reported
"Succeeded" (i.e. a Chrome-spawned process had just exited),
suggesting the crash is triggered by process-exit/reap handling
(SIGCHLD dispatch -> hashmap_remove_value on the watched-PID
table) rather than being Chrome-specific -- unrelated cron-driven
process churn (rapid scp/ssh session open/close cycles) preceded
the PID 1 crash by the same pattern.
Crash 2 (2026-09-03 18:29:14, PID 1):
kernel: systemd[1] segfault at 5d8 ip 000055dedbb3def9
error 4 in systemd[55dedbac0000+bf000]
This fault is in the main systemd binary itself (not
libsystemd-shared-245.so), though still reached via the same
sd_event_run()/sd_event_dispatch() event-loop path. Preceded by
a burst of cron-driven scp/ssh process spawns, same pattern as
crash 1.
System did NOT fully hang this time -- cron jobs continued
executing normally afterward, but `systemctl` and other
D-Bus-dependent calls became unresponsive, degrading further
over the following ~14 hours until a reboot was performed.
Crash 3 (2026-09-04 12:45:58, PID 1):
kernel: traps: systemd[1] general protection fault
ip:7f6ea40cb535 sp:7ffc0d48e810 error:0 in
libsystemd-shared-245.so[7f6ea4007000+175000]
D-Bus became unresponsive essentially immediately this time
(systemctl status timed out within minutes of the crash).
PATTERN
All three crashes occur within or immediately downstream of
systemd's event loop (sd_event_run/sd_event_dispatch), and all
are preceded by bursts of short-lived child process spawning and
reaping (either Chrome's multi-process subprocesses, or cron jobs
invoking scp/ssh in rapid succession). This is consistent with a
race condition or memory corruption bug in systemd's PID-watching/
SIGCHLD-handling code that existed prior to +esm4 but was not
triggered (or triggered exceedingly rarely) under +esm3.
REQUEST
1. Can you confirm what changed between 245.4-4ubuntu3.24+esm3 and
+esm4? (apt changelog does not return content for this ESM-only
build: "E: Failed to fetch changelog:/systemd.changelog")
2. Is this a known/tracked regression?
3. Given the crash frequency is accelerating (95 days -> ~2 days ->
~14 hours -> ~4h44m between incidents), is downgrading to +esm3
a safe interim mitigation, or is there a newer fix already
available?
I have full gdb backtraces, apport crash files, and journalctl
logs for all three incidents available on request.
** Affects: systemd (Ubuntu)
Importance: Undecided
Status: New
--
You received this bug notification because you are a member of Ubuntu
Bugs, which is subscribed to Ubuntu.
https://bugs.launchpad.net/bugs/2166539
Title:
systemd 245.4-4ubuntu3.24+esm4 crashes repeatedly and with increasing
frequency after upgrade from +esm3
To manage notifications about this bug go to:
https://bugs.launchpad.net/ubuntu/+source/systemd/+bug/2166539/+subscriptions
--
ubuntu-bugs mailing list
[email protected]
https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs
