26.7.6: unbound (and syslog-ng) segfault with signal 11 after upgrade (Hyper-V)

[mnaim]

Update - fix still holds.

Thanks Franco!

[franco]

Great to hear.  We'll pick this up in a small .7 early next week.


Cheers,
Franco

[Nex VII]

Update from my side: the test kernel (26.7.6-hyperv) is installed on the Hyper-V Gen2 VM (host Xeon E-2236, 4 vCPU, 8 GB).

- Booted 2026-10-09 17:23 CEST, uptime now 2h02m.
- No "exited on signal" entries in System > Log Files > General since boot. With the stock 26.7.6 kernel, unbound was exiting on signal 11 every 6-12 minutes and syslog-ng twice.
- No log entries above Warning since boot; all services running.

Thanks for the quick fix, Franco!

[franco]

Yay, great.  :)

[drobins1976]

I can confirm a similar problem with OPNsense 26.7.6 on Hyper-V.

My setup is a Windows 11 Hyper-V host with an AMD Ryzen 7 8745HS, and an OPNsense VM with 2 vCPUs and 4 GB RAM. Unbound has DNSSEC and the Python DNS blocklist module enabled.

After updating on 9 October, Unbound repeatedly exited with signal 11. The WAN connection remained up, but DNS requests to the router timed out. Direct internet access by IP still worked.

These are the crash timestamps recorded in the kernel log (BST):

- 09:01:57
- 09:02:03
- 09:39:15
- 09:49:58
- 09:50:15
- 10:40:20
- 10:44:44

The kernel messages were of this form:

    pid 52844 (unbound), jid 0, uid 59: exited on signal 11 (no core dump - denied by kern.coredump)

Restarting Unbound only restored DNS temporarily. Restarting the Windows host also did not stop the crashes; Unbound crashed again after the VM came back up.

I rolled back ONLY the kernel using:

    opnsense-update -kr 26.7.4

The kernel installed successfully, then I rebooted OPNsense. The application and packages remained on 26.7.6. No resolver, blocklist, firewall or VM settings were changed as part of this recovery.

Before rollback:

    FreeBSD 15.1-RELEASE-p4 stable/26.7-n284150-0a7d52ef8757

After rollback:

    FreeBSD 15.1-RELEASE-p3 stable/26.7-n283949-083dc7025377

As of 10 October at 11:27 BST, the VM has been up for approximately 24 hours 34 minutes. Unbound is running, DNS lookups through the router succeed, and HTTPS through the router works. No Unbound signal-11 entries appear in the current system log, although it has rotated, so I cannot claim complete log coverage for the whole period.

One additional point: the normal update offer now includes both opnsense 26.7.6_3 and kernel 26.7.6, which would undo this workaround. I have not installed it.

This seems consistent with the Hyper-V kernel regression discussed here. I have not tested the 26.7.6-hyperv test kernel, and I have no crash backtrace because core dumps were disabled.

[3dat]

Hi all,
I am very new to OPNsense but already a great fan :)

I think I can confirm a similar issue to 26.7.6_3 on Hyper-V, 2025 Server on DELL R630 with E5-2637v3 CPU
My Syslog-ng is not willing to start. Restart leads to:

Error opening file for reading; filename='/dev/klog', error='Device busy (16)'
Error initializong message pipeline; plugin_name='file', location='usr/local/etc/syslog.ng.conf:18:5'
/usr/local/etc/rc.d/syslog.ng: Warning: failed to start syslog_ng

What I did:
Reinstall syslog_ng

Will try with the hyperv update

Many thanks

[magnust]

On Hyper-V here too, same problem. Just did the "opnsense-update -zkr 26.7.6-hyperv" and hope for the best :-D

[magnust]

On Hyper-V here too, same problem. Just did the "opnsense-update -zkr 26.7.6-hyperv" and hope for the best :-D

Looking good so far

[Canardo96]

Hi all,

I am troubleshooting recurring Unbound crashes on OPNsense 26.7.6_3 running under Microsoft Hyper-V.

The error is:

pid XXXXX (unbound), jid 0, uid 59: exited on signal 11
(no core dump - sugid process denied by kern.sugid_coredump)

I previously had several stability problems on this OPNsense VM, including Zenarmor/netmap and syslog-ng issues.

To eliminate configuration corruption and other variables, I completely abandoned that VM.

I created a new Hyper-V VM, performed a clean OPNsense installation and manually rebuilt the configuration. No configuration backup from the previous installation was restored.

The good news:
- Zenarmor is now stable
- No previous netmap problems
- syslog-ng is stable
- The firewall itself is operating normally

Unfortunately, Unbound has started crashing with signal 11 again.

Environment:
- OPNsense 26.7.6_3
- FreeBSD 15.1-RELEASE-p4
- Microsoft Hyper-V
- 4 vCPU
- ~12 GB RAM
- hn0/hn1
- Zenarmor 2.7

Has anyone seen recurring Unbound SIGSEGV crashes on OPNsense 26.7.x under Hyper-V?

Since the problem returned on a clean VM without restoring the old configuration, I am especially interested in what debugging information I should collect to determine the cause.

Thanks,
Maxim

[franco]

There's a test kernel and 26.7.7 should fix this later today. You don't have to repost an analysis without reading the thread.


Cheers,
Franco