can't install kvm -- libvirtd brings segfault

[jeensg]

Anybody has a clue what to do with a segfault in libvirtd?
------------------------------
installed new server @ hetzner
different partitions incl. LVM for virtualization disks
---------------------
progs installed: htop mc fail2ban locate apticron mailutils postfix bash-completion bridge-utils ntp rkhunter smartmontools ntp
additionally configured: hostname and hosts, ipv6-leaveout for debian-mirror, fail2ban for ssh, firewall (iptables), rkhunter, smartd Mailing, RAID Mailing, bridge for virtualization (hetzner suggestion)
..........
then: apt install qemu-kvm libvirt-daemon-system
resulting in:
...

Code: Select all

Setting up libvirt-daemon-system (3.0.0-4) ...
Job for libvirtd.service failed because a fatal signal was delivered to the control process.
See "systemctl status libvirtd.service" and "journalctl -xe" for details.
invoke-rc.d: initscript libvirtd, action "start" failed.
● libvirtd.service - Virtualization daemon
  Loaded: loaded (/lib/systemd/system/libvirtd.service; enabled; vendor preset: enabled)
  Active: activating (auto-restart) (Result: signal) since Tue 2017-08-01 18:19:09 CEST; 7ms ago
  Docs: man:libvirtd(8)
  http://libvirt.org
  Process: 2787 ExecStart=/usr/sbin/libvirtd $libvirtd_opts (code=killed, signal=SEGV)
Main PID: 2787 (code=killed, signal=SEGV)

Aug 01 18:19:09 harry systemd[1]: libvirtd.service: Unit entered failed state.
Aug 01 18:19:09 harry systemd[1]: libvirtd.service: Failed with result 'signal'.
dpkg: error processing package libvirt-daemon-system (--configure):
subprocess installed post-installation script returned error exit status 1
...
Errors were encountered while processing:
libvirt-daemon-system

.........
systemctl status libvirtd.service brings:

Code: Select all

Aug 01 18:19:10 harry systemd[1]: Failed to start Virtualization daemon.
Aug 01 18:19:10 harry systemd[1]: libvirtd.service: Unit entered failed state.
Aug 01 18:19:10 harry systemd[1]: libvirtd.service: Failed with result 'signal'.
Aug 01 18:19:11 harry systemd[1]: libvirtd.service: Service hold-off time over, scheduling restart.
Aug 01 18:19:11 harry systemd[1]: Stopped Virtualization daemon.
Aug 01 18:19:11 harry systemd[1]: libvirtd.service: Start request repeated too quickly.
Aug 01 18:19:11 harry systemd[1]: Failed to start Virtualization daemon.
Aug 01 18:19:11 harry systemd[1]: libvirtd.service: Unit entered failed state.
Aug 01 18:19:11 harry systemd[1]: libvirtd.service: Failed with result 'signal'.

..................
grep libvirtd /var/log/syslog brings one by one retrying again and again -- segfault:

Code: Select all

Aug  1 18:19:09 harry kernel: [  358.621561] libvirtd[2787]: segfault at 0 ip 00007fd7966db24f sp 00007ffe4abaea40 error 4 in libc-2.24.so[7fd796609000+195000]
Aug  1 18:19:10 harry kernel: [  358.870880] libvirtd[2794]: segfault at 0 ip 00007f6e9635a24f sp 00007ffff97561d0 error 4 in libc-2.24.so[7f6e96288000+195000]
Aug  1 18:19:10 harry kernel: [  359.120821] libvirtd[2797]: segfault at 0 ip 00007f3e1d66124f sp 00007ffe65174d20 error 4 in libc-2.24.so[7f3e1d58f000+195000]

.............................
wanted to install package libvirt0-dbg, but it doesn't exist for stretch!? https://wiki.debian.org/libvirt/Debugging

[pylkko]

See the release notes:
https://www.debian.org/releases/stable/ ... ug-archive

I googled libvirtd and segfault and there was a 2017 redhat bug with something about using tun interface causing segfault. I don't know if that is relevant for you

[jeensg]

Thanks for the advice. I did check the release-notes and saw about the debug-symbols. but it's only for the symbols. If I include 'deb http://debug.mirrors.debian.org/debian-debug/ stretch-debug main' in my sources.list, it's possible to install libvirtd0-dbgsym, but not libvirtd0-dbg ...

As for the Redhat-bug, I saw it, too, but unfortunately it doesn't help anything...googled before especially recent problems with libvirtd/libc6(2.24) and segfault, but nothing comes up regarding stretch specifically

Found similar crashes like here, but I don't know what to do with it..

I did also find an explanation what it means ... and maybe how to solve it!? ... but unfortunately I do not understand that much of how to go on, e.g. with this addr2line ...

[pylkko]

If you have the debug symbols, then try to produce a back trace. https://wiki.debian.org/HowToGetABacktrace

Does it always fail? That is, does every time you do

Code: Select all

systemctl start libvirtd.service 

result in segfault

[jeensg]

Thanks for the advice! SOLVED (at least in the way, that I know now, where the problem resulted from)
Did the following:

Code: Select all

apt install gdb libvirt0-dbgsym
gdb libvirtd
(gdb) set pagination 0
(gdb) run
(gdb) bt

This resulted in:

Code: Select all

#0  0x00007ffff3d7424f in gaiconf_init () at ../sysdeps/posix/getaddrinfo.c:2018

From there I knew that I made changes in /etc/gai.conf, becaue ipv6 sometimes doesn't work with debian-mirrors, so I wanted to exclude them.
Finally I undid my changes to /etc/gai.conf and it works again. Any suggestions where to post this bug, if it is one? Maybe I also did some mistakes in that file, although it worked properly before:

Code: Select all

# Make IPv6 for security.debian.org undesirable
#precedence 2001:a78:5:1:216:35ff:fe7f:6ceb    5
#precedence 2a02:16a8:dc41:100::233      5
#precedence 2001:a78:5:0:216:35ff:fe7f:be4f      5

[jeensg]

Filed Bug here, found here.

[jeensg]

Anybody knows where to file a bug like this? The people from libvirtd/redhat/bugzilla said: The stack trace is showing a crash inside your C library's impl of getaddrinfo(), so I don't think this is libvirt's fault.

[GarryRicketson]

Post by jeensg » 2017-08-02 06:24
Anybody knows where to file a bug like this?

Looks like it was answered there as well:

Filing it against glibc, in your OS distros' bug tracker is your best bet i think.