Bug 1990072

This bug was initially created as a copy of Bug #1927022

I am copying this bug because: the same problem was caused by the OpenIPMI rebase in RHEL, bz1923044


Description of problem:

Activating the collectd-ipmi plugin causes a core dump

Version-Release number of selected component (if applicable):

OpenIPMI-2.0.31-1.el8

How reproducible:

Always

Steps to Reproduce:

--------------------------------
LoadPlugin ipmi

<Plugin ipmi>
  <Instance default>
        Sensor "FAN4"
        Sensor "FAN5"
        Sensor "FANB"
        IgnoreSelected true
        NotifySensorAdd true
        NotifySensorRemove true
        NotifySensorNotPresent true
  </Instance>
</Plugin>

--------------------------------

Actual results:

--------------------------------
Process 11260 (collectd) of user 0 dumped core.
                                               
                                               Stack trace of thread 11261:
                                               #0  0x00007fb40ef0c9d5 raise (libc.so.6 + 0x3d9d5)
                                               #1  0x00007fb40eef58a4 abort (libc.so.6 + 0x268a4)
                                               #2  0x00007fb40eef5789 __assert_fail_base.cold (libc.so.6 + 0x26789)
                                               #3  0x00007fb40ef05026 __assert_fail (libc.so.6 + 0x36026)
                                               #4  0x00007fb40e1d9dd4 valid_fd (libOpenIPMIpthread.so.0 + 0x4dd4)
                                               #5  0x00007fb40e1da374 sel_set_fd_write_handler (libOpenIPMIpthread.so.0 + 0x5374)
                                               #6  0x00007fb40e1db28e add_fd (libOpenIPMIpthread.so.0 + 0x628e)
                                               #7  0x00007fb40e105f15 ipmi_smi_setup_con (libOpenIPMI.so.0 + 0xb2f15)
                                               #8  0x00007fb40e1e3ecc c_ipmi_thread_main (ipmi.so + 0x2ecc)
                                               #9  0x00007fb40f0aa3f9 start_thread (libpthread.so.0 + 0x93f9)
                                               #10 0x00007fb40efd0b53 __clone (libc.so.6 + 0x101b53)
                                               
                                               Stack trace of thread 11260:
                                               #0  0x00007fb40de5f6f4 remoteConnectOpen (libvirt.so.0 + 0x30c6f4)
                                               #1  0x00007fb40de6f7a4 virConnectOpenInternal (libvirt.so.0 + 0x31c7a4)
                                               #2  0x00007fb40de70167 virConnectOpenReadOnly (libvirt.so.0 + 0x31d167)
                                               #3  0x00007fb40e034406 lv_connect (virt.so + 0x4406)
                                               #4  0x00007fb40e0348df lv_init (virt.so + 0x48df)
                                               #5  0x00005642fc871031 plugin_init_all (collectd + 0xf031)
                                               #6  0x00005642fc875bbc run_loop (collectd + 0x13bbc)
                                               #7  0x00005642fc868900 main (collectd + 0x6900)
                                               #8  0x00007fb40eef71e2 __libc_start_main (libc.so.6 + 0x281e2)
                                               #9  0x00005642fc868e1e _start (collectd + 0x6e1e)

--------------------------------


Expected results:

Should work.


Additional info:

Workaround (not really): disable IPMI plugin


According to my (pcahyna) testing, reproducing the bug is very easy: it is enough to install collectd and collectd-ipmi on a machine with an IPMI interface and start collectd (collectd -f) - it will abort after opening /dev/ipmi0.

It is also possible to test on a machine without ipmi:
copy the <Plugin ipmi> block from /etc/collectd.conf to /etc/collectd.d/ipmi.conf and uncomment the <Instance "remote"> part, i.e. append this to the file:

<Plugin ipmi>
#       <Instance "local">
#               Sensor "some_sensor"
#               Sensor "another_one"
#               IgnoreSelected false
#               NotifySensorAdd false
#               NotifySensorRemove true
#               NotifySensorNotPresent false
#               NotifyIPMIConnectionState false
#               SELEnabled false
#               SELSensor "some_sensor"
#               SELSensor "another_one"
#               SELIgnoreSelected false
#               SELClearEvent false
#       </Instance>
        <Instance "remote">
                Host "server.example.com"
                Address  "1.2.3.4"
                Username "user"
                Password "secret"
                #AuthType "md5"
                Sensor "some_sensor"
                Sensor "another_one"
                IgnoreSelected false
                NotifySensorAdd false
                NotifySensorRemove true
                NotifySensorNotPresent false
                NotifyIPMIConnectionState false
                SELEnabled false
                SELSensor "some_sensor"
                SELSensor "another_one"
                SELIgnoreSelected false
                SELClearEvent false
        </Instance>
</Plugin>