Bug 1462134

Description of problem:

The "all_devs" option was added to "/etc/multipath.conf" file to set no_path_retry count for all devices to 5. This option was working fine with device-mapper-multipath-0.4.9-85.el7_2.5.

But once the system was updated to any of the following device-mapper-multipath version, a rescan of multipath device maps, reload of multipathd service starts getting failed with "segmentation fault":

Affected versions of device-mapper-multipath:
  device-mapper-multipath-0.4.9-99.el7
  device-mapper-multipath-0.4.9-99.el7_3.1
  device-mapper-multipath-0.4.9-99.el7_3.3


$ less /etc/multipath.conf
[...]
# Remove devices entries when overrides section is available.
devices {
    device {
        # These settings overrides built-in devices settings. It does not apply
        # to devices without built-in settings (these use the settings in the
        # "defaults" section), or to devices defined in the "devices" section.
        # Note: This is not available yet on Fedora 21. For more info see
        # https://bugzilla.redhat.com/1253799
        all_devs                yes
        no_path_retry           5
    }
[...]


After updating the device-mapper-multipath package to any of the above affected versions, we get following  "segmentation fault" errors:


[root@testsystem ~]# multipath -v2
Segmentation fault


Jun 15 17:10:06 testsystem kernel: multipath[15506]: segfault at 0 ip 00007f60c204a551 sp 00007ffcd1bbee50 error 4 in libc-2.17.so[7f60c1fab000+1b7000]


[root@testsystem ~]# systemctl restart multipathd
[root@testsystem ~]# 


/var/log/messages:

Jun 15 17:11:42 testsystem systemd: Starting Device-Mapper Multipath Device Controller...
Jun 15 17:11:42 testsystem kernel: multipath[15514]: segfault at 0 ip 00007f4bd8b51551 sp 00007ffd37e89eb0 error 4 in libc-2.17.so[7f4bd8ab2000+1b7000]
Jun 15 17:11:42 testsystem kernel: multipathd[15520]: segfault at 0 ip 00007f8efd097551 sp 00007ffd99598e70 error 4 in libc-2.17.so[7f8efcff8000+1b7000]
Jun 15 17:11:42 testsystem systemd: Started Device-Mapper Multipath Device Controller.
Jun 15 17:11:42 testsystem systemd: multipathd.service: main process exited, code=killed, status=11/SEGV
Jun 15 17:11:42 testsystem systemd: Unit multipathd.service entered failed state.
Jun 15 17:11:42 testsystem systemd: multipathd.service failed.




Version-Release number of selected component (if applicable):
Affected versions of device-mapper-multipath:
  device-mapper-multipath-0.4.9-99.el7
  device-mapper-multipath-0.4.9-99.el7_3.1
  device-mapper-multipath-0.4.9-99.el7_3.3


How reproducible:
o Always


Steps to Reproduce:

1. Update device-mapper-multipath package to one of the affected versions listed above.

2. Enable "no_path_retry" count for all devices by using all_devs option as shown below:

	$ less /etc/multipath.conf
	[...]
	# Remove devices entries when overrides section is available.
	devices {
	    device {
		all_devs                yes
		no_path_retry           5
	    }
	[...]

3. Run multipath -v2, or "systemctl restart multipathd".
   This would result in "segmentation fault" errors on terminal, logs.

Actual results:
o multipathd gets crashed while processing "no_path_retry", "queue_if_no_path" option with "all_devs  yes" set

Expected results:
o multipathd does not get crashed

Additional info:

o Currently this issue could be work-around by following 2 steps:

   1. Set custom "no_path_retry" count for device section 
      corresponding to specific SAN/Storage arrays.

   2. Also, disable "no_path_retry fail" with "all_devs" option

	devices {
	    device {
		all_devs                yes
		no_path_retry           fail ###
	    }
	}