VxVM configuration daemon (vxconfigd) crashes during startup on Solaris hosts with NVME devices

Article: 100033777
Last Published: 2017-03-15
Ratings: 0 0
Product(s): InfoScale & Storage Foundation

Problem

The Veritas Volume Manager (VxVM) configuration daemon (vxconfigd) fails to complete the device discovery and crashes. It generates a core dump while starting, including during the system boot process, on Solaris systems that are equipped with NVME devices. The issue persists even when the NVME devices are excluded from the VxVM view.
 

Error Message

During startup, warning messages that are similar this excerpt can be seen in the system messages log:
  ---SNIP---<hostname> vxdmp: [ID 187204 kern.warning] WARNING: VxVM vxdmp V-5-3-2067 dmp_devno_to_devidstr ldi_open_by_dev failed for devno 0xc00000010<hostname> vxdmp: [ID 423856 kern.warning] WARNING: VxVM vxdmp V-5-0-2046 : Failed to get devid for device 0x70526620<hostname> vxdmp: [ID 337865 kern.warning] WARNING: VxVM vxdmp V-5-3-2052 Failed to create minor node for dmpnode 0x14300000000---SNIP---

Core files that are generated by the vxconfigd crash contain a stack that is similar to this excerpt:
  ddl_get_disk_given_path ()ddl_remove_ghost_entries ()ddl_reconfigure_all ()ddl_find_devices_in_system ()find_devices_in_system ()mode_set ()setup_mode ()startup ()---SNIP---

Cause

VxVM assigns incorrect device numbers for NVME devices during the initial discovery and fails to open the affected devices. At a later stage, the cleanup of the NVME device records fails because of the earlier open failure. This results in a device reconfiguration failure and causes vxconfigd to crash due to a NULL pointer dereference.
 

Solution

Code changes have been made so that VxVM calculates the device numbers that are used for opening the NVME devices correctly. Contact our technical support if a hot fix patch for an existing version of VxVM is required.

References

Etrack : 3879972

Was this content helpful?