Panic stack shows as below.
vxdmp:dmp_signal_vold+0x1ac()
vxdmp:dmp_throttle_paths+0x440()
vxdmp:dmp_process_stats+0x240(0x0)
vxdmp:dmp_daemons_loop+0x164(0x0, 0x0)
unix:thread_start+0x4()
vxdmp:dmp_signal_vold+0x6c()
vxdmp:dmp_throttle_paths+0x440()
vxdmp:dmp_process_stats+0x240(0x0)
vxdmp:dmp_daemons_loop+0x164(0x0, 0x0)
unix:thread_start+0x4()
- VxVM 6.0.3.200 and 6.0.5 and 6.1 that includes Etrack 3358407 fix.
(Etrack 3358407 issue was occurred on VxVM 6.0.1 and 6.0.3)
- VxVM 6.2 that includes Etrack 3736633 fix.
Dmp i/o statistic function (dmp_daemons_loop()) calls vxio function even though the vxio driver has been unloaded by vxvm-shutdown script. Etrack 3358407 fix "vxdmpadm iostat stop" remains incomplete for this issue.
Workaround
-comment out modunload process in /lib/svc/method/vxvm-shutdown
Details of changes:
bash-2.05$ diff ORIGINAL_vxvm-shutdown Fixed_vxvm-shutdown
104,110c104,110
< for v in vxspec vxio vxdmp
< do
< mid=`/usr/sbin/modinfo|/usr/bin/grep -w "$v"|/usr/bin/awk '{print $1}'`
< [ -n "$mid" ] && {
< /usr/sbin/modunload -i $mid >/dev/null 2>&1 || break
< }
< done &
---
> #for v in vxspec vxio vxdmp
> #do
> # mid=`/usr/sbin/modinfo|/usr/bin/grep -w "$v"|/usr/bin/awk '{print $1}'`
> # [ -n "$mid" ] && {
> # /usr/sbin/modunload -i $mid >/dev/null 2>&1 || break
> # }
> #done &
- 6.0.5.300 fixed this issue.
vm-sol11_sparc-Patch-6.0.5.300 (for solaris 11)
vm-sol10_sparc-Patch-6.0.5.300 (for solaris 10)
- 6.1.000.100 fixed this issue.
vm-sol11_sparc-6.1.0.100 (for solaris11)
vm-sol10_sparc-6.1.0.100 (for solaris10)
- 6.2 approached in different way to address this issue, but panic re-occured and reported in e3847745.
For Solaris 11, 6.2.1.200 is published.
vm-sol11_sparc-Patch-6.2.1.200
For Solaris 10, private hotfix was created, if you need this hotfix, please contact Veritas Technical Support.