misc/105241: problem with Linux NFS server up/down combined with zombied umount process and a manual mount request

[Roger Jørgensen]

>Number:         105241
>Category:       misc
>Synopsis:       problem with Linux NFS server up/down combined with zombied umount process and a manual mount request
>Confidential:   no
>Severity:       serious
>Priority:       low
>Responsible:    freebsd-bugs
>State:          open
>Quarter:        
>Keywords:       
>Date-Required:
>Class:          sw-bug
>Submitter-Id:   current-users
>Arrival-Date:   Tue Nov 07 12:30:23 GMT 2006
>Closed-Date:
>Last-Modified:
>Originator:     Roger Jørgensen
>Release:        6.1-RELEASE-p6 FreeBSD 6.1-RELEASE-p6 #7
>Organization:
>Environment:
FreeBSD lime.nhn.no 6.1-RELEASE-p6 FreeBSD 6.1-RELEASE-p6 #7: Fri Sep  8 15:29:31 CEST 2006     root at lime.nhn.no:/usr/obj/usr/src/sys/NHN-PE2650_lime  i386

But also seen on FreeBSD 6.0 boxes...
>Description:
Not sure how to describe this or document it but we've seen it several 
times here and always related to the NFS server going down/up again, AND maybe 
most important. A zombie umount command, and then someone issuing mount, 
not on the same filesystem but on another NFS mounts

This just cause the entire system to freeze up, it respond on all ports, 
ping and everything, and as long as you are not try to access any disk 
devices, or commands that access any disks you're fine. If you access any 
disks the command just freeze like the rest of the system. And there are 
nothing in the logs that give any hints that there are troubles somewhere, 
guess the disk-layer freeze before it can be logged :/

You can not log onto the system on the console, not through ssh, nothing 
work basicly and only way out are a hard poweroff/on.

>How-To-Repeat:
Have no clear idea howto recreate it but can give some idea on howto trigger it...

* linux nfs server exporting any directory, freebsd box mounting it.
* Take down the linux NFS server hard or whatever way you wish
* issue a umount request or try to access a nfs share that aren't available anymore, that command MUST be zombied, not possible to kill it
* get the NFS server up and try to remount all shares manual, umount and mount etc. 

And with some luck you might be able to trigger this. Might be just a simple error somewhere that aren't catched and handled correctly.
>Fix:
reboot the box :}
>Release-Note:
>Audit-Trail:
>Unformatted: