kern/47340: NFS resend problem with stale NFS file handles

[reinoud%NetBSD.org@localhost]

>Number:         47340
>Category:       kern
>Synopsis:       NFS resend problem with stale NFS file handles
>Confidential:   no
>Severity:       serious
>Priority:       medium
>Responsible:    kern-bug-people
>State:          open
>Class:          sw-bug
>Submitter-Id:   net
>Arrival-Date:   Sun Dec 16 18:45:01 +0000 2012
>Originator:     Reinoud Zandijk
>Release:        NetBSD 6.0_BETA2
>Organization:
NetBSD
        
>Environment:
        
        

        Two NetBSD systems; this one, the NAS running the NFS server. The other
        machine, heethofdje, mounts a /home with:

        nas:/home /home nfs rw,soft,intr,tcp 0 0

        'heethoofdje' runs: NetBSD heethoofdje.13thmonkey.org 6.0_BETA2 NetBSD 
6.0_BETA2 (GENERIC)      i386
        `heethoofdje' runs a linux binary version of openoffice3-bin-3.2.1nb2

System: NetBSD nas.13thmonkey.org 6.0_BETA2 NetBSD 6.0_BETA2 (MARVELL_NAS) #6: 
Wed Jul 11 10:23:35 CEST 2012 
root%nas.13thmonkey.org@localhost:/usr/sources/cvs.netbsd.org/src-6/sys/arch/evbarm/compile/MARVELL_NAS
 evbarm
Architecture: arm
Machine: evbarm
>Description:
        

        In normal usage, OpenOffice crashed and started recovering a document
        for restart probably.

        The NAS became unresponsive; pinging/telnet to 80 worked but ssh and
        named were almost not accessablei/usable. NFS iseems to work fine. It
        turned out that the the network was flooding:

        19:22:43.022146 IP 192.168.1.2.3203978003 > 192.168.1.1.nfs: 120 read 
fh 16,4/33664198 17720 bytes @ 0
        19:22:43.022491 IP 192.168.1.1.nfs > 192.168.1.2.3203978003: reply ok 
32 read ERROR: Stale NFS file handle
        19:22:43.022533 IP 192.168.1.2.3203978004 > 192.168.1.1.nfs: 120 read 
fh 16,4/33664198 17720 bytes @ 0
        19:22:43.022874 IP 192.168.1.1.nfs > 192.168.1.2.3203978004: reply ok 
32 read ERROR: Stale NFS file handle
        19:22:43.022916 IP 192.168.1.2.3203978005 > 192.168.1.1.nfs: 120 read 
fh 16,4/33664198 17720 bytes @ 0
        19:22:43.023261 IP 192.168.1.1.nfs > 192.168.1.2.3203978005: reply ok 
32 read ERROR: Stale NFS file handle
        .....
        etc. etc.

>How-To-Repeat:
        
        Unknown 
>Fix:
        Explicit killing the hung OpenOffice fixed the storm.

        

>Unformatted: