>Number:         18125
>Category:       kern
>Synopsis:       NFS client doesn't notice "file system full" conditions
>Confidential:   no
>Severity:       serious
>Priority:       medium
>Responsible:    kern-bug-people
>State:          open
>Class:          sw-bug
>Submitter-Id:   net
>Arrival-Date:   Fri Aug 30 08:52:01 PDT 2002
>Closed-Date:
>Last-Modified:
>Originator:     
>Release:        NetBSD 1.6_RC1
>Organization:

LIP6, Universite Paris VI.

>Environment:

System: NetBSD java 1.6_RC1 NetBSD 1.6_RC1 (JAVA) #2: Thu Aug 29 18:45:39 MEST 2002 bouyer@java:/home/NetBSD-1.6/src/sys/arch/sparc64/compile/JAVA sparc64
Architecture: sparc64
Machine: sparc64
>Description:
	A 1.6 NFS client writing to a NFS server doesn't notice when the
	remote file system is full, loosing data. I don't think the server
	(also running 1.6) is in cause because it works as expected with
	1.5.3 or linux clients.
>How-To-Repeat:
	java:/mnt/tmp>df -k .
	Filesystem          1K-blocks     Used     Avail Capacity  Mounted on
	borneo:/usr/mayotte   1062487   951405     57957    94%    /mnt
	java:/mnt/tmp>dd if=/dev/zero of=toto bs=1m count=100
	100+0 records in
	100+0 records out
	104857600 bytes transferred in 15.289 secs (6858368 bytes/sec)
	java:/mnt/tmp>ls -l
	total 115932
	-rw-r--r--  1 root    wheel      5663 Aug 26 16:36 install.log
	-rw-r--r--  1 bouyer  wheel  59310080 Aug 30 19:21 toto

	note that the dd process didn't get any error for the writes,
	and claim to have written 100M when there is only 60M really
	written to disk. On the server side a
	borneo /netbsd: uid 373 comm nfsd on /usr: file system full
	proves that the server noticed the problem
>Fix:
	unknown.
>Release-Note:
>Audit-Trail:
>Unformatted: