>Number: 167266
>Category: kern
>Synopsis: ZFS + new NFS export (sharenfs) leads to NAMEI leak
>Confidential: no
>Severity: serious
>Priority: medium
>Responsible: freebsd-bugs
>State: open
>Quarter:
>Keywords:
>Date-Required:
>Class: sw-bug
>Submitter-Id: current-users
>Arrival-Date: Tue Apr 24 14:20:13 UTC 2012
>Closed-Date:
>Last-Modified:
>Originator: Oliver Brandmueller
>Release: 9-STABLE
>Organization:
>Environment:
FreeBSD silver 9.0-STABLE FreeBSD 9.0-STABLE #9: Sun Apr 8 13:11:22 CEST 2012
root@silver:/usr/obj/usr/src/sys/BLACK amd64
>Description:
If you NFS export a ZFS, create/remove of a file or directory lead to the leak
of a NAMEI path buffer on the NFS server.
This happens with new nfs (only NFSv3 tested) only, using the old NFS server
does not trigger the NAMEI path buffer leak.
An exported UFS on the same machine/environment does not trigger the behaviour.
>How-To-Repeat:
rc.conf:
nfs_server_enable="YES"
nfs_server_flags="-u -t -n 64"
mountd_enable="YES"
mountd_flags="/etc/zfs/exports"
rpcbind_enable="YES"
lockd_enable="YES"
(mountd_flags not needed, if zfs_enable is set)
Assuming, you have a USB stick or similar as /dev/da0 (it will probably
even work with an md device):
zpool create exptest /dev/da0
zfs sharenfs="-maproot=0 localhost" exptest
# check with showmount -e exporting worked as expected
mount localhost:/exptest /mnt
cd /mnt
vmstat -z | fgrep NAMEI
echo test > foo
vmstat -z | fgrep NAMEI
rm foo
vmstat -z | fgrep NAMEI
you may repeat the creation/removal of the file several time and you will
notice the NAMEI count increasing by one on each delete operation, no matter if
file or directory.
>Fix:
No fix known.
Workaround: using old NFS server.
>Release-Note:
>Audit-Trail:
>Unformatted:
_______________________________________________
freebsd-bugs@freebsd.org mailing list
http://lists.freebsd.org/mailman/listinfo/freebsd-bugs
To unsubscribe, send any mail to "freebsd-bugs-unsubscr...@freebsd.org"
