Heads Up: NFS clients can now fail "mount -u -o udp..."

Rick Macklem rmacklem at uoguelph.ca
Wed Jan 25 00:32:03 UTC 2012


r230516, which was just committed to head/current, will result in
a "mount -u -o udp /path" failing if the mount is using an
rsize/wsize/readdirsize of greater than 16384 (which will be the
case of a default sized TCP mount).

As such, specifying "udp" or "mntudp" options in the /etc/fstab
entry for "/" on a diskless NFS client could result in the "mount -u"
failing. I'd suggest that "udp" and "mntudp" be avoided for this
case, if you are using the default of TCP for the root mnt done
when the diskless client is booted.

The discussion on freebsd-fs@ seemed to indicate returning an error
was preferable to silently ignoring the "udp/mntudp" option for this

Hopefully this won't cause many people grief, rick
ps: Since doing the above could cause errors with both the NFS
    clients for the root, I suspect few do the above.

More information about the freebsd-current mailing list