serious issue with automount daemon!

[yemi]

I am preparing to move my Xserve cluster from Panther server to Tiger server.
and I would really like to migrate from the unsupported ‘amd’ NFS automounter to Apple’s automount daemon.

I have spent the past few days reading up on ‘automount’ and trying various test configurations. Our production mount map may contain 1000’s of NFS mount points so a solid automounter is essential.

The simplest automount set-up is to provide the map in the form of a local file.
automount creates the mountpoint parent directory tree automatically and cd’ing into a mountpoint directory will trigger the mount.
However, it is also possible to trigger mounting by simply running ‘ls -l’ on the parent directory of a mount point! VERY VERY SCARY.

This means that an ‘ls -lR’ (recursive) of my directory tree (ie. ‘/nfs’)
will trigger all of the mounts. BAD BAD BAD.

amd, autofs and other NFS mount daemons have options that prevent this. What’s the solution for automount?

—-
Yemi

[isaacvetter]

yemi,

I think that Apple’s automount man page addresses this:

To make the “trigger” symbolic links used by automount distinguishable
from normal symbolic links, the sticky bit is set in the mode flags for
the link. Programs which would normally traverse symbolic links can test
for this bit and avoid triggering the mount. Various parts of the sys-
tem, including ls(1), have been modified in this way.

Isaac Vetter

[Author]

Posts