On Wed, Jan 19, 2005 at 11:44:56AM -0000, Plant, Dean wrote: > I am testing autofs on FC3 with direct mounts read from Solaris NIS maps. > > I have enabled direct mounts by editing DISABLE_DIRECT=1 in > /etc/sysconfig/autofs. > > On the NIS server I have test direct mount of: > > /admin -ro amachine:/export/admin > > Which does not work correctly in FC3 but work's fine in Solaris, > /var/log/messages show's > > cache_ghost: entry in yp:auto_direct not valid map format, key /admin > > When I add the below direct mount, it works as expected. > > /test/admin -ro amachine:/export/admin > > Is this just a feature of the Linux autofs not fully emulating Sun's autofs > or is this a bug? At some point in the past (redhat 7?, I forgot) I tried a similar setup an dcame to the same conclusion. The Linux autofs doesn't want to generate mountpoints directly in / but it works to use an intermediate directory (like /test in your example) to create the mount points in. This may very well be a safety feature; on Solaris, a misconfigure4d automounter can have something mount on e.g. /etc and make your system inoperable. Ort it may just be an unimplemented feature or bug in the Linux version. Anyway, what I did back then was modify the configuration to mount all the Solaris nfs disks under /disks, and where needed, I created symbolic links in / to point to the mountpoint in /disks (needed e.g. because some 3rd party software had hardcoded directory names). David Jansen
