> On Wed, 2005-04-20 at 09:13 -0700, Lever, Charles wrote: > > if the mount option default settings change (say it was 'udp' in > > earlier releases, and is now 'tcp') the meaning of a > missing option on > > the mount command line is different. > > > > so looking at the mount options is a completely unreliable way to > > determine which transport protocol is in use... > > True, but at least this makes some effort at it. The old code did a > UDP/NFSv2 ping no matter what. This is still what happens in > the absence of these mount options. > > Is there some way to pull this info out of the mount program > or the kernel so we could determine the appropriate default > at runtime? That would be the best way to do it.
ok, i've reviewed the code in the kernel that generates the contents of /proc/mounts. this will always show at least the transport protocol and the NFS version. /etc/mtab (and thus the mount command) does not guarantee this; if a mount is done with defaults (ie the options aren't specified), the options in effect will not show up in /etc/mtab. /proc/mounts is modulated by the namespace of the process that is reading it. if the automounter process can see all system mounts, you're golden, but any namespace-related access control will affect the ability for an automounter to read the mount options you're interested in from /proc/mounts. the kernel (post 2.6.12) can do under-the-cover mounts. thus entries may be added to or removed from /proc/mounts over time, and there may be an interesting challenge to find the right entry given this churn. _______________________________________________ autofs mailing list autofs@linux.kernel.org http://linux.kernel.org/mailman/listinfo/autofs
