H. Peter Anvin on Tue 2/01 22:10 -0800:
> And daemon?
Sorry, 3.1.4.
> Right. That would be a good example on how autofs can expose NFS
> bugs... static mounts just aren't created that often, and usually are
> only destroyed immediately before the machine is rebooted. A lot of
> corner cases in umount() especially can go unnoticed.
Well...the reason I thought autofs might be involved is because I saw a
bug report of "RPC: garbage...retrying...EIO" that appeared on much
earlier knfs list archives (1998), and one of the explanations a knfs
author suggested was that amd was munging RPC packets. I don't know
enough about autofs to gather whether it could cause this problem, but
it seemed to me that if it made sense for amd to be responsible, then so
could be autofs.
The fact that it appeared so early on mailing lists and still is related
to problems indicates to me that it could be a long-standing bug hid by
some weird border case that takes time and strange or erratic
circumstances to reproduce.
