Re: Mount-point vanishes

[H. Peter Anvin]

Jeremy Fitzhardinge wrote:
> 
> On 18-Feb-00 H. Peter Anvin wrote:
> > These kinds of issues are part of why mount-trees are *very* complex.
> > There are all kinds of funny race conditions involved, too, since most
> > of these operations can't be performed under lock.
> 
> There is a lock, of sorts, because the autofs filesystem blocks processes from
> the tree while it is being manipulated.  The issue here is that the daemon will
> pretend things worked even if not all the mounts succeeded, which is probably a
> bug.
> 

Yes, unless you do sub-automounting you need to present the whole tree
or nothing at all.

        -hpa