Re: [RFC][PATCH 1/8] namespace: use global namespace semaphore

[raven@themaw.net]

On Fri, 3 Jun 2005, Miklos Szeredi wrote:

> > > > > 
> > > > > This way the autofs mount won't be followed by the mount program,
> > > > > since looking up '.' does not follow mounts.
> > > > 
> > > > But a path walk must be carried out in order to set pwd (or
> > > > chroot), at which time the mount is carried out and remains busy
> > > > due to the reference that is held.
> > > 
> > > Huh?  Path walk of '.' will do nothing.  It does not follow mounts or
> > > anything.
> > 
> > Does "chdir ." make sense. Looks a bit like a noop to me.  Surely if
> > your current working directory is "." you've had to set it with a
> > provided path at some point.
> 
> OK, sorry, I misunderstood you.
> 
> But still I don't see why it is a problem if you have a reference to
> the mountpoint dentry/vfsmount.  You already have a reference with the
> mounted autofs, so adding another reference by opening a file, and
> another by changing cwd to it won't change anything at all.

I think we`ve lost track of what we originally started discussing here.
I think we started discussing why I needed the nameidata struct to be upto 
date when the follow link method was called, it simply isn't when a mount 
is followed at the end of the do_lookup call in link_path_walk.

> 
> > > So for direct mounts, the sibling mount (as opposed to the overmount)
> > > has the advantage of not having to do ugly recursion checking hacks in
> > > autofs.
> > 
> > I've found that all I really need to do is check that the follow_mount is 
> > successful and respond accordingly.
> 
> I don't see how that would work.

Seems to work fine so far.

> 
> Don't you have to somehow special-case the mount program to allow that
> to succeed in path lookup, but make all other lookups block until the
> mount is finished?

I haven't found that to be the case. There are a couple of mechanisms for 
this.

First is a in kernel module waitq for daemon notification that has been 
well tested over time. I've done quite a bit of work on the interaction 
between expire events and lookups. There are still a couple of patches I 
need to submit but it's getting there.

Second the user space daemon is implemented as an FSM which implies a 
considerable amount of serialisation wrt kernel communication through the 
pipe.

Ian