Re: [PATCH] utils/mount: canonicalize mountpoints

[Samuel Thibault]

Gabriele Giacone, le Sun 15 Jun 2014 02:23:49 +0200, a écrit :
> On Sun, Jun 15, 2014 at 1:59 AM, Samuel Thibault
> <samuel.thibault@gnu.org> wrote:
> > Gabriele Giacone, le Sun 15 Jun 2014 01:28:06 +0200, a écrit :
> >> ... this one would fix malformed mount/umount input paths (another
> >> source. Any else? raw settrans). It would make first and last hunks in
> >> pbuilder.diff [0] useless.
> >>
> >> [0] https://lists.debian.org/debian-hurd/2014/06/msg00032.html
> >
> > Ok, so now the goal is known.  How is it achieved?
> >
> > I'm sorry, but either you explain in details, or I have to take the time
> > to work it out by myself, it's as simple as that.  It's already late for
> > me, and I have to wake up early tomorrow, etc.
> 
> I thought it was already detailed enough: first and last hunks useless
> if patch is applied are:

That's only the goal. How does canonicalizing the paths actually
fix that?  The problem is that canonicalizing paths will have other
consequences, so we need to be convinced that it is the proper way to
fix the goal. It's the explanation you provided below that is needed.

> Such hunks are workarounds at the moment because hurd mounts paths by
> keeping superfluous '/' and '.' in pathnames if specified at mount
> time, which need to be specified at umount time as well to umount
> successfully.

But then that won't fix the case when people use settrans by hand, so
perhaps it should be rather done there?

Actually, I don't understand why canonicalizing the mountpoint in mount
would be needed, does it really fix something?  Is the mountpoint, as
provided on the mount command line, really recorded somwhere?  Isn't it
the device part (in the bind case, the source directory) which would
need a canonicalization?  Then, shouldn't it be perhaps firmlink which
should canonicalize its source directory?

> Patch simply makes mount/umount utilities fix through realpath
> mountpoints/device pathnames to pass to settrans, making workarounds
> above useless.

That's a way to fix it, yes, but is it really the proper way?  It has
other consequences on their semantic, so we have to weigh that.

Samuel