[lkml]   [1997]   [Mar]   [24]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
Messages in this thread
SubjectRe: ln weirdness
On Mon, 24 Mar 1997, Andrew Walker wrote:

> On Mon, 24 Mar 1997, Gerald Britton wrote:
> > as a normal user, the system lets me do this:
> >
> > ln /etc/shadow /tmp/testfile
> >
> > it then creates testfile as the same permissions and ownership of
> > /etc/shadow, thus i still cannot read it, but should it really be letting
> > me do this? Also, after i create the file, i cannot remove it (since i do
> > not own it). Should it really be doing this?
> >
> [ ... Explanation on link and t bit semantics ]
> Did that make sense? What you are experiencing is correct UNIX practice.
> A lot of people who are new to UNIX (I'm not saying you're a newbie,
> but you obviously weren't aware of this) don't fully grasp the permissions
> stuff, and think they have discovered huge security holes in UNIX. They
> haven't! Its designed that way. Its a feature not a bug.

Actually there is a security problem here (not exactly a hole, since
carefully written userspace programs can avoid it) - an ordinary user can
create a link to say /etc/passwd in /tmp as a name used by (for e.g.)
gcc's temporary files. If root later runs gcc, it may write to this file,
which will overwrite /etc/passwd. This, and a similar problem with
symlinks is addressed in Andrew Tridgell's symlink patch. I'd point you to

But this seems to be an old version that addresses only the symlink, not
the hard link problem.

David Gibson @ The Lorax | New from Microsoft... | THNEED 95
| Which everyone, Everyone, EVERYONE needs.

 \ /
  Last update: 2005-03-22 13:39    [from the cache]
©2003-2014 Jasper Spaans. hosted at Digital Ocean