home *** CD-ROM | disk | FTP | other *** search
- Path: sparky!uunet!olivea!sgigate!sgi!rhyolite!vjs
- From: vjs@rhyolite.wpd.sgi.com (Vernon Schryver)
- Newsgroups: comp.sys.sgi
- Subject: Re: Indigo as server for a RS600 client - was: NFS links from RS6000 server to SGI client
- Message-ID: <sk0bfes@rhyolite.wpd.sgi.com>
- Date: 20 Nov 92 19:28:29 GMT
- References: <34067@adm.brl.mil> <1992Nov14.021925.14641@ultra.com> <1992Nov20.172633.26750@cc.ic.ac.uk>
- Organization: Silicon Graphics, Inc. Mountain View, CA
- Lines: 44
-
- In article <1992Nov20.172633.26750@cc.ic.ac.uk>, vulture@imperial.ac.uk (Thomas Sippel - Dau) writes:
- > In article <1992Nov19.230854.5523@sunova.ssc.gov>, lroberts@bottom.ssc.gov (Lee Roberts) writes:
- > -
- > - Apparently, Sun's tar performs a chown on the extracted file before writing
- > - the file contents. On a BSD system such as the Sun, this chown fails for the
- > - non-superuser and uid/gid match the user. When writing to the SGI NFS server,
- > - the chown command succeeds (it's being performed on a System V system) and
- > - changes the ownership to match the uid/gid in the tar archive. Since the
- > - file ownership has changed, tar cannot write the contents.
- >
- > Sounds like "standard compliance by cock-up", which of course does not make
- > it unbelievable. The POSIX standard (IEEE 1003.1) actually addresses this
- > to some length on pp 274-9, section B10, i.e. in the rationale rather than
- > the formal standard)
- >
- > The gist there is that for system security (if nothing else) tar processors
- > should behave differently when run by root than when run by a user, users
- > should not, for example, be able to create special files. Just relying on
- > the chown to fail when executed for a user seems like a broken tar to me.
-
-
- A process can be running as root on the client, but not be root as far
- as the NFS server is concerned.
-
-
- You don't really want each and every application including backup or
- archive utility to duplicate the kernel suser() check. The messiness
- of real, effective, and other flavors of UID and GID are not the sort
- of thing you really want to spread into every application that might be
- called upon to do things that are priviledged.
-
- Moreover, creating a special file is innocent and safe in some
- circumstances. Consider creating a named pipe. Do you want all backup
- and archiving utilities to avoid making nodes just in case the thing is
- a block or character device but not if it is a directory, simple file,
- or named pipe, but ok if it is a socket and who knows for new kinds of
- nodes?
-
- Should the backup or achiving utility check to see that it is sufficiently
- priviledge to write ordinary files where directed? Or should it just
- let the operating system worry about it?
-
-
- Vernon Schryver, vjs@sgi.com
-