Cannot umount /usr filesystem (ALWAYS "busy")

[Randy Davis]

In article <1990Jun13.184338.26482 at pmafire.UUCP> rickf at pmafire.UUCP (rick furniss) writes:
|[experience in UNIX for 10 years]

  Interesting...  UNIX was not available on the machines you mention until
early to mid 1980s.  The first one *I* saw was a Plexus in 1983.

|  Name ONE or more Unix versions that install /usr as MOUNTABLE ??

  ALL stock AT&T versions as used on their 3B* series of computers from
release 2 and release 3 of System V, all AT&T licensed versions of 3.x I have
seen, which includes ISC's and a few others.  Since UNIX was written by AT&T
(or employees therof :-) as an operating system for use on larger systems on
the order of mainframes and minicomputers, your experience in desktops
computers does not really apply to the *origins* of UNIX, since UNIX was not
originally designed for use on such.

  If you don't beleive the references I cite above, the *name* itself is a
clue, as /usr was where user support programs, etc.. was stored and was by
definition not required for basic operation of the operating system.  Hence
the reason for single-user mode and the fact that /usr was not mounted in
the normal single-user mode that early machines booted into by default (back
when the machines were of considerable size and usually had physical security).

|   I,ve never seen this as an option either, though I have seen additional
|/u, /user ,/users partitions at install time.  What Version, source do you
|state does this ?  Every one I know has had to install first, then re-partition
|to get /usr as a mountable FS.

  The wording here is a bit unclear to me.   All versions I have EVER installed
normally set up the /usr partition as a separate, mounted filesystem outside
of /.  The only exception I have seen to this is certain versions designed for
the 386 and smaller machines (most notably Dell UNIX 1.1) that detects the use
of a small hard drive, I think the change-over point is 40 MB, and defaults to
no separate /usr partition.

  Versions?  AT&T System V Release 2, 2.1, 3.0, 3.1, 3.1.1, 3.2 in particular,
and I believe I remember working on a few release 1 machines also - in which
case the /usr filesystem was mounted on those also (or I would have remembered
it, since it is not common for it to be integral to root, at least in the
minicomputer enviroment).

|  My experience installing Unix is limited to Xenix & AT&T Unix derivitives 
|for Intel, Tandy & Plexus machines.  Plexus had the first ATT source license
|as far as I know.

  Probably, as I remember a Plexus in the Lab when I first started working for
AT&T in 1983.  I think the key difference between yours and my experiences are
the size of machines.  I started on mini's such as the AT&T 3B2, 3B5, 3B15, and
3B20; where the hard disk size was normally a lot greater (read: almost always
MUCH greater) than you found on any PC or home machine of the same time.  By
your statement above, your experience is primarily on microcomputers, which
UNIX was NOT ORIGINALLY designed for.  Hence, the reason I tend to doubt your
assertions of the origins and original designs of UNIX.

|  Be it by design or what ever, using /usr as a non-root filesystem will
|create some problems that need to be addressed.

  I would agree that this is so, had /usr not originally been designed to be
a mounted filesystem.  Since /usr WAS ORIGINALLY ALWAYS a separately mounted
filesystem, this issue was already addressed in the ORIGINAL DESIGN.  I AGREE
that some person who expects them to always be the same filesystem could have
since screwed up the works, but that does not change the original design of
the /usr filesystem, which is the reason for this debate.

|  I stand by my guns, /usr was never originaly designed as a non-root FS.

  And I say you don't know what you are talking about.

|  /dev, /usr, /etc, /tmp, /bin, /lib, are all considered Unix structure
|filesystem directories.  Either utilities, or kernel should be able to
|count on these being available at all times in multiuser states.

  AHA!!!  TRUE.  In *multiuser* states, you are correct...  What about in
*single* user states, which is what UNIX originally booted into??  Originally,
UNIX machines ALWAYS booted into single-user, requiring a system administrator
to perform filesystem cleanup, etc.. by hand, THEN change init levels to
multi-user.  The UNIX autoboot procedure is a relatively new addition to UNIX -
in the last seven years (in its 20-year life so far).  I believe Plexus was
the first with a semi-automated boot procedure, yet every Plexus I saw was
a single user machine, so this really doesn't apply.

|  SU, or single user modes have minimal requirments so that multiuser mode
|can be configured, fixed, or made ready.

  True.

|  If anyone can dispute this, please correct me with some history/facts.

  Just check into ANY release from AT&T (the originator of UNIX)....  Perhaps
you need to provide some facts from a little further back than the
microcomputer age of UNIX, since micros able to run UNIX are a little too
recent to provide ANY clue as to the origins of UNIX.

  It really doesn't matter to me...  I *know* I am correct, and if you *know*
you are correct, we can just go our separate ways thinking the other doesn't
have a clue as to what they are talking about.  But, from the machines you've
listed above, I would say that you need to delve a little further...

Randy Davis					UUCP: rjd at ninja.dell.com

--