[comp.sys.hp] HP <-> Sequent NFS botch

fletcher@cs.utexas.edu (Fletcher Mattox) (09/18/90)

Anybody seen this type of NFS authentication botch?

The server, taklamakan, is an HP 9000/370 running HPUX 7.0.
The client, cs.utexas.edu, is a Sequent Balance 21000 running Dynix 3.0.4.

What's weird about this is:

1.  Once the client is allowed to create the file, presumably
    by using group privileges, it gets the uid right. 
    Yet the client still cannot, for example, chmod the file
    because it thinks the owner is wrong.

2.  It only happens between the HP and the Balance; our Symmetry
    client running 3.0.12 works fine (as do all of our Suns).  There is
    *very* little difference in the NFS client code between Dynix 3.0.4
    and 3.0.12.

3.  This particular NFS problem doesn't happen with an HPUX 6.5 server.  
    It only surfaced after we upgraded to 7.0.


Script started on Thu Sep 13 17:03:40 1990
cs.utexas.edu% whoami
fletcher
cs.utexas.edu% groups
dept operator lusr wheel news uucp src bin
cs.utexas.edu% cd /mnt/fletcher
cs.utexas.edu% df .
Filesystem            kbytes    used   avail capacity  Mounted on
taklamakan:/mnt        35375      11   31826     0%    /mnt
cs.utexas.edu% mount | grep taklamakan:/mnt
taklamakan:/mnt on /mnt type nfs (rw)
cs.utexas.edu% ls -lgd .
drwxr-xr-x  2 fletcher dept         1024 Sep 13 17:03 ./
cs.utexas.edu% cp /dev/null foo
cp: foo: Permission denied
cs.utexas.edu% chmod g+w .
chmod: .: Not owner
cs.utexas.edu% rsh taklamakan chmod g+w /mnt/fletcher
cs.utexas.edu% ls -lgd .
drwxrwxr-x  2 fletcher dept         1024 Sep 13 17:03 ./
cs.utexas.edu% cp /dev/null foo
cs.utexas.edu% ls -lg foo
-rw-rw-r--  1 fletcher dept            0 Sep 13 17:06 foo
cs.utexas.edu% chmod g-w foo
chmod: foo: Not owner
cs.utexas.edu% rm foo
cs.utexas.edu% exit
cs.utexas.edu% 
script done on Thu Sep 13 17:07:19 1990

bailey@math-cs.kent.edu (Jeff Bailey) (09/18/90)

In article <12609@cs.utexas.edu>, fletcher@cs.utexas.edu (Fletcher
Mattox) writes:
> Anybody seen this type of NFS authentication botch?
> 
> The server, taklamakan, is an HP 9000/370 running HPUX 7.0.
> The client, cs.utexas.edu, is a Sequent Balance 21000 running Dynix 3.0.4.
> 
> What's weird about this is:
> 
> 1.  Once the client is allowed to create the file, presumably
>     by using group privileges, it gets the uid right. 
>     Yet the client still cannot, for example, chmod the file
>     because it thinks the owner is wrong.
> 
> 2.  It only happens between the HP and the Balance; our Symmetry
>     client running 3.0.12 works fine (as do all of our Suns).  There is
>     *very* little difference in the NFS client code between Dynix 3.0.4
>     and 3.0.12.
> 
> 3.  This particular NFS problem doesn't happen with an HPUX 6.5 server.  
>     It only surfaced after we upgraded to 7.0.
> ...

We have a similar problem here involving HP9000's and a Sequent B21000.
When using NFS, data seems to have a very difficult time flowing from
the HP's to the Sequent. NFS writes to the Sequent seem to happen in
small bursts with long pauses in between. NFS reads on the Sequent
from the HP's is similar. There is no problem with any other combination
of Suns, HP's, VAXen, or Encore but a Tektronix workstation was behaving
the same way before it died. :-(

---------------------------------------------------------------------
Jeff Bailey (JRB71) (System Administrator)      <bailey@mcs.kent.edu>
Department of Mathematics and Computer Science
Kent State University
Kent - OH 44242

liam@cs.qmw.ac.uk (William Roberts) (09/21/90)

Does it only happen with the HP machine as the client, or do you
get a similar effect when the Sequent is the client and the HP is
the server?

The detailed script showed that "fltecher" is in 8 groups... this is
a worrying number since most NFS implementations only allow for up
to 8 group ids in the NFS credentials. I'd try again after changing
things so that fletcher is in one less group and see if this helps.

Please let me know if you resolve this as we are about to upgrade to
HP/UX 7.0 and our main server is a B21000..... :-{
-- 

William Roberts                 ARPA: liam@cs.qmw.ac.uk
Queen Mary & Westfield College  UUCP: liam@qmw-cs.UUCP
Mile End Road                   AppleLink: UK0087
LONDON, E1 4NS, UK              Tel:  071-975 5250 (Fax: 081-980 6533)

fletcher@cs.utexas.edu (Fletcher Mattox) (09/23/90)

William Roberts writes:

>Does it only happen with the HP machine as the client, or do you
>get a similar effect when the Sequent is the client and the HP is
>the server?

Actually, it happens only with a Sequent client and HP server 
(I thought my script made that clear--maybe not)..

Turns out to be a problem in the Dynix 3.0.4 NFS client code.
It's fixed in 3.0.14.  If you aren't running that, you should
upgrade your Sequent before going to HPUX 7.0.