jweldon@renegade.sgi.com (Jack P. Weldon) (06/05/90)
There is a problem with Sun's backoff algorithm in 4.1 mountd, and Sun has
announced patches available for bugs #1035513 and #1035515 in Sun's database
that fix the problem. The gory details follow from a previous posting to this
newsgroup:
From: dana@tread.wpd.sgi.com (Dana Treadwell)
Newsgroups: comp.sys.sgi
Subject: Bug in SunOS4.1 mountd
Organization: Silicon Graphics, Inc., Mountain View, CA
Please take heed if you are contemplating installing SunOS4.1. There
is a bug in Sun's 4.1 mountd which will prevent you from mounting an
SGI filesystem on a SunOS4.1 client. Here's the scoop:
Sun incremented the version number of mountd in 4.1 from 1 to 2. SGI
supports its own version so that showmount(1M) can report on the additional
export options provided by SGI. We chose version number 99 in order not to
collide with Sun's version numbers.
The problem is that Sun's back-off algorithm assumes a continous range of
version numbers, though such a restriction is not documented in the protocol
specification. If
1) the server's highest version number is greater than the Sun
client's highest version number and
2) the Sun client's highest version number is not supported on the server,
the mount request will fail, even though the client and server might both
support a lower version number.
In the case with SGI, the Sun client sees that the SGI server supports #99
(which the Sun client doesn't), tries its highest version number which is #2
(which the SGI server doesn't currently support) and then the Sun client
gives up, failing to back-off and try version #1.
*********************************************************************
The result is that SunOS4.1 clients cannot NFS mount SGI filesystems.
*********************************************************************
We plan to fix this from our side in our next software release. This appears
as bugs #1035513 and #1035515 in Sun's database, if you want to talk to them
about it in the meantime.
Cheers,
Jack P. Weldon (jweldon@sgi.com)
SGI Product Support Engineeringrobert@shangri-la.gatech.edu (Robert Viduya) (06/05/90)
> There is a problem with Sun's backoff algorithm in 4.1 mountd, and Sun has > announced patches available for bugs #1035513 and #1035515 in Sun's database > that fix the problem. The gory details follow from a previous posting to this > newsgroup: ... > The result is that SunOS4.1 clients cannot NFS mount SGI filesystems. Under what conditions does this occur? I mount filesystems from various SGI hosts running 3.2 and 3.2.1 on my SparcStation running 4.1 all the time. Haven't noticed any problems yet. robert -- Robert Viduya robert@shangri-la.gatech.edu Office of Computing Services Georgia Institute of Technology (404) 894-6296 Atlanta, Georgia 30332-0275
sgf@cs.brown.edu (Sam Fulcomer) (06/06/90)
In article <8441@odin.corp.sgi.com> jweldon@renegade.sgi.com (Jack P. Weldon) writes: >There is a problem with Sun's backoff algorithm in 4.1 mountd, and Sun has > >********************************************************************* >The result is that SunOS4.1 clients cannot NFS mount SGI filesystems. >********************************************************************* > >We plan to fix this from our side in our next software release. This appears Hmmm..., Well, while we're waiting for the SGI "break" there're two easy solutions: - Run some other rpc.mountd on the SGI (if you've got src around somewhere...) - run this on the SGI right after rpc.mountd: ------------------------------------------------------------------------------ #include <rpc/rpc.h> main(argc, argv) int argc; char **argv; { /* unregister with the portmapper */ pmap_unset(100005,99); } ------------------------------------------------------------------------------ (compiled by "cc -I/usr/include/sun -I/usr/include/bsd -o unset unset.c -lsun -lbsd -lrpcsvc") Now that wasn't so hard, was it? sgf@cfm.brown.edu _/**/Sam_Fulcomer
dana@tread.sgi.com (Dana Treadwell) (06/06/90)
>> There is a problem with Sun's backoff algorithm in 4.1 mountd... >> The result is that SunOS4.1 clients cannot NFS mount SGI filesystems. > Under what conditions does this occur? I mount filesystems from various > SGI hosts running 3.2 and 3.2.1 on my SparcStation running 4.1 all the > time. Haven't noticed any problems yet. Sun's 4.1 version of automount doesn't use the same backoff algorithm as their 4.1 mountd. Automount uses version 1 of the mountd protocol unless the "posix" mount option appears in the command telling it to use version 2. So you can work around this problem by automounting SGI filesystems. Dana Treadwell dana@sgi.com
jweldon@renegade.sgi.com (Jack P. Weldon) (06/07/90)
In article <41945@brunix.UUCP> sgf@cfm.brown.edu (Sam Fulcomer) writes: >In article <8441@odin.corp.sgi.com> jweldon@renegade.sgi.com (Jack P. Weldon) writes: >>There is a problem with Sun's backoff algorithm in 4.1 mountd, and Sun has >> >>********************************************************************* >>The result is that SunOS4.1 clients cannot NFS mount SGI filesystems. >>********************************************************************* >> > >Hmmm..., Well, while we're waiting for the SGI "break" there're two easy solutions: > [code deleted] No break necessary--it is Sun's bug, not SGI's. We are just trying to create a workaround in our code for those folks who cannot get a fix from Sun. PS: Sun has fixed the bug, and will fedex a new /etc/mount to anyone who has support. Just give them a call. Cheers, Mouthletic (mowth' let ik) Jack P. Weldon adj. Having the useless talent of being able to (jweldon@sgi.com) twist a cherry stem into a knot with one's tongue.