Return to BSD News archive
Newsgroups: comp.unix.bsd
Path: sserve!manuel!munnari.oz.au!spool.mu.edu!sol.ctr.columbia.edu!e40-008-9.MIT.EDU!vikki
From: vikki@e40-008-9.MIT.EDU (Vikki King)
Subject: [386bsd] Progress!-NFS read problem
Organization: J. Random Misconfigured Site
Message-ID: <1992Sep19.180811.2219@sol.ctr.columbia.edu>
Sender: nobody@ctr.columbia.edu
X-Newsreader: Tin 1.1 PL4
Date: Sat, 19 Sep 1992 18:08:11 GMT
X-Posted-From: e40-008-9.mit.edu
NNTP-Posting-Host: sol.ctr.columbia.edu
Lines: 34
Hi Everyone!
Well, I think I have discovered the nature of the read-big-NFS-files
problem that I and at least one other person mentioned. In my /etc/fstab I
have three entries. The first is the local root file system. The next two
are NFS filesystems from a Novell Netware 3.11 server. When I first tried
mounting these filesystems, I was having major problems reading from most
of the files in them, especially ones over a few 10's of K. Eventually,
I clued in in setting rsize/wsize=4096 for the mount. To implement these
additional parameters during system bootup, I went to /etc/rc and added
the buffer settings to the existing 'mount -a -t nfs' line. I thought
I was all set, but I was still having the read problems on one, but not
both of the NFS filesystems.
I finally discovered that if I changed the order of the last two entries
in /etc/fstab, it was always the filesystem that appeared second/last that
would wind up not being readable. The first NFS mounted filesystem would
always be ok. It seems as if the buffer parameters were being applied to the
first NFS system to be mounted but not the second.
Now, a question. Is this a bug in how rc is executed/being executed
or how it is calling mount, or, have I implemented the buffer parameter
modifications in the wrong manner? My fix for now has been really hackish.
I have created 4 variations of fstab, three of them which contain a single
line with one of the filesystems to be mounted and the fourth containing
all the proper entries, as normal. Instead of running mount directly from
/etc/rc, I have rc calling a script that, one at a time, cp's each of the
three fstabs to /etc/fstab and runs /etc/mount against it. After the three
filesystems have been mounted in this manner, the last thing that's done is the
properly layed out fstab is cp'd to /etc/fstab. It's primitive, but I now seem
to have properly functioning NFS mounts. Does this evoke any startling
revelations from any of you :-)??
-John
jackson@a1.mec.mass.edu