Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Posting-Version: version B 2.10.1 6/24/83; site bgsuvax.UUCP
Path: utzoo!watmath!clyde!burl!ulysses!mhuxr!mhuxn!ihnp4!cbosgd!osu-eddie!bgsuvax!schaefer
From: schaefer@bgsuvax.UUCP (Stephen Schaefer)
Newsgroups: net.bugs.4bsd
Subject: I-node table
Message-ID: <816@bgsuvax.UUCP>
Date: Thu, 21-Nov-85 15:10:17 EST
Article-I.D.: bgsuvax.816
Posted: Thu Nov 21 15:10:17 1985
Date-Received: Sat, 23-Nov-85 06:06:57 EST
Organization: Bowling Green State University, OH
Lines: 25

4.2BSD:

Our inode table is losing slots.  pstat normally goes through the
inode table sequentially, reporting only on those slotw with non-zero
reference counts, and assumming that the others are free.  I modified
pstat to walk through the free list, printing the addresses.
Yesterday, just after a reboot of our VAX 11/785, I got a healthy
report, something like 129/788 active inodes, 659 in the free list.
This morning, I looked again and got 165/788, 456 in the free list -
165 + 456 = 621.  Arggh.  I started looking at this because we've been
getting "inode table: full" messages on the console, and the pstat at
the time told me there should have been ~600 free.

I've got the bug list from Mt. Xinu.  A few of the messages have to do
with the inode table, but none directly with the symptom of a
corrupted free list.  I'll be installing those fixes one at a time
until the problem goes away, from simplest to most complex.  Meanwhile
I appeal to the net, asking if anyone has a better guess as to what's
plaguing us.

	Thanks in advance,

	Stephen P. Schaefer
	schaefer@bgsu.CSNET
	...!cbosgd!osu-eddie!bgsuvax!schaefer