*BSD News Article 32569


Return to BSD News archive

Path: sserve!newshost.anu.edu.au!munnari.oz.au!bunyip.cc.uq.oz.au!harbinger.cc.monash.edu.au!yarrina.connect.com.au!warrane.connect.com.au!kralizec.zeta.org.au!not-for-mail
From: bde@kralizec.zeta.org.au (Bruce Evans)
Newsgroups: comp.os.386bsd.bugs
Subject: Re: 1.1 won't boot on my 486 && slow access of pcfs volumes
Date: 8 Jul 1994 19:52:47 +1000
Organization: Kralizec Dialup Unix Sydney - +61-2-837-1183, v.32bis v.42bis
Lines: 26
Message-ID: <2vj7lf$24c@kralizec.zeta.org.au>
References: <CsKz06.BD0@helios.physics.utoronto.ca>
NNTP-Posting-Host: kralizec.zeta.org.au

In article <CsKz06.BD0@helios.physics.utoronto.ca>,
Jacques Legare <jacques@helios.physics.utoronto.ca> wrote:

>2)  Anyone notice that accesses to a pcfs floppy are deathly slow.

This is because the buffer cache barely works for pcfs floppies.  The
block size for pcfs floppies is only 512 bytes, so it takes 2880
buffers to cache a single 1440K floppy.  FreeBSD allocates only about
345 buffers on a 16M system (and less than 1/4 of that on a 4M system)
so the cache often thrashes.  (FreeBSD assigns only a small number of
buffers to avoid certain memory fragmentation and overallocation
problems.  It calculates the number based on an average block size of
4K, so when the average block size is 8 times smaller, 7/8 of the cache
is wasted.)

The cache usually works better for pcfs on hard disks because the block
size is larger.

>    Even worse a command like 'cp /mnt/drive-a/* /tmp' seems to be
>    uninteruptible and worse seems to prevent access to the target drive.

It probably doesn't really hang.  Probably all the buffers are getting
assigned to reading the floppy so accesses to the hard disk have to
wait for the floppy.
-- 
Bruce Evans  bde@kralizec.zeta.org.au