From owner-freebsd-current  Fri May 23 09:42:50 1997
Return-Path: <owner-current>
Received: (from root@localhost)
          by hub.freebsd.org (8.8.5/8.8.5) id JAA00350
          for current-outgoing; Fri, 23 May 1997 09:42:50 -0700 (PDT)
Received: from phaeton.artisoft.com (phaeton.Artisoft.COM [198.17.250.50])
          by hub.freebsd.org (8.8.5/8.8.5) with SMTP id JAA00343
          for <current@FreeBSD.ORG>; Fri, 23 May 1997 09:42:47 -0700 (PDT)
Received: (from terry@localhost) by phaeton.artisoft.com (8.6.11/8.6.9) id JAA07418; Fri, 23 May 1997 09:41:01 -0700
From: Terry Lambert <terry@lambert.org>
Message-Id: <199705231641.JAA07418@phaeton.artisoft.com>
Subject: Re: cd9660 filesystem slowed down ???
To: bde@zeta.org.au (Bruce Evans)
Date: Fri, 23 May 1997 09:41:01 -0700 (MST)
Cc: current@FreeBSD.ORG, sos@sos.freebsd.dk
In-Reply-To: <199705231314.XAA27893@godzilla.zeta.org.au> from "Bruce Evans" at May 23, 97 11:14:52 pm
X-Mailer: ELM [version 2.4 PL24]
MIME-Version: 1.0
Content-Type: text/plain; charset=US-ASCII
Content-Transfer-Encoding: 7bit
Sender: owner-current@FreeBSD.ORG
X-Loop: FreeBSD.org
Precedence: bulk

> >I have this strange observation under 3.0-current:
> >
> >dd if=/dev/wcd0c of=/dev/null gives about 2MB/sec
> >
> >Mount the CD and then
> >
> >dd if=somefile of=/dev/null gives about 70K/sec
> >
> >Doing the same on 2.2.2 gives 2MB/sec / 1800K/sec...
> >
> >What have we done to degrade cd9660 performance that much ????
> 
> The first dd has the common error of operating on the block
> device instead of the raw device.  This gives a effective block
> size of 2K instead of the size requeseted by dd (512).  2K
> happens to work well for wcd0.  I get almost 1200K/s for an 8x
> drive.

Ah.  Well.  We should all remember to not use 2k blocks on a CDROM
instead of 512b (or the native size, 1k) because then we will get
2MB/sec, and that's (somehow???) undesirable.

8-|


> dd if=/dev/rwcd0c of=/dev/null gives a kernel printf "atapi0.1:
> invalid data direction", a flashing wcd0 LED and a hung IDE
> controller here.

There seems to be a lot of that going around on character devices
these days... 8-(.


> The second dd uses the too-small block size of 512.  Apparently
> read-ahead doesn't work for too-small block sizes in -current.
> I get 50K/sec for a block size of 512 and almost 1200K/sec for
> a block size of 2K.  Check rev.1.28 of cd9600_vnops.c and the
> Lite2 changes.  Not much else has changed.

Try 1K, the native size.  I think you are seeing pessimal rotational
latency.


					Regards,
					Terry Lambert
					terry@lambert.org
---
Any opinions in this posting are my own and not those of my present
or previous employers.