From mboxrd@z Thu Jan  1 00:00:00 1970
Message-ID: <9c4064ddc2535c7f841bdd71b0cb1ae2@collyer.net>
To: 9fans@cse.psu.edu
Subject: Re: [9fans] Changing disks on fs64 server
Date: Tue, 25 Jul 2006 16:12:31 -0700
From: geoff@collyer.net
In-Reply-To: <071520061223.13070.44B8DE2E0001A0CE0000330E2207021053050E0C040A900E9FD209@comcast.net>
MIME-Version: 1.0
Content-Type: text/plain; charset="US-ASCII"
Content-Transfer-Encoding: 7bit
Topicbox-Message-UUID: 8db4dfcc-ead1-11e9-9d60-3106f5b1d025

Copydev copies using the file server's block size; if it's 4K or
greater, the copy should be relatively quick.  It should also report
progress so you can gauge how long it will take.

It's possible that you don't have a bus-mastering IDE controller.
Even if your disks are capable of DMA, the file server will end up
doing programmed I/O, which is much slower.  Look for this message in
your file server console log:

	h0: disabling dma: not on a busmastering controller

Also check that you haven't set conf.idedma to 0, like this:

fs/9fsfs.c:		conf.idedma = 0; /* for old machines */

If conf.idedma is non-zero, you should see ` and IDE DMA+RWM' in your
file server log:

	Plan 9 63-bit file server with 4-deep indir blks and IDE DMA+RWM