From mboxrd@z Thu Jan  1 00:00:00 1970
Message-ID: <32a9bc49907ef0a12031cb1d17f4d17c@terzarima.net>
To: 9fans@cse.psu.edu
Subject: Re: [9fans] adding storage to ken's fs
From: Charles Forsyth <forsyth@terzarima.net>
In-Reply-To: <5d408ee5203752b249ce270e2a105bde@collyer.net>
MIME-Version: 1.0
Content-Type: text/plain; charset="US-ASCII"
Content-Transfer-Encoding: 7bit
Date: Tue,  3 Feb 2004 11:02:52 +0000
Topicbox-Message-UUID: cc130ea6-eacc-11e9-9e20-41e7f4b1d025

>>worm as appropriate.  On the other hand, I'm not sure how the
>>fake-worm's bitmap would interact with adding more space.

that's the thing you need to watch: the bitmap is at the end
of each f device, so you need to define your extended configuration
to preserve it (or you'll have a big disaster)

here's the (possibly) relevant part of my home configuration
	filsys main cp(w1.2.0)0.20(fp(w1.2.0)20.80fw1.3.0fw1.4.0)
	filsys dump o

note that it concatenates several f devices to form the worm device.
that way each existing bitmap was preserved.  alternatively you could add
some code to the fs kernel to copy it, or copy it by booting a terminal/cpu
on the machine temporarily and fiddling with dd and the /dev/sd*/data
to shuffle the blocks with the bits, but the latter two options are harder.

aha! but there's still a catch: you need to initialise the bitmap to zero first.
ream is NOT what you want.  i think i ended up doing a special fs kernel
to devream just the new device, not the file system (which would empty the super block),
but it was a long time ago.  it should be enough to clear a large chunk
of the new device to zero using a terminal/cpu kernel.

with a real worm (eg, jukebox) all that's required when adding platters is to
extend the configuration to refer to them, and it uses them.