home
***
CD-ROM
|
disk
|
FTP
|
other
***
search
/
AMIGA PD 1
/
AMIGA-PD-1.iso
/
NetBSD
/
docs-netbsd
/
Mailinglist-Archive
/
1994-08.gz
/
1994-08
/
000162_owner-current-u…s.berkeley.edu_Thu Aug 4 01:32:28 1994.msg
< prev
next >
Wrap
Text File
|
1994-10-16
|
2KB
|
36 lines
From: der Mouse <mouse@Collatz.McRCIM.McGill.EDU>
To: current-users@sun-lamp.cs.berkeley.edu
Subject: Re: how to concatenate disks
Sender: owner-current-users@sun-lamp.cs.berkeley.edu
>> [the following should probably be in some sort of FAQ, so people
>> don't have to go to the 'source', as i did.]
>> How To Concatenate Disks
>> --- -- ----------- -----
> No, it should be a man page.
Preferably, but I'd take a doc file over nothing.
However, I'm curious. What's machine-dependent about this sort of disk
trickery? More specifically, why is the ccd stuff everyone is talking
about in the hp300 port, rather than over in some machine-independent
part of the kernel? I can't see anything conceptually difficult about
a disk driver that just parcels out I/O requests to other drivers (set
via ioctls, perhaps), and while it may not be easy, I don't see any
reason why a machine-specific version of it would be any easier. (I
tried to do something very similar once under pre-Reno 4.3, and never
got it to work reliably.)
Ideally, I'd like to see a disk driver that turns around and reflects
I/O requests back to a user process, which can stripe or interleave or
whatever else it feels like. (Efficient? Who said anything about it
being efficient? I suspect it would be at least usably fast....) I'd
even write it myself, but if the concatenated disk driver is
machine-dependent, there is clearly magic I don't understand going on
somewhere.
der Mouse
mouse@collatz.mcrcim.mcgill.edu