Re: multiple scsi drives..

Dan Merillat (harik@accessorl.net)
Mon, 16 Oct 1995 18:04:04 -0400 (EDT)


On Wed, 11 Oct 1995, Eric Youngdale wrote:

> Date: Wed, 11 Oct 1995 19:48:32 -0400
> From: Eric Youngdale <eric@aib.com>
> To: Jeremiah Jahn <jeremiah@litoria.math.earlham.edu>
> Cc: linux-kernel@vger.rutgers.edu
> Subject: Re: multiple scsi drives..
>
> >when i boot the card bios scans the card going from 6-5-4-3-2-1-0 ie how
> >it should..but when lilo is fireing up the kernel...the fd scsi driver
> >scan te bus in the revers order..ie 0-1-2-3-4-5-6
>
> Who is to say how it "should" be scanned :-). FYI, all of the
> busses
> are scanned in this order.
>
> If it is of any help, you can grab the file scsidev.tar.gz from
> tsx-11 in pub/linux/ALPHA/scsi. This allows you to create a /dev/scsi/
> subdirectory and populate it with special devices whose names are
> independent
> of the other devices that are detected on the bus. The idea is that you
> run this at boot time just after you remount the root filesystem R/W, and
> it will go through and ensure that the special devices for scsi are
> correctly set.
>
> Most people don't bother with this, as they don't change their scsi
> configuration very often.
>
I take it then that there is still no way to grab the driver
by controller:deviceid? I'd rather have scsi/ have devices
of the type scsi/[a-d][1-7][1-40?] then devices that get changed
often. I also like the way extended mapping was proposed
for bsd extended/slices. A nice group of 10/extended type
should take us 15 or so months of os development. Plus
a nice absolute method of mapping drives makes for setting
the system up nicer, independent of what else happens to
be in the machine at the time. Also, has proc/scsi
been working in the newer kernels? I was going to
write a short shell script to link the volume labels
to the absolute devices (but then my system died)
so I havn't had the time (or machine) to run bleeding edge
kernels lately. I also heard rumors of treating a continous
series of blocks as a partition, for instance making
windows swapfile also be the linux swappartition (not file)
because windows has to have its partition as a contigous
file. Better then making a 32 meg dos partition with a
windows/linux swap file on it. (added filesystem overhead)

Oh well, just what I was thinking about the current methods.


Dan Merillat <Dan.Merillat@eola.accessorl.net>
Animaniac! Linux Activist and net.junkie
Also at: <drm89121@pegasus.cc.ucf.edu>