lkml.org 
[lkml]   [1996]   [Jul]   [13]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
From
SubjectRe: SCSI device numbering
Date
Quoth Kai Henningsen:
> 1. Find a way to give names to every partition, either in the partition
> table (this is already implemented on the Mac, for example), or in the
> partition itself (FAT and HPFS already have this).
>
> 2. Find a way (for example, /proc/partition) to get a list of all
> currently-available partitions.
>
> 3. Create a new syntax for mount (and /etc/fstab, and fsck, and ...) so
> you can mount partitions by name. Something like "mount part:usr /usr"
> might de an idea.

Okay, that works fairly well when we mount /usr and friends. But, even if
we don't need a complete kernel-level partitioning system, there's a case
where some kernel-level action could be quite desirable that I haven't seen
(m?)any of the plans presented here even start to address, and that is
booting off named (or uniquely-identified) disks/partitions (perhaps even
other devices -- who wants to guess which drive has the (insert favorite
distribution here) CD-ROM in it? Hell, who wants to guess what kind of
drive it is?)

That's not something you can have LILO do very easily, because it'd have to
second-guess the kernel's naming scheme of devices (and, in the
named-partition case, would have to have lots of knowledge about filesystems).
Which means it'd have to know a lot about various storage device adaptors,
and we shouldn't even think about LILO being that smart.

I propose that there be an alternate naming system for boot devices (this
could be parsed on the kernel side rather than the loader side), consisting
of either a partition name, or a unique identifier (controller, bus, target,
LUN, partition; unused fields ignored for the sake of IDE and friends) of a
partition. So

xroot=ide,1,0

would boot off of the first partition on whatever the master on the
secondary IDE channel is (if it's an ATAPI CD-ROM, tape drive, whatever...
we can grok its fullness).

xroot=tmc8xx,0,4,1,2

would boot off the second partition of the second drive inside that nasty Sun
dual-drive shoebox with an Emulex ESDI-SCSI bridge inside you picked up for
US$10 at surplus through the TMC-850 you got for $2.

xroot="root"

would boot off a partition on some device by the name of "root" which we'll
find when we start thinking about mounting the root fs and all the
controllers are up.

Come to think of it, anyone see any problems with doing this in the initrd?
(I'm not terribly familiar with initrd and its capabilities, so anyone who
_does_ know is encouraged to help me out =:^) )

> MfG Kai

-jhp

--
Jon Pickard * 149 Olive #45 * Paso Robles CA 93446 * +1 805 2399518 * 6372F5B9
*My e-mail address is (C) 1996 and is free for non-commercial use. It is not*
* to be used for solicitation or other commercial purposes of any kind. *
--
KNOW YOUR SYSADMIN: TECHNICAL MANIAC: Writes scripts that SEEM to be
monitoring the system, but are actually encrypting large lists of passwords.
Uses nearby nodes as beta test sites for worms. -Stephan Zielinski


\
 
 \ /
  Last update: 2005-03-22 13:38    [W:0.050 / U:2.492 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site