Subject: Re: Strange disklabel/fidsk info?
To: Matthew Mondor <mmondor@linuxguru.net>
From: Richard Rauch <rauch@eecs.ukans.edu>
List: port-i386
Date: 12/23/2000 21:06:51
> > I was there, and 10 years sounds about right. I can't say that the name
> > rings a bell.
>
> I was 1:167/104.2 at the time if I remember well (oldest fidonet node in
> Quebec at the time, run by Robert Williamson) we wrote a suite of fidonet
> mailers on AmigaOS (roof, porticus, gazebo), I also often used the nick
> Lord British as I was a U5 fan (-:
Robert Williamson...yes, I seem to remember that name. Lord Brittish
tolls a somewhat more distant bell, but...could be. (^&
I never used roof. I did use Michael Richardson's Welmat (later Russell
McOrmond's pet, I guess), and a variety of mail readers. (I ultimately
wrote my own mail reader, but never got around to distributing it. It
wasn't polished or graphical, but it suited my needs and tastes.)
[...]
> > important. I gave a 4GB (actually 8,000,000 sector) partition over to
> > GNU/LINUX, and the rest to NetBSD. Of course, such a round number didn't
> > come out evenly to a cylinder boundary. Hopefully, that's not going to
> > cause me any grief...
>
> I used NetBSD and GNU/Linux on two different partitions (I had /boot,
> swap, /, then NetBSD one which had it's own swap, / and /usr into that
> same partition using disklabels)
>
> I know that it never matters either the BIOS sees the partitions and disk
> size properly for linux using LILO, however I needed the bios to know the
> starting point of the NetBSD one as I used one of LILO's other= option,
> and that gets executed before any kernel is loaded, requireing the BIOS..
I had to worry about that kind of thing when dual-booting MS-WINDOWS and
NetBSD, since (I think) the boot-selector that I used at the time
basically picked one partition and marked it active. The boot-selector
that NetBSD can install presumably is a bit smarter than that. (^& (The
old boot-selector was from the Partition Magic tool for MS-WINDOWS.)
LINUX's need for discrete BIOS-level partitions to support its seperate
filesystems has caught me offguard. Maybe I should put in a second disk
for LINUX and let it chop up the drive however it likes. (^&
> I only used linux and NetBSD's fdisk however, and they reported the same
> cylender boundaries for all partitions, except one. Filesystems never
> corrupt eachother but I never could set the disklabel in a way that my
> ext2 partition could be mounted, even dough the cylender boundaries were
> (according to fdisk) properly setup, perhaps because of the fact that
> it was in an extended partition.. the other non-extended partitions (including
> my FAT32 one could be mounted properly and reflected the same boundaries on
> all OSs)
>
> > Aside from some griping from the LINUX installation about one of my
> > partitions going past the ``end'' of the disk, all is fairly well. I may
> > have to re-do the LINUX install (and even repartition the LINUX
> > segment)---I didn't know that LINUX used a BIOS-level partition for
> > swap-space.
>
> I don't know if I understand this well, the only time I had "past end" errors
The BIOS notion of my disk size is flatly wrong. I'm not sure how it got
the wrong numbers, but the numbers _do_ sound like power-of-two related
limits, and it _is_ a fake geometry. I don't know what the fake geometry
limit is, but the fake geometry being used by BIOS is (according to
NetBSD's fdisk): 1024 cylinders, 255 heads, and 63 sectors.
(The real BIOS, when asked about the disk drive, reports more credibly,
39703 cyls, 16 heads, 63 sectors. And, it is on this that my original
NetBSD disklabel was constructed, by inspection.)
That's more or less the hardware/BIOS situation.
The old NetBSD partition started at an offset of 63 (leaving one track
reserved for BIOS), and had a size filling the rest of the 20GB, more or
less.
To repartition, I started a LINUX native partition (type 131) at offset
63, with 8,000,000 sectors. I then restarted the NetBSD partition at
8,000,063, and reduced its size by 8,000,000 sectors from what it had
previously enjoyed.
I installed NetBSD on the larger partition.
Then I installed LINUX (Debian, specifically) on the 4GB partition. The
Debian installer scanned the drive and saw that I had some ``other''
partition (it doesn't recognize NetBSD, it seems?) that extended beyond
the end of the drive.
I.e., it seems that the Debian installer was only getting the 7-or-8 GB
drive size, and didn't believe that I could have a partition running from
the end of 4GB up to ~20GB.
Does that clarify?
> > (Horrors, I just had a thought. People were talking about ~30GB disks
> > having a jumper setting to mark them as smaller disks, for the sake of
[...]
> I really do not know, I used maximum 15GB ones maximum only
(^&
When I bought this system, it would by default have had a 30GB drive. I
couldn't imagine needing that much for a personal system (I still can't),
so I chopped something like $80 or $100 off the price by reducing the
drive to a 20GB drive. (At the time, 20GB drives looked like they were
about $100 apiece; I haven't followed the prices since.)
Later, I heard tales about large drives with jumpers on them to make them
pretend to be smaller drives, but I thought that at 20GB, I was safely
under the limit at which drive manufacturers were currently doing that.
If I do have such a drive, and it requires a jumper to be set for full
size, perhaps the jumper has worked loose.
Well, if it really has less than 8GB, then I can find out fairly easily
what kind of trouble this lets me in for. I just need to generate a lot
of random data and see if it (a) gets lost (b) causes errors (c) trashes
the LINUX or NetBSD partitions (due to wraparound). (^&
(Well, random data probably isn't the best choice. Structured data, such
as massive software installation and .tar.gz files probably makes more
sense...(^&)
Then I can pick up the pieces, if need be...
"I probably don't know what I'm talking about." --rauch@eecs.ukans.edu