Subject: Re: [suse-axp] Fdisk behavior is wrong in 6.3 ???
From: William H. Magill (magill@isc.upenn.edu)
Date: Wed Jun 14 2000 - 12:25:40 PDT
> From: "Mike O'Rourke" <mjoop@dominicans.it>
> References: <200006141131.e5EBVUB23575@falstaff.isc-net.upenn.edu>
> Date: Wed, 14 Jun 2000 16:45:41 +0200
>
> If you use the "-u" option with fdisk (fdisk -u /dev/sda), fdisk will show
> sectors rather than cylinders. Since this is how OSF1/Tru64 makes the
> disklabels, at least you will be on the same level. I had Tru64 4.0f
> installed on a 1000A and had to delete all partitions from the existing
> disklabel for SuSE to use the disk (it would not use the Tru64 disklabel at
> all, even though all of the partitions were the same.) Once I deleted the
> partitions and re-made the label, everything was happy; but I had to use the
> "-u" oprion to do this.
>
The partitions all delete with no problem and I can create new ones with no
trouble using either sectors or cylinders.
- the problem is that the numbers I give to fdisk are NOT the
numbers which fdisk uses (or at least plays back to me) when it
creates the partitions (using either sectors or cylinders).
And it changes these values arbitrarily with no warning, let
alone error, messages of any kind.
> If you are doing fdisk in conjunction with a new installation, I recommend
> making the disklabel and then before returning to YaST, do a mkswap
> /dev/sda? -- that way you will not have to restart YaST to mount the swap --
> it recognises the swap partition if you format it first.
>
Basically, that is what everything describes -- you flip out of YaST to do
the partition via fdisk, then reboot from the CD and "skip partitioning."
According to the SuSE 6.3 docs, YaST cannot deal with BSD partitions needed
by the SRM console. That part of the process is pretty clean, straight
forward and wll documented.
What does not perform as documented is fdisk.
> If the XP1000 is not MILO capable (since you are using BSD labels, I assume
> you are booting from the SRM console), and you follow SuSE's instructions
> and make one big partiton and a swap, the machine will not boot. There is
> some magic size limit for the partition that has the boot files on it. I
> tried a 9gb disk (on my 1000A), and had to make a small /boot partition for
> the kernel and boot files for aboot.
[I'm using SRM and booting from the 6.3 CD on both boxes. ]
This is what I was trying to do with fdisk -- make a boot partion that was
only about 10 or 20 meg. And then a couple of others for /, swap, /usr, etc.
That's the pattern described in Compaq's 2.1 "linux installation and
configuration guide..."
> A 4gb disk works fine. This is an undocumented "gotcha".
>
This doesn't surprise me -- from what I've been reading recently on the
various Alpha lists, Linux is anything but 64bit clean even deep down in
the bowels of the kernel let alone in some of the upper layers of drivers.
----- Original Message -----
From: "William H. Magill" <magill@isc.upenn.edu>
To: <suse-axp@suse.com>
Sent: Wednesday, June 14, 2000 1:31 PM
Subject: [suse-axp] Fdisk behavior is wrong in 6.3 ???
> I'm new to Suse and Linux, but old with Ultrix/OSF1/DU/Tru64.
>
> I'm trying to setup a Linux disk that can be mounted by Tru64 and
following
> the instructions in rev 2.1 of Compaq's installatin guide.
>
> I've been playing with both and XP1000 and a 433A.
>
> The XP1000 has an RZ2DD-KS [9 gig 10K RPM]
> The 433A RZ1CB-BA [4 gig 7200 RPM]
>
> I have no trouble booting from the 6.3 CD and getting to the point where
> fdisk is running, and I go to BSD labels...
>
> I try to create the new partitions trying to assing partion A
> cylinders 1 and 2.
>
> This fails.... or rather fdisk simply goes off and randomly assigns
> some number cylinders.
>
> If I look at the partition display after assigning cylinders 1 and 2,
> I get some exoticly high number. (not an obviously rational multiple of
> anything I have guessed.)
>
> If I try to assign A to 1 +5000K, it gives me yet a different number.
>
> So, in short, I see no correlation between the numbers entered in fdisk,
> and what is actually being generated by fdisk.
>
> Yes, both disk have pre-existing Tru64 labels via disklabel.
>
> So, what part of fdisk is broken --
> the calculation of the size of cylinders?
> the display of the label/partition information that it plans to
> create?
>
> I find it difficult to believe that such an important and old program can
> be so wrong (and we are talking -- not even close here gang), but have no
> other choice.
>
> --
--
www.tru64unix.compaq.com
www.tru64.org
comp.unix.tru64
T.T.F.N.
William H. Magill Senior Systems Administrator
Information Services and Computing (ISC) University of Pennsylvania
Internet: magill@isc.upenn.edu magill@acm.org
http://www.isc-net.upenn.edu/~magill/
--
To unsubscribe, e-mail: suse-axp-unsubscribe@suse.com
For additional commands, e-mail: suse-axp-help@suse.com
This archive was generated by hypermail 2a22 : Sat Jul 01 2000 - 05:31:32 PDT