Re: when is label info commited to disk?

Marty Leisner (
Sun, 13 Jul 1997 11:08:50 PDT

In message <>, you write:
>Marty Leisner:
>: Are there instructions on how to create an extended linux partition?
>An extended Linux partition differs in no way from an extended DOS
>partition, except that the IDs are 85 and 05 (hex), respectively.
>What is the use?
>DOS FDISK will crash or worse when it reads a chain of extended
>partitions passing the 1024 cylinder boundary. Since DOS does not
>recognize 85, this enables you to make DOS happy and still have a
>series of extended partitions past the 1024 cylinder mark.
>How to create?
>Any sufficiently recent *fdisk will do, I think.
>Try cfdisk, sfdisk and fdisk.
cfdisk painlessly lets you create a primary or extended partitions in
one partition.

But I think all the fdisk programs will fail when we start creating
multiple extended partitions on one disk (is there a limit?)

If we have an extended dos and extended linux partition (assuming we
already created it), how do we tell our *fdisk program which slice to
add the parititon to?

>: If I make hda2/hda3 extended linux partitions, the kernel counting
>: mechanism will screw up /dev/hda4
>Screw up in what way? What is your complaint?

If /dev/hda5 sits in /dev/hda4 now, and then I add extended partitions
to /dev/hda3, what used to be /dev/hda5 may be /dev/hda7.

I think the labeling needs to be rethought, since its too easy to
wreak havoc on a working system. The BSD file system also has
the same problem (does the newest 2.1. incarnations work with freebsd?)
I think the following would work much better:
/dev/hd*1 extended partitions would be in /dev/hd*10-19
/dev/hd*2 extended partitions would be in /dev/hd*20-29
/dev/hd*3 extended partitions would be in /dev/hd*30-39
/dev/hd*4 extended partitions would be in /dev/hd*40-49

What I'm proposing is a strategy to make disk slices independent of other
slices on the same disk...

>Do you know what you are doing when you create two chains of
>extended partitions? Sometimes that is the only solution,
>but usually it is just a very bad idea.
Actually, if you have a large disk, and create 2 extended partitions chains
(lets say, each holding several gig) I don't see it as a bad idea.