Partition renumbering under 2.4.0

Partition renumbering under 2.4.0

[David Santinoli]

Hi,
 I've noticed that some logical partitions get different numbering under 2.2.16
and 2.4.0. Here's my /dev/hdb layout:

  hdb1: fat32
  hdb2: Solaris partition (contains 4 Solaris slices)
  hdb3: ext2
  hdb4: extended partition (contains 1 ext2 logical partition)

and here's how it gets detected by the kernels:

2.2.16:
  hdb: hdb1 hdb2 <solaris: [s0] hdb5 [s1] hdb6 [s2] hdb7 [s7] hdb8 > hdb3 hdb4 <
 hdb9 >

2.4.0:
 hdb: hdb1 hdb2 hdb3 hdb4 < hdb5 >
 hdb2: <solaris: [s0] hdb6 [s1] hdb7 [s2] hdb8 [s7] hdb9 >

Note that the ext2 logical partition is called "hdb9" by 2.2.16 and "hdb5" by
2.4.0.
This makes it difficult to manage multi-boot systems with 2.2.x and 2.4.x
kernels, as it requires updating fstab between boots. Switching to other
identification strategies such as ext2 labels - as discussed in other threads -
could be a workaround, as far as I know.

Cheers,
 David
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
Please read the FAQ at http://www.tux.org/lkml/

Re: Partition renumbering under 2.4.0

[James Bottomley]

> Note that the ext2 logical partition is called "hdb9" by 2.2.16 and
> "hdb5" by 2.4.0. This makes it difficult to manage multi-boot systems
> with 2.2.x and 2.4.x kernels, as it requires updating fstab between
> boots. Switching to other identification strategies such as ext2
> labels - as discussed in other threads - could be a workaround, as far
> as I know.

Your problem is caused by the identification loop in fs/partitions/msdos.c 
having been split into two parts in 2.4.  One part identifies msdos and 
extended partitions in the first pass and then goes on to identify other 
disklabel partitions in the second.  You can restore consistency either by 
combining the two passes (as was the case in 2.2) or by moving the extended 
msdos partition identification into the second pass.

However, the underlying problem remains the same as it was when I first wrote 
the solaris code:  You only have 11 minors for all of the remaining extended 
partitions, that's why the solaris code does the slice renaming.  Under 2.4, 
if you use devfs, the solaris (and other) slice recognition code could be 
enhanced to give the correct names to all the slices.  This would turn out to 
be something like /dev/ide/hdb2s7 (or something even worse---I'm afraid I only 
really know the naming scheme for SCSI devices) but at least you can find the 
exact slice you're looking for in an easy and intuitive way.

So, would you prefer the quick fix, or the more durable solution (which would 
require you to change your fstab)?

James


-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
Please read the FAQ at http://www.tux.org/lkml/

Re: Partition renumbering under 2.4.0

[David Santinoli]

On Wed, Jan 17, 2001 at 10:56:14AM -0500, James Bottomley wrote:
> Under 2.4, if you use devfs, the solaris (and other) slice recognition code
> could be enhanced to give the correct names to all the slices.  This would
> turn out to be something like /dev/ide/hdb2s7 (or something even worse---I'm
> afraid I only really know the naming scheme for SCSI devices) but at least you
> can find the exact slice you're looking for in an easy and intuitive way.
> 
> So, would you prefer the quick fix, or the more durable solution (which would 
> require you to change your fstab)?

Personally I'd be happy with the quick hack, but the slice-enhanced naming
scheme possible with devfs looks like the way to go.

Besides, I think that documenting this issue in the "Changes" file would help
somehow.

Thanks,
 David
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
Please read the FAQ at http://www.tux.org/lkml/