Re: Problems when there are unallocated space in the beginning of an extended partition?

[Andrew Clausen]

On Tue, Jan 11, 2005 at 10:48:55PM +0100, H?kon L?vdal wrote:
> The primary hde2 partition matches a whole number of cylinders
> (like it should). The physical first and the logical partitions
> match a whole number of cylinders minus one head (like they
> should) with the exception of hde5 where the number of sectors,
> 40965750, is wrong because it should have been 63 sectors lower,
> i.e. 40965687.
> 
> hde1:   40965687 + 63 =  40965750 =  2550 * 255 * 63
> hde2:                   200282355 = 12467 * 255 * 63
> hde5:   40965750 +  0 =  40965750 =  2550 * 255 * 63 !!!
> hde6:   40949622 + 63 =  40949685 =  2549 * 255 * 63
> hde7:   56934297 + 63 =  56934360 =  3544 * 255 * 63
> 
> The unallocated space in front of hde5 is 
> 6373 - 2550 + 1 = 3824 cylinders which
> equals 3824 * 255 * 63 = 61432560 sectors.
> 
> That gives an expected relative start sector of
> 61432560 + 63 = 61432623 for hde5, but the table shows that
> it has a RSS of just 61432560, so hde5 starts on head 0
> (instead of 1) on its first cylinder.

Yes, Parted is implemented this way.  Parted's behaviour matches Linux
fdisk's behaviour (I just double-checked), and AFAIK, Windows'
behaviour.

The first logical partition can start on head 0 because the first logical
partition table is stored at the start of the extended partition,
rather than immediately before the first logical partition.

This stuff is rather strange.  For example, if the first logical
partition starts at the start of the extended partition, then it must
start on head 1 (not 0).  If a subsequent logical partition starts at
the start of the extended partition, then it must start on head 2.

Similar issues arise when the extended partition starts on the first
cylinder.

> Notice that the table that defines hde5 lies at the very
> beginning of the extended partition (on the start of cylinder
> 2550) and not at the start of the first cylinder of the space
> that hde5 occupies (i.e. cylinder 6374).
> 
> Could this be the cause?  Do we have an error here?

No error.  Of course, if this confuses Windows in any way, or is not
what Windows would normally do itself, then I'd be happy to change
Parted.

> Then for the tables for hde6 and hde7 they are skewed by 63
> sectors, and the tables are not placed on cylinder boundaries
> (CHS=8924/0/63 and CHS=11473/0/63) and hde6 and hde7 then gets
> an unusual relative start sector value of 1.

This is strange.  Perhaps this is a symptom of the old incorrect CHS
geometry problem?

Cheers,
Andrew