Comment # 2 on bug 987451 from Stanislav Brabec

Is this a regression? Did the same setup work differently in older versions?


I think that that this is a bit strange, but correct behavior.

You are using and old DOS partition table. It heavily depends on heads,
cylinders and sectors per cylinder.

When sfdisk finds an empty image of your size, it selects following geometry:

Geometry: 49 heads, 30 sectors/track, 60801 cylinders

This means 1470 sectors per one cylinder and 664471 cylinders.

This geometry makes possible to use only 976772370 of 976773168 sectors you
defined. The last 798 sectors are not usable as they don't form whole cylinder.

The smallest size of a partition is 1 cylinder, i. e. 1470 sectors.

So if I compute correctly, you need at least 2268 unused sectors to create a
partition.

The only strange thing I see:

fdisk -l -u=cylinders sda
Disk sda: 465.8 GiB, 500107862016 bytes, 976773168 sectors
Geometry: 49 heads, 30 sectors/track, 60801 cylinders
                                      ^^^^^
Units: cylinders of 1470 * 512 = 752640 bytes
Sector size (logical/physical): 512 bytes / 512 bytes
I/O size (minimum/optimal): 512 bytes / 512 bytes
Disklabel type: dos
Disk identifier: 0x55eb8806

Device Boot Start    End Cylinders   Size Id Type
sda1            1 664471    664471 465.8G 83 Linux