Testing Palimpsest and udisks
marti at juffo.org
Wed Jan 20 12:09:35 PST 2010
Is this the right mailing list for Palimpsest-related email? I
couldn't find any others, sorry if it's not!
I got a new hard drive recently and I've been using it as a chance to
check out the capabilities of Palimpsest/udisks. Overall my experience
has been very positive. This is the first partitioning program that I
actually *like* to use.
Even though I realize that the stack is still in flux, I hope you can
benefit from some feedback at this stage.
1. The slider in the "create partition" view has to go! This is a
1.5TB disk, so the smallest partition I can create on an empty disk is
21 GB! And it takes a lot of precision to get it down to this size. :)
This clearly fails when the user wants to create a small /boot or swap
2. Aligning partitions to cylinder/head size is so 1990. When I create
a GUID partition table, it should be a clear indication that I don't
care about being compatible with historic versions of MS-DOS.
There are good reasons to align to 4kbyte boundaries, as recent
Western Digital hard drives have 4kbyte physical sectors while still
emulating 512-byte sectors for the operating system. Unfortunately
most partitioning tools today start the first partition at the 63rd
sector which would be unaligned. If file system blocks do not align up
with physical sector size, then the disk has to do expensive
This may also apply to SSD disks, where eraseblocks can be a whopping
512 kbytes large. But I don't know what sort of penalty they actually
I hear that recent Windows versions always align the first partition
to 1 MiB (2048 sectors). This might be a convention worth adhering to?
Currently the only way to create properly aligned partitions on Linux
is to use command-line Parted and calculate sector numbers by hand!
3. Partitioning tools like Parted lock the whole disk while they are
running. If I try to delete a partition in Palimpsest, while Parted is
running, it actually deletes the partition on disk physically, but
then fails when telling the kernel to re-read the partition table. The
result is that udisks thinks the partition still exists (delete
failed), but any operations on the partition will fail even after
4. When adding physical volumes to a LVM2 volume group in Palimpsest,
I think it should also list allocated-but-empty partitions. I may want
to partition the whole drive up front and create the volume group
5. When I initialize a PV with "pvcreate /dev/sda9", it successfully
appears in Palimpsest as #orphan_lvm2
However, when I allocate it as a new VG with "vgcreate test
/dev/sda9", the "test" VG does not appear. I guess this is just
More information about the devkit-devel