On Tuesday, 25 Jul 2006 21:35:05 +0200, Frans Pop wrote: > As can be seen in the bug history, partman needs to create the new > smaller/larger partition with the same start sector (not start cylinder) > as the old partition. I have investigated further what happens in partman. The script that is called when resizing an NTFS partition is: /lib/partman/active_partition/70resize/do_option Before this script is executed, a partition dump in /var/log/partman shows: Partitions: # id length type fs path name (0,0,0) (0,0,62) -1 0-32255 32256 primary label /dev/scsi/host0/bus0/target0/lun0/part-1 (0,1,0) (0,32,31) -1 32256-1048575 1016320 unusable free /dev/scsi/host0/bus0/target0/lun0/part-1 (0,32,32) (2549,228,46) 1 1048576-20973617151 20972568576 primary ntfs /dev/scsi/host0/bus0/target0/lun0/part1 (2549,228,47) (2549,254,62) -1 20973617152-20974463999 846848 unusable free /dev/scsi/host0/bus0/target0/lun0/part-1 The NTFS partition and the surrounding free space were created by the Windows Vista installer. After that is an existing Linux partition. The script performs the following basic steps: - starting size of the partition is 2048-40964095 (1048576-20973617151) - ntfsresize -f -i /dev/scsi/host0/bus0/target0/lun0/part1 it greps the result for the smallest possible size for the partition - I selected 15G as the new size - ntfsresize -f --size 15000000000 /dev/scsi/host0/bus0/target0/lun0/part1 this ntfs resize operation succeeds - script calls parted_server with: open_dialog VIRTUAL_RESIZE_PARTITION 1048576-20973617151 15000000000 - in parted_server virtual_resize_partition calls resize_partition with new start and end sectors; I added a debug statement to show which start and end sectors it is being called with From this call, /var/log/partman contains: parted_server: Resizing partition with id 1048576-20973617151 parted_server: partition_with_id(1048576-20973617151) parted_server: New size: 15000000000 parted_server: resize_partition(openfs=false) parted_server: resize_partition start 2048; end 29298922 parted_server: try to check the file system for errors parted_server: successfully checked This clearly shows that resize_partition is called with correct parameters. - resize_partition calls libparted: ped_disk_set_partition_geom(disk, part, constraint, start, end) - this resize in itself is performed correctly; the next partition dump in /var/log/partman shows: Partitions: # id length type fs path name (0,0,0) (0,0,62) -1 0-32255 32256 primary label /dev/scsi/host0/bus0/target0/lun0/part-1 (0,1,0) (1823,254,62) 1 32256-15002910719 15002878464 primary ntfs /dev/scsi/host0/bus0/target0/lun0/part1 (1824,0,0) (2549,254,62) -1 15002910720-20974463999 5971553280 unusable free /dev/scsi/host0/bus0/target0/lun0/part-1 This shows that ped_disk_set_partition_geom does _not_ create the partition at the requested sectors. fdisk with the 'u' option shows the partition as: Device Boot Start End Blocks Id System /dev/sda1 * 63 29302559 14651248+ 7 HPFS/NTFS The original starting sector 2048 has been changed to 63. Instead of the requested end sector 29298922, libparted sets it at 29302559. - next, the original script does another ntfsresize to maximize on the new partition: ntfsresize -f /dev/scsi/host0/bus0/target0/lun0/part1 - this fails because the starting sector is now incorrect Conclusion: either resize_partition calls the wrong libparted function, or there is a bug in parted's ped_disk_set_partition_geom function.
Attachment:
pgpFDWzMX6Y_T.pgp
Description: PGP signature