Parted 1.6.11 croaks when 'sparse_super' is set?

[Jonathan D Reed]

(N.B. While I saw the effect of this bug reported in other messages, I
didn't see one that actually tracked it down to sparse_super -
apologies if I missed one.)

Parted appears to not be able to deal with ext2 filesystems that have
sparse_super set.  I first noticed this when running parted (and
qtparted) from the SystemRescueCD (http://sysresccd.org).  I
downloaded 1.6.11 and built it to confirm that the bug was still
present in the latest version.  Here's an example of what happens:

0) Make sure partition isn't mounted:

infinite-loop.mit.edu# mount | grep -i hda7
infinite-loop.mit.edu# grep -i hda7 /proc/mounts 
infinite-loop.mit.edu#

1) infinite-loop.mit.edu# /sbin/mke2fs /dev/hda7

[completes correctly, for the sake of brevity, I'll just list the
superblock backup]

Superblock backups stored on blocks: 
        32768, 98304, 163840, 229376, 294912, 819200, 884736, 1605632, 2654208

2) Just for fun, let's make sure it's ok:

infinite-loop.mit.edu# /sbin/e2fsck /dev/hda7
e2fsck 1.32 (09-Nov-2002)
/dev/hda7: clean, 11/1596224 files, 50103/3190902 blocks

3) It is, now we'll run parted check on it:

infinite-loop.mit.edu# parted /dev/hda check 7
No Implementation: This ext2 filesystem has a rather strange layout!  Parted 
can't resize
this (yet).

4) Now we'll recreate the same filesystem, this time with no special
   options (suitable for mounting on pre-Linux 2.2 systems)

infinite-loop.mit.edu# /sbin/mke2fs -O none /dev/hda7

[success, this time creating many more superblock backups]

5) Again, let's check it:

infinite-loop.mit.edu# /sbin/e2fsck /dev/hda7
e2fsck 1.32 (09-Nov-2002)
/dev/hda7: clean, 11/1596224 files, 50279/3190902 blocks

6) It's good.  Run parted on it:

infinite-loop.mit.edu# parted /dev/hda check 7
Information: The ext2 filesystem passed a basic check.  For a more
comprehensive check, use the e2fsck program.

7) OK, now it works.  Just to confirm it's sparse_super that's
   breaking things, we'll do it one more time.  mke2fs by default will
   enable "filetype" and "sparse_super" if no options are specified.
   So we'll enable filetype, and disable sparse_super.

infinite-loop.mit.edu# /sbin/mke2fs /dev/hda7 -O filetype,^sparse_super 
infinite-loop.mit.edu# /sbin/e2fsck /dev/hda7
e2fsck 1.32 (09-Nov-2002)
/dev/hda7: clean, 11/1596224 files, 50279/3190902 blocks
infinite-loop.mit.edu# parted /dev/hda check 7
Information: The ext2 filesystem passed a basic check.  For a more 
comprehensive check, use
the e2fsck program.


So, for some reason, parted is croaking when it encounters filesystems
with sparse superblocks.  This is unfortunate, as it prevents users
from resizing partitions that are large enough to merit
sparse_superblocks.

I was able to reproduce this behavior on 3 machines, one running
Gentoo with a 2.4.26 kernel, one running RedHat 9 with a 2.4.20 kernel
(2.4.20-30.9), and one running RedHat Enterprise Linux 3.0 (Update 2)
with a 2.4.21 kernel (2.4.21-15.EL).

Oh, and just for kicks, my partition table:

Disk geometry for /dev/hda: 0.000-38172.750 megabytes
Disk label type: msdos
Minor    Start       End     Type      Filesystem  Flags
1          0.031     47.065  primary   ext3        
2         47.065  10048.469  primary   fat32       boot, lba
3      10048.469  10566.188  primary   linux-swap  
4      10566.189  38170.063  extended              
5      10566.220  25572.216  logical   ext3        
6      25572.248  25705.568  logical   ext2        
7      25705.600  38170.063  logical   ext2        

But all 3 machines had different partition tables.

-Jon

--------
Jonathan Reed
address@hidden