Bug 78401

Confirming - I also see this behavior.

$ gfloppy
<format a floppy disk ext2>
$ mount /mnt/floppy
$ mkdir /mnt/floppy/foo
$ touch /mnt/floppy/foo/bar
$ umount /mnt/floppy
$ mount /mnt/floppy
$ cd /mnt/floppy/foo
$ ls -laF
total 0

Ironically - if you format the floppy as FAT - no such problem occurs.  "bar"
appears without a hitch.  Do this by repeating the steps and selecting "FAT"
instead of ext2.

Here's another weirdism - need to figure out where to file this bug. 
Apparently, once you've done all the above, the floppy refuses to take an ext2
file system (or at least appears to):

$ umount /mnt/floppy
$ mke2fs /dev/fd0
mke2fs 1.27 (8-Mar-2002)
Filesystem label=
OS type: Linux
Block size=1024 (log=0)
Fragment size=1024 (log=0)
184 inodes, 1440 blocks
72 blocks (5.00%) reserved for the super user
First data block=1
1 block group
8192 blocks per group, 8192 fragments per group
184 inodes per group

Writing inode tables: done                            
Writing superblocks and filesystem accounting information: done

This filesystem will be automatically checked every 33 mounts or
180 days, whichever comes first.  Use tune2fs -c or -i to override.
$ mount /mnt/floppy/
$ mount | grep floppy
/dev/fd0 on /mnt/floppy type vfat (rw,nosuid,nodev,user=myohe)

Further analysis:

$ umount /mnt/floppy
# mount -t ext2 /dev/fd0 /mnt/floppy
$ mount | grep floppy
/dev/fd0 on /mnt/floppy type ext2 (rw)
$ cd /mnt/floppy
$ ls -laF
total 17
drwxr-xr-x    3 myohe    myohe        1024 Nov 22 10:52 ./
drwxr-xr-x    8 root     root         4096 Nov 12 12:42 ../
drwx------    2 root     root        12288 Nov 22 10:52 lost+found/
$ md bar
$ cd bar
$ touch foo
$ ls -laF
total 2
drwxr-xr-x    2 myohe    myohe        1024 Nov 22 10:56 ./
drwxr-xr-x    4 myohe    myohe        1024 Nov 22 10:56 ../
-rw-r--r--    1 myohe    myohe           0 Nov 22 10:56 foo
$ cd /mnt
# umount /mnt/floppy
$ mount /mnt/floppy
$ mount | grep floppy
/dev/fd0 on /mnt/floppy type vfat (rw,nosuid,nodev,user=myohe)
$ ls -laF /mnt/floppy
total 11
drwxr-xr-x    2 myohe    myohe        7168 Dec 31  1969 ./
drwxr-xr-x    8 root     root         4096 Nov 12 12:42 ../
$ touch /mnt/floppy/f
$ ls -laF /mnt/floppy/f
total 11
drwxr-xr-x    2 myohe    myohe        7168 Nov 22 10:59 ./
drwxr-xr-x    8 root     root         4096 Nov 12 12:42 ../
-rwxr-xr-x    1 myohe    myohe           0 Nov 22 10:59 f*
$ cd /mnt
$ umount /mnt/floppy
$ mount /mnt/floppy
$ ls -laF floppy
total 11
drwxr-xr-x    2 myohe    myohe        7168 Dec 31  1969 ./
drwxr-xr-x    8 root     root         4096 Nov 12 12:42 ../
-rwxr-xr-x    1 myohe    myohe           0 Nov 22 10:59 f*
$ umount /mnt/floppy
# mount -t ext2 /dev/fd0 /mnt/floppy
$ mount | grep floppy
/dev/fd0 on /mnt/floppy type ext2 (rw)
$ ls -laF /mnt/floppy/
total 18
drwxr-xr-x    4 myohe    myohe        1024 Nov 22 10:56 ./
drwxr-xr-x    8 root     root         4096 Nov 12 12:42 ../
drwxr-xr-x    2 myohe    myohe        1024 Nov 22 10:56 bar/
drwx------    2 root     root        12288 Nov 22 10:52 lost+found/

Sounds like a SERIOUS bug...

Also discovered that the problem does not arrise when using mke2fs to format 
the disk. Suggests that the issue is with gfloppy.

What could gfloppy do that could possibly corrupt the disk so badly?  I would
have to say that the severity should by high since loss of data can occur.

Severity changed to high due to possibility od loosing data.

Ah, looks like a known problem when formatting disks of all sorts.

Basically, the problem is that different filesystems put their critical metadata
in different places.  For ext2, the superblock starts at offset 1024 into the
filesystem.  For msdos (and hence vfat), it starts at offset 512.

So if you have a disk formatted as vfat, and then reformat it as ext2, the old
vfat superblock may still be there just before you get to the ext2 one.  mke2fs
has got this right since version 1.24, though, and will clear the remnants of
the old dos filesystem in those versions (which explains why mke2fs doesn't show
the problem, only gfloppy.)

Reassigning to gfloppy.

gfloppy just calls mke2fs directly.  Am I missing an option to it?  8.0 shipped
e2fsprogs 1.27, so it should be recent enough, right?

Please verify this with a newer version of Red Hat Enterprise Linux or
Fedora Core and reopen it against the new version if it still occurs.

Closing as "not a bug" for now.