(ext3 + (2.4.18 -> 2.6.11)) = WTF?

[Michael ODonnell]

I'm trying to bring a 2.4.18 based Debian system up to
kernel 2.6.11 and so far it's been painless except for one
problem that's got me stumped.  I've got several disks
in the system (hda, hde and hdh) that are each occupied
by single ext3 filesystems, with hda1 being the root.
The 2.4.18 kernel boots from hda1 and handles all devices
and partitions just fine, mounting the ext3 filesystems
on hde1 and hdh2 without incident.

The 2.6.11 kernel also runs fine, booting from that same
ext3 partition and it is also clearly able to access the
raw data on all the other disks in question; I can fsck
hde1/hdh1 and manipulate them with fdisk, etc, etc.

The thing that's got me irked is that I can't mount
either of those non-root ext3 filesystems; the mount
command instead just complains about how either the device
appears to already be mounted or else the mount point is
busy, neither of which is true.  If I create a fresh mount
point under /tmp I still cannot mount either of those ext3
filesystems there (same complaint) while at the same time I
can readily mount any other filesystem on the mount points
in question, so they're apparently not really busy.

I clearly have all the necessary ext3 support in the 2.6.11
kernel because (A) I'm running off an ext3 root and (B)
I can create and mount new ext3 filesystems under that
kernel at will.  Note that I can create ext3 filesystem
images under that 2.6.11 kernel that are readable using
either kernel; same with the 2.4.18 kernel though, of
course, it's the same apps in both cases - only the kernel
is different.

So, it's just those two particular ext3 filesystems that
are unmountable under 2.6.11 but perfectly usable with the
2.4.18 kernel - does this sound at all familiar to anybody?

I've downloaded the sources to the filesystem manipulation
apps ("mount" and friends) and am about to embark on what
I assume will be a tedious bit of analysis that I would
love not to have to do...

Perplexed,
  --M