On Mon, Mar 01, 2021 at 07:14:00PM +0300, Благодаренко Артём wrote:

Here is how I reproduced it originally (without the patch)

# truncate -s 512MB /tmp/lustre-ost
# losetup -b 4096 /dev/loop0 /tmp/lustre-ost
# mkfs.ext4 /dev/loop0

Thanks for pointing out that this can be tested using losetup.

With your patch e2image works fine.

[root@CO82 e2fsprogs-kernel]# git rev-parse HEAD
67f2b54667e65cf5a478fcea8b85722be9ee6e8d
[root@CO82 e2fsprogs-kernel]# misc/e2image /dev/loop0 /tmp/ost-image
e2image 1.46.1 (9-Feb-2021)

So commit 67f2b54667e6 is 1.46.2, so I'm not sure that you recompiled
e2image.  More importantly, e2image in upstream doesn't use Direct
I/O.  When I tried using debugfs -D, it looks like Direct I/O on an
Advanced Format HDD isn't working correctly:

# debugfs -D /dev/loop0
debugfs 1.46.2 (28-Feb-2021)
debugfs: Bad magic number in super-block while trying to open /dev/loop0
/dev/loop0 contains a ext4 file system
        created on Mon Mar  1 12:31:43 2021

This isn't a regression since it was broken in upstream before, but it
would be good to get this fixed before e2fsprogs 1.46.3.

Cheers,

						- Ted