severity 380548 normal
thanks
On Sun, Jul 30, 2006 at 10:26:07PM +0200, Steinar H. Gunderson wrote:
> Package: e2fsprogs
> Version: 1.39-1
> Severity: important
>
> I'm trying to resize my ext3 filesystem (64 -> 80GB) online:
>
> pannekake:~# resize2fs /dev/evms/pr0n
> resize2fs 1.39 (29-May-2006)
> Filesystem at /dev/evms/pr0n is mounted on /srv/pr0n.sesse.net; on-line
> resizing required
> Performing an on-line resize of /dev/evms/pr0n to 20971519 (4k) blocks.
> resize2fs: Operation not permitted While trying to add group #512
>
> fsck.ext3 reported no errors afterwards, so I mounted the filesystem and
> again got the same error. I'm attaching an strace of the run in
> question, to possibly aid in the debugging; it looks like some ioctl is
> indeed failing. Also, the kernel reports:
>
> [4950097.452547] EXT3-fs warning (device dm-21): ext3_group_add: No
> reserved GDT blocks, can't resize
Actually, the dumpe2fs output would have been far more useful, but
looking at the kernel implementation, it's returning -EPERM in the
case where the filesystem wasn't created with -O resize_inode feature.
> FWIW; offline resizing on the same volume worked just fine. I'm unsure
> what the real bug here is; is it that resize2fs doesn't detect this
> situation itself, and should give a better warning message? (If so, it's
> probably minor, not important.)
Yes, resize2fs should have detected this situation earlier. It
actually tried to detect it, but the earlier GROUP_EXTEND ioctl()
didn't fail as anticipated.
> Is it that ext2prepare isn't included in e2fsprogs?
> (Wishlist then, I'd guess.)
Already a wishlist bug, and unfortunately it's not going to happen
until I rewrite ext2prepare from scratch. The code is too scary for
me to be willing to accept responsibility for it. :-(
> Or should resize2fs simply be able to add this itself?
Nope, although it is a bug that the kernel is returning such a
non-obvious error message.
> Or is the bug that mke2fs doesn't build the filesystems with
> -O resize_inode by default? (Wishlist again, probably.)
It does build new filesystems with -O resize_inode by default,
starting with e2fsprogs 1.39 --- or rather, starting with e2fsprogs
1.39, the default filesystem features are controlled by a
configuration file, /etc/mke2fs.conf, and the default configuration
file includes resize_inode as one of the default features that should
be enabled.
> Feel free to up- or downgrade the severity as you see fit :-)
- Ted
--
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
