From Bugzilla Helper:
User-Agent: Mozilla/5.0 (X11; U; Linux i686; en-US; rv:1.7.5)
Description of problem:
I used resize2fs to make one of my filesystems smaller and then when I
later ran a filesystem check it reported some errors
Version-Release number of selected component (if applicable):
Steps to Reproduce: (see attachment for scripted example)
1. mke2fs -j /dev/foo
2. e2fsck -f /dev/foo
3. resize2fs /dev/foo <some smaller size>
4. e2fsck -f /dev/foo
Pass 5: Checking group summary information
Block bitmap differences: +3 -259 -(531--1298) +8195 -8451 +24579
-24835 +40963 -41219 +57347 -57603 +73731 -73987
Expected Results: No errors reported by e2fsck following resize
I will attach a small test script and results log for a reproducible
test case using a loopback based filesystem.
I did some test with various mke2fs options and it seems that the
errors disappear when I disable the "resize_inode" feature. It seems
that this inode (for the ext2online resize) gets mishandled by the
legacy resize2fs code. Unfortunately - the mke2fs shipped with FC3
e2fsprogs has the resize_inode enabled by default.
Created attachment 107412 [details]
Test script which demonstrates error using loopback filesystem
Created attachment 107413 [details]
Output from running test-resize script
This bug may be related to bug #144771, which I hit. This bug
(#140762) has very similar symptoms, but I get different errors from
e2fsck. I get "Group X's inode table at Y conflicts with some other
fs block." as opposed to "Block bitmap differences".
I think that you are right and these bugs are related. From the look
of the release notes of e2fsprogs-1.36-rc1, it seems that it will fix
this problem. I have not got around to trying it myself yet.
I just hit this in FC3 with e2fsprogs-1.36-1.FC3.1 and lvm2-2.00.25-1.01.
-> goal was to resize /usr (make it larger)
-> assign a couple of GB to that parition using lvresize
(as it's an LVM partition)
-> telinit 1, unmount, resize2fs, all is well...
-> reboot - things work ok
-> the reboot after that gives filesystem errors and I get dropped to the repair
shell; errors were "Group X's inode table at Y conflicts with some other
fs block." and some complaint about the resize inode which I didn't write
But there were not too many errors and repair seems to have fixed things. Phew.
Tried the "test script which demonstrates error" but that one passes A-OK.
Fixed in the latest FC3-update packages of e2fsprogs: e2fsprogs-1.38-0.FC3.1