[lvc-project] [PATCH 2/2] ocfs2: validate cl_bpc in allocator inodes to prevent divide-by-zero

[Dmitry Antipov]

On 10/29/25 6:02 PM, Deepanshu Kartikey wrote:

> I'd like to investigate this and send a fix. Let me know what additional
> information you have, or I can create test filesystems with different
> block sizes and debug it myself.

I would suggest to start from the following:

--<cut from here>--

#!/bin/bash

BS=16M
COUNT=1024

# Blocksize 512
rm -f img-512.ocfs2
dd if=/dev/zero of=img-512.ocfs2 bs=$BS count=$COUNT
sudo losetup /dev/loop0 img-512.ocfs2
sudo mkfs.ocfs2 -b 512 -M local --fs-features=metaecc /dev/loop0
sudo losetup -d /dev/loop0
sync

# Blocksize 1K
rm -f img-1k.ocfs2
dd if=/dev/zero of=img-1k.ocfs2 bs=$BS count=$COUNT
sudo losetup /dev/loop0 img-1k.ocfs2
sudo mkfs.ocfs2 -b 1024 -M local --fs-features=metaecc /dev/loop0
sudo losetup -d /dev/loop0
sync

# Blocksize 2K
rm -f img-2k.ocfs2
dd if=/dev/zero of=img-2k.ocfs2 bs=$BS count=$COUNT
sudo losetup /dev/loop0 img-2k.ocfs2
sudo mkfs.ocfs2 -b 2048 -M local --fs-features=metaecc /dev/loop0
sudo losetup -d /dev/loop0
sync

# Blocksize 4K
rm -f img-4k.ocfs2
dd if=/dev/zero of=img-4k.ocfs2 bs=$BS count=$COUNT
sudo losetup /dev/loop0 img-4k.ocfs2
sudo mkfs.ocfs2 -b 4096 -M local --fs-features=metaecc /dev/loop0
sudo losetup -d /dev/loop0
sync

--<to here>--

and then check the things trying to mount an appropriate drive (most likely /dev/sdb) from:

qemu-system-x86_64 ... -kernel [kernel] -drive file=[rootfs] -drive file=img-512.ocfs2 ...
qemu-system-x86_64 ... -kernel [kernel] -drive file=[rootfs] -drive file=img-1k.ocfs2 ...
qemu-system-x86_64 ... -kernel [kernel] -drive file=[rootfs] -drive file=img-2k.ocfs2 ...
qemu-system-x86_64 ... -kernel [kernel] -drive file=[rootfs] -drive file=img-4k.ocfs2 ...

Dmitry