f2fs image may be corrupted after below testcase:
- mkfs.f2fs -O extra_attr,compression -f /dev/vdb
- mount /dev/vdb /mnt/f2fs
- touch /mnt/f2fs/file
- f2fs_io setflags compression /mnt/f2fs/file
- dd if=/dev/zero of=/mnt/f2fs/file bs=4k count=4
- f2fs_io release_cblocks /mnt/f2fs/file
- truncate -s 8192 /mnt/f2fs/file
- umount /mnt/f2fs
- fsck.f2fs /dev/vdb
[ASSERT] (fsck_chk_inode_blk:1256)  --> ino: 0x5 has i_blocks: 0x00000002, but has 0x3 blocks
[FSCK] valid_block_count matching with CP             [Fail] [0x4, 0x5]
[FSCK] other corrupted bugs                           [Fail]
The reason is: partial truncation assume compressed inode has reserved
blocks, after partial truncation, valid block count may change w/o
.i_blocks and .total_valid_block_count update, result in corruption.
This patch only allow cluster size aligned truncation on released
compress inode for fixing.
Fixes: c61404153eb6 ("f2fs: introduce FI_COMPRESS_RELEASED instead of using IMMUTABLE bit")
Signed-off-by: Chao Yu <chao@kernel.org>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
                                  ATTR_GID | ATTR_TIMES_SET))))
                return -EPERM;
 
-       if ((attr->ia_valid & ATTR_SIZE) &&
-               !f2fs_is_compress_backend_ready(inode))
-               return -EOPNOTSUPP;
+       if ((attr->ia_valid & ATTR_SIZE)) {
+               if (!f2fs_is_compress_backend_ready(inode))
+                       return -EOPNOTSUPP;
+               if (is_inode_flag_set(inode, FI_COMPRESS_RELEASED) &&
+                       !IS_ALIGNED(attr->ia_size,
+                       F2FS_BLK_TO_BYTES(F2FS_I(inode)->i_cluster_size)))
+                       return -EINVAL;
+       }
 
        err = setattr_prepare(idmap, dentry, attr);
        if (err)