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

[Joseph Qi]

On 2025/10/30 11:30, Deepanshu Kartikey wrote:
> Hi Joseph, Heming, and Dmitry,
> 
> Investigation complete. Here are the full test results:
> 
> **Test Setup:**
> - Kernel: Custom build with cl_bpc validation patch
> - Filesystems: Created with mkfs.ocfs2 -b [512|1024|2048|4096] -M
> local --fs-features=metaecc
> 
> **Results:**
> 
> 512-byte blocks (4KB cluster):
>   inode=20: cl_bpc=8, expected=8 (cluster_bits=12, block_bits=9)  : PASS
>   inode=23: cl_bpc=1, expected=8 (cluster_bits=12, block_bits=9) :FAIL

I've also seen this. From debugfs.ocfs2, inode 23 is global_bitmap:

[root at localhost ~]# debugfs.ocfs2 /dev/vdc
debugfs.ocfs2 1.8.7
debugfs: ls //
        18              16   1    2  .
        18              16   2    2  ..
        19              24   10   1  bad_blocks
        20              32   18   1  global_inode_alloc
        21              20   8    1  slot_map
        22              24   9    1  heartbeat
        23              28   13   1  global_bitmap
        24              28   15   2  orphan_dir:0000
        25              32   17   1  extent_alloc:0000
        26              28   16   1  inode_alloc:0000
        27              24   12   1  journal:0000
        28              28   16   1  local_alloc:0000
        29              148  17   1  truncate_log:0000

And this can be verified in mkfs:

if (rec->cluster_bitmap) {
	...
	di->id2.i_chain.cl_bpc = 1;
	...
}

So this check conflicts with global bitmap.

Joseph

>   → Validation detects corruption
> 
> 1024-byte blocks (4KB cluster):
>   inode=12: cl_bpc=4, expected=4 (cluster_bits=12, block_bits=10) :PASS
>   inode=15: cl_bpc=1, expected=4 (cluster_bits=12, block_bits=10) :FAIL
>   → Validation detects corruption
> 
> 2048-byte blocks (4KB cluster):
>   inode=10: cl_bpc=2, expected=2 (cluster_bits=12, block_bits=11) : PASS
>   inode=13: cl_bpc=1, expected=2 (cluster_bits=12, block_bits=11): FAIL
>   → Validation detects corruption
> 
> 4096-byte blocks (4KB cluster):
>   inode=8:  cl_bpc=1, expected=1 (cluster_bits=12, block_bits=12) : PASS
>   inode=11: cl_bpc=1, expected=1 (cluster_bits=12, block_bits=12) :PASS
>   → All inodes correct, mount succeeds
> 
> **Analysis:**
> 
> 1. The validation calculates expected cl_bpc as:
>    expected_bpc = 1 << (cluster_size_bits - block_size_bits)
> 
>    This matches the mkfs.ocfs2 source code you referenced:
>    di->id2.i_chain.cl_bpc = s->cluster_size / s->blocksize
> 
> 2. For non-4K block sizes, some allocator inodes (20, 12, 10, 8) have
>    correct cl_bpc values, while others (23, 15, 13) consistently have
>    cl_bpc=1.
> 
> 3. For 4K blocks, all allocator inodes have the correct value (1).
> 
> The improved error message now includes:
>   "Inode %llu has corrupted cl_bpc: ondisk=%u expected=%u
>    (cluster_size_bits=%u block_size_bits=%u)"
> 
> Please advise on next steps.
> 
> Thanks,
> Deepanshu