Re: [PATCH] fs: ocfs2: fix kernel BUG in ocfs2_find_victim_chain

[Joseph Qi <joseph.qi@linux.alibaba.com>]

On 2025/12/1 14:24, Prithvi Tambewagh wrote:
> On Mon, Dec 01, 2025 at 10:51:49AM +0800, Joseph Qi wrote:
>>
>>
>> On 2025/11/30 18:46, Prithvi Tambewagh wrote:
>>> syzbot reported a kernel BUG in ocfs2_find_victim_chain() because the
>>> `cl_next_free_rec` field of the allocation chain list is 0, triggring the
>>> BUG_ON(!cl->cl_next_free_rec) condition and panicking the kernel.
>>>
>>> To fix this, `cl_next_free_rec` is checked inside the caller of
>>> ocfs2_find_victim_chain() i.e. ocfs2_claim_suballoc_bits() and if it is
>>> equal to 0, ocfs2_error() is called, to log the corruption and force the
>>> filesystem into read-only mode, to prevent further damage.
>>>
>>> Reported-by: syzbot+96d38c6e1655c1420a72@syzkaller.appspotmail.com
>>> Closes: https://syzkaller.appspot.com/bug?extid=96d38c6e1655c1420a72
>>> Tested-by: syzbot+96d38c6e1655c1420a72@syzkaller.appspotmail.com
>>> Cc: stable@vger.kernel.org
>>> Signed-off-by: Prithvi Tambewagh <activprithvi@gmail.com>
>>> ---
>>>  fs/ocfs2/suballoc.c | 7 +++++++
>>>  1 file changed, 7 insertions(+)
>>>
>>> diff --git a/fs/ocfs2/suballoc.c b/fs/ocfs2/suballoc.c
>>> index 6ac4dcd54588..84bb2d11c2aa 100644
>>> --- a/fs/ocfs2/suballoc.c
>>> +++ b/fs/ocfs2/suballoc.c
>>> @@ -1993,6 +1993,13 @@ static int ocfs2_claim_suballoc_bits(struct ocfs2_alloc_context *ac,
>>>
>>>      cl = (struct ocfs2_chain_list *) &fe->id2.i_chain;
>>>
>>
>> This blank line can be eliminated.
>>
>>> +    if (le16_to_cpu(cl->cl_next_free_rec) == 0) {
>>
>> Better to add the upper limit check as well. e.g.
>>
>> !le16_to_cpu(cl->cl_next_free_rec) ||
>> le16_to_cpu(cl->cl_next_free_rec) > le16_to_cpu(cl->cl_count)
> 
> Hello Joseph,
> 
> I went through the code in fs/ocfs2/suballoc.c, like this function
> static inline u16 ocfs2_find_smallest_chain(struct ocfs2_chain_list *cl)
> {
>     u16 curr, best;
> 
>     best = curr = 0;
>     while (curr < le16_to_cpu(cl->cl_count)) {
>         if (le32_to_cpu(cl->cl_recs[best].c_total) >
>             le32_to_cpu(cl->cl_recs[curr].c_total))
>             best = curr;
>         curr++;
>     }
>     return best;
> }
> 
> and in function ocfs2_block_group_alloc() these lines
> if (le16_to_cpu(cl->cl_next_free_rec) < le16_to_cpu(cl->cl_count))
>     le16_add_cpu(&cl->cl_next_free_rec, 1);
> 
After this, cl_next_free_rec may equal to cl_count.


> and observed that according to the architecture of ocfs2, the chain list is in the form of 0-indexed array. In that case, the change you suggested for upper limit, could be re-written as
> le16_to_cpu(cl->cl_next_free_rec) >= le16_to_cpu(cl->cl_count)
> 
> since value of cl->cl_next_free_rec greater than or equal to cl->cl_count will indicate that there are no available chains. Can you please review this?
> 
Yes, it's full. But 'cl_next_free_rec == cl_count' is a designed behavior, see mkfs or fsck.

Joseph