Re: [PATCH] btrfs: Fix NULL pointer access in btrfs_check_leaked_roots()

[Qu Wenruo <quwenruo.btrfs@gmx.com>]

在 2025/10/15 20:32, Dewei Meng 写道:
> 
> 在 2025/10/15 16:24, Qu Wenruo 写道:
>>
>>
>> 在 2025/10/15 17:54, Dewei Meng 写道:
>>> If fs_info->super_copy or fs_info->super_for_commit is NULL in
>>> btrfs_get_tree_subvol(),
>>
>> Please reorganize this sentence. It would be way more easier to read 
>> by just saying something like "If memory allocation failed for 
>> fs_info->super_copy or fs_info->super_for_commit in 
>> btrfs_get_tree_subvol()".
> I agree, I will fix these words to make them easier to read.
>>
>>> the btrfs_check_leaked_roots() will get the
>>> btrfs_root list entry using the fs_info->allocated_roots->next
>>> which is NULL.
>>>
>>> syzkaller reported the following information:
>>>    ------------[ cut here ]------------
>>>    BUG: unable to handle page fault for address: fffffffffffffbb0
>>>    #PF: supervisor read access in kernel mode
>>>    #PF: error_code(0x0000) - not-present page
>>>    PGD 64c9067 P4D 64c9067 PUD 64cb067 PMD 0
>>>    Oops: Oops: 0000 [#1] SMP KASAN PTI
>>>    CPU: 0 UID: 0 PID: 1402 Comm: syz.1.35 Not tainted 6.15.8 #4 
>>> PREEMPT(lazy)
>>>    Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), (...)
>>>    RIP: 0010:arch_atomic_read arch/x86/include/asm/atomic.h:23 [inline]
>>>    RIP: 0010:raw_atomic_read include/linux/atomic/atomic-arch- 
>>> fallback.h:457 [inline]
>>>    RIP: 0010:atomic_read include/linux/atomic/atomic- 
>>> instrumented.h:33 [inline]
>>>    RIP: 0010:refcount_read include/linux/refcount.h:170 [inline]
>>>    RIP: 0010:btrfs_check_leaked_roots+0x18f/0x2c0 fs/btrfs/disk- 
>>> io.c:1230
>>>    [...]
>>>    Call Trace:
>>>     <TASK>
>>>     btrfs_free_fs_info+0x310/0x410 fs/btrfs/disk-io.c:1280
>>>     btrfs_get_tree_subvol+0x592/0x6b0 fs/btrfs/super.c:2029
>>>     btrfs_get_tree+0x63/0x80 fs/btrfs/super.c:2097
>>>     vfs_get_tree+0x98/0x320 fs/super.c:1759
>>>     do_new_mount+0x357/0x660 fs/namespace.c:3899
>>>     path_mount+0x716/0x19c0 fs/namespace.c:4226
>>>     do_mount fs/namespace.c:4239 [inline]
>>>     __do_sys_mount fs/namespace.c:4450 [inline]
>>>     __se_sys_mount fs/namespace.c:4427 [inline]
>>>     __x64_sys_mount+0x28c/0x310 fs/namespace.c:4427
>>>     do_syscall_x64 arch/x86/entry/syscall_64.c:63 [inline]
>>>     do_syscall_64+0x92/0x180 arch/x86/entry/syscall_64.c:94
>>>     entry_SYSCALL_64_after_hwframe+0x76/0x7e
>>>    RIP: 0033:0x7f032eaffa8d
>>>    [...]
>>>
>>> This should check if the fs_info->allocated_roots->next is NULL before
>>> accessing it.
>>>
>>> Fixes: 3bb17a25bcb0 ("btrfs: add get_tree callback for new mount API")
>>> Signed-off-by: Dewei Meng <mengdewei@cqsoftware.com.cn>
>>> ---
>>>   fs/btrfs/disk-io.c | 3 +++
>>>   1 file changed, 3 insertions(+)
>>>
>>> diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c
>>> index 0aa7e5d1b05f..76db7f98187a 100644
>>> --- a/fs/btrfs/disk-io.c
>>> +++ b/fs/btrfs/disk-io.c
>>> @@ -1213,6 +1213,9 @@ void btrfs_check_leaked_roots(const struct 
>>> btrfs_fs_info *fs_info)
>>>   #ifdef CONFIG_BTRFS_DEBUG
>>>       struct btrfs_root *root;
>>>   +    if (!fs_info->allocated_roots.next)
>>> +        return;
>>> +
>>
>> The check looks too adhoc to me.
>>
>> It would be much easier to just call kvfree() in the error handling of 
>> super_copy/super_for_commit allocation, we do not and should not call 
>> btrfs_free_fs_info() before calling btrfs_init_fs_info().
> 
> It is a good solution to fix this bug, or can we put the 
> 'btrfs_init_fs_info(fs_info)' before super_copy/super_for_commit 
> allocation?

That also sounds fine to me.

> 
> Thanks,
> 
> Dewei Meng
> 
>>
>> Thanks,
>> Qu
>>>       while (!list_empty(&fs_info->allocated_roots)) {
>>>           char buf[BTRFS_ROOT_NAME_BUF_LEN];
>>
>>
>>