Re: WARNING: filesystem loop5 was created with 512 inodes, the real maximum is 511, mounting anyway

[Randy Dunlap <rdunlap@infradead.org>]

On 11/30/20 11:47 PM, Dmitry Vyukov wrote:
> On Tue, Dec 1, 2020 at 2:03 AM Randy Dunlap <rdunlap@infradead.org> wrote:
>>
>> On 11/30/20 12:43 AM, Dmitry Vyukov wrote:
>>> On Mon, Nov 30, 2020 at 5:29 AM Randy Dunlap <rdunlap@infradead.org> wrote:
>>>>
>>>> On 11/27/20 4:32 AM, syzbot wrote:
>>>>> Hello,
>>>>>
>>>>> syzbot found the following issue on:
>>>>>
>>>>> HEAD commit:    418baf2c Linux 5.10-rc5
>>>>> git tree:       upstream
>>>>> console output: https://syzkaller.appspot.com/x/log.txt?x=171555b9500000
>>>>> kernel config:  https://syzkaller.appspot.com/x/.config?x=b81aff78c272da44
>>>>> dashboard link: https://syzkaller.appspot.com/bug?extid=3fd34060f26e766536ff
>>>>> compiler:       gcc (GCC) 10.1.0-syz 20200507
>>>>>
>>>>> Unfortunately, I don't have any reproducer for this issue yet.
>>>>>
>>>>> IMPORTANT: if you fix the issue, please add the following tag to the commit:
>>>>> Reported-by: syzbot+3fd34060f26e766536ff@syzkaller.appspotmail.com
>>>>>
>>>>> BFS-fs: bfs_fill_super(): loop5 is unclean, continuing
>>>>> BFS-fs: bfs_fill_super(): WARNING: filesystem loop5 was created with 512 inodes, the real maximum is 511, mounting anyway
>>>>> BFS-fs: bfs_fill_super(): Last block not available on loop5: 120
>>>>> BFS-fs: bfs_fill_super(): loop5 is unclean, continuing
>>>>> BFS-fs: bfs_fill_super(): WARNING: filesystem loop5 was created with 512 inodes, the real maximum is 511, mounting anyway
>>>>> BFS-fs: bfs_fill_super(): Last block not available on loop5: 120
>>>>>
>>>>>
>>>>> ---
>>>>> This report is generated by a bot. It may contain errors.
>>>>> See https://goo.gl/tpsmEJ for more information about syzbot.
>>>>> syzbot engineers can be reached at syzkaller@googlegroups.com.
>>>>>
>>>>> syzbot will keep track of this issue. See:
>>>>> https://goo.gl/tpsmEJ#status for how to communicate with syzbot.
>>>>>
>>>>
>>>> Hi,
>>>> Can you provide the BFS image file that is being mounted?
>>>> (./file0 I think.)
>>>>
>>>> --
>>>> ~Randy
>>>
>>>
>>> Hi Randy,
>>>
>>> I see this bug was reported with a reproducer:
>>> https://syzkaller.appspot.com/bug?id=a32ebd5db2f7c957b82cf54b97bdecf367bf0421
>>> I assume it's a dup of this one.
>>
>> Sure, looks the same.
>>
>>> If you need the image itself, you can dump it to a file in the C
>>> reproducer inside of syz_mount_image before mount call.
>>
>> Yes, got that.
>>
>> What outcome or result are you looking for here?
>> Or what do you see as the problem?
> 
> Hi Randy,
> 
> "WARNING:" in kernel output is supposed to mean a kernel source bug.
> Presence of that kernel bug is what syzbot has reported.
> 
> Note: the bug may be a misuse of the "WARNING:" for invalid user
> inputs in output as well :)


[adding Al Viro]

Hi Dmitry,

I expect that the "WARNING:" message is being interpreted incorrectly here,
but that's a minor issue IMO.

	if (info->si_lasti == BFS_MAX_LASTI)
		printf("WARNING: filesystem %s was created with 512 inodes, the real maximum is 511, mounting anyway\n", s->s_id);


If you/we look at fs/bfs/bfs.h, it says:

/* In theory BFS supports up to 512 inodes, numbered from 2 (for /) up to 513 inclusive.
   In actual fact, attempting to create the 512th inode (i.e. inode No. 513 or file No. 511)
   will fail with ENOSPC in bfs_add_entry(): the root directory cannot contain so many entries, counting '..'.
   So, mkfs.bfs(8) should really limit its -N option to 511 and not 512. For now, we just print a warning
   if a filesystem is mounted with such "impossible to fill up" number of inodes */

so one question is why does syzkaller try to do this at all?
Why not set number-of-inodes to 511 instead of 512 in the BFS image file?


However, in testing this, I see that the BFS image is not mounted
on /dev/loop# at all.

'mount' says:

# mount -t bfs -o loop bfsfilesyz000.img  /mnt/stand
mount: /mnt/stand: mount(2) system call failed: Not a directory.

(but it is a directory)

and I have tracked that down to fs/namespace.c::graft_tree()
returning -ENOTDIR, but I don't know why that is happening.


Al, can you provide any insights on this?

thanks.
-- 
~Randy