Re: [PATCH 1/1] ext4: fix crash on BUG_ON in ext4_alloc_group_tables

[Eric Sandeen]

On 9/26/24 3:28 AM, Baokun Li wrote:
> On 2024/9/25 23:57, Jan Kara wrote:
>> On Wed 25-09-24 16:33:24, Alexander Mikhalitsyn wrote:
>>> [   33.882936] EXT4-fs (dm-5): mounted filesystem 8aaf41b2-6ac0-4fa8-b92b-77d10e1d16ca r/w with ordered data mode. Quota mode: none.
>>> [   33.888365] EXT4-fs (dm-5): resizing filesystem from 7168 to 786432 blocks
>>> [   33.888740] ------------[ cut here ]------------
>>> [   33.888742] kernel BUG at fs/ext4/resize.c:324!
>> Ah, I was staring at this for a while before I understood what's going on
>> (it would be great to explain this in the changelog BTW).  As far as I
>> understand commit 665d3e0af4d3 ("ext4: reduce unnecessary memory allocation
>> in alloc_flex_gd()") can actually make flex_gd->resize_bg larger than
>> flexbg_size (for example when ogroup = flexbg_size, ngroup = 2*flexbg_size
>> - 1) which then confuses things. I think that was not really intended and
>> instead of fixing up ext4_alloc_group_tables() we should really change
>> the logic in alloc_flex_gd() to make sure flex_gd->resize_bg never exceeds
>> flexbg size. Baokun?
>>
>>                                 Honza
> 
> Hi Honza,
> 
> Your analysis is absolutely correct. It's a bug!
> Thank you for locating this issue！
> An extra 1 should not be added when calculating resize_bg in alloc_flex_gd().
> 
> 
> Hi Aleksandr,
> 
> Could you help test if the following changes work?
> 

I just got an internal bug report for this as well, and I can also confirm that
the patch fixes my testcase, thanks! Feel free to add:

Tested-by: Eric Sandeen <sandeen@xxxxxxxxxx>

I had been trying to debug this and it felt like an off by one but I didn't get
to a solution in time. ;) 

Can you explain what the 2 cases under

/* Avoid allocating large 'groups' array if not needed */

are doing? I *think* the first 'if' is trying not to over-allocate for the last
batch of block groups that get added during a resize. What is the "else if" case
doing?

Thanks,
-Eric

> Thanks,
> Baokun
> 
> ---
> 
> diff --git a/fs/ext4/resize.c b/fs/ext4/resize.c
> index e04eb08b9060..1f01a7632149 100644
> --- a/fs/ext4/resize.c
> +++ b/fs/ext4/resize.c
> @@ -253,10 +253,12 @@ static struct ext4_new_flex_group_data *alloc_flex_gd(unsigned int flexbg_size,
>         /* Avoid allocating large 'groups' array if not needed */
>         last_group = o_group | (flex_gd->resize_bg - 1);
>         if (n_group <= last_group)
> -               flex_gd->resize_bg = 1 << fls(n_group - o_group + 1);
> +               flex_gd->resize_bg = 1 << fls(n_group - o_group);
>         else if (n_group - last_group < flex_gd->resize_bg)
> -               flex_gd->resize_bg = 1 << max(fls(last_group - o_group + 1),
> +               flex_gd->resize_bg = 1 << max(fls(last_group - o_group),
>                                               fls(n_group - last_group));
> 
>         flex_gd->groups = kmalloc_array(flex_gd->resize_bg,
>                                         sizeof(struct ext4_new_group_data),
> 
>