Re: [PATCH v5 1/2] btrfs: always uses root memcgroup for filemap_add_folio()

[Qu Wenruo <wqu@suse.com>]

在 2024/7/19 16:46, Qu Wenruo 写道:
> 
> 
> 在 2024/7/19 16:38, Michal Hocko 写道:
>> On Fri 19-07-24 14:19:05, Qu Wenruo wrote:
>> [...]
>>> @@ -2981,8 +2982,17 @@ static int attach_eb_folio_to_filemap(struct 
>>> extent_buffer *eb, int i,
>>>       ASSERT(eb->folios[i]);
>>>
>>>   retry:
>>> +    /*
>>> +     * Btree inode is a btrfs internal inode, and not exposed to any
>>> +     * user.
>>> +     * Furthermore we do not want any cgroup limits on this inode.
>>> +     * So we always use root_mem_cgroup as our active memcg when 
>>> attaching
>>> +     * the folios.
>>> +     */
>>> +    old_memcg = set_active_memcg(root_mem_cgroup);
>>>       ret = filemap_add_folio(mapping, eb->folios[i], index + i,
>>>                   GFP_NOFS | __GFP_NOFAIL);
>>> +    set_active_memcg(old_memcg);
>>
>> I do not think this will compile with CONFIG_MEMCG=n
> 
> My bad.
> 
> I'll fix it from the btrfs part so that @root_mem_cgroup would be
> changed to NULL for the CONFIG_MEMCG=n case.
> 
> Meanwhile, can we just define root_mem_cgroup as NULL globally?

Of course, I mean for the CONFIG_MEMCG=n case.

Thanks,
Qu

> That looks more reasonable to me, and if that's fine I can send out an
> extra patch just for that.
> 
> Thanks,
> Qu
> 
>>
>>>       if (!ret)
>>>           goto finish;
>>>
>>> -- 
>>> 2.45.2
>>