Re: [PATCH] mm: fix unevictable page reclaim when calling madvise_pageout

[zhong jiang <zhongjiang@huawei.com>]

On 2019/10/31 3:33, Johannes Weiner wrote:
> On Wed, Oct 30, 2019 at 06:45:33PM +0100, Michal Hocko wrote:
>> On Wed 30-10-19 09:52:39, Minchan Kim wrote:
>>> On Tue, Oct 29, 2019 at 06:45:12PM +0800, zhong jiang wrote:
>>>> On 2019/10/29 17:40, Michal Hocko wrote:
>>>>> On Tue 29-10-19 17:30:57, zhong jiang wrote:
>>>>>> On 2019/10/29 16:11, Michal Hocko wrote:
>>>>>>> [Cc Minchan]
>>>>> [...]
>>>>>>> Removing a long existing BUG_ON begs for a much better explanation.
>>>>>>> shrink_page_list is not a trivial piece of code but I _suspect_ that
>>>>>>> removing it should be ok for mapped pages at least (try_to_unmap) but I
>>>>>>> am not so sure how unmapped unevictable pages are handled from top of my
>>>>>>> head.
>>>>>> As to the unmapped unevictable pages.  shrink_page_list has taken that into account.
>>>>>>
>>>>>> shinkr_page_list
>>>>>>      page_evictable     --> will filter the unevictable pages to putback its lru.
>>>>> Ohh, it is right there at the top. Missed it. The check has been added
>>>>> by Nick along with the BUG_ON. So it is sounds more like a "this
>>>>> shouldn't happen" bugon. I wouldn't mind to remove it with that
>>>>> justification.
>>>> As you has said,   Minchan fix the same kind of bug by checking PageUnevictable (I did not notice before)
>>>> Wait for Minchan to see whether  he has better reason. thanks,
>>> madvise_pageout could work with a shared page and one of the vmas among processes
>>> could do mlock so it could pass Unevictable LRU pages into shrink_page_list.
>>> It's pointless to try reclaim unevictable pages from the beginning so I want to fix
>>> madvise_pageout via introducing only_evictable flag into the API so that
>>> madvise_pageout uses it as "true".
>>>
>>> If we want to remove the PageUnevictable VM_BUG_ON_PAGE in shrink_page_list,
>>> I want to see more strong reason why it happens and why caller couldn't
>>> filter them out from the beginning.
>> Why is this preferable over removing the VM_BUG_ON condition? In other
>> words why should we keep PageUnevictable check there?
> The mlock LRU shuffling is a bit tricky and can race with page reclaim
> or others isolating the page from the LRU list. If another isolator
> wins, it has to move the page during putback on behalf of mlock.
>
> See the implementation and comments in __pagevec_lru_add_fn().
I see that comments in __pagevec_lru_add_fn. I have some confusion.
It will result in evictable page strand in an unevictable lru  without PageMlocked due to disorder
 
If I understand it correctly.    vmscan can see !page_evictable().   It should be PageMLocked is set
in evictable list.  Is there any race window ?

Thanks,
zhong jiang
> That's why page reclaim can see !page_evictable(), but it must not see
> pages that have the PageUnevictable lru bit already set. Because that
> would mean the isolation/putback machinery messed up somewhere and the
> page LRU state is corrupt.
>
> As that machinery is non-trivial, it's useful to have that sanity
> check in page reclaim.
>
> .
>