Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.18 as permitted sender) client-ip=23.128.96.18;
Subject: Re: [PATCH] mm/page_isolation: don't putback unisolated page
From:   David Hildenbrand <david@redhat.com>
To:     Miaohe Lin <linmiaohe@huawei.com>, akpm@linux-foundation.org
Cc:     vbabka@suse.cz, iamjoonsoo.kim@lge.com, linux-mm@kvack.org,
        linux-kernel@vger.kernel.org
References: <20210904091839.20270-1-linmiaohe@huawei.com>
 <e060283a-6295-9191-9b92-a3c183e9de02@redhat.com>
 <b533742e-e82a-a721-9a24-4087688dd812@huawei.com>
 <5f7c4ac9-d17a-5af3-0326-1152cca38465@redhat.com>
Organization: Red Hat
Message-ID: <2b103abc-2921-9d8a-7280-1d6b451c34f4@redhat.com>
Date:   Mon, 6 Sep 2021 14:01:42 +0200
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:78.0) Gecko/20100101
 Thunderbird/78.11.0
MIME-Version: 1.0
In-Reply-To: <5f7c4ac9-d17a-5af3-0326-1152cca38465@redhat.com>
Content-Type: text/plain; charset=utf-8; format=flowed
Content-Language: en-US
Content-Transfer-Encoding: 8bit
Precedence: bulk

On 06.09.21 13:50, David Hildenbrand wrote:
> On 06.09.21 13:32, Miaohe Lin wrote:
>> On 2021/9/6 17:40, David Hildenbrand wrote:
>>> On 04.09.21 11:18, Miaohe Lin wrote:
>>>> If __isolate_free_page() failed, due to zone watermark check, the page is
>>>> still on the free list. But this page will be put back to free list again
>>>> via __putback_isolated_page() now. This may trigger page->flags checks in
>>>> __free_one_page() if PageReported is set. Or we will corrupt the free list
>>>> because list_add() will be called for pages already on another list.
>>>>
>>>> Fixes: 3c605096d315 ("mm/page_alloc: restrict max order of merging on isolated pageblock")
>>>> Signed-off-by: Miaohe Lin <linmiaohe@huawei.com>
>>>> ---
>>>>     mm/page_isolation.c | 6 ++----
>>>>     1 file changed, 2 insertions(+), 4 deletions(-)
>>>>
>>>> diff --git a/mm/page_isolation.c b/mm/page_isolation.c
>>>> index 9bb562d5d194..7d70d772525c 100644
>>>> --- a/mm/page_isolation.c
>>>> +++ b/mm/page_isolation.c
>>>> @@ -93,10 +93,8 @@ static void unset_migratetype_isolate(struct page *page, unsigned migratetype)
>>>>                 buddy_pfn = __find_buddy_pfn(pfn, order);
>>>>                 buddy = page + (buddy_pfn - pfn);
>>>>     -            if (!is_migrate_isolate_page(buddy)) {
>>>> -                __isolate_free_page(page, order);
>>>> -                isolated_page = true;
>>>> -            }
>>>> +            if (!is_migrate_isolate_page(buddy))
>>>> +                isolated_page = !!__isolate_free_page(page, order);
>>>>             }
>>>>         }
>>>>    
>>>
>>> Shouldn't we much rather force to ignore watermarks here and make sure __isolate_free_page() never fails?
>>
>> It seems it is not easy to force to ignore watermarks here. And it's not a problem
>> if __isolate_free_page() fails because we can do move_freepages_block() anyway.
>> What do you think? Many thanks.
> 
> I'm wondering if all this complexity in this function is even required. What about something like this: (completely untested)
> 
> diff --git a/mm/page_isolation.c b/mm/page_isolation.c
> index bddf788f45bf..29ff2fcb339c 100644
> --- a/mm/page_isolation.c
> +++ b/mm/page_isolation.c
> @@ -66,12 +66,10 @@ static int set_migratetype_isolate(struct page *page, int migratetype, int isol_
>    
>    static void unset_migratetype_isolate(struct page *page, unsigned migratetype)
>    {
> +       bool buddy_merge_possible = false;
>           struct zone *zone;
>           unsigned long flags, nr_pages;
> -       bool isolated_page = false;
>           unsigned int order;
> -       unsigned long pfn, buddy_pfn;
> -       struct page *buddy;
>    
>           zone = page_zone(page);
>           spin_lock_irqsave(&zone->lock, flags);
> @@ -79,26 +77,15 @@ static void unset_migratetype_isolate(struct page *page, unsigned migratetype)
>                   goto out;
>    
>           /*
> -        * Because freepage with more than pageblock_order on isolated
> -        * pageblock is restricted to merge due to freepage counting problem,
> -        * it is possible that there is free buddy page.
> -        * move_freepages_block() doesn't care of merge so we need other
> -        * approach in order to merge them. Isolation and free will make
> -        * these pages to be merged.
> +        * If our free page spans at least this whole pageblock and could
> +        * eventually get merged into an even bigger page, go via
> +        * __putback_isolated_page(), because move_freepages_block() won't
> +        * trigger merging of free pages.
>            */
>           if (PageBuddy(page)) {
>                   order = buddy_order(page);
> -               if (order >= pageblock_order && order < MAX_ORDER - 1) {
> -                       pfn = page_to_pfn(page);
> -                       buddy_pfn = __find_buddy_pfn(pfn, order);
> -                       buddy = page + (buddy_pfn - pfn);
> -
> -                       if (pfn_valid_within(buddy_pfn) &&
> -                           !is_migrate_isolate_page(buddy)) {
> -                               __isolate_free_page(page, order);
> -                               isolated_page = true;
> -                       }
> -               }
> +               if (order >= pageblock_order && order < MAX_ORDER - 1)
> +                       buddy_merge_possible = true;
>           }
>    
>           /*
> @@ -111,12 +98,12 @@ static void unset_migratetype_isolate(struct page *page, unsigned migratetype)
>            * onlining - just onlined memory won't immediately be considered for
>            * allocation.
>            */
> -       if (!isolated_page) {
> +       if (!buddy_merge_possible) {
>                   nr_pages = move_freepages_block(zone, page, migratetype, NULL);
>                   __mod_zone_freepage_state(zone, nr_pages, migratetype);
>           }
>           set_pageblock_migratetype(page, migratetype);
> -       if (isolated_page)
> +       if (buddy_merge_possible)
>                   __putback_isolated_page(page, order, migratetype);
>           zone->nr_isolate_pageblock--;
>    out:

Okay, I just had another look -- that won't work because as you 
correctly said, it still is on the freelist ...

So your fix is certainly correct :)

-- 
Thanks,

David / dhildenb