Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.33 as permitted sender) client-ip=23.128.96.33;
Message-ID: <b6f0abbc-91e3-6561-a1ba-09048a999161@linux.dev>
Date:   Tue, 26 Sep 2023 15:57:10 +0800
MIME-Version: 1.0
Subject: Re: [PATCH v3 1/2] mm: pass page count and reserved to
 __init_single_page
Content-Language: en-US
To:     David Hildenbrand <david@redhat.com>, akpm@linux-foundation.org,
        rppt@kernel.org
Cc:     mike.kravetz@oracle.com, muchun.song@linux.dev,
        willy@infradead.org, linux-mm@kvack.org,
        linux-kernel@vger.kernel.org
References: <20230926023341.991124-1-yajun.deng@linux.dev>
 <20230926023341.991124-2-yajun.deng@linux.dev>
 <9777812e-c8c8-2105-cd2c-443438786172@redhat.com>
From:   Yajun Deng <yajun.deng@linux.dev>
In-Reply-To: <9777812e-c8c8-2105-cd2c-443438786172@redhat.com>
Content-Type: text/plain; charset=UTF-8; format=flowed
Content-Transfer-Encoding: 8bit
Precedence: bulk


On 2023/9/26 15:44, David Hildenbrand wrote:
> On 26.09.23 04:33, Yajun Deng wrote:
>> When we init a single page, we need to mark this page reserved if it
>> does. 
>
> I failed to parse the last part of this sentence.
>
>> And some pages may not need to set page count, such as compound
>> pages.
>
> Usually, the refcount of all tail pages *must* be zero. Otherwise, 
> get_page_unless_zero() would work on tail pages.
>
> Can you elaborate why it should be okay here?
>
>
It means the following code in memmap_init_compound().

-               __init_zone_device_page(page, pfn, zone_idx, nid, pgmap);
+               __init_zone_device_page(page, pfn, zone_idx, nid, pgmap, 0);
                 prep_compound_tail(head, pfn - head_pfn);
                 set_page_count(page, 0);

As we can see, it will reset the page count by 'set_page_count(page, 0)'.

Therefore, we don't need to set page count in __init_zone_device_page(). 
I wasn't clear enough in the commit.

Maybe we can remove the 'set_page_count(page, 0)',  But I didn't do 
that, just to be safe.