Received-SPF: pass (google.com: domain of linux-kernel+bounces-132974-linux.lists.archive=gmail.com@vger.kernel.org designates 147.75.48.161 as permitted sender) client-ip=147.75.48.161;
Message-ID: <721ca866-0c80-496b-a823-ccdbb1f488d4@linux.alibaba.com>
Date: Fri, 5 Apr 2024 20:16:35 +0800
Precedence: bulk
MIME-Version: 1.0
User-Agent: Mozilla Thunderbird
Subject: Re: [PATCH] mm: page_alloc: use the correct THP order for THP PCP
To: Barry Song <21cnbao@gmail.com>
Cc: akpm@linux-foundation.org, mgorman@techsingularity.net, vbabka@suse.cz,
 linux-mm@kvack.org, linux-kernel@vger.kernel.org
References: <a25c9e14cd03907d5978b60546a69e6aa3fc2a7d.1712151833.git.baolin.wang@linux.alibaba.com>
 <CAGsJ_4xvXT6Qy1JgP2khH0DKvTGhjsrexsSj8ddieM_XXn8bng@mail.gmail.com>
From: Baolin Wang <baolin.wang@linux.alibaba.com>
In-Reply-To: <CAGsJ_4xvXT6Qy1JgP2khH0DKvTGhjsrexsSj8ddieM_XXn8bng@mail.gmail.com>
Content-Type: text/plain; charset=UTF-8; format=flowed
Content-Transfer-Encoding: 8bit


On 2024/4/5 12:53, Barry Song wrote:
> On Thu, Apr 4, 2024 at 2:47 AM Baolin Wang
> <baolin.wang@linux.alibaba.com> wrote:
>>
>> Commit 44042b449872 ("mm/page_alloc: allow high-order pages to be stored
>> on the per-cpu lists") extends the PCP allocator to store THP pages, and
>> it determines whether to cache THP pags in PCP by comparing with pageblock_order.
>> But the pageblock_order is not always equal to THP order, it might also
>> be MAX_PAGE_ORDER, which could prevent PCP from caching THP pages.
>>
>> Therefore, using HPAGE_PMD_ORDER instead to determine the need for caching
>> THP for PCP can fix this issue
>>
>> Fixes: 44042b449872 ("mm/page_alloc: allow high-order pages to be stored on the per-cpu lists")
>> Signed-off-by: Baolin Wang <baolin.wang@linux.alibaba.com>
> 
> LGTM,
> Reviewed-by: Barry Song <baohua@kernel.org>
> 
> In the context of using mTHP, perhaps there arises a need for PCP
> allocation for frequently
> requested mTHP orders. These orders typically exceed PAGE_ALLOC_COSTLY_ORDER
> but are smaller than HPAGE_PMD_ORDER.

Yes, I have also considered this, but haven't found some time to do more 
investigation and run some benchmarks.:) May be we can create a new 
thread to talk about this first.