Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::3:7 as permitted sender) client-ip=2620:137:e000::3:7;
Message-ID: <798ddb57-ba09-e337-01b3-c80711f1e277@linux.dev>
Date:   Mon, 25 Sep 2023 11:23:03 +0800
MIME-Version: 1.0
Subject: Re: [PATCH 1/4] mm: pass set_count and set_reserved to
 __init_single_page
To:     Mike Rapoport <rppt@kernel.org>,
        David Hildenbrand <david@redhat.com>
Cc:     Matthew Wilcox <willy@infradead.org>, akpm@linux-foundation.org,
        mike.kravetz@oracle.com, muchun.song@linux.dev, glider@google.com,
        elver@google.com, dvyukov@google.com, osalvador@suse.de,
        linux-mm@kvack.org, linux-kernel@vger.kernel.org,
        kasan-dev@googlegroups.com
References: <20230922070923.355656-1-yajun.deng@linux.dev>
 <20230922070923.355656-2-yajun.deng@linux.dev>
 <ZQ1Gg533lODfqvWd@casper.infradead.org>
 <2ed9a6c5-bd36-9b9b-7022-34e7ae894f3a@redhat.com>
 <20230922080831.GH3303@kernel.org>
Content-Language: en-US
From:   Yajun Deng <yajun.deng@linux.dev>
In-Reply-To: <20230922080831.GH3303@kernel.org>
Content-Type: text/plain; charset=UTF-8; format=flowed
Content-Transfer-Encoding: 8bit
Precedence: bulk


On 2023/9/22 16:08, Mike Rapoport wrote:
> On Fri, Sep 22, 2023 at 09:48:59AM +0200, David Hildenbrand wrote:
>> On 22.09.23 09:47, Matthew Wilcox wrote:
>>> On Fri, Sep 22, 2023 at 03:09:20PM +0800, Yajun Deng wrote:
>>>> -		__init_single_page(page, pfn, zone, nid);
>>>> +		__init_single_page(page, pfn, zone, nid, true, false);
>>> So Linus has just had a big rant about not doing bool flags to
>>> functions.  And in particular _multiple_ bool flags to functions.
>>>
>>> ie this should be:
>>>
>>> #define INIT_PAGE_COUNT		(1 << 0)
>>> #define INIT_PAGE_RESERVED	(1 << 1)
>>>
>>> 		__init_single_page(page, pfn, zone, nid, INIT_PAGE_COUNT);
>>>
>>> or something similar.
>>>
>>> I have no judgement on the merits of this patch so far.  Do you have
>>> performance numbers for each of these patches?  Some of them seem quite
>>> unlikely to actually help, at least on a machine which is constrained
>>> by cacheline fetches.
>> The last patch contains
>>
>> before:
>> node 0 deferred pages initialised in 78ms
>>
>> after:
>> node 0 deferred pages initialised in 72ms
>>
>> Not earth-shattering :D Maybe with much bigger machines relevant?
> Patch 3 contains
>
> The following data was tested on an x86 machine with 190GB of RAM.
>
> before:
> free_low_memory_core_early()    342ms
>
> after:
> free_low_memory_core_early()    286ms
>
> Which is more impressive, but still I'm not convinced that it's worth the
> added complexity and potential subtle bugs.
>
I will send v2.  It will be simpler and safer.
>> -- 
>> Cheers,
>>
>> David / dhildenb
>>