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: [RFC PATCH] fs/exec: Add the support for ELF program's NUMA
 replication
To:     Huang Shijie <shijie@os.amperecomputing.com>
Cc:     viro@zeniv.linux.org.uk, akpm@linux-foundation.org,
        jlayton@kernel.org, bfields@fieldses.org,
        torvalds@linux-foundation.org, linux-fsdevel@vger.kernel.org,
        linux-kernel@vger.kernel.org, linux-mm@kvack.org,
        song.bao.hua@hisilicon.com, patches@amperecomputing.com,
        zwang@amperecomputing.com
References: <20210906161613.4249-1-shijie@os.amperecomputing.com>
 <2cb841ca-2a04-f088-cee2-6c020ecc9508@redhat.com> <YTnX7IyC420MNBLq@hsj>
From:   David Hildenbrand <david@redhat.com>
Organization: Red Hat
Message-ID: <1c115101-a549-0e88-7bbb-1b0a19621504@redhat.com>
Date:   Thu, 9 Sep 2021 10:58:49 +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: <YTnX7IyC420MNBLq@hsj>
Content-Type: text/plain; charset=utf-8; format=flowed
Content-Language: en-US
Content-Transfer-Encoding: 7bit
Precedence: bulk

On 09.09.21 11:46, Huang Shijie wrote:
> On Mon, Sep 06, 2021 at 11:35:01AM +0200, David Hildenbrand wrote:
>> On 06.09.21 18:16, Huang Shijie wrote:
>>> This patch adds AT_NUMA_REPLICATION for execveat().
>>>
>>> If this flag is set, the kernel will trigger COW(copy on write)
>>> on the mmapped ELF binary. So the program will have a copied-page
>>> on its NUMA node, even if the original page in page cache is
>>> on other NUMA nodes.
>>
>> Am I missing something important or is this just absolutely not what we
>> want?
> 
> Please see the thread:
> https://marc.info/?l=linux-kernel&m=163070220429222&w=2
> 
> Linus did not think it is a good choice to implement the "per-numa node page cache"

That doesn't make this approach any better.

I don't think we want this in the kernel. If user space wants to waste 
memory, it can happily mmap() however it wants. The advisory is to not 
do it.

-- 
Thanks,

David / dhildenb