[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Date: Tue, 2 Feb 2021 14:56:49 -0800
From: Dave Hansen <dave.hansen@...el.com>
To: Yang Shi <shy828301@...il.com>, Oscar Salvador <osalvador@...e.de>
Cc: Dave Hansen <dave.hansen@...ux.intel.com>,
Linux Kernel Mailing List <linux-kernel@...r.kernel.org>,
Linux MM <linux-mm@...ck.org>,
Yang Shi <yang.shi@...ux.alibaba.com>,
David Rientjes <rientjes@...gle.com>,
Huang Ying <ying.huang@...el.com>,
Dan Williams <dan.j.williams@...el.com>
Subject: Re: [RFC][PATCH 08/13] mm/migrate: demote pages during reclaim
On 2/2/21 2:45 PM, Yang Shi wrote:
>> Should we keep it simple for now and only try to demote those pages that are
>> free of cpusets and memory policies?
>> Actually, demoting those pages to a CPU or a NUMA node that does not fall into
>> their set, would violate those constraints right?
> Yes, this has been discussed since the very beginning. There is not an
> easy way to figure out the memory placement policy (cpuset and
> mempolicy) from "page". I think this also prevents "demote those pages
> that are free of cpusets and memory policies".
>
> The conclusion was the violation should be fine for now. And the
> demotion feature is opt'ed in by a new node reclaim mode.
This has come up a couple of times.
I'll add a bit of changelog material about it in the last patch since
that's where the opt-in is introduced.
Powered by blists - more mailing lists