Received-SPF: pass (google.com: best guess record for domain of linux-kernel-owner@vger.kernel.org designates 209.132.180.67 as permitted sender) client-ip=209.132.180.67;
Date:   Tue, 3 Mar 2020 09:19:29 +0100
From:   Michal Hocko <mhocko@kernel.org>
To:     "Huang, Ying" <ying.huang@intel.com>
Cc:     David Hildenbrand <david@redhat.com>,
        Matthew Wilcox <willy@infradead.org>,
        Andrew Morton <akpm@linux-foundation.org>, linux-mm@kvack.org,
        linux-kernel@vger.kernel.org, Mel Gorman <mgorman@suse.de>,
        Vlastimil Babka <vbabka@suse.cz>, Zi Yan <ziy@nvidia.com>,
        Peter Zijlstra <peterz@infradead.org>,
        Dave Hansen <dave.hansen@linux.intel.com>,
        Minchan Kim <minchan@kernel.org>,
        Johannes Weiner <hannes@cmpxchg.org>,
        Hugh Dickins <hughd@google.com>,
        Alexander Duyck <alexander.duyck@gmail.com>
Subject: Re: [RFC 0/3] mm: Discard lazily freed pages when migrating
Message-ID: <20200303081929.GY4380@dhcp22.suse.cz>
References: <20200228033819.3857058-1-ying.huang@intel.com>
 <20200228034248.GE29971@bombadil.infradead.org>
 <87a7538977.fsf@yhuang-dev.intel.com>
 <edae2736-3239-0bdc-499c-560fc234c974@redhat.com>
 <871rqf850z.fsf@yhuang-dev.intel.com>
 <20200228095048.GK3771@dhcp22.suse.cz>
 <87d09u7sm2.fsf@yhuang-dev.intel.com>
 <20200302142549.GO4380@dhcp22.suse.cz>
 <874kv66x8r.fsf@yhuang-dev.intel.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <874kv66x8r.fsf@yhuang-dev.intel.com>
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

On Tue 03-03-20 09:30:28, Huang, Ying wrote:
[...]
> Yes.  mmap() can control whether to populate the underlying physical
> pages.

right because many usecases benefit from it. They simply know that the
mapping will be used completely and it is worth saving overhead for #PF.
See. there is a clear justification for that policy.

> But for migrating MADV_FREE pages, there's no control, all pages
> will be populated again always by default.  Maybe we should avoid to do
> that in some situations too.

Now let's have a look here. It is the userspace that decided to mark
MADV_FREE pages. It is under its full control which pages are to be
freed lazily. If the userspace wants to move those pages then it is
likely aware they have been MADV_FREE, right? If the userspace wanted to
save migration overhead then it could either chose to not migrate those
pages or simply unmap them right away. So in the end we are talking
about saving munmap/MAMDV_DONTNEED or potentially more move_pages calls
to skip over MADV_FREE holes. Which is all nice but is there any
userspace that really does care? Because this is a fundamental question
here and it doesn't make much sense to discuss this left to right unless
this is clear.
-- 
Michal Hocko
SUSE Labs