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;
Date:   Fri, 26 Feb 2021 20:27:38 +0000
From:   Matthew Wilcox <willy@infradead.org>
To:     Yu Zhao <yuzhao@google.com>
Cc:     akpm@linux-foundation.org, alex.shi@linux.alibaba.com,
        vbabka@suse.cz, guro@fb.com, hannes@cmpxchg.org, hughd@google.com,
        linux-kernel@vger.kernel.org, linux-mm@kvack.org,
        mhocko@kernel.org, vdavydov.dev@gmail.com
Subject: Re: [PATCH v2 3/3] mm: use PF_ONLY_HEAD for PG_active and
 PG_unevictable
Message-ID: <20210226202738.GG2723601@casper.infradead.org>
References: <20210224084807.2179942-1-yuzhao@google.com>
 <20210226091718.2927291-1-yuzhao@google.com>
 <20210226091718.2927291-4-yuzhao@google.com>
 <20210226121314.GB2723601@casper.infradead.org>
 <YDlQ1VLtGsTbwp4z@google.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <YDlQ1VLtGsTbwp4z@google.com>
Precedence: bulk

On Fri, Feb 26, 2021 at 12:49:41PM -0700, Yu Zhao wrote:
> On Fri, Feb 26, 2021 at 12:13:14PM +0000, Matthew Wilcox wrote:
> > On Fri, Feb 26, 2021 at 02:17:18AM -0700, Yu Zhao wrote:
> > > All places but one test, set or clear PG_active and PG_unevictable on
> > > small or head pages. Use compound_head() explicitly for that singleton
> > > so the rest can rid of redundant compound_head().
> > 
> > How do you know it's only one place?  I really wish you'd work with me
> > on folios.  They make the compiler prove that it's not a tail page.
> 
> I hasn't been following the effort closely, so I'm rereading the very
> first discussion "Are THPs the right model for the pagecache?" and
> then I need to rewatch the recorded Zoom meeting. As I said I'm on
> board with the idea, but I can't create a roadmap based on my current
> rough understanding, unless you prefer me to just randomly throw some
> reviewed-bys at your patches in the next few days. (Our long-term plan
> for folios is to support arbitrary anon page sizes because anon memory
> is more than 90% of total user memory on Android, Chrome OS and in our
> data centers.)
> 
> That said, if you have something ready to test, we could do it for you
> in our runtime environments immediately.

I don't have anything ready to test for anonymous memory; indeed I have no
plans to work on anonymous memory myself.  My focus is on the page cache.

But, once we get the folio _concept_ into the kernel (ie a struct page
which is definitely not a tail page), it can be used more widely than
the page cache, and independently from anything I'm working on.  The
biggest risk is that we end up duplicating work ...