Received-SPF: pass (google.com: domain of linux-ext4-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
Date:   Thu, 15 Jun 2023 16:48:58 -0400
From:   Peter Xu <peterx@redhat.com>
To:     Matthew Wilcox <willy@infradead.org>
Cc:     Alex Sierra <alex.sierra@amd.com>, jgg@nvidia.com,
        david@redhat.com, Felix.Kuehling@amd.com, linux-mm@kvack.org,
        rcampbell@nvidia.com, linux-ext4@vger.kernel.org,
        linux-xfs@vger.kernel.org, amd-gfx@lists.freedesktop.org,
        dri-devel@lists.freedesktop.org, hch@lst.de, jglisse@redhat.com,
        apopple@nvidia.com, akpm@linux-foundation.org
Subject: Re: [PATCH v9 02/14] mm: move page zone helpers from mm.h to mmzone.h
Message-ID: <ZIt5Oho3enLFs+sv@x1n>
References: <20220715150521.18165-1-alex.sierra@amd.com>
 <20220715150521.18165-3-alex.sierra@amd.com>
 <ZItneGX+sqg7WApF@x1n>
 <ZItxXny9kRDq/ryf@casper.infradead.org>
MIME-Version: 1.0
Content-Type: text/plain; charset=utf-8
Content-Disposition: inline
In-Reply-To: <ZItxXny9kRDq/ryf@casper.infradead.org>
Precedence: bulk

On Thu, Jun 15, 2023 at 09:15:26PM +0100, Matthew Wilcox wrote:
> On Thu, Jun 15, 2023 at 03:33:12PM -0400, Peter Xu wrote:
> > My question is whether page_zonenum() is ready for taking all kinds of tail
> > pages?
> > 
> > Zone device tail pages all look fine, per memmap_init_zone_device().  The
> > question was other kinds of usual compound pages, like either thp or
> > hugetlb.  IIUC page->flags can be uninitialized for those tail pages.
> 
> I don't think that's true.  It's my understanding that page->flags is
> initialised for all pages in memmap at boot / hotplug / delayed-init
> time.  So you can check things like zone, node, etc on literally any
> page.  Contrariwise, those flags are not available in tail pages for
> use by the entity that has allocated a compound page / large folio.

Oh so the zone mask is special.  Fair enough.

> 
> Also, I don't believe zone device pages support compound allocation.
> I think they're always allocated as order-0.

Totally not familiar with zone device pages, but memmap_init_zone_device()
has pfns_per_compound which can be >1.  From there, memmap_init_compound()
does go ahead and setup pages as compound ones.

Thanks!

-- 
Peter Xu