Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
Date:   Fri, 14 Apr 2023 14:29:40 +0100
From:   Mel Gorman <mgorman@techsingularity.net>
To:     Matthew Wilcox <willy@infradead.org>
Cc:     Andrew Morton <akpm@linux-foundation.org>,
        Vlastimil Babka <vbabka@suse.cz>,
        Michal Hocko <mhocko@suse.com>,
        Oscar Salvador <OSalvador@suse.com>,
        Yuanxi Liu <y.liu@naruida.com>,
        David Hildenbrand <david@redhat.com>,
        Linux-MM <linux-mm@kvack.org>,
        LKML <linux-kernel@vger.kernel.org>
Subject: Re: [PATCH] mm: page_alloc: Assume huge tail pages are valid when
 allocating contiguous pages
Message-ID: <20230414132940.chjmsuiurnygajxs@techsingularity.net>
References: <20230414082222.idgw745cgcduzy37@techsingularity.net>
 <ZDlFbkwfmePEewXM@casper.infradead.org>
MIME-Version: 1.0
Content-Type: text/plain; charset=iso-8859-15
Content-Disposition: inline
In-Reply-To: <ZDlFbkwfmePEewXM@casper.infradead.org>
Precedence: bulk

On Fri, Apr 14, 2023 at 01:22:06PM +0100, Matthew Wilcox wrote:
> On Fri, Apr 14, 2023 at 09:22:22AM +0100, Mel Gorman wrote:
> > +		/*
> > +		 * Do not migrate huge pages that span the size of the region
> > +		 * being allocated contiguous. e.g. Do not migrate a 1G page
> > +		 * for a 1G allocation request. CMA is an exception as the
> > +		 * region may be reserved for hardware that requires physical
> > +		 * memory without a MMU or scatter/gather capability.
> > +		 *
> > +		 * Note that the compound check is race-prone versus
> > +		 * free/split/collapse but it should be safe and result in
> > +		 * a premature skip or a useless migration attempt.
> > +		 */
> > +		if (PageHuge(page) && compound_nr(page) >= nr_pages &&
> 
> This confuses me.  PageHuge() can be called on tail pages, but if
> compound_nr() is called on a tail page, it returns 1. 

Given the calling context is a linear scan, the head page will be scanned
first so the value for compound_nr() called on a tail page shouldn't occur.

> So I'm not
> sure why this works.  Also, do you really want PageHuge (ie only
> hugetlbfs pages), or do you really just want to check PageCompound(),
> which would also be true for THP?
> 

For now I only want hugetlbfs pages as the fix is for a regression when
allocating 1G hugetlbfs pages and previous behaviour avoided existing
hugetlbfs pages. THP pages can be split+migrated of course but the cost
of the 1G allocation attempt may be excessive relative to any benefit.

-- 
Mel Gorman
SUSE Labs