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:   Wed, 3 Feb 2021 10:53:13 -0400
From:   Jason Gunthorpe <jgg@ziepe.ca>
To:     Joao Martins <joao.m.martins@oracle.com>
Cc:     Pavel Tatashin <pasha.tatashin@soleen.com>,
        linux-kernel@vger.kernel.org, linux-mm@kvack.org,
        akpm@linux-foundation.org, vbabka@suse.cz, mhocko@suse.com,
        david@redhat.com, osalvador@suse.de, dan.j.williams@intel.com,
        sashal@kernel.org, tyhicks@linux.microsoft.com,
        iamjoonsoo.kim@lge.com, mike.kravetz@oracle.com,
        rostedt@goodmis.org, mingo@redhat.com, peterz@infradead.org,
        mgorman@suse.de, willy@infradead.org, rientjes@google.com,
        jhubbard@nvidia.com, linux-doc@vger.kernel.org,
        ira.weiny@intel.com, linux-kselftest@vger.kernel.org,
        jmorris@namei.org
Subject: Re: [PATCH v8 02/14] mm/gup: check every subpage of a compound page
 during isolation
Message-ID: <20210203145313.GQ4718@ziepe.ca>
References: <20210125194751.1275316-1-pasha.tatashin@soleen.com>
 <20210125194751.1275316-3-pasha.tatashin@soleen.com>
 <05a66361-214c-2afe-22e4-12862ea1e4e2@oracle.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <05a66361-214c-2afe-22e4-12862ea1e4e2@oracle.com>
Precedence: bulk

On Wed, Feb 03, 2021 at 01:22:18PM +0000, Joao Martins wrote:

> With this, longterm gup will 'regress' for hugetlbfs e.g. from ~6k -> 32k usecs when
> pinning a 16G hugetlb file.

Yes, but correctness demands it.

The solution is to track these pages as we discover them so we know if
a PMD/PUD points and can directly skip the duplicated work

> Splitting can only occur on THP right? If so, perhaps we could
> retain the @step increment for compound pages but when
> !is_transparent_hugepage(head) or just PageHuge(head) like:

Honestly I'd rather see it fixed properly which will give even bigger
performance gains - avoiding the entire rescan of the page list will
be a win

Jason