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;
Content-Type: text/plain;
        charset=us-ascii
MIME-Version: 1.0
Subject: Re: [PATCH] mm: hugetlb_vmemmap: fix a race between vmemmap pmd split
From:   Muchun Song <muchun.song@linux.dev>
In-Reply-To: <20230707124116.d7be4ba6504a452d11e2629d@linux-foundation.org>
Date:   Sat, 8 Jul 2023 09:40:30 +0800
Cc:     Muchun Song <songmuchun@bytedance.com>,
        Mike Kravetz <mike.kravetz@oracle.com>,
        Linux Memory Management List <linux-mm@kvack.org>,
        LKML <linux-kernel@vger.kernel.org>
Content-Transfer-Encoding: quoted-printable
Message-Id: <07A5567A-8BED-40DB-88F4-96C37DCD78BF@linux.dev>
References: <20230707033859.16148-1-songmuchun@bytedance.com>
 <20230707124116.d7be4ba6504a452d11e2629d@linux-foundation.org>
To:     Andrew Morton <akpm@linux-foundation.org>
Precedence: bulk


> On Jul 8, 2023, at 03:41, Andrew Morton <akpm@linux-foundation.org> =
wrote:
>=20
> On Fri,  7 Jul 2023 11:38:59 +0800 Muchun Song =
<songmuchun@bytedance.com> wrote:
>=20
>> And inline __split_vmemmap_huge_pmd()
>> since it only has one user.
>=20
> "open code" would be a better term than "inline" in this situation.
>=20
> If we are to offer this change to -stable then it would be better to =
do
> the open-coding of __split_vmemmap_huge_pmd() in a separate, later
> patch.
>=20

I see. Bug fix is better to "open code" instead of "inline". However, it
is a simpler and cleaner way to fix this bug by using "inline". Because
we must hold init_mm.page_table_lock to get the page table page in =
__split_vmemmap_huge_pmd(), then it is just a couple of duplicated
code from split_vmemmap_huge_pmd(). Consequently, =
split_vmemmap_huge_pmd()
is redundant, just remove it. And rename __split_vmemmap_huge_pmd()
to split_vmemmap_huge_pmd(). The result is the same as the "inline" way.
So I keep "inline" to fix this.

Thanks.