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=utf-8
MIME-Version: 1.0
Subject: Re: [PATCH] mm: hugetlb_vmemmap: use PageCompound() instead of
 PageReserved()
From:   Muchun Song <muchun.song@linux.dev>
In-Reply-To: <2b66797b-3ee9-a70f-25d0-dcab830859d2@redhat.com>
Date:   Wed, 19 Jul 2023 15:00:04 +0800
Cc:     Muchun Song <songmuchun@bytedance.com>,
        Mike Kravetz <mike.kravetz@oracle.com>,
        Andrew Morton <akpm@linux-foundation.org>,
        Linux Memory Management List <linux-mm@kvack.org>,
        linux-kernel@vger.kernel.org
Content-Transfer-Encoding: quoted-printable
Message-Id: <A363D1BA-80B6-4E71-B587-EF04530BBA77@linux.dev>
References: <20230719063132.37676-1-songmuchun@bytedance.com>
 <2b66797b-3ee9-a70f-25d0-dcab830859d2@redhat.com>
To:     =?utf-8?Q?Mika_Penttil=C3=A4?= <mpenttil@redhat.com>
Precedence: bulk


> On Jul 19, 2023, at 14:51, Mika Penttil=C3=A4 <mpenttil@redhat.com> =
wrote:
>=20
> Hi,
>=20
>=20
> On 19.7.2023 9.31, Muchun Song wrote:
>> The ckeck of PageReserved() is easy to be broken in the future, =
PageCompound()
>> is more stable to check if the page should be split.
>> Signed-off-by: Muchun Song <songmuchun@bytedance.com>
>> ---
>>  mm/hugetlb_vmemmap.c | 4 +++-
>>  1 file changed, 3 insertions(+), 1 deletion(-)
>> diff --git a/mm/hugetlb_vmemmap.c b/mm/hugetlb_vmemmap.c
>> index 4b9734777f69..8068fe890f52 100644
>> --- a/mm/hugetlb_vmemmap.c
>> +++ b/mm/hugetlb_vmemmap.c
>> @@ -73,8 +73,10 @@ static int split_vmemmap_huge_pmd(pmd_t *pmd, =
unsigned long start)
>>   * be treated as indepdenent small pages (as they can be freed
>>   * individually).
>>   */
>> - if (!PageReserved(head))
>> + if (PageCompound(head)) {
>> + 	VM_BUG_ON(compound_order(head) !=3D get_order(PMD_SIZE));
>>   	split_page(head, get_order(PMD_SIZE));
>=20
> I think vmemmap pages are not compound pages (even order > 0).
> They would bug on here trying to split_page() :

You are right. I have missed this. Now I remember why I use
PageReserved() instead of PageCompound() when I fist submit
the commit 39d35edee453. Thanks for your reminder.

Sorry for the noise. Please ignore this patch.

Thanks.

>=20
>=20
> void split_page(struct page *page, unsigned int order)
> {
>        int i;
>=20
>        VM_BUG_ON_PAGE(PageCompound(page), page);
>=20
>=20
>> + }
>>     /* Make pte visible before pmd. See comment in pmd_install(). */
>>   smp_wmb();
>=20
> --Mika
>=20