Received-SPF: pass (google.com: best guess record for domain of linux-kernel-owner@vger.kernel.org designates 209.132.180.67 as permitted sender) client-ip=209.132.180.67;
Subject: Re: [PATCH] hugetlbfs: Take read_lock on i_mmap for PMD sharing
To:     Mike Kravetz <mike.kravetz@oracle.com>,
        Andrew Morton <akpm@linux-foundation.org>
Cc:     linux-kernel@vger.kernel.org, linux-mm@kvack.org,
        Davidlohr Bueso <dave@stgolabs.net>,
        Peter Zijlstra <peterz@infradead.org>,
        Ingo Molnar <mingo@redhat.com>,
        Will Deacon <will.deacon@arm.com>
References: <20191107190628.22667-1-longman@redhat.com>
From:   Waiman Long <longman@redhat.com>
Organization: Red Hat
Message-ID: <cd7f02d2-7f20-4176-4d8a-9f663d6b8ba9@redhat.com>
Date:   Thu, 7 Nov 2019 14:15:43 -0500
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:60.0) Gecko/20100101
 Thunderbird/60.7.2
MIME-Version: 1.0
In-Reply-To: <20191107190628.22667-1-longman@redhat.com>
Content-Language: en-US
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: quoted-printable
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

On 11/7/19 2:06 PM, Waiman Long wrote:
> A customer with large SMP systems (up to 16 sockets) with application
> that uses large amount of static hugepages (~500-1500GB) are experiencing
> random multisecond delays. These delays was caused by the long time it
> took to scan the VMA interval tree with mmap_sem held.
>
> The sharing of huge PMD does not require changes to the i_mmap at all.
> As a result, we can just take the read lock and let other threads
> searching for the right VMA to share in parallel. Once the right
> VMA is found, either the PMD lock (2M huge page for x86-64) or the
> mm->page_table_lock will be acquired to perform the actual PMD sharing.
>
> Lock contention, if present, will happen in the spinlock. That is much
> better than contention in the rwsem where the time needed to scan the
> the interval tree is indeterminate.
>
> With this patch applied, the customer is seeing significant improvements
> over the unpatched kernel.

Oh, I forgot to add

Suggested-by: Mike Kravetz <mike.kravetz@oracle.com>

> Signed-off-by: Waiman Long <longman@redhat.com>
> ---
>  mm/hugetlb.c | 8 ++++++--
>  1 file changed, 6 insertions(+), 2 deletions(-)
>
> diff --git a/mm/hugetlb.c b/mm/hugetlb.c
> index b45a95363a84..087e7ff00137 100644
> --- a/mm/hugetlb.c
> +++ b/mm/hugetlb.c
> @@ -4842,7 +4842,11 @@ pte_t *huge_pmd_share(struct mm_struct *mm, unsign=
ed long addr, pud_t *pud)
>  =09if (!vma_shareable(vma, addr))
>  =09=09return (pte_t *)pmd_alloc(mm, pud, addr);
> =20
> -=09i_mmap_lock_write(mapping);
> +=09/*
> +=09 * PMD sharing does not require changes to i_mmap. So a read lock
> +=09 * is enuogh.
> +=09 */
> +=09i_mmap_lock_read(mapping);
>  =09vma_interval_tree_foreach(svma, &mapping->i_mmap, idx, idx) {
>  =09=09if (svma =3D=3D vma)
>  =09=09=09continue;
> @@ -4872,7 +4876,7 @@ pte_t *huge_pmd_share(struct mm_struct *mm, unsigne=
d long addr, pud_t *pud)
>  =09spin_unlock(ptl);
>  out:
>  =09pte =3D (pte_t *)pmd_alloc(mm, pud, addr);
> -=09i_mmap_unlock_write(mapping);
> +=09i_mmap_unlock_read(mapping);
>  =09return pte;
>  }
> =20

Cheers,
Longman