Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.36 as permitted sender) client-ip=23.128.96.36;
From:   "Huang, Ying" <ying.huang@intel.com>
To:     "zhangpeng (AS)" <zhangpeng362@huawei.com>
Cc:     Yin Fengwei <fengwei.yin@intel.com>, <linux-mm@kvack.org>,
        <linux-kernel@vger.kernel.org>, <akpm@linux-foundation.org>,
        <willy@infradead.org>, <aneesh.kumar@linux.ibm.com>,
        <shy828301@gmail.com>, <hughd@google.com>, <david@redhat.com>,
        <wangkefeng.wang@huawei.com>, <sunnanyong@huawei.com>
Subject: Re: [RFC PATCH] mm: filemap: avoid unnecessary major faults in
 filemap_fault()
In-Reply-To: <87y1en7pq3.fsf@yhuang6-desk2.ccr.corp.intel.com> (Ying Huang's
        message of "Fri, 24 Nov 2023 12:13:56 +0800")
References: <20231122140052.4092083-1-zhangpeng362@huawei.com>
        <f4dba5b5-2e6e-4c5a-a269-4abe8fe2bcd8@intel.com>
        <b0d869e4-108b-8ffe-e9f7-65c4d98f2bf8@huawei.com>
        <801bd0c9-7d0c-4231-93e5-7532e8231756@intel.com>
        <48235d73-3dc6-263d-7822-6d479b753d46@huawei.com>
        <87y1en7pq3.fsf@yhuang6-desk2.ccr.corp.intel.com>
Date:   Fri, 24 Nov 2023 12:26:36 +0800
Message-ID: <87ttpb7p4z.fsf@yhuang6-desk2.ccr.corp.intel.com>
User-Agent: Gnus/5.13 (Gnus v5.13)
MIME-Version: 1.0
Content-Type: text/plain; charset=utf-8
Content-Transfer-Encoding: quoted-printable
Precedence: bulk

"Huang, Ying" <ying.huang@intel.com> writes:

> "zhangpeng (AS)" <zhangpeng362@huawei.com> writes:
>
>> On 2023/11/23 13:26, Yin Fengwei wrote:
>>
>>> On 11/23/23 12:12, zhangpeng (AS) wrote:
>>>> On 2023/11/23 9:09, Yin Fengwei wrote:
>>>>
>>>>> Hi Peng,
>>>>>
>>>>> On 11/22/23 22:00, Peng Zhang wrote:
>>>>>> From: ZhangPeng <zhangpeng362@huawei.com>
>>>>>>
>>>>>> The major fault occurred when using mlockall(MCL_CURRENT | MCL_FUTUR=
E)
>>>>>> in application, which leading to an unexpected performance issue[1].
>>>>>>
>>>>>> This caused by temporarily cleared pte during a read/modify/write up=
date
>>>>>> of the pte, eg, do_numa_page()/change_pte_range().
>>>>>>
>>>>>> For the data segment of the user-mode program, the global variable a=
rea
>>>>>> is a private mapping. After the pagecache is loaded, the private ano=
nymous
>>>>>> page is generated after the COW is triggered. Mlockall can lock COW =
pages
>>>>>> (anonymous pages), but the original file pages cannot be locked and =
may
>>>>>> be reclaimed. If the global variable (private anon page) is accessed=
 when
>>>>>> vmf->pte is zeroed in numa fault, a file page fault will be triggere=
d.
>>>>>>
>>>>>> At this time, the original private file page may have been reclaimed.
>>>>>> If the page cache is not available at this time, a major fault will =
be
>>>>>> triggered and the file will be read, causing additional overhead.
>>>>>>
>>>>>> Fix this by rechecking the pte by holding ptl in filemap_fault() bef=
ore
>>>>>> triggering a major fault.
>>>>>>
>>>>>> [1] https://lore.kernel.org/linux-mm/9e62fd9a-bee0-52bf-50a7-498fa17=
434ee@huawei.com/
>>>>>>
>>>>>> Signed-off-by: ZhangPeng <zhangpeng362@huawei.com>
>>>>>> Signed-off-by: Kefeng Wang <wangkefeng.wang@huawei.com>
>>>>>> ---
>>>>>>  =C2=A0 mm/filemap.c | 14 ++++++++++++++
>>>>>>  =C2=A0 1 file changed, 14 insertions(+)
>>>>>>
>>>>>> diff --git a/mm/filemap.c b/mm/filemap.c
>>>>>> index 71f00539ac00..bb5e6a2790dc 100644
>>>>>> --- a/mm/filemap.c
>>>>>> +++ b/mm/filemap.c
>>>>>> @@ -3226,6 +3226,20 @@ vm_fault_t filemap_fault(struct vm_fault *vmf)
>>>>>>  =C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=
=C2=A0=C2=A0 mapping_locked =3D true;
>>>>>>  =C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0 }
>>>>>>  =C2=A0=C2=A0=C2=A0=C2=A0=C2=A0 } else {
>>>>>> +=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0 pte_t *ptep =3D pte_offs=
et_map_lock(vmf->vma->vm_mm, vmf->pmd,
>>>>>> +=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=
=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=
=A0=C2=A0 vmf->address, &vmf->ptl);
>>>>>> +=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0 if (ptep) {
>>>>>> +=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0 =
/*
>>>>>> +=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=
=C2=A0 * Recheck pte with ptl locked as the pte can be cleared
>>>>>> +=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=
=C2=A0 * temporarily during a read/modify/write update.
>>>>>> +=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=
=C2=A0 */
>>>>>> +=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0 =
if (unlikely(!pte_none(ptep_get(ptep))))
>>>>>> +=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=
=C2=A0=C2=A0=C2=A0=C2=A0 ret =3D VM_FAULT_NOPAGE;
>>>>>> +=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0 =
pte_unmap_unlock(ptep, vmf->ptl);
>>>>>> +=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0 =
if (unlikely(ret))
>>>>>> +=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=
=C2=A0=C2=A0=C2=A0=C2=A0 return ret;
>>>>>> +=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0 }
>>>>> I am curious. Did you try not to take PTL here and just check whether=
 PTE is not NONE?
>>>> Thank you for your reply.
>>>>
>>>> If we don't take PTL, the current use case won't trigger this issue ei=
ther.
>>> Is this verified by testing or just in theory?
>>
>> If we add a delay between ptep_modify_prot_start() and ptep_modify_prot_=
commit(),
>> this issue will also trigger. Without delay, we haven't reproduced this =
problem
>> so far.
>>
>>>> In most cases, if we don't take PTL, this issue won't be triggered. Ho=
wever,
>>>> there is still a possibility of triggering this issue. The corner case=
 is that
>>>> task 2 triggers a page fault when task 1 is between ptep_modify_prot_s=
tart()
>>>> and ptep_modify_prot_commit() in do_numa_page(). Furthermore,task 2 pa=
sses the
>>>> check whether the PTE is not NONE before task 1 updates PTE in
>>>> ptep_modify_prot_commit() without taking PTL.
>>> There is very limited operations between ptep_modify_prot_start() and
>>> ptep_modify_prot_commit(). While the code path from page fault to this =
check is
>>> long. My understanding is it's very likely the PTE is not NONE when do =
PTE check
>>> here without hold PTL (This is my theory. :)).
>>
>> Yes, there is a high probability that this issue won't occur without tak=
ing PTL.
>>
>>> In the other side, acquiring/releasing PTL may bring performance impact=
ion. It may
>>> not be big deal because the IO operations in this code path. But it's b=
etter to
>>> collect some performance data IMHO.
>>
>> We tested the performance of file private mapping page fault (page_fault=
2.c of
>> will-it-scale [1]) and file shared mapping page fault (page_fault3.c of =
will-it-scale).
>> The difference in performance (in operations per second) before and afte=
r patch
>> applied is about 0.7% on a x86 physical machine.
>
> Whether is it improvement or reduction?

And I think that you need to test ramdisk cases too to verify whether
this will cause performance regression and how much.

--
Best Regards,
Huang, Ying

> --
> Best Regards,
> Huang, Ying
>
>> [1] https://github.com/antonblanchard/will-it-scale/tree/master
>>
>>>
>>> Regards
>>> Yin, Fengwei
>>>
>>>>> Regards
>>>>> Yin, Fengwei
>>>>>
>>>>>> +
>>>>>>  =C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0 /* No page i=
n the page cache at all */
>>>>>>  =C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0 count_vm_eve=
nt(PGMAJFAULT);
>>>>>>  =C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0 count_memcg_=
event_mm(vmf->vma->vm_mm, PGMAJFAULT);