diff for duplicates of <9FE19350E8A7EE45B64D8D63D368C8966B85F660@SHSMSX101.ccr.corp.intel.com>
diff --git a/a/1.txt b/N1/1.txt
index ec70551..2a4493a 100644
--- a/a/1.txt
+++ b/N1/1.txt
@@ -1,613 +1,827 @@
-Hi Laurent,
-
-
-For the test result on Intel 4s skylake platform (192 CPUs, 768G Memory), the below test cases all were run 3 times.
-I check the test results, only page_fault3_thread/enable THP have 6% stddev for head commit, other tests have lower stddev.
-
-And I did not find other high variation on test case result.
-
-a). Enable THP
-testcase base stddev change head stddev metric
-page_fault3/enable THP 10519 ± 3% -20.5% 8368 ±6% will-it-scale.per_thread_ops
-page_fault2/enalbe THP 8281 ± 2% -18.8% 6728 will-it-scale.per_thread_ops
-brk1/eanble THP 998475 -2.2% 976893 will-it-scale.per_process_ops
-context_switch1/enable THP 223910 -1.3% 220930 will-it-scale.per_process_ops
-context_switch1/enable THP 233722 -1.0% 231288 will-it-scale.per_thread_ops
-
-b). Disable THP
-page_fault3/disable THP 10856 -23.1% 8344 will-it-scale.per_thread_ops
-page_fault2/disable THP 8147 -18.8% 6613 will-it-scale.per_thread_ops
-brk1/disable THP 957 -7.9% 881 will-it-scale.per_thread_ops
-context_switch1/disable THP 237006 -2.2% 231907 will-it-scale.per_thread_ops
-brk1/disable THP 997317 -2.0% 977778 will-it-scale.per_process_ops
-page_fault3/disable THP 467454 -1.8% 459251 will-it-scale.per_process_ops
-context_switch1/disable THP 224431 -1.3% 221567 will-it-scale.per_process_ops
-
-
-Best regards,
-Haiyan Song
-________________________________________
-From: Laurent Dufour [ldufour@linux.vnet.ibm.com]
-Sent: Monday, July 02, 2018 4:59 PM
-To: Song, HaiyanX
-Cc: akpm@linux-foundation.org; mhocko@kernel.org; peterz@infradead.org; kirill@shutemov.name; ak@linux.intel.com; dave@stgolabs.net; jack@suse.cz; Matthew Wilcox; khandual@linux.vnet.ibm.com; aneesh.kumar@linux.vnet.ibm.com; benh@kernel.crashing.org; mpe@ellerman.id.au; paulus@samba.org; Thomas Gleixner; Ingo Molnar; hpa@zytor.com; Will Deacon; Sergey Senozhatsky; sergey.senozhatsky.work@gmail.com; Andrea Arcangeli; Alexei Starovoitov; Wang, Kemi; Daniel Jordan; David Rientjes; Jerome Glisse; Ganesh Mahendran; Minchan Kim; Punit Agrawal; vinayak menon; Yang Shi; linux-kernel@vger.kernel.org; linux-mm@kvack.org; haren@linux.vnet.ibm.com; npiggin@gmail.com; bsingharora@gmail.com; paulmck@linux.vnet.ibm.com; Tim Chen; linuxppc-dev@lists.ozlabs.org; x86@kernel.org
-Subject: Re: [PATCH v11 00/26] Speculative page faults
-
-On 11/06/2018 09:49, Song, HaiyanX wrote:
-> Hi Laurent,
->
-> Regression test for v11 patch serials have been run, some regression is found by LKP-tools (linux kernel performance)
-> tested on Intel 4s skylake platform. This time only test the cases which have been run and found regressions on
-> V9 patch serials.
->
-> The regression result is sorted by the metric will-it-scale.per_thread_ops.
-> branch: Laurent-Dufour/Speculative-page-faults/20180520-045126
-> commit id:
-> head commit : a7a8993bfe3ccb54ad468b9f1799649e4ad1ff12
-> base commit : ba98a1cdad71d259a194461b3a61471b49b14df1
-> Benchmark: will-it-scale
-> Download link: https://github.com/antonblanchard/will-it-scale/tree/master
->
-> Metrics:
-> will-it-scale.per_process_ops=processes/nr_cpu
-> will-it-scale.per_thread_ops=threads/nr_cpu
-> test box: lkp-skl-4sp1(nr_cpu=192,memory=768G)
-> THP: enable / disable
-> nr_task:100%
->
-> 1. Regressions:
->
-> a). Enable THP
-> testcase base change head metric
-> page_fault3/enable THP 10519 -20.5% 836 will-it-scale.per_thread_ops
-> page_fault2/enalbe THP 8281 -18.8% 6728 will-it-scale.per_thread_ops
-> brk1/eanble THP 998475 -2.2% 976893 will-it-scale.per_process_ops
-> context_switch1/enable THP 223910 -1.3% 220930 will-it-scale.per_process_ops
-> context_switch1/enable THP 233722 -1.0% 231288 will-it-scale.per_thread_ops
->
-> b). Disable THP
-> page_fault3/disable THP 10856 -23.1% 8344 will-it-scale.per_thread_ops
-> page_fault2/disable THP 8147 -18.8% 6613 will-it-scale.per_thread_ops
-> brk1/disable THP 957 -7.9% 881 will-it-scale.per_thread_ops
-> context_switch1/disable THP 237006 -2.2% 231907 will-it-scale.per_thread_ops
-> brk1/disable THP 997317 -2.0% 977778 will-it-scale.per_process_ops
-> page_fault3/disable THP 467454 -1.8% 459251 will-it-scale.per_process_ops
-> context_switch1/disable THP 224431 -1.3% 221567 will-it-scale.per_process_ops
->
-> Notes: for the above values of test result, the higher is better.
-
-I tried the same tests on my PowerPC victim VM (1024 CPUs, 11TB) and I can't
-get reproducible results. The results have huge variation, even on the vanilla
-kernel, and I can't state on any changes due to that.
-
-I tried on smaller node (80 CPUs, 32G), and the tests ran better, but I didn't
-measure any changes between the vanilla and the SPF patched ones:
-
-test THP enabled 4.17.0-rc4-mm1 spf delta
-page_fault3_threads 2697.7 2683.5 -0.53%
-page_fault2_threads 170660.6 169574.1 -0.64%
-context_switch1_threads 6915269.2 6877507.3 -0.55%
-context_switch1_processes 6478076.2 6529493.5 0.79%
-brk1 243391.2 238527.5 -2.00%
-
-Tests were run 10 times, no high variation detected.
-
-Did you see high variation on your side ? How many times the test were run to
-compute the average values ?
-
-Thanks,
-Laurent.
-
-
->
-> 2. Improvement: not found improvement based on the selected test cases.
->
->
-> Best regards
-> Haiyan Song
-> ________________________________________
-> From: owner-linux-mm@kvack.org [owner-linux-mm@kvack.org] on behalf of Laurent Dufour [ldufour@linux.vnet.ibm.com]
-> Sent: Monday, May 28, 2018 4:54 PM
-> To: Song, HaiyanX
-> Cc: akpm@linux-foundation.org; mhocko@kernel.org; peterz@infradead.org; kirill@shutemov.name; ak@linux.intel.com; dave@stgolabs.net; jack@suse.cz; Matthew Wilcox; khandual@linux.vnet.ibm.com; aneesh.kumar@linux.vnet.ibm.com; benh@kernel.crashing.org; mpe@ellerman.id.au; paulus@samba.org; Thomas Gleixner; Ingo Molnar; hpa@zytor.com; Will Deacon; Sergey Senozhatsky; sergey.senozhatsky.work@gmail.com; Andrea Arcangeli; Alexei Starovoitov; Wang, Kemi; Daniel Jordan; David Rientjes; Jerome Glisse; Ganesh Mahendran; Minchan Kim; Punit Agrawal; vinayak menon; Yang Shi; linux-kernel@vger.kernel.org; linux-mm@kvack.org; haren@linux.vnet.ibm.com; npiggin@gmail.com; bsingharora@gmail.com; paulmck@linux.vnet.ibm.com; Tim Chen; linuxppc-dev@lists.ozlabs.org; x86@kernel.org
-> Subject: Re: [PATCH v11 00/26] Speculative page faults
->
-> On 28/05/2018 10:22, Haiyan Song wrote:
->> Hi Laurent,
->>
->> Yes, these tests are done on V9 patch.
->
-> Do you plan to give this V11 a run ?
->
->>
->>
->> Best regards,
->> Haiyan Song
->>
->> On Mon, May 28, 2018 at 09:51:34AM +0200, Laurent Dufour wrote:
->>> On 28/05/2018 07:23, Song, HaiyanX wrote:
->>>>
->>>> Some regression and improvements is found by LKP-tools(linux kernel performance) on V9 patch series
->>>> tested on Intel 4s Skylake platform.
->>>
->>> Hi,
->>>
->>> Thanks for reporting this benchmark results, but you mentioned the "V9 patch
->>> series" while responding to the v11 header series...
->>> Were these tests done on v9 or v11 ?
->>>
->>> Cheers,
->>> Laurent.
->>>
->>>>
->>>> The regression result is sorted by the metric will-it-scale.per_thread_ops.
->>>> Branch: Laurent-Dufour/Speculative-page-faults/20180316-151833 (V9 patch series)
->>>> Commit id:
->>>> base commit: d55f34411b1b126429a823d06c3124c16283231f
->>>> head commit: 0355322b3577eeab7669066df42c550a56801110
->>>> Benchmark suite: will-it-scale
->>>> Download link:
->>>> https://github.com/antonblanchard/will-it-scale/tree/master/tests
->>>> Metrics:
->>>> will-it-scale.per_process_ops=processes/nr_cpu
->>>> will-it-scale.per_thread_ops=threads/nr_cpu
->>>> test box: lkp-skl-4sp1(nr_cpu=192,memory=768G)
->>>> THP: enable / disable
->>>> nr_task: 100%
->>>>
->>>> 1. Regressions:
->>>> a) THP enabled:
->>>> testcase base change head metric
->>>> page_fault3/ enable THP 10092 -17.5% 8323 will-it-scale.per_thread_ops
->>>> page_fault2/ enable THP 8300 -17.2% 6869 will-it-scale.per_thread_ops
->>>> brk1/ enable THP 957.67 -7.6% 885 will-it-scale.per_thread_ops
->>>> page_fault3/ enable THP 172821 -5.3% 163692 will-it-scale.per_process_ops
->>>> signal1/ enable THP 9125 -3.2% 8834 will-it-scale.per_process_ops
->>>>
->>>> b) THP disabled:
->>>> testcase base change head metric
->>>> page_fault3/ disable THP 10107 -19.1% 8180 will-it-scale.per_thread_ops
->>>> page_fault2/ disable THP 8432 -17.8% 6931 will-it-scale.per_thread_ops
->>>> context_switch1/ disable THP 215389 -6.8% 200776 will-it-scale.per_thread_ops
->>>> brk1/ disable THP 939.67 -6.6% 877.33 will-it-scale.per_thread_ops
->>>> page_fault3/ disable THP 173145 -4.7% 165064 will-it-scale.per_process_ops
->>>> signal1/ disable THP 9162 -3.9% 8802 will-it-scale.per_process_ops
->>>>
->>>> 2. Improvements:
->>>> a) THP enabled:
->>>> testcase base change head metric
->>>> malloc1/ enable THP 66.33 +469.8% 383.67 will-it-scale.per_thread_ops
->>>> writeseek3/ enable THP 2531 +4.5% 2646 will-it-scale.per_thread_ops
->>>> signal1/ enable THP 989.33 +2.8% 1016 will-it-scale.per_thread_ops
->>>>
->>>> b) THP disabled:
->>>> testcase base change head metric
->>>> malloc1/ disable THP 90.33 +417.3% 467.33 will-it-scale.per_thread_ops
->>>> read2/ disable THP 58934 +39.2% 82060 will-it-scale.per_thread_ops
->>>> page_fault1/ disable THP 8607 +36.4% 11736 will-it-scale.per_thread_ops
->>>> read1/ disable THP 314063 +12.7% 353934 will-it-scale.per_thread_ops
->>>> writeseek3/ disable THP 2452 +12.5% 2759 will-it-scale.per_thread_ops
->>>> signal1/ disable THP 971.33 +5.5% 1024 will-it-scale.per_thread_ops
->>>>
->>>> Notes: for above values in column "change", the higher value means that the related testcase result
->>>> on head commit is better than that on base commit for this benchmark.
->>>>
->>>>
->>>> Best regards
->>>> Haiyan Song
->>>>
->>>> ________________________________________
->>>> From: owner-linux-mm@kvack.org [owner-linux-mm@kvack.org] on behalf of Laurent Dufour [ldufour@linux.vnet.ibm.com]
->>>> Sent: Thursday, May 17, 2018 7:06 PM
->>>> To: akpm@linux-foundation.org; mhocko@kernel.org; peterz@infradead.org; kirill@shutemov.name; ak@linux.intel.com; dave@stgolabs.net; jack@suse.cz; Matthew Wilcox; khandual@linux.vnet.ibm.com; aneesh.kumar@linux.vnet.ibm.com; benh@kernel.crashing.org; mpe@ellerman.id.au; paulus@samba.org; Thomas Gleixner; Ingo Molnar; hpa@zytor.com; Will Deacon; Sergey Senozhatsky; sergey.senozhatsky.work@gmail.com; Andrea Arcangeli; Alexei Starovoitov; Wang, Kemi; Daniel Jordan; David Rientjes; Jerome Glisse; Ganesh Mahendran; Minchan Kim; Punit Agrawal; vinayak menon; Yang Shi
->>>> Cc: linux-kernel@vger.kernel.org; linux-mm@kvack.org; haren@linux.vnet.ibm.com; npiggin@gmail.com; bsingharora@gmail.com; paulmck@linux.vnet.ibm.com; Tim Chen; linuxppc-dev@lists.ozlabs.org; x86@kernel.org
->>>> Subject: [PATCH v11 00/26] Speculative page faults
->>>>
->>>> This is a port on kernel 4.17 of the work done by Peter Zijlstra to handle
->>>> page fault without holding the mm semaphore [1].
->>>>
->>>> The idea is to try to handle user space page faults without holding the
->>>> mmap_sem. This should allow better concurrency for massively threaded
->>>> process since the page fault handler will not wait for other threads memory
->>>> layout change to be done, assuming that this change is done in another part
->>>> of the process's memory space. This type page fault is named speculative
->>>> page fault. If the speculative page fault fails because of a concurrency is
->>>> detected or because underlying PMD or PTE tables are not yet allocating, it
->>>> is failing its processing and a classic page fault is then tried.
->>>>
->>>> The speculative page fault (SPF) has to look for the VMA matching the fault
->>>> address without holding the mmap_sem, this is done by introducing a rwlock
->>>> which protects the access to the mm_rb tree. Previously this was done using
->>>> SRCU but it was introducing a lot of scheduling to process the VMA's
->>>> freeing operation which was hitting the performance by 20% as reported by
->>>> Kemi Wang [2]. Using a rwlock to protect access to the mm_rb tree is
->>>> limiting the locking contention to these operations which are expected to
->>>> be in a O(log n) order. In addition to ensure that the VMA is not freed in
->>>> our back a reference count is added and 2 services (get_vma() and
->>>> put_vma()) are introduced to handle the reference count. Once a VMA is
->>>> fetched from the RB tree using get_vma(), it must be later freed using
->>>> put_vma(). I can't see anymore the overhead I got while will-it-scale
->>>> benchmark anymore.
->>>>
->>>> The VMA's attributes checked during the speculative page fault processing
->>>> have to be protected against parallel changes. This is done by using a per
->>>> VMA sequence lock. This sequence lock allows the speculative page fault
->>>> handler to fast check for parallel changes in progress and to abort the
->>>> speculative page fault in that case.
->>>>
->>>> Once the VMA has been found, the speculative page fault handler would check
->>>> for the VMA's attributes to verify that the page fault has to be handled
->>>> correctly or not. Thus, the VMA is protected through a sequence lock which
->>>> allows fast detection of concurrent VMA changes. If such a change is
->>>> detected, the speculative page fault is aborted and a *classic* page fault
->>>> is tried. VMA sequence lockings are added when VMA attributes which are
->>>> checked during the page fault are modified.
->>>>
->>>> When the PTE is fetched, the VMA is checked to see if it has been changed,
->>>> so once the page table is locked, the VMA is valid, so any other changes
->>>> leading to touching this PTE will need to lock the page table, so no
->>>> parallel change is possible at this time.
->>>>
->>>> The locking of the PTE is done with interrupts disabled, this allows
->>>> checking for the PMD to ensure that there is not an ongoing collapsing
->>>> operation. Since khugepaged is firstly set the PMD to pmd_none and then is
->>>> waiting for the other CPU to have caught the IPI interrupt, if the pmd is
->>>> valid at the time the PTE is locked, we have the guarantee that the
->>>> collapsing operation will have to wait on the PTE lock to move forward.
->>>> This allows the SPF handler to map the PTE safely. If the PMD value is
->>>> different from the one recorded at the beginning of the SPF operation, the
->>>> classic page fault handler will be called to handle the operation while
->>>> holding the mmap_sem. As the PTE lock is done with the interrupts disabled,
->>>> the lock is done using spin_trylock() to avoid dead lock when handling a
->>>> page fault while a TLB invalidate is requested by another CPU holding the
->>>> PTE.
->>>>
->>>> In pseudo code, this could be seen as:
->>>> speculative_page_fault()
->>>> {
->>>> vma = get_vma()
->>>> check vma sequence count
->>>> check vma's support
->>>> disable interrupt
->>>> check pgd,p4d,...,pte
->>>> save pmd and pte in vmf
->>>> save vma sequence counter in vmf
->>>> enable interrupt
->>>> check vma sequence count
->>>> handle_pte_fault(vma)
->>>> ..
->>>> page = alloc_page()
->>>> pte_map_lock()
->>>> disable interrupt
->>>> abort if sequence counter has changed
->>>> abort if pmd or pte has changed
->>>> pte map and lock
->>>> enable interrupt
->>>> if abort
->>>> free page
->>>> abort
->>>> ...
->>>> }
->>>>
->>>> arch_fault_handler()
->>>> {
->>>> if (speculative_page_fault(&vma))
->>>> goto done
->>>> again:
->>>> lock(mmap_sem)
->>>> vma = find_vma();
->>>> handle_pte_fault(vma);
->>>> if retry
->>>> unlock(mmap_sem)
->>>> goto again;
->>>> done:
->>>> handle fault error
->>>> }
->>>>
->>>> Support for THP is not done because when checking for the PMD, we can be
->>>> confused by an in progress collapsing operation done by khugepaged. The
->>>> issue is that pmd_none() could be true either if the PMD is not already
->>>> populated or if the underlying PTE are in the way to be collapsed. So we
->>>> cannot safely allocate a PMD if pmd_none() is true.
->>>>
->>>> This series add a new software performance event named 'speculative-faults'
->>>> or 'spf'. It counts the number of successful page fault event handled
->>>> speculatively. When recording 'faults,spf' events, the faults one is
->>>> counting the total number of page fault events while 'spf' is only counting
->>>> the part of the faults processed speculatively.
->>>>
->>>> There are some trace events introduced by this series. They allow
->>>> identifying why the page faults were not processed speculatively. This
->>>> doesn't take in account the faults generated by a monothreaded process
->>>> which directly processed while holding the mmap_sem. This trace events are
->>>> grouped in a system named 'pagefault', they are:
->>>> - pagefault:spf_vma_changed : if the VMA has been changed in our back
->>>> - pagefault:spf_vma_noanon : the vma->anon_vma field was not yet set.
->>>> - pagefault:spf_vma_notsup : the VMA's type is not supported
->>>> - pagefault:spf_vma_access : the VMA's access right are not respected
->>>> - pagefault:spf_pmd_changed : the upper PMD pointer has changed in our
->>>> back.
->>>>
->>>> To record all the related events, the easier is to run perf with the
->>>> following arguments :
->>>> $ perf stat -e 'faults,spf,pagefault:*' <command>
->>>>
->>>> There is also a dedicated vmstat counter showing the number of successful
->>>> page fault handled speculatively. I can be seen this way:
->>>> $ grep speculative_pgfault /proc/vmstat
->>>>
->>>> This series builds on top of v4.16-mmotm-2018-04-13-17-28 and is functional
->>>> on x86, PowerPC and arm64.
->>>>
->>>> ---------------------
->>>> Real Workload results
->>>>
->>>> As mentioned in previous email, we did non official runs using a "popular
->>>> in memory multithreaded database product" on 176 cores SMT8 Power system
->>>> which showed a 30% improvements in the number of transaction processed per
->>>> second. This run has been done on the v6 series, but changes introduced in
->>>> this new version should not impact the performance boost seen.
->>>>
->>>> Here are the perf data captured during 2 of these runs on top of the v8
->>>> series:
->>>> vanilla spf
->>>> faults 89.418 101.364 +13%
->>>> spf n/a 97.989
->>>>
->>>> With the SPF kernel, most of the page fault were processed in a speculative
->>>> way.
->>>>
->>>> Ganesh Mahendran had backported the series on top of a 4.9 kernel and gave
->>>> it a try on an android device. He reported that the application launch time
->>>> was improved in average by 6%, and for large applications (~100 threads) by
->>>> 20%.
->>>>
->>>> Here are the launch time Ganesh mesured on Android 8.0 on top of a Qcom
->>>> MSM845 (8 cores) with 6GB (the less is better):
->>>>
->>>> Application 4.9 4.9+spf delta
->>>> com.tencent.mm 416 389 -7%
->>>> com.eg.android.AlipayGphone 1135 986 -13%
->>>> com.tencent.mtt 455 454 0%
->>>> com.qqgame.hlddz 1497 1409 -6%
->>>> com.autonavi.minimap 711 701 -1%
->>>> com.tencent.tmgp.sgame 788 748 -5%
->>>> com.immomo.momo 501 487 -3%
->>>> com.tencent.peng 2145 2112 -2%
->>>> com.smile.gifmaker 491 461 -6%
->>>> com.baidu.BaiduMap 479 366 -23%
->>>> com.taobao.taobao 1341 1198 -11%
->>>> com.baidu.searchbox 333 314 -6%
->>>> com.tencent.mobileqq 394 384 -3%
->>>> com.sina.weibo 907 906 0%
->>>> com.youku.phone 816 731 -11%
->>>> com.happyelements.AndroidAnimal.qq 763 717 -6%
->>>> com.UCMobile 415 411 -1%
->>>> com.tencent.tmgp.ak 1464 1431 -2%
->>>> com.tencent.qqmusic 336 329 -2%
->>>> com.sankuai.meituan 1661 1302 -22%
->>>> com.netease.cloudmusic 1193 1200 1%
->>>> air.tv.douyu.android 4257 4152 -2%
->>>>
->>>> ------------------
->>>> Benchmarks results
->>>>
->>>> Base kernel is v4.17.0-rc4-mm1
->>>> SPF is BASE + this series
->>>>
->>>> Kernbench:
->>>> ----------
->>>> Here are the results on a 16 CPUs X86 guest using kernbench on a 4.15
->>>> kernel (kernel is build 5 times):
->>>>
->>>> Average Half load -j 8
->>>> Run (std deviation)
->>>> BASE SPF
->>>> Elapsed Time 1448.65 (5.72312) 1455.84 (4.84951) 0.50%
->>>> User Time 10135.4 (30.3699) 10148.8 (31.1252) 0.13%
->>>> System Time 900.47 (2.81131) 923.28 (7.52779) 2.53%
->>>> Percent CPU 761.4 (1.14018) 760.2 (0.447214) -0.16%
->>>> Context Switches 85380 (3419.52) 84748 (1904.44) -0.74%
->>>> Sleeps 105064 (1240.96) 105074 (337.612) 0.01%
->>>>
->>>> Average Optimal load -j 16
->>>> Run (std deviation)
->>>> BASE SPF
->>>> Elapsed Time 920.528 (10.1212) 927.404 (8.91789) 0.75%
->>>> User Time 11064.8 (981.142) 11085 (990.897) 0.18%
->>>> System Time 979.904 (84.0615) 1001.14 (82.5523) 2.17%
->>>> Percent CPU 1089.5 (345.894) 1086.1 (343.545) -0.31%
->>>> Context Switches 159488 (78156.4) 158223 (77472.1) -0.79%
->>>> Sleeps 110566 (5877.49) 110388 (5617.75) -0.16%
->>>>
->>>>
->>>> During a run on the SPF, perf events were captured:
->>>> Performance counter stats for '../kernbench -M':
->>>> 526743764 faults
->>>> 210 spf
->>>> 3 pagefault:spf_vma_changed
->>>> 0 pagefault:spf_vma_noanon
->>>> 2278 pagefault:spf_vma_notsup
->>>> 0 pagefault:spf_vma_access
->>>> 0 pagefault:spf_pmd_changed
->>>>
->>>> Very few speculative page faults were recorded as most of the processes
->>>> involved are monothreaded (sounds that on this architecture some threads
->>>> were created during the kernel build processing).
->>>>
->>>> Here are the kerbench results on a 80 CPUs Power8 system:
->>>>
->>>> Average Half load -j 40
->>>> Run (std deviation)
->>>> BASE SPF
->>>> Elapsed Time 117.152 (0.774642) 117.166 (0.476057) 0.01%
->>>> User Time 4478.52 (24.7688) 4479.76 (9.08555) 0.03%
->>>> System Time 131.104 (0.720056) 134.04 (0.708414) 2.24%
->>>> Percent CPU 3934 (19.7104) 3937.2 (19.0184) 0.08%
->>>> Context Switches 92125.4 (576.787) 92581.6 (198.622) 0.50%
->>>> Sleeps 317923 (652.499) 318469 (1255.59) 0.17%
->>>>
->>>> Average Optimal load -j 80
->>>> Run (std deviation)
->>>> BASE SPF
->>>> Elapsed Time 107.73 (0.632416) 107.31 (0.584936) -0.39%
->>>> User Time 5869.86 (1466.72) 5871.71 (1467.27) 0.03%
->>>> System Time 153.728 (23.8573) 157.153 (24.3704) 2.23%
->>>> Percent CPU 5418.6 (1565.17) 5436.7 (1580.91) 0.33%
->>>> Context Switches 223861 (138865) 225032 (139632) 0.52%
->>>> Sleeps 330529 (13495.1) 332001 (14746.2) 0.45%
->>>>
->>>> During a run on the SPF, perf events were captured:
->>>> Performance counter stats for '../kernbench -M':
->>>> 116730856 faults
->>>> 0 spf
->>>> 3 pagefault:spf_vma_changed
->>>> 0 pagefault:spf_vma_noanon
->>>> 476 pagefault:spf_vma_notsup
->>>> 0 pagefault:spf_vma_access
->>>> 0 pagefault:spf_pmd_changed
->>>>
->>>> Most of the processes involved are monothreaded so SPF is not activated but
->>>> there is no impact on the performance.
->>>>
->>>> Ebizzy:
->>>> -------
->>>> The test is counting the number of records per second it can manage, the
->>>> higher is the best. I run it like this 'ebizzy -mTt <nrcpus>'. To get
->>>> consistent result I repeated the test 100 times and measure the average
->>>> result. The number is the record processes per second, the higher is the
->>>> best.
->>>>
->>>> BASE SPF delta
->>>> 16 CPUs x86 VM 742.57 1490.24 100.69%
->>>> 80 CPUs P8 node 13105.4 24174.23 84.46%
->>>>
->>>> Here are the performance counter read during a run on a 16 CPUs x86 VM:
->>>> Performance counter stats for './ebizzy -mTt 16':
->>>> 1706379 faults
->>>> 1674599 spf
->>>> 30588 pagefault:spf_vma_changed
->>>> 0 pagefault:spf_vma_noanon
->>>> 363 pagefault:spf_vma_notsup
->>>> 0 pagefault:spf_vma_access
->>>> 0 pagefault:spf_pmd_changed
->>>>
->>>> And the ones captured during a run on a 80 CPUs Power node:
->>>> Performance counter stats for './ebizzy -mTt 80':
->>>> 1874773 faults
->>>> 1461153 spf
->>>> 413293 pagefault:spf_vma_changed
->>>> 0 pagefault:spf_vma_noanon
->>>> 200 pagefault:spf_vma_notsup
->>>> 0 pagefault:spf_vma_access
->>>> 0 pagefault:spf_pmd_changed
->>>>
->>>> In ebizzy's case most of the page fault were handled in a speculative way,
->>>> leading the ebizzy performance boost.
->>>>
->>>> ------------------
->>>> Changes since v10 (https://lkml.org/lkml/2018/4/17/572):
->>>> - Accounted for all review feedbacks from Punit Agrawal, Ganesh Mahendran
->>>> and Minchan Kim, hopefully.
->>>> - Remove unneeded check on CONFIG_SPECULATIVE_PAGE_FAULT in
->>>> __do_page_fault().
->>>> - Loop in pte_spinlock() and pte_map_lock() when pte try lock fails
->>>> instead
->>>> of aborting the speculative page fault handling. Dropping the now
->>>> useless
->>>> trace event pagefault:spf_pte_lock.
->>>> - No more try to reuse the fetched VMA during the speculative page fault
->>>> handling when retrying is needed. This adds a lot of complexity and
->>>> additional tests done didn't show a significant performance improvement.
->>>> - Convert IS_ENABLED(CONFIG_NUMA) back to #ifdef due to build error.
->>>>
->>>> [1] http://linux-kernel.2935.n7.nabble.com/RFC-PATCH-0-6-Another-go-at-speculative-page-faults-tt965642.html#none
->>>> [2] https://patchwork.kernel.org/patch/9999687/
->>>>
->>>>
->>>> Laurent Dufour (20):
->>>> mm: introduce CONFIG_SPECULATIVE_PAGE_FAULT
->>>> x86/mm: define ARCH_SUPPORTS_SPECULATIVE_PAGE_FAULT
->>>> powerpc/mm: set ARCH_SUPPORTS_SPECULATIVE_PAGE_FAULT
->>>> mm: introduce pte_spinlock for FAULT_FLAG_SPECULATIVE
->>>> mm: make pte_unmap_same compatible with SPF
->>>> mm: introduce INIT_VMA()
->>>> mm: protect VMA modifications using VMA sequence count
->>>> mm: protect mremap() against SPF hanlder
->>>> mm: protect SPF handler against anon_vma changes
->>>> mm: cache some VMA fields in the vm_fault structure
->>>> mm/migrate: Pass vm_fault pointer to migrate_misplaced_page()
->>>> mm: introduce __lru_cache_add_active_or_unevictable
->>>> mm: introduce __vm_normal_page()
->>>> mm: introduce __page_add_new_anon_rmap()
->>>> mm: protect mm_rb tree with a rwlock
->>>> mm: adding speculative page fault failure trace events
->>>> perf: add a speculative page fault sw event
->>>> perf tools: add support for the SPF perf event
->>>> mm: add speculative page fault vmstats
->>>> powerpc/mm: add speculative page fault
->>>>
->>>> Mahendran Ganesh (2):
->>>> arm64/mm: define ARCH_SUPPORTS_SPECULATIVE_PAGE_FAULT
->>>> arm64/mm: add speculative page fault
->>>>
->>>> Peter Zijlstra (4):
->>>> mm: prepare for FAULT_FLAG_SPECULATIVE
->>>> mm: VMA sequence count
->>>> mm: provide speculative fault infrastructure
->>>> x86/mm: add speculative pagefault handling
->>>>
->>>> arch/arm64/Kconfig | 1 +
->>>> arch/arm64/mm/fault.c | 12 +
->>>> arch/powerpc/Kconfig | 1 +
->>>> arch/powerpc/mm/fault.c | 16 +
->>>> arch/x86/Kconfig | 1 +
->>>> arch/x86/mm/fault.c | 27 +-
->>>> fs/exec.c | 2 +-
->>>> fs/proc/task_mmu.c | 5 +-
->>>> fs/userfaultfd.c | 17 +-
->>>> include/linux/hugetlb_inline.h | 2 +-
->>>> include/linux/migrate.h | 4 +-
->>>> include/linux/mm.h | 136 +++++++-
->>>> include/linux/mm_types.h | 7 +
->>>> include/linux/pagemap.h | 4 +-
->>>> include/linux/rmap.h | 12 +-
->>>> include/linux/swap.h | 10 +-
->>>> include/linux/vm_event_item.h | 3 +
->>>> include/trace/events/pagefault.h | 80 +++++
->>>> include/uapi/linux/perf_event.h | 1 +
->>>> kernel/fork.c | 5 +-
->>>> mm/Kconfig | 22 ++
->>>> mm/huge_memory.c | 6 +-
->>>> mm/hugetlb.c | 2 +
->>>> mm/init-mm.c | 3 +
->>>> mm/internal.h | 20 ++
->>>> mm/khugepaged.c | 5 +
->>>> mm/madvise.c | 6 +-
->>>> mm/memory.c | 612 +++++++++++++++++++++++++++++-----
->>>> mm/mempolicy.c | 51 ++-
->>>> mm/migrate.c | 6 +-
->>>> mm/mlock.c | 13 +-
->>>> mm/mmap.c | 229 ++++++++++---
->>>> mm/mprotect.c | 4 +-
->>>> mm/mremap.c | 13 +
->>>> mm/nommu.c | 2 +-
->>>> mm/rmap.c | 5 +-
->>>> mm/swap.c | 6 +-
->>>> mm/swap_state.c | 8 +-
->>>> mm/vmstat.c | 5 +-
->>>> tools/include/uapi/linux/perf_event.h | 1 +
->>>> tools/perf/util/evsel.c | 1 +
->>>> tools/perf/util/parse-events.c | 4 +
->>>> tools/perf/util/parse-events.l | 1 +
->>>> tools/perf/util/python.c | 1 +
->>>> 44 files changed, 1161 insertions(+), 211 deletions(-)
->>>> create mode 100644 include/trace/events/pagefault.h
->>>>
->>>> --
->>>> 2.7.4
->>>>
->>>>
->>>
->>
->
\ No newline at end of file
+Hi Laurent,=0A=
+=0A=
+=0A=
+For the test result on Intel 4s skylake platform (192 CPUs, 768G Memory), t=
+he below test cases all were run 3 times.=0A=
+I check the test results, only page_fault3_thread/enable THP have 6% stddev=
+ for head commit, other tests have lower stddev.=0A=
+=0A=
+And I did not find other high variation on test case result.=0A=
+=0A=
+a). Enable THP=0A=
+testcase base stddev change head =
+ stddev metric=0A=
+page_fault3/enable THP 10519 =B1 3% -20.5% 8368 =
+ =B16% will-it-scale.per_thread_ops=0A=
+page_fault2/enalbe THP 8281 =B1 2% -18.8% 6728 =
+ will-it-scale.per_thread_ops=0A=
+brk1/eanble THP 998475 -2.2% 976893 =
+ will-it-scale.per_process_ops=0A=
+context_switch1/enable THP 223910 -1.3% 220930 =
+ will-it-scale.per_process_ops=0A=
+context_switch1/enable THP 233722 -1.0% 231288 =
+ will-it-scale.per_thread_ops=0A=
+=0A=
+b). Disable THP=0A=
+page_fault3/disable THP 10856 -23.1% 8344 =
+ will-it-scale.per_thread_ops=0A=
+page_fault2/disable THP 8147 -18.8% 6613 =
+ will-it-scale.per_thread_ops=0A=
+brk1/disable THP 957 -7.9% 881 =
+ will-it-scale.per_thread_ops=0A=
+context_switch1/disable THP 237006 -2.2% 231907 =
+ will-it-scale.per_thread_ops=0A=
+brk1/disable THP 997317 -2.0% 977778 =
+ will-it-scale.per_process_ops=0A=
+page_fault3/disable THP 467454 -1.8% 459251 =
+ will-it-scale.per_process_ops=0A=
+context_switch1/disable THP 224431 -1.3% 221567 =
+ will-it-scale.per_process_ops=0A=
+=0A=
+=0A=
+Best regards,=0A=
+Haiyan Song=0A=
+________________________________________=0A=
+From: Laurent Dufour [ldufour@linux.vnet.ibm.com]=0A=
+Sent: Monday, July 02, 2018 4:59 PM=0A=
+To: Song, HaiyanX=0A=
+Cc: akpm@linux-foundation.org; mhocko@kernel.org; peterz@infradead.org; kir=
+ill@shutemov.name; ak@linux.intel.com; dave@stgolabs.net; jack@suse.cz; Mat=
+thew Wilcox; khandual@linux.vnet.ibm.com; aneesh.kumar@linux.vnet.ibm.com; =
+benh@kernel.crashing.org; mpe@ellerman.id.au; paulus@samba.org; Thomas Glei=
+xner; Ingo Molnar; hpa@zytor.com; Will Deacon; Sergey Senozhatsky; sergey.s=
+enozhatsky.work@gmail.com; Andrea Arcangeli; Alexei Starovoitov; Wang, Kemi=
+; Daniel Jordan; David Rientjes; Jerome Glisse; Ganesh Mahendran; Minchan K=
+im; Punit Agrawal; vinayak menon; Yang Shi; linux-kernel@vger.kernel.org; l=
+inux-mm@kvack.org; haren@linux.vnet.ibm.com; npiggin@gmail.com; bsingharora=
+@gmail.com; paulmck@linux.vnet.ibm.com; Tim Chen; linuxppc-dev@lists.ozlabs=
+.org; x86@kernel.org=0A=
+Subject: Re: [PATCH v11 00/26] Speculative page faults=0A=
+=0A=
+On 11/06/2018 09:49, Song, HaiyanX wrote:=0A=
+> Hi Laurent,=0A=
+>=0A=
+> Regression test for v11 patch serials have been run, some regression is f=
+ound by LKP-tools (linux kernel performance)=0A=
+> tested on Intel 4s skylake platform. This time only test the cases which =
+have been run and found regressions on=0A=
+> V9 patch serials.=0A=
+>=0A=
+> The regression result is sorted by the metric will-it-scale.per_thread_op=
+s.=0A=
+> branch: Laurent-Dufour/Speculative-page-faults/20180520-045126=0A=
+> commit id:=0A=
+> head commit : a7a8993bfe3ccb54ad468b9f1799649e4ad1ff12=0A=
+> base commit : ba98a1cdad71d259a194461b3a61471b49b14df1=0A=
+> Benchmark: will-it-scale=0A=
+> Download link: https://github.com/antonblanchard/will-it-scale/tree/maste=
+r=0A=
+>=0A=
+> Metrics:=0A=
+> will-it-scale.per_process_ops=3Dprocesses/nr_cpu=0A=
+> will-it-scale.per_thread_ops=3Dthreads/nr_cpu=0A=
+> test box: lkp-skl-4sp1(nr_cpu=3D192,memory=3D768G)=0A=
+> THP: enable / disable=0A=
+> nr_task:100%=0A=
+>=0A=
+> 1. Regressions:=0A=
+>=0A=
+> a). Enable THP=0A=
+> testcase base change head =
+ metric=0A=
+> page_fault3/enable THP 10519 -20.5% 836 wi=
+ll-it-scale.per_thread_ops=0A=
+> page_fault2/enalbe THP 8281 -18.8% 6728 wi=
+ll-it-scale.per_thread_ops=0A=
+> brk1/eanble THP 998475 -2.2% 976893 wi=
+ll-it-scale.per_process_ops=0A=
+> context_switch1/enable THP 223910 -1.3% 220930 wi=
+ll-it-scale.per_process_ops=0A=
+> context_switch1/enable THP 233722 -1.0% 231288 wi=
+ll-it-scale.per_thread_ops=0A=
+>=0A=
+> b). Disable THP=0A=
+> page_fault3/disable THP 10856 -23.1% 8344 wi=
+ll-it-scale.per_thread_ops=0A=
+> page_fault2/disable THP 8147 -18.8% 6613 wi=
+ll-it-scale.per_thread_ops=0A=
+> brk1/disable THP 957 -7.9% 881 wi=
+ll-it-scale.per_thread_ops=0A=
+> context_switch1/disable THP 237006 -2.2% 231907 wi=
+ll-it-scale.per_thread_ops=0A=
+> brk1/disable THP 997317 -2.0% 977778 wi=
+ll-it-scale.per_process_ops=0A=
+> page_fault3/disable THP 467454 -1.8% 459251 wi=
+ll-it-scale.per_process_ops=0A=
+> context_switch1/disable THP 224431 -1.3% 221567 wi=
+ll-it-scale.per_process_ops=0A=
+>=0A=
+> Notes: for the above values of test result, the higher is better.=0A=
+=0A=
+I tried the same tests on my PowerPC victim VM (1024 CPUs, 11TB) and I can'=
+t=0A=
+get reproducible results. The results have huge variation, even on the vani=
+lla=0A=
+kernel, and I can't state on any changes due to that.=0A=
+=0A=
+I tried on smaller node (80 CPUs, 32G), and the tests ran better, but I did=
+n't=0A=
+measure any changes between the vanilla and the SPF patched ones:=0A=
+=0A=
+test THP enabled 4.17.0-rc4-mm1 spf delta=0A=
+page_fault3_threads 2697.7 2683.5 -0.53%=0A=
+page_fault2_threads 170660.6 169574.1 -0.64%=0A=
+context_switch1_threads 6915269.2 6877507.3 -0.55%=0A=
+context_switch1_processes 6478076.2 6529493.5 0.79%=0A=
+brk1 243391.2 238527.5 -2.00%=0A=
+=0A=
+Tests were run 10 times, no high variation detected.=0A=
+=0A=
+Did you see high variation on your side ? How many times the test were run =
+to=0A=
+compute the average values ?=0A=
+=0A=
+Thanks,=0A=
+Laurent.=0A=
+=0A=
+=0A=
+>=0A=
+> 2. Improvement: not found improvement based on the selected test cases.=
+=0A=
+>=0A=
+>=0A=
+> Best regards=0A=
+> Haiyan Song=0A=
+> ________________________________________=0A=
+> From: owner-linux-mm@kvack.org [owner-linux-mm@kvack.org] on behalf of La=
+urent Dufour [ldufour@linux.vnet.ibm.com]=0A=
+> Sent: Monday, May 28, 2018 4:54 PM=0A=
+> To: Song, HaiyanX=0A=
+> Cc: akpm@linux-foundation.org; mhocko@kernel.org; peterz@infradead.org; k=
+irill@shutemov.name; ak@linux.intel.com; dave@stgolabs.net; jack@suse.cz; M=
+atthew Wilcox; khandual@linux.vnet.ibm.com; aneesh.kumar@linux.vnet.ibm.com=
+; benh@kernel.crashing.org; mpe@ellerman.id.au; paulus@samba.org; Thomas Gl=
+eixner; Ingo Molnar; hpa@zytor.com; Will Deacon; Sergey Senozhatsky; sergey=
+.senozhatsky.work@gmail.com; Andrea Arcangeli; Alexei Starovoitov; Wang, Ke=
+mi; Daniel Jordan; David Rientjes; Jerome Glisse; Ganesh Mahendran; Minchan=
+ Kim; Punit Agrawal; vinayak menon; Yang Shi; linux-kernel@vger.kernel.org;=
+ linux-mm@kvack.org; haren@linux.vnet.ibm.com; npiggin@gmail.com; bsingharo=
+ra@gmail.com; paulmck@linux.vnet.ibm.com; Tim Chen; linuxppc-dev@lists.ozla=
+bs.org; x86@kernel.org=0A=
+> Subject: Re: [PATCH v11 00/26] Speculative page faults=0A=
+>=0A=
+> On 28/05/2018 10:22, Haiyan Song wrote:=0A=
+>> Hi Laurent,=0A=
+>>=0A=
+>> Yes, these tests are done on V9 patch.=0A=
+>=0A=
+> Do you plan to give this V11 a run ?=0A=
+>=0A=
+>>=0A=
+>>=0A=
+>> Best regards,=0A=
+>> Haiyan Song=0A=
+>>=0A=
+>> On Mon, May 28, 2018 at 09:51:34AM +0200, Laurent Dufour wrote:=0A=
+>>> On 28/05/2018 07:23, Song, HaiyanX wrote:=0A=
+>>>>=0A=
+>>>> Some regression and improvements is found by LKP-tools(linux kernel pe=
+rformance) on V9 patch series=0A=
+>>>> tested on Intel 4s Skylake platform.=0A=
+>>>=0A=
+>>> Hi,=0A=
+>>>=0A=
+>>> Thanks for reporting this benchmark results, but you mentioned the "V9 =
+patch=0A=
+>>> series" while responding to the v11 header series...=0A=
+>>> Were these tests done on v9 or v11 ?=0A=
+>>>=0A=
+>>> Cheers,=0A=
+>>> Laurent.=0A=
+>>>=0A=
+>>>>=0A=
+>>>> The regression result is sorted by the metric will-it-scale.per_thread=
+_ops.=0A=
+>>>> Branch: Laurent-Dufour/Speculative-page-faults/20180316-151833 (V9 pat=
+ch series)=0A=
+>>>> Commit id:=0A=
+>>>> base commit: d55f34411b1b126429a823d06c3124c16283231f=0A=
+>>>> head commit: 0355322b3577eeab7669066df42c550a56801110=0A=
+>>>> Benchmark suite: will-it-scale=0A=
+>>>> Download link:=0A=
+>>>> https://github.com/antonblanchard/will-it-scale/tree/master/tests=0A=
+>>>> Metrics:=0A=
+>>>> will-it-scale.per_process_ops=3Dprocesses/nr_cpu=0A=
+>>>> will-it-scale.per_thread_ops=3Dthreads/nr_cpu=0A=
+>>>> test box: lkp-skl-4sp1(nr_cpu=3D192,memory=3D768G)=0A=
+>>>> THP: enable / disable=0A=
+>>>> nr_task: 100%=0A=
+>>>>=0A=
+>>>> 1. Regressions:=0A=
+>>>> a) THP enabled:=0A=
+>>>> testcase base change head =
+ metric=0A=
+>>>> page_fault3/ enable THP 10092 -17.5% 8323 =
+ will-it-scale.per_thread_ops=0A=
+>>>> page_fault2/ enable THP 8300 -17.2% 6869 =
+ will-it-scale.per_thread_ops=0A=
+>>>> brk1/ enable THP 957.67 -7.6% 885 =
+ will-it-scale.per_thread_ops=0A=
+>>>> page_fault3/ enable THP 172821 -5.3% 163692 =
+ will-it-scale.per_process_ops=0A=
+>>>> signal1/ enable THP 9125 -3.2% 8834 =
+ will-it-scale.per_process_ops=0A=
+>>>>=0A=
+>>>> b) THP disabled:=0A=
+>>>> testcase base change head =
+ metric=0A=
+>>>> page_fault3/ disable THP 10107 -19.1% 8180 =
+ will-it-scale.per_thread_ops=0A=
+>>>> page_fault2/ disable THP 8432 -17.8% 6931 =
+ will-it-scale.per_thread_ops=0A=
+>>>> context_switch1/ disable THP 215389 -6.8% 200776 =
+ will-it-scale.per_thread_ops=0A=
+>>>> brk1/ disable THP 939.67 -6.6% 877.3=
+3 will-it-scale.per_thread_ops=0A=
+>>>> page_fault3/ disable THP 173145 -4.7% 165064 =
+ will-it-scale.per_process_ops=0A=
+>>>> signal1/ disable THP 9162 -3.9% 8802 =
+ will-it-scale.per_process_ops=0A=
+>>>>=0A=
+>>>> 2. Improvements:=0A=
+>>>> a) THP enabled:=0A=
+>>>> testcase base change head =
+ metric=0A=
+>>>> malloc1/ enable THP 66.33 +469.8% 383.6=
+7 will-it-scale.per_thread_ops=0A=
+>>>> writeseek3/ enable THP 2531 +4.5% 2646 =
+ will-it-scale.per_thread_ops=0A=
+>>>> signal1/ enable THP 989.33 +2.8% 1016 =
+ will-it-scale.per_thread_ops=0A=
+>>>>=0A=
+>>>> b) THP disabled:=0A=
+>>>> testcase base change head =
+ metric=0A=
+>>>> malloc1/ disable THP 90.33 +417.3% 467.3=
+3 will-it-scale.per_thread_ops=0A=
+>>>> read2/ disable THP 58934 +39.2% 82060 =
+ will-it-scale.per_thread_ops=0A=
+>>>> page_fault1/ disable THP 8607 +36.4% 11736 =
+ will-it-scale.per_thread_ops=0A=
+>>>> read1/ disable THP 314063 +12.7% 353934 =
+ will-it-scale.per_thread_ops=0A=
+>>>> writeseek3/ disable THP 2452 +12.5% 2759 =
+ will-it-scale.per_thread_ops=0A=
+>>>> signal1/ disable THP 971.33 +5.5% 1024 =
+ will-it-scale.per_thread_ops=0A=
+>>>>=0A=
+>>>> Notes: for above values in column "change", the higher value means tha=
+t the related testcase result=0A=
+>>>> on head commit is better than that on base commit for this benchmark.=
+=0A=
+>>>>=0A=
+>>>>=0A=
+>>>> Best regards=0A=
+>>>> Haiyan Song=0A=
+>>>>=0A=
+>>>> ________________________________________=0A=
+>>>> From: owner-linux-mm@kvack.org [owner-linux-mm@kvack.org] on behalf of=
+ Laurent Dufour [ldufour@linux.vnet.ibm.com]=0A=
+>>>> Sent: Thursday, May 17, 2018 7:06 PM=0A=
+>>>> To: akpm@linux-foundation.org; mhocko@kernel.org; peterz@infradead.org=
+; kirill@shutemov.name; ak@linux.intel.com; dave@stgolabs.net; jack@suse.cz=
+; Matthew Wilcox; khandual@linux.vnet.ibm.com; aneesh.kumar@linux.vnet.ibm.=
+com; benh@kernel.crashing.org; mpe@ellerman.id.au; paulus@samba.org; Thomas=
+ Gleixner; Ingo Molnar; hpa@zytor.com; Will Deacon; Sergey Senozhatsky; ser=
+gey.senozhatsky.work@gmail.com; Andrea Arcangeli; Alexei Starovoitov; Wang,=
+ Kemi; Daniel Jordan; David Rientjes; Jerome Glisse; Ganesh Mahendran; Minc=
+han Kim; Punit Agrawal; vinayak menon; Yang Shi=0A=
+>>>> Cc: linux-kernel@vger.kernel.org; linux-mm@kvack.org; haren@linux.vnet=
+.ibm.com; npiggin@gmail.com; bsingharora@gmail.com; paulmck@linux.vnet.ibm.=
+com; Tim Chen; linuxppc-dev@lists.ozlabs.org; x86@kernel.org=0A=
+>>>> Subject: [PATCH v11 00/26] Speculative page faults=0A=
+>>>>=0A=
+>>>> This is a port on kernel 4.17 of the work done by Peter Zijlstra to ha=
+ndle=0A=
+>>>> page fault without holding the mm semaphore [1].=0A=
+>>>>=0A=
+>>>> The idea is to try to handle user space page faults without holding th=
+e=0A=
+>>>> mmap_sem. This should allow better concurrency for massively threaded=
+=0A=
+>>>> process since the page fault handler will not wait for other threads m=
+emory=0A=
+>>>> layout change to be done, assuming that this change is done in another=
+ part=0A=
+>>>> of the process's memory space. This type page fault is named speculati=
+ve=0A=
+>>>> page fault. If the speculative page fault fails because of a concurren=
+cy is=0A=
+>>>> detected or because underlying PMD or PTE tables are not yet allocatin=
+g, it=0A=
+>>>> is failing its processing and a classic page fault is then tried.=0A=
+>>>>=0A=
+>>>> The speculative page fault (SPF) has to look for the VMA matching the =
+fault=0A=
+>>>> address without holding the mmap_sem, this is done by introducing a rw=
+lock=0A=
+>>>> which protects the access to the mm_rb tree. Previously this was done =
+using=0A=
+>>>> SRCU but it was introducing a lot of scheduling to process the VMA's=
+=0A=
+>>>> freeing operation which was hitting the performance by 20% as reported=
+ by=0A=
+>>>> Kemi Wang [2]. Using a rwlock to protect access to the mm_rb tree is=
+=0A=
+>>>> limiting the locking contention to these operations which are expected=
+ to=0A=
+>>>> be in a O(log n) order. In addition to ensure that the VMA is not free=
+d in=0A=
+>>>> our back a reference count is added and 2 services (get_vma() and=0A=
+>>>> put_vma()) are introduced to handle the reference count. Once a VMA is=
+=0A=
+>>>> fetched from the RB tree using get_vma(), it must be later freed using=
+=0A=
+>>>> put_vma(). I can't see anymore the overhead I got while will-it-scale=
+=0A=
+>>>> benchmark anymore.=0A=
+>>>>=0A=
+>>>> The VMA's attributes checked during the speculative page fault process=
+ing=0A=
+>>>> have to be protected against parallel changes. This is done by using a=
+ per=0A=
+>>>> VMA sequence lock. This sequence lock allows the speculative page faul=
+t=0A=
+>>>> handler to fast check for parallel changes in progress and to abort th=
+e=0A=
+>>>> speculative page fault in that case.=0A=
+>>>>=0A=
+>>>> Once the VMA has been found, the speculative page fault handler would =
+check=0A=
+>>>> for the VMA's attributes to verify that the page fault has to be handl=
+ed=0A=
+>>>> correctly or not. Thus, the VMA is protected through a sequence lock w=
+hich=0A=
+>>>> allows fast detection of concurrent VMA changes. If such a change is=
+=0A=
+>>>> detected, the speculative page fault is aborted and a *classic* page f=
+ault=0A=
+>>>> is tried. VMA sequence lockings are added when VMA attributes which a=
+re=0A=
+>>>> checked during the page fault are modified.=0A=
+>>>>=0A=
+>>>> When the PTE is fetched, the VMA is checked to see if it has been chan=
+ged,=0A=
+>>>> so once the page table is locked, the VMA is valid, so any other chang=
+es=0A=
+>>>> leading to touching this PTE will need to lock the page table, so no=
+=0A=
+>>>> parallel change is possible at this time.=0A=
+>>>>=0A=
+>>>> The locking of the PTE is done with interrupts disabled, this allows=
+=0A=
+>>>> checking for the PMD to ensure that there is not an ongoing collapsing=
+=0A=
+>>>> operation. Since khugepaged is firstly set the PMD to pmd_none and the=
+n is=0A=
+>>>> waiting for the other CPU to have caught the IPI interrupt, if the pmd=
+ is=0A=
+>>>> valid at the time the PTE is locked, we have the guarantee that the=0A=
+>>>> collapsing operation will have to wait on the PTE lock to move forward=
+.=0A=
+>>>> This allows the SPF handler to map the PTE safely. If the PMD value is=
+=0A=
+>>>> different from the one recorded at the beginning of the SPF operation,=
+ the=0A=
+>>>> classic page fault handler will be called to handle the operation whil=
+e=0A=
+>>>> holding the mmap_sem. As the PTE lock is done with the interrupts disa=
+bled,=0A=
+>>>> the lock is done using spin_trylock() to avoid dead lock when handling=
+ a=0A=
+>>>> page fault while a TLB invalidate is requested by another CPU holding =
+the=0A=
+>>>> PTE.=0A=
+>>>>=0A=
+>>>> In pseudo code, this could be seen as:=0A=
+>>>> speculative_page_fault()=0A=
+>>>> {=0A=
+>>>> vma =3D get_vma()=0A=
+>>>> check vma sequence count=0A=
+>>>> check vma's support=0A=
+>>>> disable interrupt=0A=
+>>>> check pgd,p4d,...,pte=0A=
+>>>> save pmd and pte in vmf=0A=
+>>>> save vma sequence counter in vmf=0A=
+>>>> enable interrupt=0A=
+>>>> check vma sequence count=0A=
+>>>> handle_pte_fault(vma)=0A=
+>>>> ..=0A=
+>>>> page =3D alloc_page()=0A=
+>>>> pte_map_lock()=0A=
+>>>> disable interrupt=0A=
+>>>> abort if sequence counter has chan=
+ged=0A=
+>>>> abort if pmd or pte has changed=0A=
+>>>> pte map and lock=0A=
+>>>> enable interrupt=0A=
+>>>> if abort=0A=
+>>>> free page=0A=
+>>>> abort=0A=
+>>>> ...=0A=
+>>>> }=0A=
+>>>>=0A=
+>>>> arch_fault_handler()=0A=
+>>>> {=0A=
+>>>> if (speculative_page_fault(&vma))=0A=
+>>>> goto done=0A=
+>>>> again:=0A=
+>>>> lock(mmap_sem)=0A=
+>>>> vma =3D find_vma();=0A=
+>>>> handle_pte_fault(vma);=0A=
+>>>> if retry=0A=
+>>>> unlock(mmap_sem)=0A=
+>>>> goto again;=0A=
+>>>> done:=0A=
+>>>> handle fault error=0A=
+>>>> }=0A=
+>>>>=0A=
+>>>> Support for THP is not done because when checking for the PMD, we can =
+be=0A=
+>>>> confused by an in progress collapsing operation done by khugepaged. Th=
+e=0A=
+>>>> issue is that pmd_none() could be true either if the PMD is not alread=
+y=0A=
+>>>> populated or if the underlying PTE are in the way to be collapsed. So =
+we=0A=
+>>>> cannot safely allocate a PMD if pmd_none() is true.=0A=
+>>>>=0A=
+>>>> This series add a new software performance event named 'speculative-fa=
+ults'=0A=
+>>>> or 'spf'. It counts the number of successful page fault event handled=
+=0A=
+>>>> speculatively. When recording 'faults,spf' events, the faults one is=
+=0A=
+>>>> counting the total number of page fault events while 'spf' is only cou=
+nting=0A=
+>>>> the part of the faults processed speculatively.=0A=
+>>>>=0A=
+>>>> There are some trace events introduced by this series. They allow=0A=
+>>>> identifying why the page faults were not processed speculatively. This=
+=0A=
+>>>> doesn't take in account the faults generated by a monothreaded process=
+=0A=
+>>>> which directly processed while holding the mmap_sem. This trace events=
+ are=0A=
+>>>> grouped in a system named 'pagefault', they are:=0A=
+>>>> - pagefault:spf_vma_changed : if the VMA has been changed in our back=
+=0A=
+>>>> - pagefault:spf_vma_noanon : the vma->anon_vma field was not yet set.=
+=0A=
+>>>> - pagefault:spf_vma_notsup : the VMA's type is not supported=0A=
+>>>> - pagefault:spf_vma_access : the VMA's access right are not respected=
+=0A=
+>>>> - pagefault:spf_pmd_changed : the upper PMD pointer has changed in ou=
+r=0A=
+>>>> back.=0A=
+>>>>=0A=
+>>>> To record all the related events, the easier is to run perf with the=
+=0A=
+>>>> following arguments :=0A=
+>>>> $ perf stat -e 'faults,spf,pagefault:*' <command>=0A=
+>>>>=0A=
+>>>> There is also a dedicated vmstat counter showing the number of success=
+ful=0A=
+>>>> page fault handled speculatively. I can be seen this way:=0A=
+>>>> $ grep speculative_pgfault /proc/vmstat=0A=
+>>>>=0A=
+>>>> This series builds on top of v4.16-mmotm-2018-04-13-17-28 and is funct=
+ional=0A=
+>>>> on x86, PowerPC and arm64.=0A=
+>>>>=0A=
+>>>> ---------------------=0A=
+>>>> Real Workload results=0A=
+>>>>=0A=
+>>>> As mentioned in previous email, we did non official runs using a "popu=
+lar=0A=
+>>>> in memory multithreaded database product" on 176 cores SMT8 Power syst=
+em=0A=
+>>>> which showed a 30% improvements in the number of transaction processed=
+ per=0A=
+>>>> second. This run has been done on the v6 series, but changes introduce=
+d in=0A=
+>>>> this new version should not impact the performance boost seen.=0A=
+>>>>=0A=
+>>>> Here are the perf data captured during 2 of these runs on top of the v=
+8=0A=
+>>>> series:=0A=
+>>>> vanilla spf=0A=
+>>>> faults 89.418 101.364 +13%=0A=
+>>>> spf n/a 97.989=0A=
+>>>>=0A=
+>>>> With the SPF kernel, most of the page fault were processed in a specul=
+ative=0A=
+>>>> way.=0A=
+>>>>=0A=
+>>>> Ganesh Mahendran had backported the series on top of a 4.9 kernel and =
+gave=0A=
+>>>> it a try on an android device. He reported that the application launch=
+ time=0A=
+>>>> was improved in average by 6%, and for large applications (~100 thread=
+s) by=0A=
+>>>> 20%.=0A=
+>>>>=0A=
+>>>> Here are the launch time Ganesh mesured on Android 8.0 on top of a Qco=
+m=0A=
+>>>> MSM845 (8 cores) with 6GB (the less is better):=0A=
+>>>>=0A=
+>>>> Application 4.9 4.9+spf delta=0A=
+>>>> com.tencent.mm 416 389 -7%=0A=
+>>>> com.eg.android.AlipayGphone 1135 986 -13%=0A=
+>>>> com.tencent.mtt 455 454 0%=0A=
+>>>> com.qqgame.hlddz 1497 1409 -6%=0A=
+>>>> com.autonavi.minimap 711 701 -1%=0A=
+>>>> com.tencent.tmgp.sgame 788 748 -5%=0A=
+>>>> com.immomo.momo 501 487 -3%=0A=
+>>>> com.tencent.peng 2145 2112 -2%=0A=
+>>>> com.smile.gifmaker 491 461 -6%=0A=
+>>>> com.baidu.BaiduMap 479 366 -23%=0A=
+>>>> com.taobao.taobao 1341 1198 -11%=0A=
+>>>> com.baidu.searchbox 333 314 -6%=0A=
+>>>> com.tencent.mobileqq 394 384 -3%=0A=
+>>>> com.sina.weibo 907 906 0%=0A=
+>>>> com.youku.phone 816 731 -11%=0A=
+>>>> com.happyelements.AndroidAnimal.qq 763 717 -6%=0A=
+>>>> com.UCMobile 415 411 -1%=0A=
+>>>> com.tencent.tmgp.ak 1464 1431 -2%=0A=
+>>>> com.tencent.qqmusic 336 329 -2%=0A=
+>>>> com.sankuai.meituan 1661 1302 -22%=0A=
+>>>> com.netease.cloudmusic 1193 1200 1%=0A=
+>>>> air.tv.douyu.android 4257 4152 -2%=0A=
+>>>>=0A=
+>>>> ------------------=0A=
+>>>> Benchmarks results=0A=
+>>>>=0A=
+>>>> Base kernel is v4.17.0-rc4-mm1=0A=
+>>>> SPF is BASE + this series=0A=
+>>>>=0A=
+>>>> Kernbench:=0A=
+>>>> ----------=0A=
+>>>> Here are the results on a 16 CPUs X86 guest using kernbench on a 4.15=
+=0A=
+>>>> kernel (kernel is build 5 times):=0A=
+>>>>=0A=
+>>>> Average Half load -j 8=0A=
+>>>> Run (std deviation)=0A=
+>>>> BASE SPF=0A=
+>>>> Elapsed Time 1448.65 (5.72312) 1455.84 (4.84951) 0.50%=
+=0A=
+>>>> User Time 10135.4 (30.3699) 10148.8 (31.1252) 0.13%=
+=0A=
+>>>> System Time 900.47 (2.81131) 923.28 (7.52779) 2.53%=
+=0A=
+>>>> Percent CPU 761.4 (1.14018) 760.2 (0.447214) -0.16%=
+=0A=
+>>>> Context Switches 85380 (3419.52) 84748 (1904.44) -0.74%=
+=0A=
+>>>> Sleeps 105064 (1240.96) 105074 (337.612) 0.01%=
+=0A=
+>>>>=0A=
+>>>> Average Optimal load -j 16=0A=
+>>>> Run (std deviation)=0A=
+>>>> BASE SPF=0A=
+>>>> Elapsed Time 920.528 (10.1212) 927.404 (8.91789) 0.75%=
+=0A=
+>>>> User Time 11064.8 (981.142) 11085 (990.897) 0.18%=
+=0A=
+>>>> System Time 979.904 (84.0615) 1001.14 (82.5523) 2.17%=
+=0A=
+>>>> Percent CPU 1089.5 (345.894) 1086.1 (343.545) -0.31%=
+=0A=
+>>>> Context Switches 159488 (78156.4) 158223 (77472.1) -0.79%=
+=0A=
+>>>> Sleeps 110566 (5877.49) 110388 (5617.75) -0.16%=
+=0A=
+>>>>=0A=
+>>>>=0A=
+>>>> During a run on the SPF, perf events were captured:=0A=
+>>>> Performance counter stats for '../kernbench -M':=0A=
+>>>> 526743764 faults=0A=
+>>>> 210 spf=0A=
+>>>> 3 pagefault:spf_vma_changed=0A=
+>>>> 0 pagefault:spf_vma_noanon=0A=
+>>>> 2278 pagefault:spf_vma_notsup=0A=
+>>>> 0 pagefault:spf_vma_access=0A=
+>>>> 0 pagefault:spf_pmd_changed=0A=
+>>>>=0A=
+>>>> Very few speculative page faults were recorded as most of the processe=
+s=0A=
+>>>> involved are monothreaded (sounds that on this architecture some threa=
+ds=0A=
+>>>> were created during the kernel build processing).=0A=
+>>>>=0A=
+>>>> Here are the kerbench results on a 80 CPUs Power8 system:=0A=
+>>>>=0A=
+>>>> Average Half load -j 40=0A=
+>>>> Run (std deviation)=0A=
+>>>> BASE SPF=0A=
+>>>> Elapsed Time 117.152 (0.774642) 117.166 (0.476057) 0.01%=
+=0A=
+>>>> User Time 4478.52 (24.7688) 4479.76 (9.08555) 0.03%=
+=0A=
+>>>> System Time 131.104 (0.720056) 134.04 (0.708414) 2.24%=
+=0A=
+>>>> Percent CPU 3934 (19.7104) 3937.2 (19.0184) 0.08%=
+=0A=
+>>>> Context Switches 92125.4 (576.787) 92581.6 (198.622) 0.50%=
+=0A=
+>>>> Sleeps 317923 (652.499) 318469 (1255.59) 0.17%=
+=0A=
+>>>>=0A=
+>>>> Average Optimal load -j 80=0A=
+>>>> Run (std deviation)=0A=
+>>>> BASE SPF=0A=
+>>>> Elapsed Time 107.73 (0.632416) 107.31 (0.584936) -0.39%=
+=0A=
+>>>> User Time 5869.86 (1466.72) 5871.71 (1467.27) 0.03%=
+=0A=
+>>>> System Time 153.728 (23.8573) 157.153 (24.3704) 2.23%=
+=0A=
+>>>> Percent CPU 5418.6 (1565.17) 5436.7 (1580.91) 0.33%=
+=0A=
+>>>> Context Switches 223861 (138865) 225032 (139632) 0.52%=
+=0A=
+>>>> Sleeps 330529 (13495.1) 332001 (14746.2) 0.45%=
+=0A=
+>>>>=0A=
+>>>> During a run on the SPF, perf events were captured:=0A=
+>>>> Performance counter stats for '../kernbench -M':=0A=
+>>>> 116730856 faults=0A=
+>>>> 0 spf=0A=
+>>>> 3 pagefault:spf_vma_changed=0A=
+>>>> 0 pagefault:spf_vma_noanon=0A=
+>>>> 476 pagefault:spf_vma_notsup=0A=
+>>>> 0 pagefault:spf_vma_access=0A=
+>>>> 0 pagefault:spf_pmd_changed=0A=
+>>>>=0A=
+>>>> Most of the processes involved are monothreaded so SPF is not activate=
+d but=0A=
+>>>> there is no impact on the performance.=0A=
+>>>>=0A=
+>>>> Ebizzy:=0A=
+>>>> -------=0A=
+>>>> The test is counting the number of records per second it can manage, t=
+he=0A=
+>>>> higher is the best. I run it like this 'ebizzy -mTt <nrcpus>'. To get=
+=0A=
+>>>> consistent result I repeated the test 100 times and measure the averag=
+e=0A=
+>>>> result. The number is the record processes per second, the higher is t=
+he=0A=
+>>>> best.=0A=
+>>>>=0A=
+>>>> BASE SPF delta=0A=
+>>>> 16 CPUs x86 VM 742.57 1490.24 100.69%=0A=
+>>>> 80 CPUs P8 node 13105.4 24174.23 84.46%=0A=
+>>>>=0A=
+>>>> Here are the performance counter read during a run on a 16 CPUs x86 VM=
+:=0A=
+>>>> Performance counter stats for './ebizzy -mTt 16':=0A=
+>>>> 1706379 faults=0A=
+>>>> 1674599 spf=0A=
+>>>> 30588 pagefault:spf_vma_changed=0A=
+>>>> 0 pagefault:spf_vma_noanon=0A=
+>>>> 363 pagefault:spf_vma_notsup=0A=
+>>>> 0 pagefault:spf_vma_access=0A=
+>>>> 0 pagefault:spf_pmd_changed=0A=
+>>>>=0A=
+>>>> And the ones captured during a run on a 80 CPUs Power node:=0A=
+>>>> Performance counter stats for './ebizzy -mTt 80':=0A=
+>>>> 1874773 faults=0A=
+>>>> 1461153 spf=0A=
+>>>> 413293 pagefault:spf_vma_changed=0A=
+>>>> 0 pagefault:spf_vma_noanon=0A=
+>>>> 200 pagefault:spf_vma_notsup=0A=
+>>>> 0 pagefault:spf_vma_access=0A=
+>>>> 0 pagefault:spf_pmd_changed=0A=
+>>>>=0A=
+>>>> In ebizzy's case most of the page fault were handled in a speculative =
+way,=0A=
+>>>> leading the ebizzy performance boost.=0A=
+>>>>=0A=
+>>>> ------------------=0A=
+>>>> Changes since v10 (https://lkml.org/lkml/2018/4/17/572):=0A=
+>>>> - Accounted for all review feedbacks from Punit Agrawal, Ganesh Mahen=
+dran=0A=
+>>>> and Minchan Kim, hopefully.=0A=
+>>>> - Remove unneeded check on CONFIG_SPECULATIVE_PAGE_FAULT in=0A=
+>>>> __do_page_fault().=0A=
+>>>> - Loop in pte_spinlock() and pte_map_lock() when pte try lock fails=
+=0A=
+>>>> instead=0A=
+>>>> of aborting the speculative page fault handling. Dropping the now=
+=0A=
+>>>> useless=0A=
+>>>> trace event pagefault:spf_pte_lock.=0A=
+>>>> - No more try to reuse the fetched VMA during the speculative page fa=
+ult=0A=
+>>>> handling when retrying is needed. This adds a lot of complexity and=
+=0A=
+>>>> additional tests done didn't show a significant performance improve=
+ment.=0A=
+>>>> - Convert IS_ENABLED(CONFIG_NUMA) back to #ifdef due to build error.=
+=0A=
+>>>>=0A=
+>>>> [1] http://linux-kernel.2935.n7.nabble.com/RFC-PATCH-0-6-Another-go-at=
+-speculative-page-faults-tt965642.html#none=0A=
+>>>> [2] https://patchwork.kernel.org/patch/9999687/=0A=
+>>>>=0A=
+>>>>=0A=
+>>>> Laurent Dufour (20):=0A=
+>>>> mm: introduce CONFIG_SPECULATIVE_PAGE_FAULT=0A=
+>>>> x86/mm: define ARCH_SUPPORTS_SPECULATIVE_PAGE_FAULT=0A=
+>>>> powerpc/mm: set ARCH_SUPPORTS_SPECULATIVE_PAGE_FAULT=0A=
+>>>> mm: introduce pte_spinlock for FAULT_FLAG_SPECULATIVE=0A=
+>>>> mm: make pte_unmap_same compatible with SPF=0A=
+>>>> mm: introduce INIT_VMA()=0A=
+>>>> mm: protect VMA modifications using VMA sequence count=0A=
+>>>> mm: protect mremap() against SPF hanlder=0A=
+>>>> mm: protect SPF handler against anon_vma changes=0A=
+>>>> mm: cache some VMA fields in the vm_fault structure=0A=
+>>>> mm/migrate: Pass vm_fault pointer to migrate_misplaced_page()=0A=
+>>>> mm: introduce __lru_cache_add_active_or_unevictable=0A=
+>>>> mm: introduce __vm_normal_page()=0A=
+>>>> mm: introduce __page_add_new_anon_rmap()=0A=
+>>>> mm: protect mm_rb tree with a rwlock=0A=
+>>>> mm: adding speculative page fault failure trace events=0A=
+>>>> perf: add a speculative page fault sw event=0A=
+>>>> perf tools: add support for the SPF perf event=0A=
+>>>> mm: add speculative page fault vmstats=0A=
+>>>> powerpc/mm: add speculative page fault=0A=
+>>>>=0A=
+>>>> Mahendran Ganesh (2):=0A=
+>>>> arm64/mm: define ARCH_SUPPORTS_SPECULATIVE_PAGE_FAULT=0A=
+>>>> arm64/mm: add speculative page fault=0A=
+>>>>=0A=
+>>>> Peter Zijlstra (4):=0A=
+>>>> mm: prepare for FAULT_FLAG_SPECULATIVE=0A=
+>>>> mm: VMA sequence count=0A=
+>>>> mm: provide speculative fault infrastructure=0A=
+>>>> x86/mm: add speculative pagefault handling=0A=
+>>>>=0A=
+>>>> arch/arm64/Kconfig | 1 +=0A=
+>>>> arch/arm64/mm/fault.c | 12 +=0A=
+>>>> arch/powerpc/Kconfig | 1 +=0A=
+>>>> arch/powerpc/mm/fault.c | 16 +=0A=
+>>>> arch/x86/Kconfig | 1 +=0A=
+>>>> arch/x86/mm/fault.c | 27 +-=0A=
+>>>> fs/exec.c | 2 +-=0A=
+>>>> fs/proc/task_mmu.c | 5 +-=0A=
+>>>> fs/userfaultfd.c | 17 +-=0A=
+>>>> include/linux/hugetlb_inline.h | 2 +-=0A=
+>>>> include/linux/migrate.h | 4 +-=0A=
+>>>> include/linux/mm.h | 136 +++++++-=0A=
+>>>> include/linux/mm_types.h | 7 +=0A=
+>>>> include/linux/pagemap.h | 4 +-=0A=
+>>>> include/linux/rmap.h | 12 +-=0A=
+>>>> include/linux/swap.h | 10 +-=0A=
+>>>> include/linux/vm_event_item.h | 3 +=0A=
+>>>> include/trace/events/pagefault.h | 80 +++++=0A=
+>>>> include/uapi/linux/perf_event.h | 1 +=0A=
+>>>> kernel/fork.c | 5 +-=0A=
+>>>> mm/Kconfig | 22 ++=0A=
+>>>> mm/huge_memory.c | 6 +-=0A=
+>>>> mm/hugetlb.c | 2 +=0A=
+>>>> mm/init-mm.c | 3 +=0A=
+>>>> mm/internal.h | 20 ++=0A=
+>>>> mm/khugepaged.c | 5 +=0A=
+>>>> mm/madvise.c | 6 +-=0A=
+>>>> mm/memory.c | 612 +++++++++++++++++++++++++=
+++++-----=0A=
+>>>> mm/mempolicy.c | 51 ++-=0A=
+>>>> mm/migrate.c | 6 +-=0A=
+>>>> mm/mlock.c | 13 +-=0A=
+>>>> mm/mmap.c | 229 ++++++++++---=0A=
+>>>> mm/mprotect.c | 4 +-=0A=
+>>>> mm/mremap.c | 13 +=0A=
+>>>> mm/nommu.c | 2 +-=0A=
+>>>> mm/rmap.c | 5 +-=0A=
+>>>> mm/swap.c | 6 +-=0A=
+>>>> mm/swap_state.c | 8 +-=0A=
+>>>> mm/vmstat.c | 5 +-=0A=
+>>>> tools/include/uapi/linux/perf_event.h | 1 +=0A=
+>>>> tools/perf/util/evsel.c | 1 +=0A=
+>>>> tools/perf/util/parse-events.c | 4 +=0A=
+>>>> tools/perf/util/parse-events.l | 1 +=0A=
+>>>> tools/perf/util/python.c | 1 +=0A=
+>>>> 44 files changed, 1161 insertions(+), 211 deletions(-)=0A=
+>>>> create mode 100644 include/trace/events/pagefault.h=0A=
+>>>>=0A=
+>>>> --=0A=
+>>>> 2.7.4=0A=
+>>>>=0A=
+>>>>=0A=
+>>>=0A=
+>>=0A=
+>=0A=
+=0A=
\ No newline at end of file
diff --git a/a/content_digest b/N1/content_digest
index 433e1ff..419f628 100644
--- a/a/content_digest
+++ b/N1/content_digest
@@ -80,619 +80,833 @@
"b\0"
]
[
- "Hi Laurent,\n",
- "\n",
- "\n",
- "For the test result on Intel 4s skylake platform (192 CPUs, 768G Memory), the below test cases all were run 3 times.\n",
- "I check the test results, only page_fault3_thread/enable THP have 6% stddev for head commit, other tests have lower stddev.\n",
- "\n",
- "And I did not find other high variation on test case result.\n",
- "\n",
- "a). Enable THP\n",
- "testcase base stddev change head stddev metric\n",
- "page_fault3/enable THP 10519 \302\261 3% -20.5% 8368 \302\2616% will-it-scale.per_thread_ops\n",
- "page_fault2/enalbe THP 8281 \302\261 2% -18.8% 6728 will-it-scale.per_thread_ops\n",
- "brk1/eanble THP 998475 -2.2% 976893 will-it-scale.per_process_ops\n",
- "context_switch1/enable THP 223910 -1.3% 220930 will-it-scale.per_process_ops\n",
- "context_switch1/enable THP 233722 -1.0% 231288 will-it-scale.per_thread_ops\n",
- "\n",
- "b). Disable THP\n",
- "page_fault3/disable THP 10856 -23.1% 8344 will-it-scale.per_thread_ops\n",
- "page_fault2/disable THP 8147 -18.8% 6613 will-it-scale.per_thread_ops\n",
- "brk1/disable THP 957 -7.9% 881 will-it-scale.per_thread_ops\n",
- "context_switch1/disable THP 237006 -2.2% 231907 will-it-scale.per_thread_ops\n",
- "brk1/disable THP 997317 -2.0% 977778 will-it-scale.per_process_ops\n",
- "page_fault3/disable THP 467454 -1.8% 459251 will-it-scale.per_process_ops\n",
- "context_switch1/disable THP 224431 -1.3% 221567 will-it-scale.per_process_ops\n",
- "\n",
- "\n",
- "Best regards,\n",
- "Haiyan Song\n",
- "________________________________________\n",
- "From: Laurent Dufour [ldufour\@linux.vnet.ibm.com]\n",
- "Sent: Monday, July 02, 2018 4:59 PM\n",
- "To: Song, HaiyanX\n",
- "Cc: akpm\@linux-foundation.org; mhocko\@kernel.org; peterz\@infradead.org; kirill\@shutemov.name; ak\@linux.intel.com; dave\@stgolabs.net; jack\@suse.cz; Matthew Wilcox; khandual\@linux.vnet.ibm.com; aneesh.kumar\@linux.vnet.ibm.com; benh\@kernel.crashing.org; mpe\@ellerman.id.au; paulus\@samba.org; Thomas Gleixner; Ingo Molnar; hpa\@zytor.com; Will Deacon; Sergey Senozhatsky; sergey.senozhatsky.work\@gmail.com; Andrea Arcangeli; Alexei Starovoitov; Wang, Kemi; Daniel Jordan; David Rientjes; Jerome Glisse; Ganesh Mahendran; Minchan Kim; Punit Agrawal; vinayak menon; Yang Shi; linux-kernel\@vger.kernel.org; linux-mm\@kvack.org; haren\@linux.vnet.ibm.com; npiggin\@gmail.com; bsingharora\@gmail.com; paulmck\@linux.vnet.ibm.com; Tim Chen; linuxppc-dev\@lists.ozlabs.org; x86\@kernel.org\n",
- "Subject: Re: [PATCH v11 00/26] Speculative page faults\n",
- "\n",
- "On 11/06/2018 09:49, Song, HaiyanX wrote:\n",
- "> Hi Laurent,\n",
- ">\n",
- "> Regression test for v11 patch serials have been run, some regression is found by LKP-tools (linux kernel performance)\n",
- "> tested on Intel 4s skylake platform. This time only test the cases which have been run and found regressions on\n",
- "> V9 patch serials.\n",
- ">\n",
- "> The regression result is sorted by the metric will-it-scale.per_thread_ops.\n",
- "> branch: Laurent-Dufour/Speculative-page-faults/20180520-045126\n",
- "> commit id:\n",
- "> head commit : a7a8993bfe3ccb54ad468b9f1799649e4ad1ff12\n",
- "> base commit : ba98a1cdad71d259a194461b3a61471b49b14df1\n",
- "> Benchmark: will-it-scale\n",
- "> Download link: https://github.com/antonblanchard/will-it-scale/tree/master\n",
- ">\n",
- "> Metrics:\n",
- "> will-it-scale.per_process_ops=processes/nr_cpu\n",
- "> will-it-scale.per_thread_ops=threads/nr_cpu\n",
- "> test box: lkp-skl-4sp1(nr_cpu=192,memory=768G)\n",
- "> THP: enable / disable\n",
- "> nr_task:100%\n",
- ">\n",
- "> 1. Regressions:\n",
- ">\n",
- "> a). Enable THP\n",
- "> testcase base change head metric\n",
- "> page_fault3/enable THP 10519 -20.5% 836 will-it-scale.per_thread_ops\n",
- "> page_fault2/enalbe THP 8281 -18.8% 6728 will-it-scale.per_thread_ops\n",
- "> brk1/eanble THP 998475 -2.2% 976893 will-it-scale.per_process_ops\n",
- "> context_switch1/enable THP 223910 -1.3% 220930 will-it-scale.per_process_ops\n",
- "> context_switch1/enable THP 233722 -1.0% 231288 will-it-scale.per_thread_ops\n",
- ">\n",
- "> b). Disable THP\n",
- "> page_fault3/disable THP 10856 -23.1% 8344 will-it-scale.per_thread_ops\n",
- "> page_fault2/disable THP 8147 -18.8% 6613 will-it-scale.per_thread_ops\n",
- "> brk1/disable THP 957 -7.9% 881 will-it-scale.per_thread_ops\n",
- "> context_switch1/disable THP 237006 -2.2% 231907 will-it-scale.per_thread_ops\n",
- "> brk1/disable THP 997317 -2.0% 977778 will-it-scale.per_process_ops\n",
- "> page_fault3/disable THP 467454 -1.8% 459251 will-it-scale.per_process_ops\n",
- "> context_switch1/disable THP 224431 -1.3% 221567 will-it-scale.per_process_ops\n",
- ">\n",
- "> Notes: for the above values of test result, the higher is better.\n",
- "\n",
- "I tried the same tests on my PowerPC victim VM (1024 CPUs, 11TB) and I can't\n",
- "get reproducible results. The results have huge variation, even on the vanilla\n",
- "kernel, and I can't state on any changes due to that.\n",
- "\n",
- "I tried on smaller node (80 CPUs, 32G), and the tests ran better, but I didn't\n",
- "measure any changes between the vanilla and the SPF patched ones:\n",
- "\n",
- "test THP enabled 4.17.0-rc4-mm1 spf delta\n",
- "page_fault3_threads 2697.7 2683.5 -0.53%\n",
- "page_fault2_threads 170660.6 169574.1 -0.64%\n",
- "context_switch1_threads 6915269.2 6877507.3 -0.55%\n",
- "context_switch1_processes 6478076.2 6529493.5 0.79%\n",
- "brk1 243391.2 238527.5 -2.00%\n",
- "\n",
- "Tests were run 10 times, no high variation detected.\n",
- "\n",
- "Did you see high variation on your side ? How many times the test were run to\n",
- "compute the average values ?\n",
- "\n",
- "Thanks,\n",
- "Laurent.\n",
- "\n",
- "\n",
- ">\n",
- "> 2. Improvement: not found improvement based on the selected test cases.\n",
- ">\n",
- ">\n",
- "> Best regards\n",
- "> Haiyan Song\n",
- "> ________________________________________\n",
- "> From: owner-linux-mm\@kvack.org [owner-linux-mm\@kvack.org] on behalf of Laurent Dufour [ldufour\@linux.vnet.ibm.com]\n",
- "> Sent: Monday, May 28, 2018 4:54 PM\n",
- "> To: Song, HaiyanX\n",
- "> Cc: akpm\@linux-foundation.org; mhocko\@kernel.org; peterz\@infradead.org; kirill\@shutemov.name; ak\@linux.intel.com; dave\@stgolabs.net; jack\@suse.cz; Matthew Wilcox; khandual\@linux.vnet.ibm.com; aneesh.kumar\@linux.vnet.ibm.com; benh\@kernel.crashing.org; mpe\@ellerman.id.au; paulus\@samba.org; Thomas Gleixner; Ingo Molnar; hpa\@zytor.com; Will Deacon; Sergey Senozhatsky; sergey.senozhatsky.work\@gmail.com; Andrea Arcangeli; Alexei Starovoitov; Wang, Kemi; Daniel Jordan; David Rientjes; Jerome Glisse; Ganesh Mahendran; Minchan Kim; Punit Agrawal; vinayak menon; Yang Shi; linux-kernel\@vger.kernel.org; linux-mm\@kvack.org; haren\@linux.vnet.ibm.com; npiggin\@gmail.com; bsingharora\@gmail.com; paulmck\@linux.vnet.ibm.com; Tim Chen; linuxppc-dev\@lists.ozlabs.org; x86\@kernel.org\n",
- "> Subject: Re: [PATCH v11 00/26] Speculative page faults\n",
- ">\n",
- "> On 28/05/2018 10:22, Haiyan Song wrote:\n",
- ">> Hi Laurent,\n",
- ">>\n",
- ">> Yes, these tests are done on V9 patch.\n",
- ">\n",
- "> Do you plan to give this V11 a run ?\n",
- ">\n",
- ">>\n",
- ">>\n",
- ">> Best regards,\n",
- ">> Haiyan Song\n",
- ">>\n",
- ">> On Mon, May 28, 2018 at 09:51:34AM +0200, Laurent Dufour wrote:\n",
- ">>> On 28/05/2018 07:23, Song, HaiyanX wrote:\n",
- ">>>>\n",
- ">>>> Some regression and improvements is found by LKP-tools(linux kernel performance) on V9 patch series\n",
- ">>>> tested on Intel 4s Skylake platform.\n",
- ">>>\n",
- ">>> Hi,\n",
- ">>>\n",
- ">>> Thanks for reporting this benchmark results, but you mentioned the \"V9 patch\n",
- ">>> series\" while responding to the v11 header series...\n",
- ">>> Were these tests done on v9 or v11 ?\n",
- ">>>\n",
- ">>> Cheers,\n",
- ">>> Laurent.\n",
- ">>>\n",
- ">>>>\n",
- ">>>> The regression result is sorted by the metric will-it-scale.per_thread_ops.\n",
- ">>>> Branch: Laurent-Dufour/Speculative-page-faults/20180316-151833 (V9 patch series)\n",
- ">>>> Commit id:\n",
- ">>>> base commit: d55f34411b1b126429a823d06c3124c16283231f\n",
- ">>>> head commit: 0355322b3577eeab7669066df42c550a56801110\n",
- ">>>> Benchmark suite: will-it-scale\n",
- ">>>> Download link:\n",
- ">>>> https://github.com/antonblanchard/will-it-scale/tree/master/tests\n",
- ">>>> Metrics:\n",
- ">>>> will-it-scale.per_process_ops=processes/nr_cpu\n",
- ">>>> will-it-scale.per_thread_ops=threads/nr_cpu\n",
- ">>>> test box: lkp-skl-4sp1(nr_cpu=192,memory=768G)\n",
- ">>>> THP: enable / disable\n",
- ">>>> nr_task: 100%\n",
- ">>>>\n",
- ">>>> 1. Regressions:\n",
- ">>>> a) THP enabled:\n",
- ">>>> testcase base change head metric\n",
- ">>>> page_fault3/ enable THP 10092 -17.5% 8323 will-it-scale.per_thread_ops\n",
- ">>>> page_fault2/ enable THP 8300 -17.2% 6869 will-it-scale.per_thread_ops\n",
- ">>>> brk1/ enable THP 957.67 -7.6% 885 will-it-scale.per_thread_ops\n",
- ">>>> page_fault3/ enable THP 172821 -5.3% 163692 will-it-scale.per_process_ops\n",
- ">>>> signal1/ enable THP 9125 -3.2% 8834 will-it-scale.per_process_ops\n",
- ">>>>\n",
- ">>>> b) THP disabled:\n",
- ">>>> testcase base change head metric\n",
- ">>>> page_fault3/ disable THP 10107 -19.1% 8180 will-it-scale.per_thread_ops\n",
- ">>>> page_fault2/ disable THP 8432 -17.8% 6931 will-it-scale.per_thread_ops\n",
- ">>>> context_switch1/ disable THP 215389 -6.8% 200776 will-it-scale.per_thread_ops\n",
- ">>>> brk1/ disable THP 939.67 -6.6% 877.33 will-it-scale.per_thread_ops\n",
- ">>>> page_fault3/ disable THP 173145 -4.7% 165064 will-it-scale.per_process_ops\n",
- ">>>> signal1/ disable THP 9162 -3.9% 8802 will-it-scale.per_process_ops\n",
- ">>>>\n",
- ">>>> 2. Improvements:\n",
- ">>>> a) THP enabled:\n",
- ">>>> testcase base change head metric\n",
- ">>>> malloc1/ enable THP 66.33 +469.8% 383.67 will-it-scale.per_thread_ops\n",
- ">>>> writeseek3/ enable THP 2531 +4.5% 2646 will-it-scale.per_thread_ops\n",
- ">>>> signal1/ enable THP 989.33 +2.8% 1016 will-it-scale.per_thread_ops\n",
- ">>>>\n",
- ">>>> b) THP disabled:\n",
- ">>>> testcase base change head metric\n",
- ">>>> malloc1/ disable THP 90.33 +417.3% 467.33 will-it-scale.per_thread_ops\n",
- ">>>> read2/ disable THP 58934 +39.2% 82060 will-it-scale.per_thread_ops\n",
- ">>>> page_fault1/ disable THP 8607 +36.4% 11736 will-it-scale.per_thread_ops\n",
- ">>>> read1/ disable THP 314063 +12.7% 353934 will-it-scale.per_thread_ops\n",
- ">>>> writeseek3/ disable THP 2452 +12.5% 2759 will-it-scale.per_thread_ops\n",
- ">>>> signal1/ disable THP 971.33 +5.5% 1024 will-it-scale.per_thread_ops\n",
- ">>>>\n",
- ">>>> Notes: for above values in column \"change\", the higher value means that the related testcase result\n",
- ">>>> on head commit is better than that on base commit for this benchmark.\n",
- ">>>>\n",
- ">>>>\n",
- ">>>> Best regards\n",
- ">>>> Haiyan Song\n",
- ">>>>\n",
- ">>>> ________________________________________\n",
- ">>>> From: owner-linux-mm\@kvack.org [owner-linux-mm\@kvack.org] on behalf of Laurent Dufour [ldufour\@linux.vnet.ibm.com]\n",
- ">>>> Sent: Thursday, May 17, 2018 7:06 PM\n",
- ">>>> To: akpm\@linux-foundation.org; mhocko\@kernel.org; peterz\@infradead.org; kirill\@shutemov.name; ak\@linux.intel.com; dave\@stgolabs.net; jack\@suse.cz; Matthew Wilcox; khandual\@linux.vnet.ibm.com; aneesh.kumar\@linux.vnet.ibm.com; benh\@kernel.crashing.org; mpe\@ellerman.id.au; paulus\@samba.org; Thomas Gleixner; Ingo Molnar; hpa\@zytor.com; Will Deacon; Sergey Senozhatsky; sergey.senozhatsky.work\@gmail.com; Andrea Arcangeli; Alexei Starovoitov; Wang, Kemi; Daniel Jordan; David Rientjes; Jerome Glisse; Ganesh Mahendran; Minchan Kim; Punit Agrawal; vinayak menon; Yang Shi\n",
- ">>>> Cc: linux-kernel\@vger.kernel.org; linux-mm\@kvack.org; haren\@linux.vnet.ibm.com; npiggin\@gmail.com; bsingharora\@gmail.com; paulmck\@linux.vnet.ibm.com; Tim Chen; linuxppc-dev\@lists.ozlabs.org; x86\@kernel.org\n",
- ">>>> Subject: [PATCH v11 00/26] Speculative page faults\n",
- ">>>>\n",
- ">>>> This is a port on kernel 4.17 of the work done by Peter Zijlstra to handle\n",
- ">>>> page fault without holding the mm semaphore [1].\n",
- ">>>>\n",
- ">>>> The idea is to try to handle user space page faults without holding the\n",
- ">>>> mmap_sem. This should allow better concurrency for massively threaded\n",
- ">>>> process since the page fault handler will not wait for other threads memory\n",
- ">>>> layout change to be done, assuming that this change is done in another part\n",
- ">>>> of the process's memory space. This type page fault is named speculative\n",
- ">>>> page fault. If the speculative page fault fails because of a concurrency is\n",
- ">>>> detected or because underlying PMD or PTE tables are not yet allocating, it\n",
- ">>>> is failing its processing and a classic page fault is then tried.\n",
- ">>>>\n",
- ">>>> The speculative page fault (SPF) has to look for the VMA matching the fault\n",
- ">>>> address without holding the mmap_sem, this is done by introducing a rwlock\n",
- ">>>> which protects the access to the mm_rb tree. Previously this was done using\n",
- ">>>> SRCU but it was introducing a lot of scheduling to process the VMA's\n",
- ">>>> freeing operation which was hitting the performance by 20% as reported by\n",
- ">>>> Kemi Wang [2]. Using a rwlock to protect access to the mm_rb tree is\n",
- ">>>> limiting the locking contention to these operations which are expected to\n",
- ">>>> be in a O(log n) order. In addition to ensure that the VMA is not freed in\n",
- ">>>> our back a reference count is added and 2 services (get_vma() and\n",
- ">>>> put_vma()) are introduced to handle the reference count. Once a VMA is\n",
- ">>>> fetched from the RB tree using get_vma(), it must be later freed using\n",
- ">>>> put_vma(). I can't see anymore the overhead I got while will-it-scale\n",
- ">>>> benchmark anymore.\n",
- ">>>>\n",
- ">>>> The VMA's attributes checked during the speculative page fault processing\n",
- ">>>> have to be protected against parallel changes. This is done by using a per\n",
- ">>>> VMA sequence lock. This sequence lock allows the speculative page fault\n",
- ">>>> handler to fast check for parallel changes in progress and to abort the\n",
- ">>>> speculative page fault in that case.\n",
- ">>>>\n",
- ">>>> Once the VMA has been found, the speculative page fault handler would check\n",
- ">>>> for the VMA's attributes to verify that the page fault has to be handled\n",
- ">>>> correctly or not. Thus, the VMA is protected through a sequence lock which\n",
- ">>>> allows fast detection of concurrent VMA changes. If such a change is\n",
- ">>>> detected, the speculative page fault is aborted and a *classic* page fault\n",
- ">>>> is tried. VMA sequence lockings are added when VMA attributes which are\n",
- ">>>> checked during the page fault are modified.\n",
- ">>>>\n",
- ">>>> When the PTE is fetched, the VMA is checked to see if it has been changed,\n",
- ">>>> so once the page table is locked, the VMA is valid, so any other changes\n",
- ">>>> leading to touching this PTE will need to lock the page table, so no\n",
- ">>>> parallel change is possible at this time.\n",
- ">>>>\n",
- ">>>> The locking of the PTE is done with interrupts disabled, this allows\n",
- ">>>> checking for the PMD to ensure that there is not an ongoing collapsing\n",
- ">>>> operation. Since khugepaged is firstly set the PMD to pmd_none and then is\n",
- ">>>> waiting for the other CPU to have caught the IPI interrupt, if the pmd is\n",
- ">>>> valid at the time the PTE is locked, we have the guarantee that the\n",
- ">>>> collapsing operation will have to wait on the PTE lock to move forward.\n",
- ">>>> This allows the SPF handler to map the PTE safely. If the PMD value is\n",
- ">>>> different from the one recorded at the beginning of the SPF operation, the\n",
- ">>>> classic page fault handler will be called to handle the operation while\n",
- ">>>> holding the mmap_sem. As the PTE lock is done with the interrupts disabled,\n",
- ">>>> the lock is done using spin_trylock() to avoid dead lock when handling a\n",
- ">>>> page fault while a TLB invalidate is requested by another CPU holding the\n",
- ">>>> PTE.\n",
- ">>>>\n",
- ">>>> In pseudo code, this could be seen as:\n",
- ">>>> speculative_page_fault()\n",
- ">>>> {\n",
- ">>>> vma = get_vma()\n",
- ">>>> check vma sequence count\n",
- ">>>> check vma's support\n",
- ">>>> disable interrupt\n",
- ">>>> check pgd,p4d,...,pte\n",
- ">>>> save pmd and pte in vmf\n",
- ">>>> save vma sequence counter in vmf\n",
- ">>>> enable interrupt\n",
- ">>>> check vma sequence count\n",
- ">>>> handle_pte_fault(vma)\n",
- ">>>> ..\n",
- ">>>> page = alloc_page()\n",
- ">>>> pte_map_lock()\n",
- ">>>> disable interrupt\n",
- ">>>> abort if sequence counter has changed\n",
- ">>>> abort if pmd or pte has changed\n",
- ">>>> pte map and lock\n",
- ">>>> enable interrupt\n",
- ">>>> if abort\n",
- ">>>> free page\n",
- ">>>> abort\n",
- ">>>> ...\n",
- ">>>> }\n",
- ">>>>\n",
- ">>>> arch_fault_handler()\n",
- ">>>> {\n",
- ">>>> if (speculative_page_fault(&vma))\n",
- ">>>> goto done\n",
- ">>>> again:\n",
- ">>>> lock(mmap_sem)\n",
- ">>>> vma = find_vma();\n",
- ">>>> handle_pte_fault(vma);\n",
- ">>>> if retry\n",
- ">>>> unlock(mmap_sem)\n",
- ">>>> goto again;\n",
- ">>>> done:\n",
- ">>>> handle fault error\n",
- ">>>> }\n",
- ">>>>\n",
- ">>>> Support for THP is not done because when checking for the PMD, we can be\n",
- ">>>> confused by an in progress collapsing operation done by khugepaged. The\n",
- ">>>> issue is that pmd_none() could be true either if the PMD is not already\n",
- ">>>> populated or if the underlying PTE are in the way to be collapsed. So we\n",
- ">>>> cannot safely allocate a PMD if pmd_none() is true.\n",
- ">>>>\n",
- ">>>> This series add a new software performance event named 'speculative-faults'\n",
- ">>>> or 'spf'. It counts the number of successful page fault event handled\n",
- ">>>> speculatively. When recording 'faults,spf' events, the faults one is\n",
- ">>>> counting the total number of page fault events while 'spf' is only counting\n",
- ">>>> the part of the faults processed speculatively.\n",
- ">>>>\n",
- ">>>> There are some trace events introduced by this series. They allow\n",
- ">>>> identifying why the page faults were not processed speculatively. This\n",
- ">>>> doesn't take in account the faults generated by a monothreaded process\n",
- ">>>> which directly processed while holding the mmap_sem. This trace events are\n",
- ">>>> grouped in a system named 'pagefault', they are:\n",
- ">>>> - pagefault:spf_vma_changed : if the VMA has been changed in our back\n",
- ">>>> - pagefault:spf_vma_noanon : the vma->anon_vma field was not yet set.\n",
- ">>>> - pagefault:spf_vma_notsup : the VMA's type is not supported\n",
- ">>>> - pagefault:spf_vma_access : the VMA's access right are not respected\n",
- ">>>> - pagefault:spf_pmd_changed : the upper PMD pointer has changed in our\n",
- ">>>> back.\n",
- ">>>>\n",
- ">>>> To record all the related events, the easier is to run perf with the\n",
- ">>>> following arguments :\n",
- ">>>> \$ perf stat -e 'faults,spf,pagefault:*' <command>\n",
- ">>>>\n",
- ">>>> There is also a dedicated vmstat counter showing the number of successful\n",
- ">>>> page fault handled speculatively. I can be seen this way:\n",
- ">>>> \$ grep speculative_pgfault /proc/vmstat\n",
- ">>>>\n",
- ">>>> This series builds on top of v4.16-mmotm-2018-04-13-17-28 and is functional\n",
- ">>>> on x86, PowerPC and arm64.\n",
- ">>>>\n",
- ">>>> ---------------------\n",
- ">>>> Real Workload results\n",
- ">>>>\n",
- ">>>> As mentioned in previous email, we did non official runs using a \"popular\n",
- ">>>> in memory multithreaded database product\" on 176 cores SMT8 Power system\n",
- ">>>> which showed a 30% improvements in the number of transaction processed per\n",
- ">>>> second. This run has been done on the v6 series, but changes introduced in\n",
- ">>>> this new version should not impact the performance boost seen.\n",
- ">>>>\n",
- ">>>> Here are the perf data captured during 2 of these runs on top of the v8\n",
- ">>>> series:\n",
- ">>>> vanilla spf\n",
- ">>>> faults 89.418 101.364 +13%\n",
- ">>>> spf n/a 97.989\n",
- ">>>>\n",
- ">>>> With the SPF kernel, most of the page fault were processed in a speculative\n",
- ">>>> way.\n",
- ">>>>\n",
- ">>>> Ganesh Mahendran had backported the series on top of a 4.9 kernel and gave\n",
- ">>>> it a try on an android device. He reported that the application launch time\n",
- ">>>> was improved in average by 6%, and for large applications (~100 threads) by\n",
- ">>>> 20%.\n",
- ">>>>\n",
- ">>>> Here are the launch time Ganesh mesured on Android 8.0 on top of a Qcom\n",
- ">>>> MSM845 (8 cores) with 6GB (the less is better):\n",
- ">>>>\n",
- ">>>> Application 4.9 4.9+spf delta\n",
- ">>>> com.tencent.mm 416 389 -7%\n",
- ">>>> com.eg.android.AlipayGphone 1135 986 -13%\n",
- ">>>> com.tencent.mtt 455 454 0%\n",
- ">>>> com.qqgame.hlddz 1497 1409 -6%\n",
- ">>>> com.autonavi.minimap 711 701 -1%\n",
- ">>>> com.tencent.tmgp.sgame 788 748 -5%\n",
- ">>>> com.immomo.momo 501 487 -3%\n",
- ">>>> com.tencent.peng 2145 2112 -2%\n",
- ">>>> com.smile.gifmaker 491 461 -6%\n",
- ">>>> com.baidu.BaiduMap 479 366 -23%\n",
- ">>>> com.taobao.taobao 1341 1198 -11%\n",
- ">>>> com.baidu.searchbox 333 314 -6%\n",
- ">>>> com.tencent.mobileqq 394 384 -3%\n",
- ">>>> com.sina.weibo 907 906 0%\n",
- ">>>> com.youku.phone 816 731 -11%\n",
- ">>>> com.happyelements.AndroidAnimal.qq 763 717 -6%\n",
- ">>>> com.UCMobile 415 411 -1%\n",
- ">>>> com.tencent.tmgp.ak 1464 1431 -2%\n",
- ">>>> com.tencent.qqmusic 336 329 -2%\n",
- ">>>> com.sankuai.meituan 1661 1302 -22%\n",
- ">>>> com.netease.cloudmusic 1193 1200 1%\n",
- ">>>> air.tv.douyu.android 4257 4152 -2%\n",
- ">>>>\n",
- ">>>> ------------------\n",
- ">>>> Benchmarks results\n",
- ">>>>\n",
- ">>>> Base kernel is v4.17.0-rc4-mm1\n",
- ">>>> SPF is BASE + this series\n",
- ">>>>\n",
- ">>>> Kernbench:\n",
- ">>>> ----------\n",
- ">>>> Here are the results on a 16 CPUs X86 guest using kernbench on a 4.15\n",
- ">>>> kernel (kernel is build 5 times):\n",
- ">>>>\n",
- ">>>> Average Half load -j 8\n",
- ">>>> Run (std deviation)\n",
- ">>>> BASE SPF\n",
- ">>>> Elapsed Time 1448.65 (5.72312) 1455.84 (4.84951) 0.50%\n",
- ">>>> User Time 10135.4 (30.3699) 10148.8 (31.1252) 0.13%\n",
- ">>>> System Time 900.47 (2.81131) 923.28 (7.52779) 2.53%\n",
- ">>>> Percent CPU 761.4 (1.14018) 760.2 (0.447214) -0.16%\n",
- ">>>> Context Switches 85380 (3419.52) 84748 (1904.44) -0.74%\n",
- ">>>> Sleeps 105064 (1240.96) 105074 (337.612) 0.01%\n",
- ">>>>\n",
- ">>>> Average Optimal load -j 16\n",
- ">>>> Run (std deviation)\n",
- ">>>> BASE SPF\n",
- ">>>> Elapsed Time 920.528 (10.1212) 927.404 (8.91789) 0.75%\n",
- ">>>> User Time 11064.8 (981.142) 11085 (990.897) 0.18%\n",
- ">>>> System Time 979.904 (84.0615) 1001.14 (82.5523) 2.17%\n",
- ">>>> Percent CPU 1089.5 (345.894) 1086.1 (343.545) -0.31%\n",
- ">>>> Context Switches 159488 (78156.4) 158223 (77472.1) -0.79%\n",
- ">>>> Sleeps 110566 (5877.49) 110388 (5617.75) -0.16%\n",
- ">>>>\n",
- ">>>>\n",
- ">>>> During a run on the SPF, perf events were captured:\n",
- ">>>> Performance counter stats for '../kernbench -M':\n",
- ">>>> 526743764 faults\n",
- ">>>> 210 spf\n",
- ">>>> 3 pagefault:spf_vma_changed\n",
- ">>>> 0 pagefault:spf_vma_noanon\n",
- ">>>> 2278 pagefault:spf_vma_notsup\n",
- ">>>> 0 pagefault:spf_vma_access\n",
- ">>>> 0 pagefault:spf_pmd_changed\n",
- ">>>>\n",
- ">>>> Very few speculative page faults were recorded as most of the processes\n",
- ">>>> involved are monothreaded (sounds that on this architecture some threads\n",
- ">>>> were created during the kernel build processing).\n",
- ">>>>\n",
- ">>>> Here are the kerbench results on a 80 CPUs Power8 system:\n",
- ">>>>\n",
- ">>>> Average Half load -j 40\n",
- ">>>> Run (std deviation)\n",
- ">>>> BASE SPF\n",
- ">>>> Elapsed Time 117.152 (0.774642) 117.166 (0.476057) 0.01%\n",
- ">>>> User Time 4478.52 (24.7688) 4479.76 (9.08555) 0.03%\n",
- ">>>> System Time 131.104 (0.720056) 134.04 (0.708414) 2.24%\n",
- ">>>> Percent CPU 3934 (19.7104) 3937.2 (19.0184) 0.08%\n",
- ">>>> Context Switches 92125.4 (576.787) 92581.6 (198.622) 0.50%\n",
- ">>>> Sleeps 317923 (652.499) 318469 (1255.59) 0.17%\n",
- ">>>>\n",
- ">>>> Average Optimal load -j 80\n",
- ">>>> Run (std deviation)\n",
- ">>>> BASE SPF\n",
- ">>>> Elapsed Time 107.73 (0.632416) 107.31 (0.584936) -0.39%\n",
- ">>>> User Time 5869.86 (1466.72) 5871.71 (1467.27) 0.03%\n",
- ">>>> System Time 153.728 (23.8573) 157.153 (24.3704) 2.23%\n",
- ">>>> Percent CPU 5418.6 (1565.17) 5436.7 (1580.91) 0.33%\n",
- ">>>> Context Switches 223861 (138865) 225032 (139632) 0.52%\n",
- ">>>> Sleeps 330529 (13495.1) 332001 (14746.2) 0.45%\n",
- ">>>>\n",
- ">>>> During a run on the SPF, perf events were captured:\n",
- ">>>> Performance counter stats for '../kernbench -M':\n",
- ">>>> 116730856 faults\n",
- ">>>> 0 spf\n",
- ">>>> 3 pagefault:spf_vma_changed\n",
- ">>>> 0 pagefault:spf_vma_noanon\n",
- ">>>> 476 pagefault:spf_vma_notsup\n",
- ">>>> 0 pagefault:spf_vma_access\n",
- ">>>> 0 pagefault:spf_pmd_changed\n",
- ">>>>\n",
- ">>>> Most of the processes involved are monothreaded so SPF is not activated but\n",
- ">>>> there is no impact on the performance.\n",
- ">>>>\n",
- ">>>> Ebizzy:\n",
- ">>>> -------\n",
- ">>>> The test is counting the number of records per second it can manage, the\n",
- ">>>> higher is the best. I run it like this 'ebizzy -mTt <nrcpus>'. To get\n",
- ">>>> consistent result I repeated the test 100 times and measure the average\n",
- ">>>> result. The number is the record processes per second, the higher is the\n",
- ">>>> best.\n",
- ">>>>\n",
- ">>>> BASE SPF delta\n",
- ">>>> 16 CPUs x86 VM 742.57 1490.24 100.69%\n",
- ">>>> 80 CPUs P8 node 13105.4 24174.23 84.46%\n",
- ">>>>\n",
- ">>>> Here are the performance counter read during a run on a 16 CPUs x86 VM:\n",
- ">>>> Performance counter stats for './ebizzy -mTt 16':\n",
- ">>>> 1706379 faults\n",
- ">>>> 1674599 spf\n",
- ">>>> 30588 pagefault:spf_vma_changed\n",
- ">>>> 0 pagefault:spf_vma_noanon\n",
- ">>>> 363 pagefault:spf_vma_notsup\n",
- ">>>> 0 pagefault:spf_vma_access\n",
- ">>>> 0 pagefault:spf_pmd_changed\n",
- ">>>>\n",
- ">>>> And the ones captured during a run on a 80 CPUs Power node:\n",
- ">>>> Performance counter stats for './ebizzy -mTt 80':\n",
- ">>>> 1874773 faults\n",
- ">>>> 1461153 spf\n",
- ">>>> 413293 pagefault:spf_vma_changed\n",
- ">>>> 0 pagefault:spf_vma_noanon\n",
- ">>>> 200 pagefault:spf_vma_notsup\n",
- ">>>> 0 pagefault:spf_vma_access\n",
- ">>>> 0 pagefault:spf_pmd_changed\n",
- ">>>>\n",
- ">>>> In ebizzy's case most of the page fault were handled in a speculative way,\n",
- ">>>> leading the ebizzy performance boost.\n",
- ">>>>\n",
- ">>>> ------------------\n",
- ">>>> Changes since v10 (https://lkml.org/lkml/2018/4/17/572):\n",
- ">>>> - Accounted for all review feedbacks from Punit Agrawal, Ganesh Mahendran\n",
- ">>>> and Minchan Kim, hopefully.\n",
- ">>>> - Remove unneeded check on CONFIG_SPECULATIVE_PAGE_FAULT in\n",
- ">>>> __do_page_fault().\n",
- ">>>> - Loop in pte_spinlock() and pte_map_lock() when pte try lock fails\n",
- ">>>> instead\n",
- ">>>> of aborting the speculative page fault handling. Dropping the now\n",
- ">>>> useless\n",
- ">>>> trace event pagefault:spf_pte_lock.\n",
- ">>>> - No more try to reuse the fetched VMA during the speculative page fault\n",
- ">>>> handling when retrying is needed. This adds a lot of complexity and\n",
- ">>>> additional tests done didn't show a significant performance improvement.\n",
- ">>>> - Convert IS_ENABLED(CONFIG_NUMA) back to #ifdef due to build error.\n",
- ">>>>\n",
- ">>>> [1] http://linux-kernel.2935.n7.nabble.com/RFC-PATCH-0-6-Another-go-at-speculative-page-faults-tt965642.html#none\n",
- ">>>> [2] https://patchwork.kernel.org/patch/9999687/\n",
- ">>>>\n",
- ">>>>\n",
- ">>>> Laurent Dufour (20):\n",
- ">>>> mm: introduce CONFIG_SPECULATIVE_PAGE_FAULT\n",
- ">>>> x86/mm: define ARCH_SUPPORTS_SPECULATIVE_PAGE_FAULT\n",
- ">>>> powerpc/mm: set ARCH_SUPPORTS_SPECULATIVE_PAGE_FAULT\n",
- ">>>> mm: introduce pte_spinlock for FAULT_FLAG_SPECULATIVE\n",
- ">>>> mm: make pte_unmap_same compatible with SPF\n",
- ">>>> mm: introduce INIT_VMA()\n",
- ">>>> mm: protect VMA modifications using VMA sequence count\n",
- ">>>> mm: protect mremap() against SPF hanlder\n",
- ">>>> mm: protect SPF handler against anon_vma changes\n",
- ">>>> mm: cache some VMA fields in the vm_fault structure\n",
- ">>>> mm/migrate: Pass vm_fault pointer to migrate_misplaced_page()\n",
- ">>>> mm: introduce __lru_cache_add_active_or_unevictable\n",
- ">>>> mm: introduce __vm_normal_page()\n",
- ">>>> mm: introduce __page_add_new_anon_rmap()\n",
- ">>>> mm: protect mm_rb tree with a rwlock\n",
- ">>>> mm: adding speculative page fault failure trace events\n",
- ">>>> perf: add a speculative page fault sw event\n",
- ">>>> perf tools: add support for the SPF perf event\n",
- ">>>> mm: add speculative page fault vmstats\n",
- ">>>> powerpc/mm: add speculative page fault\n",
- ">>>>\n",
- ">>>> Mahendran Ganesh (2):\n",
- ">>>> arm64/mm: define ARCH_SUPPORTS_SPECULATIVE_PAGE_FAULT\n",
- ">>>> arm64/mm: add speculative page fault\n",
- ">>>>\n",
- ">>>> Peter Zijlstra (4):\n",
- ">>>> mm: prepare for FAULT_FLAG_SPECULATIVE\n",
- ">>>> mm: VMA sequence count\n",
- ">>>> mm: provide speculative fault infrastructure\n",
- ">>>> x86/mm: add speculative pagefault handling\n",
- ">>>>\n",
- ">>>> arch/arm64/Kconfig | 1 +\n",
- ">>>> arch/arm64/mm/fault.c | 12 +\n",
- ">>>> arch/powerpc/Kconfig | 1 +\n",
- ">>>> arch/powerpc/mm/fault.c | 16 +\n",
- ">>>> arch/x86/Kconfig | 1 +\n",
- ">>>> arch/x86/mm/fault.c | 27 +-\n",
- ">>>> fs/exec.c | 2 +-\n",
- ">>>> fs/proc/task_mmu.c | 5 +-\n",
- ">>>> fs/userfaultfd.c | 17 +-\n",
- ">>>> include/linux/hugetlb_inline.h | 2 +-\n",
- ">>>> include/linux/migrate.h | 4 +-\n",
- ">>>> include/linux/mm.h | 136 +++++++-\n",
- ">>>> include/linux/mm_types.h | 7 +\n",
- ">>>> include/linux/pagemap.h | 4 +-\n",
- ">>>> include/linux/rmap.h | 12 +-\n",
- ">>>> include/linux/swap.h | 10 +-\n",
- ">>>> include/linux/vm_event_item.h | 3 +\n",
- ">>>> include/trace/events/pagefault.h | 80 +++++\n",
- ">>>> include/uapi/linux/perf_event.h | 1 +\n",
- ">>>> kernel/fork.c | 5 +-\n",
- ">>>> mm/Kconfig | 22 ++\n",
- ">>>> mm/huge_memory.c | 6 +-\n",
- ">>>> mm/hugetlb.c | 2 +\n",
- ">>>> mm/init-mm.c | 3 +\n",
- ">>>> mm/internal.h | 20 ++\n",
- ">>>> mm/khugepaged.c | 5 +\n",
- ">>>> mm/madvise.c | 6 +-\n",
- ">>>> mm/memory.c | 612 +++++++++++++++++++++++++++++-----\n",
- ">>>> mm/mempolicy.c | 51 ++-\n",
- ">>>> mm/migrate.c | 6 +-\n",
- ">>>> mm/mlock.c | 13 +-\n",
- ">>>> mm/mmap.c | 229 ++++++++++---\n",
- ">>>> mm/mprotect.c | 4 +-\n",
- ">>>> mm/mremap.c | 13 +\n",
- ">>>> mm/nommu.c | 2 +-\n",
- ">>>> mm/rmap.c | 5 +-\n",
- ">>>> mm/swap.c | 6 +-\n",
- ">>>> mm/swap_state.c | 8 +-\n",
- ">>>> mm/vmstat.c | 5 +-\n",
- ">>>> tools/include/uapi/linux/perf_event.h | 1 +\n",
- ">>>> tools/perf/util/evsel.c | 1 +\n",
- ">>>> tools/perf/util/parse-events.c | 4 +\n",
- ">>>> tools/perf/util/parse-events.l | 1 +\n",
- ">>>> tools/perf/util/python.c | 1 +\n",
- ">>>> 44 files changed, 1161 insertions(+), 211 deletions(-)\n",
- ">>>> create mode 100644 include/trace/events/pagefault.h\n",
- ">>>>\n",
- ">>>> --\n",
- ">>>> 2.7.4\n",
- ">>>>\n",
- ">>>>\n",
- ">>>\n",
- ">>\n",
- ">"
+ "Hi Laurent,=0A=\n",
+ "=0A=\n",
+ "=0A=\n",
+ "For the test result on Intel 4s skylake platform (192 CPUs, 768G Memory), t=\n",
+ "he below test cases all were run 3 times.=0A=\n",
+ "I check the test results, only page_fault3_thread/enable THP have 6% stddev=\n",
+ " for head commit, other tests have lower stddev.=0A=\n",
+ "=0A=\n",
+ "And I did not find other high variation on test case result.=0A=\n",
+ "=0A=\n",
+ "a). Enable THP=0A=\n",
+ "testcase base stddev change head =\n",
+ " stddev metric=0A=\n",
+ "page_fault3/enable THP 10519 =B1 3% -20.5% 8368 =\n",
+ " =B16% will-it-scale.per_thread_ops=0A=\n",
+ "page_fault2/enalbe THP 8281 =B1 2% -18.8% 6728 =\n",
+ " will-it-scale.per_thread_ops=0A=\n",
+ "brk1/eanble THP 998475 -2.2% 976893 =\n",
+ " will-it-scale.per_process_ops=0A=\n",
+ "context_switch1/enable THP 223910 -1.3% 220930 =\n",
+ " will-it-scale.per_process_ops=0A=\n",
+ "context_switch1/enable THP 233722 -1.0% 231288 =\n",
+ " will-it-scale.per_thread_ops=0A=\n",
+ "=0A=\n",
+ "b). Disable THP=0A=\n",
+ "page_fault3/disable THP 10856 -23.1% 8344 =\n",
+ " will-it-scale.per_thread_ops=0A=\n",
+ "page_fault2/disable THP 8147 -18.8% 6613 =\n",
+ " will-it-scale.per_thread_ops=0A=\n",
+ "brk1/disable THP 957 -7.9% 881 =\n",
+ " will-it-scale.per_thread_ops=0A=\n",
+ "context_switch1/disable THP 237006 -2.2% 231907 =\n",
+ " will-it-scale.per_thread_ops=0A=\n",
+ "brk1/disable THP 997317 -2.0% 977778 =\n",
+ " will-it-scale.per_process_ops=0A=\n",
+ "page_fault3/disable THP 467454 -1.8% 459251 =\n",
+ " will-it-scale.per_process_ops=0A=\n",
+ "context_switch1/disable THP 224431 -1.3% 221567 =\n",
+ " will-it-scale.per_process_ops=0A=\n",
+ "=0A=\n",
+ "=0A=\n",
+ "Best regards,=0A=\n",
+ "Haiyan Song=0A=\n",
+ "________________________________________=0A=\n",
+ "From: Laurent Dufour [ldufour\@linux.vnet.ibm.com]=0A=\n",
+ "Sent: Monday, July 02, 2018 4:59 PM=0A=\n",
+ "To: Song, HaiyanX=0A=\n",
+ "Cc: akpm\@linux-foundation.org; mhocko\@kernel.org; peterz\@infradead.org; kir=\n",
+ "ill\@shutemov.name; ak\@linux.intel.com; dave\@stgolabs.net; jack\@suse.cz; Mat=\n",
+ "thew Wilcox; khandual\@linux.vnet.ibm.com; aneesh.kumar\@linux.vnet.ibm.com; =\n",
+ "benh\@kernel.crashing.org; mpe\@ellerman.id.au; paulus\@samba.org; Thomas Glei=\n",
+ "xner; Ingo Molnar; hpa\@zytor.com; Will Deacon; Sergey Senozhatsky; sergey.s=\n",
+ "enozhatsky.work\@gmail.com; Andrea Arcangeli; Alexei Starovoitov; Wang, Kemi=\n",
+ "; Daniel Jordan; David Rientjes; Jerome Glisse; Ganesh Mahendran; Minchan K=\n",
+ "im; Punit Agrawal; vinayak menon; Yang Shi; linux-kernel\@vger.kernel.org; l=\n",
+ "inux-mm\@kvack.org; haren\@linux.vnet.ibm.com; npiggin\@gmail.com; bsingharora=\n",
+ "\@gmail.com; paulmck\@linux.vnet.ibm.com; Tim Chen; linuxppc-dev\@lists.ozlabs=\n",
+ ".org; x86\@kernel.org=0A=\n",
+ "Subject: Re: [PATCH v11 00/26] Speculative page faults=0A=\n",
+ "=0A=\n",
+ "On 11/06/2018 09:49, Song, HaiyanX wrote:=0A=\n",
+ "> Hi Laurent,=0A=\n",
+ ">=0A=\n",
+ "> Regression test for v11 patch serials have been run, some regression is f=\n",
+ "ound by LKP-tools (linux kernel performance)=0A=\n",
+ "> tested on Intel 4s skylake platform. This time only test the cases which =\n",
+ "have been run and found regressions on=0A=\n",
+ "> V9 patch serials.=0A=\n",
+ ">=0A=\n",
+ "> The regression result is sorted by the metric will-it-scale.per_thread_op=\n",
+ "s.=0A=\n",
+ "> branch: Laurent-Dufour/Speculative-page-faults/20180520-045126=0A=\n",
+ "> commit id:=0A=\n",
+ "> head commit : a7a8993bfe3ccb54ad468b9f1799649e4ad1ff12=0A=\n",
+ "> base commit : ba98a1cdad71d259a194461b3a61471b49b14df1=0A=\n",
+ "> Benchmark: will-it-scale=0A=\n",
+ "> Download link: https://github.com/antonblanchard/will-it-scale/tree/maste=\n",
+ "r=0A=\n",
+ ">=0A=\n",
+ "> Metrics:=0A=\n",
+ "> will-it-scale.per_process_ops=3Dprocesses/nr_cpu=0A=\n",
+ "> will-it-scale.per_thread_ops=3Dthreads/nr_cpu=0A=\n",
+ "> test box: lkp-skl-4sp1(nr_cpu=3D192,memory=3D768G)=0A=\n",
+ "> THP: enable / disable=0A=\n",
+ "> nr_task:100%=0A=\n",
+ ">=0A=\n",
+ "> 1. Regressions:=0A=\n",
+ ">=0A=\n",
+ "> a). Enable THP=0A=\n",
+ "> testcase base change head =\n",
+ " metric=0A=\n",
+ "> page_fault3/enable THP 10519 -20.5% 836 wi=\n",
+ "ll-it-scale.per_thread_ops=0A=\n",
+ "> page_fault2/enalbe THP 8281 -18.8% 6728 wi=\n",
+ "ll-it-scale.per_thread_ops=0A=\n",
+ "> brk1/eanble THP 998475 -2.2% 976893 wi=\n",
+ "ll-it-scale.per_process_ops=0A=\n",
+ "> context_switch1/enable THP 223910 -1.3% 220930 wi=\n",
+ "ll-it-scale.per_process_ops=0A=\n",
+ "> context_switch1/enable THP 233722 -1.0% 231288 wi=\n",
+ "ll-it-scale.per_thread_ops=0A=\n",
+ ">=0A=\n",
+ "> b). Disable THP=0A=\n",
+ "> page_fault3/disable THP 10856 -23.1% 8344 wi=\n",
+ "ll-it-scale.per_thread_ops=0A=\n",
+ "> page_fault2/disable THP 8147 -18.8% 6613 wi=\n",
+ "ll-it-scale.per_thread_ops=0A=\n",
+ "> brk1/disable THP 957 -7.9% 881 wi=\n",
+ "ll-it-scale.per_thread_ops=0A=\n",
+ "> context_switch1/disable THP 237006 -2.2% 231907 wi=\n",
+ "ll-it-scale.per_thread_ops=0A=\n",
+ "> brk1/disable THP 997317 -2.0% 977778 wi=\n",
+ "ll-it-scale.per_process_ops=0A=\n",
+ "> page_fault3/disable THP 467454 -1.8% 459251 wi=\n",
+ "ll-it-scale.per_process_ops=0A=\n",
+ "> context_switch1/disable THP 224431 -1.3% 221567 wi=\n",
+ "ll-it-scale.per_process_ops=0A=\n",
+ ">=0A=\n",
+ "> Notes: for the above values of test result, the higher is better.=0A=\n",
+ "=0A=\n",
+ "I tried the same tests on my PowerPC victim VM (1024 CPUs, 11TB) and I can'=\n",
+ "t=0A=\n",
+ "get reproducible results. The results have huge variation, even on the vani=\n",
+ "lla=0A=\n",
+ "kernel, and I can't state on any changes due to that.=0A=\n",
+ "=0A=\n",
+ "I tried on smaller node (80 CPUs, 32G), and the tests ran better, but I did=\n",
+ "n't=0A=\n",
+ "measure any changes between the vanilla and the SPF patched ones:=0A=\n",
+ "=0A=\n",
+ "test THP enabled 4.17.0-rc4-mm1 spf delta=0A=\n",
+ "page_fault3_threads 2697.7 2683.5 -0.53%=0A=\n",
+ "page_fault2_threads 170660.6 169574.1 -0.64%=0A=\n",
+ "context_switch1_threads 6915269.2 6877507.3 -0.55%=0A=\n",
+ "context_switch1_processes 6478076.2 6529493.5 0.79%=0A=\n",
+ "brk1 243391.2 238527.5 -2.00%=0A=\n",
+ "=0A=\n",
+ "Tests were run 10 times, no high variation detected.=0A=\n",
+ "=0A=\n",
+ "Did you see high variation on your side ? How many times the test were run =\n",
+ "to=0A=\n",
+ "compute the average values ?=0A=\n",
+ "=0A=\n",
+ "Thanks,=0A=\n",
+ "Laurent.=0A=\n",
+ "=0A=\n",
+ "=0A=\n",
+ ">=0A=\n",
+ "> 2. Improvement: not found improvement based on the selected test cases.=\n",
+ "=0A=\n",
+ ">=0A=\n",
+ ">=0A=\n",
+ "> Best regards=0A=\n",
+ "> Haiyan Song=0A=\n",
+ "> ________________________________________=0A=\n",
+ "> From: owner-linux-mm\@kvack.org [owner-linux-mm\@kvack.org] on behalf of La=\n",
+ "urent Dufour [ldufour\@linux.vnet.ibm.com]=0A=\n",
+ "> Sent: Monday, May 28, 2018 4:54 PM=0A=\n",
+ "> To: Song, HaiyanX=0A=\n",
+ "> Cc: akpm\@linux-foundation.org; mhocko\@kernel.org; peterz\@infradead.org; k=\n",
+ "irill\@shutemov.name; ak\@linux.intel.com; dave\@stgolabs.net; jack\@suse.cz; M=\n",
+ "atthew Wilcox; khandual\@linux.vnet.ibm.com; aneesh.kumar\@linux.vnet.ibm.com=\n",
+ "; benh\@kernel.crashing.org; mpe\@ellerman.id.au; paulus\@samba.org; Thomas Gl=\n",
+ "eixner; Ingo Molnar; hpa\@zytor.com; Will Deacon; Sergey Senozhatsky; sergey=\n",
+ ".senozhatsky.work\@gmail.com; Andrea Arcangeli; Alexei Starovoitov; Wang, Ke=\n",
+ "mi; Daniel Jordan; David Rientjes; Jerome Glisse; Ganesh Mahendran; Minchan=\n",
+ " Kim; Punit Agrawal; vinayak menon; Yang Shi; linux-kernel\@vger.kernel.org;=\n",
+ " linux-mm\@kvack.org; haren\@linux.vnet.ibm.com; npiggin\@gmail.com; bsingharo=\n",
+ "ra\@gmail.com; paulmck\@linux.vnet.ibm.com; Tim Chen; linuxppc-dev\@lists.ozla=\n",
+ "bs.org; x86\@kernel.org=0A=\n",
+ "> Subject: Re: [PATCH v11 00/26] Speculative page faults=0A=\n",
+ ">=0A=\n",
+ "> On 28/05/2018 10:22, Haiyan Song wrote:=0A=\n",
+ ">> Hi Laurent,=0A=\n",
+ ">>=0A=\n",
+ ">> Yes, these tests are done on V9 patch.=0A=\n",
+ ">=0A=\n",
+ "> Do you plan to give this V11 a run ?=0A=\n",
+ ">=0A=\n",
+ ">>=0A=\n",
+ ">>=0A=\n",
+ ">> Best regards,=0A=\n",
+ ">> Haiyan Song=0A=\n",
+ ">>=0A=\n",
+ ">> On Mon, May 28, 2018 at 09:51:34AM +0200, Laurent Dufour wrote:=0A=\n",
+ ">>> On 28/05/2018 07:23, Song, HaiyanX wrote:=0A=\n",
+ ">>>>=0A=\n",
+ ">>>> Some regression and improvements is found by LKP-tools(linux kernel pe=\n",
+ "rformance) on V9 patch series=0A=\n",
+ ">>>> tested on Intel 4s Skylake platform.=0A=\n",
+ ">>>=0A=\n",
+ ">>> Hi,=0A=\n",
+ ">>>=0A=\n",
+ ">>> Thanks for reporting this benchmark results, but you mentioned the \"V9 =\n",
+ "patch=0A=\n",
+ ">>> series\" while responding to the v11 header series...=0A=\n",
+ ">>> Were these tests done on v9 or v11 ?=0A=\n",
+ ">>>=0A=\n",
+ ">>> Cheers,=0A=\n",
+ ">>> Laurent.=0A=\n",
+ ">>>=0A=\n",
+ ">>>>=0A=\n",
+ ">>>> The regression result is sorted by the metric will-it-scale.per_thread=\n",
+ "_ops.=0A=\n",
+ ">>>> Branch: Laurent-Dufour/Speculative-page-faults/20180316-151833 (V9 pat=\n",
+ "ch series)=0A=\n",
+ ">>>> Commit id:=0A=\n",
+ ">>>> base commit: d55f34411b1b126429a823d06c3124c16283231f=0A=\n",
+ ">>>> head commit: 0355322b3577eeab7669066df42c550a56801110=0A=\n",
+ ">>>> Benchmark suite: will-it-scale=0A=\n",
+ ">>>> Download link:=0A=\n",
+ ">>>> https://github.com/antonblanchard/will-it-scale/tree/master/tests=0A=\n",
+ ">>>> Metrics:=0A=\n",
+ ">>>> will-it-scale.per_process_ops=3Dprocesses/nr_cpu=0A=\n",
+ ">>>> will-it-scale.per_thread_ops=3Dthreads/nr_cpu=0A=\n",
+ ">>>> test box: lkp-skl-4sp1(nr_cpu=3D192,memory=3D768G)=0A=\n",
+ ">>>> THP: enable / disable=0A=\n",
+ ">>>> nr_task: 100%=0A=\n",
+ ">>>>=0A=\n",
+ ">>>> 1. Regressions:=0A=\n",
+ ">>>> a) THP enabled:=0A=\n",
+ ">>>> testcase base change head =\n",
+ " metric=0A=\n",
+ ">>>> page_fault3/ enable THP 10092 -17.5% 8323 =\n",
+ " will-it-scale.per_thread_ops=0A=\n",
+ ">>>> page_fault2/ enable THP 8300 -17.2% 6869 =\n",
+ " will-it-scale.per_thread_ops=0A=\n",
+ ">>>> brk1/ enable THP 957.67 -7.6% 885 =\n",
+ " will-it-scale.per_thread_ops=0A=\n",
+ ">>>> page_fault3/ enable THP 172821 -5.3% 163692 =\n",
+ " will-it-scale.per_process_ops=0A=\n",
+ ">>>> signal1/ enable THP 9125 -3.2% 8834 =\n",
+ " will-it-scale.per_process_ops=0A=\n",
+ ">>>>=0A=\n",
+ ">>>> b) THP disabled:=0A=\n",
+ ">>>> testcase base change head =\n",
+ " metric=0A=\n",
+ ">>>> page_fault3/ disable THP 10107 -19.1% 8180 =\n",
+ " will-it-scale.per_thread_ops=0A=\n",
+ ">>>> page_fault2/ disable THP 8432 -17.8% 6931 =\n",
+ " will-it-scale.per_thread_ops=0A=\n",
+ ">>>> context_switch1/ disable THP 215389 -6.8% 200776 =\n",
+ " will-it-scale.per_thread_ops=0A=\n",
+ ">>>> brk1/ disable THP 939.67 -6.6% 877.3=\n",
+ "3 will-it-scale.per_thread_ops=0A=\n",
+ ">>>> page_fault3/ disable THP 173145 -4.7% 165064 =\n",
+ " will-it-scale.per_process_ops=0A=\n",
+ ">>>> signal1/ disable THP 9162 -3.9% 8802 =\n",
+ " will-it-scale.per_process_ops=0A=\n",
+ ">>>>=0A=\n",
+ ">>>> 2. Improvements:=0A=\n",
+ ">>>> a) THP enabled:=0A=\n",
+ ">>>> testcase base change head =\n",
+ " metric=0A=\n",
+ ">>>> malloc1/ enable THP 66.33 +469.8% 383.6=\n",
+ "7 will-it-scale.per_thread_ops=0A=\n",
+ ">>>> writeseek3/ enable THP 2531 +4.5% 2646 =\n",
+ " will-it-scale.per_thread_ops=0A=\n",
+ ">>>> signal1/ enable THP 989.33 +2.8% 1016 =\n",
+ " will-it-scale.per_thread_ops=0A=\n",
+ ">>>>=0A=\n",
+ ">>>> b) THP disabled:=0A=\n",
+ ">>>> testcase base change head =\n",
+ " metric=0A=\n",
+ ">>>> malloc1/ disable THP 90.33 +417.3% 467.3=\n",
+ "3 will-it-scale.per_thread_ops=0A=\n",
+ ">>>> read2/ disable THP 58934 +39.2% 82060 =\n",
+ " will-it-scale.per_thread_ops=0A=\n",
+ ">>>> page_fault1/ disable THP 8607 +36.4% 11736 =\n",
+ " will-it-scale.per_thread_ops=0A=\n",
+ ">>>> read1/ disable THP 314063 +12.7% 353934 =\n",
+ " will-it-scale.per_thread_ops=0A=\n",
+ ">>>> writeseek3/ disable THP 2452 +12.5% 2759 =\n",
+ " will-it-scale.per_thread_ops=0A=\n",
+ ">>>> signal1/ disable THP 971.33 +5.5% 1024 =\n",
+ " will-it-scale.per_thread_ops=0A=\n",
+ ">>>>=0A=\n",
+ ">>>> Notes: for above values in column \"change\", the higher value means tha=\n",
+ "t the related testcase result=0A=\n",
+ ">>>> on head commit is better than that on base commit for this benchmark.=\n",
+ "=0A=\n",
+ ">>>>=0A=\n",
+ ">>>>=0A=\n",
+ ">>>> Best regards=0A=\n",
+ ">>>> Haiyan Song=0A=\n",
+ ">>>>=0A=\n",
+ ">>>> ________________________________________=0A=\n",
+ ">>>> From: owner-linux-mm\@kvack.org [owner-linux-mm\@kvack.org] on behalf of=\n",
+ " Laurent Dufour [ldufour\@linux.vnet.ibm.com]=0A=\n",
+ ">>>> Sent: Thursday, May 17, 2018 7:06 PM=0A=\n",
+ ">>>> To: akpm\@linux-foundation.org; mhocko\@kernel.org; peterz\@infradead.org=\n",
+ "; kirill\@shutemov.name; ak\@linux.intel.com; dave\@stgolabs.net; jack\@suse.cz=\n",
+ "; Matthew Wilcox; khandual\@linux.vnet.ibm.com; aneesh.kumar\@linux.vnet.ibm.=\n",
+ "com; benh\@kernel.crashing.org; mpe\@ellerman.id.au; paulus\@samba.org; Thomas=\n",
+ " Gleixner; Ingo Molnar; hpa\@zytor.com; Will Deacon; Sergey Senozhatsky; ser=\n",
+ "gey.senozhatsky.work\@gmail.com; Andrea Arcangeli; Alexei Starovoitov; Wang,=\n",
+ " Kemi; Daniel Jordan; David Rientjes; Jerome Glisse; Ganesh Mahendran; Minc=\n",
+ "han Kim; Punit Agrawal; vinayak menon; Yang Shi=0A=\n",
+ ">>>> Cc: linux-kernel\@vger.kernel.org; linux-mm\@kvack.org; haren\@linux.vnet=\n",
+ ".ibm.com; npiggin\@gmail.com; bsingharora\@gmail.com; paulmck\@linux.vnet.ibm.=\n",
+ "com; Tim Chen; linuxppc-dev\@lists.ozlabs.org; x86\@kernel.org=0A=\n",
+ ">>>> Subject: [PATCH v11 00/26] Speculative page faults=0A=\n",
+ ">>>>=0A=\n",
+ ">>>> This is a port on kernel 4.17 of the work done by Peter Zijlstra to ha=\n",
+ "ndle=0A=\n",
+ ">>>> page fault without holding the mm semaphore [1].=0A=\n",
+ ">>>>=0A=\n",
+ ">>>> The idea is to try to handle user space page faults without holding th=\n",
+ "e=0A=\n",
+ ">>>> mmap_sem. This should allow better concurrency for massively threaded=\n",
+ "=0A=\n",
+ ">>>> process since the page fault handler will not wait for other threads m=\n",
+ "emory=0A=\n",
+ ">>>> layout change to be done, assuming that this change is done in another=\n",
+ " part=0A=\n",
+ ">>>> of the process's memory space. This type page fault is named speculati=\n",
+ "ve=0A=\n",
+ ">>>> page fault. If the speculative page fault fails because of a concurren=\n",
+ "cy is=0A=\n",
+ ">>>> detected or because underlying PMD or PTE tables are not yet allocatin=\n",
+ "g, it=0A=\n",
+ ">>>> is failing its processing and a classic page fault is then tried.=0A=\n",
+ ">>>>=0A=\n",
+ ">>>> The speculative page fault (SPF) has to look for the VMA matching the =\n",
+ "fault=0A=\n",
+ ">>>> address without holding the mmap_sem, this is done by introducing a rw=\n",
+ "lock=0A=\n",
+ ">>>> which protects the access to the mm_rb tree. Previously this was done =\n",
+ "using=0A=\n",
+ ">>>> SRCU but it was introducing a lot of scheduling to process the VMA's=\n",
+ "=0A=\n",
+ ">>>> freeing operation which was hitting the performance by 20% as reported=\n",
+ " by=0A=\n",
+ ">>>> Kemi Wang [2]. Using a rwlock to protect access to the mm_rb tree is=\n",
+ "=0A=\n",
+ ">>>> limiting the locking contention to these operations which are expected=\n",
+ " to=0A=\n",
+ ">>>> be in a O(log n) order. In addition to ensure that the VMA is not free=\n",
+ "d in=0A=\n",
+ ">>>> our back a reference count is added and 2 services (get_vma() and=0A=\n",
+ ">>>> put_vma()) are introduced to handle the reference count. Once a VMA is=\n",
+ "=0A=\n",
+ ">>>> fetched from the RB tree using get_vma(), it must be later freed using=\n",
+ "=0A=\n",
+ ">>>> put_vma(). I can't see anymore the overhead I got while will-it-scale=\n",
+ "=0A=\n",
+ ">>>> benchmark anymore.=0A=\n",
+ ">>>>=0A=\n",
+ ">>>> The VMA's attributes checked during the speculative page fault process=\n",
+ "ing=0A=\n",
+ ">>>> have to be protected against parallel changes. This is done by using a=\n",
+ " per=0A=\n",
+ ">>>> VMA sequence lock. This sequence lock allows the speculative page faul=\n",
+ "t=0A=\n",
+ ">>>> handler to fast check for parallel changes in progress and to abort th=\n",
+ "e=0A=\n",
+ ">>>> speculative page fault in that case.=0A=\n",
+ ">>>>=0A=\n",
+ ">>>> Once the VMA has been found, the speculative page fault handler would =\n",
+ "check=0A=\n",
+ ">>>> for the VMA's attributes to verify that the page fault has to be handl=\n",
+ "ed=0A=\n",
+ ">>>> correctly or not. Thus, the VMA is protected through a sequence lock w=\n",
+ "hich=0A=\n",
+ ">>>> allows fast detection of concurrent VMA changes. If such a change is=\n",
+ "=0A=\n",
+ ">>>> detected, the speculative page fault is aborted and a *classic* page f=\n",
+ "ault=0A=\n",
+ ">>>> is tried. VMA sequence lockings are added when VMA attributes which a=\n",
+ "re=0A=\n",
+ ">>>> checked during the page fault are modified.=0A=\n",
+ ">>>>=0A=\n",
+ ">>>> When the PTE is fetched, the VMA is checked to see if it has been chan=\n",
+ "ged,=0A=\n",
+ ">>>> so once the page table is locked, the VMA is valid, so any other chang=\n",
+ "es=0A=\n",
+ ">>>> leading to touching this PTE will need to lock the page table, so no=\n",
+ "=0A=\n",
+ ">>>> parallel change is possible at this time.=0A=\n",
+ ">>>>=0A=\n",
+ ">>>> The locking of the PTE is done with interrupts disabled, this allows=\n",
+ "=0A=\n",
+ ">>>> checking for the PMD to ensure that there is not an ongoing collapsing=\n",
+ "=0A=\n",
+ ">>>> operation. Since khugepaged is firstly set the PMD to pmd_none and the=\n",
+ "n is=0A=\n",
+ ">>>> waiting for the other CPU to have caught the IPI interrupt, if the pmd=\n",
+ " is=0A=\n",
+ ">>>> valid at the time the PTE is locked, we have the guarantee that the=0A=\n",
+ ">>>> collapsing operation will have to wait on the PTE lock to move forward=\n",
+ ".=0A=\n",
+ ">>>> This allows the SPF handler to map the PTE safely. If the PMD value is=\n",
+ "=0A=\n",
+ ">>>> different from the one recorded at the beginning of the SPF operation,=\n",
+ " the=0A=\n",
+ ">>>> classic page fault handler will be called to handle the operation whil=\n",
+ "e=0A=\n",
+ ">>>> holding the mmap_sem. As the PTE lock is done with the interrupts disa=\n",
+ "bled,=0A=\n",
+ ">>>> the lock is done using spin_trylock() to avoid dead lock when handling=\n",
+ " a=0A=\n",
+ ">>>> page fault while a TLB invalidate is requested by another CPU holding =\n",
+ "the=0A=\n",
+ ">>>> PTE.=0A=\n",
+ ">>>>=0A=\n",
+ ">>>> In pseudo code, this could be seen as:=0A=\n",
+ ">>>> speculative_page_fault()=0A=\n",
+ ">>>> {=0A=\n",
+ ">>>> vma =3D get_vma()=0A=\n",
+ ">>>> check vma sequence count=0A=\n",
+ ">>>> check vma's support=0A=\n",
+ ">>>> disable interrupt=0A=\n",
+ ">>>> check pgd,p4d,...,pte=0A=\n",
+ ">>>> save pmd and pte in vmf=0A=\n",
+ ">>>> save vma sequence counter in vmf=0A=\n",
+ ">>>> enable interrupt=0A=\n",
+ ">>>> check vma sequence count=0A=\n",
+ ">>>> handle_pte_fault(vma)=0A=\n",
+ ">>>> ..=0A=\n",
+ ">>>> page =3D alloc_page()=0A=\n",
+ ">>>> pte_map_lock()=0A=\n",
+ ">>>> disable interrupt=0A=\n",
+ ">>>> abort if sequence counter has chan=\n",
+ "ged=0A=\n",
+ ">>>> abort if pmd or pte has changed=0A=\n",
+ ">>>> pte map and lock=0A=\n",
+ ">>>> enable interrupt=0A=\n",
+ ">>>> if abort=0A=\n",
+ ">>>> free page=0A=\n",
+ ">>>> abort=0A=\n",
+ ">>>> ...=0A=\n",
+ ">>>> }=0A=\n",
+ ">>>>=0A=\n",
+ ">>>> arch_fault_handler()=0A=\n",
+ ">>>> {=0A=\n",
+ ">>>> if (speculative_page_fault(&vma))=0A=\n",
+ ">>>> goto done=0A=\n",
+ ">>>> again:=0A=\n",
+ ">>>> lock(mmap_sem)=0A=\n",
+ ">>>> vma =3D find_vma();=0A=\n",
+ ">>>> handle_pte_fault(vma);=0A=\n",
+ ">>>> if retry=0A=\n",
+ ">>>> unlock(mmap_sem)=0A=\n",
+ ">>>> goto again;=0A=\n",
+ ">>>> done:=0A=\n",
+ ">>>> handle fault error=0A=\n",
+ ">>>> }=0A=\n",
+ ">>>>=0A=\n",
+ ">>>> Support for THP is not done because when checking for the PMD, we can =\n",
+ "be=0A=\n",
+ ">>>> confused by an in progress collapsing operation done by khugepaged. Th=\n",
+ "e=0A=\n",
+ ">>>> issue is that pmd_none() could be true either if the PMD is not alread=\n",
+ "y=0A=\n",
+ ">>>> populated or if the underlying PTE are in the way to be collapsed. So =\n",
+ "we=0A=\n",
+ ">>>> cannot safely allocate a PMD if pmd_none() is true.=0A=\n",
+ ">>>>=0A=\n",
+ ">>>> This series add a new software performance event named 'speculative-fa=\n",
+ "ults'=0A=\n",
+ ">>>> or 'spf'. It counts the number of successful page fault event handled=\n",
+ "=0A=\n",
+ ">>>> speculatively. When recording 'faults,spf' events, the faults one is=\n",
+ "=0A=\n",
+ ">>>> counting the total number of page fault events while 'spf' is only cou=\n",
+ "nting=0A=\n",
+ ">>>> the part of the faults processed speculatively.=0A=\n",
+ ">>>>=0A=\n",
+ ">>>> There are some trace events introduced by this series. They allow=0A=\n",
+ ">>>> identifying why the page faults were not processed speculatively. This=\n",
+ "=0A=\n",
+ ">>>> doesn't take in account the faults generated by a monothreaded process=\n",
+ "=0A=\n",
+ ">>>> which directly processed while holding the mmap_sem. This trace events=\n",
+ " are=0A=\n",
+ ">>>> grouped in a system named 'pagefault', they are:=0A=\n",
+ ">>>> - pagefault:spf_vma_changed : if the VMA has been changed in our back=\n",
+ "=0A=\n",
+ ">>>> - pagefault:spf_vma_noanon : the vma->anon_vma field was not yet set.=\n",
+ "=0A=\n",
+ ">>>> - pagefault:spf_vma_notsup : the VMA's type is not supported=0A=\n",
+ ">>>> - pagefault:spf_vma_access : the VMA's access right are not respected=\n",
+ "=0A=\n",
+ ">>>> - pagefault:spf_pmd_changed : the upper PMD pointer has changed in ou=\n",
+ "r=0A=\n",
+ ">>>> back.=0A=\n",
+ ">>>>=0A=\n",
+ ">>>> To record all the related events, the easier is to run perf with the=\n",
+ "=0A=\n",
+ ">>>> following arguments :=0A=\n",
+ ">>>> \$ perf stat -e 'faults,spf,pagefault:*' <command>=0A=\n",
+ ">>>>=0A=\n",
+ ">>>> There is also a dedicated vmstat counter showing the number of success=\n",
+ "ful=0A=\n",
+ ">>>> page fault handled speculatively. I can be seen this way:=0A=\n",
+ ">>>> \$ grep speculative_pgfault /proc/vmstat=0A=\n",
+ ">>>>=0A=\n",
+ ">>>> This series builds on top of v4.16-mmotm-2018-04-13-17-28 and is funct=\n",
+ "ional=0A=\n",
+ ">>>> on x86, PowerPC and arm64.=0A=\n",
+ ">>>>=0A=\n",
+ ">>>> ---------------------=0A=\n",
+ ">>>> Real Workload results=0A=\n",
+ ">>>>=0A=\n",
+ ">>>> As mentioned in previous email, we did non official runs using a \"popu=\n",
+ "lar=0A=\n",
+ ">>>> in memory multithreaded database product\" on 176 cores SMT8 Power syst=\n",
+ "em=0A=\n",
+ ">>>> which showed a 30% improvements in the number of transaction processed=\n",
+ " per=0A=\n",
+ ">>>> second. This run has been done on the v6 series, but changes introduce=\n",
+ "d in=0A=\n",
+ ">>>> this new version should not impact the performance boost seen.=0A=\n",
+ ">>>>=0A=\n",
+ ">>>> Here are the perf data captured during 2 of these runs on top of the v=\n",
+ "8=0A=\n",
+ ">>>> series:=0A=\n",
+ ">>>> vanilla spf=0A=\n",
+ ">>>> faults 89.418 101.364 +13%=0A=\n",
+ ">>>> spf n/a 97.989=0A=\n",
+ ">>>>=0A=\n",
+ ">>>> With the SPF kernel, most of the page fault were processed in a specul=\n",
+ "ative=0A=\n",
+ ">>>> way.=0A=\n",
+ ">>>>=0A=\n",
+ ">>>> Ganesh Mahendran had backported the series on top of a 4.9 kernel and =\n",
+ "gave=0A=\n",
+ ">>>> it a try on an android device. He reported that the application launch=\n",
+ " time=0A=\n",
+ ">>>> was improved in average by 6%, and for large applications (~100 thread=\n",
+ "s) by=0A=\n",
+ ">>>> 20%.=0A=\n",
+ ">>>>=0A=\n",
+ ">>>> Here are the launch time Ganesh mesured on Android 8.0 on top of a Qco=\n",
+ "m=0A=\n",
+ ">>>> MSM845 (8 cores) with 6GB (the less is better):=0A=\n",
+ ">>>>=0A=\n",
+ ">>>> Application 4.9 4.9+spf delta=0A=\n",
+ ">>>> com.tencent.mm 416 389 -7%=0A=\n",
+ ">>>> com.eg.android.AlipayGphone 1135 986 -13%=0A=\n",
+ ">>>> com.tencent.mtt 455 454 0%=0A=\n",
+ ">>>> com.qqgame.hlddz 1497 1409 -6%=0A=\n",
+ ">>>> com.autonavi.minimap 711 701 -1%=0A=\n",
+ ">>>> com.tencent.tmgp.sgame 788 748 -5%=0A=\n",
+ ">>>> com.immomo.momo 501 487 -3%=0A=\n",
+ ">>>> com.tencent.peng 2145 2112 -2%=0A=\n",
+ ">>>> com.smile.gifmaker 491 461 -6%=0A=\n",
+ ">>>> com.baidu.BaiduMap 479 366 -23%=0A=\n",
+ ">>>> com.taobao.taobao 1341 1198 -11%=0A=\n",
+ ">>>> com.baidu.searchbox 333 314 -6%=0A=\n",
+ ">>>> com.tencent.mobileqq 394 384 -3%=0A=\n",
+ ">>>> com.sina.weibo 907 906 0%=0A=\n",
+ ">>>> com.youku.phone 816 731 -11%=0A=\n",
+ ">>>> com.happyelements.AndroidAnimal.qq 763 717 -6%=0A=\n",
+ ">>>> com.UCMobile 415 411 -1%=0A=\n",
+ ">>>> com.tencent.tmgp.ak 1464 1431 -2%=0A=\n",
+ ">>>> com.tencent.qqmusic 336 329 -2%=0A=\n",
+ ">>>> com.sankuai.meituan 1661 1302 -22%=0A=\n",
+ ">>>> com.netease.cloudmusic 1193 1200 1%=0A=\n",
+ ">>>> air.tv.douyu.android 4257 4152 -2%=0A=\n",
+ ">>>>=0A=\n",
+ ">>>> ------------------=0A=\n",
+ ">>>> Benchmarks results=0A=\n",
+ ">>>>=0A=\n",
+ ">>>> Base kernel is v4.17.0-rc4-mm1=0A=\n",
+ ">>>> SPF is BASE + this series=0A=\n",
+ ">>>>=0A=\n",
+ ">>>> Kernbench:=0A=\n",
+ ">>>> ----------=0A=\n",
+ ">>>> Here are the results on a 16 CPUs X86 guest using kernbench on a 4.15=\n",
+ "=0A=\n",
+ ">>>> kernel (kernel is build 5 times):=0A=\n",
+ ">>>>=0A=\n",
+ ">>>> Average Half load -j 8=0A=\n",
+ ">>>> Run (std deviation)=0A=\n",
+ ">>>> BASE SPF=0A=\n",
+ ">>>> Elapsed Time 1448.65 (5.72312) 1455.84 (4.84951) 0.50%=\n",
+ "=0A=\n",
+ ">>>> User Time 10135.4 (30.3699) 10148.8 (31.1252) 0.13%=\n",
+ "=0A=\n",
+ ">>>> System Time 900.47 (2.81131) 923.28 (7.52779) 2.53%=\n",
+ "=0A=\n",
+ ">>>> Percent CPU 761.4 (1.14018) 760.2 (0.447214) -0.16%=\n",
+ "=0A=\n",
+ ">>>> Context Switches 85380 (3419.52) 84748 (1904.44) -0.74%=\n",
+ "=0A=\n",
+ ">>>> Sleeps 105064 (1240.96) 105074 (337.612) 0.01%=\n",
+ "=0A=\n",
+ ">>>>=0A=\n",
+ ">>>> Average Optimal load -j 16=0A=\n",
+ ">>>> Run (std deviation)=0A=\n",
+ ">>>> BASE SPF=0A=\n",
+ ">>>> Elapsed Time 920.528 (10.1212) 927.404 (8.91789) 0.75%=\n",
+ "=0A=\n",
+ ">>>> User Time 11064.8 (981.142) 11085 (990.897) 0.18%=\n",
+ "=0A=\n",
+ ">>>> System Time 979.904 (84.0615) 1001.14 (82.5523) 2.17%=\n",
+ "=0A=\n",
+ ">>>> Percent CPU 1089.5 (345.894) 1086.1 (343.545) -0.31%=\n",
+ "=0A=\n",
+ ">>>> Context Switches 159488 (78156.4) 158223 (77472.1) -0.79%=\n",
+ "=0A=\n",
+ ">>>> Sleeps 110566 (5877.49) 110388 (5617.75) -0.16%=\n",
+ "=0A=\n",
+ ">>>>=0A=\n",
+ ">>>>=0A=\n",
+ ">>>> During a run on the SPF, perf events were captured:=0A=\n",
+ ">>>> Performance counter stats for '../kernbench -M':=0A=\n",
+ ">>>> 526743764 faults=0A=\n",
+ ">>>> 210 spf=0A=\n",
+ ">>>> 3 pagefault:spf_vma_changed=0A=\n",
+ ">>>> 0 pagefault:spf_vma_noanon=0A=\n",
+ ">>>> 2278 pagefault:spf_vma_notsup=0A=\n",
+ ">>>> 0 pagefault:spf_vma_access=0A=\n",
+ ">>>> 0 pagefault:spf_pmd_changed=0A=\n",
+ ">>>>=0A=\n",
+ ">>>> Very few speculative page faults were recorded as most of the processe=\n",
+ "s=0A=\n",
+ ">>>> involved are monothreaded (sounds that on this architecture some threa=\n",
+ "ds=0A=\n",
+ ">>>> were created during the kernel build processing).=0A=\n",
+ ">>>>=0A=\n",
+ ">>>> Here are the kerbench results on a 80 CPUs Power8 system:=0A=\n",
+ ">>>>=0A=\n",
+ ">>>> Average Half load -j 40=0A=\n",
+ ">>>> Run (std deviation)=0A=\n",
+ ">>>> BASE SPF=0A=\n",
+ ">>>> Elapsed Time 117.152 (0.774642) 117.166 (0.476057) 0.01%=\n",
+ "=0A=\n",
+ ">>>> User Time 4478.52 (24.7688) 4479.76 (9.08555) 0.03%=\n",
+ "=0A=\n",
+ ">>>> System Time 131.104 (0.720056) 134.04 (0.708414) 2.24%=\n",
+ "=0A=\n",
+ ">>>> Percent CPU 3934 (19.7104) 3937.2 (19.0184) 0.08%=\n",
+ "=0A=\n",
+ ">>>> Context Switches 92125.4 (576.787) 92581.6 (198.622) 0.50%=\n",
+ "=0A=\n",
+ ">>>> Sleeps 317923 (652.499) 318469 (1255.59) 0.17%=\n",
+ "=0A=\n",
+ ">>>>=0A=\n",
+ ">>>> Average Optimal load -j 80=0A=\n",
+ ">>>> Run (std deviation)=0A=\n",
+ ">>>> BASE SPF=0A=\n",
+ ">>>> Elapsed Time 107.73 (0.632416) 107.31 (0.584936) -0.39%=\n",
+ "=0A=\n",
+ ">>>> User Time 5869.86 (1466.72) 5871.71 (1467.27) 0.03%=\n",
+ "=0A=\n",
+ ">>>> System Time 153.728 (23.8573) 157.153 (24.3704) 2.23%=\n",
+ "=0A=\n",
+ ">>>> Percent CPU 5418.6 (1565.17) 5436.7 (1580.91) 0.33%=\n",
+ "=0A=\n",
+ ">>>> Context Switches 223861 (138865) 225032 (139632) 0.52%=\n",
+ "=0A=\n",
+ ">>>> Sleeps 330529 (13495.1) 332001 (14746.2) 0.45%=\n",
+ "=0A=\n",
+ ">>>>=0A=\n",
+ ">>>> During a run on the SPF, perf events were captured:=0A=\n",
+ ">>>> Performance counter stats for '../kernbench -M':=0A=\n",
+ ">>>> 116730856 faults=0A=\n",
+ ">>>> 0 spf=0A=\n",
+ ">>>> 3 pagefault:spf_vma_changed=0A=\n",
+ ">>>> 0 pagefault:spf_vma_noanon=0A=\n",
+ ">>>> 476 pagefault:spf_vma_notsup=0A=\n",
+ ">>>> 0 pagefault:spf_vma_access=0A=\n",
+ ">>>> 0 pagefault:spf_pmd_changed=0A=\n",
+ ">>>>=0A=\n",
+ ">>>> Most of the processes involved are monothreaded so SPF is not activate=\n",
+ "d but=0A=\n",
+ ">>>> there is no impact on the performance.=0A=\n",
+ ">>>>=0A=\n",
+ ">>>> Ebizzy:=0A=\n",
+ ">>>> -------=0A=\n",
+ ">>>> The test is counting the number of records per second it can manage, t=\n",
+ "he=0A=\n",
+ ">>>> higher is the best. I run it like this 'ebizzy -mTt <nrcpus>'. To get=\n",
+ "=0A=\n",
+ ">>>> consistent result I repeated the test 100 times and measure the averag=\n",
+ "e=0A=\n",
+ ">>>> result. The number is the record processes per second, the higher is t=\n",
+ "he=0A=\n",
+ ">>>> best.=0A=\n",
+ ">>>>=0A=\n",
+ ">>>> BASE SPF delta=0A=\n",
+ ">>>> 16 CPUs x86 VM 742.57 1490.24 100.69%=0A=\n",
+ ">>>> 80 CPUs P8 node 13105.4 24174.23 84.46%=0A=\n",
+ ">>>>=0A=\n",
+ ">>>> Here are the performance counter read during a run on a 16 CPUs x86 VM=\n",
+ ":=0A=\n",
+ ">>>> Performance counter stats for './ebizzy -mTt 16':=0A=\n",
+ ">>>> 1706379 faults=0A=\n",
+ ">>>> 1674599 spf=0A=\n",
+ ">>>> 30588 pagefault:spf_vma_changed=0A=\n",
+ ">>>> 0 pagefault:spf_vma_noanon=0A=\n",
+ ">>>> 363 pagefault:spf_vma_notsup=0A=\n",
+ ">>>> 0 pagefault:spf_vma_access=0A=\n",
+ ">>>> 0 pagefault:spf_pmd_changed=0A=\n",
+ ">>>>=0A=\n",
+ ">>>> And the ones captured during a run on a 80 CPUs Power node:=0A=\n",
+ ">>>> Performance counter stats for './ebizzy -mTt 80':=0A=\n",
+ ">>>> 1874773 faults=0A=\n",
+ ">>>> 1461153 spf=0A=\n",
+ ">>>> 413293 pagefault:spf_vma_changed=0A=\n",
+ ">>>> 0 pagefault:spf_vma_noanon=0A=\n",
+ ">>>> 200 pagefault:spf_vma_notsup=0A=\n",
+ ">>>> 0 pagefault:spf_vma_access=0A=\n",
+ ">>>> 0 pagefault:spf_pmd_changed=0A=\n",
+ ">>>>=0A=\n",
+ ">>>> In ebizzy's case most of the page fault were handled in a speculative =\n",
+ "way,=0A=\n",
+ ">>>> leading the ebizzy performance boost.=0A=\n",
+ ">>>>=0A=\n",
+ ">>>> ------------------=0A=\n",
+ ">>>> Changes since v10 (https://lkml.org/lkml/2018/4/17/572):=0A=\n",
+ ">>>> - Accounted for all review feedbacks from Punit Agrawal, Ganesh Mahen=\n",
+ "dran=0A=\n",
+ ">>>> and Minchan Kim, hopefully.=0A=\n",
+ ">>>> - Remove unneeded check on CONFIG_SPECULATIVE_PAGE_FAULT in=0A=\n",
+ ">>>> __do_page_fault().=0A=\n",
+ ">>>> - Loop in pte_spinlock() and pte_map_lock() when pte try lock fails=\n",
+ "=0A=\n",
+ ">>>> instead=0A=\n",
+ ">>>> of aborting the speculative page fault handling. Dropping the now=\n",
+ "=0A=\n",
+ ">>>> useless=0A=\n",
+ ">>>> trace event pagefault:spf_pte_lock.=0A=\n",
+ ">>>> - No more try to reuse the fetched VMA during the speculative page fa=\n",
+ "ult=0A=\n",
+ ">>>> handling when retrying is needed. This adds a lot of complexity and=\n",
+ "=0A=\n",
+ ">>>> additional tests done didn't show a significant performance improve=\n",
+ "ment.=0A=\n",
+ ">>>> - Convert IS_ENABLED(CONFIG_NUMA) back to #ifdef due to build error.=\n",
+ "=0A=\n",
+ ">>>>=0A=\n",
+ ">>>> [1] http://linux-kernel.2935.n7.nabble.com/RFC-PATCH-0-6-Another-go-at=\n",
+ "-speculative-page-faults-tt965642.html#none=0A=\n",
+ ">>>> [2] https://patchwork.kernel.org/patch/9999687/=0A=\n",
+ ">>>>=0A=\n",
+ ">>>>=0A=\n",
+ ">>>> Laurent Dufour (20):=0A=\n",
+ ">>>> mm: introduce CONFIG_SPECULATIVE_PAGE_FAULT=0A=\n",
+ ">>>> x86/mm: define ARCH_SUPPORTS_SPECULATIVE_PAGE_FAULT=0A=\n",
+ ">>>> powerpc/mm: set ARCH_SUPPORTS_SPECULATIVE_PAGE_FAULT=0A=\n",
+ ">>>> mm: introduce pte_spinlock for FAULT_FLAG_SPECULATIVE=0A=\n",
+ ">>>> mm: make pte_unmap_same compatible with SPF=0A=\n",
+ ">>>> mm: introduce INIT_VMA()=0A=\n",
+ ">>>> mm: protect VMA modifications using VMA sequence count=0A=\n",
+ ">>>> mm: protect mremap() against SPF hanlder=0A=\n",
+ ">>>> mm: protect SPF handler against anon_vma changes=0A=\n",
+ ">>>> mm: cache some VMA fields in the vm_fault structure=0A=\n",
+ ">>>> mm/migrate: Pass vm_fault pointer to migrate_misplaced_page()=0A=\n",
+ ">>>> mm: introduce __lru_cache_add_active_or_unevictable=0A=\n",
+ ">>>> mm: introduce __vm_normal_page()=0A=\n",
+ ">>>> mm: introduce __page_add_new_anon_rmap()=0A=\n",
+ ">>>> mm: protect mm_rb tree with a rwlock=0A=\n",
+ ">>>> mm: adding speculative page fault failure trace events=0A=\n",
+ ">>>> perf: add a speculative page fault sw event=0A=\n",
+ ">>>> perf tools: add support for the SPF perf event=0A=\n",
+ ">>>> mm: add speculative page fault vmstats=0A=\n",
+ ">>>> powerpc/mm: add speculative page fault=0A=\n",
+ ">>>>=0A=\n",
+ ">>>> Mahendran Ganesh (2):=0A=\n",
+ ">>>> arm64/mm: define ARCH_SUPPORTS_SPECULATIVE_PAGE_FAULT=0A=\n",
+ ">>>> arm64/mm: add speculative page fault=0A=\n",
+ ">>>>=0A=\n",
+ ">>>> Peter Zijlstra (4):=0A=\n",
+ ">>>> mm: prepare for FAULT_FLAG_SPECULATIVE=0A=\n",
+ ">>>> mm: VMA sequence count=0A=\n",
+ ">>>> mm: provide speculative fault infrastructure=0A=\n",
+ ">>>> x86/mm: add speculative pagefault handling=0A=\n",
+ ">>>>=0A=\n",
+ ">>>> arch/arm64/Kconfig | 1 +=0A=\n",
+ ">>>> arch/arm64/mm/fault.c | 12 +=0A=\n",
+ ">>>> arch/powerpc/Kconfig | 1 +=0A=\n",
+ ">>>> arch/powerpc/mm/fault.c | 16 +=0A=\n",
+ ">>>> arch/x86/Kconfig | 1 +=0A=\n",
+ ">>>> arch/x86/mm/fault.c | 27 +-=0A=\n",
+ ">>>> fs/exec.c | 2 +-=0A=\n",
+ ">>>> fs/proc/task_mmu.c | 5 +-=0A=\n",
+ ">>>> fs/userfaultfd.c | 17 +-=0A=\n",
+ ">>>> include/linux/hugetlb_inline.h | 2 +-=0A=\n",
+ ">>>> include/linux/migrate.h | 4 +-=0A=\n",
+ ">>>> include/linux/mm.h | 136 +++++++-=0A=\n",
+ ">>>> include/linux/mm_types.h | 7 +=0A=\n",
+ ">>>> include/linux/pagemap.h | 4 +-=0A=\n",
+ ">>>> include/linux/rmap.h | 12 +-=0A=\n",
+ ">>>> include/linux/swap.h | 10 +-=0A=\n",
+ ">>>> include/linux/vm_event_item.h | 3 +=0A=\n",
+ ">>>> include/trace/events/pagefault.h | 80 +++++=0A=\n",
+ ">>>> include/uapi/linux/perf_event.h | 1 +=0A=\n",
+ ">>>> kernel/fork.c | 5 +-=0A=\n",
+ ">>>> mm/Kconfig | 22 ++=0A=\n",
+ ">>>> mm/huge_memory.c | 6 +-=0A=\n",
+ ">>>> mm/hugetlb.c | 2 +=0A=\n",
+ ">>>> mm/init-mm.c | 3 +=0A=\n",
+ ">>>> mm/internal.h | 20 ++=0A=\n",
+ ">>>> mm/khugepaged.c | 5 +=0A=\n",
+ ">>>> mm/madvise.c | 6 +-=0A=\n",
+ ">>>> mm/memory.c | 612 +++++++++++++++++++++++++=\n",
+ "++++-----=0A=\n",
+ ">>>> mm/mempolicy.c | 51 ++-=0A=\n",
+ ">>>> mm/migrate.c | 6 +-=0A=\n",
+ ">>>> mm/mlock.c | 13 +-=0A=\n",
+ ">>>> mm/mmap.c | 229 ++++++++++---=0A=\n",
+ ">>>> mm/mprotect.c | 4 +-=0A=\n",
+ ">>>> mm/mremap.c | 13 +=0A=\n",
+ ">>>> mm/nommu.c | 2 +-=0A=\n",
+ ">>>> mm/rmap.c | 5 +-=0A=\n",
+ ">>>> mm/swap.c | 6 +-=0A=\n",
+ ">>>> mm/swap_state.c | 8 +-=0A=\n",
+ ">>>> mm/vmstat.c | 5 +-=0A=\n",
+ ">>>> tools/include/uapi/linux/perf_event.h | 1 +=0A=\n",
+ ">>>> tools/perf/util/evsel.c | 1 +=0A=\n",
+ ">>>> tools/perf/util/parse-events.c | 4 +=0A=\n",
+ ">>>> tools/perf/util/parse-events.l | 1 +=0A=\n",
+ ">>>> tools/perf/util/python.c | 1 +=0A=\n",
+ ">>>> 44 files changed, 1161 insertions(+), 211 deletions(-)=0A=\n",
+ ">>>> create mode 100644 include/trace/events/pagefault.h=0A=\n",
+ ">>>>=0A=\n",
+ ">>>> --=0A=\n",
+ ">>>> 2.7.4=0A=\n",
+ ">>>>=0A=\n",
+ ">>>>=0A=\n",
+ ">>>=0A=\n",
+ ">>=0A=\n",
+ ">=0A=\n",
+ "=0A="
]
-a20814b647c614b36a938ba0dabb3d7fc985be6f043bf93ba9ce31c5cbfc3db2
+3fe57aa64e67bec846efb9e40f093021b4adbcaceb036655e3916a57dd15f081
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.