Date:   Tue, 15 Jun 2021 18:39:39 +0800
From:   Ian Kent <raven@...maw.net>
To:     Greg Kroah-Hartman <gregkh@...uxfoundation.org>,
        Tejun Heo <tj@...nel.org>
Cc:     Eric Sandeen <sandeen@...deen.net>, Fox Chen <foxhlchen@...il.com>,
        Brice Goglin <brice.goglin@...il.com>,
        Al Viro <viro@...IV.linux.org.uk>,
        Rick Lindsley <ricklind@...ux.vnet.ibm.com>,
        David Howells <dhowells@...hat.com>,
        Miklos Szeredi <miklos@...redi.hu>,
        Marcelo Tosatti <mtosatti@...hat.com>,
        "Eric W. Biederman" <ebiederm@...ssion.com>,
        Carlos Maiolino <cmaiolino@...hat.com>,
        linux-fsdevel <linux-fsdevel@...r.kernel.org>,
        Kernel Mailing List <linux-kernel@...r.kernel.org>
Subject: Re: [PATCH v7 0/6] kernfs: proposed locking and concurrency
 improvement

Hi,

Here are benchmark test runs against patch set v7.
The locking change has made a bit of a defference but the results are
not much different to previous runs.

1) Run with 43 distinct sysfs file paths on 86 cpu machine, perf graphs
in base-files-cpu-86-perf.txt and patched-files-cpu-86-perf.txt.

Base (5.12.2, unpatched)
------------------------
single: total 37.537578ms per 3.753758us
concur: total 4796.665718ms per 479.666572us  CPU 65
concur: total 4800.220672ms per 480.022067us  CPU 22
concur: total 4828.959645ms per 482.895964us  CPU 43
concur: total 4848.001744ms per 484.800174us  CPU 0
concur: total 4859.682650ms per 485.968265us  CPU 31
concur: total 4865.884472ms per 486.588447us  CPU 67
concur: total 4876.658191ms per 487.665819us  CPU 30
concur: total 4877.914506ms per 487.791451us  CPU 72
concur: total 4880.408600ms per 488.040860us  CPU 24
concur: total 4884.697273ms per 488.469727us  CPU 29
concur: total 4886.702702ms per 488.670270us  CPU 34
concur: total 4887.505107ms per 488.750511us  CPU 25
concur: total 4889.623774ms per 488.962377us  CPU 69
concur: total 4894.103764ms per 489.410376us  CPU 23
concur: total 4903.002239ms per 490.300224us  CPU 81
concur: total 4906.873196ms per 490.687320us  CPU 82
concur: total 4907.692111ms per 490.769211us  CPU 68
concur: total 4908.913453ms per 490.891345us  CPU 74
concur: total 4909.243539ms per 490.924354us  CPU 73
concur: total 4916.646302ms per 491.664630us  CPU 78
concur: total 4916.962985ms per 491.696299us  CPU 27
concur: total 4917.448436ms per 491.744844us  CPU 75
concur: total 4920.916201ms per 492.091620us  CPU 70
concur: total 4922.719147ms per 492.271915us  CPU 84
concur: total 4922.971086ms per 492.297109us  CPU 26
concur: total 4923.825391ms per 492.382539us  CPU 64
concur: total 4924.360433ms per 492.436043us  CPU 80
concur: total 4927.506957ms per 492.750696us  CPU 71
concur: total 4928.302267ms per 492.830227us  CPU 85
concur: total 4928.619595ms per 492.861959us  CPU 38
concur: total 4929.667220ms per 492.966722us  CPU 41
concur: total 4929.783827ms per 492.978383us  CPU 39
concur: total 4929.820893ms per 492.982089us  CPU 77
concur: total 4929.869480ms per 492.986948us  CPU 36
concur: total 4931.021352ms per 493.102135us  CPU 76
concur: total 4931.837530ms per 493.183753us  CPU 28
concur: total 4932.697037ms per 493.269704us  CPU 5
concur: total 4932.854097ms per 493.285410us  CPU 35
concur: total 4933.293678ms per 493.329368us  CPU 49
concur: total 4933.634401ms per 493.363440us  CPU 79
concur: total 4933.826684ms per 493.382668us  CPU 40
concur: total 4933.903000ms per 493.390300us  CPU 51
concur: total 4935.303136ms per 493.530314us  CPU 21
concur: total 4935.369630ms per 493.536963us  CPU 44
concur: total 4935.510909ms per 493.551091us  CPU 32
concur: total 4935.723570ms per 493.572357us  CPU 33
concur: total 4936.612243ms per 493.661224us  CPU 42
concur: total 4936.901434ms per 493.690143us  CPU 37
concur: total 4937.042596ms per 493.704260us  CPU 48
concur: total 4937.138242ms per 493.713824us  CPU 83
concur: total 4937.141487ms per 493.714149us  CPU 47
concur: total 4937.326596ms per 493.732660us  CPU 1
concur: total 4938.003521ms per 493.800352us  CPU 61
concur: total 4939.302626ms per 493.930263us  CPU 8
concur: total 4939.499698ms per 493.949970us  CPU 18
concur: total 4940.621054ms per 494.062105us  CPU 9
concur: total 4941.746319ms per 494.174632us  CPU 50
concur: total 4941.781419ms per 494.178142us  CPU 6
concur: total 4943.453629ms per 494.345363us  CPU 45
concur: total 4943.543980ms per 494.354398us  CPU 59
concur: total 4943.976473ms per 494.397647us  CPU 52
concur: total 4944.397692ms per 494.439769us  CPU 3
concur: total 4944.411995ms per 494.441200us  CPU 7
concur: total 4944.431336ms per 494.443134us  CPU 4
concur: total 4944.968336ms per 494.496834us  CPU 46
concur: total 4945.390966ms per 494.539097us  CPU 17
concur: total 4948.021887ms per 494.802189us  CPU 10
concur: total 4948.039212ms per 494.803921us  CPU 2
concur: total 4948.123469ms per 494.812347us  CPU 16
concur: total 4948.180674ms per 494.818067us  CPU 60
concur: total 4948.294916ms per 494.829492us  CPU 20
concur: total 4948.585694ms per 494.858569us  CPU 54
concur: total 4948.905652ms per 494.890565us  CPU 55
concur: total 4948.958036ms per 494.895804us  CPU 53
concur: total 4949.346910ms per 494.934691us  CPU 15
concur: total 4949.456923ms per 494.945692us  CPU 12
concur: total 4949.469902ms per 494.946990us  CPU 19
concur: total 4949.518404ms per 494.951840us  CPU 57
concur: total 4949.541110ms per 494.954111us  CPU 63
concur: total 4949.622395ms per 494.962239us  CPU 14
concur: total 4949.718346ms per 494.971835us  CPU 56
concur: total 4950.237632ms per 495.023763us  CPU 58
concur: total 4950.356322ms per 495.035632us  CPU 13
concur: total 4950.452444ms per 495.045244us  CPU 62
concur: total 4951.023635ms per 495.102363us  CPU 11
concur: total 4954.519297ms per 495.451930us  CPU 66
times: 10000 threads: 86 cpus: 86

patched (5.12.2)
----------------
single: total 40.276214ms per 4.027621us
concur: total 1688.316678ms per 168.831668us  CPU 62
concur: total 1689.238692ms per 168.923869us  CPU 10
concur: total 1690.603833ms per 169.060383us  CPU 12
concur: total 1691.032655ms per 169.103265us  CPU 61
concur: total 1691.472857ms per 169.147286us  CPU 16
concur: total 1693.003579ms per 169.300358us  CPU 58
concur: total 1694.125776ms per 169.412578us  CPU 3
concur: total 1694.490338ms per 169.449034us  CPU 46
concur: total 1694.602632ms per 169.460263us  CPU 56
concur: total 1695.086541ms per 169.508654us  CPU 44
concur: total 1695.554273ms per 169.555427us  CPU 59
concur: total 1695.667461ms per 169.566746us  CPU 19
concur: total 1695.690350ms per 169.569035us  CPU 57
concur: total 1696.535623ms per 169.653562us  CPU 60
concur: total 1696.690512ms per 169.669051us  CPU 14
concur: total 1697.387634ms per 169.738763us  CPU 18
concur: total 1698.343510ms per 169.834351us  CPU 13
concur: total 1698.681114ms per 169.868111us  CPU 6
concur: total 1698.836863ms per 169.883686us  CPU 55
concur: total 1699.167141ms per 169.916714us  CPU 15
concur: total 1699.676994ms per 169.967699us  CPU 43
concur: total 1700.591012ms per 170.059101us  CPU 2
concur: total 1700.934749ms per 170.093475us  CPU 4
concur: total 1701.024945ms per 170.102495us  CPU 51
concur: total 1701.226732ms per 170.122673us  CPU 45
concur: total 1701.660419ms per 170.166042us  CPU 49
concur: total 1701.678262ms per 170.167826us  CPU 11
concur: total 1701.686480ms per 170.168648us  CPU 25
concur: total 1701.844570ms per 170.184457us  CPU 24
concur: total 1702.128359ms per 170.212836us  CPU 21
concur: total 1702.610757ms per 170.261076us  CPU 17
concur: total 1702.619791ms per 170.261979us  CPU 48
concur: total 1702.791868ms per 170.279187us  CPU 50
concur: total 1703.042799ms per 170.304280us  CPU 68
concur: total 1703.183104ms per 170.318310us  CPU 53
concur: total 1703.507021ms per 170.350702us  CPU 54
concur: total 1704.226830ms per 170.422683us  CPU 7
concur: total 1704.812777ms per 170.481278us  CPU 66
concur: total 1704.930442ms per 170.493044us  CPU 47
concur: total 1705.165520ms per 170.516552us  CPU 20
concur: total 1705.177711ms per 170.517771us  CPU 78
concur: total 1705.419976ms per 170.541998us  CPU 1
concur: total 1705.803125ms per 170.580312us  CPU 83
concur: total 1705.833619ms per 170.583362us  CPU 76
concur: total 1705.890534ms per 170.589053us  CPU 77
concur: total 1706.080503ms per 170.608050us  CPU 0
concur: total 1706.463447ms per 170.646345us  CPU 5
concur: total 1706.510373ms per 170.651037us  CPU 67
concur: total 1706.530695ms per 170.653070us  CPU 39
concur: total 1706.567113ms per 170.656711us  CPU 85
concur: total 1706.574973ms per 170.657497us  CPU 63
concur: total 1707.226897ms per 170.722690us  CPU 22
concur: total 1707.361517ms per 170.736152us  CPU 8
concur: total 1707.628961ms per 170.762896us  CPU 72
concur: total 1707.759502ms per 170.775950us  CPU 27
concur: total 1707.896622ms per 170.789662us  CPU 73
concur: total 1707.955414ms per 170.795541us  CPU 41
concur: total 1708.020238ms per 170.802024us  CPU 37
concur: total 1708.212862ms per 170.821286us  CPU 36
concur: total 1708.525288ms per 170.852529us  CPU 23
concur: total 1708.600291ms per 170.860029us  CPU 79
concur: total 1708.701518ms per 170.870152us  CPU 29
concur: total 1708.810124ms per 170.881012us  CPU 32
concur: total 1708.838772ms per 170.883877us  CPU 84
concur: total 1708.879647ms per 170.887965us  CPU 69
concur: total 1708.972612ms per 170.897261us  CPU 40
concur: total 1708.985762ms per 170.898576us  CPU 71
concur: total 1709.034237ms per 170.903424us  CPU 65
concur: total 1709.074527ms per 170.907453us  CPU 38
concur: total 1709.142994ms per 170.914299us  CPU 28
concur: total 1709.160142ms per 170.916014us  CPU 34
concur: total 1709.204206ms per 170.920421us  CPU 26
concur: total 1709.228029ms per 170.922803us  CPU 33
concur: total 1709.248037ms per 170.924804us  CPU 42
concur: total 1709.326463ms per 170.932646us  CPU 70
concur: total 1709.542704ms per 170.954270us  CPU 35
concur: total 1709.920521ms per 170.992052us  CPU 30
concur: total 1709.924509ms per 170.992451us  CPU 80
concur: total 1712.591681ms per 171.259168us  CPU 82
concur: total 1712.606558ms per 171.260656us  CPU 64
concur: total 1712.649220ms per 171.264922us  CPU 75
concur: total 1715.379022ms per 171.537902us  CPU 52
concur: total 1715.620116ms per 171.562012us  CPU 9
concur: total 1722.125475ms per 172.212547us  CPU 31
concur: total 1723.295082ms per 172.329508us  CPU 74
concur: total 1723.751586ms per 172.375159us  CPU 81
times: 10000 threads: 86 cpus: 86

2) Run with a single sysfs file path on 86 cpu machine, perf graphs in
base-missing-cpu-86-perf.txt and patched-missing-cpu-86-perf.txt.

Base (5.12.2, unpatched)
------------------------
single: total 25.602588ms per 2.560259us
concur: total 7671.072803ms per 767.107280us  CPU 22
concur: total 7676.655038ms per 767.665504us  CPU 73
concur: total 7694.228695ms per 769.422870us  CPU 23
concur: total 7804.388342ms per 780.438834us  CPU 71
concur: total 7809.647276ms per 780.964728us  CPU 70
concur: total 7902.296857ms per 790.229686us  CPU 72
concur: total 7965.755246ms per 796.575525us  CPU 24
concur: total 7994.052914ms per 799.405291us  CPU 30
concur: total 8012.053435ms per 801.205343us  CPU 65
concur: total 8046.237160ms per 804.623716us  CPU 66
concur: total 8166.590955ms per 816.659096us  CPU 28
concur: total 8188.160332ms per 818.816033us  CPU 68
concur: total 8223.517444ms per 822.351744us  CPU 29
concur: total 8316.801754ms per 831.680175us  CPU 27
concur: total 8339.528925ms per 833.952892us  CPU 21
concur: total 8356.378000ms per 835.637800us  CPU 64
concur: total 8442.124512ms per 844.212451us  CPU 83
concur: total 8601.109595ms per 860.110960us  CPU 31
concur: total 8716.851975ms per 871.685197us  CPU 67
concur: total 8719.306654ms per 871.930665us  CPU 25
concur: total 8745.745263ms per 874.574526us  CPU 40
concur: total 8781.849696ms per 878.184970us  CPU 74
concur: total 8966.314242ms per 896.631424us  CPU 69
concur: total 8974.189988ms per 897.418999us  CPU 34
concur: total 9205.936603ms per 920.593660us  CPU 77
concur: total 9209.980188ms per 920.998019us  CPU 26
concur: total 9312.050816ms per 931.205082us  CPU 39
concur: total 9337.596881ms per 933.759688us  CPU 80
concur: total 9367.971342ms per 936.797134us  CPU 82
concur: total 9425.680953ms per 942.568095us  CPU 76
concur: total 9445.991013ms per 944.599101us  CPU 81
concur: total 9494.789554ms per 949.478955us  CPU 85
concur: total 9497.486906ms per 949.748691us  CPU 78
concur: total 9575.471616ms per 957.547162us  CPU 84
concur: total 9583.711012ms per 958.371101us  CPU 41
concur: total 9603.143505ms per 960.314351us  CPU 37
concur: total 9659.819786ms per 965.981979us  CPU 33
concur: total 9660.746579ms per 966.074658us  CPU 79
concur: total 9672.471465ms per 967.247146us  CPU 32
concur: total 9673.978144ms per 967.397814us  CPU 75
concur: total 9690.083225ms per 969.008322us  CPU 38
concur: total 9718.957048ms per 971.895705us  CPU 42
concur: total 9729.803443ms per 972.980344us  CPU 35
concur: total 9736.271248ms per 973.627125us  CPU 36
concur: total 9846.191441ms per 984.619144us  CPU 6
concur: total 9858.151722ms per 985.815172us  CPU 51
concur: total 9871.108661ms per 987.110866us  CPU 45
concur: total 9876.793398ms per 987.679340us  CPU 44
concur: total 9877.221388ms per 987.722139us  CPU 43
concur: total 9881.354352ms per 988.135435us  CPU 7
concur: total 9883.292263ms per 988.329226us  CPU 48
concur: total 9893.701193ms per 989.370119us  CPU 1
concur: total 9893.797272ms per 989.379727us  CPU 50
concur: total 9893.950988ms per 989.395099us  CPU 2
concur: total 9898.109350ms per 989.810935us  CPU 8
concur: total 9899.264544ms per 989.926454us  CPU 49
concur: total 9914.472096ms per 991.447210us  CPU 61
concur: total 9918.795033ms per 991.879503us  CPU 0
concur: total 9929.303538ms per 992.930354us  CPU 3
concur: total 9930.259153ms per 993.025915us  CPU 9
concur: total 9931.281750ms per 993.128175us  CPU 46
concur: total 9933.713011ms per 993.371301us  CPU 5
concur: total 9939.194669ms per 993.919467us  CPU 18
concur: total 9943.118290ms per 994.311829us  CPU 52
concur: total 9943.595571ms per 994.359557us  CPU 47
concur: total 9945.999798ms per 994.599980us  CPU 55
concur: total 9951.944744ms per 995.194474us  CPU 12
concur: total 9959.544673ms per 995.954467us  CPU 54
concur: total 9963.678139ms per 996.367814us  CPU 62
concur: total 9963.679840ms per 996.367984us  CPU 4
concur: total 9963.968986ms per 996.396899us  CPU 53
concur: total 9964.919007ms per 996.491901us  CPU 60
concur: total 9971.413697ms per 997.141370us  CPU 59
concur: total 9973.607985ms per 997.360798us  CPU 63
concur: total 9974.159664ms per 997.415966us  CPU 16
concur: total 9974.520640ms per 997.452064us  CPU 14
concur: total 9974.931335ms per 997.493134us  CPU 17
concur: total 9974.962836ms per 997.496284us  CPU 57
concur: total 9975.494649ms per 997.549465us  CPU 11
concur: total 9976.683023ms per 997.668302us  CPU 10
concur: total 9976.757297ms per 997.675730us  CPU 56
concur: total 9977.161874ms per 997.716187us  CPU 19
concur: total 9977.290610ms per 997.729061us  CPU 20
concur: total 9978.145954ms per 997.814595us  CPU 13
concur: total 9978.815314ms per 997.881531us  CPU 58
concur: total 9979.288565ms per 997.928856us  CPU 15
times: 10000 threads: 86 cpus: 86

patched (5.12.2)
----------------
single: total 19.675378ms per 1.967538us
concur: total 1267.301099ms per 126.730110us  CPU 57
concur: total 1267.676731ms per 126.767673us  CPU 58
concur: total 1268.688528ms per 126.868853us  CPU 14
concur: total 1269.572284ms per 126.957228us  CPU 56
concur: total 1273.534346ms per 127.353435us  CPU 13
concur: total 1273.615124ms per 127.361512us  CPU 15
concur: total 1275.602230ms per 127.560223us  CPU 62
concur: total 1275.658943ms per 127.565894us  CPU 19
concur: total 1275.789693ms per 127.578969us  CPU 20
concur: total 1276.271081ms per 127.627108us  CPU 63
concur: total 1277.278733ms per 127.727873us  CPU 61
concur: total 1278.056971ms per 127.805697us  CPU 17
concur: total 1278.083690ms per 127.808369us  CPU 18
concur: total 1278.446545ms per 127.844654us  CPU 16
concur: total 1279.284518ms per 127.928452us  CPU 60
concur: total 1279.931614ms per 127.993161us  CPU 59
concur: total 1282.073914ms per 128.207391us  CPU 12
concur: total 1282.711558ms per 128.271156us  CPU 55
concur: total 1283.261985ms per 128.326199us  CPU 1
concur: total 1283.325370ms per 128.332537us  CPU 53
concur: total 1283.571147ms per 128.357115us  CPU 10
concur: total 1284.082425ms per 128.408243us  CPU 44
concur: total 1284.714728ms per 128.471473us  CPU 11
concur: total 1284.829587ms per 128.482959us  CPU 54
concur: total 1288.556189ms per 128.855619us  CPU 45
concur: total 1289.402348ms per 128.940235us  CPU 47
concur: total 1289.483215ms per 128.948322us  CPU 50
concur: total 1289.517709ms per 128.951771us  CPU 7
concur: total 1289.576335ms per 128.957633us  CPU 2
concur: total 1289.865537ms per 128.986554us  CPU 49
concur: total 1290.402884ms per 129.040288us  CPU 4
concur: total 1290.635575ms per 129.063558us  CPU 51
concur: total 1290.960194ms per 129.096019us  CPU 8
concur: total 1291.037710ms per 129.103771us  CPU 5
concur: total 1291.164969ms per 129.116497us  CPU 43
concur: total 1291.269961ms per 129.126996us  CPU 48
concur: total 1291.328620ms per 129.132862us  CPU 3
concur: total 1291.512777ms per 129.151278us  CPU 6
concur: total 1292.126213ms per 129.212621us  CPU 46
concur: total 1292.459740ms per 129.245974us  CPU 0
concur: total 1319.990689ms per 131.999069us  CPU 36
concur: total 1320.534768ms per 132.053477us  CPU 79
concur: total 1321.478834ms per 132.147883us  CPU 52
concur: total 1321.688430ms per 132.168843us  CPU 9
concur: total 1322.367295ms per 132.236729us  CPU 41
concur: total 1322.774626ms per 132.277463us  CPU 84
concur: total 1322.858841ms per 132.285884us  CPU 37
concur: total 1323.004662ms per 132.300466us  CPU 80
concur: total 1323.509692ms per 132.350969us  CPU 35
concur: total 1324.168083ms per 132.416808us  CPU 78
concur: total 1324.349236ms per 132.434924us  CPU 40
concur: total 1324.390289ms per 132.439029us  CPU 83
concur: total 1324.582900ms per 132.458290us  CPU 39
concur: total 1324.827884ms per 132.482788us  CPU 82
concur: total 1325.039595ms per 132.503960us  CPU 38
concur: total 1325.178815ms per 132.517881us  CPU 81
concur: total 1325.988735ms per 132.598873us  CPU 76
concur: total 1326.019984ms per 132.601998us  CPU 77
concur: total 1326.032461ms per 132.603246us  CPU 34
concur: total 1326.344117ms per 132.634412us  CPU 32
concur: total 1326.633867ms per 132.663387us  CPU 75
concur: total 1326.717049ms per 132.671705us  CPU 23
concur: total 1326.808905ms per 132.680891us  CPU 33
concur: total 1327.019730ms per 132.701973us  CPU 66
concur: total 1327.332877ms per 132.733288us  CPU 25
concur: total 1327.359688ms per 132.735969us  CPU 68
concur: total 1327.583548ms per 132.758355us  CPU 26
concur: total 1327.881813ms per 132.788181us  CPU 69
concur: total 1333.479850ms per 133.347985us  CPU 67
concur: total 1333.685477ms per 133.368548us  CPU 65
concur: total 1334.340429ms per 133.434043us  CPU 70
concur: total 1334.459344ms per 133.445934us  CPU 73
concur: total 1334.721884ms per 133.472188us  CPU 24
concur: total 1329.700073ms per 132.970007us  CPU 29
concur: total 1329.780331ms per 132.978033us  CPU 72
concur: total 1334.903260ms per 133.490326us  CPU 30
concur: total 1334.930622ms per 133.493062us  CPU 22
concur: total 1334.987178ms per 133.498718us  CPU 27
concur: total 1335.053335ms per 133.505334us  CPU 21
concur: total 1335.098157ms per 133.509816us  CPU 71
concur: total 1335.130695ms per 133.513070us  CPU 28
concur: total 1335.131987ms per 133.513199us  CPU 42
concur: total 1335.168902ms per 133.516890us  CPU 64
concur: total 1343.899304ms per 134.389930us  CPU 74
concur: total 1343.952212ms per 134.395221us  CPU 31
concur: total 1351.058043ms per 135.105804us  CPU 85
times: 10000 threads: 86 cpus: 86


View attachment "base-files-cpus-86-perf.txt" of type "text/plain" (435440 bytes)

View attachment "base-missing-cpus-86-perf.txt" of type "text/plain" (398373 bytes)

View attachment "patched-files-cpus-86-perf.txt" of type "text/plain" (451091 bytes)

View attachment "patched-missing-cpus-86-perf.txt" of type "text/plain" (553056 bytes)

Powered by blists - more mailing lists