| Message ID | 162642752894.63632.5596341704463755308.stgit@web.messagingengine.com (mailing list archive) |
|---|---|
| Headers | show |
| Series | kernfs: proposed locking and concurrency improvement | expand |
Hi, Here are benchmark test runs against patch set v8. There pretty much the same a the v7 results as they should be. 1) Run with 40 distinct sysfs file paths on 80 cpu machine, perf graphs in base-files-cpu-80-perf.txt and patched-files-cpu-80-perf.txt. Base (5.12.3, unpatched) ------------------------ single: total 43.610095ms per 4.361009us concur: total 4515.483148ms per 451.548315us CPU 69 concur: total 4518.304581ms per 451.830458us CPU 17 concur: total 4525.275675ms per 452.527567us CPU 73 concur: total 4531.404107ms per 453.140411us CPU 77 concur: total 4543.011580ms per 454.301158us CPU 57 concur: total 4547.066040ms per 454.706604us CPU 45 concur: total 4547.652917ms per 454.765292us CPU 33 concur: total 4550.700660ms per 455.070066us CPU 5 concur: total 4552.377974ms per 455.237797us CPU 53 concur: total 4557.183667ms per 455.718367us CPU 15 concur: total 4560.088854ms per 456.008885us CPU 25 concur: total 4561.096541ms per 456.109654us CPU 49 concur: total 4562.542284ms per 456.254228us CPU 65 concur: total 4567.247924ms per 456.724792us CPU 29 concur: total 4567.551375ms per 456.755137us CPU 11 concur: total 4576.361518ms per 457.636152us CPU 35 concur: total 4579.072483ms per 457.907248us CPU 13 concur: total 4580.456910ms per 458.045691us CPU 21 concur: total 4581.441238ms per 458.144124us CPU 55 concur: total 4582.808775ms per 458.280878us CPU 37 concur: total 4583.284351ms per 458.328435us CPU 9 concur: total 4583.644571ms per 458.364457us CPU 3 concur: total 4584.435959ms per 458.443596us CPU 51 concur: total 4588.450469ms per 458.845047us CPU 61 concur: total 4588.784765ms per 458.878476us CPU 63 concur: total 4603.455415ms per 460.345542us CPU 27 concur: total 4605.229148ms per 460.522915us CPU 39 concur: total 4605.412438ms per 460.541244us CPU 71 concur: total 4607.079973ms per 460.707997us CPU 23 concur: total 4608.041174ms per 460.804117us CPU 79 concur: total 4608.345038ms per 460.834504us CPU 1 concur: total 4616.631547ms per 461.663155us CPU 19 concur: total 4616.845962ms per 461.684596us CPU 43 concur: total 4617.521385ms per 461.752139us CPU 59 concur: total 4620.077938ms per 462.007794us CPU 41 concur: total 4620.766014ms per 462.076601us CPU 47 concur: total 4622.365869ms per 462.236587us CPU 67 concur: total 4622.546362ms per 462.254636us CPU 31 concur: total 4623.371457ms per 462.337146us CPU 7 concur: total 4673.657084ms per 467.365708us CPU 75 concur: total 4703.545988ms per 470.354599us CPU 48 concur: total 4704.692544ms per 470.469254us CPU 56 concur: total 4705.308174ms per 470.530817us CPU 12 concur: total 4708.179087ms per 470.817909us CPU 52 concur: total 4710.879479ms per 471.087948us CPU 76 concur: total 4712.478473ms per 471.247847us CPU 16 concur: total 4712.897185ms per 471.289718us CPU 64 concur: total 4713.066302ms per 471.306630us CPU 32 concur: total 4713.309041ms per 471.330904us CPU 28 concur: total 4713.775133ms per 471.377513us CPU 8 concur: total 4714.142315ms per 471.414232us CPU 4 concur: total 4714.408679ms per 471.440868us CPU 0 concur: total 4714.823031ms per 471.482303us CPU 68 concur: total 4715.911133ms per 471.591113us CPU 72 concur: total 4716.356891ms per 471.635689us CPU 36 concur: total 4717.148390ms per 471.714839us CPU 74 concur: total 4717.956787ms per 471.795679us CPU 30 concur: total 4718.550180ms per 471.855018us CPU 62 concur: total 4718.780680ms per 471.878068us CPU 20 concur: total 4718.804903ms per 471.880490us CPU 14 concur: total 4719.527173ms per 471.952717us CPU 34 concur: total 4719.610769ms per 471.961077us CPU 44 concur: total 4720.205242ms per 472.020524us CPU 46 concur: total 4720.442429ms per 472.044243us CPU 60 concur: total 4720.844723ms per 472.084472us CPU 24 concur: total 4720.885926ms per 472.088593us CPU 40 concur: total 4721.157172ms per 472.115717us CPU 22 concur: total 4721.325226ms per 472.132523us CPU 42 concur: total 4722.194975ms per 472.219497us CPU 54 concur: total 4722.198238ms per 472.219824us CPU 38 concur: total 4722.270686ms per 472.227069us CPU 10 concur: total 4722.520311ms per 472.252031us CPU 70 concur: total 4722.697182ms per 472.269718us CPU 50 concur: total 4722.814473ms per 472.281447us CPU 2 concur: total 4723.047319ms per 472.304732us CPU 78 concur: total 4723.211099ms per 472.321110us CPU 58 concur: total 4723.325139ms per 472.332514us CPU 6 concur: total 4723.517066ms per 472.351707us CPU 66 concur: total 4723.540041ms per 472.354004us CPU 18 concur: total 4723.532642ms per 472.353264us CPU 26 times: 10000 threads: 80 cpus: 80 patched (5.12.3) ---------------- single: total 42.800654ms per 4.280065us concur: total 1497.315162ms per 149.731516us CPU 40 concur: total 1523.384944ms per 152.338494us CPU 6 concur: total 1527.974937ms per 152.797494us CPU 41 concur: total 1528.611772ms per 152.861177us CPU 4 concur: total 1530.640917ms per 153.064092us CPU 2 concur: total 1531.255960ms per 153.125596us CPU 28 concur: total 1531.502709ms per 153.150271us CPU 22 concur: total 1531.550394ms per 153.155039us CPU 54 concur: total 1531.951529ms per 153.195153us CPU 60 concur: total 1532.197299ms per 153.219730us CPU 78 concur: total 1532.279187ms per 153.227919us CPU 48 concur: total 1533.034026ms per 153.303403us CPU 14 concur: total 1533.558031ms per 153.355803us CPU 32 concur: total 1534.782403ms per 153.478240us CPU 62 concur: total 1534.997189ms per 153.499719us CPU 56 concur: total 1536.109295ms per 153.610929us CPU 38 concur: total 1536.214181ms per 153.621418us CPU 18 concur: total 1537.096808ms per 153.709681us CPU 36 concur: total 1537.612459ms per 153.761246us CPU 42 concur: total 1538.681882ms per 153.868188us CPU 12 concur: total 1538.912144ms per 153.891214us CPU 20 concur: total 1539.133370ms per 153.913337us CPU 76 concur: total 1539.732430ms per 153.973243us CPU 8 concur: total 1540.007044ms per 154.000704us CPU 66 concur: total 1540.497045ms per 154.049704us CPU 44 concur: total 1540.982660ms per 154.098266us CPU 16 concur: total 1541.037031ms per 154.103703us CPU 26 concur: total 1541.441290ms per 154.144129us CPU 72 concur: total 1541.445028ms per 154.144503us CPU 34 concur: total 1541.899099ms per 154.189910us CPU 46 concur: total 1542.395210ms per 154.239521us CPU 74 concur: total 1542.641833ms per 154.264183us CPU 58 concur: total 1543.768222ms per 154.376822us CPU 1 concur: total 1543.781564ms per 154.378156us CPU 10 concur: total 1545.325536ms per 154.532554us CPU 64 concur: total 1545.602875ms per 154.560287us CPU 70 concur: total 1545.628720ms per 154.562872us CPU 30 concur: total 1547.650530ms per 154.765053us CPU 68 concur: total 1547.925833ms per 154.792583us CPU 52 concur: total 1550.700993ms per 155.070099us CPU 24 concur: total 1551.551400ms per 155.155140us CPU 55 concur: total 1552.062602ms per 155.206260us CPU 49 concur: total 1552.084444ms per 155.208444us CPU 33 concur: total 1552.080191ms per 155.208019us CPU 7 concur: total 1552.108824ms per 155.210882us CPU 35 concur: total 1552.169043ms per 155.216904us CPU 50 concur: total 1552.246985ms per 155.224699us CPU 17 concur: total 1552.266063ms per 155.226606us CPU 21 concur: total 1552.322319ms per 155.232232us CPU 23 concur: total 1552.921723ms per 155.292172us CPU 3 concur: total 1553.085462ms per 155.308546us CPU 45 concur: total 1553.360574ms per 155.336057us CPU 37 concur: total 1553.736812ms per 155.373681us CPU 69 concur: total 1553.810163ms per 155.381016us CPU 63 concur: total 1554.245252ms per 155.424525us CPU 75 concur: total 1554.587717ms per 155.458772us CPU 47 concur: total 1554.811591ms per 155.481159us CPU 9 concur: total 1555.156676ms per 155.515668us CPU 5 concur: total 1555.240540ms per 155.524054us CPU 27 concur: total 1555.270924ms per 155.527092us CPU 79 concur: total 1555.502580ms per 155.550258us CPU 57 concur: total 1555.650301ms per 155.565030us CPU 43 concur: total 1555.684979ms per 155.568498us CPU 77 concur: total 1555.784989ms per 155.578499us CPU 67 concur: total 1555.892789ms per 155.589279us CPU 15 concur: total 1555.894896ms per 155.589490us CPU 73 concur: total 1556.029098ms per 155.602910us CPU 29 concur: total 1556.136688ms per 155.613669us CPU 59 concur: total 1556.184156ms per 155.618416us CPU 61 concur: total 1556.272459ms per 155.627246us CPU 19 concur: total 1556.316365ms per 155.631637us CPU 39 concur: total 1556.448166ms per 155.644817us CPU 31 concur: total 1556.494722ms per 155.649472us CPU 25 concur: total 1556.874991ms per 155.687499us CPU 13 concur: total 1556.964956ms per 155.696496us CPU 65 concur: total 1557.505884ms per 155.750588us CPU 53 concur: total 1557.670220ms per 155.767022us CPU 71 concur: total 1558.641962ms per 155.864196us CPU 0 concur: total 1559.141537ms per 155.914154us CPU 51 concur: total 1559.180879ms per 155.918088us CPU 11 times: 10000 threads: 80 cpus: 80 2) Run with a single sysfs file path on 80 cpu machine, perf graphs in base-missing-cpu-80-perf.txt and patched-missing-cpu-80-perf.txt. Base (5.12.3, unpatched) ------------------------ single: total 24.746542ms per 2.474654us concur: total 11642.525326ms per 1164.252533us CPU 41 concur: total 11753.266354ms per 1175.326635us CPU 21 concur: total 11882.784250ms per 1188.278425us CPU 5 concur: total 11897.657687ms per 1189.765769us CPU 13 concur: total 11915.176939ms per 1191.517694us CPU 61 concur: total 12032.739153ms per 1203.273915us CPU 45 concur: total 12095.349631ms per 1209.534963us CPU 33 concur: total 12147.225465ms per 1214.722546us CPU 65 concur: total 12236.584495ms per 1223.658449us CPU 73 concur: total 12273.499836ms per 1227.349984us CPU 25 concur: total 12327.559782ms per 1232.755978us CPU 1 concur: total 12339.221934ms per 1233.922193us CPU 53 concur: total 12592.935435ms per 1259.293543us CPU 69 concur: total 12751.932850ms per 1275.193285us CPU 57 concur: total 12789.601281ms per 1278.960128us CPU 49 concur: total 12800.329994ms per 1280.032999us CPU 15 concur: total 12899.047414ms per 1289.904741us CPU 29 concur: total 12950.772402ms per 1295.077240us CPU 55 concur: total 13031.263191ms per 1303.126319us CPU 17 concur: total 13077.573490ms per 1307.757349us CPU 59 concur: total 13086.450703ms per 1308.645070us CPU 19 concur: total 13112.144105ms per 1311.214410us CPU 47 concur: total 13167.152306ms per 1316.715231us CPU 32 concur: total 13203.910956ms per 1320.391096us CPU 9 concur: total 13230.538364ms per 1323.053836us CPU 60 concur: total 13232.286187ms per 1323.228619us CPU 23 concur: total 13244.948925ms per 1324.494893us CPU 75 concur: total 13260.586625ms per 1326.058663us CPU 24 concur: total 13279.711923ms per 1327.971192us CPU 4 concur: total 13290.308549ms per 1329.030855us CPU 64 concur: total 13297.234721ms per 1329.723472us CPU 37 concur: total 13297.585100ms per 1329.758510us CPU 7 concur: total 13300.929348ms per 1330.092935us CPU 20 concur: total 13311.528772ms per 1331.152877us CPU 46 concur: total 13315.521790ms per 1331.552179us CPU 52 concur: total 13331.269808ms per 1333.126981us CPU 28 concur: total 13333.563808ms per 1333.356381us CPU 26 concur: total 13334.923225ms per 1333.492323us CPU 39 concur: total 13340.811453ms per 1334.081145us CPU 72 concur: total 13346.192625ms per 1334.619263us CPU 63 concur: total 13349.415339ms per 1334.941534us CPU 68 concur: total 13352.414423ms per 1335.241442us CPU 0 concur: total 13354.401586ms per 1335.440159us CPU 44 concur: total 13356.828605ms per 1335.682861us CPU 54 concur: total 13358.754397ms per 1335.875440us CPU 16 concur: total 13362.014261ms per 1336.201426us CPU 12 concur: total 13364.189431ms per 1336.418943us CPU 6 concur: total 13364.252331ms per 1336.425233us CPU 67 concur: total 13365.159491ms per 1336.515949us CPU 76 concur: total 13369.198367ms per 1336.919837us CPU 48 concur: total 13369.574959ms per 1336.957496us CPU 36 concur: total 13377.299530ms per 1337.729953us CPU 40 concur: total 13377.717747ms per 1337.771775us CPU 35 concur: total 13377.886941ms per 1337.788694us CPU 56 concur: total 13382.748789ms per 1338.274879us CPU 66 concur: total 13384.125526ms per 1338.412553us CPU 79 concur: total 13384.216216ms per 1338.421622us CPU 58 concur: total 13388.371680ms per 1338.837168us CPU 18 concur: total 13393.289323ms per 1339.328932us CPU 10 concur: total 13394.423134ms per 1339.442313us CPU 8 concur: total 13394.537986ms per 1339.453799us CPU 14 concur: total 13396.152597ms per 1339.615260us CPU 30 concur: total 13397.312727ms per 1339.731273us CPU 34 concur: total 13397.555389ms per 1339.755539us CPU 70 concur: total 13398.055759ms per 1339.805576us CPU 78 concur: total 13399.219289ms per 1339.921929us CPU 77 concur: total 13401.160948ms per 1340.116095us CPU 38 concur: total 13401.177905ms per 1340.117790us CPU 27 concur: total 13401.652556ms per 1340.165256us CPU 50 concur: total 13403.311058ms per 1340.331106us CPU 22 concur: total 13406.279259ms per 1340.627926us CPU 42 concur: total 13406.714708ms per 1340.671471us CPU 2 concur: total 13409.186649ms per 1340.918665us CPU 74 concur: total 13409.591594ms per 1340.959159us CPU 62 concur: total 13410.246135ms per 1341.024613us CPU 31 concur: total 13412.007054ms per 1341.200705us CPU 11 concur: total 13412.140675ms per 1341.214068us CPU 71 concur: total 13412.737053ms per 1341.273705us CPU 51 concur: total 13413.214418ms per 1341.321442us CPU 3 concur: total 13414.303830ms per 1341.430383us CPU 43 times: 10000 threads: 80 cpus: 80 patched (5.12.3) ---------------- single: total 23.208077ms per 2.320808us concur: total 1206.802099ms per 120.680210us CPU 22 concur: total 1208.930779ms per 120.893078us CPU 8 concur: total 1211.247205ms per 121.124721us CPU 34 concur: total 1212.998353ms per 121.299835us CPU 62 concur: total 1213.389780ms per 121.338978us CPU 28 concur: total 1215.414555ms per 121.541455us CPU 48 concur: total 1218.962627ms per 121.896263us CPU 36 concur: total 1219.112096ms per 121.911210us CPU 74 concur: total 1219.292732ms per 121.929273us CPU 68 concur: total 1219.454262ms per 121.945426us CPU 2 concur: total 1219.937964ms per 121.993796us CPU 20 concur: total 1221.773152ms per 122.177315us CPU 42 concur: total 1222.383925ms per 122.238393us CPU 16 concur: total 1222.753728ms per 122.275373us CPU 14 concur: total 1222.824826ms per 122.282483us CPU 0 concur: total 1224.441811ms per 122.444181us CPU 56 concur: total 1224.779489ms per 122.477949us CPU 6 concur: total 1224.973746ms per 122.497375us CPU 40 concur: total 1225.042053ms per 122.504205us CPU 54 concur: total 1225.513107ms per 122.551311us CPU 4 concur: total 1225.848993ms per 122.584899us CPU 76 concur: total 1226.191945ms per 122.619195us CPU 32 concur: total 1226.556633ms per 122.655663us CPU 38 concur: total 1226.642992ms per 122.664299us CPU 60 concur: total 1227.444816ms per 122.744482us CPU 46 concur: total 1227.981028ms per 122.798103us CPU 18 concur: total 1228.193518ms per 122.819352us CPU 44 concur: total 1228.550486ms per 122.855049us CPU 26 concur: total 1228.840427ms per 122.884043us CPU 24 concur: total 1229.017209ms per 122.901721us CPU 78 concur: total 1229.724519ms per 122.972452us CPU 72 concur: total 1229.878627ms per 122.987863us CPU 12 concur: total 1230.770167ms per 123.077017us CPU 66 concur: total 1231.876743ms per 123.187674us CPU 58 concur: total 1232.688921ms per 123.268892us CPU 64 concur: total 1233.311151ms per 123.331115us CPU 52 concur: total 1236.349393ms per 123.634939us CPU 30 concur: total 1240.084640ms per 124.008464us CPU 70 concur: total 1247.767398ms per 124.776740us CPU 10 concur: total 1248.471436ms per 124.847144us CPU 50 concur: total 1249.292940ms per 124.929294us CPU 39 concur: total 1295.617635ms per 129.561764us CPU 23 concur: total 1297.170489ms per 129.717049us CPU 9 concur: total 1297.964528ms per 129.796453us CPU 63 concur: total 1299.431947ms per 129.943195us CPU 35 concur: total 1299.701030ms per 129.970103us CPU 49 concur: total 1299.703983ms per 129.970398us CPU 29 concur: total 1300.535158ms per 130.053516us CPU 3 concur: total 1300.902632ms per 130.090263us CPU 1 concur: total 1301.034574ms per 130.103457us CPU 5 concur: total 1301.105484ms per 130.110548us CPU 17 concur: total 1301.141575ms per 130.114158us CPU 75 concur: total 1301.499277ms per 130.149928us CPU 69 concur: total 1301.970890ms per 130.197089us CPU 41 concur: total 1302.582629ms per 130.258263us CPU 57 concur: total 1302.751012ms per 130.275101us CPU 43 concur: total 1302.816857ms per 130.281686us CPU 45 concur: total 1302.909810ms per 130.290981us CPU 15 concur: total 1303.018114ms per 130.301811us CPU 21 concur: total 1303.023366ms per 130.302337us CPU 7 concur: total 1303.218080ms per 130.321808us CPU 37 concur: total 1303.578443ms per 130.357844us CPU 27 concur: total 1303.954501ms per 130.395450us CPU 47 concur: total 1304.038096ms per 130.403810us CPU 55 concur: total 1304.040952ms per 130.404095us CPU 33 concur: total 1304.156534ms per 130.415653us CPU 77 concur: total 1304.188698ms per 130.418870us CPU 19 concur: total 1304.792850ms per 130.479285us CPU 67 concur: total 1304.894882ms per 130.489488us CPU 25 concur: total 1304.996763ms per 130.499676us CPU 61 concur: total 1305.107675ms per 130.510768us CPU 13 concur: total 1305.370941ms per 130.537094us CPU 73 concur: total 1305.487118ms per 130.548712us CPU 59 concur: total 1305.654535ms per 130.565453us CPU 65 concur: total 1305.891900ms per 130.589190us CPU 53 concur: total 1306.513791ms per 130.651379us CPU 31 concur: total 1307.221019ms per 130.722102us CPU 71 concur: total 1309.522870ms per 130.952287us CPU 11 concur: total 1309.526910ms per 130.952691us CPU 51 concur: total 1316.416007ms per 131.641601us CPU 79 times: 10000 threads: 80 cpus: 80 Ian
On Fri, Jul 16, 2021 at 05:28:13PM +0800, Ian Kent wrote: > There have been a few instances of contention on the kernfs_mutex during > path walks, a case on very large IBM systems seen by myself, a report by > Brice Goglin and followed up by Fox Chen, and I've since seen a couple > of other reports by CoreOS users. > > The common thread is a large number of kernfs path walks leading to > slowness of path walks due to kernfs_mutex contention. > > The problem being that changes to the VFS over some time have increased > it's concurrency capabilities to an extent that kernfs's use of a mutex > is no longer appropriate. There's also an issue of walks for non-existent > paths causing contention if there are quite a few of them which is a less > common problem. > > This patch series is relatively straight forward. > > All it does is add the ability to take advantage of VFS negative dentry > caching to avoid needless dentry alloc/free cycles for lookups of paths > that don't exit and change the kernfs_mutex to a read/write semaphore. > > The patch that tried to stay in VFS rcu-walk mode during path walks has > been dropped for two reasons. First, it doesn't actually give very much > improvement and, second, if there's a place where mistakes could go > unnoticed it would be in that path. This makes the patch series simpler > to review and reduces the likelihood of problems going unnoticed and > popping up later. > > Changes since v7: > - remove extra tab in helper kernfs_dir_changed. > - fix thinko adding an unnecessary kernfs_inc_rev() in kernfs_rename_ns(). Thanks for sticking with this, I've applied this to my testing branch and let's see how 0-day does with it :) greg k-h
On Tue, 2021-07-27 at 12:02 +0200, Greg Kroah-Hartman wrote: > On Fri, Jul 16, 2021 at 05:28:13PM +0800, Ian Kent wrote: > > There have been a few instances of contention on the kernfs_mutex > > during > > path walks, a case on very large IBM systems seen by myself, a > > report by > > Brice Goglin and followed up by Fox Chen, and I've since seen a > > couple > > of other reports by CoreOS users. > > > > The common thread is a large number of kernfs path walks leading to > > slowness of path walks due to kernfs_mutex contention. > > > > The problem being that changes to the VFS over some time have > > increased > > it's concurrency capabilities to an extent that kernfs's use of a > > mutex > > is no longer appropriate. There's also an issue of walks for non- > > existent > > paths causing contention if there are quite a few of them which is > > a less > > common problem. > > > > This patch series is relatively straight forward. > > > > All it does is add the ability to take advantage of VFS negative > > dentry > > caching to avoid needless dentry alloc/free cycles for lookups of > > paths > > that don't exit and change the kernfs_mutex to a read/write > > semaphore. > > > > The patch that tried to stay in VFS rcu-walk mode during path walks > > has > > been dropped for two reasons. First, it doesn't actually give very > > much > > improvement and, second, if there's a place where mistakes could go > > unnoticed it would be in that path. This makes the patch series > > simpler > > to review and reduces the likelihood of problems going unnoticed > > and > > popping up later. > > > > Changes since v7: > > - remove extra tab in helper kernfs_dir_changed. > > - fix thinko adding an unnecessary kernfs_inc_rev() in > > kernfs_rename_ns(). > > Thanks for sticking with this, I've applied this to my testing branch > and let's see how 0-day does with it :) That's great news Greg, and thanks for putting up with me too, ;) Ian
There have been a few instances of contention on the kernfs_mutex during path walks, a case on very large IBM systems seen by myself, a report by Brice Goglin and followed up by Fox Chen, and I've since seen a couple of other reports by CoreOS users. The common thread is a large number of kernfs path walks leading to slowness of path walks due to kernfs_mutex contention. The problem being that changes to the VFS over some time have increased it's concurrency capabilities to an extent that kernfs's use of a mutex is no longer appropriate. There's also an issue of walks for non-existent paths causing contention if there are quite a few of them which is a less common problem. This patch series is relatively straight forward. All it does is add the ability to take advantage of VFS negative dentry caching to avoid needless dentry alloc/free cycles for lookups of paths that don't exit and change the kernfs_mutex to a read/write semaphore. The patch that tried to stay in VFS rcu-walk mode during path walks has been dropped for two reasons. First, it doesn't actually give very much improvement and, second, if there's a place where mistakes could go unnoticed it would be in that path. This makes the patch series simpler to review and reduces the likelihood of problems going unnoticed and popping up later. Changes since v7: - remove extra tab in helper kernfs_dir_changed. - fix thinko adding an unnecessary kernfs_inc_rev() in kernfs_rename_ns(). Changes since v6: - ensure negative dentry as rename target gets invalidated. - don't bother checking if node is a directory in revision helpers. - don't use dget_parent(), just use dentry d_lock to ensure parent is stable. - fix kernfs_iop_getattr() and kernfs_iop_permission() locking. - drop add kernfs_need_inode_refresh() patch, it can't be used. Changes since v5: - change kernfs_dir_changed() comparison. - move negative dentry out from under kernfs node lock in revalidate. - only set d_time for negative dentries. - add patch to move d_splice_alias() out from under kernfs node lock in lookup. Changes since v4: - fixed kernfs_active() naming. - added back kernfs_node revision patch to use for negative dentry validation. - minor updates to patch descriptions. Changes since v3: - remove unneeded indirection when referencing the super block. - check if inode attribute update is actually needed. Changes since v2: - actually fix the inode attribute update locking. - drop the patch that tried to stay in rcu-walk mode. - drop the use a revision to identify if a directory has changed patch. Changes since v1: - fix locking in .permission() and .gated() by re-factoring the attribute handling code. --- Ian Kent (5): kernfs: add a revision to identify directory node changes kernfs: use VFS negative dentry caching kernfs: switch kernfs to use an rwsem kernfs: use i_lock to protect concurrent inode updates kernfs: dont call d_splice_alias() under kernfs node lock fs/kernfs/dir.c | 151 ++++++++++++++++++++---------------- fs/kernfs/file.c | 4 +- fs/kernfs/inode.c | 26 ++++--- fs/kernfs/kernfs-internal.h | 5 +- fs/kernfs/mount.c | 12 +-- fs/kernfs/symlink.c | 4 +- include/linux/kernfs.h | 2 +- 7 files changed, 112 insertions(+), 92 deletions(-) -- Ian