LinuxLists.cc - [memcg] 0f12156dff: will-it-scale.per_process

2021-09-07 14:52:54

Subject: [memcg] 0f12156dff: will-it-scale.per_process_ops -33.6% regression

Greeting,

FYI, we noticed a

commit: 0f12156dff2862ac542 In addition to that,
+------------------+------- | testcase: change | test machine | test parameters | | mode=process | | nr_task=50% | | test=lock1 | | ucode=0xd000280 +------------------+-------

If you fix the issue, Reported-by: kernel

Details are as below:
---------------------------

To reproduce:

git clone bin/lkp install bin/lkp split-job bin/lkp run
=========================== compiler/cpufreq_governor/k gcc-9/performance/x86_64-
commit:
b655843444 ("memcg: 0f12156dff ("memcg:
b655843444152c0a 0f12156dff ---------------- ---------- %stddev %change \ | 65855410 1266449 -33.6% 65855410 141099 ? 3% 140875 ? 3% 28.79 +6.7 20.49 -6.7 138801 ? 3% 138689 ? 3% 34681 ? 3% +8.3% 34681 ? 3% +8.3% 4480 ? 3% -85.9% 4480 ? 3% -85.9% 2957456 ?214% -96.5% 103.99 19.83 -34.5% 2487 ? 2% +4.4% 35219 ? 3% +8.2% 10977 +1.8% 40640 ? 3% +7.6% 35219 ? 3% +8.2% 53061 ? 3% +6.5% 0.02 ? 19% -24.0% 9.26 ? 21% -39.8% 172.83 ? 11% -29.4% 127.50 ? 4% +47.8% 1080 ? 19% +66.6% 7204 ? 3% -12.0% 7204 ? 3% -12.0% 7209 ? 3% -12.0% 6825 ? 4% -11.5% 9.26 ? 21% -39.8% 7204 ? 3% -12.0% 7204 ? 3% -12.0% 0.05 ? 7% +751.3% 7209 ? 3% -12.0% 6825 ? 4% -11.5% 827.67 ? 9% -16.3% 3623 ? 54% +66.8% 3623 ? 54% +66.8% 239.50 ? 18% -33.1% 261.33 ? 18% -54.4% 7894 -30.2% 7894 -30.2% 3374 ? 40% +90.6% 3374 ? 40% +90.6% 272.83 ? 8% -36.4% 2890 ? 37% +61.0% 2890 ? 37% +61.0% 272.67 ? 9% -29.0% 3475 ? 47% +109.3% 3475 ? 47% +109.3% 3054 ? 37% +91.7% 3054 ? 37% +91.7% 0.05 ? 5% +1601.9% 12.52 ? 2% +20.9 647128 ? 2% +4855.0% 5130146 ? 2% +1777.3% 2302 +6.1% 259518 ? 2% -98.1% 0.17 -0.1 65644478 51892 -27.4% 68181011 1933 +58.2% 81.87 ? 12% +782.2% 126326 ? 3% +6119.4% 26893 ? 6% +3382.5% 87.04 +12.8 33838 ? 3% +28351.8% 0.04 ? 2% +1709.7% 0.52 -0.0 12.65 ? 2% +20.7 222708 ? 2% -98.0% 0.17 -0.1 0.00 -0.0 1929 +56.0% 82.22 +7.1 78.75 +21.2 603681 +56.8% 649626 ? 2% +4816.5% 5137402 ? 2% +1767.5% 2313 +5.8% 65424262 51749 -27.4% 67951468 126213 ? 3% +6102.7% 27281 ? 6% +3321.9% 33783 ? 3% +28302.8% 15207 ? 10% -34.0% 16968 ? 17% -36.3% 17526 ? 19% -36.4% 17232 ? 16% -48.1% 15770 ? 16% -43.7% 14396 ? 15% -44.6% 18909 ? 5% -46.4% 17822 ? 23% -45.3% 18056 ? 22% -52.5% 14576 ? 17% -33.3% 18836 ? 15% -47.1% 20542 ? 17% -54.3% 17485 ? 19% -42.0% 17387 ? 17% -37.8% 18473 ? 15% -49.9% 19232 ? 19% -44.1% 17052 ? 21% -42.9% 17330 ? 17% -43.5% 18624 ? 15% -43.9% 18206 ? 18% -44.4% 18047 ? 12% -44.5% 19351 ? 18% -47.7% 18629 ? 17% -51.4% 15355 ? 14% -44.8% 15129 ? 14% -39.7% 14744 ? 10% -39.0% 13973 ? 11% -39.4% 35779 ? 10% -27.5% 18703 ? 16% -43.4% 17000 ? 19% -40.8% 18602 ? 16% -40.6% 17242 ? 23% -44.0% 17841 ? 15% -43.1% 17867 ? 14% -44.6% 19083 ? 16% -46.7% 19161 ? 11% -49.8% 19335 ? 15% -47.7% 20460 ? 11% -54.3% 10641 ? 58% +117.0% 20165 ? 8% -53.5% 20158 ? 9% -47.3% 18503 ? 20% -40.7% 18917 ? 11% -51.1% 19989 ? 9% -57.3% 10393 ? 58% +155.8% 16499 ? 11% -45.5% 18599 ? 16% -49.3% 20721 ? 8% -56.4% 6210 ? 26% +213.4% 16912 ? 21% -43.6% 19085 ? 14% -49.9% 16524 ? 21% -46.7% 18590 ? 13% -52.0% 19473 ? 15% -47.6% 19807 ? 16% -51.7% 14621 ? 7% -44.6% 21579 ? 9% -49.4% 17758 ? 25% -50.6% 18495 ? 15% -47.1% 16729 ? 20% -46.4% 15031 ? 10% -44.4% 15721 ? 17% -42.2% 14757 ? 3% -32.1% 37022 ? 9% -41.4% 16077 ? 6% -41.1% 20064 ? 11% -54.4% 14811 ? 18% -46.8% 15191 ? 8% -49.8% 16592 ? 13% -49.0% 15392 ? 21% -44.3% 18606 ? 16% -52.5% 14770 ? 22% -40.4% 14465 ? 11% -32.3% 16993 ? 16% -48.1% 18034 ? 14% -52.9% 18628 ? 16% -47.9% 16564 ? 17% -42.4% 16333 ? 19% -45.3% 17985 ? 19% -52.0% 16575 ? 17% -50.7% 17243 ? 18% -48.5% 17293 ? 12% -44.8% 16269 ? 20% -46.7% 17326 ? 18% -45.8% 18826 ? 18% -50.6% 19893 ? 8% -52.4% 18812 ? 19% -43.6% 20001 ? 12% -53.0% 20756 ? 5% -53.9% 18840 ? 15% -53.5% 17159 ? 21% -52.3% 18671 ? 16% -52.4% 18929 ? 11% -55.5% 17559 ? 15% -51.0% 17343 ? 14% -53.0% 15481 ? 13% -43.6% 16364 ? 14% -43.9% 18492 ? 19% -43.5% 14056 ? 7% -38.3% 14189 ? 6% -37.9% 15154 ? 14% -48.1% 15589 ? 14% -42.4% 14492 ? 10% -34.4% 34198 ? 16% -52.1% 17203 ? 7% -46.2% 15367 ? 11% -44.0% 14328 ? 13% -37.8% 16134 ? 16% -46.3% 15282 ? 17% -44.5% 1808149 ? 6% -46.3% 36250 ? 5% +36.3% 17.47 ? 9% -5.9 9.44 ? 9% -3.2 8.57 ? 9% -2.9 2.27 ? 9% -1.0 2.68 ? 9% -1.0 2.65 ? 8% -0.9 2.52 ? 9% -0.9 0.80 ? 8% -0.5 1.34 ? 9% -0.5 1.62 ? 9% -0.5 0.82 ? 10% -0.4 0.66 ? 11% -0.4 1.04 ? 9% -0.3 0.86 ? 9% -0.3 0.00 +0.6 1.06 ? 9% +0.9 0.00 +0.9 0.67 ? 9% +1.0 0.00 +1.4 0.00 +1.8 0.00 +2.2 0.00 +2.2 0.00 +2.4 0.00 +2.8 0.00 +2.8 0.00 +2.9 1.10 ? 9% +2.9 0.00 +3.0 1.69 ? 9% +3.8 3.39 ? 8% +4.0 3.00 ? 8% +4.2 0.00 +4.6 0.00 +5.0 6.48 ? 9% +6.3 5.71 ? 9% +6.5 0.00 +7.3 17.27 ? 9% +8.5 13.80 ? 9% +9.6 28.29 ? 9% +13.3 26.73 ? 9% +13.8 23.74 ? 9% +14.8 17.55 ? 9% -5.9 9.74 ? 9% -3.3 9.02 ? 8% -3.1 4.02 ? 9% -1.5 2.58 ? 9% -0.9 2.68 ? 8% -0.9 2.02 ? 7% -0.7 1.40 ? 8% -0.6 1.67 ? 9% -0.5 1.48 ? 9% -0.5 1.23 ? 9% -0.4 1.04 ? 9% -0.4 0.86 ? 10% -0.3 0.86 ? 9% -0.3 0.76 ? 7% -0.3 0.77 ? 8% -0.3 0.59 ? 8% -0.2 0.66 ? 9% -0.2 0.56 ? 9% -0.2 0.58 ? 9% -0.2 0.69 ? 11% -0.1 0.56 ? 10% -0.1 0.38 ? 8% -0.1 0.33 ? 11% -0.1 0.35 ? 8% -0.1 0.27 ? 21% -0.1 0.23 ? 9% -0.1 0.31 ? 8% -0.1 0.24 ? 9% -0.1 0.19 ? 11% -0.1 0.15 ? 10% -0.1 0.16 ? 7% -0.1 0.13 ? 9% -0.0 0.12 ? 12% -0.0 0.14 ? 9% -0.0 0.08 ? 10% -0.0 0.13 ? 8% -0.0 0.10 ? 10% -0.0 0.10 ? 10% -0.0 0.00 +0.1 0.00 +0.2 0.00 +0.6 0.00 +0.6 1.10 ? 9% +0.9 0.00 +1.7 0.00 +2.4 0.00 +2.5 0.00 +4.7 0.00 +5.8 0.00 +5.8 0.00 +7.4 0.00 +7.4 3.47 ? 9% +7.8 0.00 +8.1 17.35 ? 9% +8.5 14.01 ? 9% +9.5 10.00 ? 8% +10.2 9.00 ? 8% +10.7 28.35 ? 9% +13.3 26.89 ? 9% +13.7 23.92 ? 9% +14.7 17.12 ? 9% -5.7 9.72 ? 9% -3.3 7.81 ? 8% -2.6 3.90 ? 9% -1.4 1.97 ? 7% -0.7 1.78 ? 9% -0.6 1.34 ? 8% -0.5 1.35 ? 9% -0.4 1.19 ? 9% -0.4 1.30 ? 8% -0.3 0.91 ? 9% -0.3 1.08 ? 9% -0.3 0.74 ? 8% -0.3 0.79 ? 9% -0.3 0.76 ? 9% -0.3 0.83 ? 9% -0.3 0.70 ? 9% -0.2 0.66 ? 7% -0.2 0.64 ? 9% -0.2 0.52 ? 9% -0.2 0.47 ? 10% -0.2 0.54 ? 9% -0.2 0.46 ? 9% -0.2 0.40 ? 11% -0.2 0.33 ? 11% -0.1 0.38 ? 9% -0.1 0.25 ? 7% -0.1 0.24 ? 9% -0.1 0.28 ? 7% -0.1 0.22 ? 24% -0.1 0.28 ? 8% -0.1 0.25 ? 9% -0.1 0.26 ? 8% -0.1 0.18 ? 9% -0.1 0.15 ? 8% -0.1 0.14 ? 9% -0.1 0.12 ? 12% -0.1 0.18 ? 12% -0.1 0.08 ? 12% -0.0 0.12 ? 9% -0.0 0.10 ? 10% -0.0 0.10 ? 10% -0.0 0.10 ? 8% -0.0 0.11 ? 8% -0.0 0.00 +0.1 0.00 +0.2 0.00 +0.3 0.00 +0.6 0.00 +0.6 3.14 ? 8% +1.3 0.00 +1.7 0.00 +2.2 0.00 +2.4 0.00 +4.7 0.00 +5.9

7e+07 +---------------- | .+ 6.5e+07 |+++ ++++++.++++ | 6e+07 |-+ | 5.5e+07 |-+ | 5e+07 |-+ O | OO O O 4.5e+07 |O+O OOOO O O | O 4e+07 |-+ O | 3.5e+07 +---------------- 1.3e+06 +---------------- |+++.+++++++.++++ 1.2e+06 |-+ | | 1.1e+06 |-+ | 1e+06 |-+ | O 900000 |-+ O |O O OOO O O OO O O O O O | O 800000 |-O O O | 700000 +---------------- 7e+07 +---------------- | .+ 6.5e+07 |+++ ++++++.++++ | 6e+07 |-+ | 5.5e+07 |-+ | 5e+07 |-+ O | OO O O 4.5e+07 |O+O OOOO O O | O 4e+07 |-+ O | 3.5e+07 +---------------- [*] bisect-good sample
[O] bisect-bad sample

*************************** lkp-icl-2sp2: 128 =========================== compiler/cpufreq_governor/k gcc-9/performance/x86_64-
commit:
b655843444 ("memcg: 0f12156dff ("memcg:
b655843444152c0a 0f12156dff ---------------- ---------- %stddev %change \ | 1.785e+08 2788911 1.785e+08 366571 ? 7% +15.2% 2807 +2.4% 0.02 ? 2% -0.0 5.55 -2.4 5364 ? 2% -85.5% 5364 ? 2% -85.5% 49038 +5.1% 55209 +5.1% 49038 +5.1% 183.83 ? 28% +320.7% 836.50 ? 10% +46.3% 1348 ? 83% +208.2% 1089 ? 77% +158.6% 183.83 ? 28% +320.7% 806.17 ? 14% +31.2% 723.17 ? 74% +417.0% 4847 ? 94% -97.1% 4847 ? 94% -97.1% 8069 ? 63% -80.7% 8069 ? 63% -80.7% 2345 -12.0% 2345 -12.0% 262.68 ? 7% -42.8% 561.17 ? 9% -16.5% 183.33 ? 7% +43.2% 113.33 ? 11% +78.7% 7031 -12.0% 7031 -12.0% 7035 -12.0% 2345 -12.0% 2345 -12.0% 2.50 ? 7% -47.6% 262.68 ? 7% -42.8% 7031 -12.0% 7031 -12.0% 7035 -12.0% 0.04 ? 22% +1112.5% 0.02 ? 3% +1574.6% 7.076e+10 0.05 +0.0 35955313 ? 2% 10.31 +18.0 492492 +4777.7% 4620608 +1768.8% 2651 +2.5% 0.47 +8.5% 425867 -98.4% 0.00 ? 4% +0.0 1.03e+11 0.00 +0.0 6.534e+10 3.516e+11 2.11 -7.1% 37.50 ? 2% +2090.1% 1867 -6.7% 104632 ? 2% +6707.0% 14377 ? 21% +3209.4% 45.30 ? 16% +51.5 48602 ? 14% +22917.3% 65851 ? 9% +432.6% 0.01 +1902.5% 0.05 +0.0 10.65 +17.3 0.47 +7.7% 336119 -97.9% 0.00 ? 4% +0.0 0.00 +0.0 2.11 -7.1% 87.73 ? 2% +6.0 42.34 ? 13% +54.6 594415 +56.9% 7.05e+10 35867377 493933 +4747.8% 4639810 +1756.0% 2653 +2.6% 1.027e+11 6.51e+10 3.504e+11 104869 ? 2% +6668.1% 14698 ? 21% +3134.0% 48497 ? 14% +22887.1% 65971 ? 9% +430.7% 1.061e+14 11738 ? 11% -32.5% 12768 ? 12% -39.4% 8853 ? 10% -33.9% 9695 ? 6% -40.1% 9277 ? 7% -38.0% 9585 ? 8% -33.5% 9838 ? 7% -40.5% 9315 ? 8% -40.6% 9661 ? 10% -40.5% 9531 ? 6% -35.6% 9472 ? 7% -34.4% 11502 ? 13% -41.5% 9003 ? 11% -34.3% 9739 ? 7% -36.5% 9453 ? 9% -34.2% 9891 ? 7% -31.3% 10008 ? 9% -39.1% 10132 ? 17% -41.7% 9765 ? 8% -37.5% 8960 ? 11% -33.9% 10910 ? 17% -34.0% 9577 ? 6% -37.6% 9565 ? 10% -42.9% 9137 ? 10% -38.7% 9418 ? 13% -34.3% 9151 ? 15% -37.2% 10800 ? 7% -34.7% 9815 ? 9% -29.1% 11463 ? 15% -39.2% 10757 ? 15% -37.6% 10809 ? 17% -38.5% 11169 ? 14% -42.4% 12215 ? 11% -40.1% 11500 ? 12% -39.2% 11027 ? 18% -32.7% 11155 ? 17% -39.5% 11083 ? 16% -32.8% 11841 ? 15% -44.9% 11415 ? 16% -38.7% 11325 ? 5% -37.7% 12111 ? 11% -44.5% 11627 ? 11% -43.1% 11113 ? 3% -40.1% 10333 ? 13% -33.2% 12191 ? 13% -42.5% 10265 ? 11% -31.7% 11440 ? 13% -39.6% 10389 ? 12% -33.8% 11224 ? 10% -37.3% 10886 ? 13% -32.8% 10588 ? 9% -32.7% 9719 ? 16% -31.8% 11099 ? 12% -33.2% 10338 ? 9% -25.6% 10238 ? 11% -27.1% 11660 ? 15% -35.5% 10552 ? 13% -31.6% 10552 ? 10% -31.1% 10103 ? 13% -31.8% 10974 ? 8% -36.9% 10730 ? 13% -37.2% 11205 ? 12% -35.4% 11136 ? 12% -39.4% 10649 ? 11% -33.6% 9988 ? 10% -36.5% 9801 ? 9% -36.9% 12102 ? 9% -37.5% 9884 ? 12% -36.2% 9347 ? 9% -34.8% 10164 ? 11% -35.3% 10518 ? 15% -37.2% 9857 ? 12% -32.5% 10795 ? 13% -44.0% 10397 ? 12% -38.9% 9513 ? 6% -30.5% 10400 ? 10% -37.6% 10434 ? 8% -39.1% 12055 ? 14% -32.8% 10580 ? 12% -40.1% 10099 ? 11% -41.4% 10021 ? 10% -35.8% 9856 ? 7% -31.3% 9963 ? 12% -36.0% 9664 ? 8% -30.1% 9673 ? 5% -33.2% 9910 ? 8% -36.4% 9589 ? 5% -34.2% 11733 ? 9% -42.0% 9603 ? 42% +105.6% 9881 ? 9% -34.0% 9620 ? 6% -33.2% 35921 ? 13% -39.4% 10659 ? 13% -38.8% 9917 ? 7% -38.9% 9572 ? 7% -40.0% 9683 ? 9% -36.2% 10099 ? 7% -38.4% 9726 ? 5% -28.6% 9463 ? 6% -37.1% 10789 ? 12% -34.1% 9917 ? 7% -35.1% 9821 ? 4% -34.5% 9938 ? 12% -38.6% 9838 ? 4% -35.2% 10401 ? 8% -40.1% 10051 ? 6% -36.4% 9792 ? 4% -35.3% 10059 ? 9% -36.3% 9972 ? 8% -39.2% 9447 ? 9% -30.8% 11511 ? 24% -37.0% 9563 ? 6% -30.9% 36244 ? 11% -29.7% 9671 ? 8% -34.4% 9848 ? 7% -33.7% 9578 ? 8% -32.4% 9519 ? 5% -36.3% 9480 ? 9% -31.0% 9025 ? 8% -35.2% 9656 ? 19% -39.8% 10008 ? 7% -39.4% 9619 ? 6% -41.1% 1308899 ? 3% -36.2% 47568 ? 2% +20.6% 10.02 ? 3% -4.9 8.20 -3.8 5.28 -2.7 4.78 -2.3 3.68 -1.8 3.36 -1.4 2.98 -1.2 2.51 -1.2 2.49 -1.1 2.02 -1.0 2.02 -1.0 1.14 -0.8 1.70 -0.7 1.24 -0.6 2.15 -0.6 1.34 -0.6 1.32 -0.6 1.22 -0.5 1.21 -0.5 1.41 -0.5 1.21 -0.5 1.08 -0.5 0.76 -0.4 1.27 -0.4 1.04 -0.4 0.76 -0.1 97.66 +0.4 0.00 +0.5 0.00 +0.8 0.00 +0.9 0.00 +0.9 0.00 +1.3 2.15 +1.4 0.00 +1.6 1.14 +1.8 0.00 +1.8 0.00 +2.1 0.00 +2.2 0.00 +2.4 0.00 +2.6 0.00 +2.7 0.00 +3.0 2.86 ? 2% +3.6 0.00 +3.6 0.00 +3.6 0.00 +3.7 0.00 +4.3 3.76 +5.7 8.70 +6.2 7.40 +6.3 83.98 +6.8 82.41 +7.4 0.00 +7.8 46.22 +7.9 78.68 +8.9 0.00 +9.2 73.64 +11.0 0.00 +11.2 38.04 +11.7 16.21 +11.8 13.60 +12.1 63.80 +15.2 7.76 -3.4 8.05 -3.4 5.70 ? 2% -2.8 6.00 -2.5 5.10 -2.4 5.00 -2.2 3.90 -1.9 3.94 -1.7 3.74 -1.7 3.95 -1.6 2.72 -1.3 2.58 -1.3 3.10 -1.3 2.18 -1.0 2.18 -1.0 2.03 -0.9 2.19 -0.8 1.54 -0.7 1.54 -0.7 1.68 -0.6 1.32 -0.6 1.19 -0.5 1.10 -0.4 0.81 -0.4 0.94 -0.3 0.74 -0.3 0.75 -0.3 0.70 -0.3 0.53 -0.3 0.78 ? 2% -0.3 0.61 -0.3 0.60 -0.2 0.53 -0.2 0.43 ? 2% -0.2 0.43 ? 2% -0.2 0.50 -0.2 0.37 -0.2 0.32 ? 2% -0.1 0.36 -0.1 0.22 -0.1 0.41 -0.1 0.20 ? 2% -0.1 0.00 +0.1 0.00 +0.1 98.10 +0.1 0.00 +0.3 0.00 +0.5 2.41 +1.3 0.00 +2.0 0.32 ? 2% +2.1 0.00 +3.4 0.00 +4.9 0.00 +5.6 0.00 +5.9 84.29 +6.7 0.00 +7.1 83.10 +7.2 0.00 +7.3 46.87 +7.6 79.26 +8.6 74.02 +10.8 39.59 +11.0 0.00 +11.2 0.00 +11.4 8.04 +11.5 0.00 +13.6 64.58 +14.8 25.73 +17.7 22.71 +18.8 7.42 -3.3 6.55 -3.2 5.72 -2.4 5.14 -2.4 5.00 -2.2 3.49 -1.7 3.61 -1.5 3.20 -1.4 2.97 -1.2 2.35 -1.2 3.02 -1.2 2.25 -1.1 2.38 -0.9 1.80 -0.8 1.39 -0.7 1.61 -0.7 1.60 -0.6 1.43 -0.6 1.40 -0.6 1.31 -0.6 1.16 -0.5 1.08 -0.5 0.97 -0.4 1.02 -0.4 1.04 -0.4 0.81 -0.4 0.76 -0.4 0.71 -0.3 0.64 -0.3 0.75 ? 2% -0.3 0.70 -0.3 0.58 ? 2% -0.3 0.43 -0.3 0.52 -0.2 0.53 -0.2 0.63 ? 2% -0.2 0.43 -0.2 0.56 ? 2% -0.2 0.43 -0.2 0.32 ? 2% -0.2 0.33 -0.2 0.33 -0.1 0.42 ? 2% -0.1 0.33 ? 2% -0.1 0.32 -0.1 0.32 ? 2% -0.1 0.30 ? 3% -0.1 0.26 ? 3% -0.1 0.17 ? 2% -0.1 0.11 ? 4% -0.0 0.20 ? 3% -0.0 0.10 ? 4% -0.0 0.00 +0.1 0.00 +0.1 0.00 +0.1 0.00 +0.4 0.11 ? 3% +0.8 0.00 +1.8 0.00 +2.0 9.35 +3.0 0.00 +3.4 0.00 +4.0 0.00 +5.2 0.00 +5.9 0.00 +9.0

Disclaimer:
Results have been for informational design or configuration

---
0DAY/LKP+ Test Infrastructure https://git.kernel.org/cgit/linux/kernel/git/torvalds/linux.git master
threads 2 sockets Skylake with 192G memory
Will It Scale takes a testcase and runs it from 1 through to n parallel copies to see if the testcase will scale. It builds both a process and threads based test in order to see any differences between the two.
/github.com/antonblanchard/will-it-scale">https://github.com/antonblanchard/will-it-scale
the commit also has significant impact on the following tests:
--------------------------------------------------------------------------+
| will-it-scale: will-it-scale.per_process_ops -40.9% regression |
| 128 threads 2 sockets Intel(R) Xeon(R) Gold 6338 CPU @ 2.00GHz with 256G memory |
| cpufreq_governor=performance |
|
|
|
|
--------------------------------------------------------------------------+
kindly add following tag
test robot <[email protected]>
----------------------------------------------------------------------->
href="https://github.com/intel/lkp-tests.git">https://github.com/intel/lkp-tests.git
job.yaml # job file is attached in this email
--compatible job.yaml # generate the yaml file for lkp run
generated-yaml-file
==============================================================
config/mode/nr_task/rootfs/tbox_group/test/testcase/ucode:
rhel-8.3/process/50%/debian-10.4-x86_64-20200603.cgz/lkp-skl-fpga01/lock1/will-it-scale/0x2006a0a
enable accounting for pollfd and select bits arrays")
enable accounting for file lock caches")
2862ac54235fc7270
-----------------
%stddev
\
-33.6% 43722413 ? 5% will-it-scale.52.processes
840815 ? 5% will-it-scale.per_process_ops
-33.6% 43722413 ? 5% will-it-scale.workload
+8.2% 152680 ? 2% meminfo.Active
+8.2% 152456 ? 2% meminfo.Active(anon)
35.47 mpstat.cpu.all.sys%
13.76 ? 5% mpstat.cpu.all.usr%
+8.4% 150456 ? 2% numa-meminfo.node1.Active
+8.4% 150381 ? 2% numa-meminfo.node1.Active(anon)
37570 ? 2% numa-vmstat.node1.nr_active_anon
37570 ? 2% numa-vmstat.node1.nr_zone_active_anon
632.67 ? 12% slabinfo.Acpi-Parse.active_objs
632.67 ? 12% slabinfo.Acpi-Parse.num_objs
104939 ? 5% turbostat.C1
+5.8% 110.04 turbostat.RAMWatt
13.00 ? 4% vmstat.cpu.us
2596 vmstat.system.cs
38114 ? 2% proc-vmstat.nr_active_anon
11171 proc-vmstat.nr_mapped
43745 ? 2% proc-vmstat.nr_shmem
38114 ? 2% proc-vmstat.nr_zone_active_anon
56488 ? 3% proc-vmstat.pgactivate
0.01 ? 18% perf-sched.sch_delay.max.ms.devkmsg_read.vfs_read.ksys_read.do_syscall_64
5.58 ? 14% perf-sched.wait_and_delay.avg.ms.schedule_timeout.rcu_gp_kthread.kthread.ret_from_fork
122.00 ? 5% perf-sched.wait_and_delay.count.preempt_schedule_common.__cond_resched.kmem_cache_alloc.locks_alloc_lock.fcntl_setlk
188.50 ? 7% perf-sched.wait_and_delay.count.preempt_schedule_common.__cond_resched.kmem_cache_alloc.locks_alloc_lock.posix_lock_inode
1799 ? 13% perf-sched.wait_and_delay.count.schedule_timeout.rcu_gp_kthread.kthread.ret_from_fork
6337 perf-sched.wait_and_delay.max.ms.devkmsg_read.vfs_read.ksys_read.do_syscall_64
6337 perf-sched.wait_and_delay.max.ms.do_syslog.part.0.kmsg_read.vfs_read
6342 perf-sched.wait_and_delay.max.ms.pipe_read.new_sync_read.vfs_read.ksys_read
6036 perf-sched.wait_and_delay.max.ms.schedule_hrtimeout_range_clock.ep_poll.do_epoll_wait.__x64_sys_epoll_wait
5.57 ? 14% perf-sched.wait_time.avg.ms.schedule_timeout.rcu_gp_kthread.kthread.ret_from_fork
6337 perf-sched.wait_time.max.ms.devkmsg_read.vfs_read.ksys_read.do_syscall_64
6337 perf-sched.wait_time.max.ms.do_syslog.part.0.kmsg_read.vfs_read
0.40 ?180% perf-sched.wait_time.max.ms.exit_to_user_mode_prepare.irqentry_exit_to_user_mode.asm_sysvec_reschedule_ipi.[unknown]
6342 perf-sched.wait_time.max.ms.pipe_read.new_sync_read.vfs_read.ksys_read
6036 perf-sched.wait_time.max.ms.schedule_hrtimeout_range_clock.ep_poll.do_epoll_wait.__x64_sys_epoll_wait
692.50 ? 4% interrupts.CPU13.CAL:Function_call_interrupts
6045 ? 33% interrupts.CPU26.NMI:Non-maskable_interrupts
6045 ? 33% interrupts.CPU26.PMI:Performance_monitoring_interrupts
160.33 ? 38% interrupts.CPU41.RES:Rescheduling_interrupts
119.17 ? 49% interrupts.CPU42.RES:Rescheduling_interrupts
5511 ? 37% interrupts.CPU45.NMI:Non-maskable_interrupts
5511 ? 37% interrupts.CPU45.PMI:Performance_monitoring_interrupts
6432 ? 23% interrupts.CPU58.NMI:Non-maskable_interrupts
6432 ? 23% interrupts.CPU58.PMI:Performance_monitoring_interrupts
173.50 ? 40% interrupts.CPU6.RES:Rescheduling_interrupts
4652 ? 37% interrupts.CPU76.NMI:Non-maskable_interrupts
4652 ? 37% interrupts.CPU76.PMI:Performance_monitoring_interrupts
193.67 ? 27% interrupts.CPU81.RES:Rescheduling_interrupts
7273 ? 12% interrupts.CPU93.NMI:Non-maskable_interrupts
7273 ? 12% interrupts.CPU93.PMI:Performance_monitoring_interrupts
5857 ? 38% interrupts.CPU96.NMI:Non-maskable_interrupts
5857 ? 38% interrupts.CPU96.PMI:Performance_monitoring_interrupts
0.79 ? 20% perf-stat.i.MPKI
33.42 ? 4% perf-stat.i.cache-miss-rate%
32065121 ? 8% perf-stat.i.cache-misses
96305803 ? 9% perf-stat.i.cache-references
2443 perf-stat.i.context-switches
4827 ? 6% perf-stat.i.cycles-between-cache-misses
0.11 perf-stat.i.dTLB-load-miss-rate%
-33.6% 43563819 ? 5% perf-stat.i.dTLB-load-misses
37675 ? 3% perf-stat.i.dTLB-store-misses
-33.4% 45405593 ? 5% perf-stat.i.iTLB-load-misses
3058 ? 3% perf-stat.i.instructions-per-iTLB-miss
722.30 ? 16% perf-stat.i.metric.K/sec
7856755 ? 2% perf-stat.i.node-load-misses
936567 ? 3% perf-stat.i.node-loads
99.79 perf-stat.i.node-store-miss-rate%
9627739 perf-stat.i.node-store-misses
0.71 ? 14% perf-stat.overall.MPKI
0.49 ? 3% perf-stat.overall.branch-miss-rate%
33.35 ? 4% perf-stat.overall.cache-miss-rate%
4560 ? 8% perf-stat.overall.cycles-between-cache-misses
0.10 perf-stat.overall.dTLB-load-miss-rate%
0.00 ? 2% perf-stat.overall.dTLB-store-miss-rate%
3010 perf-stat.overall.instructions-per-iTLB-miss
89.34 perf-stat.overall.node-load-miss-rate%
99.91 perf-stat.overall.node-store-miss-rate%
946452 perf-stat.overall.path-length
31938893 ? 8% perf-stat.ps.cache-misses
95941923 ? 9% perf-stat.ps.cache-references
2448 perf-stat.ps.context-switches
-33.6% 43419292 ? 5% perf-stat.ps.dTLB-load-misses
37573 ? 3% perf-stat.ps.dTLB-store-misses
-33.4% 45252180 ? 5% perf-stat.ps.iTLB-load-misses
7828694 ? 2% perf-stat.ps.node-load-misses
933553 ? 3% perf-stat.ps.node-loads
9595407 perf-stat.ps.node-store-misses
10039 ? 12% softirqs.CPU0.RCU
10810 ? 19% softirqs.CPU1.RCU
11142 ? 14% softirqs.CPU10.RCU
8946 ? 13% softirqs.CPU100.RCU
8874 ? 28% softirqs.CPU101.RCU
7969 ? 13% softirqs.CPU102.RCU
10130 ? 17% softirqs.CPU103.RCU
9744 ? 22% softirqs.CPU11.RCU
8582 ? 22% softirqs.CPU12.RCU
9723 ? 15% softirqs.CPU13.RCU
9961 ? 8% softirqs.CPU14.RCU
9394 ? 11% softirqs.CPU15.RCU
10141 ? 19% softirqs.CPU16.RCU
10807 ? 13% softirqs.CPU17.RCU
9250 ? 18% softirqs.CPU18.RCU
10751 ? 10% softirqs.CPU19.RCU
9744 ? 22% softirqs.CPU2.RCU
9797 ? 16% softirqs.CPU20.RCU
10445 ? 24% softirqs.CPU21.RCU
10123 ? 26% softirqs.CPU22.RCU
10022 ? 23% softirqs.CPU23.RCU
10115 ? 21% softirqs.CPU24.RCU
9057 ? 18% softirqs.CPU25.RCU
8480 ? 13% softirqs.CPU26.RCU
9117 ? 16% softirqs.CPU27.RCU
8996 ? 21% softirqs.CPU28.RCU
8470 ? 18% softirqs.CPU29.RCU
25930 ? 30% softirqs.CPU29.SCHED
10577 ? 19% softirqs.CPU3.RCU
10057 ? 21% softirqs.CPU30.RCU
11040 ? 15% softirqs.CPU31.RCU
9662 ? 19% softirqs.CPU32.RCU
10144 ? 21% softirqs.CPU33.RCU
9890 ? 21% softirqs.CPU34.RCU
10177 ? 15% softirqs.CPU35.RCU
9616 ? 24% softirqs.CPU36.RCU
10106 ? 16% softirqs.CPU37.RCU
9341 ? 17% softirqs.CPU38.RCU
23087 ? 43% softirqs.CPU38.SCHED
9381 ? 17% softirqs.CPU39.RCU
10630 ? 20% softirqs.CPU4.RCU
10980 ? 20% softirqs.CPU40.RCU
9241 ? 17% softirqs.CPU41.RCU
8544 ? 12% softirqs.CPU42.RCU
26585 ? 30% softirqs.CPU42.SCHED
8991 ? 19% softirqs.CPU43.RCU
9433 ? 24% softirqs.CPU44.RCU
9035 ? 14% softirqs.CPU45.RCU
19464 ? 52% softirqs.CPU45.SCHED
9547 ? 24% softirqs.CPU46.RCU
9571 ? 15% softirqs.CPU47.RCU
8815 ? 20% softirqs.CPU48.RCU
8921 ? 11% softirqs.CPU49.RCU
10198 ? 12% softirqs.CPU5.RCU
9574 ? 20% softirqs.CPU50.RCU
8100 ? 14% softirqs.CPU51.RCU
10911 ? 16% softirqs.CPU52.RCU
8778 ? 25% softirqs.CPU53.RCU
9776 ? 18% softirqs.CPU54.RCU
8963 ? 20% softirqs.CPU55.RCU
8358 ? 18% softirqs.CPU56.RCU
9080 ? 30% softirqs.CPU57.RCU
10022 ? 23% softirqs.CPU58.RCU
21679 ? 41% softirqs.CPU58.SCHED
9466 ? 32% softirqs.CPU59.RCU
9150 ? 18% softirqs.CPU6.RCU
7880 ? 15% softirqs.CPU60.RCU
7623 ? 14% softirqs.CPU61.RCU
8455 ? 12% softirqs.CPU62.RCU
8566 ? 17% softirqs.CPU63.RCU
8837 ? 19% softirqs.CPU65.RCU
8802 ? 29% softirqs.CPU66.RCU
9797 ? 24% softirqs.CPU67.RCU
8818 ? 12% softirqs.CPU68.RCU
8497 ? 18% softirqs.CPU69.RCU
9700 ? 18% softirqs.CPU7.RCU
9545 ? 15% softirqs.CPU70.RCU
8933 ? 25% softirqs.CPU71.RCU
8624 ? 25% softirqs.CPU72.RCU
8172 ? 15% softirqs.CPU73.RCU
8875 ? 15% softirqs.CPU74.RCU
9547 ? 15% softirqs.CPU75.RCU
8673 ? 24% softirqs.CPU76.RCU
9383 ? 22% softirqs.CPU77.RCU
9292 ? 24% softirqs.CPU78.RCU
9460 ? 18% softirqs.CPU79.RCU
10616 ? 19% softirqs.CPU8.RCU
9393 ? 14% softirqs.CPU80.RCU
9562 ? 18% softirqs.CPU81.RCU
8761 ? 17% softirqs.CPU82.RCU
8189 ? 20% softirqs.CPU83.RCU
8881 ? 19% softirqs.CPU84.RCU
8421 ? 21% softirqs.CPU85.RCU
8604 ? 13% softirqs.CPU86.RCU
8154 ? 18% softirqs.CPU87.RCU
8727 ? 14% softirqs.CPU88.RCU
9187 ? 24% softirqs.CPU89.RCU
10441 ? 14% softirqs.CPU9.RCU
8666 ? 21% softirqs.CPU90.RCU
8815 ? 24% softirqs.CPU91.RCU
7867 ? 15% softirqs.CPU92.RCU
8979 ? 25% softirqs.CPU93.RCU
9504 ? 19% softirqs.CPU94.RCU
16384 ? 44% softirqs.CPU94.SCHED
9249 ? 16% softirqs.CPU95.RCU
8601 ? 12% softirqs.CPU96.RCU
8911 ? 27% softirqs.CPU97.RCU
8672 ? 19% softirqs.CPU98.RCU
8486 ? 18% softirqs.CPU99.RCU
971467 ? 15% softirqs.RCU
49404 ? 2% softirqs.TIMER
11.61 ? 9% perf-profile.calltrace.cycles-pp.syscall_exit_to_user_mode.do_syscall_64.entry_SYSCALL_64_after_hwframe.__libc_fcntl64
6.23 ? 9% perf-profile.calltrace.cycles-pp.__entry_text_start.__libc_fcntl64
5.64 ? 9% perf-profile.calltrace.cycles-pp.syscall_return_via_sysret.__libc_fcntl64
1.27 ? 29% perf-profile.calltrace.cycles-pp.entry_SYSCALL_64_safe_stack.__libc_fcntl64
1.70 ? 9% perf-profile.calltrace.cycles-pp.memset_erms.kmem_cache_alloc.locks_alloc_lock.posix_lock_inode.do_lock_file_wait
1.78 ? 9% perf-profile.calltrace.cycles-pp.security_file_lock.do_lock_file_wait.fcntl_setlk.do_fcntl.__x64_sys_fcntl
1.65 ? 9% perf-profile.calltrace.cycles-pp._copy_from_user.do_fcntl.__x64_sys_fcntl.do_syscall_64.entry_SYSCALL_64_after_hwframe
0.28 ?100% perf-profile.calltrace.cycles-pp._raw_spin_lock.posix_lock_inode.do_lock_file_wait.fcntl_setlk.do_fcntl
0.83 ? 9% perf-profile.calltrace.cycles-pp.memset_erms.kmem_cache_alloc.locks_alloc_lock.fcntl_setlk.do_fcntl
1.12 ? 10% perf-profile.calltrace.cycles-pp.common_file_perm.security_file_lock.do_lock_file_wait.fcntl_setlk.do_fcntl
0.38 ? 71% perf-profile.calltrace.cycles-pp.syscall_enter_from_user_mode.do_syscall_64.entry_SYSCALL_64_after_hwframe.__libc_fcntl64
0.28 ?100% perf-profile.calltrace.cycles-pp.locks_delete_lock_ctx.posix_lock_inode.do_lock_file_wait.fcntl_setlk.do_fcntl
0.69 ? 9% perf-profile.calltrace.cycles-pp.copy_user_generic_unrolled._copy_from_user.do_fcntl.__x64_sys_fcntl.do_syscall_64
0.59 ? 8% perf-profile.calltrace.cycles-pp.__fget_light.__x64_sys_fcntl.do_syscall_64.entry_SYSCALL_64_after_hwframe.__libc_fcntl64
0.59 ? 11% perf-profile.calltrace.cycles-pp.mod_objcg_state.kmem_cache_alloc.locks_alloc_lock.posix_lock_inode.do_lock_file_wait
1.92 ? 12% perf-profile.calltrace.cycles-pp.locks_dispose_list.posix_lock_inode.do_lock_file_wait.fcntl_setlk.do_fcntl
0.92 ? 10% perf-profile.calltrace.cycles-pp.get_obj_cgroup_from_current.kmem_cache_alloc.locks_alloc_lock.fcntl_setlk.do_fcntl
1.68 ? 14% perf-profile.calltrace.cycles-pp.kmem_cache_free.locks_dispose_list.posix_lock_inode.do_lock_file_wait.fcntl_setlk
1.38 ? 8% perf-profile.calltrace.cycles-pp.get_obj_cgroup_from_current.kmem_cache_alloc.locks_alloc_lock.posix_lock_inode.do_lock_file_wait
1.79 ? 59% perf-profile.calltrace.cycles-pp.page_counter_cancel.page_counter_uncharge.obj_cgroup_uncharge_pages.kmem_cache_free.fcntl_setlk
2.18 ? 58% perf-profile.calltrace.cycles-pp.page_counter_uncharge.obj_cgroup_uncharge_pages.kmem_cache_free.fcntl_setlk.do_fcntl
2.21 ? 58% perf-profile.calltrace.cycles-pp.obj_cgroup_uncharge_pages.kmem_cache_free.fcntl_setlk.do_fcntl.__x64_sys_fcntl
2.35 ? 43% perf-profile.calltrace.cycles-pp.page_counter_cancel.page_counter_uncharge.obj_cgroup_uncharge_pages.kmem_cache_free.posix_lock_inode
2.80 ? 59% perf-profile.calltrace.cycles-pp.obj_cgroup_charge_pages.obj_cgroup_charge.kmem_cache_alloc.locks_alloc_lock.fcntl_setlk
2.84 ? 41% perf-profile.calltrace.cycles-pp.page_counter_uncharge.obj_cgroup_uncharge_pages.kmem_cache_free.posix_lock_inode.do_lock_file_wait
2.86 ? 41% perf-profile.calltrace.cycles-pp.obj_cgroup_uncharge_pages.kmem_cache_free.posix_lock_inode.do_lock_file_wait.fcntl_setlk
3.98 ? 32% perf-profile.calltrace.cycles-pp.kmem_cache_free.fcntl_setlk.do_fcntl.__x64_sys_fcntl.do_syscall_64
2.99 ? 55% perf-profile.calltrace.cycles-pp.obj_cgroup_charge.kmem_cache_alloc.locks_alloc_lock.fcntl_setlk.do_fcntl
5.49 ? 21% perf-profile.calltrace.cycles-pp.kmem_cache_free.posix_lock_inode.do_lock_file_wait.fcntl_setlk.do_fcntl
7.42 ? 22% perf-profile.calltrace.cycles-pp.locks_alloc_lock.fcntl_setlk.do_fcntl.__x64_sys_fcntl.do_syscall_64
7.17 ? 23% perf-profile.calltrace.cycles-pp.kmem_cache_alloc.locks_alloc_lock.fcntl_setlk.do_fcntl.__x64_sys_fcntl
4.61 ? 29% perf-profile.calltrace.cycles-pp.obj_cgroup_charge_pages.obj_cgroup_charge.kmem_cache_alloc.locks_alloc_lock.posix_lock_inode
5.00 ? 27% perf-profile.calltrace.cycles-pp.obj_cgroup_charge.kmem_cache_alloc.locks_alloc_lock.posix_lock_inode.do_lock_file_wait
12.76 ? 12% perf-profile.calltrace.cycles-pp.locks_alloc_lock.posix_lock_inode.do_lock_file_wait.fcntl_setlk.do_fcntl
12.21 ? 13% perf-profile.calltrace.cycles-pp.kmem_cache_alloc.locks_alloc_lock.posix_lock_inode.do_lock_file_wait.fcntl_setlk
7.35 ? 35% perf-profile.calltrace.cycles-pp.page_counter_try_charge.obj_cgroup_charge_pages.obj_cgroup_charge.kmem_cache_alloc.locks_alloc_lock
25.78 ? 11% perf-profile.calltrace.cycles-pp.do_lock_file_wait.fcntl_setlk.do_fcntl.__x64_sys_fcntl.do_syscall_64
23.35 ? 12% perf-profile.calltrace.cycles-pp.posix_lock_inode.do_lock_file_wait.fcntl_setlk.do_fcntl.__x64_sys_fcntl
41.58 ? 12% perf-profile.calltrace.cycles-pp.__x64_sys_fcntl.do_syscall_64.entry_SYSCALL_64_after_hwframe.__libc_fcntl64
40.53 ? 12% perf-profile.calltrace.cycles-pp.do_fcntl.__x64_sys_fcntl.do_syscall_64.entry_SYSCALL_64_after_hwframe.__libc_fcntl64
38.56 ? 13% perf-profile.calltrace.cycles-pp.fcntl_setlk.do_fcntl.__x64_sys_fcntl.do_syscall_64.entry_SYSCALL_64_after_hwframe
11.68 ? 9% perf-profile.children.cycles-pp.syscall_exit_to_user_mode
6.41 ? 9% perf-profile.children.cycles-pp.syscall_return_via_sysret
5.96 ? 9% perf-profile.children.cycles-pp.__entry_text_start
2.53 ? 9% perf-profile.children.cycles-pp.memset_erms
1.69 ? 9% perf-profile.children.cycles-pp._copy_from_user
1.80 ? 9% perf-profile.children.cycles-pp.security_file_lock
1.33 ? 9% perf-profile.children.cycles-pp._raw_spin_lock
0.85 ? 10% perf-profile.children.cycles-pp.___might_sleep
1.16 ? 10% perf-profile.children.cycles-pp.common_file_perm
0.98 ? 9% perf-profile.children.cycles-pp.entry_SYSCALL_64_safe_stack
0.81 ? 9% perf-profile.children.cycles-pp.copy_user_generic_unrolled
0.67 ? 9% perf-profile.children.cycles-pp.__might_sleep
0.56 ? 9% perf-profile.children.cycles-pp.syscall_enter_from_user_mode
0.59 ? 8% perf-profile.children.cycles-pp.__fget_light
0.50 ? 8% perf-profile.children.cycles-pp.apparmor_file_lock
0.51 ? 9% perf-profile.children.cycles-pp.__cond_resched
0.38 ? 9% perf-profile.children.cycles-pp.__might_fault
0.46 ? 9% perf-profile.children.cycles-pp.locks_release_private
0.36 ? 9% perf-profile.children.cycles-pp.copy_user_enhanced_fast_string
0.44 ? 10% perf-profile.children.cycles-pp.locks_insert_lock_ctx
0.54 ? 9% perf-profile.children.cycles-pp.locks_delete_lock_ctx
0.44 ? 8% perf-profile.children.cycles-pp.locks_unlink_lock_ctx
0.26 ? 10% perf-profile.children.cycles-pp.rcu_all_qs
0.21 ? 9% perf-profile.children.cycles-pp.flock64_to_posix_lock
0.24 ? 10% perf-profile.children.cycles-pp.aa_file_perm
0.16 ? 9% perf-profile.children.cycles-pp.__x86_indirect_thunk_rax
0.13 ? 8% perf-profile.children.cycles-pp.__init_waitqueue_head
0.21 ? 9% perf-profile.children.cycles-pp.exit_to_user_mode_prepare
0.17 ? 6% perf-profile.children.cycles-pp.testcase
0.12 ? 10% perf-profile.children.cycles-pp.should_failslab
0.09 ? 10% perf-profile.children.cycles-pp.fcntl@plt
0.10 ? 14% perf-profile.children.cycles-pp.syscall_exit_to_user_mode_prepare
0.08 ? 5% perf-profile.children.cycles-pp.security_file_fcntl
0.08 ? 8% perf-profile.children.cycles-pp.__list_del_entry_valid
0.10 ? 10% perf-profile.children.cycles-pp.locks_get_lock_context
0.03 ? 70% perf-profile.children.cycles-pp.vfs_lock_file
0.10 ? 7% perf-profile.children.cycles-pp.locks_delete_block
0.06 ? 14% perf-profile.children.cycles-pp.__list_add_valid
0.06 ? 7% perf-profile.children.cycles-pp.memset
0.08 ? 9% perf-profile.children.cycles-pp.obj_cgroup_uncharge
0.21 ? 18% perf-profile.children.cycles-pp.mem_cgroup_from_task
0.55 ? 9% perf-profile.children.cycles-pp.rcu_read_unlock_strict
0.60 ? 9% perf-profile.children.cycles-pp.refill_obj_stock
1.96 ? 12% perf-profile.children.cycles-pp.locks_dispose_list
1.67 ? 10% perf-profile.children.cycles-pp.mod_objcg_state
2.42 ? 35% perf-profile.children.cycles-pp.propagate_protected_usage
2.47 ? 9% perf-profile.children.cycles-pp.get_obj_cgroup_from_current
4.74 ? 36% perf-profile.children.cycles-pp.page_counter_cancel
5.77 ? 35% perf-profile.children.cycles-pp.page_counter_uncharge
5.82 ? 35% perf-profile.children.cycles-pp.obj_cgroup_uncharge_pages
7.36 ? 35% perf-profile.children.cycles-pp.page_counter_try_charge
7.42 ? 35% perf-profile.children.cycles-pp.obj_cgroup_charge_pages
11.27 ? 18% perf-profile.children.cycles-pp.kmem_cache_free
8.05 ? 32% perf-profile.children.cycles-pp.obj_cgroup_charge
25.81 ? 11% perf-profile.children.cycles-pp.do_lock_file_wait
23.56 ? 12% perf-profile.children.cycles-pp.posix_lock_inode
20.24 ? 14% perf-profile.children.cycles-pp.locks_alloc_lock
19.67 ? 14% perf-profile.children.cycles-pp.kmem_cache_alloc
41.61 ? 12% perf-profile.children.cycles-pp.__x64_sys_fcntl
40.63 ? 12% perf-profile.children.cycles-pp.do_fcntl
38.67 ? 13% perf-profile.children.cycles-pp.fcntl_setlk
11.40 ? 9% perf-profile.self.cycles-pp.syscall_exit_to_user_mode
6.39 ? 9% perf-profile.self.cycles-pp.syscall_return_via_sysret
5.16 ? 9% perf-profile.self.cycles-pp.__entry_text_start
2.45 ? 9% perf-profile.self.cycles-pp.memset_erms
1.32 ? 10% perf-profile.self.cycles-pp._raw_spin_lock
1.21 ? 9% perf-profile.self.cycles-pp.__libc_fcntl64
0.82 ? 11% perf-profile.self.cycles-pp.___might_sleep
0.94 ? 10% perf-profile.self.cycles-pp.common_file_perm
0.79 ? 9% perf-profile.self.cycles-pp.copy_user_generic_unrolled
0.96 ? 8% perf-profile.self.cycles-pp.posix_lock_inode
0.58 ? 9% perf-profile.self.cycles-pp.__might_sleep
0.76 ? 8% perf-profile.self.cycles-pp.fcntl_setlk
0.43 ? 10% perf-profile.self.cycles-pp.locks_alloc_lock
0.51 ? 10% perf-profile.self.cycles-pp.syscall_enter_from_user_mode
0.49 ? 6% perf-profile.self.cycles-pp.entry_SYSCALL_64_after_hwframe
0.56 ? 8% perf-profile.self.cycles-pp.__fget_light
0.46 ? 8% perf-profile.self.cycles-pp.entry_SYSCALL_64_safe_stack
0.43 ? 7% perf-profile.self.cycles-pp.apparmor_file_lock
0.42 ? 9% perf-profile.self.cycles-pp.locks_release_private
0.34 ? 9% perf-profile.self.cycles-pp.copy_user_enhanced_fast_string
0.30 ? 9% perf-profile.self.cycles-pp.do_lock_file_wait
0.38 ? 9% perf-profile.self.cycles-pp.do_fcntl
0.30 ? 8% perf-profile.self.cycles-pp.__x64_sys_fcntl
0.25 ? 11% perf-profile.self.cycles-pp.do_syscall_64
0.19 ? 10% perf-profile.self.cycles-pp.flock64_to_posix_lock
0.25 ? 9% perf-profile.self.cycles-pp.__cond_resched
0.14 ? 10% perf-profile.self.cycles-pp.locks_dispose_list
0.13 ? 8% perf-profile.self.cycles-pp.testcase
0.19 ? 11% perf-profile.self.cycles-pp.aa_file_perm
0.13 ? 8% perf-profile.self.cycles-pp.__x86_indirect_thunk_rax
0.18 ? 10% perf-profile.self.cycles-pp.exit_to_user_mode_prepare
0.16 ? 10% perf-profile.self.cycles-pp._copy_from_user
0.17 ? 10% perf-profile.self.cycles-pp.rcu_all_qs
0.10 ? 10% perf-profile.self.cycles-pp.__init_waitqueue_head
0.07 ? 14% perf-profile.self.cycles-pp.syscall_exit_to_user_mode_prepare
0.08 ? 11% perf-profile.self.cycles-pp.fcntl@plt
0.07 ? 7% perf-profile.self.cycles-pp.__list_del_entry_valid
0.13 ? 8% perf-profile.self.cycles-pp.security_file_lock
0.03 ?100% perf-profile.self.cycles-pp.__list_add_valid
0.08 ? 11% perf-profile.self.cycles-pp.locks_get_lock_context
0.06 ? 11% perf-profile.self.cycles-pp.__might_fault
0.06 ? 7% perf-profile.self.cycles-pp.memset
0.06 ? 7% perf-profile.self.cycles-pp.security_file_fcntl
0.08 ? 8% perf-profile.self.cycles-pp.locks_delete_block
0.06 ? 7% perf-profile.self.cycles-pp.obj_cgroup_uncharge
0.16 ? 24% perf-profile.self.cycles-pp.mem_cgroup_from_task
0.34 ? 9% perf-profile.self.cycles-pp.rcu_read_unlock_strict
0.56 ? 9% perf-profile.self.cycles-pp.obj_cgroup_charge
0.58 ? 10% perf-profile.self.cycles-pp.refill_obj_stock
4.46 ? 10% perf-profile.self.cycles-pp.kmem_cache_alloc
1.65 ? 10% perf-profile.self.cycles-pp.mod_objcg_state
2.17 ? 9% perf-profile.self.cycles-pp.get_obj_cgroup_from_current
2.40 ? 35% perf-profile.self.cycles-pp.propagate_protected_usage
4.70 ? 36% perf-profile.self.cycles-pp.page_counter_cancel
5.92 ? 35% perf-profile.self.cycles-pp.page_counter_try_charge

will-it-scale.52.processes

-------------------------------------------------+
+ + + + |
+++.+ +++++.+++++++.++++ ++.++ +++++.+++++++ |
|
|
|
|
|
|
OOO OO O O OOO O |
O O OOO O OOOOO O O O O OO O OOOO O |
O O OO O O O|
O |
|
-------------------------------------------------+

will-it-scale.per_process_ops

-------------------------------------------------+
++++.+++++++.+++++++.+++++++.++++++++.+++++++ |
|
|
|
|
|
|
|
O O O O O O O O O |
O OOOO O O O OO O O OOOO |
O O O OO OO O OO O|
O |
|
-------------------------------------------------+

will-it-scale.workload

-------------------------------------------------+
+ + + + |
+++.+ +++++.+++++++.++++ ++.++ +++++.+++++++ |
|
|
|
|
|
|
OOO OO O O OOO O |
O O OOO O OOOOO O O O O OO O OOOO O |
O O OO O O O|
O |
|
-------------------------------------------------+

************************************************************************
threads 2 sockets Intel(R) Xeon(R) Gold 6338 CPU @ 2.00GHz with 256G memory
==============================================================
config/mode/nr_task/rootfs/tbox_group/test/testcase/ucode:
rhel-8.3/process/50%/debian-10.4-x86_64-20200603.cgz/lkp-icl-2sp2/lock1/will-it-scale/0xd000280
enable accounting for pollfd and select bits arrays")
enable accounting for file lock caches")
2862ac54235fc7270
-----------------
%stddev
\
-40.9% 1.055e+08 ? 2% will-it-scale.64.processes
-40.9% 1649196 ? 2% will-it-scale.per_process_ops
-40.9% 1.055e+08 ? 2% will-it-scale.workload
422206 ? 9% numa-numastat.node0.local_node
2874 vmstat.system.cs
0.02 ? 2% mpstat.cpu.all.soft%
3.15 ? 7% mpstat.cpu.all.usr%
778.67 ? 13% slabinfo.Acpi-Parse.active_objs
778.67 ? 13% slabinfo.Acpi-Parse.num_objs
51541 proc-vmstat.nr_active_anon
58037 proc-vmstat.nr_shmem
51541 proc-vmstat.nr_zone_active_anon
773.33 ?101% interrupts.154:IR-PCI-MSI.25690117-edge.eth0-TxRx-5
1223 ? 30% interrupts.CPU103.CAL:Function_call_interrupts
4155 ? 50% interrupts.CPU2.RES:Rescheduling_interrupts
2818 ? 53% interrupts.CPU4.RES:Rescheduling_interrupts
773.33 ?101% interrupts.CPU5.154:IR-PCI-MSI.25690117-edge.eth0-TxRx-5
1058 ? 30% interrupts.CPU55.CAL:Function_call_interrupts
3738 ? 82% interrupts.CPU55.RES:Rescheduling_interrupts
138.17 ? 19% interrupts.CPU86.NMI:Non-maskable_interrupts
138.17 ? 19% interrupts.CPU86.PMI:Performance_monitoring_interrupts
1561 ?200% interrupts.CPU99.NMI:Non-maskable_interrupts
1561 ?200% interrupts.CPU99.PMI:Performance_monitoring_interrupts
2063 perf-sched.wait_and_delay.avg.ms.devkmsg_read.vfs_read.ksys_read.do_syscall_64
2063 perf-sched.wait_and_delay.avg.ms.do_syslog.part.0.kmsg_read.vfs_read
150.13 ? 12% perf-sched.wait_and_delay.avg.ms.schedule_hrtimeout_range_clock.poll_schedule_timeout.constprop.0.do_sys_poll
468.33 ? 5% perf-sched.wait_and_delay.count.exit_to_user_mode_prepare.syscall_exit_to_user_mode.do_syscall_64.entry_SYSCALL_64_after_hwframe
262.50 ? 8% perf-sched.wait_and_delay.count.preempt_schedule_common.__cond_resched.kmem_cache_alloc.locks_alloc_lock.posix_lock_inode
202.50 ? 4% perf-sched.wait_and_delay.count.schedule_hrtimeout_range_clock.poll_schedule_timeout.constprop.0.do_sys_poll
6185 perf-sched.wait_and_delay.max.ms.devkmsg_read.vfs_read.ksys_read.do_syscall_64
6185 perf-sched.wait_and_delay.max.ms.do_syslog.part.0.kmsg_read.vfs_read
6189 perf-sched.wait_and_delay.max.ms.pipe_read.new_sync_read.vfs_read.ksys_read
2063 perf-sched.wait_time.avg.ms.devkmsg_read.vfs_read.ksys_read.do_syscall_64
2063 perf-sched.wait_time.avg.ms.do_syslog.part.0.kmsg_read.vfs_read
1.31 ? 18% perf-sched.wait_time.avg.ms.rcu_gp_kthread.kthread.ret_from_fork
150.13 ? 12% perf-sched.wait_time.avg.ms.schedule_hrtimeout_range_clock.poll_schedule_timeout.constprop.0.do_sys_poll
6185 perf-sched.wait_time.max.ms.devkmsg_read.vfs_read.ksys_read.do_syscall_64
6185 perf-sched.wait_time.max.ms.do_syslog.part.0.kmsg_read.vfs_read
6189 perf-sched.wait_time.max.ms.pipe_read.new_sync_read.vfs_read.ksys_read
0.50 ?128% perf-sched.wait_time.max.ms.preempt_schedule_common.__cond_resched.posix_lock_inode.do_lock_file_wait.fcntl_setlk
0.27 ? 6% perf-stat.i.MPKI
-6.5% 6.615e+10 ? 2% perf-stat.i.branch-instructions
0.08 ? 5% perf-stat.i.branch-miss-rate%
+45.6% 52354202 ? 5% perf-stat.i.branch-misses
28.33 ? 8% perf-stat.i.cache-miss-rate%
24022051 ? 4% perf-stat.i.cache-misses
86349315 ? 4% perf-stat.i.cache-references
2716 perf-stat.i.context-switches
0.51 ? 2% perf-stat.i.cpi
6992 ? 5% perf-stat.i.cycles-between-cache-misses
0.00 ? 11% perf-stat.i.dTLB-load-miss-rate%
-3.8% 9.907e+10 ? 2% perf-stat.i.dTLB-loads
0.00 ? 8% perf-stat.i.dTLB-store-miss-rate%
-11.3% 5.796e+10 ? 2% perf-stat.i.dTLB-stores
-7.2% 3.263e+11 ? 2% perf-stat.i.instructions
1.96 ? 2% perf-stat.i.ipc
821.35 ? 3% perf-stat.i.metric.K/sec
1743 ? 2% perf-stat.i.metric.M/sec
7122311 ? 3% perf-stat.i.node-load-misses
475820 ? 29% perf-stat.i.node-loads
96.80 perf-stat.i.node-store-miss-rate%
11186859 ? 6% perf-stat.i.node-store-misses
350742 ? 9% perf-stat.i.node-stores
0.27 ? 6% perf-stat.overall.MPKI
0.08 ? 6% perf-stat.overall.branch-miss-rate%
27.92 ? 8% perf-stat.overall.cache-miss-rate%
0.51 ? 2% perf-stat.overall.cpi
6940 ? 4% perf-stat.overall.cycles-between-cache-misses
0.00 ? 10% perf-stat.overall.dTLB-load-miss-rate%
0.00 ? 4% perf-stat.overall.dTLB-store-miss-rate%
1.96 ? 2% perf-stat.overall.ipc
93.76 perf-stat.overall.node-load-miss-rate%
96.95 perf-stat.overall.node-store-miss-rate%
932831 perf-stat.overall.path-length
-6.5% 6.592e+10 ? 2% perf-stat.ps.branch-instructions
+45.6% 52233576 ? 5% perf-stat.ps.branch-misses
23944727 ? 4% perf-stat.ps.cache-misses
86115077 ? 4% perf-stat.ps.cache-references
2721 perf-stat.ps.context-switches
-3.8% 9.872e+10 ? 2% perf-stat.ps.dTLB-loads
-11.3% 5.775e+10 ? 2% perf-stat.ps.dTLB-stores
-7.2% 3.252e+11 ? 2% perf-stat.ps.instructions
7097622 ? 3% perf-stat.ps.node-load-misses
475371 ? 29% perf-stat.ps.node-loads
11148161 ? 6% perf-stat.ps.node-store-misses
350139 ? 8% perf-stat.ps.node-stores
-7.2% 9.845e+13 ? 2% perf-stat.total.instructions
7926 ? 19% softirqs.CPU1.RCU
7741 ? 11% softirqs.CPU10.RCU
5851 ? 8% softirqs.CPU100.RCU
5811 ? 5% softirqs.CPU101.RCU
5752 ? 7% softirqs.CPU102.RCU
6376 ? 17% softirqs.CPU103.RCU
5857 ? 9% softirqs.CPU104.RCU
5529 ? 7% softirqs.CPU105.RCU
5749 ? 5% softirqs.CPU106.RCU
6138 ? 7% softirqs.CPU108.RCU
6210 ? 16% softirqs.CPU109.RCU
6726 ? 14% softirqs.CPU11.RCU
5916 ? 12% softirqs.CPU111.RCU
6187 ? 10% softirqs.CPU113.RCU
6220 ? 13% softirqs.CPU114.RCU
6798 ? 18% softirqs.CPU115.RCU
6090 ? 12% softirqs.CPU116.RCU
5912 ? 15% softirqs.CPU117.RCU
6103 ? 8% softirqs.CPU118.RCU
5923 ? 8% softirqs.CPU119.RCU
7197 ? 6% softirqs.CPU12.RCU
5973 ? 12% softirqs.CPU120.RCU
5457 ? 7% softirqs.CPU121.RCU
5604 ? 11% softirqs.CPU124.RCU
6186 ? 13% softirqs.CPU125.RCU
5745 ? 11% softirqs.CPU126.RCU
7057 ? 10% softirqs.CPU127.RCU
6958 ? 6% softirqs.CPU13.RCU
6967 ? 12% softirqs.CPU14.RCU
6716 ? 10% softirqs.CPU15.RCU
6648 ? 16% softirqs.CPU16.RCU
6435 ? 13% softirqs.CPU17.RCU
7319 ? 13% softirqs.CPU19.RCU
6996 ? 10% softirqs.CPU2.RCU
7416 ? 9% softirqs.CPU20.RCU
6745 ? 5% softirqs.CPU21.RCU
7451 ? 12% softirqs.CPU22.RCU
6523 ? 13% softirqs.CPU23.RCU
6997 ? 15% softirqs.CPU24.RCU
7061 ? 8% softirqs.CPU25.RCU
6715 ? 11% softirqs.CPU26.RCU
6613 ? 11% softirqs.CPU27.RCU
6658 ? 10% softirqs.CPU28.RCU
6907 ? 8% softirqs.CPU29.RCU
7009 ? 13% softirqs.CPU3.RCU
7012 ? 9% softirqs.CPU30.RCU
6910 ? 17% softirqs.CPU31.RCU
6874 ? 11% softirqs.CPU32.RCU
7032 ? 19% softirqs.CPU33.RCU
7317 ? 16% softirqs.CPU34.RCU
7125 ? 6% softirqs.CPU35.RCU
6625 ? 14% softirqs.CPU36.RCU
7414 ? 12% softirqs.CPU37.RCU
7690 ? 14% softirqs.CPU38.RCU
7461 ? 9% softirqs.CPU39.RCU
7516 ? 12% softirqs.CPU4.RCU
7220 ? 12% softirqs.CPU40.RCU
7273 ? 7% softirqs.CPU41.RCU
6891 ? 9% softirqs.CPU42.RCU
6921 ? 10% softirqs.CPU43.RCU
6739 ? 17% softirqs.CPU44.RCU
7233 ? 11% softirqs.CPU45.RCU
6751 ? 12% softirqs.CPU46.RCU
7073 ? 8% softirqs.CPU47.RCU
6346 ? 8% softirqs.CPU48.RCU
6180 ? 6% softirqs.CPU49.RCU
7563 ? 17% softirqs.CPU5.RCU
6303 ? 7% softirqs.CPU50.RCU
6096 ? 7% softirqs.CPU51.RCU
6575 ? 7% softirqs.CPU52.RCU
6609 ? 10% softirqs.CPU53.RCU
6655 ? 5% softirqs.CPU54.RCU
6047 ? 5% softirqs.CPU55.RCU
6350 ? 7% softirqs.CPU56.RCU
6615 ? 8% softirqs.CPU57.RCU
6492 ? 5% softirqs.CPU58.RCU
6351 ? 9% softirqs.CPU59.RCU
8102 ? 22% softirqs.CPU6.RCU
6340 ? 11% softirqs.CPU60.RCU
5915 ? 7% softirqs.CPU61.RCU
6436 ? 3% softirqs.CPU62.RCU
6774 ? 9% softirqs.CPU64.RCU
6372 ? 8% softirqs.CPU65.RCU
6752 ? 5% softirqs.CPU66.RCU
6463 ? 10% softirqs.CPU67.RCU
6307 ? 6% softirqs.CPU68.RCU
6310 ? 7% softirqs.CPU69.RCU
6809 ? 12% softirqs.CPU7.RCU
19745 ? 32% softirqs.CPU7.SCHED
6521 ? 7% softirqs.CPU70.RCU
6424 ? 2% softirqs.CPU71.RCU
21764 ? 30% softirqs.CPU71.SCHED
6524 ? 7% softirqs.CPU72.RCU
6055 ? 7% softirqs.CPU73.RCU
5741 ? 9% softirqs.CPU74.RCU
6176 ? 6% softirqs.CPU75.RCU
6220 ? 10% softirqs.CPU76.RCU
6946 ? 14% softirqs.CPU78.RCU
5956 ? 12% softirqs.CPU79.RCU
7105 ? 20% softirqs.CPU8.RCU
6439 ? 11% softirqs.CPU80.RCU
6430 ? 3% softirqs.CPU81.RCU
6098 ? 10% softirqs.CPU82.RCU
6370 ? 7% softirqs.CPU83.RCU
6229 ? 6% softirqs.CPU84.RCU
6396 ? 9% softirqs.CPU85.RCU
6339 ? 14% softirqs.CPU86.RCU
6408 ? 8% softirqs.CPU87.RCU
6059 ? 3% softirqs.CPU88.RCU
6534 ? 12% softirqs.CPU89.RCU
7246 ? 11% softirqs.CPU9.RCU
6612 ? 6% softirqs.CPU90.RCU
25496 ? 25% softirqs.CPU90.SCHED
6340 ? 7% softirqs.CPU91.RCU
6531 ? 5% softirqs.CPU92.RCU
6479 ? 5% softirqs.CPU93.RCU
6064 ? 7% softirqs.CPU94.RCU
6544 ? 12% softirqs.CPU95.RCU
5845 ? 5% softirqs.CPU96.RCU
5814 ? 8% softirqs.CPU97.RCU
6064 ? 9% softirqs.CPU98.RCU
5661 ? 3% softirqs.CPU99.RCU
835466 ? 3% softirqs.RCU
57361 ? 4% softirqs.TIMER
5.08 ? 18% perf-profile.calltrace.cycles-pp.__entry_text_start.__libc_fcntl64
4.38 ? 6% perf-profile.calltrace.cycles-pp._copy_from_user.do_fcntl.__x64_sys_fcntl.do_syscall_64.entry_SYSCALL_64_after_hwframe
2.58 ? 5% perf-profile.calltrace.cycles-pp.memset_erms.kmem_cache_alloc.locks_alloc_lock.posix_lock_inode.do_lock_file_wait
2.48 ? 7% perf-profile.calltrace.cycles-pp.security_file_lock.do_lock_file_wait.fcntl_setlk.do_fcntl.__x64_sys_fcntl
1.88 ? 7% perf-profile.calltrace.cycles-pp.common_file_perm.security_file_lock.do_lock_file_wait.fcntl_setlk.do_fcntl
1.93 ? 6% perf-profile.calltrace.cycles-pp.copy_user_generic_unrolled._copy_from_user.do_fcntl.__x64_sys_fcntl.do_syscall_64
1.75 ? 6% perf-profile.calltrace.cycles-pp.__fget_light.__x64_sys_fcntl.do_syscall_64.entry_SYSCALL_64_after_hwframe.__libc_fcntl64
1.30 ? 7% perf-profile.calltrace.cycles-pp.locks_delete_lock_ctx.posix_lock_inode.do_lock_file_wait.fcntl_setlk.do_fcntl
1.36 ? 6% perf-profile.calltrace.cycles-pp._raw_spin_lock.posix_lock_inode.do_lock_file_wait.fcntl_setlk.do_fcntl
1.05 ? 6% perf-profile.calltrace.cycles-pp.locks_unlink_lock_ctx.locks_delete_lock_ctx.posix_lock_inode.do_lock_file_wait.fcntl_setlk
1.06 ? 7% perf-profile.calltrace.cycles-pp.locks_insert_lock_ctx.posix_lock_inode.do_lock_file_wait.fcntl_setlk.do_fcntl
0.37 ? 70% perf-profile.calltrace.cycles-pp.locks_release_private.posix_lock_inode.do_lock_file_wait.fcntl_setlk.do_fcntl
0.98 ? 7% perf-profile.calltrace.cycles-pp.copy_user_enhanced_fast_string._copy_from_user.do_fcntl.__x64_sys_fcntl.do_syscall_64
0.64 ? 7% perf-profile.calltrace.cycles-pp._raw_spin_lock.locks_insert_lock_ctx.posix_lock_inode.do_lock_file_wait.fcntl_setlk
1.57 ? 8% perf-profile.calltrace.cycles-pp.memset_erms.kmem_cache_alloc.locks_alloc_lock.fcntl_setlk.do_fcntl
0.77 ? 6% perf-profile.calltrace.cycles-pp.___might_sleep.kmem_cache_alloc.locks_alloc_lock.fcntl_setlk.do_fcntl
0.75 ? 7% perf-profile.calltrace.cycles-pp.__might_fault._copy_from_user.do_fcntl.__x64_sys_fcntl.do_syscall_64
0.68 ? 6% perf-profile.calltrace.cycles-pp._raw_spin_lock.locks_unlink_lock_ctx.locks_delete_lock_ctx.posix_lock_inode.do_lock_file_wait
0.68 ? 7% perf-profile.calltrace.cycles-pp.aa_file_perm.common_file_perm.security_file_lock.do_lock_file_wait.fcntl_setlk
0.90 ? 9% perf-profile.calltrace.cycles-pp.syscall_exit_to_user_mode.do_syscall_64.entry_SYSCALL_64_after_hwframe.__libc_fcntl64
0.70 ? 6% perf-profile.calltrace.cycles-pp.syscall_enter_from_user_mode.do_syscall_64.entry_SYSCALL_64_after_hwframe.__libc_fcntl64
0.61 ? 7% perf-profile.calltrace.cycles-pp.flock64_to_posix_lock.fcntl_setlk.do_fcntl.__x64_sys_fcntl.do_syscall_64
0.35 ? 70% perf-profile.calltrace.cycles-pp.___might_sleep.kmem_cache_alloc.locks_alloc_lock.posix_lock_inode.do_lock_file_wait
0.86 ? 5% perf-profile.calltrace.cycles-pp.__might_sleep.kmem_cache_alloc.locks_alloc_lock.posix_lock_inode.do_lock_file_wait
0.66 ? 10% perf-profile.calltrace.cycles-pp.syscall_return_via_sysret.__libc_fcntl64
0.63 ? 5% perf-profile.calltrace.cycles-pp._raw_spin_lock.fcntl_setlk.do_fcntl.__x64_sys_fcntl.do_syscall_64
98.07 perf-profile.calltrace.cycles-pp.__libc_fcntl64
0.54 ? 4% perf-profile.calltrace.cycles-pp.refill_obj_stock.kmem_cache_free.fcntl_setlk.do_fcntl.__x64_sys_fcntl
0.83 ? 7% perf-profile.calltrace.cycles-pp.mod_objcg_state.kmem_cache_free.fcntl_setlk.do_fcntl.__x64_sys_fcntl
0.90 ? 7% perf-profile.calltrace.cycles-pp.mod_objcg_state.kmem_cache_alloc.locks_alloc_lock.fcntl_setlk.do_fcntl
0.92 ? 5% perf-profile.calltrace.cycles-pp.refill_obj_stock.kmem_cache_free.posix_lock_inode.do_lock_file_wait.fcntl_setlk
1.26 ? 7% perf-profile.calltrace.cycles-pp.mod_objcg_state.kmem_cache_free.posix_lock_inode.do_lock_file_wait.fcntl_setlk
3.54 ? 14% perf-profile.calltrace.cycles-pp.locks_dispose_list.posix_lock_inode.do_lock_file_wait.fcntl_setlk.do_fcntl
1.59 ? 11% perf-profile.calltrace.cycles-pp.get_obj_cgroup_from_current.kmem_cache_alloc.locks_alloc_lock.fcntl_setlk.do_fcntl
2.90 ? 18% perf-profile.calltrace.cycles-pp.kmem_cache_free.locks_dispose_list.posix_lock_inode.do_lock_file_wait.fcntl_setlk
1.83 ? 6% perf-profile.calltrace.cycles-pp.mod_objcg_state.kmem_cache_alloc.locks_alloc_lock.posix_lock_inode.do_lock_file_wait
2.13 ? 29% perf-profile.calltrace.cycles-pp.propagate_protected_usage.page_counter_try_charge.obj_cgroup_charge_pages.obj_cgroup_charge.kmem_cache_alloc
2.16 ? 29% perf-profile.calltrace.cycles-pp.page_counter_cancel.page_counter_uncharge.obj_cgroup_uncharge_pages.kmem_cache_free.fcntl_setlk
2.42 ? 7% perf-profile.calltrace.cycles-pp.get_obj_cgroup_from_current.kmem_cache_alloc.locks_alloc_lock.posix_lock_inode.do_lock_file_wait
2.60 ? 28% perf-profile.calltrace.cycles-pp.page_counter_uncharge.obj_cgroup_uncharge_pages.kmem_cache_free.fcntl_setlk.do_fcntl
2.66 ? 27% perf-profile.calltrace.cycles-pp.obj_cgroup_uncharge_pages.kmem_cache_free.fcntl_setlk.do_fcntl.__x64_sys_fcntl
3.04 ? 37% perf-profile.calltrace.cycles-pp.page_counter_cancel.page_counter_uncharge.obj_cgroup_uncharge_pages.kmem_cache_free.posix_lock_inode
6.42 ? 10% perf-profile.calltrace.cycles-pp.kmem_cache_free.fcntl_setlk.do_fcntl.__x64_sys_fcntl.do_syscall_64
3.59 ? 33% perf-profile.calltrace.cycles-pp.obj_cgroup_charge_pages.obj_cgroup_charge.kmem_cache_alloc.locks_alloc_lock.fcntl_setlk
3.65 ? 36% perf-profile.calltrace.cycles-pp.page_counter_uncharge.obj_cgroup_uncharge_pages.kmem_cache_free.posix_lock_inode.do_lock_file_wait
3.72 ? 36% perf-profile.calltrace.cycles-pp.obj_cgroup_uncharge_pages.kmem_cache_free.posix_lock_inode.do_lock_file_wait.fcntl_setlk
4.29 ? 27% perf-profile.calltrace.cycles-pp.obj_cgroup_charge.kmem_cache_alloc.locks_alloc_lock.fcntl_setlk.do_fcntl
9.47 ? 12% perf-profile.calltrace.cycles-pp.kmem_cache_free.posix_lock_inode.do_lock_file_wait.fcntl_setlk.do_fcntl
14.88 ? 6% perf-profile.calltrace.cycles-pp.locks_alloc_lock.fcntl_setlk.do_fcntl.__x64_sys_fcntl.do_syscall_64
13.74 ? 7% perf-profile.calltrace.cycles-pp.kmem_cache_alloc.locks_alloc_lock.fcntl_setlk.do_fcntl.__x64_sys_fcntl
90.74 perf-profile.calltrace.cycles-pp.entry_SYSCALL_64_after_hwframe.__libc_fcntl64
89.82 perf-profile.calltrace.cycles-pp.do_syscall_64.entry_SYSCALL_64_after_hwframe.__libc_fcntl64
7.75 ? 29% perf-profile.calltrace.cycles-pp.obj_cgroup_charge_pages.obj_cgroup_charge.kmem_cache_alloc.locks_alloc_lock.posix_lock_inode
54.10 ? 3% perf-profile.calltrace.cycles-pp.do_lock_file_wait.fcntl_setlk.do_fcntl.__x64_sys_fcntl.do_syscall_64
87.56 perf-profile.calltrace.cycles-pp.__x64_sys_fcntl.do_syscall_64.entry_SYSCALL_64_after_hwframe.__libc_fcntl64
9.15 ? 23% perf-profile.calltrace.cycles-pp.obj_cgroup_charge.kmem_cache_alloc.locks_alloc_lock.posix_lock_inode.do_lock_file_wait
84.60 perf-profile.calltrace.cycles-pp.do_fcntl.__x64_sys_fcntl.do_syscall_64.entry_SYSCALL_64_after_hwframe.__libc_fcntl64
11.21 ? 25% perf-profile.calltrace.cycles-pp.page_counter_try_charge.obj_cgroup_charge_pages.obj_cgroup_charge.kmem_cache_alloc.locks_alloc_lock
49.70 ? 4% perf-profile.calltrace.cycles-pp.posix_lock_inode.do_lock_file_wait.fcntl_setlk.do_fcntl.__x64_sys_fcntl
28.04 ? 5% perf-profile.calltrace.cycles-pp.locks_alloc_lock.posix_lock_inode.do_lock_file_wait.fcntl_setlk.do_fcntl
25.71 ? 6% perf-profile.calltrace.cycles-pp.kmem_cache_alloc.locks_alloc_lock.posix_lock_inode.do_lock_file_wait.fcntl_setlk
78.96 perf-profile.calltrace.cycles-pp.fcntl_setlk.do_fcntl.__x64_sys_fcntl.do_syscall_64.entry_SYSCALL_64_after_hwframe
4.33 ? 6% perf-profile.children.cycles-pp.memset_erms
4.62 ? 7% perf-profile.children.cycles-pp._copy_from_user
2.89 ? 18% perf-profile.children.cycles-pp.__entry_text_start
3.47 ? 6% perf-profile.children.cycles-pp._raw_spin_lock
2.66 ? 6% perf-profile.children.cycles-pp.security_file_lock
2.75 ? 8% perf-profile.children.cycles-pp.syscall_return_via_sysret
2.01 ? 7% perf-profile.children.cycles-pp.common_file_perm
2.28 ? 6% perf-profile.children.cycles-pp.copy_user_generic_unrolled
2.08 ? 5% perf-profile.children.cycles-pp.___might_sleep
2.40 ? 6% perf-profile.children.cycles-pp.__might_sleep
1.41 ? 6% perf-profile.children.cycles-pp.locks_delete_lock_ctx
1.29 ? 5% perf-profile.children.cycles-pp.locks_release_private
1.81 ? 6% perf-profile.children.cycles-pp.__fget_light
1.14 ? 6% perf-profile.children.cycles-pp.locks_unlink_lock_ctx
1.15 ? 7% perf-profile.children.cycles-pp.locks_insert_lock_ctx
1.17 ? 7% perf-profile.children.cycles-pp.copy_user_enhanced_fast_string
1.34 ? 5% perf-profile.children.cycles-pp.__cond_resched
0.86 ? 7% perf-profile.children.cycles-pp.aa_file_perm
0.87 ? 7% perf-profile.children.cycles-pp.__might_fault
1.07 ? 9% perf-profile.children.cycles-pp.syscall_exit_to_user_mode
0.76 ? 6% perf-profile.children.cycles-pp.syscall_enter_from_user_mode
0.67 ? 6% perf-profile.children.cycles-pp.flock64_to_posix_lock
0.65 ? 5% perf-profile.children.cycles-pp.rcu_all_qs
0.45 ? 7% perf-profile.children.cycles-pp.entry_SYSCALL_64_safe_stack
0.60 ? 6% perf-profile.children.cycles-pp.__init_waitqueue_head
0.41 ? 4% perf-profile.children.cycles-pp.locks_delete_block
0.42 ? 5% perf-profile.children.cycles-pp.locks_get_lock_context
0.39 ? 7% perf-profile.children.cycles-pp.locks_copy_lock
0.23 ? 5% perf-profile.children.cycles-pp.__list_del_entry_valid
0.49 ? 9% perf-profile.children.cycles-pp.exit_to_user_mode_prepare
0.35 ? 8% perf-profile.children.cycles-pp.security_file_fcntl
0.35 ? 11% perf-profile.children.cycles-pp.memset
0.29 ? 6% perf-profile.children.cycles-pp.apparmor_file_lock
0.24 ? 7% perf-profile.children.cycles-pp.__list_add_valid
0.25 ? 7% perf-profile.children.cycles-pp.vfs_lock_file
0.33 ? 8% perf-profile.children.cycles-pp.testcase
0.21 ? 6% perf-profile.children.cycles-pp.locks_copy_conflock
0.18 ? 7% perf-profile.children.cycles-pp.__fdget_raw
0.24 ? 7% perf-profile.children.cycles-pp.syscall_exit_to_user_mode_prepare
0.12 ? 6% perf-profile.children.cycles-pp.locks_move_blocks
0.32 ? 6% perf-profile.children.cycles-pp.should_failslab
0.13 ? 10% perf-profile.children.cycles-pp.fcntl@plt
0.08 ? 6% perf-profile.children.cycles-pp.get_mem_cgroup_from_objcg
0.08 ? 11% perf-profile.children.cycles-pp.refill_stock
98.23 perf-profile.children.cycles-pp.__libc_fcntl64
0.30 ? 6% perf-profile.children.cycles-pp.obj_cgroup_uncharge
0.53 ? 7% perf-profile.children.cycles-pp.mem_cgroup_from_task
3.69 ? 14% perf-profile.children.cycles-pp.locks_dispose_list
1.99 ? 5% perf-profile.children.cycles-pp.refill_obj_stock
2.40 ? 6% perf-profile.children.cycles-pp.rcu_read_unlock_strict
3.39 ? 21% perf-profile.children.cycles-pp.propagate_protected_usage
4.90 ? 8% perf-profile.children.cycles-pp.get_obj_cgroup_from_current
5.60 ? 6% perf-profile.children.cycles-pp.mod_objcg_state
5.94 ? 26% perf-profile.children.cycles-pp.page_counter_cancel
90.98 perf-profile.children.cycles-pp.entry_SYSCALL_64_after_hwframe
7.12 ? 24% perf-profile.children.cycles-pp.page_counter_uncharge
90.30 perf-profile.children.cycles-pp.do_syscall_64
7.27 ? 24% perf-profile.children.cycles-pp.obj_cgroup_uncharge_pages
54.45 ? 3% perf-profile.children.cycles-pp.do_lock_file_wait
87.90 perf-profile.children.cycles-pp.__x64_sys_fcntl
84.84 perf-profile.children.cycles-pp.do_fcntl
50.57 ? 4% perf-profile.children.cycles-pp.posix_lock_inode
11.23 ? 25% perf-profile.children.cycles-pp.page_counter_try_charge
11.37 ? 24% perf-profile.children.cycles-pp.obj_cgroup_charge_pages
19.54 ? 6% perf-profile.children.cycles-pp.kmem_cache_free
13.63 ? 19% perf-profile.children.cycles-pp.obj_cgroup_charge
79.39 perf-profile.children.cycles-pp.fcntl_setlk
43.45 ? 2% perf-profile.children.cycles-pp.locks_alloc_lock
41.55 ? 3% perf-profile.children.cycles-pp.kmem_cache_alloc
4.14 ? 6% perf-profile.self.cycles-pp.memset_erms
3.35 ? 8% perf-profile.self.cycles-pp.__libc_fcntl64
3.30 ? 6% perf-profile.self.cycles-pp._raw_spin_lock
2.76 ? 5% perf-profile.self.cycles-pp.posix_lock_inode
2.75 ? 8% perf-profile.self.cycles-pp.syscall_return_via_sysret
1.77 ? 4% perf-profile.self.cycles-pp.fcntl_setlk
2.09 ? 6% perf-profile.self.cycles-pp.copy_user_generic_unrolled
1.77 ? 5% perf-profile.self.cycles-pp.___might_sleep
1.74 ? 6% perf-profile.self.cycles-pp.__fget_light
1.14 ? 7% perf-profile.self.cycles-pp.common_file_perm
1.84 ? 6% perf-profile.self.cycles-pp.__might_sleep
1.14 ? 5% perf-profile.self.cycles-pp.locks_release_private
1.47 ? 8% perf-profile.self.cycles-pp.locks_alloc_lock
1.04 ? 7% perf-profile.self.cycles-pp.copy_user_enhanced_fast_string
0.71 ? 21% perf-profile.self.cycles-pp.__entry_text_start
0.94 ? 7% perf-profile.self.cycles-pp.__x64_sys_fcntl
0.95 ? 8% perf-profile.self.cycles-pp.do_fcntl
0.79 ? 6% perf-profile.self.cycles-pp.do_lock_file_wait
0.81 ? 7% perf-profile.self.cycles-pp.entry_SYSCALL_64_after_hwframe
0.73 ? 7% perf-profile.self.cycles-pp.aa_file_perm
0.67 ? 6% perf-profile.self.cycles-pp.syscall_enter_from_user_mode
0.60 ? 6% perf-profile.self.cycles-pp.flock64_to_posix_lock
0.56 ? 6% perf-profile.self.cycles-pp._copy_from_user
0.61 ? 8% perf-profile.self.cycles-pp.do_syscall_64
0.65 ? 4% perf-profile.self.cycles-pp.__cond_resched
0.45 ? 7% perf-profile.self.cycles-pp.entry_SYSCALL_64_safe_stack
0.42 ? 5% perf-profile.self.cycles-pp.security_file_lock
0.38 ? 8% perf-profile.self.cycles-pp.locks_insert_lock_ctx
0.32 ? 6% perf-profile.self.cycles-pp.locks_unlink_lock_ctx
0.44 ? 6% perf-profile.self.cycles-pp.rcu_all_qs
0.42 ? 10% perf-profile.self.cycles-pp.exit_to_user_mode_prepare
0.32 ? 5% perf-profile.self.cycles-pp.locks_dispose_list
0.18 ? 7% perf-profile.self.cycles-pp.__list_del_entry_valid
0.29 ? 5% perf-profile.self.cycles-pp.locks_delete_block
0.30 ? 5% perf-profile.self.cycles-pp.locks_get_lock_context
0.42 ? 5% perf-profile.self.cycles-pp.__init_waitqueue_head
0.23 ? 9% perf-profile.self.cycles-pp.locks_delete_lock_ctx
0.36 ? 11% perf-profile.self.cycles-pp.syscall_exit_to_user_mode
0.25 ? 9% perf-profile.self.cycles-pp.security_file_fcntl
0.17 ? 7% perf-profile.self.cycles-pp.__list_add_valid
0.18 ? 9% perf-profile.self.cycles-pp.locks_copy_lock
0.18 ? 7% perf-profile.self.cycles-pp.__might_fault
0.27 ? 9% perf-profile.self.cycles-pp.testcase
0.18 ? 5% perf-profile.self.cycles-pp.vfs_lock_file
0.18 ? 7% perf-profile.self.cycles-pp.apparmor_file_lock
0.18 ? 6% perf-profile.self.cycles-pp.locks_copy_conflock
0.18 ? 10% perf-profile.self.cycles-pp.memset
0.18 ? 7% perf-profile.self.cycles-pp.syscall_exit_to_user_mode_prepare
0.09 ? 7% perf-profile.self.cycles-pp.locks_move_blocks
0.06 ? 6% perf-profile.self.cycles-pp.__fdget_raw
0.17 ? 6% perf-profile.self.cycles-pp.should_failslab
0.07 ? 14% perf-profile.self.cycles-pp.fcntl@plt
0.06 ? 6% perf-profile.self.cycles-pp.get_mem_cgroup_from_objcg
0.08 ? 11% perf-profile.self.cycles-pp.refill_stock
0.14 ? 10% perf-profile.self.cycles-pp.obj_cgroup_uncharge
0.36 ? 7% perf-profile.self.cycles-pp.mem_cgroup_from_task
0.92 ? 5% perf-profile.self.cycles-pp.rcu_read_unlock_strict
1.85 ? 5% perf-profile.self.cycles-pp.refill_obj_stock
2.05 ? 7% perf-profile.self.cycles-pp.obj_cgroup_charge
12.32 ? 4% perf-profile.self.cycles-pp.kmem_cache_alloc
3.36 ? 21% perf-profile.self.cycles-pp.propagate_protected_usage
4.01 ? 8% perf-profile.self.cycles-pp.get_obj_cgroup_from_current
5.23 ? 7% perf-profile.self.cycles-pp.mod_objcg_state
5.90 ? 26% perf-profile.self.cycles-pp.page_counter_cancel
8.97 ? 25% perf-profile.self.cycles-pp.page_counter_try_charge
estimated based on internal Intel analysis and are provided
purposes only. Any difference in system hardware or software
may affect actual performance.
Open Source Technology Center
org/hyperkitty/list/lkp@lists.01.org">https://lists.01.org/hyperkitty/list/[email protected] Intel Corporation

Attachments:

(No filename) (78.96 kB)
config-5.14.0-00066-g0f12156dff28 (178.19 kB)
job-script (7.66 kB)
job.yaml (5.22 kB)
reproduce (347.00 B)
Download all attachments

2021-09-07 15:48:22

by Jens Axboe

[permalink] [raw]

Subject: Re: [memcg] 0f12156dff: will-it-scale.per_process_ops -33.6% regression

On 9/7/21 9:07 AM, kernel test robot wrote:
>
>
> Greeting,
>
> FYI, we noticed a -33.6% regression of will-it-scale.per_process_ops due to commit:
>
>
> commit: 0f12156dff2862ac54235fc72703f18770769042 ("memcg: enable accounting for file lock caches")
> https://git.kernel.org/cgit/linux/kernel/git/torvalds/linux.git master

Are we at all worried about these? There's been a number of them
reported, basically for all the accounting enablements that have been
done in this merge window.

When io_uring was switched to use accounted memory, we did a bunch of
work to ameliorate the inevitable slowdowns that happen if you do
repeated allocs and/or frees and have memcg accounting enabled.

--
Jens Axboe

2021-09-07 16:16:38

by Shakeel Butt

[permalink] [raw]

Subject: Re: [memcg] 0f12156dff: will-it-scale.per_process_ops -33.6% regression

On Tue, Sep 7, 2021 at 8:46 AM Jens Axboe <[email protected]> wrote:
>
> On 9/7/21 9:07 AM, kernel test robot wrote:
> >
> >
> > Greeting,
> >
> > FYI, we noticed a -33.6% regression of will-it-scale.per_process_ops due to commit:
> >
> >
> > commit: 0f12156dff2862ac54235fc72703f18770769042 ("memcg: enable accounting for file lock caches")
> > https://git.kernel.org/cgit/linux/kernel/git/torvalds/linux.git master
>
> Are we at all worried about these? There's been a number of them
> reported, basically for all the accounting enablements that have been
> done in this merge window.
>
> When io_uring was switched to use accounted memory, we did a bunch of
> work to ameliorate the inevitable slowdowns that happen if you do
> repeated allocs and/or frees and have memcg accounting enabled.
>

I think these are important and we should aim to continuously improve
performance with memcg accounting. I would like to know more about the
io_uring work done to improve memcg accounting. Maybe we can
generalize it to others as well.

2021-09-07 16:18:21

by Jens Axboe

[permalink] [raw]

Subject: Re: [memcg] 0f12156dff: will-it-scale.per_process_ops -33.6% regression

On 9/7/21 9:57 AM, Shakeel Butt wrote:
> On Tue, Sep 7, 2021 at 8:46 AM Jens Axboe <[email protected]> wrote:
>>
>> On 9/7/21 9:07 AM, kernel test robot wrote:
>>>
>>>
>>> Greeting,
>>>
>>> FYI, we noticed a -33.6% regression of will-it-scale.per_process_ops due to commit:
>>>
>>>
>>> commit: 0f12156dff2862ac54235fc72703f18770769042 ("memcg: enable accounting for file lock caches")
>>> https://git.kernel.org/cgit/linux/kernel/git/torvalds/linux.git master
>>
>> Are we at all worried about these? There's been a number of them
>> reported, basically for all the accounting enablements that have been
>> done in this merge window.
>>
>> When io_uring was switched to use accounted memory, we did a bunch of
>> work to ameliorate the inevitable slowdowns that happen if you do
>> repeated allocs and/or frees and have memcg accounting enabled.
>>
>
> I think these are important and we should aim to continuously improve
> performance with memcg accounting. I would like to know more about the
> io_uring work done to improve memcg accounting. Maybe we can
> generalize it to others as well.

It's pretty basic and may not be applicable to all cases, we simply hang
on to our allocations for longer periods and reuse them. Hence instead
of always going through alloc+free to each "unit", they are recycled and
reused until no longer needed.

Now this is more efficient in general for us, as we can have a very high
rate of requests (and hence allocs+frees). I suspect most use cases
would benefit from simply having a cache in front of memcg slabs, but
that seems like solving the issue at the wrong layer. IMHO it'd be
better to have the memcg accounting be done in batches, eg have some
notion of deferred frees. If someone allocates before the deferred frees
are accounted, then that would have saved two pieces of accounting.

It is of course possible that a lot of these regressions are simply
accounting the alloc, in which case it seems like accounting in batches
might help there. All depends on the slack that is acceptable for memcg.

--
Jens Axboe

2021-09-07 17:37:24

by Linus Torvalds

[permalink] [raw]

Subject: Re: [memcg] 0f12156dff: will-it-scale.per_process_ops -33.6% regression

On Tue, Sep 7, 2021 at 8:46 AM Jens Axboe <[email protected]> wrote:
>
> Are we at all worried about these? There's been a number of them
> reported, basically for all the accounting enablements that have been
> done in this merge window.

We are worried about them. I'm considering reverting several of them
because I think the problems are

(a) big

(b) nontrivial

and the patches clearly weren't ready and people weren't aware of this issue.

Linus

2021-09-07 18:47:32

by Shakeel Butt

[permalink] [raw]

Subject: Re: [memcg] 0f12156dff: will-it-scale.per_process_ops -33.6% regression

On Tue, Sep 7, 2021 at 9:40 AM Linus Torvalds
<[email protected]> wrote:
>
> On Tue, Sep 7, 2021 at 8:46 AM Jens Axboe <[email protected]> wrote:
> >
> > Are we at all worried about these? There's been a number of them
> > reported, basically for all the accounting enablements that have been
> > done in this merge window.
>
> We are worried about them. I'm considering reverting several of them
> because I think the problems are
>
> (a) big
>
> (b) nontrivial
>
> and the patches clearly weren't ready and people weren't aware of this issue.
>

Sounds good to me. Please let me know which patches you are planning
to revert. I will work on the followup to make those acceptable.

thanks,
Shakeel

2021-09-07 18:52:44

by Linus Torvalds

[permalink] [raw]

Subject: Re: [memcg] 0f12156dff: will-it-scale.per_process_ops -33.6% regression

On Tue, Sep 7, 2021 at 9:49 AM Shakeel Butt <[email protected]> wrote:
>
> On Tue, Sep 7, 2021 at 9:40 AM Linus Torvalds
> <[email protected]> wrote:
> >
> > We are worried about them. I'm considering reverting several of them
> > because I think the problems are
> >
> > (a) big
> >
> > (b) nontrivial
> >
> > and the patches clearly weren't ready and people weren't aware of this issue.
>
> Sounds good to me. Please let me know which patches you are planning
> to revert. I will work on the followup to make those acceptable.

The one that looks clear-cut is the one in this thread:

0f12156dff28 memcg: enable accounting for file lock caches

which seems to result in regressions on multiple machines and just be
very bad for anything that uses file locking. I'm not entirely sure
how much that would show up in real life, but I can most definitely
imagine it being a problem on a real load.

There's a few other regression reports I've seen, like

5387c90490f7 mm/memcg: improve refill_obj_stock() performance

but that one had mixed reports (it improved another benchmark), and it
looks like Minchan has a fix for the regression already.

And

aa48e47e3906 memcg: infrastructure to flush memcg stats
b65584344415 memcg: enable accounting for pollfd and select bits arrays

were reported as a regression in -mm, but not in mainline yet.

I assume (but didn't check) that aa48e47e3906 is a bigger deal to revert.

So _right_now_ my plan is to revert the two obvious cases:

0f12156dff28 memcg: enable accounting for file lock caches
b65584344415 memcg: enable accounting for pollfd and select bits arrays

on the assumption that the memcg accounting code needs some work to
make it less of a performance hog.

Does anybody have other commits they want to highlight (or other
comments) about this?

Linus

2021-09-07 19:24:42

by Shakeel Butt

[permalink] [raw]

Subject: Re: [memcg] 0f12156dff: will-it-scale.per_process_ops -33.6% regression

On Tue, Sep 7, 2021 at 10:54 AM Linus Torvalds
<[email protected]> wrote:
>
> On Tue, Sep 7, 2021 at 9:49 AM Shakeel Butt <[email protected]> wrote:
> >
> > On Tue, Sep 7, 2021 at 9:40 AM Linus Torvalds
> > <[email protected]> wrote:
> > >
> > > We are worried about them. I'm considering reverting several of them
> > > because I think the problems are
> > >
> > > (a) big
> > >
> > > (b) nontrivial
> > >
> > > and the patches clearly weren't ready and people weren't aware of this issue.
> >
> > Sounds good to me. Please let me know which patches you are planning
> > to revert. I will work on the followup to make those acceptable.
>
> The one that looks clear-cut is the one in this thread:
>
> 0f12156dff28 memcg: enable accounting for file lock caches
>
> which seems to result in regressions on multiple machines and just be
> very bad for anything that uses file locking. I'm not entirely sure
> how much that would show up in real life, but I can most definitely
> imagine it being a problem on a real load.
>
> There's a few other regression reports I've seen, like
>
> 5387c90490f7 mm/memcg: improve refill_obj_stock() performance
>
> but that one had mixed reports (it improved another benchmark), and it
> looks like Minchan has a fix for the regression already.
>
> And
>
> aa48e47e3906 memcg: infrastructure to flush memcg stats
> b65584344415 memcg: enable accounting for pollfd and select bits arrays
>
> were reported as a regression in -mm, but not in mainline yet.
>
> I assume (but didn't check) that aa48e47e3906 is a bigger deal to revert.
>

I am suspecting aa48e47e3906 might be similar to [1]. I am testing that theory.

[1] https://lore.kernel.org/all/20210811031734.GA5193@xsang-OptiPlex-9020/T/#u