Hi all, This patch series are some optimization and extension for PSI. patch 1/9 fix periodic aggregation shut off problem introduced by earlier commit 4117cebf1a9f ("psi: Optimize task switch inside shared cgroups"). patch 2/9 optimize task switch inside shared cgroups when in_memstall status of prev task and next task are different. patch 3-4 optimize and simplify PSI status tracking by don't change task psi_flags when migrate CPU/cgroup. patch 7-8 introduce new kernel cmdline parameter "psi_inner_cgroup=" to configure whether or not to track PSI stall information for inner cgroups. patch 9/9 introduce new PSI resource PSI_IRQ to track IRQ/SOFTIRQ pressure stall information when CONFIG_IRQ_TIME_ACCOUNTING. Performance test on Intel Xeon Platinum with 3 levels of cgroup, in which run mmtests config-scheduler-perfpipe: tip tip tip patched patched patched patched default cgroup_disable=pressure IRQ_TIME_ACCOUNTING default psi_inner_cgroup=off PSI_IRQ PSI_IRQ + psi_inner_cgroup=off Min Time 9.89 ( 0.00%) 8.99 ( 9.12%) 10.04 ( -1.53%) 9.63 ( 2.58%) 9.27 ( 6.22%) 10.09 ( -2.04%) 9.45 ( 4.41%) 1st-qrtle Time 10.01 ( 0.00%) 9.15 ( 8.66%) 10.16 ( -1.45%) 9.72 ( 2.89%) 9.35 ( 6.61%) 10.20 ( -1.81%) 9.54 ( 4.77%) 2nd-qrtle Time 10.07 ( 0.00%) 9.25 ( 8.12%) 10.19 ( -1.21%) 9.79 ( 2.73%) 9.38 ( 6.78%) 10.24 ( -1.75%) 9.59 ( 4.68%) 3rd-qrtle Time 10.14 ( 0.00%) 9.30 ( 8.32%) 10.23 ( -0.88%) 9.84 ( 3.00%) 9.44 ( 6.92%) 10.27 ( -1.21%) 9.62 ( 5.18%) Max-1 Time 9.89 ( 0.00%) 8.99 ( 9.12%) 10.04 ( -1.53%) 9.63 ( 2.58%) 9.27 ( 6.22%) 10.09 ( -2.04%) 9.45 ( 4.41%) Max-5 Time 9.89 ( 0.00%) 8.99 ( 9.12%) 10.04 ( -1.53%) 9.63 ( 2.58%) 9.27 ( 6.22%) 10.09 ( -2.04%) 9.45 ( 4.41%) Max-10 Time 9.92 ( 0.00%) 9.09 ( 8.33%) 10.11 ( -1.97%) 9.67 ( 2.51%) 9.29 ( 6.29%) 10.15 ( -2.30%) 9.48 ( 4.46%) Max-90 Time 10.20 ( 0.00%) 9.33 ( 8.53%) 10.33 ( -1.24%) 9.87 ( 3.29%) 9.49 ( 6.99%) 10.29 ( -0.85%) 9.66 ( 5.32%) Max-95 Time 10.23 ( 0.00%) 9.34 ( 8.70%) 10.37 ( -1.39%) 9.94 ( 2.83%) 9.53 ( 6.88%) 10.30 ( -0.65%) 9.67 ( 5.51%) Max-99 Time 10.23 ( 0.00%) 9.37 ( 8.43%) 10.40 ( -1.63%) 9.99 ( 2.41%) 9.76 ( 4.57%) 10.31 ( -0.74%) 9.69 ( 5.25%) Max Time 10.34 ( 0.00%) 9.46 ( 8.50%) 10.43 ( -0.83%) 17.04 ( -64.80%) 9.79 ( 5.36%) 10.32 ( 0.20%) 9.71 ( 6.07%) Amean Time 10.08 ( 0.00%) 9.23 * 8.39%* 10.21 * -1.33%* 10.03 ( 0.47%) 9.41 * 6.59%* 10.23 * -1.53%* 9.59 * 4.87%* Thanks! Chengming Zhou (9): sched/psi: fix periodic aggregation shut off sched/psi: optimize task switch inside shared cgroups again sched/psi: move private helpers to sched/stats.h sched/psi: don't change task psi_flags when migrate CPU/group sched/psi: don't create cgroup PSI files when psi_disabled sched/psi: save percpu memory when !psi_cgroups_enabled sched/psi: cache parent psi_group to speed up groups iterate sched/psi: add kernel cmdline parameter psi_inner_cgroup sched/psi: add PSI_IRQ to track IRQ/SOFTIRQ pressure .../admin-guide/kernel-parameters.txt | 11 + include/linux/psi.h | 5 +- include/linux/psi_types.h | 9 +- include/linux/sched.h | 3 - kernel/cgroup/cgroup.c | 30 +++ kernel/sched/core.c | 2 + kernel/sched/psi.c | 194 +++++++++++++----- kernel/sched/stats.h | 71 ++++--- 8 files changed, 232 insertions(+), 93 deletions(-) -- 2.36.1