Greeting, FYI, we noticed a 20.1% regression of fio.latency_10us% due to commit: commit: 51f2c7c0900521da299f5b28f642582ea97dc47a ("mm, vmscan: consider eligible zones in get_scan_count") https://git.kernel.org/cgit/linux/kernel/git/mhocko/mm.git to_test/linus-tree/oom_hickups in testcase: fio-basic on test machine: 88 threads Intel(R) Xeon(R) CPU E5-2699 v4 @ 2.20GHz with 128G memory with following parameters: disk: 1SSD fs: ext4 runtime: 300s nr_task: 8 rw: write bs: 4k ioengine: sync test_size: 512g cpufreq_governor: performance test-description: Fio is a tool that will spawn a number of threads or processes doing a particular type of I/O action as specified by the user. test-url: https://github.com/axboe/fio Details are as below: --------------------------------------------------------------------------------------------------> To reproduce: git clone git://git.kernel.org/pub/scm/linux/kernel/git/wfg/lkp-tests.git cd lkp-tests bin/lkp install job.yaml # job file is attached in this email bin/lkp run job.yaml testcase/path_params/tbox_group/run: fio-basic/1SSD-ext4-300s-8-write-4k-sync-512g-performance/lkp-bdw-ep2 c675d2353ec11e10 51f2c7c0900521da299f5b28f6 ---------------- -------------------------- 14.19 20% 17.04 ± 5% fio.latency_10us% 0.50 ± 3% 12% 0.56 ± 3% fio.latency_50us% 636 628 fio.write_clat_stddev 83.22 81.09 fio.latency_4us% 1.41 ± 10% -55% 0.63 fio.latency_2us% 386 5% 406 fio.time.system_time 152 5% 159 fio.time.percent_of_cpu_this_job_got 95250 4% 99321 fio.time.voluntary_context_switches 4.315e+11 7% 4.6e+11 ± 3% perf-stat.branch-instructions 2.178e+12 6% 2.314e+12 perf-stat.instructions 8887 6% 9424 ± 3% perf-stat.instructions-per-iTLB-miss 6.109e+11 6% 6.459e+11 perf-stat.dTLB-loads 2.404e+09 4% 2.502e+09 perf-stat.branch-misses 3.751e+11 4% 3.894e+11 perf-stat.dTLB-stores 7.819e+08 4% 8.111e+08 ± 3% perf-stat.dTLB-load-misses 1.663e+08 1.638e+08 perf-stat.dTLB-store-misses 0.04 -5% 0.04 perf-stat.dTLB-store-miss-rate% 0 1e+04 14354 ±100% latency_stats.avg.balance_dirty_pages.balance_dirty_pages_ratelimited.generic_perform_write.__generic_file_write_iter.ext4_file_write_iter.__vfs_write.vfs_write.SyS_write.do_syscall_64.return_from_SYSCALL_64 0 8e+04 82365 ± 5% latency_stats.max.call_rwsem_down_read_failed.ext4_da_get_block_prep.__block_write_begin_int.__block_write_begin.ext4_da_write_begin.generic_perform_write.__generic_file_write_iter.ext4_file_write_iter.__vfs_write.vfs_write.SyS_write.entry_SYSCALL_64_fastpath 0 1e+04 14354 ±100% latency_stats.max.balance_dirty_pages.balance_dirty_pages_ratelimited.generic_perform_write.__generic_file_write_iter.ext4_file_write_iter.__vfs_write.vfs_write.SyS_write.do_syscall_64.return_from_SYSCALL_64 0 1e+06 1353230 ± 11% latency_stats.sum.call_rwsem_down_read_failed.ext4_da_get_block_prep.__block_write_begin_int.__block_write_begin.ext4_da_write_begin.generic_perform_write.__generic_file_write_iter.ext4_file_write_iter.__vfs_write.vfs_write.SyS_write.entry_SYSCALL_64_fastpath 0 1e+04 14354 ±100% latency_stats.sum.balance_dirty_pages.balance_dirty_pages_ratelimited.generic_perform_write.__generic_file_write_iter.ext4_file_write_iter.__vfs_write.vfs_write.SyS_write.do_syscall_64.return_from_SYSCALL_64 perf-stat.dTLB-stores 3.94e+11 ++---------------------------------------------------------------+ 3.92e+11 ++O O O O O O O O O | O O O O O O O O O O O | 3.9e+11 ++ O | 3.88e+11 ++ O O O O | 3.86e+11 ++ | 3.84e+11 ++ | | | 3.82e+11 ++ | 3.8e+11 ++ | 3.78e+11 ++ * *. * * | 3.76e+11 ++ + : *.* + *.*.* + : + : *.* | | .* : .. + .* + .* : .* : + +| 3.74e+11 *+* *.* * *.*. *.*.*.* *.*..*.*.* * 3.72e+11 ++---------------------------------------------------------------+ fio.time.voluntary_context_switches 101000 ++-----------------------------------------------------------------+ | O O | 100000 ++ O O O O O | | O O O O O O O O O O | 99000 O+O O O O O O O | | | 98000 ++ | | | 97000 ++ | | | 96000 ++ *.* * * | | .*.*.. : + .*.*. .*.* : + .*.* : + .*.. .*.* 95000 *+* : *. .*..* *.*.*. + : *.*. + : * *.* | | * * * * | 94000 ++-----------------------------------------------------------------+ [*] bisect-good sample [O] bisect-bad sample Disclaimer: Results have been estimated based on internal Intel analysis and are provided for informational purposes only. Any difference in system hardware or software design or configuration may affect actual performance. Thanks, Xiaolong