FYI, we noticed the below changes on https://git.kernel.org/pub/scm/linux/kernel/git/next/linux-next.git master commit 46cfb125d673f2f30acf06b3c02485bf693d54e9 ("mm: increase SWAP_CLUSTER_MAX to batch TLB flushes") ========================================================================================= tbox_group/testcase/rootfs/kconfig/compiler/cpufreq_governor/iterations/nr_threads/disk/fs/filesize/test_size/sync_method/nr_directories/nr_files_per_directory: nhm4/fsmark/debian-x86_64-2015-02-07.cgz/x86_64-rhel/gcc-4.9/performance/1x/32t/1HDD/btrfs/16MB/60G/fsyncBeforeClose/16d/256fpd commit: 284f74b8c06045dcc1b5caaf5f412219f18cfc60 46cfb125d673f2f30acf06b3c02485bf693d54e9 284f74b8c06045dc 46cfb125d673f2f30acf06b3c0 ---------------- -------------------------- %stddev %change %stddev \ | \ 14126 ± 3% +11.3% 15720 ± 4% fsmark.time.minor_page_faults 17.00 ± 0% +5.9% 18.00 ± 0% fsmark.time.percent_of_cpu_this_job_got 88.97 ± 0% +8.8% 96.80 ± 0% fsmark.time.system_time 52682 ± 0% -18.1% 43155 ± 0% softirqs.BLOCK 14126 ± 3% +11.3% 15720 ± 4% time.minor_page_faults 25179737 ± 8% -18.5% 20528816 ± 5% cpuidle.C1E-NHM.time 133847 ± 3% -14.1% 114908 ± 4% cpuidle.C3-NHM.usage 92678 ± 2% -20.8% 73391 ± 1% meminfo.Dirty 124930 ± 1% +21.4% 151629 ± 0% meminfo.Writeback 3.45 ± 1% +3.4% 3.56 ± 0% turbostat.%Busy 111.50 ± 1% +3.8% 115.75 ± 0% turbostat.Avg_MHz 101775 ±157% +329.9% 437551 ±169% latency_stats.max.btrfs_tree_lock.[btrfs].btrfs_lock_root_node.[btrfs].btrfs_search_slot.[btrfs].btrfs_insert_empty_items.[btrfs].btrfs_new_inode.[btrfs].btrfs_create.[btrfs].vfs_create.path_openat.do_filp_open.do_sys_open.SyS_open.entry_SYSCALL_64_fastpath 8005 ± 19% +5364.1% 437413 ±169% latency_stats.max.btrfs_tree_lock.[btrfs].btrfs_lock_root_node.[btrfs].btrfs_search_slot.[btrfs].btrfs_insert_empty_items.[btrfs].insert_with_overflow.[btrfs].btrfs_insert_dir_item.[btrfs].btrfs_add_link.[btrfs].btrfs_create.[btrfs].vfs_create.path_openat.do_filp_open.do_sys_open 469382 ± 66% -98.7% 6266 ± 18% latency_stats.max.btrfs_tree_lock.[btrfs].btrfs_search_slot.[btrfs].btrfs_lookup_inode.[btrfs].__btrfs_update_delayed_inode.[btrfs].btrfs_commit_inode_delayed_items.[btrfs].btrfs_log_inode.[btrfs].btrfs_log_inode_parent.[btrfs].btrfs_log_dentry_safe.[btrfs].btrfs_sync_file.[btrfs].vfs_fsync_range.do_fsync.SyS_fsync 363661 ±103% -100.0% 84.50 ± 19% latency_stats.max.btrfs_tree_read_lock.[btrfs].btrfs_read_lock_root_node.[btrfs].btrfs_search_slot.[btrfs].btrfs_insert_empty_items.[btrfs].copy_items.[btrfs].btrfs_log_inode.[btrfs].btrfs_log_inode_parent.[btrfs].btrfs_log_dentry_safe.[btrfs].btrfs_sync_file.[btrfs].vfs_fsync_range.do_fsync.SyS_fsync 404767 ±170% -98.6% 5611 ±114% latency_stats.max.btrfs_tree_read_lock.[btrfs].btrfs_read_lock_root_node.[btrfs].btrfs_search_slot.[btrfs].btrfs_lookup_dir_item.[btrfs].btrfs_lookup_dentry.[btrfs].btrfs_lookup.[btrfs].lookup_real.path_openat.do_filp_open.do_sys_open.SyS_open.entry_SYSCALL_64_fastpath 35516 ±145% +1126.2% 435514 ±170% latency_stats.max.btrfs_tree_read_lock.[btrfs].btrfs_read_lock_root_node.[btrfs].btrfs_search_slot.[btrfs].btrfs_lookup_xattr.[btrfs].__btrfs_getxattr.[btrfs].btrfs_getxattr.[btrfs].cap_inode_need_killpriv.security_inode_need_killpriv.dentry_needs_remove_privs.file_remove_privs.btrfs_file_write_iter.[btrfs].__vfs_write 1107650 ±130% -72.4% 305514 ± 96% latency_stats.max.wait_on_page_bit.__filemap_fdatawait_range.filemap_fdatawait_range.btrfs_wait_marked_extents.[btrfs].btrfs_write_and_wait_transaction.[btrfs].btrfs_commit_transaction.[btrfs].btrfs_sync_file.[btrfs].vfs_fsync_range.do_fsync.SyS_fsync.entry_SYSCALL_64_fastpath 199353 ± 84% +377.4% 951697 ±154% latency_stats.sum.btrfs_tree_lock.[btrfs].btrfs_lock_root_node.[btrfs].btrfs_search_slot.[btrfs].btrfs_insert_empty_items.[btrfs].btrfs_new_inode.[btrfs].btrfs_create.[btrfs].vfs_create.path_openat.do_filp_open.do_sys_open.SyS_open.entry_SYSCALL_64_fastpath 62503 ± 33% +657.0% 473167 ±155% latency_stats.sum.btrfs_tree_lock.[btrfs].btrfs_lock_root_node.[btrfs].btrfs_search_slot.[btrfs].btrfs_insert_empty_items.[btrfs].insert_with_overflow.[btrfs].btrfs_insert_dir_item.[btrfs].btrfs_add_link.[btrfs].btrfs_create.[btrfs].vfs_create.path_openat.do_filp_open.do_sys_open 5568453 ± 73% -75.9% 1342137 ± 11% latency_stats.sum.btrfs_tree_lock.[btrfs].btrfs_search_slot.[btrfs].btrfs_lookup_inode.[btrfs].__btrfs_update_delayed_inode.[btrfs].btrfs_commit_inode_delayed_items.[btrfs].btrfs_log_inode.[btrfs].btrfs_log_inode_parent.[btrfs].btrfs_log_dentry_safe.[btrfs].btrfs_sync_file.[btrfs].vfs_fsync_range.do_fsync.SyS_fsync 1647770 ±118% -99.5% 7733 ± 11% latency_stats.sum.btrfs_tree_read_lock.[btrfs].btrfs_read_lock_root_node.[btrfs].btrfs_search_slot.[btrfs].btrfs_insert_empty_items.[btrfs].copy_items.[btrfs].btrfs_log_inode.[btrfs].btrfs_log_inode_parent.[btrfs].btrfs_log_dentry_safe.[btrfs].btrfs_sync_file.[btrfs].vfs_fsync_range.do_fsync.SyS_fsync 4780008 ±170% -99.2% 36367 ± 61% latency_stats.sum.btrfs_tree_read_lock.[btrfs].btrfs_read_lock_root_node.[btrfs].btrfs_search_slot.[btrfs].btrfs_lookup_dir_item.[btrfs].btrfs_lookup_dentry.[btrfs].btrfs_lookup.[btrfs].lookup_real.path_openat.do_filp_open.do_sys_open.SyS_open.entry_SYSCALL_64_fastpath 88081 ±118% +919.9% 898339 ±164% latency_stats.sum.btrfs_tree_read_lock.[btrfs].btrfs_read_lock_root_node.[btrfs].btrfs_search_slot.[btrfs].btrfs_lookup_xattr.[btrfs].__btrfs_getxattr.[btrfs].btrfs_getxattr.[btrfs].cap_inode_need_killpriv.security_inode_need_killpriv.dentry_needs_remove_privs.file_remove_privs.btrfs_file_write_iter.[btrfs].__vfs_write 20472 ± 0% -24.3% 15507 ± 0% proc-vmstat.allocstall 1021 ± 34% -73.5% 271.00 ± 26% proc-vmstat.kswapd_high_wmark_hit_quickly 23142 ± 2% -20.9% 18301 ± 1% proc-vmstat.nr_dirty 2441 ± 28% -82.8% 419.50 ± 29% proc-vmstat.nr_vmscan_immediate_reclaim 31236 ± 1% +21.5% 37946 ± 1% proc-vmstat.nr_writeback 8479 ± 5% +18.8% 10075 ± 7% proc-vmstat.numa_hint_faults 8479 ± 5% +18.8% 10075 ± 7% proc-vmstat.numa_hint_faults_local 9293 ± 4% +18.1% 10977 ± 6% proc-vmstat.numa_pte_updates 683.50 ± 95% -86.0% 96.00 ± 67% proc-vmstat.pgdeactivate 1298 ± 11% -39.7% 783.50 ± 45% proc-vmstat.pgmigrate_success 796.75 ± 81% -71.9% 224.25 ± 30% proc-vmstat.pgrefill_dma 2969680 ± 0% +110.2% 6241700 ± 0% proc-vmstat.pgscan_direct_dma32 9735 ± 15% -86.3% 1332 ± 20% proc-vmstat.pgscan_kswapd_dma 12295232 ± 0% -26.7% 9017741 ± 0% proc-vmstat.pgscan_kswapd_dma32 2893713 ± 1% +113.0% 6162694 ± 0% proc-vmstat.pgsteal_direct_dma32 12278505 ± 0% -26.7% 9003970 ± 0% proc-vmstat.pgsteal_kswapd_dma32 7.00 ±141% +225.0% 22.75 ± 30% sched_debug.cfs_rq[0]:/.load 88.75 ± 47% +166.2% 236.25 ± 19% sched_debug.cfs_rq[0]:/.tg_load_avg 0.50 ±100% +3150.0% 16.25 ± 47% sched_debug.cfs_rq[1]:/.runnable_load_avg 86.25 ± 42% +175.1% 237.25 ± 19% sched_debug.cfs_rq[1]:/.tg_load_avg 87.25 ± 41% +175.4% 240.25 ± 20% sched_debug.cfs_rq[2]:/.tg_load_avg 22.75 ± 35% +217.6% 72.25 ± 44% sched_debug.cfs_rq[2]:/.util_avg 7.00 ± 82% +128.6% 16.00 ± 32% sched_debug.cfs_rq[3]:/.load_avg -533786 ± -7% -15.0% -453453 ± -6% sched_debug.cfs_rq[3]:/.spread0 88.75 ± 40% +169.6% 239.25 ± 19% sched_debug.cfs_rq[3]:/.tg_load_avg 7.00 ± 82% +135.7% 16.50 ± 35% sched_debug.cfs_rq[3]:/.tg_load_avg_contrib 23.50 ± 58% +217.0% 74.50 ± 44% sched_debug.cfs_rq[3]:/.util_avg 89.00 ± 40% +168.8% 239.25 ± 19% sched_debug.cfs_rq[4]:/.tg_load_avg 89.50 ± 39% +169.6% 241.25 ± 18% sched_debug.cfs_rq[5]:/.tg_load_avg 28.75 ± 90% +173.0% 78.50 ± 61% sched_debug.cfs_rq[5]:/.util_avg 5799 ± 5% +14.2% 6619 ± 2% sched_debug.cfs_rq[6]:/.exec_clock 90.50 ± 39% +167.7% 242.25 ± 19% sched_debug.cfs_rq[6]:/.tg_load_avg 13.00 ±105% +142.3% 31.50 ± 60% sched_debug.cfs_rq[7]:/.load_avg 93.00 ± 40% +162.4% 244.00 ± 19% sched_debug.cfs_rq[7]:/.tg_load_avg 13.00 ±105% +165.4% 34.50 ± 64% sched_debug.cfs_rq[7]:/.tg_load_avg_contrib 0.75 ±110% +733.3% 6.25 ± 60% sched_debug.cpu#0.cpu_load[3] 0.75 ±110% +1033.3% 8.50 ± 45% sched_debug.cpu#0.cpu_load[4] 34.33 ±141% +759.2% 295.00 ± 42% sched_debug.cpu#0.curr->pid 5.75 ±153% +791.3% 51.25 ± 83% sched_debug.cpu#0.load -821.75 ± -3% -13.2% -713.25 ± -2% sched_debug.cpu#0.nr_uninterruptible 1.00 ±100% +750.0% 8.50 ± 48% sched_debug.cpu#1.cpu_load[1] 1.75 ± 84% +400.0% 8.75 ± 48% sched_debug.cpu#1.cpu_load[2] 30329 ± 51% -46.2% 16309 ± 5% sched_debug.cpu#2.nr_load_updates 4.00 ± 98% +918.8% 40.75 ± 69% sched_debug.cpu#6.cpu_load[0] 2.25 ± 85% +1400.0% 33.75 ± 85% sched_debug.cpu#6.cpu_load[1] 1.75 ± 84% +1557.1% 29.00 ± 90% sched_debug.cpu#6.cpu_load[2] 1.25 ± 87% +1960.0% 25.75 ± 93% sched_debug.cpu#6.cpu_load[3] 149.50 ± 20% -29.3% 105.75 ± 15% sched_debug.cpu#6.nr_uninterruptible 36817 ± 8% +4975.8% 1868779 ±169% sched_debug.cpu#7.nr_switches 37382 ± 9% +4900.3% 1869233 ±169% sched_debug.cpu#7.sched_count 16197 ± 9% +5605.7% 924151 ±169% sched_debug.cpu#7.sched_goidle 0.49 ± 2% +1020.2% 5.53 ±103% sched_debug.rt_rq[0]:/.rt_time nhm4: Nehalem Memory: 4G fsmark.time.system_time 100 ++--------------------------------------------------------------------+ | O | 98 O+ O O O O O O | | O O O O O O O | 96 ++ O O O O | | | 94 ++ | | | 92 ++ | | .*. .*.*.. .*.. | 90 ++ .*.*. *. *.*..* * | *..*.*..*. .*. .*. + .*.*..*.*..*.*..*.* 88 ++ *. *..* *. | | | 86 ++--------------------------------------------------------------------+ [*] bisect-good sample [O] bisect-bad sample 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 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, Ying Huang