Hi Trond, I attached perf-profile part big changes, hope it is useful for analyzing the issue. In testcase: fsmark on test machine: 40 threads Intel(R) Xeon(R) CPU E5-2690 v2 @ 3.00GHz with 384G memory with following parameters: iterations: 20x nr_threads: 64t disk: 1BRD_48G fs: xfs fs2: nfsv4 filesize: 4M test_size: 80G sync_method: fsyncBeforeClose cpufreq_governor: performance test-description: The fsmark is a file system benchmark to test synchronous write workloads, for example, mail servers workload. test-url: https://sourceforge.net/projects/fsmark/ commit: e791f8e938 ("SUNRPC: Convert xs_send_kvec() to use iov_iter_kvec()") 0472e47660 ("SUNRPC: Convert socket page send code to use iov_iter()") e791f8e9380d945e 0472e476604998c127f3c80d291 ---------------- --------------------------- %stddev %change %stddev \ | \ 527.29 -22.6% 407.96 fsmark.files_per_sec 1.97 ± 11% +0.9 2.88 ± 4% perf-profile.calltrace.cycles-pp.smp_apic_timer_interrupt.apic_timer_interrupt.cpuidle_enter_state.do_idle.cpu_startup_entry 0.00 +0.9 0.93 ± 4% perf-profile.calltrace.cycles-pp.tcp_write_xmit.tcp_sendmsg_locked.tcp_sendmsg.sock_sendmsg.xs_sendpages 2.11 ± 10% +0.9 3.05 ± 4% perf-profile.calltrace.cycles-pp.apic_timer_interrupt.cpuidle_enter_state.do_idle.cpu_startup_entry.start_secondary 5.29 ± 2% +1.2 6.46 ± 7% perf-profile.calltrace.cycles-pp.svc_recv.nfsd.kthread.ret_from_fork 9.61 ± 5% +3.1 12.70 ± 2% perf-profile.calltrace.cycles-pp.worker_thread.kthread.ret_from_fork 9.27 ± 5% +3.1 12.40 ± 2% perf-profile.calltrace.cycles-pp.process_one_work.worker_thread.kthread.ret_from_fork 34.52 ± 4% +3.3 37.78 ± 2% perf-profile.calltrace.cycles-pp.ret_from_fork 34.52 ± 4% +3.3 37.78 ± 2% perf-profile.calltrace.cycles-pp.kthread.ret_from_fork 0.00 +3.4 3.41 ± 4% perf-profile.calltrace.cycles-pp.memcpy_erms.memcpy_from_page._copy_from_iter_full.tcp_sendmsg_locked.tcp_sendmsg 0.00 +3.4 3.44 ± 4% perf-profile.calltrace.cycles-pp.memcpy_from_page._copy_from_iter_full.tcp_sendmsg_locked.tcp_sendmsg.sock_sendmsg 0.00 +3.5 3.54 ± 4% perf-profile.calltrace.cycles-pp._copy_from_iter_full.tcp_sendmsg_locked.tcp_sendmsg.sock_sendmsg.xs_sendpages 2.30 ± 5% +3.7 6.02 ± 3% perf-profile.calltrace.cycles-pp.__rpc_execute.rpc_async_schedule.process_one_work.worker_thread.kthread 2.30 ± 5% +3.7 6.02 ± 3% perf-profile.calltrace.cycles-pp.rpc_async_schedule.process_one_work.worker_thread.kthread.ret_from_fork 1.81 ± 4% +3.8 5.59 ± 4% perf-profile.calltrace.cycles-pp.call_transmit.__rpc_execute.rpc_async_schedule.process_one_work.worker_thread 1.80 ± 3% +3.8 5.59 ± 3% perf-profile.calltrace.cycles-pp.xprt_transmit.call_transmit.__rpc_execute.rpc_async_schedule.process_one_work 1.73 ± 4% +3.8 5.54 ± 4% perf-profile.calltrace.cycles-pp.xs_tcp_send_request.xprt_transmit.call_transmit.__rpc_execute.rpc_async_schedule 1.72 ± 4% +3.8 5.54 ± 4% perf-profile.calltrace.cycles-pp.xs_sendpages.xs_tcp_send_request.xprt_transmit.call_transmit.__rpc_execute 0.00 +5.4 5.42 ± 4% perf-profile.calltrace.cycles-pp.tcp_sendmsg_locked.tcp_sendmsg.sock_sendmsg.xs_sendpages.xs_tcp_send_request 0.00 +5.5 5.52 ± 4% perf-profile.calltrace.cycles-pp.tcp_sendmsg.sock_sendmsg.xs_sendpages.xs_tcp_send_request.xprt_transmit 0.00 +5.5 5.53 ± 4% perf-profile.calltrace.cycles-pp.sock_sendmsg.xs_sendpages.xs_tcp_send_request.xprt_transmit.call_transmit 9.61 ± 5% +3.1 12.70 ± 2% perf-profile.children.cycles-pp.worker_thread 9.27 ± 5% +3.1 12.40 ± 2% perf-profile.children.cycles-pp.process_one_work 6.19 +3.2 9.40 ± 4% perf-profile.children.cycles-pp.memcpy_erms 34.53 ± 4% +3.3 37.78 ± 2% perf-profile.children.cycles-pp.ret_from_fork 34.52 ± 4% +3.3 37.78 ± 2% perf-profile.children.cycles-pp.kthread 0.00 +3.5 3.46 ± 4% perf-profile.children.cycles-pp.memcpy_from_page 0.00 +3.6 3.56 ± 4% perf-profile.children.cycles-pp._copy_from_iter_full 2.47 ± 4% +3.7 6.18 ± 3% perf-profile.children.cycles-pp.__rpc_execute 2.30 ± 5% +3.7 6.02 ± 3% perf-profile.children.cycles-pp.rpc_async_schedule 1.90 ± 4% +3.8 5.67 ± 3% perf-profile.children.cycles-pp.call_transmit 1.89 ± 3% +3.8 5.66 ± 3% perf-profile.children.cycles-pp.xprt_transmit 1.82 ± 4% +3.8 5.62 ± 3% perf-profile.children.cycles-pp.xs_tcp_send_request 1.81 ± 4% +3.8 5.62 ± 3% perf-profile.children.cycles-pp.xs_sendpages 0.21 ± 17% +5.3 5.48 ± 4% perf-profile.children.cycles-pp.tcp_sendmsg_locked 0.25 ± 18% +5.3 5.59 ± 3% perf-profile.children.cycles-pp.tcp_sendmsg 0.26 ± 16% +5.3 5.60 ± 3% perf-profile.children.cycles-pp.sock_sendmsg 1.19 ± 5% +0.5 1.68 ± 3% perf-profile.self.cycles-pp.get_page_from_freelist 6.10 +3.2 9.27 ± 4% perf-profile.self.cycles-pp.memcpy_erms On 7/9/2019 10:39 AM, Xing Zhengjun wrote: > Hi Trond, > > On 7/8/2019 7:44 PM, Trond Myklebust wrote: >> I've asked several times now about how to interpret your results. As >> far as I can tell from your numbers, the overhead appears to be >> entirely contained in the NUMA section of your results. >> IOW: it would appear to be a scheduling overhead due to NUMA. I've >> been asking whether or not that is a correct interpretation of the >> numbers you published. > Thanks for your feedback. I used the same hardware and the same test > parameters to test the two commits: >    e791f8e938 ("SUNRPC: Convert xs_send_kvec() to use iov_iter_kvec()") >    0472e47660 ("SUNRPC: Convert socket page send code to use iov_iter()") > > If it is caused by NUMA, why only commit 0472e47660 throughput is > decreased? The filesystem we test is NFS, commit 0472e47660 is related > with the network, could you help to check if have any other clues for > the regression. Thanks. > -- Zhengjun Xing