linux-dev/tools/perf/bench, branch master

perf bench: Update use of pthread mutex/cond

2022-10-04T11:55:19Z

Switch to the use of mutex wrappers that provide better error checking. Signed-off-by: Ian Rogers Reviewed-by: Adrian Hunter Acked-by: Namhyung Kim Cc: Alexander Shishkin Cc: Alexandre Truong Cc: Alexey Bayduraev Cc: Andi Kleen Cc: Andres Freund Cc: Andrii Nakryiko Cc: André Almeida Cc: Athira Jajeev Cc: Christophe JAILLET Cc: Colin Ian King Cc: Dario Petrillo Cc: Darren Hart Cc: Dave Marchevsky Cc: Davidlohr Bueso Cc: Fangrui Song Cc: Hewenliang Cc: Ingo Molnar Cc: James Clark Cc: Jason Wang Cc: Jiri Olsa Cc: Kajol Jain Cc: Kim Phillips Cc: Leo Yan Cc: Mark Rutland Cc: Martin Liška Cc: Masami Hiramatsu Cc: Nathan Chancellor Cc: Nick Desaulniers Cc: Pavithra Gurushankar Cc: Peter Zijlstra Cc: Quentin Monnet Cc: Ravi Bangoria Cc: Remi Bernon Cc: Riccardo Mancini Cc: Song Liu Cc: Stephane Eranian Cc: Thomas Gleixner Cc: Thomas Richter Cc: Tom Rix Cc: Weiguo Li Cc: Wenyu Liu Cc: William Cohen Cc: Zechuan Chen Cc: bpf@vger.kernel.org Cc: llvm@lists.linux.dev Cc: yaowenbin Link: https://lore.kernel.org/r/20220826164242.43412-3-irogers@google.com Signed-off-by: Arnaldo Carvalho de Melo

Merge remote-tracking branch 'torvalds/master' into perf/core

2022-05-23T12:32:49Z

To get the rest of 5.18. Signed-off-by: Arnaldo Carvalho de Melo

perf bench numa: Address compiler error on s390

2022-05-21T17:45:19Z

The compilation on s390 results in this error: # make DEBUG=y bench/numa.o ... bench/numa.c: In function ‘__bench_numa’: bench/numa.c:1749:81: error: ‘%d’ directive output may be truncated writing between 1 and 11 bytes into a region of size between 10 and 20 [-Werror=format-truncation=] 1749 | snprintf(tname, sizeof(tname), "process%d:thread%d", p, t); ^~ ... bench/numa.c:1749:64: note: directive argument in the range [-2147483647, 2147483646] ... # The maximum length of the %d replacement is 11 characters because of the negative sign. Therefore extend the array by two more characters. Output after: # make DEBUG=y bench/numa.o > /dev/null 2>&1; ll bench/numa.o -rw-r--r-- 1 root root 418320 May 19 09:11 bench/numa.o # Fixes: 3aff8ba0a4c9c919 ("perf bench numa: Avoid possible truncation when using snprintf()") Suggested-by: Namhyung Kim Signed-off-by: Thomas Richter Cc: Heiko Carstens Cc: Sumanth Korikkar Cc: Sven Schnelle Cc: Vasily Gorbik Link: https://lore.kernel.org/r/20220520081158.2990006-1-tmricht@linux.ibm.com Signed-off-by: Arnaldo Carvalho de Melo

perf bench breakpoint: Fix build on 32-bit arches

2022-05-17T00:23:09Z

Cast pointers to unsigned long instead of to uint64_t to avoid this problem on 32-bit arches: 31 6.89 debian:experimental-x-mips : FAIL gcc version 11.2.0 (Debian 11.2.0-18) bench/breakpoint.c: In function 'breakpoint_setup': bench/breakpoint.c:56:24: error: cast from pointer to integer of different size [-Werror=pointer-to-int-cast] 56 | attr.bp_addr = (uint64_t)addr; | ^ cc1: all warnings being treated as errors make[3]: *** [/git/perf-5.18.0-rc7/tools/build/Makefile.build:139: bench] Error 2 Fixes: 68a6772f11dbb1ed ("perf bench: Add breakpoint benchmarks") Cc: Adrian Hunter Cc: Alexander Shishkin Cc: Dmitriy Vyukov Cc: Ian Rogers Cc: Ingo Molnar Cc: Jiri Olsa Cc: Marco Elver Cc: Mark Rutland Cc: Namhyung Kim Cc: Peter Zijlstra Link: http://lore.kernel.org/lkml/YoLq1nHx1doi+VWl@kernel.org Signed-off-by: Arnaldo Carvalho de Melo

Merge remote-tracking branch 'torvalds/master' into perf/core

2022-05-16T13:14:59Z

To pick up fixes from perf/urgent. Signed-off-by: Arnaldo Carvalho de Melo

perf bench: Add breakpoint benchmarks

2022-05-13T14:00:38Z

Add 2 benchmarks: 1. Performance of thread creation/exiting in presence of breakpoints. 2. Performance of breakpoint modification in presence of threads. The benchmarks capture use cases that we are interested in: using inheritable breakpoints in large highly-threaded applications. The benchmarks show significant slowdown imposed by breakpoints (even when they don't fire). Testing on Intel 8173M with 112 HW threads show: perf bench --repeat=56 breakpoint thread --breakpoints=0 --parallelism=56 --threads=20 78.675000 usecs/op perf bench --repeat=56 breakpoint thread --breakpoints=4 --parallelism=56 --threads=20 12967.135714 usecs/op That's 165x slowdown due to presence of the breakpoints. perf bench --repeat=20000 breakpoint enable --passive=0 --active=0 1.433250 usecs/op perf bench --repeat=20000 breakpoint enable --passive=224 --active=0 585.318400 usecs/op perf bench --repeat=20000 breakpoint enable --passive=0 --active=111 635.953000 usecs/op That's 408x and 444x slowdown due to presence of threads. Profiles show some overhead in toggle_bp_slot, but also very high contention: 90.83% breakpoint-thre [kernel.kallsyms] [k] osq_lock 4.69% breakpoint-thre [kernel.kallsyms] [k] mutex_spin_on_owner 2.06% breakpoint-thre [kernel.kallsyms] [k] __reserve_bp_slot 2.04% breakpoint-thre [kernel.kallsyms] [k] toggle_bp_slot 79.01% breakpoint-enab [kernel.kallsyms] [k] smp_call_function_single 9.94% breakpoint-enab [kernel.kallsyms] [k] llist_add_batch 5.70% breakpoint-enab [kernel.kallsyms] [k] _raw_spin_lock_irq 1.84% breakpoint-enab [kernel.kallsyms] [k] event_function_call 1.12% breakpoint-enab [kernel.kallsyms] [k] send_call_function_single_ipi 0.37% breakpoint-enab [kernel.kallsyms] [k] generic_exec_single 0.24% breakpoint-enab [kernel.kallsyms] [k] __perf_event_disable 0.20% breakpoint-enab [kernel.kallsyms] [k] _perf_event_enable 0.18% breakpoint-enab [kernel.kallsyms] [k] toggle_bp_slot Committer notes: Fixup struct init for older compilers: 3 32.90 alpine:3.5 : FAIL clang version 3.8.1 (tags/RELEASE_381/final) bench/breakpoint.c:49:34: error: missing field 'size' initializer [-Werror,-Wmissing-field-initializers] struct perf_event_attr attr = {0}; ^ 1 error generated. 7 37.31 alpine:3.9 : FAIL gcc version 8.3.0 (Alpine 8.3.0) bench/breakpoint.c:49:34: error: missing field 'size' initializer [-Werror,-Wmissing-field-initializers] struct perf_event_attr attr = {0}; ^ 1 error generated. Signed-off-by: Dmitriy Vyukov Tested-by: Arnaldo Carvalho de Melo Acked-by: Ian Rogers Cc: Alexander Shishkin Cc: Ingo Molnar Cc: Jiri Olsa Cc: Marco Elver Cc: Mark Rutland Cc: Namhyung Kim Cc: Peter Zijlstra Link: https://lore.kernel.org/r/20220505155745.1690906-1-dvyukov@google.com Signed-off-by: Arnaldo Carvalho de Melo

perf bench: Fix two numa NDEBUG warnings

2022-05-09T00:48:49Z

BUG_ON is a no-op if NDEBUG is defined, otherwise it is an assert. Compiling with NDEBUG yields: bench/numa.c: In function ‘bind_to_cpu’: bench/numa.c:314:1: error: control reaches end of non-void function [-Werror=return-type] 314 | } | ^ bench/numa.c: In function ‘bind_to_node’: bench/numa.c:367:1: error: control reaches end of non-void function [-Werror=return-type] 367 | } | ^ Add return statements to cover this case. Reviewed-by: Athira Jajeev Signed-off-by: Ian Rogers Cc: Alexander Shishkin Cc: Andi Kleen Cc: Ingo Molnar Cc: Jin Yao Cc: Jiri Olsa Cc: Mark Rutland Cc: Namhyung Kim Cc: Peter Zijlstra Cc: Xing Zhengjun Link: https://lore.kernel.org/r/20220428202912.1056444-1-irogers@google.com Signed-off-by: Arnaldo Carvalho de Melo

perf bench: Fix numa bench to fix usage of affinity for machines with #CPUs > 1K

2022-04-14T12:15:10Z

The 'perf bench numa' testcase fails on systems with more than 1K CPUs. Testcase: perf bench numa mem -p 1 -t 3 -P 512 -s 100 -zZ0qcm --thp 1 Snippet of code: <<>> perf: bench/numa.c:302: bind_to_node: Assertion `!(ret)' failed. Aborted (core dumped) <<>> bind_to_node() uses "sched_getaffinity" to save the original cpumask and this call is returning EINVAL ((invalid argument). This happens because the default mask size in glibc is 1024. To overcome this 1024 CPUs mask size limitation of cpu_set_t, change the mask size using the CPU_*_S macros ie, use CPU_ALLOC to allocate cpumask, CPU_ALLOC_SIZE for size. Apart from fixing this for "orig_mask", apply same logic to "mask" as well which is used to setaffinity so that mask size is large enough to represent number of possible CPU's in the system. sched_getaffinity is used in one more place in perf numa bench. It is in "bind_to_cpu" function. Apply the same logic there also. Though currently no failure is reported from there, it is ideal to change getaffinity to work with such system configurations having CPU's more than default mask size supported by glibc. Also fix "sched_setaffinity" to use mask size which is large enough to represent number of possible CPU's in the system. Fixed all places where "bind_cpumask" which is part of "struct thread_data" is used such that bind_cpumask works in all configuration. Reported-by: Disha Goel Signed-off-by: Athira Jajeev Cc: Ian Rogers Cc: Jiri Olsa Cc: Kajol Jain Cc: Madhavan Srinivasan Cc: Michael Ellerman Cc: Nageswara R Sastry Cc: Srikar Dronamraju Cc: linuxppc-dev@lists.ozlabs.org Link: https://lore.kernel.org/r/20220412164059.42654-3-atrajeev@linux.vnet.ibm.com Signed-off-by: Arnaldo Carvalho de Melo

perf bench: Fix numa testcase to check if CPU used to bind task is online

2022-04-14T12:13:41Z

Perf numa bench test fails with error: Testcase: ./perf bench numa mem -p 2 -t 1 -P 1024 -C 0,8 -M 1,0 -s 20 -zZq --thp 1 --no-data_rand_walk Failure snippet: <<>> Running 'numa/mem' benchmark: # Running main, "perf bench numa numa-mem -p 2 -t 1 -P 1024 -C 0,8 -M 1,0 -s 20 -zZq --thp 1 --no-data_rand_walk" perf: bench/numa.c:333: bind_to_cpumask: Assertion `!(ret)' failed. <<>> The Testcases uses CPU's 0 and 8. In function "parse_setup_cpu_list", There is check to see if cpu number is greater than max cpu's possible in the system ie via "if (bind_cpu_0 >= g->p.nr_cpus || bind_cpu_1 >= g->p.nr_cpus) {". But it could happen that system has say 48 CPU's, but only number of online CPU's is 0-7. Other CPU's are offlined. Since "g->p.nr_cpus" is 48, so function will go ahead and set bit for CPU 8 also in cpumask ( td->bind_cpumask). bind_to_cpumask function is called to set affinity using sched_setaffinity and the cpumask. Since the CPU8 is not present, set affinity will fail here with EINVAL. Fix this issue by adding a check to make sure that, CPU's provided in the input argument values are online before proceeding further and skip the test. For this, include new helper function "is_cpu_online" in "tools/perf/util/header.c". Since "BIT(x)" definition will get included from header.h, remove that from bench/numa.c Reported-by: Disha Goel Signed-off-by: Athira Jajeev Tested-by: Disha Goel Cc: Ian Rogers Cc: Jiri Olsa Cc: Kajol Jain Cc: Madhavan Srinivasan Cc: Michael Ellerman Cc: Nageswara R Sastry Cc: Srikar Dronamraju Cc: linuxppc-dev@lists.ozlabs.org Link: https://lore.kernel.org/r/20220412164059.42654-2-atrajeev@linux.vnet.ibm.com Signed-off-by: Arnaldo Carvalho de Melo

perf bench: Fix epoll bench to correct usage of affinity for machines with #CPUs > 1K

2022-04-09T15:34:29Z

The 'perf bench epoll' testcase fails on systems with more than 1K CPUs. Testcase: perf bench epoll all Result snippet: <<>> Run summary [PID 106497]: 1399 threads monitoring on 64 file-descriptors for 8 secs. perf: pthread_create: No such file or directory <<>> In epoll benchmarks (ctl, wait) pthread_create is invoked in do_threads from respective bench_epoll_* function. Though the logs shows direct failure from pthread_create, the actual failure is from "sched_setaffinity" returning EINVAL (invalid argument). This happens because the default mask size in glibc is 1024. To overcome this 1024 CPUs mask size limitation of cpu_set_t, change the mask size using the CPU_*_S macros. Patch addresses this by fixing all the epoll benchmarks to use CPU_ALLOC to allocate cpumask, CPU_ALLOC_SIZE for size, and CPU_SET_S to set the mask. Reported-by: Disha Goel Signed-off-by: Athira Jajeev Tested-by: Disha Goel Acked-by: Ian Rogers Cc: Jiri Olsa Cc: Kajol Jain Cc: Madhavan Srinivasan Cc: Michael Ellerman Cc: Nageswara R Sastry Cc: Srikar Dronamraju Cc: linuxppc-dev@lists.ozlabs.org Link: https://lore.kernel.org/r/20220406175113.87881-3-atrajeev@linux.vnet.ibm.com Signed-off-by: Arnaldo Carvalho de Melo