From: Ingo Molnar Date: Mon, 23 Mar 2009 20:49:25 +0000 (+0100) Subject: perf_counter tools: tidy up in-kernel dependencies X-Git-Tag: firefly_0821_release~13973^2~443 X-Git-Url: http://demsky.eecs.uci.edu/git/?a=commitdiff_plain;h=383c5f8cd7d253be0017d8d5a39cbb78aaf70206;p=firefly-linux-kernel-4.4.55.git perf_counter tools: tidy up in-kernel dependencies Remove now unified perfstat.c and perf_counter.h, and link to the in-kernel perf_counter.h. Cc: Wu Fengguang Cc: Paul Mackerras Acked-by: Peter Zijlstra Orig-LKML-Reference: <20090323172417.677932499@chello.nl> Signed-off-by: Ingo Molnar --- diff --git a/Documentation/perf_counter/Makefile b/Documentation/perf_counter/Makefile index b45749753fcb..666da95a7877 100644 --- a/Documentation/perf_counter/Makefile +++ b/Documentation/perf_counter/Makefile @@ -2,7 +2,7 @@ BINS = kerneltop perfstat all: $(BINS) -kerneltop: kerneltop.c perfcounters.h +kerneltop: kerneltop.c ../../include/linux/perf_counter.h cc -O6 -Wall -lrt `pkg-config --cflags --libs glib-2.0` -o $@ $< perfstat: kerneltop diff --git a/Documentation/perf_counter/kerneltop.c b/Documentation/perf_counter/kerneltop.c index 80b790553ec8..25e80bc44556 100644 --- a/Documentation/perf_counter/kerneltop.c +++ b/Documentation/perf_counter/kerneltop.c @@ -88,7 +88,7 @@ #include -#include "include/linux/perf_counter.h" +#include "../../include/linux/perf_counter.h" /* diff --git a/Documentation/perf_counter/perfcounters.h b/Documentation/perf_counter/perfcounters.h deleted file mode 100644 index 32e24b9154ab..000000000000 --- a/Documentation/perf_counter/perfcounters.h +++ /dev/null @@ -1,142 +0,0 @@ -/* - * Ioctls that can be done on a perf counter fd: - */ -#define PERF_COUNTER_IOC_ENABLE _IO('$', 0) -#define PERF_COUNTER_IOC_DISABLE _IO('$', 1) - -/* - * prctl(PR_TASK_PERF_COUNTERS_DISABLE) will (cheaply) disable all - * counters in the current task. - */ -#define PR_TASK_PERF_COUNTERS_DISABLE 31 -#define PR_TASK_PERF_COUNTERS_ENABLE 32 - -#define ARRAY_SIZE(x) (sizeof(x) / sizeof((x)[0])) - -#define rdclock() \ -({ \ - struct timespec ts; \ - \ - clock_gettime(CLOCK_MONOTONIC, &ts); \ - ts.tv_sec * 1000000000ULL + ts.tv_nsec; \ -}) - -/* - * Pick up some kernel type conventions: - */ -#define __user -#define asmlinkage - -typedef unsigned int __u32; -typedef unsigned long long __u64; -typedef long long __s64; - -/* - * User-space ABI bits: - */ - -/* - * Generalized performance counter event types, used by the hw_event.type - * parameter of the sys_perf_counter_open() syscall: - */ -enum hw_event_types { - /* - * Common hardware events, generalized by the kernel: - */ - PERF_COUNT_CPU_CYCLES = 0, - PERF_COUNT_INSTRUCTIONS = 1, - PERF_COUNT_CACHE_REFERENCES = 2, - PERF_COUNT_CACHE_MISSES = 3, - PERF_COUNT_BRANCH_INSTRUCTIONS = 4, - PERF_COUNT_BRANCH_MISSES = 5, - PERF_COUNT_BUS_CYCLES = 6, - - PERF_HW_EVENTS_MAX = 7, - - /* - * Special "software" counters provided by the kernel, even if - * the hardware does not support performance counters. These - * counters measure various physical and sw events of the - * kernel (and allow the profiling of them as well): - */ - PERF_COUNT_CPU_CLOCK = -1, - PERF_COUNT_TASK_CLOCK = -2, - PERF_COUNT_PAGE_FAULTS = -3, - PERF_COUNT_CONTEXT_SWITCHES = -4, - PERF_COUNT_CPU_MIGRATIONS = -5, - - PERF_SW_EVENTS_MIN = -6, -}; - -/* - * IRQ-notification data record type: - */ -enum perf_counter_record_type { - PERF_RECORD_SIMPLE = 0, - PERF_RECORD_IRQ = 1, - PERF_RECORD_GROUP = 2, -}; - -/* - * Hardware event to monitor via a performance monitoring counter: - */ -struct perf_counter_hw_event { - __s64 type; - - __u64 irq_period; - __u64 record_type; - __u64 read_format; - - __u64 disabled : 1, /* off by default */ - nmi : 1, /* NMI sampling */ - raw : 1, /* raw event type */ - inherit : 1, /* children inherit it */ - pinned : 1, /* must always be on PMU */ - exclusive : 1, /* only group on PMU */ - exclude_user : 1, /* don't count user */ - exclude_kernel : 1, /* ditto kernel */ - exclude_hv : 1, /* ditto hypervisor */ - exclude_idle : 1, /* don't count when idle */ - - __reserved_1 : 54; - - __u32 extra_config_len; - __u32 __reserved_4; - - __u64 __reserved_2; - __u64 __reserved_3; -}; - - -#ifdef __x86_64__ -# define __NR_perf_counter_open 295 -#endif - -#ifdef __i386__ -# define __NR_perf_counter_open 333 -#endif - -#ifdef __powerpc__ -#define __NR_perf_counter_open 319 -#endif - -asmlinkage int sys_perf_counter_open( - - struct perf_counter_hw_event *hw_event_uptr __user, - pid_t pid, - int cpu, - int group_fd, - unsigned long flags) -{ - int ret; - - ret = syscall( - __NR_perf_counter_open, hw_event_uptr, pid, cpu, group_fd, flags); -#if defined(__x86_64__) || defined(__i386__) - if (ret < 0 && ret > -4096) { - errno = -ret; - ret = -1; - } -#endif - return ret; -} diff --git a/Documentation/perf_counter/perfstat.c b/Documentation/perf_counter/perfstat.c deleted file mode 100644 index fd594468e655..000000000000 --- a/Documentation/perf_counter/perfstat.c +++ /dev/null @@ -1,251 +0,0 @@ -/* - * perfstat: /usr/bin/time -alike performance counter statistics utility - * - * It summarizes the counter events of all tasks (and child tasks), - * covering all CPUs that the command (or workload) executes on. - * It only counts the per-task events of the workload started, - * independent of how many other tasks run on those CPUs. - * - * Build with: cc -O2 -g -lrt -Wall -W -o perfstat perfstat.c - * - * Sample output: - * - - $ ./perfstat -e 1 -e 3 -e 5 ls -lR /usr/include/ >/dev/null - - Performance counter stats for 'ls': - - 163516953 instructions - 2295 cache-misses - 2855182 branch-misses - - * - * Copyright (C) 2008, Red Hat Inc, Ingo Molnar - * - * Released under the GPLv2 (not later). - * - * Percpu counter support by: Yanmin Zhang - * Symbolic event options by: Wu Fengguang - */ -#define _GNU_SOURCE - -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include - -#include -#include -#include -#include -#include -#include -#include -#include - -#include - -#include "perfcounters.h" - -static int nr_cpus = 0; - -static int system_wide = 0; - -static void display_help(void) -{ - unsigned int i; - int e; - - printf( - "Usage: perfstat [] \n\n" - "PerfStat Options (up to %d event types can be specified):\n\n", - MAX_COUNTERS); - printf( - " -e EVENT --event=EVENT # symbolic-name abbreviations"); - - for (i = 0, e = PERF_HW_EVENTS_MAX; i < ARRAY_SIZE(event_symbols); i++) { - if (e != event_symbols[i].event) { - e = event_symbols[i].event; - printf( - "\n %2d: %-20s", e, event_symbols[i].symbol); - } else - printf(" %s", event_symbols[i].symbol); - } - - printf("\n" - " rNNN: raw event type\n\n" - " -s # system-wide collection\n\n" - " -c --command= # command+arguments to be timed.\n" - "\n"); - exit(0); -} - -static void process_options(int argc, char *argv[]) -{ - for (;;) { - int option_index = 0; - /** Options for getopt */ - static struct option long_options[] = { - {"event", required_argument, NULL, 'e'}, - {"help", no_argument, NULL, 'h'}, - {"command", no_argument, NULL, 'c'}, - {NULL, 0, NULL, 0 } - }; - int c = getopt_long(argc, argv, "+:e:c:s", - long_options, &option_index); - if (c == -1) - break; - - switch (c) { - case 'c': - break; - case 's': - system_wide = 1; - break; - case 'e': - parse_events(optarg); - break; - default: - break; - } - } - if (optind == argc) - goto err; - - if (!nr_counters) - nr_counters = 8; - return; - -err: - display_help(); -} - -char fault_here[1000000]; - -static int fd[MAX_NR_CPUS][MAX_COUNTERS]; - -static void create_counter(int counter) -{ - struct perf_counter_hw_event hw_event; - - memset(&hw_event, 0, sizeof(hw_event)); - hw_event.type = event_id[counter]; - hw_event.raw = event_raw[counter]; - hw_event.record_type = PERF_RECORD_SIMPLE; - hw_event.nmi = 0; - - if (system_wide) { - int cpu; - for (cpu = 0; cpu < nr_cpus; cpu ++) { - fd[cpu][counter] = sys_perf_counter_open(&hw_event, -1, cpu, -1, 0); - if (fd[cpu][counter] < 0) { - printf("perfstat error: syscall returned with %d (%s)\n", - fd[cpu][counter], strerror(errno)); - exit(-1); - } - - } - } else { - hw_event.inherit = 1; - hw_event.disabled = 1; - - fd[0][counter] = sys_perf_counter_open(&hw_event, 0, -1, -1, 0); - if (fd[0][counter] < 0) { - printf("perfstat error: syscall returned with %d (%s)\n", - fd[0][counter], strerror(errno)); - exit(-1); - } - } -} - - -int main(int argc, char *argv[]) -{ - unsigned long long t0, t1; - int counter; - ssize_t res; - int status; - int pid; - - process_options(argc, argv); - - if (system_wide) { - nr_cpus = sysconf(_SC_NPROCESSORS_ONLN); - assert(nr_cpus <= MAX_NR_CPUS); - assert(nr_cpus >= 0); - } else - nr_cpus = 1; - - for (counter = 0; counter < nr_counters; counter++) - create_counter(counter); - - argc -= optind; - argv += optind; - - /* - * Enable counters and exec the command: - */ - t0 = rdclock(); - prctl(PR_TASK_PERF_COUNTERS_ENABLE); - - if ((pid = fork()) < 0) - perror("failed to fork"); - if (!pid) { - if (execvp(argv[0], argv)) { - perror(argv[0]); - exit(-1); - } - } - while (wait(&status) >= 0) - ; - prctl(PR_TASK_PERF_COUNTERS_DISABLE); - t1 = rdclock(); - - fflush(stdout); - - fprintf(stderr, "\n"); - fprintf(stderr, " Performance counter stats for \'%s\':\n", - argv[0]); - fprintf(stderr, "\n"); - - for (counter = 0; counter < nr_counters; counter++) { - int cpu; - __u64 count, single_count; - - count = 0; - for (cpu = 0; cpu < nr_cpus; cpu ++) { - res = read(fd[cpu][counter], - (char *) &single_count, sizeof(single_count)); - assert(res == sizeof(single_count)); - count += single_count; - } - - if (!event_raw[counter] && - (event_id[counter] == PERF_COUNT_CPU_CLOCK || - event_id[counter] == PERF_COUNT_TASK_CLOCK)) { - - double msecs = (double)count / 1000000; - - fprintf(stderr, " %14.6f %-20s (msecs)\n", - msecs, event_name(counter)); - } else { - fprintf(stderr, " %14Ld %-20s (events)\n", - count, event_name(counter)); - } - if (!counter) - fprintf(stderr, "\n"); - } - fprintf(stderr, "\n"); - fprintf(stderr, " Wall-clock time elapsed: %12.6f msecs\n", - (double)(t1-t0)/1e6); - fprintf(stderr, "\n"); - - return 0; -}