From 308197b9474bcde2cafba2cd19bef46e0c0428bd Mon Sep 17 00:00:00 2001 From: Rabin Vincent Date: Tue, 2 Dec 2014 16:50:39 +0100 Subject: perf bench: Prepare memcpy for merge The memset benchmark is largely copy-pasted from the memcpy benchmark. Prepare the memcpy file for merge with memset by extracting out a generic function. Signed-off-by: Rabin Vincent Cc: Ingo Molnar Cc: Paul Mackerras Cc: Peter Zijlstra Cc: Rabin Vincent Link: http://lkml.kernel.org/r/1417535441-3965-1-git-send-email-rabin.vincent@axis.com Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/bench/mem-memcpy.c | 182 ++++++++++++++++++++++++------------------ 1 file changed, 104 insertions(+), 78 deletions(-) (limited to 'tools/perf') diff --git a/tools/perf/bench/mem-memcpy.c b/tools/perf/bench/mem-memcpy.c index 2465141b554b..4a4493a7a1f8 100644 --- a/tools/perf/bench/mem-memcpy.c +++ b/tools/perf/bench/mem-memcpy.c @@ -52,16 +52,18 @@ typedef void *(*memcpy_t)(void *, const void *, size_t); struct routine { const char *name; const char *desc; - memcpy_t fn; + union { + memcpy_t memcpy; + } fn; }; -struct routine routines[] = { - { "default", - "Default memcpy() provided by glibc", - memcpy }, +struct routine memcpy_routines[] = { + { .name = "default", + .desc = "Default memcpy() provided by glibc", + .fn.memcpy = memcpy }, #ifdef HAVE_ARCH_X86_64_SUPPORT -#define MEMCPY_FN(fn, name, desc) { name, desc, fn }, +#define MEMCPY_FN(_fn, _name, _desc) {.name = _name, .desc = _desc, .fn.memcpy = _fn}, #include "mem-memcpy-x86-64-asm-def.h" #undef MEMCPY_FN @@ -69,7 +71,7 @@ struct routine routines[] = { { NULL, NULL, - NULL } + {NULL} } }; static const char * const bench_mem_memcpy_usage[] = { @@ -110,63 +112,6 @@ static double timeval2double(struct timeval *ts) (double)ts->tv_usec / (double)1000000; } -static void alloc_mem(void **dst, void **src, size_t length) -{ - *dst = zalloc(length); - if (!*dst) - die("memory allocation failed - maybe length is too large?\n"); - - *src = zalloc(length); - if (!*src) - die("memory allocation failed - maybe length is too large?\n"); - /* Make sure to always replace the zero pages even if MMAP_THRESH is crossed */ - memset(*src, 0, length); -} - -static u64 do_memcpy_cycle(memcpy_t fn, size_t len, bool prefault) -{ - u64 cycle_start = 0ULL, cycle_end = 0ULL; - void *src = NULL, *dst = NULL; - int i; - - alloc_mem(&src, &dst, len); - - if (prefault) - fn(dst, src, len); - - cycle_start = get_cycle(); - for (i = 0; i < iterations; ++i) - fn(dst, src, len); - cycle_end = get_cycle(); - - free(src); - free(dst); - return cycle_end - cycle_start; -} - -static double do_memcpy_gettimeofday(memcpy_t fn, size_t len, bool prefault) -{ - struct timeval tv_start, tv_end, tv_diff; - void *src = NULL, *dst = NULL; - int i; - - alloc_mem(&src, &dst, len); - - if (prefault) - fn(dst, src, len); - - BUG_ON(gettimeofday(&tv_start, NULL)); - for (i = 0; i < iterations; ++i) - fn(dst, src, len); - BUG_ON(gettimeofday(&tv_end, NULL)); - - timersub(&tv_end, &tv_start, &tv_diff); - - free(src); - free(dst); - return (double)((double)len / timeval2double(&tv_diff)); -} - #define pf (no_prefault ? 0 : 1) #define print_bps(x) do { \ @@ -180,8 +125,16 @@ static double do_memcpy_gettimeofday(memcpy_t fn, size_t len, bool prefault) printf(" %14lf GB/Sec", x / K / K / K); \ } while (0) -int bench_mem_memcpy(int argc, const char **argv, - const char *prefix __maybe_unused) +struct bench_mem_info { + const struct routine *routines; + u64 (*do_cycle)(const struct routine *r, size_t len, bool prefault); + double (*do_gettimeofday)(const struct routine *r, size_t len, bool prefault); + const char *const *usage; +}; + +static int bench_mem_common(int argc, const char **argv, + const char *prefix __maybe_unused, + struct bench_mem_info *info) { int i; size_t len; @@ -189,7 +142,7 @@ int bench_mem_memcpy(int argc, const char **argv, u64 result_cycle[2]; argc = parse_options(argc, argv, options, - bench_mem_memcpy_usage, 0); + info->usage, 0); if (no_prefault && only_prefault) { fprintf(stderr, "Invalid options: -o and -n are mutually exclusive\n"); @@ -213,16 +166,16 @@ int bench_mem_memcpy(int argc, const char **argv, if (only_prefault && no_prefault) only_prefault = no_prefault = false; - for (i = 0; routines[i].name; i++) { - if (!strcmp(routines[i].name, routine)) + for (i = 0; info->routines[i].name; i++) { + if (!strcmp(info->routines[i].name, routine)) break; } - if (!routines[i].name) { + if (!info->routines[i].name) { printf("Unknown routine:%s\n", routine); printf("Available routines...\n"); - for (i = 0; routines[i].name; i++) { + for (i = 0; info->routines[i].name; i++) { printf("\t%s ... %s\n", - routines[i].name, routines[i].desc); + info->routines[i].name, info->routines[i].desc); } return 1; } @@ -234,25 +187,25 @@ int bench_mem_memcpy(int argc, const char **argv, /* show both of results */ if (use_cycle) { result_cycle[0] = - do_memcpy_cycle(routines[i].fn, len, false); + info->do_cycle(&info->routines[i], len, false); result_cycle[1] = - do_memcpy_cycle(routines[i].fn, len, true); + info->do_cycle(&info->routines[i], len, true); } else { result_bps[0] = - do_memcpy_gettimeofday(routines[i].fn, + info->do_gettimeofday(&info->routines[i], len, false); result_bps[1] = - do_memcpy_gettimeofday(routines[i].fn, + info->do_gettimeofday(&info->routines[i], len, true); } } else { if (use_cycle) { result_cycle[pf] = - do_memcpy_cycle(routines[i].fn, + info->do_cycle(&info->routines[i], len, only_prefault); } else { result_bps[pf] = - do_memcpy_gettimeofday(routines[i].fn, + info->do_gettimeofday(&info->routines[i], len, only_prefault); } } @@ -310,3 +263,76 @@ int bench_mem_memcpy(int argc, const char **argv, return 0; } + +static void memcpy_alloc_mem(void **dst, void **src, size_t length) +{ + *dst = zalloc(length); + if (!*dst) + die("memory allocation failed - maybe length is too large?\n"); + + *src = zalloc(length); + if (!*src) + die("memory allocation failed - maybe length is too large?\n"); + /* Make sure to always replace the zero pages even if MMAP_THRESH is crossed */ + memset(*src, 0, length); +} + +static u64 do_memcpy_cycle(const struct routine *r, size_t len, bool prefault) +{ + u64 cycle_start = 0ULL, cycle_end = 0ULL; + void *src = NULL, *dst = NULL; + memcpy_t fn = r->fn.memcpy; + int i; + + memcpy_alloc_mem(&src, &dst, len); + + if (prefault) + fn(dst, src, len); + + cycle_start = get_cycle(); + for (i = 0; i < iterations; ++i) + fn(dst, src, len); + cycle_end = get_cycle(); + + free(src); + free(dst); + return cycle_end - cycle_start; +} + +static double do_memcpy_gettimeofday(const struct routine *r, size_t len, + bool prefault) +{ + struct timeval tv_start, tv_end, tv_diff; + memcpy_t fn = r->fn.memcpy; + void *src = NULL, *dst = NULL; + int i; + + memcpy_alloc_mem(&src, &dst, len); + + if (prefault) + fn(dst, src, len); + + BUG_ON(gettimeofday(&tv_start, NULL)); + for (i = 0; i < iterations; ++i) + fn(dst, src, len); + BUG_ON(gettimeofday(&tv_end, NULL)); + + timersub(&tv_end, &tv_start, &tv_diff); + + free(src); + free(dst); + return (double)((double)len / timeval2double(&tv_diff)); +} + +int bench_mem_memcpy(int argc, const char **argv, + const char *prefix __maybe_unused) +{ + struct bench_mem_info info = { + .routines = memcpy_routines, + .do_cycle = do_memcpy_cycle, + .do_gettimeofday = do_memcpy_gettimeofday, + .usage = bench_mem_memcpy_usage, + }; + + return bench_mem_common(argc, argv, prefix, &info); +} -- cgit v1.2.3 From 5bce1a5772cb52aad7e0466484ba07cfbfec2478 Mon Sep 17 00:00:00 2001 From: Rabin Vincent Date: Tue, 2 Dec 2014 16:50:40 +0100 Subject: perf bench: Merge memset into memcpy The memset benchmark is largely copy-pasted from the memcpy benchmark. Merge the two now that memcpy is made more generic. Signed-off-by: Rabin Vincent Cc: Ingo Molnar Cc: Paul Mackerras Cc: Peter Zijlstra Cc: Rabin Vincent Link: http://lkml.kernel.org/r/1417535441-3965-2-git-send-email-rabin.vincent@axis.com Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/Makefile.perf | 1 - tools/perf/bench/mem-memcpy.c | 90 +++++++++++++ tools/perf/bench/mem-memset.c | 304 ------------------------------------------ 3 files changed, 90 insertions(+), 305 deletions(-) delete mode 100644 tools/perf/bench/mem-memset.c (limited to 'tools/perf') diff --git a/tools/perf/Makefile.perf b/tools/perf/Makefile.perf index 478efa9b2364..763e68fb5767 100644 --- a/tools/perf/Makefile.perf +++ b/tools/perf/Makefile.perf @@ -458,7 +458,6 @@ BUILTIN_OBJS += $(OUTPUT)bench/mem-memcpy-x86-64-asm.o BUILTIN_OBJS += $(OUTPUT)bench/mem-memset-x86-64-asm.o endif BUILTIN_OBJS += $(OUTPUT)bench/mem-memcpy.o -BUILTIN_OBJS += $(OUTPUT)bench/mem-memset.o BUILTIN_OBJS += $(OUTPUT)bench/futex-hash.o BUILTIN_OBJS += $(OUTPUT)bench/futex-wake.o BUILTIN_OBJS += $(OUTPUT)bench/futex-requeue.o diff --git a/tools/perf/bench/mem-memcpy.c b/tools/perf/bench/mem-memcpy.c index 4a4493a7a1f8..e18be70c8a47 100644 --- a/tools/perf/bench/mem-memcpy.c +++ b/tools/perf/bench/mem-memcpy.c @@ -13,6 +13,7 @@ #include "../util/cloexec.h" #include "bench.h" #include "mem-memcpy-arch.h" +#include "mem-memset-arch.h" #include #include @@ -48,12 +49,14 @@ static const struct option options[] = { }; typedef void *(*memcpy_t)(void *, const void *, size_t); +typedef void *(*memset_t)(void *, int, size_t); struct routine { const char *name; const char *desc; union { memcpy_t memcpy; + memset_t memset; } fn; }; @@ -336,3 +339,90 @@ int bench_mem_memcpy(int argc, const char **argv, return bench_mem_common(argc, argv, prefix, &info); } + +static void memset_alloc_mem(void **dst, size_t length) +{ + *dst = zalloc(length); + if (!*dst) + die("memory allocation failed - maybe length is too large?\n"); +} + +static u64 do_memset_cycle(const struct routine *r, size_t len, bool prefault) +{ + u64 cycle_start = 0ULL, cycle_end = 0ULL; + memset_t fn = r->fn.memset; + void *dst = NULL; + int i; + + memset_alloc_mem(&dst, len); + + if (prefault) + fn(dst, -1, len); + + cycle_start = get_cycle(); + for (i = 0; i < iterations; ++i) + fn(dst, i, len); + cycle_end = get_cycle(); + + free(dst); + return cycle_end - cycle_start; +} + +static double do_memset_gettimeofday(const struct routine *r, size_t len, + bool prefault) +{ + struct timeval tv_start, tv_end, tv_diff; + memset_t fn = r->fn.memset; + void *dst = NULL; + int i; + + memset_alloc_mem(&dst, len); + + if (prefault) + fn(dst, -1, len); + + BUG_ON(gettimeofday(&tv_start, NULL)); + for (i = 0; i < iterations; ++i) + fn(dst, i, len); + BUG_ON(gettimeofday(&tv_end, NULL)); + + timersub(&tv_end, &tv_start, &tv_diff); + + free(dst); + return (double)((double)len / timeval2double(&tv_diff)); +} + +static const char * const bench_mem_memset_usage[] = { + "perf bench mem memset ", + NULL +}; + +static const struct routine memset_routines[] = { + { .name ="default", + .desc = "Default memset() provided by glibc", + .fn.memset = memset }, +#ifdef HAVE_ARCH_X86_64_SUPPORT + +#define MEMSET_FN(_fn, _name, _desc) { .name = _name, .desc = _desc, .fn.memset = _fn }, +#include "mem-memset-x86-64-asm-def.h" +#undef MEMSET_FN + +#endif + + { .name = NULL, + .desc = NULL, + .fn.memset = NULL } +}; + +int bench_mem_memset(int argc, const char **argv, + const char *prefix __maybe_unused) +{ + struct bench_mem_info info = { + .routines = memset_routines, + .do_cycle = do_memset_cycle, + .do_gettimeofday = do_memset_gettimeofday, + .usage = bench_mem_memset_usage, + }; + + return bench_mem_common(argc, argv, prefix, &info); +} diff --git a/tools/perf/bench/mem-memset.c b/tools/perf/bench/mem-memset.c deleted file mode 100644 index 75fc3e65fb2a..000000000000 --- a/tools/perf/bench/mem-memset.c +++ /dev/null @@ -1,304 +0,0 @@ -/* - * mem-memset.c - * - * memset: Simple memory set in various ways - * - * Trivial clone of mem-memcpy.c. - */ - -#include "../perf.h" -#include "../util/util.h" -#include "../util/parse-options.h" -#include "../util/header.h" -#include "../util/cloexec.h" -#include "bench.h" -#include "mem-memset-arch.h" - -#include -#include -#include -#include -#include - -#define K 1024 - -static const char *length_str = "1MB"; -static const char *routine = "default"; -static int iterations = 1; -static bool use_cycle; -static int cycle_fd; -static bool only_prefault; -static bool no_prefault; - -static const struct option options[] = { - OPT_STRING('l', "length", &length_str, "1MB", - "Specify length of memory to set. " - "Available units: B, KB, MB, GB and TB (upper and lower)"), - OPT_STRING('r', "routine", &routine, "default", - "Specify routine to set"), - OPT_INTEGER('i', "iterations", &iterations, - "repeat memset() invocation this number of times"), - OPT_BOOLEAN('c', "cycle", &use_cycle, - "Use cycles event instead of gettimeofday() for measuring"), - OPT_BOOLEAN('o', "only-prefault", &only_prefault, - "Show only the result with page faults before memset()"), - OPT_BOOLEAN('n', "no-prefault", &no_prefault, - "Show only the result without page faults before memset()"), - OPT_END() -}; - -typedef void *(*memset_t)(void *, int, size_t); - -struct routine { - const char *name; - const char *desc; - memset_t fn; -}; - -static const struct routine routines[] = { - { "default", - "Default memset() provided by glibc", - memset }, -#ifdef HAVE_ARCH_X86_64_SUPPORT - -#define MEMSET_FN(fn, name, desc) { name, desc, fn }, -#include "mem-memset-x86-64-asm-def.h" -#undef MEMSET_FN - -#endif - - { NULL, - NULL, - NULL } -}; - -static const char * const bench_mem_memset_usage[] = { - "perf bench mem memset ", - NULL -}; - -static struct perf_event_attr cycle_attr = { - .type = PERF_TYPE_HARDWARE, - .config = PERF_COUNT_HW_CPU_CYCLES -}; - -static void init_cycle(void) -{ - cycle_fd = sys_perf_event_open(&cycle_attr, getpid(), -1, -1, - perf_event_open_cloexec_flag()); - - if (cycle_fd < 0 && errno == ENOSYS) - die("No CONFIG_PERF_EVENTS=y kernel support configured?\n"); - else - BUG_ON(cycle_fd < 0); -} - -static u64 get_cycle(void) -{ - int ret; - u64 clk; - - ret = read(cycle_fd, &clk, sizeof(u64)); - BUG_ON(ret != sizeof(u64)); - - return clk; -} - -static double timeval2double(struct timeval *ts) -{ - return (double)ts->tv_sec + - (double)ts->tv_usec / (double)1000000; -} - -static void alloc_mem(void **dst, size_t length) -{ - *dst = zalloc(length); - if (!*dst) - die("memory allocation failed - maybe length is too large?\n"); -} - -static u64 do_memset_cycle(memset_t fn, size_t len, bool prefault) -{ - u64 cycle_start = 0ULL, cycle_end = 0ULL; - void *dst = NULL; - int i; - - alloc_mem(&dst, len); - - if (prefault) - fn(dst, -1, len); - - cycle_start = get_cycle(); - for (i = 0; i < iterations; ++i) - fn(dst, i, len); - cycle_end = get_cycle(); - - free(dst); - return cycle_end - cycle_start; -} - -static double do_memset_gettimeofday(memset_t fn, size_t len, bool prefault) -{ - struct timeval tv_start, tv_end, tv_diff; - void *dst = NULL; - int i; - - alloc_mem(&dst, len); - - if (prefault) - fn(dst, -1, len); - - BUG_ON(gettimeofday(&tv_start, NULL)); - for (i = 0; i < iterations; ++i) - fn(dst, i, len); - BUG_ON(gettimeofday(&tv_end, NULL)); - - timersub(&tv_end, &tv_start, &tv_diff); - - free(dst); - return (double)((double)len / timeval2double(&tv_diff)); -} - -#define pf (no_prefault ? 0 : 1) - -#define print_bps(x) do { \ - if (x < K) \ - printf(" %14lf B/Sec", x); \ - else if (x < K * K) \ - printf(" %14lfd KB/Sec", x / K); \ - else if (x < K * K * K) \ - printf(" %14lf MB/Sec", x / K / K); \ - else \ - printf(" %14lf GB/Sec", x / K / K / K); \ - } while (0) - -int bench_mem_memset(int argc, const char **argv, - const char *prefix __maybe_unused) -{ - int i; - size_t len; - double result_bps[2]; - u64 result_cycle[2]; - - argc = parse_options(argc, argv, options, - bench_mem_memset_usage, 0); - - if (no_prefault && only_prefault) { - fprintf(stderr, "Invalid options: -o and -n are mutually exclusive\n"); - return 1; - } - - if (use_cycle) - init_cycle(); - - len = (size_t)perf_atoll((char *)length_str); - - result_cycle[0] = result_cycle[1] = 0ULL; - result_bps[0] = result_bps[1] = 0.0; - - if ((s64)len <= 0) { - fprintf(stderr, "Invalid length:%s\n", length_str); - return 1; - } - - /* same to without specifying either of prefault and no-prefault */ - if (only_prefault && no_prefault) - only_prefault = no_prefault = false; - - for (i = 0; routines[i].name; i++) { - if (!strcmp(routines[i].name, routine)) - break; - } - if (!routines[i].name) { - printf("Unknown routine:%s\n", routine); - printf("Available routines...\n"); - for (i = 0; routines[i].name; i++) { - printf("\t%s ... %s\n", - routines[i].name, routines[i].desc); - } - return 1; - } - - if (bench_format == BENCH_FORMAT_DEFAULT) - printf("# Copying %s Bytes ...\n\n", length_str); - - if (!only_prefault && !no_prefault) { - /* show both of results */ - if (use_cycle) { - result_cycle[0] = - do_memset_cycle(routines[i].fn, len, false); - result_cycle[1] = - do_memset_cycle(routines[i].fn, len, true); - } else { - result_bps[0] = - do_memset_gettimeofday(routines[i].fn, - len, false); - result_bps[1] = - do_memset_gettimeofday(routines[i].fn, - len, true); - } - } else { - if (use_cycle) { - result_cycle[pf] = - do_memset_cycle(routines[i].fn, - len, only_prefault); - } else { - result_bps[pf] = - do_memset_gettimeofday(routines[i].fn, - len, only_prefault); - } - } - - switch (bench_format) { - case BENCH_FORMAT_DEFAULT: - if (!only_prefault && !no_prefault) { - if (use_cycle) { - printf(" %14lf Cycle/Byte\n", - (double)result_cycle[0] - / (double)len); - printf(" %14lf Cycle/Byte (with prefault)\n ", - (double)result_cycle[1] - / (double)len); - } else { - print_bps(result_bps[0]); - printf("\n"); - print_bps(result_bps[1]); - printf(" (with prefault)\n"); - } - } else { - if (use_cycle) { - printf(" %14lf Cycle/Byte", - (double)result_cycle[pf] - / (double)len); - } else - print_bps(result_bps[pf]); - - printf("%s\n", only_prefault ? " (with prefault)" : ""); - } - break; - case BENCH_FORMAT_SIMPLE: - if (!only_prefault && !no_prefault) { - if (use_cycle) { - printf("%lf %lf\n", - (double)result_cycle[0] / (double)len, - (double)result_cycle[1] / (double)len); - } else { - printf("%lf %lf\n", - result_bps[0], result_bps[1]); - } - } else { - if (use_cycle) { - printf("%lf\n", (double)result_cycle[pf] - / (double)len); - } else - printf("%lf\n", result_bps[pf]); - } - break; - default: - /* reaching this means there's some disaster: */ - die("unknown format: %d\n", bench_format); - break; - } - - return 0; -} -- cgit v1.2.3 From 1182f883113483cefbc3be0178a2df2dc9ae8b77 Mon Sep 17 00:00:00 2001 From: Rabin Vincent Date: Tue, 2 Dec 2014 16:50:41 +0100 Subject: perf bench: Fix memcpy/memset output The memcpy and memset benchmarks return bogus results when iterations > 0 because the iterations value is not taken into account when calculating the final result: $ perf bench mem memset --only-prefault --length 1GB --iterations 1 # Running 'mem/memset' benchmark: # Copying 1GB Bytes ... 20.798669 GB/Sec (with prefault) $ perf bench mem memset --only-prefault --length 1GB --iterations 10 # Running 'mem/memset' benchmark: # Copying 1GB Bytes ... 2.086576 GB/Sec (with prefault) $ perf bench mem memset --only-prefault --length 1GB --iterations 100 # Running 'mem/memset' benchmark: # Copying 1GB Bytes ... 212.840917 MB/Sec (with prefault) Fix this. Signed-off-by: Rabin Vincent Acked-by: Ingo Molnar Cc: Ingo Molnar Cc: Paul Mackerras Cc: Peter Zijlstra Cc: Rabin Vincent Cc: Rabin Vincent Link: http://lkml.kernel.org/r/1417535441-3965-3-git-send-email-rabin.vincent@axis.com Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/bench/mem-memcpy.c | 18 ++++++++++-------- 1 file changed, 10 insertions(+), 8 deletions(-) (limited to 'tools/perf') diff --git a/tools/perf/bench/mem-memcpy.c b/tools/perf/bench/mem-memcpy.c index e18be70c8a47..6c14afe8c1b1 100644 --- a/tools/perf/bench/mem-memcpy.c +++ b/tools/perf/bench/mem-memcpy.c @@ -141,6 +141,7 @@ static int bench_mem_common(int argc, const char **argv, { int i; size_t len; + double totallen; double result_bps[2]; u64 result_cycle[2]; @@ -156,6 +157,7 @@ static int bench_mem_common(int argc, const char **argv, init_cycle(); len = (size_t)perf_atoll((char *)length_str); + totallen = (double)len * iterations; result_cycle[0] = result_cycle[1] = 0ULL; result_bps[0] = result_bps[1] = 0.0; @@ -219,10 +221,10 @@ static int bench_mem_common(int argc, const char **argv, if (use_cycle) { printf(" %14lf Cycle/Byte\n", (double)result_cycle[0] - / (double)len); + / totallen); printf(" %14lf Cycle/Byte (with prefault)\n", (double)result_cycle[1] - / (double)len); + / totallen); } else { print_bps(result_bps[0]); printf("\n"); @@ -233,7 +235,7 @@ static int bench_mem_common(int argc, const char **argv, if (use_cycle) { printf(" %14lf Cycle/Byte", (double)result_cycle[pf] - / (double)len); + / totallen); } else print_bps(result_bps[pf]); @@ -244,8 +246,8 @@ static int bench_mem_common(int argc, const char **argv, if (!only_prefault && !no_prefault) { if (use_cycle) { printf("%lf %lf\n", - (double)result_cycle[0] / (double)len, - (double)result_cycle[1] / (double)len); + (double)result_cycle[0] / totallen, + (double)result_cycle[1] / totallen); } else { printf("%lf %lf\n", result_bps[0], result_bps[1]); @@ -253,7 +255,7 @@ static int bench_mem_common(int argc, const char **argv, } else { if (use_cycle) { printf("%lf\n", (double)result_cycle[pf] - / (double)len); + / totallen); } else printf("%lf\n", result_bps[pf]); } @@ -324,7 +326,7 @@ static double do_memcpy_gettimeofday(const struct routine *r, size_t len, free(src); free(dst); - return (double)((double)len / timeval2double(&tv_diff)); + return (double)(((double)len * iterations) / timeval2double(&tv_diff)); } int bench_mem_memcpy(int argc, const char **argv, @@ -389,7 +391,7 @@ static double do_memset_gettimeofday(const struct routine *r, size_t len, timersub(&tv_end, &tv_start, &tv_diff); free(dst); - return (double)((double)len / timeval2double(&tv_diff)); + return (double)(((double)len * iterations) / timeval2double(&tv_diff)); } static const char * const bench_mem_memset_usage[] = { -- cgit v1.2.3 From e641f696ebe4514efac00d1f4a485a06c8e94eea Mon Sep 17 00:00:00 2001 From: Tom Huynh Date: Tue, 2 Dec 2014 11:37:22 -0600 Subject: perf hists browser: Change print format from %lu to %PRIu64 The nr_events variable in tools/perf/ui/browsers/hists.c is of type u64, so the print format (%lu) causes 'perf report' to show 0 event count when running with 32-bit userspace without redirection. This patch fixes that problem by printing nr_events as PRIu64. Signed-off-by: Tom Huynh Cc: Ingo Molnar Cc: Jiri Olsa Cc: Kim Phillips Cc: Matt Mullins Cc: Namhyung Kim Cc: Paul Mackerras Cc: Peter Zijlstra Link: http://lkml.kernel.org/r/1417541842-9747-1-git-send-email-tom.huynh@freescale.com Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/ui/browsers/hists.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'tools/perf') diff --git a/tools/perf/ui/browsers/hists.c b/tools/perf/ui/browsers/hists.c index 502daff76ceb..e6bb04b5b09b 100644 --- a/tools/perf/ui/browsers/hists.c +++ b/tools/perf/ui/browsers/hists.c @@ -1252,7 +1252,7 @@ static int hists__browser_title(struct hists *hists, nr_samples = convert_unit(nr_samples, &unit); printed = scnprintf(bf, size, - "Samples: %lu%c of event '%s', Event count (approx.): %lu", + "Samples: %lu%c of event '%s', Event count (approx.): %" PRIu64, nr_samples, unit, ev_name, nr_events); -- cgit v1.2.3 From cfd31d85bb42b96449157bd57c638dc779070753 Mon Sep 17 00:00:00 2001 From: Jiri Olsa Date: Mon, 1 Dec 2014 20:06:22 +0100 Subject: perf tools: Use single strcmp call instead of two There's no need to use 2 strcmp calls, one is enough. Signed-off-by: Jiri Olsa Cc: Corey Ashford Cc: David Ahern Cc: Frederic Weisbecker Cc: Ingo Molnar Cc: Namhyung Kim Cc: Paul Mackerras Cc: Peter Zijlstra Cc: Stephane Eranian Cc: Steven Rostedt Link: http://lkml.kernel.org/r/1417460789-13874-2-git-send-email-jolsa@kernel.org Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/util/config.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'tools/perf') diff --git a/tools/perf/util/config.c b/tools/perf/util/config.c index 57ff826f150b..c802236b2bfe 100644 --- a/tools/perf/util/config.c +++ b/tools/perf/util/config.c @@ -522,7 +522,7 @@ static int buildid_dir_command_config(const char *var, const char *value, const char *v; /* same dir for all commands */ - if (!prefixcmp(var, "buildid.") && !strcmp(var + 8, "dir")) { + if (!strcmp(var, "buildid.dir")) { v = perf_config_dirname(var, value); if (!v) return -1; -- cgit v1.2.3 From 498922adf1173ddeebd155f82646d4a9d518d606 Mon Sep 17 00:00:00 2001 From: Jiri Olsa Date: Mon, 1 Dec 2014 20:06:23 +0100 Subject: perf buildid-cache: Remove extra debugdir variables There's no need to copy over the buildid_dir into separate variable with no change. This is leftover from commit: 45de34bbe3e1 perf buildid: add perfconfig option to specify buildid cache dir that added global buildid_dir variable that holds cache directory, but did not cleanup the debugdir copies. Signed-off-by: Jiri Olsa Cc: Corey Ashford Cc: David Ahern Cc: Frederic Weisbecker Cc: Ingo Molnar Cc: Namhyung Kim Cc: Paul Mackerras Cc: Peter Zijlstra Cc: Stephane Eranian Cc: Steven Rostedt Link: http://lkml.kernel.org/r/1417460789-13874-3-git-send-email-jolsa@kernel.org Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/builtin-buildid-cache.c | 11 ++++------- tools/perf/util/build-id.c | 9 +++------ 2 files changed, 7 insertions(+), 13 deletions(-) (limited to 'tools/perf') diff --git a/tools/perf/builtin-buildid-cache.c b/tools/perf/builtin-buildid-cache.c index 70385756da63..29f24c071bc6 100644 --- a/tools/perf/builtin-buildid-cache.c +++ b/tools/perf/builtin-buildid-cache.c @@ -285,7 +285,6 @@ int cmd_buildid_cache(int argc, const char **argv, struct str_node *pos; int ret = 0; bool force = false; - char debugdir[PATH_MAX]; char const *add_name_list_str = NULL, *remove_name_list_str = NULL, *missing_filename = NULL, @@ -335,13 +334,11 @@ int cmd_buildid_cache(int argc, const char **argv, setup_pager(); - snprintf(debugdir, sizeof(debugdir), "%s", buildid_dir); - if (add_name_list_str) { list = strlist__new(true, add_name_list_str); if (list) { strlist__for_each(pos, list) - if (build_id_cache__add_file(pos->s, debugdir)) { + if (build_id_cache__add_file(pos->s, buildid_dir)) { if (errno == EEXIST) { pr_debug("%s already in the cache\n", pos->s); @@ -359,7 +356,7 @@ int cmd_buildid_cache(int argc, const char **argv, list = strlist__new(true, remove_name_list_str); if (list) { strlist__for_each(pos, list) - if (build_id_cache__remove_file(pos->s, debugdir)) { + if (build_id_cache__remove_file(pos->s, buildid_dir)) { if (errno == ENOENT) { pr_debug("%s wasn't in the cache\n", pos->s); @@ -380,7 +377,7 @@ int cmd_buildid_cache(int argc, const char **argv, list = strlist__new(true, update_name_list_str); if (list) { strlist__for_each(pos, list) - if (build_id_cache__update_file(pos->s, debugdir)) { + if (build_id_cache__update_file(pos->s, buildid_dir)) { if (errno == ENOENT) { pr_debug("%s wasn't in the cache\n", pos->s); @@ -395,7 +392,7 @@ int cmd_buildid_cache(int argc, const char **argv, } if (kcore_filename && - build_id_cache__add_kcore(kcore_filename, debugdir, force)) + build_id_cache__add_kcore(kcore_filename, buildid_dir, force)) pr_warning("Couldn't add %s\n", kcore_filename); out: diff --git a/tools/perf/util/build-id.c b/tools/perf/util/build-id.c index e8d79e5bfaf7..0c72680a977f 100644 --- a/tools/perf/util/build-id.c +++ b/tools/perf/util/build-id.c @@ -410,21 +410,18 @@ int perf_session__cache_build_ids(struct perf_session *session) { struct rb_node *nd; int ret; - char debugdir[PATH_MAX]; if (no_buildid_cache) return 0; - snprintf(debugdir, sizeof(debugdir), "%s", buildid_dir); - - if (mkdir(debugdir, 0755) != 0 && errno != EEXIST) + if (mkdir(buildid_dir, 0755) != 0 && errno != EEXIST) return -1; - ret = machine__cache_build_ids(&session->machines.host, debugdir); + ret = machine__cache_build_ids(&session->machines.host, buildid_dir); for (nd = rb_first(&session->machines.guests); nd; nd = rb_next(nd)) { struct machine *pos = rb_entry(nd, struct machine, rb_node); - ret |= machine__cache_build_ids(pos, debugdir); + ret |= machine__cache_build_ids(pos, buildid_dir); } return ret ? -1 : 0; } -- cgit v1.2.3 From eec5a688f426d6fe4097feb0b916ad41803d2ebb Mon Sep 17 00:00:00 2001 From: Jiri Olsa Date: Mon, 1 Dec 2014 20:06:27 +0100 Subject: perf buildid cache: Fix -a segfault related to kcore handling The kcore_filename is uninitialized and trash value could trigger build_id_cache__add_kcore function ending up with segfault. Signed-off-by: Jiri Olsa Cc: Adrian Hunter Cc: Corey Ashford Cc: David Ahern Cc: Frederic Weisbecker Cc: Ingo Molnar Cc: Namhyung Kim Cc: Paul Mackerras Cc: Peter Zijlstra Cc: Stephane Eranian Cc: Steven Rostedt Link: http://lkml.kernel.org/r/1417460789-13874-7-git-send-email-jolsa@kernel.org Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/builtin-buildid-cache.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'tools/perf') diff --git a/tools/perf/builtin-buildid-cache.c b/tools/perf/builtin-buildid-cache.c index 29f24c071bc6..77d5cae54c6a 100644 --- a/tools/perf/builtin-buildid-cache.c +++ b/tools/perf/builtin-buildid-cache.c @@ -289,7 +289,7 @@ int cmd_buildid_cache(int argc, const char **argv, *remove_name_list_str = NULL, *missing_filename = NULL, *update_name_list_str = NULL, - *kcore_filename; + *kcore_filename = NULL; char sbuf[STRERR_BUFSIZE]; struct perf_data_file file = { -- cgit v1.2.3 From 99ce8e9fce99147f865cda8a8e471900518c9a49 Mon Sep 17 00:00:00 2001 From: Jiri Olsa Date: Mon, 1 Dec 2014 20:06:24 +0100 Subject: perf tools: Add --buildid-dir option to set cache directory Adding --buildid-dir to be able to set specific cache directory. It's going to be handy for buildid tests coming in shortly. Signed-off-by: Jiri Olsa Cc: Corey Ashford Cc: David Ahern Cc: Frederic Weisbecker Cc: Ingo Molnar Cc: Namhyung Kim Cc: Paul Mackerras Cc: Peter Zijlstra Cc: Stephane Eranian Cc: Steven Rostedt Link: http://lkml.kernel.org/r/1417460789-13874-4-git-send-email-jolsa@kernel.org Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/Documentation/perf.txt | 4 ++++ tools/perf/perf.c | 14 ++++++++++++-- tools/perf/util/config.c | 8 +++++--- tools/perf/util/util.h | 2 +- 4 files changed, 22 insertions(+), 6 deletions(-) (limited to 'tools/perf') diff --git a/tools/perf/Documentation/perf.txt b/tools/perf/Documentation/perf.txt index d240bb2e5b22..1e8e400b4493 100644 --- a/tools/perf/Documentation/perf.txt +++ b/tools/perf/Documentation/perf.txt @@ -18,6 +18,10 @@ OPTIONS --debug verbose # sets verbose = 1 --debug verbose=2 # sets verbose = 2 +--buildid-dir:: + Setup buildid cache directory. It has higher priority than + buildid.dir config file option. + DESCRIPTION ----------- Performance counters for Linux are a new kernel-based subsystem diff --git a/tools/perf/perf.c b/tools/perf/perf.c index 452a8474d29d..3700a7faca6c 100644 --- a/tools/perf/perf.c +++ b/tools/perf/perf.c @@ -200,6 +200,16 @@ static int handle_options(const char ***argv, int *argc, int *envchanged) *envchanged = 1; (*argv)++; (*argc)--; + } else if (!strcmp(cmd, "--buildid-dir")) { + if (*argc < 2) { + fprintf(stderr, "No directory given for --buildid-dir.\n"); + usage(perf_usage_string); + } + set_buildid_dir((*argv)[1]); + if (envchanged) + *envchanged = 1; + (*argv)++; + (*argc)--; } else if (!prefixcmp(cmd, CMD_DEBUGFS_DIR)) { perf_debugfs_set_path(cmd + strlen(CMD_DEBUGFS_DIR)); fprintf(stderr, "dir: %s\n", debugfs_mountpoint); @@ -499,7 +509,7 @@ int main(int argc, const char **argv) } if (!prefixcmp(cmd, "trace")) { #ifdef HAVE_LIBAUDIT_SUPPORT - set_buildid_dir(); + set_buildid_dir(NULL); setup_path(); argv[0] = "trace"; return cmd_trace(argc, argv, NULL); @@ -514,7 +524,7 @@ int main(int argc, const char **argv) argc--; handle_options(&argv, &argc, NULL); commit_pager_choice(); - set_buildid_dir(); + set_buildid_dir(NULL); if (argc > 0) { if (!prefixcmp(argv[0], "--")) diff --git a/tools/perf/util/config.c b/tools/perf/util/config.c index c802236b2bfe..e18f653cd7db 100644 --- a/tools/perf/util/config.c +++ b/tools/perf/util/config.c @@ -539,12 +539,14 @@ static void check_buildid_dir_config(void) perf_config(buildid_dir_command_config, &c); } -void set_buildid_dir(void) +void set_buildid_dir(const char *dir) { - buildid_dir[0] = '\0'; + if (dir) + scnprintf(buildid_dir, MAXPATHLEN-1, "%s", dir); /* try config file */ - check_buildid_dir_config(); + if (buildid_dir[0] == '\0') + check_buildid_dir_config(); /* default to $HOME/.debug */ if (buildid_dir[0] == '\0') { diff --git a/tools/perf/util/util.h b/tools/perf/util/util.h index 419bee030f83..abc445ee4f60 100644 --- a/tools/perf/util/util.h +++ b/tools/perf/util/util.h @@ -153,7 +153,7 @@ extern void warning(const char *err, ...) __attribute__((format (printf, 1, 2))) extern void set_die_routine(void (*routine)(const char *err, va_list params) NORETURN); extern int prefixcmp(const char *str, const char *prefix); -extern void set_buildid_dir(void); +extern void set_buildid_dir(const char *dir); static inline const char *skip_prefix(const char *str, const char *prefix) { -- cgit v1.2.3 From f70b4e39de4ef25aade966c0dfc69cfb97091be9 Mon Sep 17 00:00:00 2001 From: Kan Liang Date: Tue, 2 Dec 2014 10:06:52 -0500 Subject: perf callchain: Fixup parameter handling error message Fix up parse_callchain_record_opt error message for 'fp', in the past using '-g fp' was a valid alternative to '--call-graph fp', which is not the case since: commit 09b0fd45ff63413df94cbd832a765076b201edbb Author: Jiri Olsa Date: Sat Oct 26 16:25:33 2013 +0200 perf record: Split -g and --call-graph I.e. -g means "use the configured unwind data collection method" which has as default 'fp', while --call-graph requires passing the method to use. Signed-off-by: Kan Liang Acked-by: Jiri Olsa Cc: Andi Kleen Cc: Ingo Molnar Cc: Jiri Olsa Cc: Namhyung Kim Cc: Paul Mackerras Cc: Peter Zijlstra Cc: Stephane Eranian Link: http://lkml.kernel.org/r/1417532814-26208-2-git-send-email-kan.liang@intel.com [ split this from a larger patch related to LBR based unwinding ] Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/util/callchain.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'tools/perf') diff --git a/tools/perf/util/callchain.c b/tools/perf/util/callchain.c index cf524a35cc84..64b377e591e4 100644 --- a/tools/perf/util/callchain.c +++ b/tools/perf/util/callchain.c @@ -77,7 +77,7 @@ int parse_callchain_record_opt(const char *arg) ret = 0; } else pr_err("callchain: No more arguments " - "needed for -g fp\n"); + "needed for --call-graph fp\n"); break; #ifdef HAVE_DWARF_UNWIND_SUPPORT -- cgit v1.2.3 From 2e77784bb7d882647c33d8e75a650625e6df0f8b Mon Sep 17 00:00:00 2001 From: Kan Liang Date: Tue, 2 Dec 2014 10:06:53 -0500 Subject: perf callchain: Move cpumode resolve code to add_callchain_ip Using flag to distinguish between branch_history and normal callchain. Move the cpumode to add_callchain_ip function. No change in behavior. Signed-off-by: Kan Liang Acked-by: Jiri Olsa Cc: Andi Kleen Cc: Ingo Molnar Cc: Jiri Olsa Cc: Namhyung Kim Cc: Paul Mackerras Cc: Peter Zijlstra Cc: Stephane Eranian Link: http://lkml.kernel.org/r/1417532814-26208-3-git-send-email-kan.liang@intel.com Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/util/machine.c | 72 +++++++++++++++++++++++------------------------ 1 file changed, 35 insertions(+), 37 deletions(-) (limited to 'tools/perf') diff --git a/tools/perf/util/machine.c b/tools/perf/util/machine.c index 15dd0a9691ce..94de3e48b490 100644 --- a/tools/perf/util/machine.c +++ b/tools/perf/util/machine.c @@ -1385,19 +1385,46 @@ struct mem_info *sample__resolve_mem(struct perf_sample *sample, static int add_callchain_ip(struct thread *thread, struct symbol **parent, struct addr_location *root_al, - int cpumode, + bool branch_history, u64 ip) { struct addr_location al; al.filtered = 0; al.sym = NULL; - if (cpumode == -1) + if (branch_history) thread__find_cpumode_addr_location(thread, MAP__FUNCTION, ip, &al); - else + else { + u8 cpumode = PERF_RECORD_MISC_USER; + + if (ip >= PERF_CONTEXT_MAX) { + switch (ip) { + case PERF_CONTEXT_HV: + cpumode = PERF_RECORD_MISC_HYPERVISOR; + break; + case PERF_CONTEXT_KERNEL: + cpumode = PERF_RECORD_MISC_KERNEL; + break; + case PERF_CONTEXT_USER: + cpumode = PERF_RECORD_MISC_USER; + break; + default: + pr_debug("invalid callchain context: " + "%"PRId64"\n", (s64) ip); + /* + * It seems the callchain is corrupted. + * Discard all. + */ + callchain_cursor_reset(&callchain_cursor); + return 1; + } + return 0; + } thread__find_addr_location(thread, cpumode, MAP__FUNCTION, ip, &al); + } + if (al.sym != NULL) { if (sort__has_parent && !*parent && symbol__match_regex(al.sym, &parent_regex)) @@ -1480,11 +1507,8 @@ static int thread__resolve_callchain_sample(struct thread *thread, struct addr_location *root_al, int max_stack) { - u8 cpumode = PERF_RECORD_MISC_USER; int chain_nr = min(max_stack, (int)chain->nr); - int i; - int j; - int err; + int i, j, err; int skip_idx = -1; int first_call = 0; @@ -1542,10 +1566,10 @@ static int thread__resolve_callchain_sample(struct thread *thread, for (i = 0; i < nr; i++) { err = add_callchain_ip(thread, parent, root_al, - -1, be[i].to); + true, be[i].to); if (!err) err = add_callchain_ip(thread, parent, root_al, - -1, be[i].from); + true, be[i].from); if (err == -EINVAL) break; if (err) @@ -1574,36 +1598,10 @@ check_calls: #endif ip = chain->ips[j]; - if (ip >= PERF_CONTEXT_MAX) { - switch (ip) { - case PERF_CONTEXT_HV: - cpumode = PERF_RECORD_MISC_HYPERVISOR; - break; - case PERF_CONTEXT_KERNEL: - cpumode = PERF_RECORD_MISC_KERNEL; - break; - case PERF_CONTEXT_USER: - cpumode = PERF_RECORD_MISC_USER; - break; - default: - pr_debug("invalid callchain context: " - "%"PRId64"\n", (s64) ip); - /* - * It seems the callchain is corrupted. - * Discard all. - */ - callchain_cursor_reset(&callchain_cursor); - return 0; - } - continue; - } + err = add_callchain_ip(thread, parent, root_al, false, ip); - err = add_callchain_ip(thread, parent, root_al, - cpumode, ip); - if (err == -EINVAL) - break; if (err) - return err; + return (err < 0) ? err : 0; } return 0; -- cgit v1.2.3 From e4e458b45c5861808674eebfea94cee2258bb2ea Mon Sep 17 00:00:00 2001 From: Arjun Sreedharan Date: Sat, 6 Dec 2014 17:10:43 +0530 Subject: calloc/xcalloc: Fix argument order The calloc() and xcalloc() functions takes @nmemb first and then @size. Fix all w/ pattern "calloc\s*(\s*sizeof". Signed-off-by: Arjun Sreedharan Cc: "Yann E. MORIN" Cc: Ingo Molnar Cc: Paul Mackerras Cc: Peter Zijlstra Link: http://lkml.kernel.org/r/1417866043-1877-1-git-send-email-arjun024@gmail.com Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/ui/hist.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'tools/perf') diff --git a/tools/perf/ui/hist.c b/tools/perf/ui/hist.c index 2af18376b077..dc0d095f318c 100644 --- a/tools/perf/ui/hist.c +++ b/tools/perf/ui/hist.c @@ -162,8 +162,8 @@ static int __hpp__sort(struct hist_entry *a, struct hist_entry *b, return ret; nr_members = evsel->nr_members; - fields_a = calloc(sizeof(*fields_a), nr_members); - fields_b = calloc(sizeof(*fields_b), nr_members); + fields_a = calloc(nr_members, sizeof(*fields_a)); + fields_b = calloc(nr_members, sizeof(*fields_b)); if (!fields_a || !fields_b) goto out; -- cgit v1.2.3 From 75226c577c8869ae1449cf92d781edda0177f1cf Mon Sep 17 00:00:00 2001 From: Jiri Olsa Date: Tue, 9 Dec 2014 14:53:01 +0100 Subject: perf tests: Fix attr tests size values to cope with machine state on interrupt ABI changes Following change adjusted 'struct perf_event_attr', but let the attr test's sizes untouched: 60e2364e60e8 perf: Add ability to sample machine state on interrupt [jolsa@krava perf]$ ./perf test attr -vv --- start --- test child forked, pid 9719 running './tests/attr/test-stat-group1' 'PERF_TEST_ATTR=/tmp/tmp4drvul ./perf stat -o /tmp/tmp4drvul/perf.data -e '{cycles,instructions}' kill >/dev/null 2>&1' ret 1 expected size=96, got 104 FAILED './tests/attr/test-stat-group1' - match failure Adjusting test size values for attr test. Reported-by: Arnaldo Carvalho de Melo Signed-off-by: Jiri Olsa Cc: Andi Kleen Cc: Ingo Molnar Cc: Peter Zijlstra Cc: Stephane Eranian Link: http://lkml.kernel.org/r/20141209135301.GC6784@krava.brq.redhat.com Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/tests/attr/base-record | 2 +- tools/perf/tests/attr/base-stat | 2 +- 2 files changed, 2 insertions(+), 2 deletions(-) (limited to 'tools/perf') diff --git a/tools/perf/tests/attr/base-record b/tools/perf/tests/attr/base-record index f710b92ccff6..d3095dafed36 100644 --- a/tools/perf/tests/attr/base-record +++ b/tools/perf/tests/attr/base-record @@ -5,7 +5,7 @@ group_fd=-1 flags=0|8 cpu=* type=0|1 -size=96 +size=104 config=0 sample_period=4000 sample_type=263 diff --git a/tools/perf/tests/attr/base-stat b/tools/perf/tests/attr/base-stat index dc3ada2470c0..872ed7e24c7c 100644 --- a/tools/perf/tests/attr/base-stat +++ b/tools/perf/tests/attr/base-stat @@ -5,7 +5,7 @@ group_fd=-1 flags=0|8 cpu=* type=0 -size=96 +size=104 config=0 sample_period=0 sample_type=0 -- cgit v1.2.3 From 99d348a84c2118ed04c9b72168787f55e2fe33a5 Mon Sep 17 00:00:00 2001 From: Alexander Yarygin Date: Wed, 10 Dec 2014 14:28:10 +0100 Subject: perf kvm stat live: Mark events as (x86 only) in help output The mmio and ioport events are useful only on x86. Signed-off-by: Alexander Yarygin Acked-by: David Ahern Cc: Christian Borntraeger Cc: Jiri Olsa Link: http://lkml.kernel.org/r/1418218090-20718-1-git-send-email-borntraeger@de.ibm.com Signed-off-by: Christian Borntraeger Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/builtin-kvm.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) (limited to 'tools/perf') diff --git a/tools/perf/builtin-kvm.c b/tools/perf/builtin-kvm.c index 3c0f3d4fb021..0894a817f67e 100644 --- a/tools/perf/builtin-kvm.c +++ b/tools/perf/builtin-kvm.c @@ -1293,7 +1293,8 @@ static int kvm_events_live(struct perf_kvm_stat *kvm, OPT_UINTEGER('d', "display", &kvm->display_time, "time in seconds between display updates"), OPT_STRING(0, "event", &kvm->report_event, "report event", - "event for reporting: vmexit, mmio, ioport"), + "event for reporting: " + "vmexit, mmio (x86 only), ioport (x86 only)"), OPT_INTEGER(0, "vcpu", &kvm->trace_vcpu, "vcpu id to report"), OPT_STRING('k', "key", &kvm->sort_key, "sort-key", -- cgit v1.2.3 From 3a351127cbc682c3a0ae7383c4ff6fd0e8fee83c Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Thu, 11 Dec 2014 13:17:46 -0300 Subject: tools lib fs: Adopt filename__read_int from tools/perf/ Will be useful for new helpers to read sysctl values. Cc: Adrian Hunter Cc: Borislav Petkov Cc: David Ahern Cc: Don Zickus Cc: Frederic Weisbecker Cc: Jiri Olsa Cc: Mike Galbraith Cc: Namhyung Kim Cc: Paul Mackerras Cc: Peter Zijlstra Cc: Stephane Eranian Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/util/util.c | 17 ----------------- tools/perf/util/util.h | 1 - 2 files changed, 18 deletions(-) (limited to 'tools/perf') diff --git a/tools/perf/util/util.c b/tools/perf/util/util.c index d5eab3f3323f..41125e5b011e 100644 --- a/tools/perf/util/util.c +++ b/tools/perf/util/util.c @@ -442,23 +442,6 @@ unsigned long parse_tag_value(const char *str, struct parse_tag *tags) return (unsigned long) -1; } -int filename__read_int(const char *filename, int *value) -{ - char line[64]; - int fd = open(filename, O_RDONLY), err = -1; - - if (fd < 0) - return -1; - - if (read(fd, line, sizeof(line)) > 0) { - *value = atoi(line); - err = 0; - } - - close(fd); - return err; -} - int filename__read_str(const char *filename, char **buf, size_t *sizep) { size_t size = 0, alloc_size = 0; diff --git a/tools/perf/util/util.h b/tools/perf/util/util.h index abc445ee4f60..008b361b1758 100644 --- a/tools/perf/util/util.h +++ b/tools/perf/util/util.h @@ -343,7 +343,6 @@ char *get_srcline(struct dso *dso, unsigned long addr, struct symbol *sym, bool show_sym); void free_srcline(char *srcline); -int filename__read_int(const char *filename, int *value); int filename__read_str(const char *filename, char **buf, size_t *sizep); int perf_event_paranoid(void); -- cgit v1.2.3 From ce27309f67c347e6e49ebd45042f5362b0ccbdcc Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Thu, 11 Dec 2014 13:37:59 -0300 Subject: perf tools: Use sysctl__read_int instead of ad-hoc copies Cc: Adrian Hunter Cc: Borislav Petkov Cc: David Ahern Cc: Don Zickus Cc: Frederic Weisbecker Cc: Jiri Olsa Cc: Mike Galbraith Cc: Namhyung Kim Cc: Paul Mackerras Cc: Peter Zijlstra Cc: Stephane Eranian Link: http://lkml.kernel.org/n/tip-pt2u7a3b50oddggecx7rwq2n@git.kernel.org Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/util/record.c | 11 +---------- tools/perf/util/util.c | 9 +-------- 2 files changed, 2 insertions(+), 18 deletions(-) (limited to 'tools/perf') diff --git a/tools/perf/util/record.c b/tools/perf/util/record.c index cf69325b985f..8acd0df88b5c 100644 --- a/tools/perf/util/record.c +++ b/tools/perf/util/record.c @@ -137,16 +137,7 @@ void perf_evlist__config(struct perf_evlist *evlist, struct record_opts *opts) static int get_max_rate(unsigned int *rate) { - char path[PATH_MAX]; - const char *procfs = procfs__mountpoint(); - - if (!procfs) - return -1; - - snprintf(path, PATH_MAX, - "%s/sys/kernel/perf_event_max_sample_rate", procfs); - - return filename__read_int(path, (int *) rate); + return sysctl__read_int("kernel/perf_event_max_sample_rate", (int *)rate); } static int record_opts__config_freq(struct record_opts *opts) diff --git a/tools/perf/util/util.c b/tools/perf/util/util.c index 41125e5b011e..b86744f29eef 100644 --- a/tools/perf/util/util.c +++ b/tools/perf/util/util.c @@ -506,16 +506,9 @@ const char *get_filename_for_perf_kvm(void) int perf_event_paranoid(void) { - char path[PATH_MAX]; - const char *procfs = procfs__mountpoint(); int value; - if (!procfs) - return INT_MAX; - - scnprintf(path, PATH_MAX, "%s/sys/kernel/perf_event_paranoid", procfs); - - if (filename__read_int(path, &value)) + if (sysctl__read_int("kernel/perf_event_paranoid", &value)) return INT_MAX; return value; -- cgit v1.2.3 From 956fa57106d1ad3ffbe73f4534b9f3b3bc92dd6c Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Thu, 11 Dec 2014 18:03:01 -0300 Subject: perf evlist: Introduce strerror_mmap method To pretty print hints about perf_evlist__mmap errors. Will be used in 'trace' in the next patch. Cc: Adrian Hunter Cc: Borislav Petkov Cc: David Ahern Cc: Don Zickus Cc: Frederic Weisbecker Cc: Jiri Olsa Cc: Mike Galbraith Cc: Namhyung Kim Cc: Paul Mackerras Cc: Peter Zijlstra Cc: Stephane Eranian Link: http://lkml.kernel.org/n/tip-2g3gczfwyz0xt3we0s15mqqt@git.kernel.org Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/util/evlist.c | 23 +++++++++++++++++++++++ tools/perf/util/evlist.h | 1 + 2 files changed, 24 insertions(+) (limited to 'tools/perf') diff --git a/tools/perf/util/evlist.c b/tools/perf/util/evlist.c index cfbe2b99b9aa..bb5dfc5d1e75 100644 --- a/tools/perf/util/evlist.c +++ b/tools/perf/util/evlist.c @@ -8,6 +8,7 @@ */ #include "util.h" #include +#include #include #include "cpumap.h" #include "thread_map.h" @@ -1483,6 +1484,28 @@ int perf_evlist__strerror_open(struct perf_evlist *evlist __maybe_unused, return 0; } +int perf_evlist__strerror_mmap(struct perf_evlist *evlist, int err, char *buf, size_t size) +{ + char sbuf[STRERR_BUFSIZE], *emsg = strerror_r(err, sbuf, sizeof(sbuf)); + int value; + + switch (err) { + case EPERM: + sysctl__read_int("kernel/perf_event_mlock_kb", &value); + scnprintf(buf, size, "Error:\t%s.\n" + "Hint:\tCheck /proc/sys/kernel/perf_event_mlock_kb (%d kB) setting.\n" + "Hint:\tTried using %zd kB.\n" + "Hint:\tTry using a bigger -m/--mmap-pages value.", + emsg, value, evlist->mmap_len / 1024); + break; + default: + scnprintf(buf, size, "%s", emsg); + break; + } + + return 0; +} + void perf_evlist__to_front(struct perf_evlist *evlist, struct perf_evsel *move_evsel) { diff --git a/tools/perf/util/evlist.h b/tools/perf/util/evlist.h index 649b0c597283..0ba93f67ab94 100644 --- a/tools/perf/util/evlist.h +++ b/tools/perf/util/evlist.h @@ -185,6 +185,7 @@ size_t perf_evlist__fprintf(struct perf_evlist *evlist, FILE *fp); int perf_evlist__strerror_tp(struct perf_evlist *evlist, int err, char *buf, size_t size); int perf_evlist__strerror_open(struct perf_evlist *evlist, int err, char *buf, size_t size); +int perf_evlist__strerror_mmap(struct perf_evlist *evlist, int err, char *buf, size_t size); static inline unsigned int perf_mmap__read_head(struct perf_mmap *mm) { -- cgit v1.2.3 From e09b18d4907992d3d615b215c1abf585721b2810 Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Thu, 11 Dec 2014 18:04:10 -0300 Subject: perf trace: Provide a better explanation when mmap fails If we ask for a mmap lenght than the max configured via the relevant sysctl, provide a better warning, instead of just expanding the EPERM returned: [acme@ssdandy ~]$ trace -m 256 -e nanosleep sleep 2 Error: Operation not permitted. Hint: Check /proc/sys/kernel/perf_event_mlock_kb (516 kB) setting. Hint: Tried using 1028 kB. Hint: Try using a bigger -m/--mmap-pages value. [acme@ssdandy ~]$ trace -m 128 -e nanosleep sleep 2 2001.280 (2000.403 ms): nanosleep(rqtp: 0x7fff89a8a7f0) = 0 [acme@ssdandy ~]$ An upcoming patch will autotune the request for non-root users when -m is not used. Cc: Adrian Hunter Cc: Borislav Petkov Cc: David Ahern Cc: Don Zickus Cc: Frederic Weisbecker Cc: Jiri Olsa Cc: Mike Galbraith Cc: Namhyung Kim Cc: Paul Mackerras Cc: Peter Zijlstra Cc: Stephane Eranian Link: http://lkml.kernel.org/n/tip-cdvxfz2gycetbkopm9sna1qp@git.kernel.org Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/builtin-trace.c | 12 ++++++------ 1 file changed, 6 insertions(+), 6 deletions(-) (limited to 'tools/perf') diff --git a/tools/perf/builtin-trace.c b/tools/perf/builtin-trace.c index 83a4835c8118..327541e43c7f 100644 --- a/tools/perf/builtin-trace.c +++ b/tools/perf/builtin-trace.c @@ -2045,7 +2045,6 @@ static int trace__run(struct trace *trace, int argc, const char **argv) unsigned long before; const bool forks = argc > 0; bool draining = false; - char sbuf[STRERR_BUFSIZE]; trace->live = true; @@ -2106,11 +2105,8 @@ static int trace__run(struct trace *trace, int argc, const char **argv) goto out_error_open; err = perf_evlist__mmap(evlist, trace->opts.mmap_pages, false); - if (err < 0) { - fprintf(trace->output, "Couldn't mmap the events: %s\n", - strerror_r(errno, sbuf, sizeof(sbuf))); - goto out_delete_evlist; - } + if (err < 0) + goto out_error_mmap; perf_evlist__enable(evlist); @@ -2210,6 +2206,10 @@ out_error_tp: perf_evlist__strerror_tp(evlist, errno, errbuf, sizeof(errbuf)); goto out_error; +out_error_mmap: + perf_evlist__strerror_mmap(evlist, errno, errbuf, sizeof(errbuf)); + goto out_error; + out_error_open: perf_evlist__strerror_open(evlist, errno, errbuf, sizeof(errbuf)); -- cgit v1.2.3