From 67540759151aefafddade3e27c4671ab7b3d230f Mon Sep 17 00:00:00 2001 From: Milian Wolff Date: Tue, 16 Aug 2016 17:39:26 +0200 Subject: [PATCH 01/38] perf unwind: Use addr_location::addr instead of ip for entries This fixes the srcline translation for call chains of user space applications. Before we got: perf report --stdio --no-children -s sym,srcline -g address 8.92% [.] main mandelbrot.h:41 | |--3.70%--main +8390240 | __libc_start_main +139950056726769 | _start +8388650 | |--2.74%--main +8390189 | --2.08%--main +8390296 __libc_start_main +139950056726769 _start +8388650 7.59% [.] main complex:1326 | |--4.79%--main +8390203 | __libc_start_main +139950056726769 | _start +8388650 | --2.80%--main +8390219 7.12% [.] __muldc3 libgcc2.c:1945 | |--3.76%--__muldc3 +139950060519490 | main +8390224 | __libc_start_main +139950056726769 | _start +8388650 | --3.32%--__muldc3 +139950060519512 main +8390224 With this patch applied, we instead get: perf report --stdio --no-children -s sym,srcline -g address 8.92% [.] main mandelbrot.h:41 | |--3.70%--main mandelbrot.h:41 | __libc_start_main +241 | _start +4194346 | |--2.74%--main mandelbrot.h:41 | --2.08%--main mandelbrot.h:41 __libc_start_main +241 _start +4194346 7.59% [.] main complex:1326 | |--4.79%--main complex:1326 | __libc_start_main +241 | _start +4194346 | --2.80%--main complex:1326 7.12% [.] __muldc3 libgcc2.c:1945 | |--3.76%--__muldc3 libgcc2.c:1945 | main mandelbrot.h:39 | __libc_start_main +241 | _start +4194346 | --3.32%--__muldc3 libgcc2.c:1945 main mandelbrot.h:39 Suggested-and-Acked-by: Namhyung Kim Signed-off-by: Milian Wolff Tested-by: Arnaldo Carvalho de Melo LPU-Reference: 20160816153926.11288-1-milian.wolff@kdab.com Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/util/unwind-libdw.c | 2 +- tools/perf/util/unwind-libunwind-local.c | 2 +- 2 files changed, 2 insertions(+), 2 deletions(-) diff --git a/tools/perf/util/unwind-libdw.c b/tools/perf/util/unwind-libdw.c index cf5e250bc78e..783a53fb7a4e 100644 --- a/tools/perf/util/unwind-libdw.c +++ b/tools/perf/util/unwind-libdw.c @@ -66,7 +66,7 @@ static int entry(u64 ip, struct unwind_info *ui) if (__report_module(&al, ip, ui)) return -1; - e->ip = ip; + e->ip = al.addr; e->map = al.map; e->sym = al.sym; diff --git a/tools/perf/util/unwind-libunwind-local.c b/tools/perf/util/unwind-libunwind-local.c index 97c0f8fc5561..20c2e5743903 100644 --- a/tools/perf/util/unwind-libunwind-local.c +++ b/tools/perf/util/unwind-libunwind-local.c @@ -542,7 +542,7 @@ static int entry(u64 ip, struct thread *thread, thread__find_addr_location(thread, PERF_RECORD_MISC_USER, MAP__FUNCTION, ip, &al); - e.ip = ip; + e.ip = al.addr; e.map = al.map; e.sym = al.sym; From 0215d59b154ab90c56c4fe49bc1deefe8bca18f1 Mon Sep 17 00:00:00 2001 From: Vineet Gupta Date: Thu, 18 Aug 2016 09:28:23 -0700 Subject: [PATCH 02/38] tools lib: Reinstate strlcpy() header guard with __UCLIBC__ MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit perf tools build in recent kernels spews splat when cross compiling with uClibc: | CC util/alias.o | In file included from tools/perf/util/../ui/../util/cache.h:8:0, | from tools/perf/util/../ui/helpline.h:7, | from tools/perf/util/debug.h:8, | from arch/../util/cpumap.h:9, | from arch/../util/env.h:5, | from arch/common.h:4, | from arch/common.c:3: | tools/include/linux/string.h:12:15: warning: redundant redeclaration of ‘strlcpy’ [-Wredundant-decls] | extern size_t strlcpy(char *dest, const char *src, size_t size); ^ This is after commit 61a6445e463a31 ("tools lib: Guard the strlcpy() header with __GLIBC__"). The problem is uClibc also defines __GLIBC__ for exported headers for applications. So add that specific check to not trip for uClibc. Signed-off-by: Vineet Gupta Cc: Adrian Hunter Cc: Alexey Brodkin Cc: David Ahern Cc: Jiri Olsa Cc: Josh Poimboeuf Cc: Namhyung Kim Cc: Peter Zijlstra Cc: Petri Gynther Cc: Wang Nan Cc: linux-snps-arc@lists.infradead.org Link: http://lkml.kernel.org/r/1471537703-16439-1-git-send-email-vgupta@synopsys.com Signed-off-by: Arnaldo Carvalho de Melo --- tools/include/linux/string.h | 6 +++++- 1 file changed, 5 insertions(+), 1 deletion(-) diff --git a/tools/include/linux/string.h b/tools/include/linux/string.h index b96879477311..f436d2420a18 100644 --- a/tools/include/linux/string.h +++ b/tools/include/linux/string.h @@ -8,7 +8,11 @@ void *memdup(const void *src, size_t len); int strtobool(const char *s, bool *res); -#ifdef __GLIBC__ +/* + * glibc based builds needs the extern while uClibc doesn't. + * However uClibc headers also define __GLIBC__ hence the hack below + */ +#if defined(__GLIBC__) && !defined(__UCLIBC__) extern size_t strlcpy(char *dest, const char *src, size_t size); #endif From c53412ee8c7ec31373a4176ff7f3a6b79296c05c Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Thu, 18 Aug 2016 16:30:28 -0300 Subject: [PATCH 03/38] perf evsel: Do not access outside hw cache name arrays We have to check if the values are >= *_MAX, not just >, fix it. From the bugzilla report: ''In file /tools/perf/util/evsel.c function __perf_evsel__hw_cache_name it appears that there is a bug that reads beyond the end of the buffer. The statement "if (type > PERF_COUNT_HW_CACHE_MAX)" allows type to be equal to the maximum value. Later, when statement "if (!perf_evsel__is_cache_op_valid(type, op))" is executed, the function can access array perf_evsel__hw_cache_stat[type] beyond the end of the buffer. It appears to me that the statement "if (type > PERF_COUNT_HW_CACHE_MAX)" should be "if (type >= PERF_COUNT_HW_CACHE_MAX)" Bug found with Coverity and manual code review. No attempts were made to execute the code with a maximum type value.'' Committer note: Testing it: $ perf record -e $(echo $(perf list cache | cut -d \[ -f1) | sed 's/ /,/g') usleep 1 [ perf record: Woken up 16 times to write data ] [ perf record: Captured and wrote 0.023 MB perf.data (34 samples) ] $ perf evlist L1-dcache-load-misses L1-dcache-loads L1-dcache-stores L1-icache-load-misses LLC-load-misses LLC-loads LLC-store-misses LLC-stores branch-load-misses branch-loads dTLB-load-misses dTLB-loads dTLB-store-misses dTLB-stores iTLB-load-misses iTLB-loads node-load-misses node-loads node-store-misses node-stores $ perf list cache List of pre-defined events (to be used in -e): L1-dcache-load-misses [Hardware cache event] L1-dcache-loads [Hardware cache event] L1-dcache-stores [Hardware cache event] L1-icache-load-misses [Hardware cache event] LLC-load-misses [Hardware cache event] LLC-loads [Hardware cache event] LLC-store-misses [Hardware cache event] LLC-stores [Hardware cache event] branch-load-misses [Hardware cache event] branch-loads [Hardware cache event] dTLB-load-misses [Hardware cache event] dTLB-loads [Hardware cache event] dTLB-store-misses [Hardware cache event] dTLB-stores [Hardware cache event] iTLB-load-misses [Hardware cache event] iTLB-loads [Hardware cache event] node-load-misses [Hardware cache event] node-loads [Hardware cache event] node-store-misses [Hardware cache event] node-stores [Hardware cache event] $ Reported-by: Brian Sweeney Tested-by: Arnaldo Carvalho de Melo Cc: Jiri Olsa Bugzilla: https://bugzilla.kernel.org/show_bug.cgi?id=153351 Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/util/evsel.c | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) diff --git a/tools/perf/util/evsel.c b/tools/perf/util/evsel.c index d9b80ef881cd..21fd573106ed 100644 --- a/tools/perf/util/evsel.c +++ b/tools/perf/util/evsel.c @@ -507,17 +507,17 @@ static int __perf_evsel__hw_cache_name(u64 config, char *bf, size_t size) u8 op, result, type = (config >> 0) & 0xff; const char *err = "unknown-ext-hardware-cache-type"; - if (type > PERF_COUNT_HW_CACHE_MAX) + if (type >= PERF_COUNT_HW_CACHE_MAX) goto out_err; op = (config >> 8) & 0xff; err = "unknown-ext-hardware-cache-op"; - if (op > PERF_COUNT_HW_CACHE_OP_MAX) + if (op >= PERF_COUNT_HW_CACHE_OP_MAX) goto out_err; result = (config >> 16) & 0xff; err = "unknown-ext-hardware-cache-result"; - if (result > PERF_COUNT_HW_CACHE_RESULT_MAX) + if (result >= PERF_COUNT_HW_CACHE_RESULT_MAX) goto out_err; err = "invalid-cache"; From bd48c63eb0afc28b29fb342f215cdd77b995c02e Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Fri, 5 Aug 2016 15:40:30 -0300 Subject: [PATCH 04/38] tools: Introduce tools/include/linux/time64.h for *SEC_PER_*SEC macros And remove it from tools/perf/{perf,util}.h, making code that needs these macros to include linux/time64.h instead, to match how this is used in the kernel sources. Cc: Adrian Hunter Cc: David Ahern Cc: Jiri Olsa Cc: Namhyung Kim Cc: Steven Rostedt Cc: Wang Nan Link: http://lkml.kernel.org/n/tip-e69fc1pvkgt57yvxqt6eunyg@git.kernel.org Signed-off-by: Arnaldo Carvalho de Melo --- tools/include/linux/time64.h | 12 ++++++++++++ tools/perf/MANIFEST | 1 + tools/perf/builtin-kvm.c | 1 + tools/perf/builtin-script.c | 7 ++++--- tools/perf/builtin-stat.c | 7 ++++--- tools/perf/builtin-trace.c | 1 + tools/perf/perf.h | 7 ------- tools/perf/util/debug.c | 10 ++++------ tools/perf/util/scripting-engines/trace-event-perl.c | 5 +++-- .../perf/util/scripting-engines/trace-event-python.c | 5 +++-- tools/perf/util/util.c | 1 + tools/perf/util/util.h | 4 ---- 12 files changed, 34 insertions(+), 27 deletions(-) create mode 100644 tools/include/linux/time64.h diff --git a/tools/include/linux/time64.h b/tools/include/linux/time64.h new file mode 100644 index 000000000000..df9265483d65 --- /dev/null +++ b/tools/include/linux/time64.h @@ -0,0 +1,12 @@ +#ifndef _TOOLS_LINUX_TIME64_H +#define _TOOLS_LINUX_TIME64_H + +#define MSEC_PER_SEC 1000L +#define USEC_PER_MSEC 1000L +#define NSEC_PER_USEC 1000L +#define NSEC_PER_MSEC 1000000L +#define USEC_PER_SEC 1000000L +#define NSEC_PER_SEC 1000000000L +#define FSEC_PER_SEC 1000000000000000LL + +#endif /* _LINUX_TIME64_H */ diff --git a/tools/perf/MANIFEST b/tools/perf/MANIFEST index ad2534df4ba6..f23a5e72aaa5 100644 --- a/tools/perf/MANIFEST +++ b/tools/perf/MANIFEST @@ -77,4 +77,5 @@ tools/include/linux/stringify.h tools/include/linux/types.h tools/include/linux/err.h tools/include/linux/bitmap.h +tools/include/linux/time64.h tools/arch/*/include/uapi/asm/perf_regs.h diff --git a/tools/perf/builtin-kvm.c b/tools/perf/builtin-kvm.c index 5e2127e04f83..a0040f74e935 100644 --- a/tools/perf/builtin-kvm.c +++ b/tools/perf/builtin-kvm.c @@ -24,6 +24,7 @@ #include #endif +#include #include #include #include diff --git a/tools/perf/builtin-script.c b/tools/perf/builtin-script.c index c859e59dfe3e..6b3c8b0d3276 100644 --- a/tools/perf/builtin-script.c +++ b/tools/perf/builtin-script.c @@ -24,6 +24,7 @@ #include "util/thread-stack.h" #include #include +#include #include "asm/bug.h" #include "util/mem-events.h" @@ -464,9 +465,9 @@ static void print_sample_start(struct perf_sample *sample, if (PRINT_FIELD(TIME)) { nsecs = sample->time; - secs = nsecs / NSECS_PER_SEC; - nsecs -= secs * NSECS_PER_SEC; - usecs = nsecs / NSECS_PER_USEC; + secs = nsecs / NSEC_PER_SEC; + nsecs -= secs * NSEC_PER_SEC; + usecs = nsecs / NSEC_PER_USEC; if (nanosecs) printf("%5lu.%09llu: ", secs, nsecs); else diff --git a/tools/perf/builtin-stat.c b/tools/perf/builtin-stat.c index 3c7452b39f57..e33a66bffbdb 100644 --- a/tools/perf/builtin-stat.c +++ b/tools/perf/builtin-stat.c @@ -65,6 +65,7 @@ #include "util/group.h" #include "asm/bug.h" +#include #include #include #include @@ -354,7 +355,7 @@ static void process_interval(void) diff_timespec(&rs, &ts, &ref_time); if (STAT_RECORD) { - if (WRITE_STAT_ROUND_EVENT(rs.tv_sec * NSECS_PER_SEC + rs.tv_nsec, INTERVAL)) + if (WRITE_STAT_ROUND_EVENT(rs.tv_sec * NSEC_PER_SEC + rs.tv_nsec, INTERVAL)) pr_err("failed to write stat round event\n"); } @@ -2175,8 +2176,8 @@ static int process_stat_round_event(struct perf_tool *tool __maybe_unused, update_stats(&walltime_nsecs_stats, stat_round->time); if (stat_config.interval && stat_round->time) { - tsh.tv_sec = stat_round->time / NSECS_PER_SEC; - tsh.tv_nsec = stat_round->time % NSECS_PER_SEC; + tsh.tv_sec = stat_round->time / NSEC_PER_SEC; + tsh.tv_nsec = stat_round->time % NSEC_PER_SEC; ts = &tsh; } diff --git a/tools/perf/builtin-trace.c b/tools/perf/builtin-trace.c index b8c6766301db..b4fc1ab3d2a7 100644 --- a/tools/perf/builtin-trace.c +++ b/tools/perf/builtin-trace.c @@ -45,6 +45,7 @@ #include #include #include +#include #ifndef O_CLOEXEC # define O_CLOEXEC 02000000 diff --git a/tools/perf/perf.h b/tools/perf/perf.h index cb0f1356ff81..9a0236a4cf95 100644 --- a/tools/perf/perf.h +++ b/tools/perf/perf.h @@ -14,13 +14,6 @@ void test_attr__open(struct perf_event_attr *attr, pid_t pid, int cpu, #define HAVE_ATTR_TEST #include "perf-sys.h" -#ifndef NSEC_PER_SEC -# define NSEC_PER_SEC 1000000000ULL -#endif -#ifndef NSEC_PER_USEC -# define NSEC_PER_USEC 1000ULL -#endif - static inline unsigned long long rdclock(void) { struct timespec ts; diff --git a/tools/perf/util/debug.c b/tools/perf/util/debug.c index 8c4212abd19b..c1838b643108 100644 --- a/tools/perf/util/debug.c +++ b/tools/perf/util/debug.c @@ -6,6 +6,7 @@ #include #include #include +#include #include "cache.h" #include "color.h" @@ -14,9 +15,6 @@ #include "util.h" #include "target.h" -#define NSECS_PER_SEC 1000000000ULL -#define NSECS_PER_USEC 1000ULL - int verbose; bool dump_trace = false, quiet = false; int debug_ordered_events; @@ -54,9 +52,9 @@ static int veprintf_time(u64 t, const char *fmt, va_list args) int ret = 0; u64 secs, usecs, nsecs = t; - secs = nsecs / NSECS_PER_SEC; - nsecs -= secs * NSECS_PER_SEC; - usecs = nsecs / NSECS_PER_USEC; + secs = nsecs / NSEC_PER_SEC; + nsecs -= secs * NSEC_PER_SEC; + usecs = nsecs / NSEC_PER_USEC; ret = fprintf(stderr, "[%13" PRIu64 ".%06" PRIu64 "] ", secs, usecs); diff --git a/tools/perf/util/scripting-engines/trace-event-perl.c b/tools/perf/util/scripting-engines/trace-event-perl.c index 5d1eb1ccd96c..e55a132f69b7 100644 --- a/tools/perf/util/scripting-engines/trace-event-perl.c +++ b/tools/perf/util/scripting-engines/trace-event-perl.c @@ -25,6 +25,7 @@ #include #include #include +#include #include "../util.h" #include @@ -359,8 +360,8 @@ static void perl_process_tracepoint(struct perf_sample *sample, if (!test_and_set_bit(event->id, events_defined)) define_event_symbols(event, handler, event->print_fmt.args); - s = nsecs / NSECS_PER_SEC; - ns = nsecs - s * NSECS_PER_SEC; + s = nsecs / NSEC_PER_SEC; + ns = nsecs - s * NSEC_PER_SEC; scripting_context->event_data = data; scripting_context->pevent = evsel->tp_format->pevent; diff --git a/tools/perf/util/scripting-engines/trace-event-python.c b/tools/perf/util/scripting-engines/trace-event-python.c index e0203b979474..089438da1f7f 100644 --- a/tools/perf/util/scripting-engines/trace-event-python.c +++ b/tools/perf/util/scripting-engines/trace-event-python.c @@ -27,6 +27,7 @@ #include #include #include +#include #include "../../perf.h" #include "../debug.h" @@ -426,8 +427,8 @@ static void python_process_tracepoint(struct perf_sample *sample, if (!dict) Py_FatalError("couldn't create Python dict"); } - s = nsecs / NSECS_PER_SEC; - ns = nsecs - s * NSECS_PER_SEC; + s = nsecs / NSEC_PER_SEC; + ns = nsecs - s * NSEC_PER_SEC; scripting_context->event_data = data; scripting_context->pevent = evsel->tp_format->pevent; diff --git a/tools/perf/util/util.c b/tools/perf/util/util.c index cee559d8c9e8..85c56800f17a 100644 --- a/tools/perf/util/util.c +++ b/tools/perf/util/util.c @@ -15,6 +15,7 @@ #include #include #include +#include #include #include "callchain.h" #include "strlist.h" diff --git a/tools/perf/util/util.h b/tools/perf/util/util.h index e5f55477491d..43899e0d6fa1 100644 --- a/tools/perf/util/util.h +++ b/tools/perf/util/util.h @@ -179,10 +179,6 @@ static inline void *zalloc(size_t size) #undef tolower #undef toupper -#ifndef NSEC_PER_MSEC -#define NSEC_PER_MSEC 1000000L -#endif - int parse_nsec_time(const char *str, u64 *ptime); extern unsigned char sane_ctype[256]; From a8ad8329b91551bbfc3a317f3655dffcd8c18591 Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Mon, 8 Aug 2016 11:55:38 -0300 Subject: [PATCH 05/38] perf bench numa: Use NSEC_PER_U?SEC Following kernel practices, using linux/time64.h Cc: Adrian Hunter Cc: David Ahern Cc: Jiri Olsa Cc: Namhyung Kim Cc: Steven Rostedt Cc: Wang Nan Link: http://lkml.kernel.org/n/tip-7vnv15263y50qku76p4w5xk6@git.kernel.org Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/bench/numa.c | 53 +++++++++++++++++++++-------------------- 1 file changed, 27 insertions(+), 26 deletions(-) diff --git a/tools/perf/bench/numa.c b/tools/perf/bench/numa.c index f7f530081aa9..8efe904e486b 100644 --- a/tools/perf/bench/numa.c +++ b/tools/perf/bench/numa.c @@ -30,6 +30,7 @@ #include #include #include +#include #include #include @@ -1004,7 +1005,7 @@ static void calc_convergence(double runtime_ns_max, double *convergence) if (strong && process_groups == g->p.nr_proc) { if (!*convergence) { *convergence = runtime_ns_max; - tprintf(" (%6.1fs converged)\n", *convergence/1e9); + tprintf(" (%6.1fs converged)\n", *convergence / NSEC_PER_SEC); if (g->p.measure_convergence) { g->all_converged = true; g->stop_work = true; @@ -1012,7 +1013,7 @@ static void calc_convergence(double runtime_ns_max, double *convergence) } } else { if (*convergence) { - tprintf(" (%6.1fs de-converged)", runtime_ns_max/1e9); + tprintf(" (%6.1fs de-converged)", runtime_ns_max / NSEC_PER_SEC); *convergence = 0; } tprintf("\n"); @@ -1022,7 +1023,7 @@ static void calc_convergence(double runtime_ns_max, double *convergence) static void show_summary(double runtime_ns_max, int l, double *convergence) { tprintf("\r # %5.1f%% [%.1f mins]", - (double)(l+1)/g->p.nr_loops*100.0, runtime_ns_max/1e9 / 60.0); + (double)(l+1)/g->p.nr_loops*100.0, runtime_ns_max / NSEC_PER_SEC / 60.0); calc_convergence(runtime_ns_max, convergence); @@ -1179,8 +1180,8 @@ static void *worker_thread(void *__tdata) if (details >= 3) { timersub(&stop, &start, &diff); - runtime_ns_max = diff.tv_sec * 1000000000; - runtime_ns_max += diff.tv_usec * 1000; + runtime_ns_max = diff.tv_sec * NSEC_PER_SEC; + runtime_ns_max += diff.tv_usec * NSEC_PER_USEC; if (details >= 0) { printf(" #%2d / %2d: %14.2lf nsecs/op [val: %016"PRIx64"]\n", @@ -1192,23 +1193,23 @@ static void *worker_thread(void *__tdata) continue; timersub(&stop, &start0, &diff); - runtime_ns_max = diff.tv_sec * 1000000000ULL; - runtime_ns_max += diff.tv_usec * 1000ULL; + runtime_ns_max = diff.tv_sec * NSEC_PER_SEC; + runtime_ns_max += diff.tv_usec * NSEC_PER_USEC; show_summary(runtime_ns_max, l, &convergence); } gettimeofday(&stop, NULL); timersub(&stop, &start0, &diff); - td->runtime_ns = diff.tv_sec * 1000000000ULL; - td->runtime_ns += diff.tv_usec * 1000ULL; - td->speed_gbs = bytes_done / (td->runtime_ns / 1e9) / 1e9; + td->runtime_ns = diff.tv_sec * NSEC_PER_SEC; + td->runtime_ns += diff.tv_usec * NSEC_PER_USEC; + td->speed_gbs = bytes_done / (td->runtime_ns / NSEC_PER_SEC) / 1e9; getrusage(RUSAGE_THREAD, &rusage); - td->system_time_ns = rusage.ru_stime.tv_sec * 1000000000ULL; - td->system_time_ns += rusage.ru_stime.tv_usec * 1000ULL; - td->user_time_ns = rusage.ru_utime.tv_sec * 1000000000ULL; - td->user_time_ns += rusage.ru_utime.tv_usec * 1000ULL; + td->system_time_ns = rusage.ru_stime.tv_sec * NSEC_PER_SEC; + td->system_time_ns += rusage.ru_stime.tv_usec * NSEC_PER_USEC; + td->user_time_ns = rusage.ru_utime.tv_sec * NSEC_PER_SEC; + td->user_time_ns += rusage.ru_utime.tv_usec * NSEC_PER_USEC; free_data(thread_data, g->p.bytes_thread); @@ -1469,7 +1470,7 @@ static int __bench_numa(const char *name) } /* Wait for all the threads to start up: */ while (g->nr_tasks_started != g->p.nr_tasks) - usleep(1000); + usleep(USEC_PER_MSEC); BUG_ON(g->nr_tasks_started != g->p.nr_tasks); @@ -1488,9 +1489,9 @@ static int __bench_numa(const char *name) timersub(&stop, &start, &diff); - startup_sec = diff.tv_sec * 1000000000.0; - startup_sec += diff.tv_usec * 1000.0; - startup_sec /= 1e9; + startup_sec = diff.tv_sec * NSEC_PER_SEC; + startup_sec += diff.tv_usec * NSEC_PER_USEC; + startup_sec /= NSEC_PER_SEC; tprintf(" threads initialized in %.6f seconds.\n", startup_sec); tprintf(" #\n"); @@ -1529,14 +1530,14 @@ static int __bench_numa(const char *name) tprintf("\n ###\n"); tprintf("\n"); - runtime_sec_max = diff.tv_sec * 1000000000.0; - runtime_sec_max += diff.tv_usec * 1000.0; - runtime_sec_max /= 1e9; + runtime_sec_max = diff.tv_sec * NSEC_PER_SEC; + runtime_sec_max += diff.tv_usec * NSEC_PER_USEC; + runtime_sec_max /= NSEC_PER_SEC; - runtime_sec_min = runtime_ns_min/1e9; + runtime_sec_min = runtime_ns_min / NSEC_PER_SEC; bytes = g->bytes_done; - runtime_avg = (double)runtime_ns_sum / g->p.nr_tasks / 1e9; + runtime_avg = (double)runtime_ns_sum / g->p.nr_tasks / NSEC_PER_SEC; if (g->p.measure_convergence) { print_res(name, runtime_sec_max, @@ -1562,7 +1563,7 @@ static int __bench_numa(const char *name) print_res(name, bytes / 1e9, "GB,", "data-total", "GB data processed, total"); - print_res(name, runtime_sec_max * 1e9 / (bytes / g->p.nr_tasks), + print_res(name, runtime_sec_max * NSEC_PER_SEC / (bytes / g->p.nr_tasks), "nsecs,", "runtime/byte/thread","nsecs/byte/thread runtime"); print_res(name, bytes / g->p.nr_tasks / 1e9 / runtime_sec_max, @@ -1581,9 +1582,9 @@ static int __bench_numa(const char *name) snprintf(tname, 32, "process%d:thread%d", p, t); print_res(tname, td->speed_gbs, "GB/sec", "thread-speed", "GB/sec/thread speed"); - print_res(tname, td->system_time_ns / 1e9, + print_res(tname, td->system_time_ns / NSEC_PER_SEC, "secs", "thread-system-time", "system CPU time/thread"); - print_res(tname, td->user_time_ns / 1e9, + print_res(tname, td->user_time_ns / NSEC_PER_SEC, "secs", "thread-user-time", "user CPU time/thread"); } } From 4fc76e495b60da343b94dadeea1001f878ceb955 Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Mon, 8 Aug 2016 12:23:49 -0300 Subject: [PATCH 06/38] perf sched: Use linux/time64.h Probably the next step is to introduce linux/time.h and use timespec_to_ns(), etc. Cc: Adrian Hunter Cc: David Ahern Cc: Jiri Olsa Cc: Namhyung Kim Cc: Steven Rostedt Cc: Wang Nan Link: http://lkml.kernel.org/n/tip-4nqhskn27fn93cz3ukbc8drf@git.kernel.org Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/builtin-sched.c | 37 +++++++++++++++++++------------------ 1 file changed, 19 insertions(+), 18 deletions(-) diff --git a/tools/perf/builtin-sched.c b/tools/perf/builtin-sched.c index 0dfe8df2ab9b..f5503ca22e1c 100644 --- a/tools/perf/builtin-sched.c +++ b/tools/perf/builtin-sched.c @@ -26,6 +26,7 @@ #include #include #include +#include #define PR_SET_NAME 15 /* Set process name */ #define MAX_CPUS 4096 @@ -199,7 +200,7 @@ static u64 get_nsecs(void) clock_gettime(CLOCK_MONOTONIC, &ts); - return ts.tv_sec * 1000000000ULL + ts.tv_nsec; + return ts.tv_sec * NSEC_PER_SEC + ts.tv_nsec; } static void burn_nsecs(struct perf_sched *sched, u64 nsecs) @@ -223,7 +224,7 @@ static void sleep_nsecs(u64 nsecs) static void calibrate_run_measurement_overhead(struct perf_sched *sched) { - u64 T0, T1, delta, min_delta = 1000000000ULL; + u64 T0, T1, delta, min_delta = NSEC_PER_SEC; int i; for (i = 0; i < 10; i++) { @@ -240,7 +241,7 @@ static void calibrate_run_measurement_overhead(struct perf_sched *sched) static void calibrate_sleep_measurement_overhead(struct perf_sched *sched) { - u64 T0, T1, delta, min_delta = 1000000000ULL; + u64 T0, T1, delta, min_delta = NSEC_PER_SEC; int i; for (i = 0; i < 10; i++) { @@ -452,8 +453,8 @@ static u64 get_cpu_usage_nsec_parent(void) err = getrusage(RUSAGE_SELF, &ru); BUG_ON(err); - sum = ru.ru_utime.tv_sec*1e9 + ru.ru_utime.tv_usec*1e3; - sum += ru.ru_stime.tv_sec*1e9 + ru.ru_stime.tv_usec*1e3; + sum = ru.ru_utime.tv_sec * NSEC_PER_SEC + ru.ru_utime.tv_usec * NSEC_PER_USEC; + sum += ru.ru_stime.tv_sec * NSEC_PER_SEC + ru.ru_stime.tv_usec * NSEC_PER_USEC; return sum; } @@ -667,12 +668,12 @@ static void run_one_test(struct perf_sched *sched) sched->run_avg = delta; sched->run_avg = (sched->run_avg * (sched->replay_repeat - 1) + delta) / sched->replay_repeat; - printf("#%-3ld: %0.3f, ", sched->nr_runs, (double)delta / 1000000.0); + printf("#%-3ld: %0.3f, ", sched->nr_runs, (double)delta / NSEC_PER_MSEC); - printf("ravg: %0.2f, ", (double)sched->run_avg / 1e6); + printf("ravg: %0.2f, ", (double)sched->run_avg / NSEC_PER_MSEC); printf("cpu: %0.2f / %0.2f", - (double)sched->cpu_usage / 1e6, (double)sched->runavg_cpu_usage / 1e6); + (double)sched->cpu_usage / NSEC_PER_MSEC, (double)sched->runavg_cpu_usage / NSEC_PER_MSEC); #if 0 /* @@ -680,8 +681,8 @@ static void run_one_test(struct perf_sched *sched) * accurate than the sched->sum_exec_runtime based statistics: */ printf(" [%0.2f / %0.2f]", - (double)sched->parent_cpu_usage/1e6, - (double)sched->runavg_parent_cpu_usage/1e6); + (double)sched->parent_cpu_usage / NSEC_PER_MSEC, + (double)sched->runavg_parent_cpu_usage / NSEC_PER_MSEC); #endif printf("\n"); @@ -696,13 +697,13 @@ static void test_calibrations(struct perf_sched *sched) u64 T0, T1; T0 = get_nsecs(); - burn_nsecs(sched, 1e6); + burn_nsecs(sched, NSEC_PER_MSEC); T1 = get_nsecs(); printf("the run test took %" PRIu64 " nsecs\n", T1 - T0); T0 = get_nsecs(); - sleep_nsecs(1e6); + sleep_nsecs(NSEC_PER_MSEC); T1 = get_nsecs(); printf("the sleep test took %" PRIu64 " nsecs\n", T1 - T0); @@ -1213,10 +1214,10 @@ static void output_lat_thread(struct perf_sched *sched, struct work_atoms *work_ avg = work_list->total_lat / work_list->nb_atoms; printf("|%11.3f ms |%9" PRIu64 " | avg:%9.3f ms | max:%9.3f ms | max at: %13.6f s\n", - (double)work_list->total_runtime / 1e6, - work_list->nb_atoms, (double)avg / 1e6, - (double)work_list->max_lat / 1e6, - (double)work_list->max_lat_at / 1e9); + (double)work_list->total_runtime / NSEC_PER_MSEC, + work_list->nb_atoms, (double)avg / NSEC_PER_MSEC, + (double)work_list->max_lat / NSEC_PER_MSEC, + (double)work_list->max_lat_at / NSEC_PER_SEC); } static int pid_cmp(struct work_atoms *l, struct work_atoms *r) @@ -1491,7 +1492,7 @@ static int map_switch_event(struct perf_sched *sched, struct perf_evsel *evsel, if (sched->map.cpus && !cpu_map__has(sched->map.cpus, this_cpu)) goto out; - color_fprintf(stdout, color, " %12.6f secs ", (double)timestamp/1e9); + color_fprintf(stdout, color, " %12.6f secs ", (double)timestamp / NSEC_PER_SEC); if (new_shortname) { const char *pid_color = color; @@ -1753,7 +1754,7 @@ static int perf_sched__lat(struct perf_sched *sched) printf(" -----------------------------------------------------------------------------------------------------------------\n"); printf(" TOTAL: |%11.3f ms |%9" PRIu64 " |\n", - (double)sched->all_runtime / 1e6, sched->all_count); + (double)sched->all_runtime / NSEC_PER_MSEC, sched->all_count); printf(" ---------------------------------------------------\n"); From af4b2c972a5fc9358486d946d15f32510534ccbf Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Mon, 8 Aug 2016 12:45:58 -0300 Subject: [PATCH 07/38] perf timechart: Use NSEC_PER_U?SEC Following kernel practices, using linux/time64.h Cc: Adrian Hunter Cc: Arjan van de Ven Cc: David Ahern Cc: Jiri Olsa Cc: Namhyung Kim Cc: Stanislav Fomichev Cc: Steven Rostedt Cc: Wang Nan Link: http://lkml.kernel.org/n/tip-5l1md8lsdhfnrlsqyejzo9w2@git.kernel.org Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/builtin-timechart.c | 13 +++++++------ tools/perf/util/svghelper.c | 11 ++++++----- 2 files changed, 13 insertions(+), 11 deletions(-) diff --git a/tools/perf/builtin-timechart.c b/tools/perf/builtin-timechart.c index 733a55422d03..e7eaa298d34a 100644 --- a/tools/perf/builtin-timechart.c +++ b/tools/perf/builtin-timechart.c @@ -24,6 +24,7 @@ #include "util/evlist.h" #include "util/evsel.h" #include +#include #include "util/symbol.h" #include "util/callchain.h" #include "util/strlist.h" @@ -1288,9 +1289,9 @@ static void draw_process_bars(struct timechart *tchart) if (c->comm) { char comm[256]; if (c->total_time > 5000000000) /* 5 seconds */ - sprintf(comm, "%s:%i (%2.2fs)", c->comm, p->pid, c->total_time / 1000000000.0); + sprintf(comm, "%s:%i (%2.2fs)", c->comm, p->pid, c->total_time / (double)NSEC_PER_SEC); else - sprintf(comm, "%s:%i (%3.1fms)", c->comm, p->pid, c->total_time / 1000000.0); + sprintf(comm, "%s:%i (%3.1fms)", c->comm, p->pid, c->total_time / (double)NSEC_PER_MSEC); svg_text(Y, c->start_time, comm); } @@ -1637,7 +1638,7 @@ static int __cmd_timechart(struct timechart *tchart, const char *output_name) write_svg_file(tchart, output_name); pr_info("Written %2.1f seconds of trace to %s.\n", - (tchart->last_time - tchart->first_time) / 1000000000.0, output_name); + (tchart->last_time - tchart->first_time) / (double)NSEC_PER_SEC, output_name); out_delete: perf_session__delete(session); return ret; @@ -1901,10 +1902,10 @@ parse_time(const struct option *opt, const char *arg, int __maybe_unused unset) if (sscanf(arg, "%" PRIu64 "%cs", value, &unit) > 0) { switch (unit) { case 'm': - *value *= 1000000; + *value *= NSEC_PER_MSEC; break; case 'u': - *value *= 1000; + *value *= NSEC_PER_USEC; break; case 'n': break; @@ -1928,7 +1929,7 @@ int cmd_timechart(int argc, const char **argv, .ordered_events = true, }, .proc_num = 15, - .min_time = 1000000, + .min_time = NSEC_PER_MSEC, .merge_dist = 1000, }; const char *output_name = "output.svg"; diff --git a/tools/perf/util/svghelper.c b/tools/perf/util/svghelper.c index eec6c1149f44..1cbada2dc6be 100644 --- a/tools/perf/util/svghelper.c +++ b/tools/perf/util/svghelper.c @@ -18,6 +18,7 @@ #include #include #include +#include #include "perf.h" #include "svghelper.h" @@ -274,14 +275,14 @@ static char *time_to_string(u64 duration) text[0] = 0; - if (duration < 1000) /* less than 1 usec */ + if (duration < NSEC_PER_USEC) /* less than 1 usec */ return text; - if (duration < 1000 * 1000) { /* less than 1 msec */ - sprintf(text, "%.1f us", duration / 1000.0); + if (duration < NSEC_PER_MSEC) { /* less than 1 msec */ + sprintf(text, "%.1f us", duration / (double)NSEC_PER_USEC); return text; } - sprintf(text, "%.1f ms", duration / 1000.0 / 1000); + sprintf(text, "%.1f ms", duration / (double)NSEC_PER_MSEC); return text; } @@ -297,7 +298,7 @@ void svg_waiting(int Yslot, int cpu, u64 start, u64 end, const char *backtrace) style = "waiting"; - if (end-start > 10 * 1000000) /* 10 msec */ + if (end-start > 10 * NSEC_PER_MSEC) /* 10 msec */ style = "WAITING"; text = time_to_string(end-start); From 16633ccff091dacb2138b95e8911a1892983c6fd Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Mon, 8 Aug 2016 14:51:30 -0300 Subject: [PATCH 08/38] perf bench sched-pipe: Use linux/time64.h, USEC_PER_SEC Following kernel practices. Cc: Adrian Hunter Cc: David Ahern Cc: Jiri Olsa Cc: Namhyung Kim Cc: Steven Rostedt Cc: Wang Nan Link: http://lkml.kernel.org/n/tip-wgfu1h1pnw8lc919o2tan58y@git.kernel.org Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/bench/sched-pipe.c | 9 +++++---- 1 file changed, 5 insertions(+), 4 deletions(-) diff --git a/tools/perf/bench/sched-pipe.c b/tools/perf/bench/sched-pipe.c index 1dc2d13cc272..2243f0150d76 100644 --- a/tools/perf/bench/sched-pipe.c +++ b/tools/perf/bench/sched-pipe.c @@ -25,6 +25,7 @@ #include #include #include +#include #include @@ -153,24 +154,24 @@ int bench_sched_pipe(int argc, const char **argv, const char *prefix __maybe_unu printf("# Executed %d pipe operations between two %s\n\n", loops, threaded ? "threads" : "processes"); - result_usec = diff.tv_sec * 1000000; + result_usec = diff.tv_sec * USEC_PER_SEC; result_usec += diff.tv_usec; printf(" %14s: %lu.%03lu [sec]\n\n", "Total time", diff.tv_sec, - (unsigned long) (diff.tv_usec/1000)); + (unsigned long) (diff.tv_usec / USEC_PER_MSEC)); printf(" %14lf usecs/op\n", (double)result_usec / (double)loops); printf(" %14d ops/sec\n", (int)((double)loops / - ((double)result_usec / (double)1000000))); + ((double)result_usec / (double)USEC_PER_SEC))); break; case BENCH_FORMAT_SIMPLE: printf("%lu.%03lu\n", diff.tv_sec, - (unsigned long) (diff.tv_usec / 1000)); + (unsigned long) (diff.tv_usec / USEC_PER_MSEC)); break; default: From 310ebb93676d2106b4c2e68bbf1b2811461bb9f3 Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Mon, 8 Aug 2016 14:57:04 -0300 Subject: [PATCH 09/38] perf stat: Use *SEC_PER_*SEC macros To match how this is done in the kernel. Cc: Adrian Hunter Cc: David Ahern Cc: Jiri Olsa Cc: Namhyung Kim Cc: Steven Rostedt Cc: Wang Nan Link: http://lkml.kernel.org/n/tip-gym6yshewpdegt153u8v2q5r@git.kernel.org Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/builtin-stat.c | 12 ++++++------ 1 file changed, 6 insertions(+), 6 deletions(-) diff --git a/tools/perf/builtin-stat.c b/tools/perf/builtin-stat.c index e33a66bffbdb..90882b1d6a91 100644 --- a/tools/perf/builtin-stat.c +++ b/tools/perf/builtin-stat.c @@ -173,7 +173,7 @@ static inline void diff_timespec(struct timespec *r, struct timespec *a, { r->tv_sec = a->tv_sec - b->tv_sec; if (a->tv_nsec < b->tv_nsec) { - r->tv_nsec = a->tv_nsec + 1000000000L - b->tv_nsec; + r->tv_nsec = a->tv_nsec + NSEC_PER_SEC - b->tv_nsec; r->tv_sec--; } else { r->tv_nsec = a->tv_nsec - b->tv_nsec ; @@ -365,7 +365,7 @@ static void process_interval(void) static void enable_counters(void) { if (initial_delay) - usleep(initial_delay * 1000); + usleep(initial_delay * USEC_PER_MSEC); /* * We need to enable counters only if: @@ -542,8 +542,8 @@ static int __run_perf_stat(int argc, const char **argv) bool is_pipe = STAT_RECORD ? perf_stat.file.is_pipe : false; if (interval) { - ts.tv_sec = interval / 1000; - ts.tv_nsec = (interval % 1000) * 1000000; + ts.tv_sec = interval / USEC_PER_MSEC; + ts.tv_nsec = (interval % USEC_PER_MSEC) * NSEC_PER_MSEC; } else { ts.tv_sec = 1; ts.tv_nsec = 0; @@ -972,7 +972,7 @@ static void print_metric_header(void *ctx, const char *color __maybe_unused, static void nsec_printout(int id, int nr, struct perf_evsel *evsel, double avg) { FILE *output = stat_config.output; - double msecs = avg / 1e6; + double msecs = avg / NSEC_PER_MSEC; const char *fmt_v, *fmt_n; char name[25]; @@ -1461,7 +1461,7 @@ static void print_footer(void) if (!null_run) fprintf(output, "\n"); fprintf(output, " %17.9f seconds time elapsed", - avg_stats(&walltime_nsecs_stats)/1e9); + avg_stats(&walltime_nsecs_stats) / NSEC_PER_SEC); if (run_count > 1) { fprintf(output, " "); print_noise_pct(stddev_stats(&walltime_nsecs_stats), From f2b91be731a6c1aff9333c1ce3fece7a311f1e0d Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Mon, 8 Aug 2016 14:59:21 -0300 Subject: [PATCH 10/38] perf bench mem: Use USEC_PER_SEC Following kernel practices, using linux/time64.h Cc: Adrian Hunter Cc: David Ahern Cc: Jiri Olsa Cc: Namhyung Kim Cc: Steven Rostedt Cc: Wang Nan Link: http://lkml.kernel.org/n/tip-xdtmguafva17wp023sxojiib@git.kernel.org Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/bench/mem-functions.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) diff --git a/tools/perf/bench/mem-functions.c b/tools/perf/bench/mem-functions.c index 2b54d0f2672a..c684910e5a48 100644 --- a/tools/perf/bench/mem-functions.c +++ b/tools/perf/bench/mem-functions.c @@ -21,6 +21,7 @@ #include #include #include +#include #define K 1024 @@ -89,7 +90,7 @@ static u64 get_cycles(void) static double timeval2double(struct timeval *ts) { - return (double)ts->tv_sec + (double)ts->tv_usec / (double)1000000; + return (double)ts->tv_sec + (double)ts->tv_usec / (double)USEC_PER_SEC; } #define print_bps(x) do { \ From af15e67e8f8572072167bfb193ceabff04f4b21e Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Mon, 8 Aug 2016 15:04:23 -0300 Subject: [PATCH 11/38] perf bench sched-messaging: Use USEC_PER_MSEC Cc: Adrian Hunter Cc: David Ahern Cc: Jiri Olsa Cc: Namhyung Kim Cc: Steven Rostedt Cc: Wang Nan Link: http://lkml.kernel.org/n/tip-xhyoyxejvorrgmwjx9k3j8k2@git.kernel.org Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/bench/sched-messaging.c | 5 +++-- 1 file changed, 3 insertions(+), 2 deletions(-) diff --git a/tools/perf/bench/sched-messaging.c b/tools/perf/bench/sched-messaging.c index bfaf9503de8e..6a111e775210 100644 --- a/tools/perf/bench/sched-messaging.c +++ b/tools/perf/bench/sched-messaging.c @@ -29,6 +29,7 @@ #include #include #include +#include #define DATASIZE 100 @@ -312,11 +313,11 @@ int bench_sched_messaging(int argc, const char **argv, thread_mode ? "threads" : "processes"); printf(" %14s: %lu.%03lu [sec]\n", "Total time", diff.tv_sec, - (unsigned long) (diff.tv_usec/1000)); + (unsigned long) (diff.tv_usec / USEC_PER_MSEC)); break; case BENCH_FORMAT_SIMPLE: printf("%lu.%03lu\n", diff.tv_sec, - (unsigned long) (diff.tv_usec/1000)); + (unsigned long) (diff.tv_usec / USEC_PER_MSEC)); break; default: /* reaching here is something disaster */ From 0693e680d32e5c7415666ccfcff57ad4cd976294 Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Mon, 8 Aug 2016 15:05:46 -0300 Subject: [PATCH 12/38] perf record: Use USEC_PER_MSEC Instead of a naked 1000. Cc: Adrian Hunter Cc: David Ahern Cc: Jiri Olsa Cc: Namhyung Kim Cc: Steven Rostedt Cc: Wang Nan Link: http://lkml.kernel.org/n/tip-7v6be7jhvstbkvk3rsytjw0o@git.kernel.org Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/builtin-record.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/tools/perf/builtin-record.c b/tools/perf/builtin-record.c index 6355902fbfc8..a3792e8c3e03 100644 --- a/tools/perf/builtin-record.c +++ b/tools/perf/builtin-record.c @@ -42,7 +42,7 @@ #include #include #include - +#include struct record { struct perf_tool tool; @@ -954,7 +954,7 @@ static int __cmd_record(struct record *rec, int argc, const char **argv) } if (opts->initial_delay) { - usleep(opts->initial_delay * 1000); + usleep(opts->initial_delay * USEC_PER_MSEC); perf_evlist__enable(rec->evlist); } From c05a6e14150f81a8cb93d83af55582e05afc05ae Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Mon, 8 Aug 2016 15:10:44 -0300 Subject: [PATCH 13/38] perf kvm: Use NSEC_PER_USEC Following kernel practices and better documenting units of time. Cc: Adrian Hunter Cc: Alexander Yarygin Cc: David Ahern Cc: Hemant Kumar Cc: Jiri Olsa Cc: Namhyung Kim Cc: Steven Rostedt Cc: Wang Nan Link: http://lkml.kernel.org/n/tip-5x6p6fmzrogonpbnkkkw4usk@git.kernel.org Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/builtin-kvm.c | 10 +++++----- 1 file changed, 5 insertions(+), 5 deletions(-) diff --git a/tools/perf/builtin-kvm.c b/tools/perf/builtin-kvm.c index a0040f74e935..08fa88f62a24 100644 --- a/tools/perf/builtin-kvm.c +++ b/tools/perf/builtin-kvm.c @@ -363,7 +363,7 @@ static bool handle_end_event(struct perf_kvm_stat *kvm, if (!skip_event(decode)) { pr_info("%" PRIu64 " VM %d, vcpu %d: %s event took %" PRIu64 "usec\n", sample->time, sample->pid, vcpu_record->vcpu_id, - decode, time_diff/1000); + decode, time_diff / NSEC_PER_USEC); } } @@ -609,15 +609,15 @@ static void print_result(struct perf_kvm_stat *kvm) pr_info("%10llu ", (unsigned long long)ecount); pr_info("%8.2f%% ", (double)ecount / kvm->total_count * 100); pr_info("%8.2f%% ", (double)etime / kvm->total_time * 100); - pr_info("%9.2fus ", (double)min / 1e3); - pr_info("%9.2fus ", (double)max / 1e3); - pr_info("%9.2fus ( +-%7.2f%% )", (double)etime / ecount/1e3, + pr_info("%9.2fus ", (double)min / NSEC_PER_USEC); + pr_info("%9.2fus ", (double)max / NSEC_PER_USEC); + pr_info("%9.2fus ( +-%7.2f%% )", (double)etime / ecount / NSEC_PER_USEC, kvm_event_rel_stddev(vcpu, event)); pr_info("\n"); } pr_info("\nTotal Samples:%" PRIu64 ", Total events handled time:%.2fus.\n\n", - kvm->total_count, kvm->total_time / 1e3); + kvm->total_count, kvm->total_time / (double)NSEC_PER_USEC); if (kvm->lost_events) pr_info("\nLost events: %" PRIu64 "\n\n", kvm->lost_events); From 565e69114e4aa42664b0e7da135181651c3402f2 Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Mon, 8 Aug 2016 15:35:21 -0300 Subject: [PATCH 14/38] perf bench futex: Use NSEC_PER_USEC Following kernel practices and better documentin Cc: Adrian Hunter Cc: David Ahern Cc: Jiri Olsa Cc: Namhyung Kim Cc: Steven Rostedt Cc: Wang Nan Link: http://lkml.kernel.org/n/tip-xncwqxegjp13g2nxih3lp9mx@git.kernel.org Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/bench/futex-requeue.c | 5 +++-- tools/perf/bench/futex-wake-parallel.c | 5 +++-- tools/perf/bench/futex-wake.c | 5 +++-- 3 files changed, 9 insertions(+), 6 deletions(-) diff --git a/tools/perf/bench/futex-requeue.c b/tools/perf/bench/futex-requeue.c index f96e22ed9f87..2b9705a8734c 100644 --- a/tools/perf/bench/futex-requeue.c +++ b/tools/perf/bench/futex-requeue.c @@ -16,6 +16,7 @@ #include #include #include +#include #include #include "bench.h" #include "futex.h" @@ -62,7 +63,7 @@ static void print_summary(void) printf("Requeued %d of %d threads in %.4f ms (+-%.2f%%)\n", requeued_avg, nthreads, - requeuetime_avg/1e3, + requeuetime_avg / USEC_PER_MSEC, rel_stddev_stats(requeuetime_stddev, requeuetime_avg)); } @@ -184,7 +185,7 @@ int bench_futex_requeue(int argc, const char **argv, if (!silent) { printf("[Run %d]: Requeued %d of %d threads in %.4f ms\n", - j + 1, nrequeued, nthreads, runtime.tv_usec/1e3); + j + 1, nrequeued, nthreads, runtime.tv_usec / (double)USEC_PER_MSEC); } /* everybody should be blocked on futex2, wake'em up */ diff --git a/tools/perf/bench/futex-wake-parallel.c b/tools/perf/bench/futex-wake-parallel.c index 4a2ecd7438ca..2c8fa67ad537 100644 --- a/tools/perf/bench/futex-wake-parallel.c +++ b/tools/perf/bench/futex-wake-parallel.c @@ -15,6 +15,7 @@ #include #include #include +#include #include #include "bench.h" #include "futex.h" @@ -156,7 +157,7 @@ static void print_run(struct thread_data *waking_worker, unsigned int run_num) printf("[Run %d]: Avg per-thread latency (waking %d/%d threads) " "in %.4f ms (+-%.2f%%)\n", run_num + 1, wakeup_avg, - nblocked_threads, waketime_avg/1e3, + nblocked_threads, waketime_avg / USEC_PER_MSEC, rel_stddev_stats(waketime_stddev, waketime_avg)); } @@ -172,7 +173,7 @@ static void print_summary(void) printf("Avg per-thread latency (waking %d/%d threads) in %.4f ms (+-%.2f%%)\n", wakeup_avg, nblocked_threads, - waketime_avg/1e3, + waketime_avg / USEC_PER_MSEC, rel_stddev_stats(waketime_stddev, waketime_avg)); } diff --git a/tools/perf/bench/futex-wake.c b/tools/perf/bench/futex-wake.c index 87d8f4f292d9..e246b1b8388a 100644 --- a/tools/perf/bench/futex-wake.c +++ b/tools/perf/bench/futex-wake.c @@ -16,6 +16,7 @@ #include #include #include +#include #include #include "bench.h" #include "futex.h" @@ -81,7 +82,7 @@ static void print_summary(void) printf("Wokeup %d of %d threads in %.4f ms (+-%.2f%%)\n", wakeup_avg, nthreads, - waketime_avg/1e3, + waketime_avg / USEC_PER_MSEC, rel_stddev_stats(waketime_stddev, waketime_avg)); } @@ -182,7 +183,7 @@ int bench_futex_wake(int argc, const char **argv, if (!silent) { printf("[Run %d]: Wokeup %d of %d threads in %.4f ms\n", - j + 1, nwoken, nthreads, runtime.tv_usec/1e3); + j + 1, nwoken, nthreads, runtime.tv_usec / (double)USEC_PER_MSEC); } for (i = 0; i < nthreads; i++) { From b9c4b0f40d22d4b1d29540f5faf6ca4269f25848 Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Mon, 8 Aug 2016 15:37:58 -0300 Subject: [PATCH 15/38] perf top: Use MSEC_PER_SEC Cc: Adrian Hunter Cc: David Ahern Cc: Jiri Olsa Cc: Namhyung Kim Cc: Steven Rostedt Cc: Wang Nan Link: http://lkml.kernel.org/n/tip-iof4j6mutyogdeie1sj98dhv@git.kernel.org Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/builtin-top.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) diff --git a/tools/perf/builtin-top.c b/tools/perf/builtin-top.c index 418ed94756d3..a3223aa22213 100644 --- a/tools/perf/builtin-top.c +++ b/tools/perf/builtin-top.c @@ -68,6 +68,7 @@ #include #include +#include #include static volatile int done; @@ -624,7 +625,7 @@ static void *display_thread(void *arg) display_setup_sig(); pthread__unblock_sigwinch(); repeat: - delay_msecs = top->delay_secs * 1000; + delay_msecs = top->delay_secs * MSEC_PER_SEC; set_term_quiet_input(&save); /* trash return*/ getc(stdin); From f8e6710de859e1ac3a5df294bddeca19f60cec9a Mon Sep 17 00:00:00 2001 From: Jiri Olsa Date: Sun, 7 Aug 2016 17:28:26 +0200 Subject: [PATCH 16/38] perf hists: Introduce nr_header_lines into struct perf_hpp_list Currently we support just single line headers, this is first step to allow more. Store the number of header lines in perf_hpp_list, which encompasses all the display/sort entries and is thus suitable to hold this value. Signed-off-by: Jiri Olsa Cc: David Ahern Cc: Namhyung Kim Cc: Peter Zijlstra Link: http://lkml.kernel.org/r/1470583710-1649-2-git-send-email-jolsa@kernel.org Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/ui/browsers/hists.c | 15 ++++++++++++--- tools/perf/ui/hist.c | 1 + tools/perf/util/hist.h | 1 + 3 files changed, 14 insertions(+), 3 deletions(-) diff --git a/tools/perf/ui/browsers/hists.c b/tools/perf/ui/browsers/hists.c index 13d414384739..95f7cf1991fc 100644 --- a/tools/perf/ui/browsers/hists.c +++ b/tools/perf/ui/browsers/hists.c @@ -69,8 +69,11 @@ static u32 hist_browser__nr_entries(struct hist_browser *hb) static void hist_browser__update_rows(struct hist_browser *hb) { struct ui_browser *browser = &hb->b; - u16 header_offset = hb->show_headers ? 1 : 0, index_row; + struct hists *hists = hb->hists; + struct perf_hpp_list *hpp_list = hists->hpp_list; + u16 header_offset, index_row; + header_offset = hb->show_headers ? hpp_list->nr_header_lines : 0; browser->rows = browser->height - header_offset; /* * Verify if we were at the last line and that line isn't @@ -99,8 +102,11 @@ static void hist_browser__refresh_dimensions(struct ui_browser *browser) static void hist_browser__gotorc(struct hist_browser *browser, int row, int column) { - u16 header_offset = browser->show_headers ? 1 : 0; + struct hists *hists = browser->hists; + struct perf_hpp_list *hpp_list = hists->hpp_list; + u16 header_offset; + header_offset = browser->show_headers ? hpp_list->nr_header_lines : 0; ui_browser__gotorc(&browser->b, row + header_offset, column); } @@ -1656,10 +1662,13 @@ static unsigned int hist_browser__refresh(struct ui_browser *browser) u16 header_offset = 0; struct rb_node *nd; struct hist_browser *hb = container_of(browser, struct hist_browser, b); + struct hists *hists = hb->hists; if (hb->show_headers) { + struct perf_hpp_list *hpp_list = hists->hpp_list; + hist_browser__show_headers(hb); - header_offset = 1; + header_offset = hpp_list->nr_header_lines; } ui_browser__hists_init_top(browser); diff --git a/tools/perf/ui/hist.c b/tools/perf/ui/hist.c index 4274969ddc89..be7a17fec889 100644 --- a/tools/perf/ui/hist.c +++ b/tools/perf/ui/hist.c @@ -441,6 +441,7 @@ struct perf_hpp_fmt perf_hpp__format[] = { struct perf_hpp_list perf_hpp_list = { .fields = LIST_HEAD_INIT(perf_hpp_list.fields), .sorts = LIST_HEAD_INIT(perf_hpp_list.sorts), + .nr_header_lines = 1, }; #undef HPP__COLOR_PRINT_FNS diff --git a/tools/perf/util/hist.h b/tools/perf/util/hist.h index 0a1edf1ab450..8cc5d33d2405 100644 --- a/tools/perf/util/hist.h +++ b/tools/perf/util/hist.h @@ -259,6 +259,7 @@ struct perf_hpp_list { struct list_head fields; struct list_head sorts; + int nr_header_lines; int need_collapse; int parent; int sym; From 74bb43f29ec80bc998804fa7399930d86c4bae67 Mon Sep 17 00:00:00 2001 From: Jiri Olsa Date: Sun, 7 Aug 2016 17:28:27 +0200 Subject: [PATCH 17/38] perf hists: Add line argument into perf_hpp_fmt's header callback Adding line argument into perf_hpp_fmt's header callback to be able to request specific header line. Signed-off-by: Jiri Olsa Cc: David Ahern Cc: Namhyung Kim Cc: Peter Zijlstra Link: http://lkml.kernel.org/r/1470583710-1649-3-git-send-email-jolsa@kernel.org Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/builtin-diff.c | 3 ++- tools/perf/ui/browsers/hists.c | 6 +++--- tools/perf/ui/gtk/hists.c | 2 +- tools/perf/ui/hist.c | 2 +- tools/perf/ui/stdio/hist.c | 6 +++--- tools/perf/util/hist.h | 2 +- tools/perf/util/sort.c | 5 +++-- 7 files changed, 14 insertions(+), 12 deletions(-) diff --git a/tools/perf/builtin-diff.c b/tools/perf/builtin-diff.c index 21ee753211ad..75da965df343 100644 --- a/tools/perf/builtin-diff.c +++ b/tools/perf/builtin-diff.c @@ -1033,7 +1033,8 @@ static int hpp__entry_global(struct perf_hpp_fmt *_fmt, struct perf_hpp *hpp, } static int hpp__header(struct perf_hpp_fmt *fmt, struct perf_hpp *hpp, - struct hists *hists __maybe_unused) + struct hists *hists __maybe_unused, + int line __maybe_unused) { struct diff_hpp_fmt *dfmt = container_of(fmt, struct diff_hpp_fmt, fmt); diff --git a/tools/perf/ui/browsers/hists.c b/tools/perf/ui/browsers/hists.c index 95f7cf1991fc..303ed62a8f53 100644 --- a/tools/perf/ui/browsers/hists.c +++ b/tools/perf/ui/browsers/hists.c @@ -1523,7 +1523,7 @@ static int hists_browser__scnprintf_headers(struct hist_browser *browser, char * if (perf_hpp__should_skip(fmt, hists) || column++ < browser->b.horiz_scroll) continue; - ret = fmt->header(fmt, &dummy_hpp, hists); + ret = fmt->header(fmt, &dummy_hpp, hists, 0); if (advance_hpp_check(&dummy_hpp, ret)) break; @@ -1560,7 +1560,7 @@ static int hists_browser__scnprintf_hierarchy_headers(struct hist_browser *brows if (column++ < browser->b.horiz_scroll) continue; - ret = fmt->header(fmt, &dummy_hpp, hists); + ret = fmt->header(fmt, &dummy_hpp, hists, 0); if (advance_hpp_check(&dummy_hpp, ret)) break; @@ -1597,7 +1597,7 @@ static int hists_browser__scnprintf_hierarchy_headers(struct hist_browser *brows } first_col = false; - ret = fmt->header(fmt, &dummy_hpp, hists); + ret = fmt->header(fmt, &dummy_hpp, hists, 0); dummy_hpp.buf[ret] = '\0'; start = trim(dummy_hpp.buf); diff --git a/tools/perf/ui/gtk/hists.c b/tools/perf/ui/gtk/hists.c index c5f3677f6679..79cb5c4d5e2e 100644 --- a/tools/perf/ui/gtk/hists.c +++ b/tools/perf/ui/gtk/hists.c @@ -549,7 +549,7 @@ static void perf_gtk__show_hierarchy(GtkWidget *window, struct hists *hists, strcat(buf, "+"); first_col = false; - fmt->header(fmt, &hpp, hists); + fmt->header(fmt, &hpp, hists, 0); strcat(buf, ltrim(rtrim(hpp.buf))); } } diff --git a/tools/perf/ui/hist.c b/tools/perf/ui/hist.c index be7a17fec889..30457c65980c 100644 --- a/tools/perf/ui/hist.c +++ b/tools/perf/ui/hist.c @@ -230,7 +230,7 @@ static int hpp__width_fn(struct perf_hpp_fmt *fmt, } static int hpp__header_fn(struct perf_hpp_fmt *fmt, struct perf_hpp *hpp, - struct hists *hists) + struct hists *hists, int line __maybe_unused) { int len = hpp__width_fn(fmt, hpp, hists); return scnprintf(hpp->buf, hpp->size, "%*s", len, fmt->name); diff --git a/tools/perf/ui/stdio/hist.c b/tools/perf/ui/stdio/hist.c index f04a63112079..91b8e10cabd5 100644 --- a/tools/perf/ui/stdio/hist.c +++ b/tools/perf/ui/stdio/hist.c @@ -549,7 +549,7 @@ static int print_hierarchy_header(struct hists *hists, struct perf_hpp *hpp, struct perf_hpp_list_node, list); perf_hpp_list__for_each_format(&fmt_node->hpp, fmt) { - fmt->header(fmt, hpp, hists); + fmt->header(fmt, hpp, hists, 0); fprintf(fp, "%s%s", hpp->buf, sep ?: " "); } @@ -569,7 +569,7 @@ static int print_hierarchy_header(struct hists *hists, struct perf_hpp *hpp, header_width += fprintf(fp, "+"); first_col = false; - fmt->header(fmt, hpp, hists); + fmt->header(fmt, hpp, hists, 0); header_width += fprintf(fp, "%s", trim(hpp->buf)); } @@ -658,7 +658,7 @@ hists__fprintf_standard_headers(struct hists *hists, else first = false; - fmt->header(fmt, hpp, hists); + fmt->header(fmt, hpp, hists, 0); fprintf(fp, "%s", hpp->buf); } diff --git a/tools/perf/util/hist.h b/tools/perf/util/hist.h index 8cc5d33d2405..71a44c1c44de 100644 --- a/tools/perf/util/hist.h +++ b/tools/perf/util/hist.h @@ -230,7 +230,7 @@ struct perf_hpp { struct perf_hpp_fmt { const char *name; int (*header)(struct perf_hpp_fmt *fmt, struct perf_hpp *hpp, - struct hists *hists); + struct hists *hists, int line); int (*width)(struct perf_hpp_fmt *fmt, struct perf_hpp *hpp, struct hists *hists); int (*color)(struct perf_hpp_fmt *fmt, struct perf_hpp *hpp, diff --git a/tools/perf/util/sort.c b/tools/perf/util/sort.c index 3d3cb8392c86..a556de751018 100644 --- a/tools/perf/util/sort.c +++ b/tools/perf/util/sort.c @@ -1492,7 +1492,7 @@ void perf_hpp__reset_sort_width(struct perf_hpp_fmt *fmt, struct hists *hists) } static int __sort__hpp_header(struct perf_hpp_fmt *fmt, struct perf_hpp *hpp, - struct hists *hists) + struct hists *hists, int line __maybe_unused) { struct hpp_sort_entry *hse; size_t len = fmt->user_len; @@ -1797,7 +1797,8 @@ static void update_dynamic_len(struct hpp_dynamic_entry *hde, } static int __sort__hde_header(struct perf_hpp_fmt *fmt, struct perf_hpp *hpp, - struct hists *hists __maybe_unused) + struct hists *hists __maybe_unused, + int line __maybe_unused) { struct hpp_dynamic_entry *hde; size_t len = fmt->user_len; From 69705b35859195b5640861fff91cc936d1baf4eb Mon Sep 17 00:00:00 2001 From: Jiri Olsa Date: Sun, 7 Aug 2016 17:28:28 +0200 Subject: [PATCH 18/38] perf tools tui: Display multiple header lines Display multiple header lines in TUI browser, if it's configured within struct perf_hpp_list::nr_header_lines. Signed-off-by: Jiri Olsa Cc: David Ahern Cc: Namhyung Kim Cc: Peter Zijlstra Link: http://lkml.kernel.org/r/1470583710-1649-4-git-send-email-jolsa@kernel.org Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/ui/browsers/hists.c | 25 +++++++++++++++++-------- 1 file changed, 17 insertions(+), 8 deletions(-) diff --git a/tools/perf/ui/browsers/hists.c b/tools/perf/ui/browsers/hists.c index 303ed62a8f53..4a7aac8ed5ee 100644 --- a/tools/perf/ui/browsers/hists.c +++ b/tools/perf/ui/browsers/hists.c @@ -1502,7 +1502,9 @@ static int advance_hpp_check(struct perf_hpp *hpp, int inc) return hpp->size <= 0; } -static int hists_browser__scnprintf_headers(struct hist_browser *browser, char *buf, size_t size) +static int +hists_browser__scnprintf_headers(struct hist_browser *browser, char *buf, + size_t size, int line) { struct hists *hists = browser->hists; struct perf_hpp dummy_hpp = { @@ -1523,7 +1525,7 @@ static int hists_browser__scnprintf_headers(struct hist_browser *browser, char * if (perf_hpp__should_skip(fmt, hists) || column++ < browser->b.horiz_scroll) continue; - ret = fmt->header(fmt, &dummy_hpp, hists, 0); + ret = fmt->header(fmt, &dummy_hpp, hists, line); if (advance_hpp_check(&dummy_hpp, ret)) break; @@ -1628,14 +1630,21 @@ static void hists_browser__hierarchy_headers(struct hist_browser *browser) static void hists_browser__headers(struct hist_browser *browser) { - char headers[1024]; + struct hists *hists = browser->hists; + struct perf_hpp_list *hpp_list = hists->hpp_list; - hists_browser__scnprintf_headers(browser, headers, - sizeof(headers)); + int line; - ui_browser__gotorc(&browser->b, 0, 0); - ui_browser__set_color(&browser->b, HE_COLORSET_ROOT); - ui_browser__write_nstring(&browser->b, headers, browser->b.width + 1); + for (line = 0; line < hpp_list->nr_header_lines; line++) { + char headers[1024]; + + hists_browser__scnprintf_headers(browser, headers, + sizeof(headers), line); + + ui_browser__gotorc(&browser->b, line, 0); + ui_browser__set_color(&browser->b, HE_COLORSET_ROOT); + ui_browser__write_nstring(&browser->b, headers, browser->b.width + 1); + } } static void hist_browser__show_headers(struct hist_browser *browser) From f3705b062eafc2867eb0e9ee3502bd59564f103c Mon Sep 17 00:00:00 2001 From: Jiri Olsa Date: Sun, 7 Aug 2016 17:28:29 +0200 Subject: [PATCH 19/38] perf tools stdio: Display multiple header lines Display multiple header lines in stdio output , if it's configured within struct perf_hpp_list::nr_header_lines. Signed-off-by: Jiri Olsa Cc: David Ahern Cc: Namhyung Kim Cc: Peter Zijlstra Link: http://lkml.kernel.org/r/1470583710-1649-5-git-send-email-jolsa@kernel.org Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/ui/stdio/hist.c | 34 +++++++++++++++++++++++++--------- 1 file changed, 25 insertions(+), 9 deletions(-) diff --git a/tools/perf/ui/stdio/hist.c b/tools/perf/ui/stdio/hist.c index 91b8e10cabd5..9b807cbb56b1 100644 --- a/tools/perf/ui/stdio/hist.c +++ b/tools/perf/ui/stdio/hist.c @@ -639,13 +639,10 @@ hists__fprintf_hierarchy_headers(struct hists *hists, return print_hierarchy_header(hists, hpp, symbol_conf.field_sep, fp); } -static int -hists__fprintf_standard_headers(struct hists *hists, - struct perf_hpp *hpp, - FILE *fp) +static void fprintf_line(struct hists *hists, struct perf_hpp *hpp, + int line, FILE *fp) { struct perf_hpp_fmt *fmt; - unsigned int width; const char *sep = symbol_conf.field_sep; bool first = true; @@ -658,14 +655,33 @@ hists__fprintf_standard_headers(struct hists *hists, else first = false; - fmt->header(fmt, hpp, hists, 0); + fmt->header(fmt, hpp, hists, line); fprintf(fp, "%s", hpp->buf); } +} - fprintf(fp, "\n"); +static int +hists__fprintf_standard_headers(struct hists *hists, + struct perf_hpp *hpp, + FILE *fp) +{ + struct perf_hpp_list *hpp_list = hists->hpp_list; + struct perf_hpp_fmt *fmt; + unsigned int width; + const char *sep = symbol_conf.field_sep; + bool first = true; + int line; + + for (line = 0; line < hpp_list->nr_header_lines; line++) { + /* first # is displayed one level up */ + if (line) + fprintf(fp, "# "); + fprintf_line(hists, hpp, line, fp); + fprintf(fp, "\n"); + } if (sep) - return 1; + return hpp_list->nr_header_lines; first = true; @@ -689,7 +705,7 @@ hists__fprintf_standard_headers(struct hists *hists, fprintf(fp, "\n"); fprintf(fp, "#\n"); - return 3; + return hpp_list->nr_header_lines + 2; } static int hists__fprintf_headers(struct hists *hists, FILE *fp) From 29659ab4e7d02c08d8f2d08db0b0f708bd8b6771 Mon Sep 17 00:00:00 2001 From: Jiri Olsa Date: Sun, 7 Aug 2016 17:28:30 +0200 Subject: [PATCH 20/38] perf hists: Add support for header span Add span argument for header callback function. The handling of this argument is completely in the hands of the callback. The only thing the caller ensures is it's zeroed on the beginning. Omitting span skipping in hierarchy headers and gtk code. The c2c code use this to span header lines based on the entries span configuration. Signed-off-by: Jiri Olsa Cc: David Ahern Cc: Namhyung Kim Cc: Peter Zijlstra Link: http://lkml.kernel.org/r/1470583710-1649-6-git-send-email-jolsa@kernel.org Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/builtin-diff.c | 3 ++- tools/perf/ui/browsers/hists.c | 10 +++++++--- tools/perf/ui/gtk/hists.c | 2 +- tools/perf/ui/hist.c | 3 ++- tools/perf/ui/stdio/hist.c | 13 ++++++++----- tools/perf/util/hist.h | 2 +- tools/perf/util/sort.c | 6 ++++-- 7 files changed, 25 insertions(+), 14 deletions(-) diff --git a/tools/perf/builtin-diff.c b/tools/perf/builtin-diff.c index 75da965df343..9ff0db4e2d0c 100644 --- a/tools/perf/builtin-diff.c +++ b/tools/perf/builtin-diff.c @@ -1034,7 +1034,8 @@ static int hpp__entry_global(struct perf_hpp_fmt *_fmt, struct perf_hpp *hpp, static int hpp__header(struct perf_hpp_fmt *fmt, struct perf_hpp *hpp, struct hists *hists __maybe_unused, - int line __maybe_unused) + int line __maybe_unused, + int *span __maybe_unused) { struct diff_hpp_fmt *dfmt = container_of(fmt, struct diff_hpp_fmt, fmt); diff --git a/tools/perf/ui/browsers/hists.c b/tools/perf/ui/browsers/hists.c index 4a7aac8ed5ee..22e48a3ebe29 100644 --- a/tools/perf/ui/browsers/hists.c +++ b/tools/perf/ui/browsers/hists.c @@ -1514,6 +1514,7 @@ hists_browser__scnprintf_headers(struct hist_browser *browser, char *buf, struct perf_hpp_fmt *fmt; size_t ret = 0; int column = 0; + int span = 0; if (symbol_conf.use_callchain) { ret = scnprintf(buf, size, " "); @@ -1525,10 +1526,13 @@ hists_browser__scnprintf_headers(struct hist_browser *browser, char *buf, if (perf_hpp__should_skip(fmt, hists) || column++ < browser->b.horiz_scroll) continue; - ret = fmt->header(fmt, &dummy_hpp, hists, line); + ret = fmt->header(fmt, &dummy_hpp, hists, line, &span); if (advance_hpp_check(&dummy_hpp, ret)) break; + if (span) + continue; + ret = scnprintf(dummy_hpp.buf, dummy_hpp.size, " "); if (advance_hpp_check(&dummy_hpp, ret)) break; @@ -1562,7 +1566,7 @@ static int hists_browser__scnprintf_hierarchy_headers(struct hist_browser *brows if (column++ < browser->b.horiz_scroll) continue; - ret = fmt->header(fmt, &dummy_hpp, hists, 0); + ret = fmt->header(fmt, &dummy_hpp, hists, 0, NULL); if (advance_hpp_check(&dummy_hpp, ret)) break; @@ -1599,7 +1603,7 @@ static int hists_browser__scnprintf_hierarchy_headers(struct hist_browser *brows } first_col = false; - ret = fmt->header(fmt, &dummy_hpp, hists, 0); + ret = fmt->header(fmt, &dummy_hpp, hists, 0, NULL); dummy_hpp.buf[ret] = '\0'; start = trim(dummy_hpp.buf); diff --git a/tools/perf/ui/gtk/hists.c b/tools/perf/ui/gtk/hists.c index 79cb5c4d5e2e..a4f02de7c1b5 100644 --- a/tools/perf/ui/gtk/hists.c +++ b/tools/perf/ui/gtk/hists.c @@ -549,7 +549,7 @@ static void perf_gtk__show_hierarchy(GtkWidget *window, struct hists *hists, strcat(buf, "+"); first_col = false; - fmt->header(fmt, &hpp, hists, 0); + fmt->header(fmt, &hpp, hists, 0, NULL); strcat(buf, ltrim(rtrim(hpp.buf))); } } diff --git a/tools/perf/ui/hist.c b/tools/perf/ui/hist.c index 30457c65980c..b47fafc8ee2a 100644 --- a/tools/perf/ui/hist.c +++ b/tools/perf/ui/hist.c @@ -230,7 +230,8 @@ static int hpp__width_fn(struct perf_hpp_fmt *fmt, } static int hpp__header_fn(struct perf_hpp_fmt *fmt, struct perf_hpp *hpp, - struct hists *hists, int line __maybe_unused) + struct hists *hists, int line __maybe_unused, + int *span __maybe_unused) { int len = hpp__width_fn(fmt, hpp, hists); return scnprintf(hpp->buf, hpp->size, "%*s", len, fmt->name); diff --git a/tools/perf/ui/stdio/hist.c b/tools/perf/ui/stdio/hist.c index 9b807cbb56b1..9b65f4a6b35a 100644 --- a/tools/perf/ui/stdio/hist.c +++ b/tools/perf/ui/stdio/hist.c @@ -549,7 +549,7 @@ static int print_hierarchy_header(struct hists *hists, struct perf_hpp *hpp, struct perf_hpp_list_node, list); perf_hpp_list__for_each_format(&fmt_node->hpp, fmt) { - fmt->header(fmt, hpp, hists, 0); + fmt->header(fmt, hpp, hists, 0, NULL); fprintf(fp, "%s%s", hpp->buf, sep ?: " "); } @@ -569,7 +569,7 @@ static int print_hierarchy_header(struct hists *hists, struct perf_hpp *hpp, header_width += fprintf(fp, "+"); first_col = false; - fmt->header(fmt, hpp, hists, 0); + fmt->header(fmt, hpp, hists, 0, NULL); header_width += fprintf(fp, "%s", trim(hpp->buf)); } @@ -645,18 +645,21 @@ static void fprintf_line(struct hists *hists, struct perf_hpp *hpp, struct perf_hpp_fmt *fmt; const char *sep = symbol_conf.field_sep; bool first = true; + int span = 0; hists__for_each_format(hists, fmt) { if (perf_hpp__should_skip(fmt, hists)) continue; - if (!first) + if (!first && !span) fprintf(fp, "%s", sep ?: " "); else first = false; - fmt->header(fmt, hpp, hists, line); - fprintf(fp, "%s", hpp->buf); + fmt->header(fmt, hpp, hists, line, &span); + + if (!span) + fprintf(fp, "%s", hpp->buf); } } diff --git a/tools/perf/util/hist.h b/tools/perf/util/hist.h index 71a44c1c44de..a002c93fe422 100644 --- a/tools/perf/util/hist.h +++ b/tools/perf/util/hist.h @@ -230,7 +230,7 @@ struct perf_hpp { struct perf_hpp_fmt { const char *name; int (*header)(struct perf_hpp_fmt *fmt, struct perf_hpp *hpp, - struct hists *hists, int line); + struct hists *hists, int line, int *span); int (*width)(struct perf_hpp_fmt *fmt, struct perf_hpp *hpp, struct hists *hists); int (*color)(struct perf_hpp_fmt *fmt, struct perf_hpp *hpp, diff --git a/tools/perf/util/sort.c b/tools/perf/util/sort.c index a556de751018..5ae3d32a21d6 100644 --- a/tools/perf/util/sort.c +++ b/tools/perf/util/sort.c @@ -1492,7 +1492,8 @@ void perf_hpp__reset_sort_width(struct perf_hpp_fmt *fmt, struct hists *hists) } static int __sort__hpp_header(struct perf_hpp_fmt *fmt, struct perf_hpp *hpp, - struct hists *hists, int line __maybe_unused) + struct hists *hists, int line __maybe_unused, + int *span __maybe_unused) { struct hpp_sort_entry *hse; size_t len = fmt->user_len; @@ -1798,7 +1799,8 @@ static void update_dynamic_len(struct hpp_dynamic_entry *hde, static int __sort__hde_header(struct perf_hpp_fmt *fmt, struct perf_hpp *hpp, struct hists *hists __maybe_unused, - int line __maybe_unused) + int line __maybe_unused, + int *span __maybe_unused) { struct hpp_dynamic_entry *hde; size_t len = fmt->user_len; From c12944f7faa7f76441d83c1413f13e8bc70162b2 Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Tue, 9 Aug 2016 14:56:13 -0300 Subject: [PATCH 21/38] perf disassemble: Move check for kallsyms + !kcore We don't need to do all that filename logic to then just have to test something unrelated and bail out, move it to the start of the function. Cc: Adrian Hunter Cc: David Ahern Cc: Jiri Olsa Cc: Namhyung Kim Cc: Wang Nan Link: http://lkml.kernel.org/n/tip-lk1v4srtsktonnyp6t1o0uhx@git.kernel.org Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/util/annotate.c | 16 ++++++++-------- 1 file changed, 8 insertions(+), 8 deletions(-) diff --git a/tools/perf/util/annotate.c b/tools/perf/util/annotate.c index 4024d309bb00..9882bc6ac566 100644 --- a/tools/perf/util/annotate.c +++ b/tools/perf/util/annotate.c @@ -1165,11 +1165,10 @@ int symbol__strerror_disassemble(struct symbol *sym __maybe_unused, struct map * int symbol__disassemble(struct symbol *sym, struct map *map, size_t privsize) { struct dso *dso = map->dso; - char *filename = dso__build_id_filename(dso, NULL, 0); + char *filename; bool free_filename = true; char command[PATH_MAX * 2]; FILE *file; - int err = 0; char symfs_filename[PATH_MAX]; struct kcore_extract kce; bool delete_extract = false; @@ -1177,7 +1176,13 @@ int symbol__disassemble(struct symbol *sym, struct map *map, size_t privsize) int lineno = 0; int nline; pid_t pid; + int err = SYMBOL_ANNOTATE_ERRNO__NO_VMLINUX; + if (dso->symtab_type == DSO_BINARY_TYPE__KALLSYMS && + !dso__is_kcore(dso)) + goto out; + + filename = dso__build_id_filename(dso, NULL, 0); if (filename) symbol__join_symfs(symfs_filename, filename); @@ -1201,12 +1206,6 @@ fallback: free_filename = false; } - if (dso->symtab_type == DSO_BINARY_TYPE__KALLSYMS && - !dso__is_kcore(dso)) { - err = SYMBOL_ANNOTATE_ERRNO__NO_VMLINUX; - goto out_free_filename; - } - pr_debug("%s: filename=%s, sym=%s, start=%#" PRIx64 ", end=%#" PRIx64 "\n", __func__, filename, sym->name, map->unmap_ip(map, sym->start), map->unmap_ip(map, sym->end)); @@ -1338,6 +1337,7 @@ out_free_filename: kcore_extract__delete(&kce); if (free_filename) free(filename); +out: return err; out_close_stdout: From 3caee094d160b0ef92988099105e9a173a3805b0 Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Tue, 9 Aug 2016 15:16:37 -0300 Subject: [PATCH 22/38] perf disassemble: Simplify logic for picking the filename to disassemble Lots of changes to support kcore, compressed modules, build-id files left us with some spaguetti code, simplify it a bit, more to come. Cc: Adrian Hunter Cc: David Ahern Cc: Jiri Olsa Cc: Namhyung Kim Cc: Wang Nan Link: http://lkml.kernel.org/n/tip-h70p7x451li3f2fhs44vzmm8@git.kernel.org Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/util/annotate.c | 41 +++++++++++++++----------------------- 1 file changed, 16 insertions(+), 25 deletions(-) diff --git a/tools/perf/util/annotate.c b/tools/perf/util/annotate.c index 9882bc6ac566..8a9949399112 100644 --- a/tools/perf/util/annotate.c +++ b/tools/perf/util/annotate.c @@ -1166,7 +1166,6 @@ int symbol__disassemble(struct symbol *sym, struct map *map, size_t privsize) { struct dso *dso = map->dso; char *filename; - bool free_filename = true; char command[PATH_MAX * 2]; FILE *file; char symfs_filename[PATH_MAX]; @@ -1183,31 +1182,30 @@ int symbol__disassemble(struct symbol *sym, struct map *map, size_t privsize) goto out; filename = dso__build_id_filename(dso, NULL, 0); - if (filename) + if (filename) { symbol__join_symfs(symfs_filename, filename); - - if (filename == NULL) { + free(filename); + } else { if (dso->has_build_id) return ENOMEM; goto fallback; - } else if (dso__is_kcore(dso) || - readlink(symfs_filename, command, sizeof(command)) < 0 || - strstr(command, DSO__NAME_KALLSYMS) || - access(symfs_filename, R_OK)) { - free(filename); + } + + if (dso__is_kcore(dso) || + readlink(symfs_filename, command, sizeof(command)) < 0 || + strstr(command, DSO__NAME_KALLSYMS) || + access(symfs_filename, R_OK)) { fallback: /* * If we don't have build-ids or the build-id file isn't in the * cache, or is just a kallsyms file, well, lets hope that this * DSO is the same as when 'perf record' ran. */ - filename = (char *)dso->long_name; - symbol__join_symfs(symfs_filename, filename); - free_filename = false; + symbol__join_symfs(symfs_filename, dso->long_name); } pr_debug("%s: filename=%s, sym=%s, start=%#" PRIx64 ", end=%#" PRIx64 "\n", __func__, - filename, sym->name, map->unmap_ip(map, sym->start), + symfs_filename, sym->name, map->unmap_ip(map, sym->start), map->unmap_ip(map, sym->end)); pr_debug("annotating [%p] %30s : [%p] %30s\n", @@ -1222,11 +1220,6 @@ fallback: delete_extract = true; strlcpy(symfs_filename, kce.extract_filename, sizeof(symfs_filename)); - if (free_filename) { - free(filename); - free_filename = false; - } - filename = symfs_filename; } } else if (dso__needs_decompress(dso)) { char tmp[PATH_MAX]; @@ -1235,14 +1228,14 @@ fallback: bool ret; if (kmod_path__parse_ext(&m, symfs_filename)) - goto out_free_filename; + goto out; snprintf(tmp, PATH_MAX, "/tmp/perf-kmod-XXXXXX"); fd = mkstemp(tmp); if (fd < 0) { free(m.ext); - goto out_free_filename; + goto out; } ret = decompress_to_file(m.ext, symfs_filename, fd); @@ -1254,7 +1247,7 @@ fallback: close(fd); if (!ret) - goto out_free_filename; + goto out; strcpy(symfs_filename, tmp); } @@ -1270,7 +1263,7 @@ fallback: map__rip_2objdump(map, sym->end), symbol_conf.annotate_asm_raw ? "" : "--no-show-raw", symbol_conf.annotate_src ? "-S" : "", - symfs_filename, filename); + symfs_filename, symfs_filename); pr_debug("Executing: %s\n", command); @@ -1332,11 +1325,9 @@ out_remove_tmp: if (dso__needs_decompress(dso)) unlink(symfs_filename); -out_free_filename: + if (delete_extract) kcore_extract__delete(&kce); - if (free_filename) - free(filename); out: return err; From 05ed3ac9417dac9be9dd63f3cf97416c80bad359 Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Tue, 9 Aug 2016 15:32:53 -0300 Subject: [PATCH 23/38] perf disassemble: Extract logic to find file to pass to objdump to a separate function Disentangling this a bit further, more to come. Cc: Adrian Hunter Cc: David Ahern Cc: Jiri Olsa Cc: Namhyung Kim Cc: Wang Nan Link: http://lkml.kernel.org/n/tip-7bjv2xazuyzs0xw01mlwosn5@git.kernel.org Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/util/annotate.c | 66 ++++++++++++++++++++++---------------- 1 file changed, 38 insertions(+), 28 deletions(-) diff --git a/tools/perf/util/annotate.c b/tools/perf/util/annotate.c index 8a9949399112..25a9259a6a6e 100644 --- a/tools/perf/util/annotate.c +++ b/tools/perf/util/annotate.c @@ -1162,10 +1162,44 @@ int symbol__strerror_disassemble(struct symbol *sym __maybe_unused, struct map * return 0; } +static int dso__disassemble_filename(struct dso *dso, char *filename, size_t filename_size) +{ + char linkname[PATH_MAX]; + char *build_id_filename; + + if (dso->symtab_type == DSO_BINARY_TYPE__KALLSYMS && + !dso__is_kcore(dso)) + return SYMBOL_ANNOTATE_ERRNO__NO_VMLINUX; + + build_id_filename = dso__build_id_filename(dso, NULL, 0); + if (build_id_filename) { + __symbol__join_symfs(filename, filename_size, build_id_filename); + free(build_id_filename); + } else { + if (dso->has_build_id) + return ENOMEM; + goto fallback; + } + + if (dso__is_kcore(dso) || + readlink(filename, linkname, sizeof(linkname)) < 0 || + strstr(linkname, DSO__NAME_KALLSYMS) || + access(filename, R_OK)) { +fallback: + /* + * If we don't have build-ids or the build-id file isn't in the + * cache, or is just a kallsyms file, well, lets hope that this + * DSO is the same as when 'perf record' ran. + */ + __symbol__join_symfs(filename, filename_size, dso->long_name); + } + + return 0; +} + int symbol__disassemble(struct symbol *sym, struct map *map, size_t privsize) { struct dso *dso = map->dso; - char *filename; char command[PATH_MAX * 2]; FILE *file; char symfs_filename[PATH_MAX]; @@ -1175,34 +1209,10 @@ int symbol__disassemble(struct symbol *sym, struct map *map, size_t privsize) int lineno = 0; int nline; pid_t pid; - int err = SYMBOL_ANNOTATE_ERRNO__NO_VMLINUX; + int err = dso__disassemble_filename(dso, symfs_filename, sizeof(symfs_filename)); - if (dso->symtab_type == DSO_BINARY_TYPE__KALLSYMS && - !dso__is_kcore(dso)) - goto out; - - filename = dso__build_id_filename(dso, NULL, 0); - if (filename) { - symbol__join_symfs(symfs_filename, filename); - free(filename); - } else { - if (dso->has_build_id) - return ENOMEM; - goto fallback; - } - - if (dso__is_kcore(dso) || - readlink(symfs_filename, command, sizeof(command)) < 0 || - strstr(command, DSO__NAME_KALLSYMS) || - access(symfs_filename, R_OK)) { -fallback: - /* - * If we don't have build-ids or the build-id file isn't in the - * cache, or is just a kallsyms file, well, lets hope that this - * DSO is the same as when 'perf record' ran. - */ - symbol__join_symfs(symfs_filename, dso->long_name); - } + if (err) + return err; pr_debug("%s: filename=%s, sym=%s, start=%#" PRIx64 ", end=%#" PRIx64 "\n", __func__, symfs_filename, sym->name, map->unmap_ip(map, sym->start), From 39ff526350059e61234d58676c13bcfcaac3a451 Mon Sep 17 00:00:00 2001 From: Mathieu Poirier Date: Thu, 11 Aug 2016 10:20:56 -0600 Subject: [PATCH 24/38] tools: Copy coresight-pmu.h header file needed by perf tools Directly accessing kernel files is not allowed anymore. As such making file coresight-pmu.h accessible by the perf tools and complain if this copy strays from the one found in the main kernel tree. Signed-off-by: Mathieu Poirier Cc: Jiri Olsa Cc: Peter Zijlstra Cc: Vince Weaver Link: http://lkml.kernel.org/r/1470932464-726-2-git-send-email-mathieu.poirier@linaro.org Signed-off-by: Arnaldo Carvalho de Melo --- tools/include/linux/coresight-pmu.h | 39 +++++++++++++++++++++++++++++ tools/perf/MANIFEST | 1 + tools/perf/Makefile.perf | 3 +++ 3 files changed, 43 insertions(+) create mode 100644 tools/include/linux/coresight-pmu.h diff --git a/tools/include/linux/coresight-pmu.h b/tools/include/linux/coresight-pmu.h new file mode 100644 index 000000000000..7d410260661b --- /dev/null +++ b/tools/include/linux/coresight-pmu.h @@ -0,0 +1,39 @@ +/* + * Copyright(C) 2015 Linaro Limited. All rights reserved. + * Author: Mathieu Poirier + * + * This program is free software; you can redistribute it and/or modify it + * under the terms of the GNU General Public License version 2 as published by + * the Free Software Foundation. + * + * This program is distributed in the hope that it will be useful, but WITHOUT + * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or + * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for + * more details. + * + * You should have received a copy of the GNU General Public License along with + * this program. If not, see . + */ + +#ifndef _LINUX_CORESIGHT_PMU_H +#define _LINUX_CORESIGHT_PMU_H + +#define CORESIGHT_ETM_PMU_NAME "cs_etm" +#define CORESIGHT_ETM_PMU_SEED 0x10 + +/* ETMv3.5/PTM's ETMCR config bit */ +#define ETM_OPT_CYCACC 12 +#define ETM_OPT_TS 28 + +static inline int coresight_get_trace_id(int cpu) +{ + /* + * A trace ID of value 0 is invalid, so let's start at some + * random value that fits in 7 bits and go from there. Since + * the common convention is to have data trace IDs be I(N) + 1, + * set instruction trace IDs as a function of the CPU number. + */ + return (CORESIGHT_ETM_PMU_SEED + (cpu * 2)); +} + +#endif diff --git a/tools/perf/MANIFEST b/tools/perf/MANIFEST index f23a5e72aaa5..ff200c6cb790 100644 --- a/tools/perf/MANIFEST +++ b/tools/perf/MANIFEST @@ -60,6 +60,7 @@ tools/include/asm-generic/bitops.h tools/include/linux/atomic.h tools/include/linux/bitops.h tools/include/linux/compiler.h +tools/include/linux/coresight-pmu.h tools/include/linux/filter.h tools/include/linux/hash.h tools/include/linux/kernel.h diff --git a/tools/perf/Makefile.perf b/tools/perf/Makefile.perf index 2d9087501633..aa7ab23fbc9f 100644 --- a/tools/perf/Makefile.perf +++ b/tools/perf/Makefile.perf @@ -429,6 +429,9 @@ $(PERF_IN): prepare FORCE @(test -f ../../include/asm-generic/bitops/fls64.h && ( \ (diff -B ../include/asm-generic/bitops/fls64.h ../../include/asm-generic/bitops/fls64.h >/dev/null) \ || echo "Warning: tools/include/asm-generic/bitops/fls64.h differs from kernel" >&2 )) || true + @(test -f ../../include/linux/coresight-pmu.h && ( \ + (diff -B ../include/linux/coresight-pmu.h ../../include/linux/coresight-pmu.h >/dev/null) \ + || echo "Warning: tools/include/linux/coresight-pmu.h differs from kernel" >&2 )) || true $(Q)$(MAKE) $(build)=perf $(OUTPUT)perf: $(PERFLIBS) $(PERF_IN) $(LIBTRACEEVENT_DYNAMIC_LIST) From fa1f456592347c6f40c9d37ea407b029fda5324a Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Fri, 12 Aug 2016 20:41:01 -0300 Subject: [PATCH 25/38] perf report: Allow configuring the default sort order in ~/.perfconfig Allows changing the default sort order from "comm,dso,symbol" to some other default, for instance "sym,dso" may be more fitting for kernel developers. Cc: Adrian Hunter Cc: David Ahern Cc: Jiri Olsa Cc: Linus Torvalds Cc: Namhyung Kim Cc: Wang Nan Link: http://lkml.kernel.org/n/tip-pm1h5puxua8nsxksd68fjm8r@git.kernel.org Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/Documentation/perf-config.txt | 4 ++++ tools/perf/builtin-report.c | 4 ++++ tools/perf/util/sort.c | 2 +- tools/perf/util/sort.h | 2 +- 4 files changed, 10 insertions(+), 2 deletions(-) diff --git a/tools/perf/Documentation/perf-config.txt b/tools/perf/Documentation/perf-config.txt index 15949e2a7805..68c89190e79c 100644 --- a/tools/perf/Documentation/perf-config.txt +++ b/tools/perf/Documentation/perf-config.txt @@ -382,6 +382,10 @@ call-graph.*:: histogram entry. Default is 0 which means no limitation. report.*:: + report.sort_order:: + Allows changing the default sort order from "comm,dso,symbol" to + some other default, for instance "sym,dso" may be more fitting for + kernel developers. report.percent-limit:: This one is mostly the same as call-graph.threshold but works for histogram entries. Entries having an overhead lower than this diff --git a/tools/perf/builtin-report.c b/tools/perf/builtin-report.c index 949e5a15c960..b9e046baa5fc 100644 --- a/tools/perf/builtin-report.c +++ b/tools/perf/builtin-report.c @@ -89,6 +89,10 @@ static int report__config(const char *var, const char *value, void *cb) rep->queue_size = perf_config_u64(var, value); return 0; } + if (!strcmp(var, "report.sort_order")) { + default_sort_order = strdup(value); + return 0; + } return 0; } diff --git a/tools/perf/util/sort.c b/tools/perf/util/sort.c index 5ae3d32a21d6..1884d7f9b9d2 100644 --- a/tools/perf/util/sort.c +++ b/tools/perf/util/sort.c @@ -11,7 +11,7 @@ regex_t parent_regex; const char default_parent_pattern[] = "^sys_|^do_page_fault"; const char *parent_pattern = default_parent_pattern; -const char default_sort_order[] = "comm,dso,symbol"; +const char *default_sort_order = "comm,dso,symbol"; const char default_branch_sort_order[] = "comm,dso_from,symbol_from,symbol_to,cycles"; const char default_mem_sort_order[] = "local_weight,mem,sym,dso,symbol_daddr,dso_daddr,snoop,tlb,locked"; const char default_top_sort_order[] = "dso,symbol"; diff --git a/tools/perf/util/sort.h b/tools/perf/util/sort.h index 7ca37ea17395..28c0524c8702 100644 --- a/tools/perf/util/sort.h +++ b/tools/perf/util/sort.h @@ -28,7 +28,7 @@ extern const char *sort_order; extern const char *field_order; extern const char default_parent_pattern[]; extern const char *parent_pattern; -extern const char default_sort_order[]; +extern const char *default_sort_order; extern regex_t ignore_callees_regex; extern int have_ignore_callees; extern enum sort_mode sort__mode; From 11d8f870c8eb8dd41ade688fbad6946ad69243fe Mon Sep 17 00:00:00 2001 From: Rui Teng Date: Thu, 28 Jul 2016 10:05:57 +0800 Subject: [PATCH 26/38] perf tools: Use __weak definition from linux/compiler.h Replace __attribute__((weak)) with __weak definition Signed-off-by: Rui Teng Cc: Alexander Shishkin Cc: Peter Zijlstra Link: http://lkml.kernel.org/r/1469671557-2256-2-git-send-email-rui.teng@linux.vnet.ibm.com Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/util/header.c | 3 +-- 1 file changed, 1 insertion(+), 2 deletions(-) diff --git a/tools/perf/util/header.c b/tools/perf/util/header.c index 8f0db4007282..85dd0db0a127 100644 --- a/tools/perf/util/header.c +++ b/tools/perf/util/header.c @@ -828,8 +828,7 @@ static int write_group_desc(int fd, struct perf_header *h __maybe_unused, * default get_cpuid(): nothing gets recorded * actual implementation must be in arch/$(ARCH)/util/header.c */ -int __attribute__ ((weak)) get_cpuid(char *buffer __maybe_unused, - size_t sz __maybe_unused) +int __weak get_cpuid(char *buffer __maybe_unused, size_t sz __maybe_unused) { return -1; } From 11196b79164c95b8e5968ff63fa6b59536c9748f Mon Sep 17 00:00:00 2001 From: Rui Teng Date: Wed, 10 Aug 2016 16:49:08 +0800 Subject: [PATCH 27/38] perf tools: Skip running the feature tests for 'make install-doc' It is a requirement from the perf TODO list[1]: ''The feature tests should be performed only when a file that needs those tests, or at least only when some .c or .h file will be rebuilt. An initial step would be for 'make install-doc' not to run the feature tests, there it is not needed at all.'' By adding 'install-doc' to the NON_CONFIG_TARGETS, it will skip running the feature tests for such target. The Auto-detecting system features list will not be displayed: $ make install-doc BUILD: Doing 'make -j2' parallel build SUBDIR Documentation make[2]: Nothing to be done for 'install'. [1] https://perf.wiki.kernel.org/index.php/Todo Signed-off-by: Rui Teng Tested-by: Arnaldo Carvalho de Melo Cc: Alexander Shishkin Cc: Peter Zijlstra Link: http://lkml.kernel.org/r/1470818948-17784-1-git-send-email-rui.teng@linux.vnet.ibm.com Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/Makefile.perf | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/tools/perf/Makefile.perf b/tools/perf/Makefile.perf index aa7ab23fbc9f..828cfd766e31 100644 --- a/tools/perf/Makefile.perf +++ b/tools/perf/Makefile.perf @@ -165,7 +165,7 @@ SUBCMD_DIR = $(srctree)/tools/lib/subcmd/ # non-config cases config := 1 -NON_CONFIG_TARGETS := clean TAGS tags cscope help +NON_CONFIG_TARGETS := clean TAGS tags cscope help install-doc ifdef MAKECMDGOALS ifeq ($(filter-out $(NON_CONFIG_TARGETS),$(MAKECMDGOALS)),) From 6637e6f1ac62e62a353612844db21cff9a17758b Mon Sep 17 00:00:00 2001 From: Colin Ian King Date: Mon, 15 Aug 2016 17:25:43 +0100 Subject: [PATCH 28/38] perf hists browser: Remove superfluous null check on map 'map' is being already checked if it is NULL at the start of do_zoom_dso(), so the second subsequent check is superfluous and can be removed. Signed-off-by: Colin King Acked-by: Namhyung Kim Cc: Alexander Shishkin Cc: Jiri Olsa Cc: Pekka Enberg Cc: Peter Zijlstra Cc: Wang Nan Link: http://lkml.kernel.org/r/1471278343-14999-1-git-send-email-colin.king@canonical.com Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/ui/browsers/hists.c | 2 -- 1 file changed, 2 deletions(-) diff --git a/tools/perf/ui/browsers/hists.c b/tools/perf/ui/browsers/hists.c index 22e48a3ebe29..f0611c937d4b 100644 --- a/tools/perf/ui/browsers/hists.c +++ b/tools/perf/ui/browsers/hists.c @@ -2440,8 +2440,6 @@ do_zoom_dso(struct hist_browser *browser, struct popup_action *act) browser->hists->dso_filter = NULL; ui_helpline__pop(); } else { - if (map == NULL) - return 0; ui_helpline__fpush("To zoom out press ESC or ENTER + \"Zoom out of %s DSO\"", __map__is_kernel(map) ? "the Kernel" : map->dso->short_name); browser->hists->dso_filter = map->dso; From 17ce3dc7e5a0e4796cc7838d1f7b2531d0bca130 Mon Sep 17 00:00:00 2001 From: Masami Hiramatsu Date: Thu, 18 Aug 2016 17:57:50 +0900 Subject: [PATCH 29/38] ftrace: kprobe: uprobe: Add x8/x16/x32/x64 for hexadecimal types Add x8/x16/x32/x64 for hexadecimal type casting to kprobe/uprobe event tracer. These type casts can be used for integer arguments for explicitly showing them in hexadecimal digits in formatted text. Signed-off-by: Masami Hiramatsu Acked-by: Steven Rostedt Cc: Alexander Shishkin Cc: Hemant Kumar Cc: Naohiro Aota Cc: Peter Zijlstra Cc: Wang Nan Link: http://lkml.kernel.org/r/147151067029.12957.11591314629326414783.stgit@devbox Signed-off-by: Arnaldo Carvalho de Melo --- Documentation/trace/kprobetrace.txt | 4 ++-- Documentation/trace/uprobetracer.txt | 4 ++-- kernel/trace/trace_kprobe.c | 4 ++++ kernel/trace/trace_probe.c | 28 ++++++++++++++----------- kernel/trace/trace_probe.h | 9 ++++++++ kernel/trace/trace_uprobe.c | 4 ++++ tools/perf/Documentation/perf-probe.txt | 4 ++-- 7 files changed, 39 insertions(+), 18 deletions(-) diff --git a/Documentation/trace/kprobetrace.txt b/Documentation/trace/kprobetrace.txt index ea52ec1f8484..9109c8e8a5a1 100644 --- a/Documentation/trace/kprobetrace.txt +++ b/Documentation/trace/kprobetrace.txt @@ -44,8 +44,8 @@ Synopsis of kprobe_events +|-offs(FETCHARG) : Fetch memory at FETCHARG +|- offs address.(**) NAME=FETCHARG : Set NAME as the argument name of FETCHARG. FETCHARG:TYPE : Set TYPE as the type of FETCHARG. Currently, basic types - (u8/u16/u32/u64/s8/s16/s32/s64), "string" and bitfield - are supported. + (u8/u16/u32/u64/s8/s16/s32/s64), hexadecimal types + (x8/x16/x32/x64), "string" and bitfield are supported. (*) only for return probe. (**) this is useful for fetching a field of data structures. diff --git a/Documentation/trace/uprobetracer.txt b/Documentation/trace/uprobetracer.txt index 72d1cd4f7bf3..7e6d28c54b75 100644 --- a/Documentation/trace/uprobetracer.txt +++ b/Documentation/trace/uprobetracer.txt @@ -40,8 +40,8 @@ Synopsis of uprobe_tracer +|-offs(FETCHARG) : Fetch memory at FETCHARG +|- offs address.(**) NAME=FETCHARG : Set NAME as the argument name of FETCHARG. FETCHARG:TYPE : Set TYPE as the type of FETCHARG. Currently, basic types - (u8/u16/u32/u64/s8/s16/s32/s64), "string" and bitfield - are supported. + (u8/u16/u32/u64/s8/s16/s32/s64), hexadecimal types + (x8/x16/x32/x64), "string" and bitfield are supported. (*) only for return probe. (**) this is useful for fetching a field of data structures. diff --git a/kernel/trace/trace_kprobe.c b/kernel/trace/trace_kprobe.c index 9aedb0b06683..eb6c9f1d3a93 100644 --- a/kernel/trace/trace_kprobe.c +++ b/kernel/trace/trace_kprobe.c @@ -253,6 +253,10 @@ static const struct fetch_type kprobes_fetch_type_table[] = { ASSIGN_FETCH_TYPE(s16, u16, 1), ASSIGN_FETCH_TYPE(s32, u32, 1), ASSIGN_FETCH_TYPE(s64, u64, 1), + ASSIGN_FETCH_TYPE_ALIAS(x8, u8, u8, 0), + ASSIGN_FETCH_TYPE_ALIAS(x16, u16, u16, 0), + ASSIGN_FETCH_TYPE_ALIAS(x32, u32, u32, 0), + ASSIGN_FETCH_TYPE_ALIAS(x64, u64, u64, 0), ASSIGN_FETCH_TYPE_END }; diff --git a/kernel/trace/trace_probe.c b/kernel/trace/trace_probe.c index 74e80a582c28..725af9dcbdff 100644 --- a/kernel/trace/trace_probe.c +++ b/kernel/trace/trace_probe.c @@ -36,24 +36,28 @@ const char *reserved_field_names[] = { }; /* Printing in basic type function template */ -#define DEFINE_BASIC_PRINT_TYPE_FUNC(type, fmt) \ -int PRINT_TYPE_FUNC_NAME(type)(struct trace_seq *s, const char *name, \ +#define DEFINE_BASIC_PRINT_TYPE_FUNC(tname, type, fmt) \ +int PRINT_TYPE_FUNC_NAME(tname)(struct trace_seq *s, const char *name, \ void *data, void *ent) \ { \ trace_seq_printf(s, " %s=" fmt, name, *(type *)data); \ return !trace_seq_has_overflowed(s); \ } \ -const char PRINT_TYPE_FMT_NAME(type)[] = fmt; \ -NOKPROBE_SYMBOL(PRINT_TYPE_FUNC_NAME(type)); +const char PRINT_TYPE_FMT_NAME(tname)[] = fmt; \ +NOKPROBE_SYMBOL(PRINT_TYPE_FUNC_NAME(tname)); -DEFINE_BASIC_PRINT_TYPE_FUNC(u8 , "0x%x") -DEFINE_BASIC_PRINT_TYPE_FUNC(u16, "0x%x") -DEFINE_BASIC_PRINT_TYPE_FUNC(u32, "0x%x") -DEFINE_BASIC_PRINT_TYPE_FUNC(u64, "0x%Lx") -DEFINE_BASIC_PRINT_TYPE_FUNC(s8, "%d") -DEFINE_BASIC_PRINT_TYPE_FUNC(s16, "%d") -DEFINE_BASIC_PRINT_TYPE_FUNC(s32, "%d") -DEFINE_BASIC_PRINT_TYPE_FUNC(s64, "%Ld") +DEFINE_BASIC_PRINT_TYPE_FUNC(u8, u8, "0x%x") +DEFINE_BASIC_PRINT_TYPE_FUNC(u16, u16, "0x%x") +DEFINE_BASIC_PRINT_TYPE_FUNC(u32, u32, "0x%x") +DEFINE_BASIC_PRINT_TYPE_FUNC(u64, u64, "0x%Lx") +DEFINE_BASIC_PRINT_TYPE_FUNC(s8, s8, "%d") +DEFINE_BASIC_PRINT_TYPE_FUNC(s16, s16, "%d") +DEFINE_BASIC_PRINT_TYPE_FUNC(s32, s32, "%d") +DEFINE_BASIC_PRINT_TYPE_FUNC(s64, s64, "%Ld") +DEFINE_BASIC_PRINT_TYPE_FUNC(x8, u8, "0x%x") +DEFINE_BASIC_PRINT_TYPE_FUNC(x16, u16, "0x%x") +DEFINE_BASIC_PRINT_TYPE_FUNC(x32, u32, "0x%x") +DEFINE_BASIC_PRINT_TYPE_FUNC(x64, u64, "0x%Lx") /* Print type function for string type */ int PRINT_TYPE_FUNC_NAME(string)(struct trace_seq *s, const char *name, diff --git a/kernel/trace/trace_probe.h b/kernel/trace/trace_probe.h index 45400ca5ded1..f0c470a10edd 100644 --- a/kernel/trace/trace_probe.h +++ b/kernel/trace/trace_probe.h @@ -149,6 +149,11 @@ DECLARE_BASIC_PRINT_TYPE_FUNC(s8); DECLARE_BASIC_PRINT_TYPE_FUNC(s16); DECLARE_BASIC_PRINT_TYPE_FUNC(s32); DECLARE_BASIC_PRINT_TYPE_FUNC(s64); +DECLARE_BASIC_PRINT_TYPE_FUNC(x8); +DECLARE_BASIC_PRINT_TYPE_FUNC(x16); +DECLARE_BASIC_PRINT_TYPE_FUNC(x32); +DECLARE_BASIC_PRINT_TYPE_FUNC(x64); + DECLARE_BASIC_PRINT_TYPE_FUNC(string); #define FETCH_FUNC_NAME(method, type) fetch_##method##_##type @@ -234,6 +239,10 @@ ASSIGN_FETCH_FUNC(file_offset, ftype), \ #define ASSIGN_FETCH_TYPE(ptype, ftype, sign) \ __ASSIGN_FETCH_TYPE(#ptype, ptype, ftype, sizeof(ftype), sign, #ptype) +/* If ptype is an alias of atype, use this macro (show atype in format) */ +#define ASSIGN_FETCH_TYPE_ALIAS(ptype, atype, ftype, sign) \ + __ASSIGN_FETCH_TYPE(#ptype, ptype, ftype, sizeof(ftype), sign, #atype) + #define ASSIGN_FETCH_TYPE_END {} #define FETCH_TYPE_STRING 0 diff --git a/kernel/trace/trace_uprobe.c b/kernel/trace/trace_uprobe.c index c53485441c88..7a687320f867 100644 --- a/kernel/trace/trace_uprobe.c +++ b/kernel/trace/trace_uprobe.c @@ -211,6 +211,10 @@ static const struct fetch_type uprobes_fetch_type_table[] = { ASSIGN_FETCH_TYPE(s16, u16, 1), ASSIGN_FETCH_TYPE(s32, u32, 1), ASSIGN_FETCH_TYPE(s64, u64, 1), + ASSIGN_FETCH_TYPE_ALIAS(x8, u8, u8, 0), + ASSIGN_FETCH_TYPE_ALIAS(x16, u16, u16, 0), + ASSIGN_FETCH_TYPE_ALIAS(x32, u32, u32, 0), + ASSIGN_FETCH_TYPE_ALIAS(x64, u64, u64, 0), ASSIGN_FETCH_TYPE_END }; diff --git a/tools/perf/Documentation/perf-probe.txt b/tools/perf/Documentation/perf-probe.txt index b303bcdd8ed1..aa333258b1b7 100644 --- a/tools/perf/Documentation/perf-probe.txt +++ b/tools/perf/Documentation/perf-probe.txt @@ -176,13 +176,13 @@ Each probe argument follows below syntax. 'NAME' specifies the name of this argument (optional). You can use the name of local variable, local data structure member (e.g. var->field, var.field2), local array with fixed index (e.g. array[1], var->array[0], var->pointer[2]), or kprobe-tracer argument format (e.g. $retval, %ax, etc). Note that the name of this argument will be set as the last member name if you specify a local data structure member (e.g. field2 for 'var->field1.field2'.) '$vars' and '$params' special arguments are also available for NAME, '$vars' is expanded to the local variables (including function parameters) which can access at given probe point. '$params' is expanded to only the function parameters. -'TYPE' casts the type of this argument (optional). If omitted, perf probe automatically set the type based on debuginfo. Currently, basic types (u8/u16/u32/u64/s8/s16/s32/s64), signedness casting (u/s), "string" and bitfield are supported. (see TYPES for detail) +'TYPE' casts the type of this argument (optional). If omitted, perf probe automatically set the type based on debuginfo. Currently, basic types (u8/u16/u32/u64/s8/s16/s32/s64), hexadecimal integers (x8/x16/x32/x64), signedness casting (u/s), "string" and bitfield are supported. (see TYPES for detail) On x86 systems %REG is always the short form of the register: for example %AX. %RAX or %EAX is not valid. TYPES ----- -Basic types (u8/u16/u32/u64/s8/s16/s32/s64) are integer types. Prefix 's' and 'u' means those types are signed and unsigned respectively. Traced arguments are shown in decimal (signed) or hex (unsigned). You can also use 's' or 'u' to specify only signedness and leave its size auto-detected by perf probe. +Basic types (u8/u16/u32/u64/s8/s16/s32/s64) and hexadecimal integers (x8/x16/x32/x64) are integer types. Prefix 's' and 'u' means those types are signed and unsigned respectively, and 'x' means that is shown in hexadecimal format. Traced arguments are shown in decimal (signed) or hex (unsigned). You can also use 's' or 'u' to specify only signedness and leave its size auto-detected by perf probe. String type is a special type, which fetches a "null-terminated" string from kernel space. This means it will fail and store NULL if the string container has been paged out. You can specify 'string' type only for the local variable or structure member which is an array of or a pointer to 'char' or 'unsigned char' type. Bitfield is another special type, which takes 3 parameters, bit-width, bit-offset, and container-size (usually 32). The syntax is; From 864256255597aad86abcecbe6c53da8852ded15b Mon Sep 17 00:00:00 2001 From: Masami Hiramatsu Date: Thu, 18 Aug 2016 17:58:15 +0900 Subject: [PATCH 30/38] ftrace: probe: Add README entries for k/uprobe-events Add README entries for kprobe-events and uprobe-events. This allows user to check what options can be acceptable for running kernel. E.g. perf tools can choose correct types for the kernel. Signed-off-by: Masami Hiramatsu Acked-by: Steven Rostedt Cc: Alexander Shishkin Cc: Hemant Kumar Cc: Naohiro Aota Cc: Peter Zijlstra Cc: Wang Nan Link: http://lkml.kernel.org/r/147151069524.12957.12957179170304055028.stgit@devbox Signed-off-by: Arnaldo Carvalho de Melo --- kernel/trace/trace.c | 24 ++++++++++++++++++++++++ 1 file changed, 24 insertions(+) diff --git a/kernel/trace/trace.c b/kernel/trace/trace.c index dade4c9559cc..1e2ce3b52e51 100644 --- a/kernel/trace/trace.c +++ b/kernel/trace/trace.c @@ -4123,6 +4123,30 @@ static const char readme_msg[] = "\t\t\t traces\n" #endif #endif /* CONFIG_STACK_TRACER */ +#ifdef CONFIG_KPROBE_EVENT + " kprobe_events\t\t- Add/remove/show the kernel dynamic events\n" + "\t\t\t Write into this file to define/undefine new trace events.\n" +#endif +#ifdef CONFIG_UPROBE_EVENT + " uprobe_events\t\t- Add/remove/show the userspace dynamic events\n" + "\t\t\t Write into this file to define/undefine new trace events.\n" +#endif +#if defined(CONFIG_KPROBE_EVENT) || defined(CONFIG_UPROBE_EVENT) + "\t accepts: event-definitions (one definition per line)\n" + "\t Format: p|r[:[/]] []\n" + "\t -:[/]\n" +#ifdef CONFIG_KPROBE_EVENT + "\t place: [:][+]|\n" +#endif +#ifdef CONFIG_UPROBE_EVENT + "\t place: :\n" +#endif + "\t args: =fetcharg[:type]\n" + "\t fetcharg: %, @
, @[+|-],\n" + "\t $stack, $stack, $retval, $comm\n" + "\t type: s8/16/32/64, u8/16/32/64, x8/16/32/64, string,\n" + "\t b@/\n" +#endif " events/\t\t- Directory containing all trace event subsystems:\n" " enable\t\t- Write 0/1 to enable/disable tracing of all events\n" " events//\t- Directory containing all trace events for :\n" From 180b20616ce57e93eb692170c793be94c456b1e2 Mon Sep 17 00:00:00 2001 From: Masami Hiramatsu Date: Thu, 18 Aug 2016 17:58:31 +0900 Subject: [PATCH 31/38] perf probe: Add supported for type casting by the running kernel Add a checking routine what types are supported by the running kernel by finding the pattern in /tracing/README. Signed-off-by: Masami Hiramatsu Cc: Alexander Shishkin Cc: Hemant Kumar Cc: Naohiro Aota Cc: Peter Zijlstra Cc: Steven Rostedt Cc: Wang Nan Link: http://lkml.kernel.org/r/147151071172.12957.3340095690753291085.stgit@devbox [ 'enum probe_type' has no negative entries, so ends up as 'unsigned', remove '< 0' test to fix the build on at least centos:5, debian:7 & ubuntu:12.04.5 ] Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/util/probe-file.c | 57 ++++++++++++++++++++++++++++++++++ tools/perf/util/probe-file.h | 10 ++++++ tools/perf/util/probe-finder.c | 1 + 3 files changed, 68 insertions(+) diff --git a/tools/perf/util/probe-file.c b/tools/perf/util/probe-file.c index 9c3b9ed5b3c3..697ef66bff91 100644 --- a/tools/perf/util/probe-file.c +++ b/tools/perf/util/probe-file.c @@ -877,3 +877,60 @@ int probe_cache__show_all_caches(struct strfilter *filter) return 0; } + +static struct { + const char *pattern; + bool avail; + bool checked; +} probe_type_table[] = { +#define DEFINE_TYPE(idx, pat, def_avail) \ + [idx] = {.pattern = pat, .avail = (def_avail)} + DEFINE_TYPE(PROBE_TYPE_U, "* u8/16/32/64,*", true), + DEFINE_TYPE(PROBE_TYPE_S, "* s8/16/32/64,*", true), + DEFINE_TYPE(PROBE_TYPE_X, "* x8/16/32/64,*", false), + DEFINE_TYPE(PROBE_TYPE_STRING, "* string,*", true), + DEFINE_TYPE(PROBE_TYPE_BITFIELD, + "* b@/", true), +}; + +bool probe_type_is_available(enum probe_type type) +{ + FILE *fp; + char *buf = NULL; + size_t len = 0; + bool target_line = false; + bool ret = probe_type_table[type].avail; + + if (type >= PROBE_TYPE_END) + return false; + /* We don't have to check the type which supported by default */ + if (ret || probe_type_table[type].checked) + return ret; + + if (asprintf(&buf, "%s/README", tracing_path) < 0) + return ret; + + fp = fopen(buf, "r"); + if (!fp) + goto end; + + zfree(&buf); + while (getline(&buf, &len, fp) > 0 && !ret) { + if (!target_line) { + target_line = !!strstr(buf, " type: "); + if (!target_line) + continue; + } else if (strstr(buf, "\t ") != buf) + break; + ret = strglobmatch(buf, probe_type_table[type].pattern); + } + /* Cache the result */ + probe_type_table[type].checked = true; + probe_type_table[type].avail = ret; + + fclose(fp); +end: + free(buf); + + return ret; +} diff --git a/tools/perf/util/probe-file.h b/tools/perf/util/probe-file.h index 9577b5c0b487..eba44c3e9dca 100644 --- a/tools/perf/util/probe-file.h +++ b/tools/perf/util/probe-file.h @@ -19,6 +19,15 @@ struct probe_cache { struct list_head entries; }; +enum probe_type { + PROBE_TYPE_U = 0, + PROBE_TYPE_S, + PROBE_TYPE_X, + PROBE_TYPE_STRING, + PROBE_TYPE_BITFIELD, + PROBE_TYPE_END, +}; + #define PF_FL_UPROBE 1 #define PF_FL_RW 2 #define for_each_probe_cache_entry(entry, pcache) \ @@ -54,6 +63,7 @@ struct probe_cache_entry *probe_cache__find(struct probe_cache *pcache, struct probe_cache_entry *probe_cache__find_by_name(struct probe_cache *pcache, const char *group, const char *event); int probe_cache__show_all_caches(struct strfilter *filter); +bool probe_type_is_available(enum probe_type type); #else /* ! HAVE_LIBELF_SUPPORT */ static inline struct probe_cache *probe_cache__new(const char *tgt __maybe_unused) { diff --git a/tools/perf/util/probe-finder.c b/tools/perf/util/probe-finder.c index 5c290c682afe..24dbe23e010c 100644 --- a/tools/perf/util/probe-finder.c +++ b/tools/perf/util/probe-finder.c @@ -39,6 +39,7 @@ #include "util.h" #include "symbol.h" #include "probe-finder.h" +#include "probe-file.h" /* Kprobe tracer basic type is up to u64 */ #define MAX_BASIC_TYPE_BITS 64 From 925437872525ee229736a9a8bdf804fc98f75b44 Mon Sep 17 00:00:00 2001 From: Masami Hiramatsu Date: Thu, 18 Aug 2016 17:58:47 +0900 Subject: [PATCH 32/38] perf probe: Support hexadecimal casting Support hexadecimal unsigned integer casting by 'x'. This allows user to explicitly specify the output format of the probe arguments as hexadecimal. Signed-off-by: Masami Hiramatsu Cc: Alexander Shishkin Cc: Hemant Kumar Cc: Naohiro Aota Cc: Peter Zijlstra Cc: Steven Rostedt Cc: Wang Nan Link: http://lkml.kernel.org/r/147151072679.12957.4458656416765710753.stgit@devbox Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/Documentation/perf-probe.txt | 4 ++-- tools/perf/util/probe-finder.c | 17 ++++++++++------- 2 files changed, 12 insertions(+), 9 deletions(-) diff --git a/tools/perf/Documentation/perf-probe.txt b/tools/perf/Documentation/perf-probe.txt index aa333258b1b7..d217617108a1 100644 --- a/tools/perf/Documentation/perf-probe.txt +++ b/tools/perf/Documentation/perf-probe.txt @@ -176,13 +176,13 @@ Each probe argument follows below syntax. 'NAME' specifies the name of this argument (optional). You can use the name of local variable, local data structure member (e.g. var->field, var.field2), local array with fixed index (e.g. array[1], var->array[0], var->pointer[2]), or kprobe-tracer argument format (e.g. $retval, %ax, etc). Note that the name of this argument will be set as the last member name if you specify a local data structure member (e.g. field2 for 'var->field1.field2'.) '$vars' and '$params' special arguments are also available for NAME, '$vars' is expanded to the local variables (including function parameters) which can access at given probe point. '$params' is expanded to only the function parameters. -'TYPE' casts the type of this argument (optional). If omitted, perf probe automatically set the type based on debuginfo. Currently, basic types (u8/u16/u32/u64/s8/s16/s32/s64), hexadecimal integers (x8/x16/x32/x64), signedness casting (u/s), "string" and bitfield are supported. (see TYPES for detail) +'TYPE' casts the type of this argument (optional). If omitted, perf probe automatically set the type based on debuginfo. Currently, basic types (u8/u16/u32/u64/s8/s16/s32/s64), hexadecimal integers (x/x8/x16/x32/x64), signedness casting (u/s), "string" and bitfield are supported. (see TYPES for detail) On x86 systems %REG is always the short form of the register: for example %AX. %RAX or %EAX is not valid. TYPES ----- -Basic types (u8/u16/u32/u64/s8/s16/s32/s64) and hexadecimal integers (x8/x16/x32/x64) are integer types. Prefix 's' and 'u' means those types are signed and unsigned respectively, and 'x' means that is shown in hexadecimal format. Traced arguments are shown in decimal (signed) or hex (unsigned). You can also use 's' or 'u' to specify only signedness and leave its size auto-detected by perf probe. +Basic types (u8/u16/u32/u64/s8/s16/s32/s64) and hexadecimal integers (x8/x16/x32/x64) are integer types. Prefix 's' and 'u' means those types are signed and unsigned respectively, and 'x' means that is shown in hexadecimal format. Traced arguments are shown in decimal (signed) or hex (unsigned). You can also use 's' or 'u' to specify only signedness and leave its size auto-detected by perf probe. Moreover, you can use 'x' to explicitly specify to be shown in hexadecimal (the size is also auto-detected). String type is a special type, which fetches a "null-terminated" string from kernel space. This means it will fail and store NULL if the string container has been paged out. You can specify 'string' type only for the local variable or structure member which is an array of or a pointer to 'char' or 'unsigned char' type. Bitfield is another special type, which takes 3 parameters, bit-width, bit-offset, and container-size (usually 32). The syntax is; diff --git a/tools/perf/util/probe-finder.c b/tools/perf/util/probe-finder.c index 24dbe23e010c..f18cd6bbada9 100644 --- a/tools/perf/util/probe-finder.c +++ b/tools/perf/util/probe-finder.c @@ -298,13 +298,13 @@ static int convert_variable_type(Dwarf_Die *vr_die, char sbuf[STRERR_BUFSIZE]; int bsize, boffs, total; int ret; - char sign; + char prefix; /* TODO: check all types */ - if (cast && strcmp(cast, "string") != 0 && + if (cast && strcmp(cast, "string") != 0 && strcmp(cast, "x") != 0 && strcmp(cast, "s") != 0 && strcmp(cast, "u") != 0) { /* Non string type is OK */ - /* and respect signedness cast */ + /* and respect signedness/hexadecimal cast */ tvar->type = strdup(cast); return (tvar->type == NULL) ? -ENOMEM : 0; } @@ -366,11 +366,14 @@ static int convert_variable_type(Dwarf_Die *vr_die, } if (cast && (strcmp(cast, "u") == 0)) - sign = 'u'; + prefix = 'u'; else if (cast && (strcmp(cast, "s") == 0)) - sign = 's'; + prefix = 's'; + else if (cast && (strcmp(cast, "x") == 0) && + probe_type_is_available(PROBE_TYPE_X)) + prefix = 'x'; else - sign = die_is_signed_type(&type) ? 's' : 'u'; + prefix = die_is_signed_type(&type) ? 's' : 'u'; ret = dwarf_bytesize(&type); if (ret <= 0) @@ -384,7 +387,7 @@ static int convert_variable_type(Dwarf_Die *vr_die, dwarf_diename(&type), MAX_BASIC_TYPE_BITS); ret = MAX_BASIC_TYPE_BITS; } - ret = snprintf(buf, 16, "%c%d", sign, ret); + ret = snprintf(buf, 16, "%c%d", prefix, ret); formatted: if (ret < 0 || ret >= 16) { From 9880ce4a69ba5c66a5ffdd711fe446bd0226bd8c Mon Sep 17 00:00:00 2001 From: Masami Hiramatsu Date: Thu, 18 Aug 2016 17:59:07 +0900 Subject: [PATCH 33/38] perf probe: Use hexadecimal type by default if possible Use hexadecimal type by default if it is available on current running kernel. This keeps the default behavior of perf probe after changing the output format of 'u8/16/32/64' to unsigned decimal number. Signed-off-by: Masami Hiramatsu Cc: Alexander Shishkin Cc: Hemant Kumar Cc: Naohiro Aota Cc: Peter Zijlstra Cc: Steven Rostedt Cc: Wang Nan Link: http://lkml.kernel.org/r/147151074685.12957.16415861010796255514.stgit@devbox Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/util/probe-finder.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) diff --git a/tools/perf/util/probe-finder.c b/tools/perf/util/probe-finder.c index f18cd6bbada9..ac4740f8ee3a 100644 --- a/tools/perf/util/probe-finder.c +++ b/tools/perf/util/probe-finder.c @@ -373,7 +373,8 @@ static int convert_variable_type(Dwarf_Die *vr_die, probe_type_is_available(PROBE_TYPE_X)) prefix = 'x'; else - prefix = die_is_signed_type(&type) ? 's' : 'u'; + prefix = die_is_signed_type(&type) ? 's' : + probe_type_is_available(PROBE_TYPE_X) ? 'x' : 'u'; ret = dwarf_bytesize(&type); if (ret <= 0) From bdca79c2bf40556b664c9b1c32aec103e9bdb4a9 Mon Sep 17 00:00:00 2001 From: Masami Hiramatsu Date: Thu, 18 Aug 2016 17:59:21 +0900 Subject: [PATCH 34/38] ftrace: kprobe: uprobe: Show u8/u16/u32/u64 types in decimal Change kprobe/uprobe-tracer to show the arguments type-casted with u8/u16/u32/u64 in decimal digits instead of hexadecimal. To minimize compatibility issue, the arguments without type casting are typed by x64 (or x32 for 32bit arch) by default. Note: all arguments set by old perf probe without types are shown in decimal by default. Signed-off-by: Masami Hiramatsu Acked-by: Steven Rostedt Cc: Alexander Shishkin Cc: Hemant Kumar Cc: Naohiro Aota Cc: Peter Zijlstra Cc: Wang Nan Link: http://lkml.kernel.org/r/147151076135.12957.14684546093034343894.stgit@devbox Signed-off-by: Arnaldo Carvalho de Melo --- Documentation/trace/kprobetrace.txt | 5 ++++- Documentation/trace/uprobetracer.txt | 5 ++++- kernel/trace/trace_probe.c | 8 ++++---- kernel/trace/trace_probe.h | 2 +- tools/perf/Documentation/perf-probe.txt | 5 ++--- 5 files changed, 15 insertions(+), 10 deletions(-) diff --git a/Documentation/trace/kprobetrace.txt b/Documentation/trace/kprobetrace.txt index 9109c8e8a5a1..e4991fb1eedc 100644 --- a/Documentation/trace/kprobetrace.txt +++ b/Documentation/trace/kprobetrace.txt @@ -54,7 +54,10 @@ Types ----- Several types are supported for fetch-args. Kprobe tracer will access memory by given type. Prefix 's' and 'u' means those types are signed and unsigned -respectively. Traced arguments are shown in decimal (signed) or hex (unsigned). +respectively. 'x' prefix implies it is unsigned. Traced arguments are shown +in decimal ('s' and 'u') or hexadecimal ('x'). Without type casting, 'x32' +or 'x64' is used depends on the architecture (e.g. x86-32 uses x32, and +x86-64 uses x64). String type is a special type, which fetches a "null-terminated" string from kernel space. This means it will fail and store NULL if the string container has been paged out. diff --git a/Documentation/trace/uprobetracer.txt b/Documentation/trace/uprobetracer.txt index 7e6d28c54b75..94b6b4581763 100644 --- a/Documentation/trace/uprobetracer.txt +++ b/Documentation/trace/uprobetracer.txt @@ -50,7 +50,10 @@ Types ----- Several types are supported for fetch-args. Uprobe tracer will access memory by given type. Prefix 's' and 'u' means those types are signed and unsigned -respectively. Traced arguments are shown in decimal (signed) or hex (unsigned). +respectively. 'x' prefix implies it is unsigned. Traced arguments are shown +in decimal ('s' and 'u') or hexadecimal ('x'). Without type casting, 'x32' +or 'x64' is used depends on the architecture (e.g. x86-32 uses x32, and +x86-64 uses x64). String type is a special type, which fetches a "null-terminated" string from user space. Bitfield is another special type, which takes 3 parameters, bit-width, bit- diff --git a/kernel/trace/trace_probe.c b/kernel/trace/trace_probe.c index 725af9dcbdff..8c0553d9afd3 100644 --- a/kernel/trace/trace_probe.c +++ b/kernel/trace/trace_probe.c @@ -46,10 +46,10 @@ int PRINT_TYPE_FUNC_NAME(tname)(struct trace_seq *s, const char *name, \ const char PRINT_TYPE_FMT_NAME(tname)[] = fmt; \ NOKPROBE_SYMBOL(PRINT_TYPE_FUNC_NAME(tname)); -DEFINE_BASIC_PRINT_TYPE_FUNC(u8, u8, "0x%x") -DEFINE_BASIC_PRINT_TYPE_FUNC(u16, u16, "0x%x") -DEFINE_BASIC_PRINT_TYPE_FUNC(u32, u32, "0x%x") -DEFINE_BASIC_PRINT_TYPE_FUNC(u64, u64, "0x%Lx") +DEFINE_BASIC_PRINT_TYPE_FUNC(u8, u8, "%u") +DEFINE_BASIC_PRINT_TYPE_FUNC(u16, u16, "%u") +DEFINE_BASIC_PRINT_TYPE_FUNC(u32, u32, "%u") +DEFINE_BASIC_PRINT_TYPE_FUNC(u64, u64, "%Lu") DEFINE_BASIC_PRINT_TYPE_FUNC(s8, s8, "%d") DEFINE_BASIC_PRINT_TYPE_FUNC(s16, s16, "%d") DEFINE_BASIC_PRINT_TYPE_FUNC(s32, s32, "%d") diff --git a/kernel/trace/trace_probe.h b/kernel/trace/trace_probe.h index f0c470a10edd..0c0ae54d44c6 100644 --- a/kernel/trace/trace_probe.h +++ b/kernel/trace/trace_probe.h @@ -208,7 +208,7 @@ DEFINE_FETCH_##method(u32) \ DEFINE_FETCH_##method(u64) /* Default (unsigned long) fetch type */ -#define __DEFAULT_FETCH_TYPE(t) u##t +#define __DEFAULT_FETCH_TYPE(t) x##t #define _DEFAULT_FETCH_TYPE(t) __DEFAULT_FETCH_TYPE(t) #define DEFAULT_FETCH_TYPE _DEFAULT_FETCH_TYPE(BITS_PER_LONG) #define DEFAULT_FETCH_TYPE_STR __stringify(DEFAULT_FETCH_TYPE) diff --git a/tools/perf/Documentation/perf-probe.txt b/tools/perf/Documentation/perf-probe.txt index d217617108a1..f37d123d5dac 100644 --- a/tools/perf/Documentation/perf-probe.txt +++ b/tools/perf/Documentation/perf-probe.txt @@ -176,13 +176,12 @@ Each probe argument follows below syntax. 'NAME' specifies the name of this argument (optional). You can use the name of local variable, local data structure member (e.g. var->field, var.field2), local array with fixed index (e.g. array[1], var->array[0], var->pointer[2]), or kprobe-tracer argument format (e.g. $retval, %ax, etc). Note that the name of this argument will be set as the last member name if you specify a local data structure member (e.g. field2 for 'var->field1.field2'.) '$vars' and '$params' special arguments are also available for NAME, '$vars' is expanded to the local variables (including function parameters) which can access at given probe point. '$params' is expanded to only the function parameters. -'TYPE' casts the type of this argument (optional). If omitted, perf probe automatically set the type based on debuginfo. Currently, basic types (u8/u16/u32/u64/s8/s16/s32/s64), hexadecimal integers (x/x8/x16/x32/x64), signedness casting (u/s), "string" and bitfield are supported. (see TYPES for detail) - +'TYPE' casts the type of this argument (optional). If omitted, perf probe automatically set the type based on debuginfo (*). Currently, basic types (u8/u16/u32/u64/s8/s16/s32/s64), hexadecimal integers (x/x8/x16/x32/x64), signedness casting (u/s), "string" and bitfield are supported. (see TYPES for detail) On x86 systems %REG is always the short form of the register: for example %AX. %RAX or %EAX is not valid. TYPES ----- -Basic types (u8/u16/u32/u64/s8/s16/s32/s64) and hexadecimal integers (x8/x16/x32/x64) are integer types. Prefix 's' and 'u' means those types are signed and unsigned respectively, and 'x' means that is shown in hexadecimal format. Traced arguments are shown in decimal (signed) or hex (unsigned). You can also use 's' or 'u' to specify only signedness and leave its size auto-detected by perf probe. Moreover, you can use 'x' to explicitly specify to be shown in hexadecimal (the size is also auto-detected). +Basic types (u8/u16/u32/u64/s8/s16/s32/s64) and hexadecimal integers (x8/x16/x32/x64) are integer types. Prefix 's' and 'u' means those types are signed and unsigned respectively, and 'x' means that is shown in hexadecimal format. Traced arguments are shown in decimal (sNN/uNN) or hex (xNN). You can also use 's' or 'u' to specify only signedness and leave its size auto-detected by perf probe. Moreover, you can use 'x' to explicitly specify to be shown in hexadecimal (the size is also auto-detected). String type is a special type, which fetches a "null-terminated" string from kernel space. This means it will fail and store NULL if the string container has been paged out. You can specify 'string' type only for the local variable or structure member which is an array of or a pointer to 'char' or 'unsigned char' type. Bitfield is another special type, which takes 3 parameters, bit-width, bit-offset, and container-size (usually 32). The syntax is; From dd6fa4e197f0123bcd751e47c171121b2451f148 Mon Sep 17 00:00:00 2001 From: Colin Ian King Date: Sun, 21 Aug 2016 15:12:56 +0100 Subject: [PATCH 35/38] perf tools: Fix typo: "ehough" -> "enough" Trivial typo fix in pr_debug message Signed-off-by: Colin King Cc: Alexander Shishkin Cc: He Kuang Cc: Jiri Olsa Cc: Peter Zijlstra Cc: Wang Nan Link: http://lkml.kernel.org/r/20160821141256.7530-1-colin.king@canonical.com Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/tests/backward-ring-buffer.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/tools/perf/tests/backward-ring-buffer.c b/tools/perf/tests/backward-ring-buffer.c index 615780cbfe1d..e6d1816e431a 100644 --- a/tools/perf/tests/backward-ring-buffer.c +++ b/tools/perf/tests/backward-ring-buffer.c @@ -97,7 +97,7 @@ int test__backward_ring_buffer(int subtest __maybe_unused) evlist = perf_evlist__new(); if (!evlist) { - pr_debug("No ehough memory to create evlist\n"); + pr_debug("No enough memory to create evlist\n"); return TEST_FAIL; } From 17d4666f0613dfd1e2a3919d9b4c724acba8f8b1 Mon Sep 17 00:00:00 2001 From: Colin Ian King Date: Sun, 21 Aug 2016 15:16:03 +0100 Subject: [PATCH 36/38] perf test bpf: Fix typo: "ehough" -> "enough" Trivial typo fix in pr_debug message Signed-off-by: Colin King Cc: Alexander Shishkin Cc: Peter Zijlstra Cc: Wang Nan Link: http://lkml.kernel.org/r/20160821141603.7832-1-colin.king@canonical.com Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/tests/bpf.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/tools/perf/tests/bpf.c b/tools/perf/tests/bpf.c index fc54064b9186..2673e86ed50f 100644 --- a/tools/perf/tests/bpf.c +++ b/tools/perf/tests/bpf.c @@ -125,7 +125,7 @@ static int do_test(struct bpf_object *obj, int (*func)(void), /* Instead of perf_evlist__new_default, don't add default events */ evlist = perf_evlist__new(); if (!evlist) { - pr_debug("No ehough memory to create evlist\n"); + pr_debug("No enough memory to create evlist\n"); return TEST_FAIL; } From c77ce225d5ed1c5bfd4c6d58d3333d859ae49fd7 Mon Sep 17 00:00:00 2001 From: Colin Ian King Date: Sun, 21 Aug 2016 15:19:24 +0100 Subject: [PATCH 37/38] perf bpf: Fix typo: "ehough" -> "enough" Trivial typo fix in pr_debug message Signed-off-by: Colin King Cc: Alexander Shishkin Cc: Alexei Starovoitov Cc: He Kuang Cc: Peter Zijlstra Cc: Wang Nan Link: http://lkml.kernel.org/r/20160821141924.8056-1-colin.king@canonical.com Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/util/bpf-loader.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/tools/perf/util/bpf-loader.c b/tools/perf/util/bpf-loader.c index 1f12e4e40006..2b2c9b82f5ab 100644 --- a/tools/perf/util/bpf-loader.c +++ b/tools/perf/util/bpf-loader.c @@ -531,7 +531,7 @@ static int map_prologue(struct perf_probe_event *pev, int *mapping, ptevs = malloc(array_sz); if (!ptevs) { - pr_debug("No ehough memory: alloc ptevs failed\n"); + pr_debug("No enough memory: alloc ptevs failed\n"); return -ENOMEM; } From 5e30d55c71de058e4156080fe32d426c22d094cb Mon Sep 17 00:00:00 2001 From: Colin Ian King Date: Mon, 22 Aug 2016 19:30:08 +0100 Subject: [PATCH 38/38] perf record: Fix spelling mistake "Finshed" -> "Finished" Trivial fix to spelling mistake in pr_debug message. Signed-off-by: Colin King Cc: Alexander Shishkin Cc: Peter Zijlstra Link: http://lkml.kernel.org/r/20160822183008.26368-1-colin.king@canonical.com Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/builtin-record.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/tools/perf/builtin-record.c b/tools/perf/builtin-record.c index a3792e8c3e03..03251c7f14ec 100644 --- a/tools/perf/builtin-record.c +++ b/tools/perf/builtin-record.c @@ -96,7 +96,7 @@ backward_rb_find_range(void *buf, int mask, u64 head, u64 *start, u64 *end) *start = head; while (true) { if (evt_head - head >= (unsigned int)size) { - pr_debug("Finshed reading backward ring buffer: rewind\n"); + pr_debug("Finished reading backward ring buffer: rewind\n"); if (evt_head - head > (unsigned int)size) evt_head -= pheader->size; *end = evt_head; @@ -106,7 +106,7 @@ backward_rb_find_range(void *buf, int mask, u64 head, u64 *start, u64 *end) pheader = (struct perf_event_header *)(buf + (evt_head & mask)); if (pheader->size == 0) { - pr_debug("Finshed reading backward ring buffer: get start\n"); + pr_debug("Finished reading backward ring buffer: get start\n"); *end = evt_head; return 0; }