From 3c581a7f94542341bf0da496a226b44ac63521a8 Mon Sep 17 00:00:00 2001 From: Ingo Molnar Date: Tue, 11 Aug 2009 10:47:36 +0200 Subject: [PATCH 01/17] perf_counter, x86: Fix lapic printk message Instead of this garbled bootup on UP Pentium-M systems: [ 0.015048] Performance Counters: [ 0.016004] no Local APIC, try rebooting with lapicno PMU driver, software counters only. Print: [ 0.015050] Performance Counters: [ 0.016004] no APIC, boot with the "lapic" boot parameter to force-enable it. [ 0.017003] no PMU driver, software counters only. Cf: Frederic Weisbecker Cc: Peter Zijlstra Cc: Paul Mackerras Cc: Arnaldo Carvalho de Melo Cc: Frederic Weisbecker LKML-Reference: Signed-off-by: Ingo Molnar --- arch/x86/kernel/cpu/perf_counter.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/arch/x86/kernel/cpu/perf_counter.c b/arch/x86/kernel/cpu/perf_counter.c index a7aa8f900954..40e233a24d9f 100644 --- a/arch/x86/kernel/cpu/perf_counter.c +++ b/arch/x86/kernel/cpu/perf_counter.c @@ -1590,7 +1590,7 @@ static int p6_pmu_init(void) } if (!cpu_has_apic) { - pr_info("no Local APIC, try rebooting with lapic"); + pr_info("no APIC, boot with the \"lapic\" boot parameter to force-enable it.\n"); return -ENODEV; } From f64ccccb8afa43abdd63fcbd230f818d6ea0883f Mon Sep 17 00:00:00 2001 From: Ingo Molnar Date: Tue, 11 Aug 2009 10:26:33 +0200 Subject: [PATCH 02/17] perf_counter, x86: Fix generic cache events on P6-mobile CPUs Johannes Stezenbach reported that 'perf stat' does not count cache-miss and cache-references events on his Pentium-M based laptop. This is because we left them blank in p6_perfmon_event_map[], fill them in. Reported-by: Johannes Stezenbach Cc: Peter Zijlstra Cc: Mike Galbraith Cc: Paul Mackerras Cc: Arnaldo Carvalho de Melo Cc: Frederic Weisbecker LKML-Reference: Signed-off-by: Ingo Molnar --- arch/x86/kernel/cpu/perf_counter.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/arch/x86/kernel/cpu/perf_counter.c b/arch/x86/kernel/cpu/perf_counter.c index 40e233a24d9f..fffc126dbdf0 100644 --- a/arch/x86/kernel/cpu/perf_counter.c +++ b/arch/x86/kernel/cpu/perf_counter.c @@ -72,8 +72,8 @@ static const u64 p6_perfmon_event_map[] = { [PERF_COUNT_HW_CPU_CYCLES] = 0x0079, [PERF_COUNT_HW_INSTRUCTIONS] = 0x00c0, - [PERF_COUNT_HW_CACHE_REFERENCES] = 0x0000, - [PERF_COUNT_HW_CACHE_MISSES] = 0x0000, + [PERF_COUNT_HW_CACHE_REFERENCES] = 0x0f2e, + [PERF_COUNT_HW_CACHE_MISSES] = 0x012e, [PERF_COUNT_HW_BRANCH_INSTRUCTIONS] = 0x00c4, [PERF_COUNT_HW_BRANCH_MISSES] = 0x00c5, [PERF_COUNT_HW_BUS_CYCLES] = 0x0062, From 2a8083f063472f27c253545dd64e1a7bbbb1ab61 Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Tue, 11 Aug 2009 16:22:00 -0300 Subject: [PATCH 03/17] perf record: Fix .tid and .pid fill-in when synthesizing events MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Noticed when trying to record events for a firefox thread. We were synthesizing both .tid and .pid with the pid passed via --pid. Fix it by reading /proc/PID/status and getting the tgid to use in .pid, .tid gets the specified "pid". Signed-off-by: Arnaldo Carvalho de Melo Cc: "H. Peter Anvin" Cc: Frédéric Weisbecker Cc: Peter Zijlstra Cc: Mike Galbraith LKML-Reference: <20090811192200.GF18061@ghostprotocols.net> Signed-off-by: Ingo Molnar --- tools/perf/builtin-record.c | 69 ++++++++++++++++++++----------------- 1 file changed, 37 insertions(+), 32 deletions(-) diff --git a/tools/perf/builtin-record.c b/tools/perf/builtin-record.c index 0345aad8eba5..30b83def03d4 100644 --- a/tools/perf/builtin-record.c +++ b/tools/perf/builtin-record.c @@ -203,46 +203,48 @@ static void sig_atexit(void) kill(getpid(), signr); } -static void pid_synthesize_comm_event(pid_t pid, int full) +static pid_t pid_synthesize_comm_event(pid_t pid, int full) { struct comm_event comm_ev; char filename[PATH_MAX]; char bf[BUFSIZ]; - int fd; - size_t size; - char *field, *sep; + FILE *fp; + size_t size = 0; DIR *tasks; struct dirent dirent, *next; + pid_t tgid = 0; - snprintf(filename, sizeof(filename), "/proc/%d/stat", pid); + snprintf(filename, sizeof(filename), "/proc/%d/status", pid); - fd = open(filename, O_RDONLY); - if (fd < 0) { + fp = fopen(filename, "r"); + if (fd == NULL) { /* * We raced with a task exiting - just return: */ if (verbose) fprintf(stderr, "couldn't open %s\n", filename); - return; + return 0; } - if (read(fd, bf, sizeof(bf)) < 0) { - fprintf(stderr, "couldn't read %s\n", filename); - exit(EXIT_FAILURE); - } - close(fd); - /* 9027 (cat) R 6747 9027 6747 34816 9027 ... */ memset(&comm_ev, 0, sizeof(comm_ev)); - field = strchr(bf, '('); - if (field == NULL) - goto out_failure; - sep = strchr(++field, ')'); - if (sep == NULL) - goto out_failure; - size = sep - field; - memcpy(comm_ev.comm, field, size++); + while (!comm_ev.comm[0] || !comm_ev.pid) { + if (fgets(bf, sizeof(bf), fp) == NULL) + goto out_failure; + + if (memcmp(bf, "Name:", 5) == 0) { + char *name = bf + 5; + while (*name && isspace(*name)) + ++name; + size = strlen(name) - 1; + memcpy(comm_ev.comm, name, size++); + } else if (memcmp(bf, "Tgid:", 5) == 0) { + char *tgids = bf + 5; + while (*tgids && isspace(*tgids)) + ++tgids; + tgid = comm_ev.pid = atoi(tgids); + } + } - comm_ev.pid = pid; comm_ev.header.type = PERF_EVENT_COMM; size = ALIGN(size, sizeof(u64)); comm_ev.header.size = sizeof(comm_ev) - (sizeof(comm_ev.comm) - size); @@ -251,7 +253,7 @@ static void pid_synthesize_comm_event(pid_t pid, int full) comm_ev.tid = pid; write_output(&comm_ev, comm_ev.header.size); - return; + goto out_fclose; } snprintf(filename, sizeof(filename), "/proc/%d/task", pid); @@ -268,7 +270,10 @@ static void pid_synthesize_comm_event(pid_t pid, int full) write_output(&comm_ev, comm_ev.header.size); } closedir(tasks); - return; + +out_fclose: + fclose(fp); + return tgid; out_failure: fprintf(stderr, "couldn't get COMM and pgid, malformed %s\n", @@ -276,7 +281,7 @@ out_failure: exit(EXIT_FAILURE); } -static void pid_synthesize_mmap_samples(pid_t pid) +static void pid_synthesize_mmap_samples(pid_t pid, pid_t tgid) { char filename[PATH_MAX]; FILE *fp; @@ -328,7 +333,7 @@ static void pid_synthesize_mmap_samples(pid_t pid) mmap_ev.len -= mmap_ev.start; mmap_ev.header.size = (sizeof(mmap_ev) - (sizeof(mmap_ev.filename) - size)); - mmap_ev.pid = pid; + mmap_ev.pid = tgid; mmap_ev.tid = pid; write_output(&mmap_ev, mmap_ev.header.size); @@ -347,14 +352,14 @@ static void synthesize_all(void) while (!readdir_r(proc, &dirent, &next) && next) { char *end; - pid_t pid; + pid_t pid, tgid; pid = strtol(dirent.d_name, &end, 10); if (*end) /* only interested in proper numerical dirents */ continue; - pid_synthesize_comm_event(pid, 1); - pid_synthesize_mmap_samples(pid); + tgid = pid_synthesize_comm_event(pid, 1); + pid_synthesize_mmap_samples(pid, tgid); } closedir(proc); @@ -567,8 +572,8 @@ static int __cmd_record(int argc, const char **argv) perf_header__write(header, output); if (!system_wide) { - pid_synthesize_comm_event(pid, 0); - pid_synthesize_mmap_samples(pid); + pid_t tgid = pid_synthesize_comm_event(pid, 0); + pid_synthesize_mmap_samples(pid, tgid); } else synthesize_all(); From 94a24752fe95ca1e7f98b197052d44e6a207740d Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Tue, 11 Aug 2009 16:21:38 -0300 Subject: [PATCH 04/17] perf report: Show the tid too in -D MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit This made it easier to find the firefox threading related bug. Signed-off-by: Arnaldo Carvalho de Melo Cc: "H. Peter Anvin" Cc: Peter Zijlstra Cc: Mike Galbraith Cc: Frédéric Weisbecker LKML-Reference: <20090811192138.GE18061@ghostprotocols.net> Signed-off-by: Ingo Molnar --- tools/perf/builtin-report.c | 7 ++++--- 1 file changed, 4 insertions(+), 3 deletions(-) diff --git a/tools/perf/builtin-report.c b/tools/perf/builtin-report.c index 99274cec0adb..23e1457f2409 100644 --- a/tools/perf/builtin-report.c +++ b/tools/perf/builtin-report.c @@ -1526,11 +1526,11 @@ process_sample_event(event_t *event, unsigned long offset, unsigned long head) more_data += sizeof(u64); } - dprintf("%p [%p]: PERF_EVENT_SAMPLE (IP, %d): %d: %p period: %Ld\n", + dprintf("%p [%p]: PERF_EVENT_SAMPLE (IP, %d): %d/%d: %p period: %Ld\n", (void *)(offset + head), (void *)(long)(event->header.size), event->header.misc, - event->ip.pid, + event->ip.pid, event->ip.tid, (void *)(long)ip, (long long)period); @@ -1612,10 +1612,11 @@ process_mmap_event(event_t *event, unsigned long offset, unsigned long head) struct thread *thread = threads__findnew(event->mmap.pid); struct map *map = map__new(&event->mmap); - dprintf("%p [%p]: PERF_EVENT_MMAP %d: [%p(%p) @ %p]: %s\n", + dprintf("%p [%p]: PERF_EVENT_MMAP %d/%d: [%p(%p) @ %p]: %s\n", (void *)(offset + head), (void *)(long)(event->header.size), event->mmap.pid, + event->mmap.tid, (void *)(long)event->mmap.start, (void *)(long)event->mmap.len, (void *)(long)event->mmap.pgoff, From 247648e3742ded01e42a4b14c2da330b13cbb47f Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Tue, 11 Aug 2009 16:22:11 -0300 Subject: [PATCH 05/17] perf tools: Fix fallback to cplus_demangle() when bfd_demangle() is not available MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit In old binutils we can't access bfd_demangle(), use cplus_demangle() just like oprofile. Signed-off-by: Arnaldo Carvalho de Melo Cc: Luis Claudio R. Gonçalves Cc: H. Peter Anvin Cc: Peter Zijlstra Cc: Mike Galbraith Cc: Frédéric Weisbecker LKML-Reference: <20090811192211.GG18061@ghostprotocols.net> Signed-off-by: Ingo Molnar --- tools/perf/Makefile | 29 ++++++++++++++++++----------- tools/perf/util/symbol.c | 15 --------------- tools/perf/util/symbol.h | 24 ++++++++++++++++++++++++ 3 files changed, 42 insertions(+), 26 deletions(-) diff --git a/tools/perf/Makefile b/tools/perf/Makefile index 60411e94113b..c045b4271e57 100644 --- a/tools/perf/Makefile +++ b/tools/perf/Makefile @@ -382,22 +382,29 @@ endif ifdef NO_DEMANGLE BASIC_CFLAGS += -DNO_DEMANGLE else - has_bfd := $(shell sh -c "(echo '\#include '; echo 'int main(void) { bfd_demangle(0, 0, 0); return 0; }') | $(CC) -x c - $(ALL_CFLAGS) -o /dev/null $(ALL_LDFLAGS) -lbfd > /dev/null 2>&1 && echo y") - has_bfd_iberty := $(shell sh -c "(echo '\#include '; echo 'int main(void) { bfd_demangle(0, 0, 0); return 0; }') | $(CC) -x c - $(ALL_CFLAGS) -o /dev/null $(ALL_LDFLAGS) -lbfd -liberty > /dev/null 2>&1 && echo y") - - has_bfd_iberty_z := $(shell sh -c "(echo '\#include '; echo 'int main(void) { bfd_demangle(0, 0, 0); return 0; }') | $(CC) -x c - $(ALL_CFLAGS) -o /dev/null $(ALL_LDFLAGS) -lbfd -liberty -lz > /dev/null 2>&1 && echo y") - ifeq ($(has_bfd),y) EXTLIBS += -lbfd - else ifeq ($(has_bfd_iberty),y) - EXTLIBS += -lbfd -liberty - else ifeq ($(has_bfd_iberty_z),y) - EXTLIBS += -lbfd -liberty -lz else - msg := $(warning No bfd.h/libbfd found, install binutils-dev[el] to gain symbol demangling) - BASIC_CFLAGS += -DNO_DEMANGLE + has_bfd_iberty := $(shell sh -c "(echo '\#include '; echo 'int main(void) { bfd_demangle(0, 0, 0); return 0; }') | $(CC) -x c - $(ALL_CFLAGS) -o /dev/null $(ALL_LDFLAGS) -lbfd -liberty > /dev/null 2>&1 && echo y") + ifeq ($(has_bfd_iberty),y) + EXTLIBS += -lbfd -liberty + else + has_bfd_iberty_z := $(shell sh -c "(echo '\#include '; echo 'int main(void) { bfd_demangle(0, 0, 0); return 0; }') | $(CC) -x c - $(ALL_CFLAGS) -o /dev/null $(ALL_LDFLAGS) -lbfd -liberty -lz > /dev/null 2>&1 && echo y") + ifeq ($(has_bfd_iberty_z),y) + EXTLIBS += -lbfd -liberty -lz + else + has_cplus_demangle := $(shell sh -c "(echo 'extern char *cplus_demangle(const char *, int);'; echo 'int main(void) { cplus_demangle(0, 0); return 0; }') | $(CC) -x c - $(ALL_CFLAGS) -o /dev/null $(ALL_LDFLAGS) -liberty > /dev/null 2>&1 && echo y") + ifeq ($(has_cplus_demangle),y) + EXTLIBS += -liberty + BASIC_CFLAGS += -DHAVE_CPLUS_DEMANGLE + else + msg := $(warning No bfd.h/libbfd found, install binutils-dev[el] to gain symbol demangling) + BASIC_CFLAGS += -DNO_DEMANGLE + endif + endif + endif endif endif diff --git a/tools/perf/util/symbol.c b/tools/perf/util/symbol.c index f1dcede14307..2473fd427beb 100644 --- a/tools/perf/util/symbol.c +++ b/tools/perf/util/symbol.c @@ -7,23 +7,8 @@ #include #include -#ifndef NO_DEMANGLE -#include -#else -static inline -char *bfd_demangle(void __used *v, const char __used *c, int __used i) -{ - return NULL; -} -#endif - const char *sym_hist_filter; -#ifndef DMGL_PARAMS -#define DMGL_PARAMS (1 << 0) /* Include function args */ -#define DMGL_ANSI (1 << 1) /* Include const, volatile, etc */ -#endif - enum dso_origin { DSO__ORIG_KERNEL = 0, DSO__ORIG_JAVA_JIT, diff --git a/tools/perf/util/symbol.h b/tools/perf/util/symbol.h index 1e003ec2f4b1..b53bf0125c1b 100644 --- a/tools/perf/util/symbol.h +++ b/tools/perf/util/symbol.h @@ -7,6 +7,30 @@ #include #include "module.h" +#ifdef HAVE_CPLUS_DEMANGLE +extern char *cplus_demangle(const char *, int); + +static inline char *bfd_demangle(void __used *v, const char *c, int i) +{ + return cplus_demangle(c, i); +} +#else +#ifdef NO_DEMANGLE +static inline char *bfd_demangle(void __used *v, const char __used *c, + int __used i) +{ + return NULL; +} +#else +#include +#endif +#endif + +#ifndef DMGL_PARAMS +#define DMGL_PARAMS (1 << 0) /* Include function args */ +#define DMGL_ANSI (1 << 1) /* Include const, volatile, etc */ +#endif + struct symbol { struct rb_node rb_node; u64 start; From 1340e6bbaff7ff7f6f75eb4a5c34933efce84a84 Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Tue, 11 Aug 2009 17:04:36 -0300 Subject: [PATCH 06/17] perf tools: Fix dso__new handle() to handle deleted DSOs MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit It is better than showing the map addr, this way at least we know that we can't get the symtabs because the DSO was deleted (system update) while an app still used such DSO. Yeah, don't do that, but if you do, you'll figure it out quicker this way. [acme@doppio linux-2.6-tip]$ perf report | head -15 # Samples: 3796 # # Overhead Command Shared Object Symbol # ........ ....... ................................................................... ...... # 23.55% pidgin /lib64/libglib-2.0.so.0.2000.4.#prelink#.Pd98lu (deleted) [.] 0x00000000038844 21.55% pidgin /lib64/libpthread-2.10.1.so.#prelink#.AFwK8Q (deleted) [.] 0x0000000000a42d 10.85% pidgin [kernel] [.] vread_hpet 7.85% pidgin /lib64/libgobject-2.0.so.0.2000.4.#prelink#.o1vpU7 (deleted) [.] 0x00000000014de8 3.35% pidgin /lib64/libc-2.10.1.so (deleted) [.] 0x0000000007a875 3.19% pidgin /lib64/libdbus-1.so.3.4.0.#prelink#.6mwgZP (deleted) [.] 0x0000000001d254 3.06% pidgin /usr/lib64/libgtk-x11-2.0.so.0.1600.5.#prelink#.511hAl (deleted) [.] 0x000000002334e7 2.90% pidgin /usr/lib64/libgdk-x11-2.0.so.0.1600.5.#prelink#.5qlMo1 (deleted) [.] 0x00000000037b2d 1.84% pidgin [kernel] [k] do_sys_poll 1.45% pidgin /usr/lib64/libX11.so.6.2.0.#prelink#.iR59Rx (deleted) [.] 0x0000000004c751 [acme@doppio linux-2.6-tip]$ Signed-off-by: Arnaldo Carvalho de Melo Cc: Luis Claudio R. Gonçalves Cc: Clark Williams Cc: H. Peter Anvin Cc: Peter Zijlstra Cc: Mike Galbraith Cc: Frédéric Weisbecker LKML-Reference: <20090811200436.GA3478@ghostprotocols.net> Signed-off-by: Ingo Molnar --- tools/perf/util/symbol.c | 2 ++ 1 file changed, 2 insertions(+) diff --git a/tools/perf/util/symbol.c b/tools/perf/util/symbol.c index 2473fd427beb..5c0f42e6b33b 100644 --- a/tools/perf/util/symbol.c +++ b/tools/perf/util/symbol.c @@ -801,6 +801,8 @@ more: } out: free(name); + if (ret < 0 && strstr(self->name, " (deleted)") != NULL) + return 0; return ret; } From 0a5ac84650fb7a7f226814103d95724e34b012ae Mon Sep 17 00:00:00 2001 From: Jens Axboe Date: Wed, 12 Aug 2009 11:18:01 +0200 Subject: [PATCH 07/17] perf record: Add missing -C option support for specifying profile cpu perf top supports a -C for setting the profile CPU, but perf record does not. This adds the same option for perf record, allowing the user to specify a specific target profile CPU. Signed-off-by: Jens Axboe Cc: Peter Zijlstra Cc: Mike Galbraith Cc: Paul Mackerras Cc: Arnaldo Carvalho de Melo Cc: Frederic Weisbecker LKML-Reference: <20090812091801.GC12579@kernel.dk> Signed-off-by: Ingo Molnar --- tools/perf/builtin-record.c | 17 ++++++++++++++--- 1 file changed, 14 insertions(+), 3 deletions(-) diff --git a/tools/perf/builtin-record.c b/tools/perf/builtin-record.c index 30b83def03d4..de76008fbaa2 100644 --- a/tools/perf/builtin-record.c +++ b/tools/perf/builtin-record.c @@ -35,6 +35,7 @@ static const char *output_name = "perf.data"; static int group = 0; static unsigned int realtime_prio = 0; static int system_wide = 0; +static int profile_cpu = -1; static pid_t target_pid = -1; static int inherit = 1; static int force = 0; @@ -431,6 +432,8 @@ try_again: if (err == EPERM) die("Permission error - are you root?\n"); + else if (err == ENODEV && profile_cpu != -1) + die("No such device - did you specify an out-of-range profile CPU?\n"); /* * If it's cycles then fall back to hrtimer @@ -564,9 +567,15 @@ static int __cmd_record(int argc, const char **argv) if (pid == -1) pid = getpid(); - open_counters(-1, pid); - } else for (i = 0; i < nr_cpus; i++) - open_counters(i, target_pid); + open_counters(profile_cpu, pid); + } else { + if (profile_cpu != -1) { + open_counters(profile_cpu, target_pid); + } else { + for (i = 0; i < nr_cpus; i++) + open_counters(i, target_pid); + } + } if (file_new) perf_header__write(header, output); @@ -645,6 +654,8 @@ static const struct option options[] = { "system-wide collection from all CPUs"), OPT_BOOLEAN('A', "append", &append_file, "append to the output file to do incremental profiling"), + OPT_INTEGER('C', "profile_cpu", &profile_cpu, + "CPU to profile on"), OPT_BOOLEAN('f', "force", &force, "overwrite existing data file"), OPT_LONG('c', "count", &default_interval, From 04da8a43da804723a550f00dd158fd5b5e25ae35 Mon Sep 17 00:00:00 2001 From: Ingo Molnar Date: Tue, 11 Aug 2009 10:40:08 +0200 Subject: [PATCH 08/17] perf_counter, x86: Fix/improve apic fallback Johannes Stezenbach reported that his Pentium-M based laptop does not have the local APIC enabled by default, and hence perfcounters do not get initialized. Add a fallback for this case: allow non-sampled counters and return with an error on sampled counters. This allows 'perf stat' to work out of box - and allows 'perf top' and 'perf record' to fall back on a hrtimer based sampling method. ( Passing 'lapic' on the boot line will allow hardware sampling to occur - but if the APIC is disabled permanently by the hardware then this fallback still allows more systems to use perfcounters. ) Also decouple perfcounter support from X86_LOCAL_APIC. -v2: fix typo breaking counters on all other systems ... Reported-by: Johannes Stezenbach Cc: Peter Zijlstra Cc: Mike Galbraith Cc: Paul Mackerras Cc: Arnaldo Carvalho de Melo Cc: Frederic Weisbecker LKML-Reference: Signed-off-by: Ingo Molnar --- arch/x86/Kconfig | 2 +- arch/x86/kernel/cpu/perf_counter.c | 34 ++++++++++++++++++++++++++---- 2 files changed, 31 insertions(+), 5 deletions(-) diff --git a/arch/x86/Kconfig b/arch/x86/Kconfig index 738bdc6b0f8b..13ffa5df37d7 100644 --- a/arch/x86/Kconfig +++ b/arch/x86/Kconfig @@ -24,6 +24,7 @@ config X86 select HAVE_UNSTABLE_SCHED_CLOCK select HAVE_IDE select HAVE_OPROFILE + select HAVE_PERF_COUNTERS if (!M386 && !M486) select HAVE_IOREMAP_PROT select HAVE_KPROBES select ARCH_WANT_OPTIONAL_GPIOLIB @@ -742,7 +743,6 @@ config X86_UP_IOAPIC config X86_LOCAL_APIC def_bool y depends on X86_64 || SMP || X86_32_NON_STANDARD || X86_UP_APIC - select HAVE_PERF_COUNTERS if (!M386 && !M486) config X86_IO_APIC def_bool y diff --git a/arch/x86/kernel/cpu/perf_counter.c b/arch/x86/kernel/cpu/perf_counter.c index fffc126dbdf0..900332b800f8 100644 --- a/arch/x86/kernel/cpu/perf_counter.c +++ b/arch/x86/kernel/cpu/perf_counter.c @@ -55,6 +55,7 @@ struct x86_pmu { int num_counters_fixed; int counter_bits; u64 counter_mask; + int apic; u64 max_period; u64 intel_ctrl; }; @@ -613,6 +614,7 @@ static DEFINE_MUTEX(pmc_reserve_mutex); static bool reserve_pmc_hardware(void) { +#ifdef CONFIG_X86_LOCAL_APIC int i; if (nmi_watchdog == NMI_LOCAL_APIC) @@ -627,9 +629,11 @@ static bool reserve_pmc_hardware(void) if (!reserve_evntsel_nmi(x86_pmu.eventsel + i)) goto eventsel_fail; } +#endif return true; +#ifdef CONFIG_X86_LOCAL_APIC eventsel_fail: for (i--; i >= 0; i--) release_evntsel_nmi(x86_pmu.eventsel + i); @@ -644,10 +648,12 @@ perfctr_fail: enable_lapic_nmi_watchdog(); return false; +#endif } static void release_pmc_hardware(void) { +#ifdef CONFIG_X86_LOCAL_APIC int i; for (i = 0; i < x86_pmu.num_counters; i++) { @@ -657,6 +663,7 @@ static void release_pmc_hardware(void) if (nmi_watchdog == NMI_LOCAL_APIC) enable_lapic_nmi_watchdog(); +#endif } static void hw_perf_counter_destroy(struct perf_counter *counter) @@ -748,6 +755,15 @@ static int __hw_perf_counter_init(struct perf_counter *counter) hwc->sample_period = x86_pmu.max_period; hwc->last_period = hwc->sample_period; atomic64_set(&hwc->period_left, hwc->sample_period); + } else { + /* + * If we have a PMU initialized but no APIC + * interrupts, we cannot sample hardware + * counters (user-space has to fall back and + * sample via a hrtimer based software counter): + */ + if (!x86_pmu.apic) + return -EOPNOTSUPP; } counter->destroy = hw_perf_counter_destroy; @@ -1449,18 +1465,22 @@ void smp_perf_pending_interrupt(struct pt_regs *regs) void set_perf_counter_pending(void) { +#ifdef CONFIG_X86_LOCAL_APIC apic->send_IPI_self(LOCAL_PENDING_VECTOR); +#endif } void perf_counters_lapic_init(void) { - if (!x86_pmu_initialized()) +#ifdef CONFIG_X86_LOCAL_APIC + if (!x86_pmu.apic || !x86_pmu_initialized()) return; /* * Always use NMI for PMU */ apic_write(APIC_LVTPC, APIC_DM_NMI); +#endif } static int __kprobes @@ -1484,7 +1504,9 @@ perf_counter_nmi_handler(struct notifier_block *self, regs = args->regs; +#ifdef CONFIG_X86_LOCAL_APIC apic_write(APIC_LVTPC, APIC_DM_NMI); +#endif /* * Can't rely on the handled return value to say it was our NMI, two * counters could trigger 'simultaneously' raising two back-to-back NMIs. @@ -1515,6 +1537,7 @@ static struct x86_pmu p6_pmu = { .event_map = p6_pmu_event_map, .raw_event = p6_pmu_raw_event, .max_events = ARRAY_SIZE(p6_perfmon_event_map), + .apic = 1, .max_period = (1ULL << 31) - 1, .version = 0, .num_counters = 2, @@ -1541,6 +1564,7 @@ static struct x86_pmu intel_pmu = { .event_map = intel_pmu_event_map, .raw_event = intel_pmu_raw_event, .max_events = ARRAY_SIZE(intel_perfmon_event_map), + .apic = 1, /* * Intel PMCs cannot be accessed sanely above 32 bit width, * so we install an artificial 1<<31 period regardless of @@ -1564,6 +1588,7 @@ static struct x86_pmu amd_pmu = { .num_counters = 4, .counter_bits = 48, .counter_mask = (1ULL << 48) - 1, + .apic = 1, /* use highest bit to detect overflow */ .max_period = (1ULL << 47) - 1, }; @@ -1589,13 +1614,14 @@ static int p6_pmu_init(void) return -ENODEV; } + x86_pmu = p6_pmu; + if (!cpu_has_apic) { pr_info("no APIC, boot with the \"lapic\" boot parameter to force-enable it.\n"); - return -ENODEV; + pr_info("no hardware sampling interrupt available.\n"); + x86_pmu.apic = 0; } - x86_pmu = p6_pmu; - return 0; } From 8f7a0dc51674ad0dd06155291b0aed60d655943c Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Wed, 12 Aug 2009 14:44:59 -0300 Subject: [PATCH 09/17] perf list: Fix large list output by using the pager When /sys/kernel/debug is mounted the list can be imense, so use the pager like the other tools. Signed-off-by: Arnaldo Carvalho de Melo Acked-by: Frederic Weisbecker Cc: Peter Zijlstra LKML-Reference: <20090812174459.GB3495@ghostprotocols.net> Signed-off-by: Ingo Molnar --- tools/perf/builtin-list.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) diff --git a/tools/perf/builtin-list.c b/tools/perf/builtin-list.c index f990fa8a35c9..d88c6961274c 100644 --- a/tools/perf/builtin-list.c +++ b/tools/perf/builtin-list.c @@ -10,11 +10,12 @@ #include "perf.h" -#include "util/parse-options.h" #include "util/parse-events.h" +#include "util/cache.h" int cmd_list(int argc __used, const char **argv __used, const char *prefix __used) { + setup_pager(); print_events(); return 0; } From 28402971d869e26271b25301011f667d3a5640c3 Mon Sep 17 00:00:00 2001 From: Ingo Molnar Date: Thu, 13 Aug 2009 10:13:22 +0200 Subject: [PATCH 10/17] perf_counter: Provide hw_perf_counter_setup_online() APIs Provide weak aliases for hw_perf_counter_setup_online(). This is used by the BTS patches (for v2.6.32), but it interacts with fixes so propagate this upstream. (it has no effect as of yet) Also export perf_counter_output() to architecture code. Cc: Peter Zijlstra Cc: Paul Mackerras Cc: Frederic Weisbecker LKML-Reference: Signed-off-by: Ingo Molnar --- include/linux/perf_counter.h | 2 ++ kernel/perf_counter.c | 10 +++++++++- 2 files changed, 11 insertions(+), 1 deletion(-) diff --git a/include/linux/perf_counter.h b/include/linux/perf_counter.h index a9d823a93fe8..2b36afe6ae0f 100644 --- a/include/linux/perf_counter.h +++ b/include/linux/perf_counter.h @@ -694,6 +694,8 @@ struct perf_sample_data { extern int perf_counter_overflow(struct perf_counter *counter, int nmi, struct perf_sample_data *data); +extern void perf_counter_output(struct perf_counter *counter, int nmi, + struct perf_sample_data *data); /* * Return 1 for a software counter, 0 for a hardware counter diff --git a/kernel/perf_counter.c b/kernel/perf_counter.c index b0b20a07f394..e26d2fcfa320 100644 --- a/kernel/perf_counter.c +++ b/kernel/perf_counter.c @@ -88,6 +88,7 @@ void __weak hw_perf_disable(void) { barrier(); } void __weak hw_perf_enable(void) { barrier(); } void __weak hw_perf_counter_setup(int cpu) { barrier(); } +void __weak hw_perf_counter_setup_online(int cpu) { barrier(); } int __weak hw_perf_group_sched_in(struct perf_counter *group_leader, @@ -2630,7 +2631,7 @@ static u32 perf_counter_tid(struct perf_counter *counter, struct task_struct *p) return task_pid_nr_ns(p, counter->ns); } -static void perf_counter_output(struct perf_counter *counter, int nmi, +void perf_counter_output(struct perf_counter *counter, int nmi, struct perf_sample_data *data) { int ret; @@ -4592,6 +4593,11 @@ perf_cpu_notify(struct notifier_block *self, unsigned long action, void *hcpu) perf_counter_init_cpu(cpu); break; + case CPU_ONLINE: + case CPU_ONLINE_FROZEN: + hw_perf_counter_setup_online(cpu); + break; + case CPU_DOWN_PREPARE: case CPU_DOWN_PREPARE_FROZEN: perf_counter_exit_cpu(cpu); @@ -4616,6 +4622,8 @@ void __init perf_counter_init(void) { perf_cpu_notify(&perf_cpu_nb, (unsigned long)CPU_UP_PREPARE, (void *)(long)smp_processor_id()); + perf_cpu_notify(&perf_cpu_nb, (unsigned long)CPU_ONLINE, + (void *)(long)smp_processor_id()); register_cpu_notifier(&perf_cpu_nb); } From 3a9f131fb00b8ac5950a11ad1599e45edfb5ae44 Mon Sep 17 00:00:00 2001 From: Frederic Weisbecker Date: Thu, 13 Aug 2009 10:27:18 +0200 Subject: [PATCH 11/17] perf tools: Add a per tracepoint counter attribute to get raw sample Add a new flag field while opening a tracepoint perf counter: -e tracepoint_subsystem:tracepoint_name:flags This is intended to be generic although for now it only supports the r[e[c[o[r[d]]]]] flag: ./perf record -e workqueue:workqueue_insertion:record ./perf record -e workqueue:workqueue_insertion:r will have the same effect: enabling the raw samples record for the given tracepoint counter. In the future, we may want to support further flags, separated by commas. Signed-off-by: Frederic Weisbecker Cc: Peter Zijlstra Cc: Arnaldo Carvalho de Melo Cc: Mike Galbraith LKML-Reference: <1250152039-7284-1-git-send-email-fweisbec@gmail.com> Signed-off-by: Ingo Molnar --- tools/perf/builtin-record.c | 2 +- tools/perf/util/parse-events.c | 10 ++++++++++ 2 files changed, 11 insertions(+), 1 deletion(-) diff --git a/tools/perf/builtin-record.c b/tools/perf/builtin-record.c index de76008fbaa2..78adc47da869 100644 --- a/tools/perf/builtin-record.c +++ b/tools/perf/builtin-record.c @@ -398,7 +398,7 @@ static void create_counter(int counter, int cpu, pid_t pid) PERF_FORMAT_TOTAL_TIME_RUNNING | PERF_FORMAT_ID; - attr->sample_type = PERF_SAMPLE_IP | PERF_SAMPLE_TID; + attr->sample_type |= PERF_SAMPLE_IP | PERF_SAMPLE_TID; if (freq) { attr->sample_type |= PERF_SAMPLE_PERIOD; diff --git a/tools/perf/util/parse-events.c b/tools/perf/util/parse-events.c index 4858d83b3b67..044178408783 100644 --- a/tools/perf/util/parse-events.c +++ b/tools/perf/util/parse-events.c @@ -379,6 +379,7 @@ static int parse_tracepoint_event(const char **strp, struct perf_counter_attr *attr) { const char *evt_name; + char *flags; char sys_name[MAX_EVENT_LENGTH]; char id_buf[4]; int fd; @@ -400,6 +401,15 @@ static int parse_tracepoint_event(const char **strp, strncpy(sys_name, *strp, sys_length); sys_name[sys_length] = '\0'; evt_name = evt_name + 1; + + flags = strchr(evt_name, ':'); + if (flags) { + *flags = '\0'; + flags++; + if (!strncmp(flags, "record", strlen(flags))) + attr->sample_type |= PERF_SAMPLE_RAW; + } + evt_length = strlen(evt_name); if (evt_length >= MAX_EVENT_LENGTH) return 0; From daac07b2e6b77f1bd44104aa2f0593e5505f27d4 Mon Sep 17 00:00:00 2001 From: Frederic Weisbecker Date: Thu, 13 Aug 2009 10:27:19 +0200 Subject: [PATCH 12/17] perf tools: Add a general option to enable raw sample records While we can enable the perf sample records per tracepoint counter, we may also want to enable this option for every tracepoint counters to open, so that we don't need to add a :record flag for all of them. Add the -R, --raw-samples options for this purpose. Signed-off-by: Frederic Weisbecker Cc: Peter Zijlstra Cc: Arnaldo Carvalho de Melo Cc: Mike Galbraith LKML-Reference: <1250152039-7284-2-git-send-email-fweisbec@gmail.com> Signed-off-by: Ingo Molnar --- tools/perf/builtin-record.c | 5 +++++ 1 file changed, 5 insertions(+) diff --git a/tools/perf/builtin-record.c b/tools/perf/builtin-record.c index 78adc47da869..3d051b9cf25f 100644 --- a/tools/perf/builtin-record.c +++ b/tools/perf/builtin-record.c @@ -34,6 +34,7 @@ static int output; static const char *output_name = "perf.data"; static int group = 0; static unsigned int realtime_prio = 0; +static int raw_samples = 0; static int system_wide = 0; static int profile_cpu = -1; static pid_t target_pid = -1; @@ -418,6 +419,8 @@ static void create_counter(int counter, int cpu, pid_t pid) if (call_graph) attr->sample_type |= PERF_SAMPLE_CALLCHAIN; + if (raw_samples) + attr->sample_type |= PERF_SAMPLE_RAW; attr->mmap = track; attr->comm = track; @@ -650,6 +653,8 @@ static const struct option options[] = { "record events on existing pid"), OPT_INTEGER('r', "realtime", &realtime_prio, "collect data with this RT SCHED_FIFO priority"), + OPT_BOOLEAN('R', "raw-samples", &raw_samples, + "collect raw sample records from all opened counters"), OPT_BOOLEAN('a', "all-cpus", &system_wide, "system-wide collection from all CPUs"), OPT_BOOLEAN('A', "append", &append_file, From 8fd101f20bdf771949a8f3a5a779877d09b2fb56 Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Wed, 12 Aug 2009 18:19:57 -0300 Subject: [PATCH 13/17] perf report: Don't show unresolved DSOs and symbols when -S/-d is used We're interested in just those symbols/DSOs, so filter out the unresolved ones. Signed-off-by: Arnaldo Carvalho de Melo Cc: Peter Zijlstra LKML-Reference: <20090812211957.GE3495@ghostprotocols.net> Signed-off-by: Ingo Molnar --- tools/perf/builtin-report.c | 5 +++-- 1 file changed, 3 insertions(+), 2 deletions(-) diff --git a/tools/perf/builtin-report.c b/tools/perf/builtin-report.c index 23e1457f2409..b53a60fc12de 100644 --- a/tools/perf/builtin-report.c +++ b/tools/perf/builtin-report.c @@ -1590,10 +1590,11 @@ process_sample_event(event_t *event, unsigned long offset, unsigned long head) if (show & show_mask) { struct symbol *sym = resolve_symbol(thread, &map, &dso, &ip); - if (dso_list && dso && dso->name && !strlist__has_entry(dso_list, dso->name)) + if (dso_list && (!dso || !dso->name || + !strlist__has_entry(dso_list, dso->name))) return 0; - if (sym_list && sym && !strlist__has_entry(sym_list, sym->name)) + if (sym_list && (!sym || !strlist__has_entry(sym_list, sym->name))) return 0; if (hist_entry__add(thread, map, dso, sym, ip, chain, level, period)) { From bcfc2602e8541ac13b1def38e2591dca072cff7a Mon Sep 17 00:00:00 2001 From: Peter Zijlstra Date: Thu, 13 Aug 2009 09:51:55 +0200 Subject: [PATCH 14/17] perf_counter: Fix swcounter context invariance perf_swcounter_is_counting() uses a lock, which means we cannot use swcounters from NMI or when holding that particular lock, this is unintended. The below removes the lock, this opens up race window, but not worse than the swcounters already experience due to RCU traversal of the context in perf_swcounter_ctx_event(). This also fixes the hard lockups while opening a lockdep tracepoint counter. Signed-off-by: Peter Zijlstra Acked-by: Frederic Weisbecker Cc: Paul Mackerras Cc: stephane eranian Cc: Corey J Ashford LKML-Reference: <1250149915.10001.66.camel@twins> Signed-off-by: Ingo Molnar --- kernel/perf_counter.c | 44 ++++++++++++++++++------------------------- 1 file changed, 18 insertions(+), 26 deletions(-) diff --git a/kernel/perf_counter.c b/kernel/perf_counter.c index e26d2fcfa320..3dd4339589a0 100644 --- a/kernel/perf_counter.c +++ b/kernel/perf_counter.c @@ -3444,40 +3444,32 @@ static void perf_swcounter_add(struct perf_counter *counter, u64 nr, static int perf_swcounter_is_counting(struct perf_counter *counter) { - struct perf_counter_context *ctx; - unsigned long flags; - int count; - + /* + * The counter is active, we're good! + */ if (counter->state == PERF_COUNTER_STATE_ACTIVE) return 1; + /* + * The counter is off/error, not counting. + */ if (counter->state != PERF_COUNTER_STATE_INACTIVE) return 0; /* - * If the counter is inactive, it could be just because - * its task is scheduled out, or because it's in a group - * which could not go on the PMU. We want to count in - * the first case but not the second. If the context is - * currently active then an inactive software counter must - * be the second case. If it's not currently active then - * we need to know whether the counter was active when the - * context was last active, which we can determine by - * comparing counter->tstamp_stopped with ctx->time. - * - * We are within an RCU read-side critical section, - * which protects the existence of *ctx. + * The counter is inactive, if the context is active + * we're part of a group that didn't make it on the 'pmu', + * not counting. */ - ctx = counter->ctx; - spin_lock_irqsave(&ctx->lock, flags); - count = 1; - /* Re-check state now we have the lock */ - if (counter->state < PERF_COUNTER_STATE_INACTIVE || - counter->ctx->is_active || - counter->tstamp_stopped < ctx->time) - count = 0; - spin_unlock_irqrestore(&ctx->lock, flags); - return count; + if (counter->ctx->is_active) + return 0; + + /* + * We're inactive and the context is too, this means the + * task is scheduled out, we're counting events that happen + * to us, like migration events. + */ + return 1; } static int perf_swcounter_match(struct perf_counter *counter, From 3dab77fb1bf89664bb1c9544607159dcab6f7d57 Mon Sep 17 00:00:00 2001 From: Peter Zijlstra Date: Thu, 13 Aug 2009 11:47:53 +0200 Subject: [PATCH 15/17] perf: Rework/fix the whole read vs group stuff Replace PERF_SAMPLE_GROUP with PERF_SAMPLE_READ and introduce PERF_FORMAT_GROUP to deal with group reads in a more generic way. This allows you to get group reads out of read() as well. Signed-off-by: Peter Zijlstra Cc: Corey J Ashford Cc: Paul Mackerras Cc: stephane eranian LKML-Reference: <20090813103655.117411814@chello.nl> Signed-off-by: Ingo Molnar --- include/linux/perf_counter.h | 47 ++++-- kernel/perf_counter.c | 274 ++++++++++++++++++++++++++--------- 2 files changed, 238 insertions(+), 83 deletions(-) diff --git a/include/linux/perf_counter.h b/include/linux/perf_counter.h index 2b36afe6ae0f..b53f7006cc4e 100644 --- a/include/linux/perf_counter.h +++ b/include/linux/perf_counter.h @@ -115,7 +115,7 @@ enum perf_counter_sample_format { PERF_SAMPLE_TID = 1U << 1, PERF_SAMPLE_TIME = 1U << 2, PERF_SAMPLE_ADDR = 1U << 3, - PERF_SAMPLE_GROUP = 1U << 4, + PERF_SAMPLE_READ = 1U << 4, PERF_SAMPLE_CALLCHAIN = 1U << 5, PERF_SAMPLE_ID = 1U << 6, PERF_SAMPLE_CPU = 1U << 7, @@ -127,16 +127,32 @@ enum perf_counter_sample_format { }; /* - * Bits that can be set in attr.read_format to request that - * reads on the counter should return the indicated quantities, - * in increasing order of bit value, after the counter value. + * The format of the data returned by read() on a perf counter fd, + * as specified by attr.read_format: + * + * struct read_format { + * { u64 value; + * { u64 time_enabled; } && PERF_FORMAT_ENABLED + * { u64 time_running; } && PERF_FORMAT_RUNNING + * { u64 id; } && PERF_FORMAT_ID + * } && !PERF_FORMAT_GROUP + * + * { u64 nr; + * { u64 time_enabled; } && PERF_FORMAT_ENABLED + * { u64 time_running; } && PERF_FORMAT_RUNNING + * { u64 value; + * { u64 id; } && PERF_FORMAT_ID + * } cntr[nr]; + * } && PERF_FORMAT_GROUP + * }; */ enum perf_counter_read_format { PERF_FORMAT_TOTAL_TIME_ENABLED = 1U << 0, PERF_FORMAT_TOTAL_TIME_RUNNING = 1U << 1, PERF_FORMAT_ID = 1U << 2, + PERF_FORMAT_GROUP = 1U << 3, - PERF_FORMAT_MAX = 1U << 3, /* non-ABI */ + PERF_FORMAT_MAX = 1U << 4, /* non-ABI */ }; #define PERF_ATTR_SIZE_VER0 64 /* sizeof first published struct */ @@ -343,10 +359,8 @@ enum perf_event_type { * struct { * struct perf_event_header header; * u32 pid, tid; - * u64 value; - * { u64 time_enabled; } && PERF_FORMAT_ENABLED - * { u64 time_running; } && PERF_FORMAT_RUNNING - * { u64 parent_id; } && PERF_FORMAT_ID + * + * struct read_format values; * }; */ PERF_EVENT_READ = 8, @@ -364,11 +378,22 @@ enum perf_event_type { * { u32 cpu, res; } && PERF_SAMPLE_CPU * { u64 period; } && PERF_SAMPLE_PERIOD * - * { u64 nr; - * { u64 id, val; } cnt[nr]; } && PERF_SAMPLE_GROUP + * { struct read_format values; } && PERF_SAMPLE_READ * * { u64 nr, * u64 ips[nr]; } && PERF_SAMPLE_CALLCHAIN + * + * # + * # The RAW record below is opaque data wrt the ABI + * # + * # That is, the ABI doesn't make any promises wrt to + * # the stability of its content, it may vary depending + * # on event, hardware, kernel version and phase of + * # the moon. + * # + * # In other words, PERF_SAMPLE_RAW contents are not an ABI. + * # + * * { u32 size; * char data[size];}&& PERF_SAMPLE_RAW * }; diff --git a/kernel/perf_counter.c b/kernel/perf_counter.c index 3dd4339589a0..b8c6b97a20a3 100644 --- a/kernel/perf_counter.c +++ b/kernel/perf_counter.c @@ -1692,7 +1692,32 @@ static int perf_release(struct inode *inode, struct file *file) return 0; } -static u64 perf_counter_read_tree(struct perf_counter *counter) +static int perf_counter_read_size(struct perf_counter *counter) +{ + int entry = sizeof(u64); /* value */ + int size = 0; + int nr = 1; + + if (counter->attr.read_format & PERF_FORMAT_TOTAL_TIME_ENABLED) + size += sizeof(u64); + + if (counter->attr.read_format & PERF_FORMAT_TOTAL_TIME_RUNNING) + size += sizeof(u64); + + if (counter->attr.read_format & PERF_FORMAT_ID) + entry += sizeof(u64); + + if (counter->attr.read_format & PERF_FORMAT_GROUP) { + nr += counter->group_leader->nr_siblings; + size += sizeof(u64); + } + + size += entry * nr; + + return size; +} + +static u64 perf_counter_read_value(struct perf_counter *counter) { struct perf_counter *child; u64 total = 0; @@ -1704,14 +1729,96 @@ static u64 perf_counter_read_tree(struct perf_counter *counter) return total; } +static int perf_counter_read_entry(struct perf_counter *counter, + u64 read_format, char __user *buf) +{ + int n = 0, count = 0; + u64 values[2]; + + values[n++] = perf_counter_read_value(counter); + if (read_format & PERF_FORMAT_ID) + values[n++] = primary_counter_id(counter); + + count = n * sizeof(u64); + + if (copy_to_user(buf, values, count)) + return -EFAULT; + + return count; +} + +static int perf_counter_read_group(struct perf_counter *counter, + u64 read_format, char __user *buf) +{ + struct perf_counter *leader = counter->group_leader, *sub; + int n = 0, size = 0, err = -EFAULT; + u64 values[3]; + + values[n++] = 1 + leader->nr_siblings; + if (read_format & PERF_FORMAT_TOTAL_TIME_ENABLED) { + values[n++] = leader->total_time_enabled + + atomic64_read(&leader->child_total_time_enabled); + } + if (read_format & PERF_FORMAT_TOTAL_TIME_RUNNING) { + values[n++] = leader->total_time_running + + atomic64_read(&leader->child_total_time_running); + } + + size = n * sizeof(u64); + + if (copy_to_user(buf, values, size)) + return -EFAULT; + + err = perf_counter_read_entry(leader, read_format, buf + size); + if (err < 0) + return err; + + size += err; + + list_for_each_entry(sub, &leader->sibling_list, list_entry) { + err = perf_counter_read_entry(counter, read_format, + buf + size); + if (err < 0) + return err; + + size += err; + } + + return size; +} + +static int perf_counter_read_one(struct perf_counter *counter, + u64 read_format, char __user *buf) +{ + u64 values[4]; + int n = 0; + + values[n++] = perf_counter_read_value(counter); + if (read_format & PERF_FORMAT_TOTAL_TIME_ENABLED) { + values[n++] = counter->total_time_enabled + + atomic64_read(&counter->child_total_time_enabled); + } + if (read_format & PERF_FORMAT_TOTAL_TIME_RUNNING) { + values[n++] = counter->total_time_running + + atomic64_read(&counter->child_total_time_running); + } + if (read_format & PERF_FORMAT_ID) + values[n++] = primary_counter_id(counter); + + if (copy_to_user(buf, values, n * sizeof(u64))) + return -EFAULT; + + return n * sizeof(u64); +} + /* * Read the performance counter - simple non blocking version for now */ static ssize_t perf_read_hw(struct perf_counter *counter, char __user *buf, size_t count) { - u64 values[4]; - int n; + u64 read_format = counter->attr.read_format; + int ret; /* * Return end-of-file for a read on a counter that is in @@ -1721,28 +1828,18 @@ perf_read_hw(struct perf_counter *counter, char __user *buf, size_t count) if (counter->state == PERF_COUNTER_STATE_ERROR) return 0; + if (count < perf_counter_read_size(counter)) + return -ENOSPC; + WARN_ON_ONCE(counter->ctx->parent_ctx); mutex_lock(&counter->child_mutex); - values[0] = perf_counter_read_tree(counter); - n = 1; - if (counter->attr.read_format & PERF_FORMAT_TOTAL_TIME_ENABLED) - values[n++] = counter->total_time_enabled + - atomic64_read(&counter->child_total_time_enabled); - if (counter->attr.read_format & PERF_FORMAT_TOTAL_TIME_RUNNING) - values[n++] = counter->total_time_running + - atomic64_read(&counter->child_total_time_running); - if (counter->attr.read_format & PERF_FORMAT_ID) - values[n++] = primary_counter_id(counter); + if (read_format & PERF_FORMAT_GROUP) + ret = perf_counter_read_group(counter, read_format, buf); + else + ret = perf_counter_read_one(counter, read_format, buf); mutex_unlock(&counter->child_mutex); - if (count < n * sizeof(u64)) - return -EINVAL; - count = n * sizeof(u64); - - if (copy_to_user(buf, values, count)) - return -EFAULT; - - return count; + return ret; } static ssize_t @@ -2631,6 +2728,79 @@ static u32 perf_counter_tid(struct perf_counter *counter, struct task_struct *p) return task_pid_nr_ns(p, counter->ns); } +static void perf_output_read_one(struct perf_output_handle *handle, + struct perf_counter *counter) +{ + u64 read_format = counter->attr.read_format; + u64 values[4]; + int n = 0; + + values[n++] = atomic64_read(&counter->count); + if (read_format & PERF_FORMAT_TOTAL_TIME_ENABLED) { + values[n++] = counter->total_time_enabled + + atomic64_read(&counter->child_total_time_enabled); + } + if (read_format & PERF_FORMAT_TOTAL_TIME_RUNNING) { + values[n++] = counter->total_time_running + + atomic64_read(&counter->child_total_time_running); + } + if (read_format & PERF_FORMAT_ID) + values[n++] = primary_counter_id(counter); + + perf_output_copy(handle, values, n * sizeof(u64)); +} + +/* + * XXX PERF_FORMAT_GROUP vs inherited counters seems difficult. + */ +static void perf_output_read_group(struct perf_output_handle *handle, + struct perf_counter *counter) +{ + struct perf_counter *leader = counter->group_leader, *sub; + u64 read_format = counter->attr.read_format; + u64 values[5]; + int n = 0; + + values[n++] = 1 + leader->nr_siblings; + + if (read_format & PERF_FORMAT_TOTAL_TIME_ENABLED) + values[n++] = leader->total_time_enabled; + + if (read_format & PERF_FORMAT_TOTAL_TIME_RUNNING) + values[n++] = leader->total_time_running; + + if (leader != counter) + leader->pmu->read(leader); + + values[n++] = atomic64_read(&leader->count); + if (read_format & PERF_FORMAT_ID) + values[n++] = primary_counter_id(leader); + + perf_output_copy(handle, values, n * sizeof(u64)); + + list_for_each_entry(sub, &leader->sibling_list, list_entry) { + n = 0; + + if (sub != counter) + sub->pmu->read(sub); + + values[n++] = atomic64_read(&sub->count); + if (read_format & PERF_FORMAT_ID) + values[n++] = primary_counter_id(sub); + + perf_output_copy(handle, values, n * sizeof(u64)); + } +} + +static void perf_output_read(struct perf_output_handle *handle, + struct perf_counter *counter) +{ + if (counter->attr.read_format & PERF_FORMAT_GROUP) + perf_output_read_group(handle, counter); + else + perf_output_read_one(handle, counter); +} + void perf_counter_output(struct perf_counter *counter, int nmi, struct perf_sample_data *data) { @@ -2642,10 +2812,6 @@ void perf_counter_output(struct perf_counter *counter, int nmi, struct { u32 pid, tid; } tid_entry; - struct { - u64 id; - u64 counter; - } group_entry; struct perf_callchain_entry *callchain = NULL; int callchain_size = 0; u64 time; @@ -2700,10 +2866,8 @@ void perf_counter_output(struct perf_counter *counter, int nmi, if (sample_type & PERF_SAMPLE_PERIOD) header.size += sizeof(u64); - if (sample_type & PERF_SAMPLE_GROUP) { - header.size += sizeof(u64) + - counter->nr_siblings * sizeof(group_entry); - } + if (sample_type & PERF_SAMPLE_READ) + header.size += perf_counter_read_size(counter); if (sample_type & PERF_SAMPLE_CALLCHAIN) { callchain = perf_callchain(data->regs); @@ -2760,26 +2924,8 @@ void perf_counter_output(struct perf_counter *counter, int nmi, if (sample_type & PERF_SAMPLE_PERIOD) perf_output_put(&handle, data->period); - /* - * XXX PERF_SAMPLE_GROUP vs inherited counters seems difficult. - */ - if (sample_type & PERF_SAMPLE_GROUP) { - struct perf_counter *leader, *sub; - u64 nr = counter->nr_siblings; - - perf_output_put(&handle, nr); - - leader = counter->group_leader; - list_for_each_entry(sub, &leader->sibling_list, list_entry) { - if (sub != counter) - sub->pmu->read(sub); - - group_entry.id = primary_counter_id(sub); - group_entry.counter = atomic64_read(&sub->count); - - perf_output_put(&handle, group_entry); - } - } + if (sample_type & PERF_SAMPLE_READ) + perf_output_read(&handle, counter); if (sample_type & PERF_SAMPLE_CALLCHAIN) { if (callchain) @@ -2818,8 +2964,6 @@ struct perf_read_event { u32 pid; u32 tid; - u64 value; - u64 format[3]; }; static void @@ -2831,34 +2975,20 @@ perf_counter_read_event(struct perf_counter *counter, .header = { .type = PERF_EVENT_READ, .misc = 0, - .size = sizeof(event) - sizeof(event.format), + .size = sizeof(event) + perf_counter_read_size(counter), }, .pid = perf_counter_pid(counter, task), .tid = perf_counter_tid(counter, task), - .value = atomic64_read(&counter->count), }; - int ret, i = 0; - - if (counter->attr.read_format & PERF_FORMAT_TOTAL_TIME_ENABLED) { - event.header.size += sizeof(u64); - event.format[i++] = counter->total_time_enabled; - } - - if (counter->attr.read_format & PERF_FORMAT_TOTAL_TIME_RUNNING) { - event.header.size += sizeof(u64); - event.format[i++] = counter->total_time_running; - } - - if (counter->attr.read_format & PERF_FORMAT_ID) { - event.header.size += sizeof(u64); - event.format[i++] = primary_counter_id(counter); - } + int ret; ret = perf_output_begin(&handle, counter, event.header.size, 0, 0); if (ret) return; - perf_output_copy(&handle, &event, event.header.size); + perf_output_put(&handle, event); + perf_output_read(&handle, counter); + perf_output_end(&handle); } @@ -3921,9 +4051,9 @@ perf_counter_alloc(struct perf_counter_attr *attr, atomic64_set(&hwc->period_left, hwc->sample_period); /* - * we currently do not support PERF_SAMPLE_GROUP on inherited counters + * we currently do not support PERF_FORMAT_GROUP on inherited counters */ - if (attr->inherit && (attr->sample_type & PERF_SAMPLE_GROUP)) + if (attr->inherit && (attr->read_format & PERF_FORMAT_GROUP)) goto done; switch (attr->type) { From 970892a9031a5dc7217bd394fb9d89fa75a4a7bc Mon Sep 17 00:00:00 2001 From: Peter Zijlstra Date: Thu, 13 Aug 2009 11:47:54 +0200 Subject: [PATCH 16/17] perf_counter: Fix an ipi-deadlock perf_pending_counter() is called from IRQ context and will call perf_counter_disable(), however perf_counter_disable() uses smp_call_function_single() which doesn't fancy being used with IRQs disabled due to IPI deadlocks. Fix this by making it use the local __perf_counter_disable() call and teaching the counter_sched_out() code about pending disables as well. This should cover the case where a counter migrates before the pending queue gets processed. Signed-off-by: Peter Zijlstra Cc: Corey J Ashford Cc: Paul Mackerras Cc: stephane eranian LKML-Reference: <20090813103655.244097721@chello.nl> Signed-off-by: Ingo Molnar --- kernel/perf_counter.c | 6 +++++- 1 file changed, 5 insertions(+), 1 deletion(-) diff --git a/kernel/perf_counter.c b/kernel/perf_counter.c index b8c6b97a20a3..3f841beefc7b 100644 --- a/kernel/perf_counter.c +++ b/kernel/perf_counter.c @@ -307,6 +307,10 @@ counter_sched_out(struct perf_counter *counter, return; counter->state = PERF_COUNTER_STATE_INACTIVE; + if (counter->pending_disable) { + counter->pending_disable = 0; + counter->state = PERF_COUNTER_STATE_OFF; + } counter->tstamp_stopped = ctx->time; counter->pmu->disable(counter); counter->oncpu = -1; @@ -2343,7 +2347,7 @@ static void perf_pending_counter(struct perf_pending_entry *entry) if (counter->pending_disable) { counter->pending_disable = 0; - perf_counter_disable(counter); + __perf_counter_disable(counter); } if (counter->pending_wakeup) { From 94d5d1b2d891f1fd5205f978246b7864d998b25c Mon Sep 17 00:00:00 2001 From: Peter Zijlstra Date: Thu, 13 Aug 2009 16:14:42 +0200 Subject: [PATCH 17/17] perf_counter: Report the cloning task as parent on perf_counter_fork() A bug in (9f498cc: perf_counter: Full task tracing) makes profiling multi-threaded apps it go belly up. [ output as: (PID:TID):(PPID:PTID) ] # ./perf report -D | grep FORK 0x4b0 [0x18]: PERF_EVENT_FORK: (3237:3237):(3236:3236) 0xa10 [0x18]: PERF_EVENT_FORK: (3237:3238):(3236:3236) 0xa70 [0x18]: PERF_EVENT_FORK: (3237:3239):(3236:3236) 0xad0 [0x18]: PERF_EVENT_FORK: (3237:3240):(3236:3236) 0xb18 [0x18]: PERF_EVENT_FORK: (3237:3241):(3236:3236) Shows us that the test (27d028d perf report: Update for the new FORK/EXIT events) in builtin-report.c: /* * A thread clone will have the same PID for both * parent and child. */ if (thread == parent) return 0; Will clearly fail. The problem is that perf_counter_fork() reports the actual parent, instead of the cloning thread. Fixing that (with the below patch), yields: # ./perf report -D | grep FORK 0x4c8 [0x18]: PERF_EVENT_FORK: (1590:1590):(1589:1589) 0xbd8 [0x18]: PERF_EVENT_FORK: (1590:1591):(1590:1590) 0xc80 [0x18]: PERF_EVENT_FORK: (1590:1592):(1590:1590) 0x3338 [0x18]: PERF_EVENT_FORK: (1590:1593):(1590:1590) 0x66b0 [0x18]: PERF_EVENT_FORK: (1590:1594):(1590:1590) Which both makes more sense and doesn't confuse perf report anymore. Reported-by: Pekka Enberg Signed-off-by: Peter Zijlstra Cc: Arnaldo Carvalho de Melo Cc: paulus@samba.org Cc: Anton Blanchard Cc: Arjan van de Ven LKML-Reference: <1250172882.5241.62.camel@twins> Signed-off-by: Ingo Molnar --- kernel/perf_counter.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/kernel/perf_counter.c b/kernel/perf_counter.c index 3f841beefc7b..534e20d14d63 100644 --- a/kernel/perf_counter.c +++ b/kernel/perf_counter.c @@ -3028,10 +3028,10 @@ static void perf_counter_task_output(struct perf_counter *counter, return; task_event->event.pid = perf_counter_pid(counter, task); - task_event->event.ppid = perf_counter_pid(counter, task->real_parent); + task_event->event.ppid = perf_counter_pid(counter, current); task_event->event.tid = perf_counter_tid(counter, task); - task_event->event.ptid = perf_counter_tid(counter, task->real_parent); + task_event->event.ptid = perf_counter_tid(counter, current); perf_output_put(&handle, task_event->event); perf_output_end(&handle);