Merge branch 'perf/core' of git://git.kernel.org/pub/scm/linux/kernel/git/acme/linux-2.6 into perf/core

tools/perf/Documentation/perf-probe.txt

@@ -73,6 +73,10 @@ OPTIONS
 	(Only for --vars) Show external defined variables in addition to local
 	variables.
 
+-F::
+--funcs::
+	Show available functions in given module or kernel.
+
 -f::
 --force::
 	Forcibly add events with existing name.

tools/perf/Makefile

@@ -402,6 +402,7 @@ LIB_H += util/debug.h
 LIB_H += util/debugfs.h
 LIB_H += util/event.h
 LIB_H += util/evsel.h
+LIB_H += util/evlist.h
 LIB_H += util/exec_cmd.h
 LIB_H += util/types.h
 LIB_H += util/levenshtein.h
@@ -425,6 +426,7 @@ LIB_H += util/values.h
 LIB_H += util/sort.h
 LIB_H += util/hist.h
 LIB_H += util/thread.h
+LIB_H += util/thread_map.h
 LIB_H += util/trace-event.h
 LIB_H += util/probe-finder.h
 LIB_H += util/probe-event.h
@@ -440,6 +442,7 @@ LIB_OBJS += $(OUTPUT)util/ctype.o
 LIB_OBJS += $(OUTPUT)util/debugfs.o
 LIB_OBJS += $(OUTPUT)util/environment.o
 LIB_OBJS += $(OUTPUT)util/event.o
+LIB_OBJS += $(OUTPUT)util/evlist.o
 LIB_OBJS += $(OUTPUT)util/evsel.o
 LIB_OBJS += $(OUTPUT)util/exec_cmd.o
 LIB_OBJS += $(OUTPUT)util/help.o
@@ -469,6 +472,7 @@ LIB_OBJS += $(OUTPUT)util/map.o
 LIB_OBJS += $(OUTPUT)util/pstack.o
 LIB_OBJS += $(OUTPUT)util/session.o
 LIB_OBJS += $(OUTPUT)util/thread.o
+LIB_OBJS += $(OUTPUT)util/thread_map.o
 LIB_OBJS += $(OUTPUT)util/trace-event-parse.o
 LIB_OBJS += $(OUTPUT)util/trace-event-read.o
 LIB_OBJS += $(OUTPUT)util/trace-event-info.o

tools/perf/builtin-probe.c

@@ -52,6 +52,7 @@ static struct {
 	bool show_lines;
 	bool show_vars;
 	bool show_ext_vars;
+	bool show_funcs;
 	bool mod_events;
 	int nevents;
 	struct perf_probe_event events[MAX_PROBES];
@@ -221,6 +222,8 @@ static const struct option options[] = {
 	OPT__DRY_RUN(&probe_event_dry_run),
 	OPT_INTEGER('\0', "max-probes", &params.max_probe_points,
 		 "Set how many probe points can be found for a probe."),
+	OPT_BOOLEAN('F', "funcs", &params.show_funcs,
+		    "Show potential probe-able functions."),
 	OPT_END()
 };
 
@@ -246,7 +249,7 @@ int cmd_probe(int argc, const char **argv, const char *prefix __used)
 		params.max_probe_points = MAX_PROBES;
 
 	if ((!params.nevents && !params.dellist && !params.list_events &&
-	     !params.show_lines))
+	     !params.show_lines && !params.show_funcs))
 		usage_with_options(probe_usage, options);
 
 	/*
@@ -267,12 +270,36 @@ int cmd_probe(int argc, const char **argv, const char *prefix __used)
 			pr_err(" Error: Don't use --list with --vars.\n");
 			usage_with_options(probe_usage, options);
 		}
+		if (params.show_funcs) {
+			pr_err("  Error: Don't use --list with --funcs.\n");
+			usage_with_options(probe_usage, options);
+		}
 		ret = show_perf_probe_events();
 		if (ret < 0)
 			pr_err("  Error: Failed to show event list. (%d)\n",
 			       ret);
 		return ret;
 	}
+	if (params.show_funcs) {
+		if (params.nevents != 0 || params.dellist) {
+			pr_err("  Error: Don't use --funcs with"
+			       " --add/--del.\n");
+			usage_with_options(probe_usage, options);
+		}
+		if (params.show_lines) {
+			pr_err("  Error: Don't use --funcs with --line.\n");
+			usage_with_options(probe_usage, options);
+		}
+		if (params.show_vars) {
+			pr_err("  Error: Don't use --funcs with --vars.\n");
+			usage_with_options(probe_usage, options);
+		}
+		ret = show_available_funcs(params.target_module);
+		if (ret < 0)
+			pr_err("  Error: Failed to show functions."
+			       " (%d)\n", ret);
+		return ret;
+	}
 
 #ifdef DWARF_SUPPORT
 	if (params.show_lines) {

tools/perf/builtin-record.c

@@ -18,17 +18,20 @@
 
 #include "util/header.h"
 #include "util/event.h"
+#include "util/evlist.h"
 #include "util/evsel.h"
 #include "util/debug.h"
 #include "util/session.h"
 #include "util/symbol.h"
 #include "util/cpumap.h"
+#include "util/thread_map.h"
 
 #include <unistd.h>
 #include <sched.h>
 #include <sys/mman.h>
 
 #define FD(e, x, y) (*(int *)xyarray__entry(e->fd, x, y))
+#define SID(e, x, y) xyarray__entry(e->id, x, y)
 
 enum write_mode_t {
 	WRITE_FORCE,
@@ -46,7 +49,7 @@ static unsigned int		user_freq 			= UINT_MAX;
 static int			freq				=   1000;
 static int			output;
 static int			pipe_output			=      0;
-static const char		*output_name			= "perf.data";
+static const char		*output_name			= NULL;
 static int			group				=      0;
 static int			realtime_prio			=      0;
 static bool			nodelay				=  false;
@@ -66,51 +69,17 @@ static bool			sample_address			=  false;
 static bool			sample_time			=  false;
 static bool			no_buildid			=  false;
 static bool			no_buildid_cache		=  false;
+static struct perf_evlist	*evsel_list;
 
 static long			samples				=      0;
 static u64			bytes_written			=      0;
 
-static struct pollfd		*event_array;
-
-static int			nr_poll				=      0;
-static int			nr_cpu				=      0;
-
 static int			file_new			=      1;
 static off_t			post_processing_offset;
 
 static struct perf_session	*session;
 static const char		*cpu_list;
 
-struct mmap_data {
-	void			*base;
-	unsigned int		mask;
-	unsigned int		prev;
-};
-
-static struct mmap_data		mmap_array[MAX_NR_CPUS];
-
-static unsigned long mmap_read_head(struct mmap_data *md)
-{
-	struct perf_event_mmap_page *pc = md->base;
-	long head;
-
-	head = pc->data_head;
-	rmb();
-
-	return head;
-}
-
-static void mmap_write_tail(struct mmap_data *md, unsigned long tail)
-{
-	struct perf_event_mmap_page *pc = md->base;
-
-	/*
-	 * ensure all reads are done before we write the tail out.
-	 */
-	/* mb(); */
-	pc->data_tail = tail;
-}
-
 static void advance_output(size_t size)
 {
 	bytes_written += size;
@@ -139,9 +108,9 @@ static int process_synthesized_event(event_t *event,
 	return 0;
 }
 
-static void mmap_read(struct mmap_data *md)
+static void mmap_read(struct perf_mmap *md)
 {
-	unsigned int head = mmap_read_head(md);
+	unsigned int head = perf_mmap__read_head(md);
 	unsigned int old = md->prev;
 	unsigned char *data = md->base + page_size;
 	unsigned long size;
@@ -185,7 +154,7 @@ static void mmap_read(struct mmap_data *md)
 	write_output(buf, size);
 
 	md->prev = old;
-	mmap_write_tail(md, old);
+	perf_mmap__write_tail(md, old);
 }
 
 static volatile int done = 0;
@@ -209,8 +178,6 @@ static void sig_atexit(void)
 	kill(getpid(), signr);
 }
 
-static int group_fd;
-
 static struct perf_header_attr *get_header_attr(struct perf_event_attr *a, int nr)
 {
 	struct perf_header_attr *h_attr;
@@ -234,28 +201,47 @@ static void create_counter(struct perf_evsel *evsel, int cpu)
 	char *filter = evsel->filter;
 	struct perf_event_attr *attr = &evsel->attr;
 	struct perf_header_attr *h_attr;
-	int track = !evsel->idx; /* only the first counter needs these */
+	struct perf_sample_id *sid;
 	int thread_index;
 	int ret;
-	struct {
-		u64 count;
-		u64 time_enabled;
-		u64 time_running;
-		u64 id;
-	} read_data;
-	/*
- 	 * Check if parse_single_tracepoint_event has already asked for
- 	 * PERF_SAMPLE_TIME.
- 	 *
-	 * XXX this is kludgy but short term fix for problems introduced by
-	 * eac23d1c that broke 'perf script' by having different sample_types
-	 * when using multiple tracepoint events when we use a perf binary
-	 * that tries to use sample_id_all on an older kernel.
- 	 *
- 	 * We need to move counter creation to perf_session, support
- 	 * different sample_types, etc.
- 	 */
-	bool time_needed = attr->sample_type & PERF_SAMPLE_TIME;
+
+	for (thread_index = 0; thread_index < threads->nr; thread_index++) {
+		h_attr = get_header_attr(attr, evsel->idx);
+		if (h_attr == NULL)
+			die("nomem\n");
+
+		if (!file_new) {
+			if (memcmp(&h_attr->attr, attr, sizeof(*attr))) {
+				fprintf(stderr, "incompatible append\n");
+				exit(-1);
+			}
+		}
+
+		sid = SID(evsel, cpu, thread_index);
+		if (perf_header_attr__add_id(h_attr, sid->id) < 0) {
+			pr_warning("Not enough memory to add id\n");
+			exit(-1);
+		}
+
+		if (filter != NULL) {
+			ret = ioctl(FD(evsel, cpu, thread_index),
+				    PERF_EVENT_IOC_SET_FILTER, filter);
+			if (ret) {
+				error("failed to set filter with %d (%s)\n", errno,
+						strerror(errno));
+				exit(-1);
+			}
+		}
+	}
+
+	if (!sample_type)
+		sample_type = attr->sample_type;
+}
+
+static void config_attr(struct perf_evsel *evsel, struct perf_evlist *evlist)
+{
+	struct perf_event_attr *attr = &evsel->attr;
+	int track = !evsel->idx; /* only the first counter needs these */
 
 	attr->read_format	= PERF_FORMAT_TOTAL_TIME_ENABLED |
 				  PERF_FORMAT_TOTAL_TIME_RUNNING |
@@ -263,7 +249,7 @@ static void create_counter(struct perf_evsel *evsel, int cpu)
 
 	attr->sample_type	|= PERF_SAMPLE_IP | PERF_SAMPLE_TID;
 
-	if (nr_counters > 1)
+	if (evlist->nr_entries > 1)
 		attr->sample_type |= PERF_SAMPLE_ID;
 
 	/*
@@ -315,19 +301,39 @@ static void create_counter(struct perf_evsel *evsel, int cpu)
 
 	attr->mmap		= track;
 	attr->comm		= track;
-	attr->inherit		= !no_inherit;
+
 	if (target_pid == -1 && target_tid == -1 && !system_wide) {
 		attr->disabled = 1;
 		attr->enable_on_exec = 1;
 	}
-retry_sample_id:
-	attr->sample_id_all = sample_id_all_avail ? 1 : 0;
+}
 
-	for (thread_index = 0; thread_index < threads->nr; thread_index++) {
-try_again:
-		FD(evsel, nr_cpu, thread_index) = sys_perf_event_open(attr, threads->map[thread_index], cpu, group_fd, 0);
+static void open_counters(struct perf_evlist *evlist)
+{
+	struct perf_evsel *pos;
+	int cpu;
 
-		if (FD(evsel, nr_cpu, thread_index) < 0) {
+	list_for_each_entry(pos, &evlist->entries, node) {
+		struct perf_event_attr *attr = &pos->attr;
+		/*
+		 * Check if parse_single_tracepoint_event has already asked for
+		 * PERF_SAMPLE_TIME.
+		 *
+		 * XXX this is kludgy but short term fix for problems introduced by
+		 * eac23d1c that broke 'perf script' by having different sample_types
+		 * when using multiple tracepoint events when we use a perf binary
+		 * that tries to use sample_id_all on an older kernel.
+		 *
+		 * We need to move counter creation to perf_session, support
+		 * different sample_types, etc.
+		 */
+		bool time_needed = attr->sample_type & PERF_SAMPLE_TIME;
+
+		config_attr(pos, evlist);
+retry_sample_id:
+		attr->sample_id_all = sample_id_all_avail ? 1 : 0;
+try_again:
+		if (perf_evsel__open(pos, cpus, threads, group, !no_inherit) < 0) {
 			int err = errno;
 
 			if (err == EPERM || err == EACCES)
@@ -364,7 +370,7 @@ try_again:
 			}
 			printf("\n");
 			error("sys_perf_event_open() syscall returned with %d (%s).  /bin/dmesg may provide additional information.\n",
-			      FD(evsel, nr_cpu, thread_index), strerror(err));
+			      err, strerror(err));
 
 #if defined(__i386__) || defined(__x86_64__)
 			if (attr->type == PERF_TYPE_HARDWARE && err == EOPNOTSUPP)
@@ -375,90 +381,16 @@ try_again:
 #endif
 
 			die("No CONFIG_PERF_EVENTS=y kernel support configured?\n");
-			exit(-1);
-		}
-
-		h_attr = get_header_attr(attr, evsel->idx);
-		if (h_attr == NULL)
-			die("nomem\n");
-
-		if (!file_new) {
-			if (memcmp(&h_attr->attr, attr, sizeof(*attr))) {
-				fprintf(stderr, "incompatible append\n");
-				exit(-1);
-			}
-		}
-
-		if (read(FD(evsel, nr_cpu, thread_index), &read_data, sizeof(read_data)) == -1) {
-			perror("Unable to read perf file descriptor");
-			exit(-1);
-		}
-
-		if (perf_header_attr__add_id(h_attr, read_data.id) < 0) {
-			pr_warning("Not enough memory to add id\n");
-			exit(-1);
-		}
-
-		assert(FD(evsel, nr_cpu, thread_index) >= 0);
-		fcntl(FD(evsel, nr_cpu, thread_index), F_SETFL, O_NONBLOCK);
-
-		/*
-		 * First counter acts as the group leader:
-		 */
-		if (group && group_fd == -1)
-			group_fd = FD(evsel, nr_cpu, thread_index);
-
-		if (evsel->idx || thread_index) {
-			struct perf_evsel *first;
-			first = list_entry(evsel_list.next, struct perf_evsel, node);
-			ret = ioctl(FD(evsel, nr_cpu, thread_index),
-				    PERF_EVENT_IOC_SET_OUTPUT,
-				    FD(first, nr_cpu, 0));
-			if (ret) {
-				error("failed to set output: %d (%s)\n", errno,
-						strerror(errno));
-				exit(-1);
-			}
-		} else {
-			mmap_array[nr_cpu].prev = 0;
-			mmap_array[nr_cpu].mask = mmap_pages*page_size - 1;
-			mmap_array[nr_cpu].base = mmap(NULL, (mmap_pages+1)*page_size,
-				PROT_READ | PROT_WRITE, MAP_SHARED, FD(evsel, nr_cpu, thread_index), 0);
-			if (mmap_array[nr_cpu].base == MAP_FAILED) {
-				error("failed to mmap with %d (%s)\n", errno, strerror(errno));
-				exit(-1);
-			}
-
-			event_array[nr_poll].fd = FD(evsel, nr_cpu, thread_index);
-			event_array[nr_poll].events = POLLIN;
-			nr_poll++;
-		}
-
-		if (filter != NULL) {
-			ret = ioctl(FD(evsel, nr_cpu, thread_index),
-				    PERF_EVENT_IOC_SET_FILTER, filter);
-			if (ret) {
-				error("failed to set filter with %d (%s)\n", errno,
-						strerror(errno));
-				exit(-1);
-			}
 		}
 	}
 
-	if (!sample_type)
-		sample_type = attr->sample_type;
-}
-
-static void open_counters(int cpu)
-{
-	struct perf_evsel *pos;
-
-	group_fd = -1;
-
-	list_for_each_entry(pos, &evsel_list, node)
-		create_counter(pos, cpu);
+	if (perf_evlist__mmap(evlist, cpus, threads, mmap_pages, false) < 0)
+		die("failed to mmap with %d (%s)\n", errno, strerror(errno));
 
-	nr_cpu++;
+	for (cpu = 0; cpu < cpus->nr; ++cpu) {
+		list_for_each_entry(pos, &evlist->entries, node)
+			create_counter(pos, cpu);
+	}
 }
 
 static int process_buildids(void)
@@ -481,9 +413,9 @@ static void atexit_header(void)
 
 		if (!no_buildid)
 			process_buildids();
-		perf_header__write(&session->header, output, true);
+		perf_header__write(&session->header, evsel_list, output, true);
 		perf_session__delete(session);
-		perf_evsel_list__delete();
+		perf_evlist__delete(evsel_list);
 		symbol__exit();
 	}
 }
@@ -533,9 +465,9 @@ static void mmap_read_all(void)
 {
 	int i;
 
-	for (i = 0; i < nr_cpu; i++) {
-		if (mmap_array[i].base)
-			mmap_read(&mmap_array[i]);
+	for (i = 0; i < cpus->nr; i++) {
+		if (evsel_list->mmap[i].base)
+			mmap_read(&evsel_list->mmap[i]);
 	}
 
 	if (perf_header__has_feat(&session->header, HEADER_TRACE_INFO))
@@ -566,18 +498,26 @@ static int __cmd_record(int argc, const char **argv)
 		exit(-1);
 	}
 
-	if (!strcmp(output_name, "-"))
-		pipe_output = 1;
-	else if (!stat(output_name, &st) && st.st_size) {
-		if (write_mode == WRITE_FORCE) {
-			char oldname[PATH_MAX];
-			snprintf(oldname, sizeof(oldname), "%s.old",
-				 output_name);
-			unlink(oldname);
-			rename(output_name, oldname);
+	if (!output_name) {
+		if (!fstat(STDOUT_FILENO, &st) && S_ISFIFO(st.st_mode))
+			pipe_output = 1;
+		else
+			output_name = "perf.data";
+	}
+	if (output_name) {
+		if (!strcmp(output_name, "-"))
+			pipe_output = 1;
+		else if (!stat(output_name, &st) && st.st_size) {
+			if (write_mode == WRITE_FORCE) {
+				char oldname[PATH_MAX];
+				snprintf(oldname, sizeof(oldname), "%s.old",
+					 output_name);
+				unlink(oldname);
+				rename(output_name, oldname);
+			}
+		} else if (write_mode == WRITE_APPEND) {
+			write_mode = WRITE_FORCE;
 		}
-	} else if (write_mode == WRITE_APPEND) {
-		write_mode = WRITE_FORCE;
 	}
 
 	flags = O_CREAT|O_RDWR;
@@ -611,7 +551,7 @@ static int __cmd_record(int argc, const char **argv)
 			goto out_delete_session;
 	}
 
-	if (have_tracepoints(&evsel_list))
+	if (have_tracepoints(&evsel_list->entries))
 		perf_header__set_feat(&session->header, HEADER_TRACE_INFO);
 
 	/*
@@ -673,12 +613,7 @@ static int __cmd_record(int argc, const char **argv)
 		close(child_ready_pipe[0]);
 	}
 
-	if (!system_wide && no_inherit && !cpu_list) {
-		open_counters(-1);
-	} else {
-		for (i = 0; i < cpus->nr; i++)
-			open_counters(cpus->map[i]);
-	}
+	open_counters(evsel_list);
 
 	perf_session__set_sample_type(session, sample_type);
 
@@ -687,7 +622,8 @@ static int __cmd_record(int argc, const char **argv)
 		if (err < 0)
 			return err;
 	} else if (file_new) {
-		err = perf_header__write(&session->header, output, false);
+		err = perf_header__write(&session->header, evsel_list,
+					 output, false);
 		if (err < 0)
 			return err;
 	}
@@ -712,7 +648,7 @@ static int __cmd_record(int argc, const char **argv)
 			return err;
 		}
 
-		if (have_tracepoints(&evsel_list)) {
+		if (have_tracepoints(&evsel_list->entries)) {
 			/*
 			 * FIXME err <= 0 here actually means that
 			 * there were no tracepoints so its not really
@@ -721,7 +657,7 @@ static int __cmd_record(int argc, const char **argv)
 			 * return this more properly and also
 			 * propagate errors that now are calling die()
 			 */
-			err = event__synthesize_tracing_data(output, &evsel_list,
+			err = event__synthesize_tracing_data(output, evsel_list,
 							     process_synthesized_event,
 							     session);
 			if (err <= 0) {
@@ -789,15 +725,15 @@ static int __cmd_record(int argc, const char **argv)
 		if (hits == samples) {
 			if (done)
 				break;
-			err = poll(event_array, nr_poll, -1);
+			err = poll(evsel_list->pollfd, evsel_list->nr_fds, -1);
 			waking++;
 		}
 
 		if (done) {
-			for (i = 0; i < nr_cpu; i++) {
+			for (i = 0; i < cpus->nr; i++) {
 				struct perf_evsel *pos;
 
-				list_for_each_entry(pos, &evsel_list, node) {
+				list_for_each_entry(pos, &evsel_list->entries, node) {
 					for (thread = 0;
 						thread < threads->nr;
 						thread++)
@@ -838,10 +774,10 @@ static const char * const record_usage[] = {
 static bool force, append_file;
 
 const struct option record_options[] = {
-	OPT_CALLBACK('e', "event", NULL, "event",
+	OPT_CALLBACK('e', "event", &evsel_list, "event",
 		     "event selector. use 'perf list' to list available events",
 		     parse_events),
-	OPT_CALLBACK(0, "filter", NULL, "filter",
+	OPT_CALLBACK(0, "filter", &evsel_list, "filter",
 		     "event filter", parse_filter),
 	OPT_INTEGER('p', "pid", &target_pid,
 		    "record events on existing process id"),
@@ -892,6 +828,10 @@ int cmd_record(int argc, const char **argv, const char *prefix __used)
 	int err = -ENOMEM;
 	struct perf_evsel *pos;
 
+	evsel_list = perf_evlist__new();
+	if (evsel_list == NULL)
+		return -ENOMEM;
+
 	argc = parse_options(argc, argv, record_options, record_usage,
 			    PARSE_OPT_STOP_AT_NON_OPTION);
 	if (!argc && target_pid == -1 && target_tid == -1 &&
@@ -913,7 +853,8 @@ int cmd_record(int argc, const char **argv, const char *prefix __used)
 	if (no_buildid_cache || no_buildid)
 		disable_buildid_cache();
 
-	if (list_empty(&evsel_list) && perf_evsel_list__create_default() < 0) {
+	if (evsel_list->nr_entries == 0 &&
+	    perf_evlist__add_default(evsel_list) < 0) {
 		pr_err("Not enough memory for event selector list\n");
 		goto out_symbol_exit;
 	}
@@ -927,21 +868,22 @@ int cmd_record(int argc, const char **argv, const char *prefix __used)
 		usage_with_options(record_usage, record_options);
 	}
 
-	cpus = cpu_map__new(cpu_list);
-	if (cpus == NULL) {
-		perror("failed to parse CPUs map");
-		return -1;
-	}
+	if (target_tid != -1)
+		cpus = cpu_map__dummy_new();
+	else
+		cpus = cpu_map__new(cpu_list);
 
-	list_for_each_entry(pos, &evsel_list, node) {
+	if (cpus == NULL)
+		usage_with_options(record_usage, record_options);
+
+	list_for_each_entry(pos, &evsel_list->entries, node) {
 		if (perf_evsel__alloc_fd(pos, cpus->nr, threads->nr) < 0)
 			goto out_free_fd;
 		if (perf_header__push_event(pos->attr.config, event_name(pos)))
 			goto out_free_fd;
 	}
-	event_array = malloc((sizeof(struct pollfd) * MAX_NR_CPUS *
-			      MAX_COUNTERS * threads->nr));
-	if (!event_array)
+
+	if (perf_evlist__alloc_pollfd(evsel_list, cpus->nr, threads->nr) < 0)
 		goto out_free_fd;
 
 	if (user_interval != ULLONG_MAX)
@@ -959,13 +901,11 @@ int cmd_record(int argc, const char **argv, const char *prefix __used)
 	} else {
 		fprintf(stderr, "frequency and count are zero, aborting\n");
 		err = -EINVAL;
-		goto out_free_event_array;
+		goto out_free_fd;
 	}
 
 	err = __cmd_record(argc, argv);
 
-out_free_event_array:
-	free(event_array);
 out_free_fd:
 	thread_map__delete(threads);
 	threads = NULL;

tools/perf/builtin-report.c

@@ -81,18 +81,17 @@ static int perf_session__add_hist_entry(struct perf_session *self,
 					struct addr_location *al,
 					struct sample_data *data)
 {
-	struct map_symbol *syms = NULL;
 	struct symbol *parent = NULL;
-	int err = -ENOMEM;
+	int err = 0;
 	struct hist_entry *he;
 	struct hists *hists;
 	struct perf_event_attr *attr;
 
 	if ((sort__has_parent || symbol_conf.use_callchain) && data->callchain) {
-		syms = perf_session__resolve_callchain(self, al->thread,
-						       data->callchain, &parent);
-		if (syms == NULL)
-			return -ENOMEM;
+		err = perf_session__resolve_callchain(self, al->thread,
+						      data->callchain, &parent);
+		if (err)
+			return err;
 	}
 
 	attr = perf_header__find_attr(data->id, &self->header);
@@ -101,16 +100,17 @@ static int perf_session__add_hist_entry(struct perf_session *self,
 	else
 		hists = perf_session__hists_findnew(self, data->id, 0, 0);
 	if (hists == NULL)
-		goto out_free_syms;
+		return -ENOMEM;
+
 	he = __hists__add_entry(hists, al, parent, data->period);
 	if (he == NULL)
-		goto out_free_syms;
-	err = 0;
+		return -ENOMEM;
+
 	if (symbol_conf.use_callchain) {
-		err = callchain_append(he->callchain, data->callchain, syms,
+		err = callchain_append(he->callchain, &self->callchain_cursor,
 				       data->period);
 		if (err)
-			goto out_free_syms;
+			return err;
 	}
 	/*
 	 * Only in the newt browser we are doing integrated annotation,
@@ -119,8 +119,7 @@ static int perf_session__add_hist_entry(struct perf_session *self,
 	 */
 	if (use_browser > 0)
 		err = hist_entry__inc_addr_samples(he, al->addr);
-out_free_syms:
-	free(syms);
+
 	return err;
 }
 
@@ -222,7 +221,7 @@ static int perf_session__setup_sample_type(struct perf_session *self)
 	} else if (!dont_use_callchains && callchain_param.mode != CHAIN_NONE &&
 		   !symbol_conf.use_callchain) {
 			symbol_conf.use_callchain = true;
-			if (register_callchain_param(&callchain_param) < 0) {
+			if (callchain_register_param(&callchain_param) < 0) {
 				fprintf(stderr, "Can't register callchain"
 						" params\n");
 				return -EINVAL;
@@ -424,7 +423,7 @@ parse_callchain_opt(const struct option *opt __used, const char *arg,
 	if (tok2)
 		callchain_param.print_limit = strtod(tok2, &endptr);
 setup:
-	if (register_callchain_param(&callchain_param) < 0) {
+	if (callchain_register_param(&callchain_param) < 0) {
 		fprintf(stderr, "Can't register callchain params\n");
 		return -1;
 	}

tools/perf/builtin-stat.c

@@ -43,11 +43,13 @@
 #include "util/parse-options.h"
 #include "util/parse-events.h"
 #include "util/event.h"
+#include "util/evlist.h"
 #include "util/evsel.h"
 #include "util/debug.h"
 #include "util/header.h"
 #include "util/cpumap.h"
 #include "util/thread.h"
+#include "util/thread_map.h"
 
 #include <sys/prctl.h>
 #include <math.h>
@@ -71,6 +73,8 @@ static struct perf_event_attr default_attrs[] = {
 
 };
 
+struct perf_evlist		*evsel_list;
+
 static bool			system_wide			=  false;
 static struct cpu_map		*cpus;
 static int			run_idx				=  0;
@@ -166,7 +170,7 @@ static int create_perf_stat_counter(struct perf_evsel *evsel)
 				    PERF_FORMAT_TOTAL_TIME_RUNNING;
 
 	if (system_wide)
-		return perf_evsel__open_per_cpu(evsel, cpus);
+		return perf_evsel__open_per_cpu(evsel, cpus, false, false);
 
 	attr->inherit = !no_inherit;
 	if (target_pid == -1 && target_tid == -1) {
@@ -174,7 +178,7 @@ static int create_perf_stat_counter(struct perf_evsel *evsel)
 		attr->enable_on_exec = 1;
 	}
 
-	return perf_evsel__open_per_thread(evsel, threads);
+	return perf_evsel__open_per_thread(evsel, threads, false, false);
 }
 
 /*
@@ -309,7 +313,7 @@ static int run_perf_stat(int argc __used, const char **argv)
 		close(child_ready_pipe[0]);
 	}
 
-	list_for_each_entry(counter, &evsel_list, node) {
+	list_for_each_entry(counter, &evsel_list->entries, node) {
 		if (create_perf_stat_counter(counter) < 0) {
 			if (errno == -EPERM || errno == -EACCES) {
 				error("You may not have permission to collect %sstats.\n"
@@ -347,12 +351,12 @@ static int run_perf_stat(int argc __used, const char **argv)
 	update_stats(&walltime_nsecs_stats, t1 - t0);
 
 	if (no_aggr) {
-		list_for_each_entry(counter, &evsel_list, node) {
+		list_for_each_entry(counter, &evsel_list->entries, node) {
 			read_counter(counter);
 			perf_evsel__close_fd(counter, cpus->nr, 1);
 		}
 	} else {
-		list_for_each_entry(counter, &evsel_list, node) {
+		list_for_each_entry(counter, &evsel_list->entries, node) {
 			read_counter_aggr(counter);
 			perf_evsel__close_fd(counter, cpus->nr, threads->nr);
 		}
@@ -555,10 +559,10 @@ static void print_stat(int argc, const char **argv)
 	}
 
 	if (no_aggr) {
-		list_for_each_entry(counter, &evsel_list, node)
+		list_for_each_entry(counter, &evsel_list->entries, node)
 			print_counter(counter);
 	} else {
-		list_for_each_entry(counter, &evsel_list, node)
+		list_for_each_entry(counter, &evsel_list->entries, node)
 			print_counter_aggr(counter);
 	}
 
@@ -610,7 +614,7 @@ static int stat__set_big_num(const struct option *opt __used,
 }
 
 static const struct option options[] = {
-	OPT_CALLBACK('e', "event", NULL, "event",
+	OPT_CALLBACK('e', "event", &evsel_list, "event",
 		     "event selector. use 'perf list' to list available events",
 		     parse_events),
 	OPT_BOOLEAN('i', "no-inherit", &no_inherit,
@@ -648,6 +652,10 @@ int cmd_stat(int argc, const char **argv, const char *prefix __used)
 
 	setlocale(LC_ALL, "");
 
+	evsel_list = perf_evlist__new();
+	if (evsel_list == NULL)
+		return -ENOMEM;
+
 	argc = parse_options(argc, argv, options, stat_usage,
 		PARSE_OPT_STOP_AT_NON_OPTION);
 
@@ -679,17 +687,14 @@ int cmd_stat(int argc, const char **argv, const char *prefix __used)
 		usage_with_options(stat_usage, options);
 
 	/* Set attrs and nr_counters if no event is selected and !null_run */
-	if (!null_run && !nr_counters) {
+	if (!null_run && !evsel_list->nr_entries) {
 		size_t c;
 
-		nr_counters = ARRAY_SIZE(default_attrs);
-
 		for (c = 0; c < ARRAY_SIZE(default_attrs); ++c) {
-			pos = perf_evsel__new(&default_attrs[c],
-					      nr_counters);
+			pos = perf_evsel__new(&default_attrs[c], c);
 			if (pos == NULL)
 				goto out;
-			list_add(&pos->node, &evsel_list);
+			perf_evlist__add(evsel_list, pos);
 		}
 	}
 
@@ -713,7 +718,7 @@ int cmd_stat(int argc, const char **argv, const char *prefix __used)
 		return -1;
 	}
 
-	list_for_each_entry(pos, &evsel_list, node) {
+	list_for_each_entry(pos, &evsel_list->entries, node) {
 		if (perf_evsel__alloc_stat_priv(pos) < 0 ||
 		    perf_evsel__alloc_counts(pos, cpus->nr) < 0 ||
 		    perf_evsel__alloc_fd(pos, cpus->nr, threads->nr) < 0)
@@ -741,9 +746,9 @@ int cmd_stat(int argc, const char **argv, const char *prefix __used)
 	if (status != -1)
 		print_stat(argc, argv);
 out_free_fd:
-	list_for_each_entry(pos, &evsel_list, node)
+	list_for_each_entry(pos, &evsel_list->entries, node)
 		perf_evsel__free_stat_priv(pos);
-	perf_evsel_list__delete();
+	perf_evlist__delete(evsel_list);
 out:
 	thread_map__delete(threads);
 	threads = NULL;

tools/perf/builtin-test.c

@@ -7,10 +7,11 @@
 
 #include "util/cache.h"
 #include "util/debug.h"
+#include "util/evlist.h"
 #include "util/parse-options.h"
-#include "util/session.h"
+#include "util/parse-events.h"
 #include "util/symbol.h"
-#include "util/thread.h"
+#include "util/thread_map.h"
 
 static long page_size;
 
@@ -238,14 +239,14 @@ out:
 #include "util/evsel.h"
 #include <sys/types.h>
 
-static int trace_event__id(const char *event_name)
+static int trace_event__id(const char *evname)
 {
 	char *filename;
 	int err = -1, fd;
 
 	if (asprintf(&filename,
 		     "/sys/kernel/debug/tracing/events/syscalls/%s/id",
-		     event_name) < 0)
+		     evname) < 0)
 		return -1;
 
 	fd = open(filename, O_RDONLY);
@@ -289,7 +290,7 @@ static int test__open_syscall_event(void)
 		goto out_thread_map_delete;
 	}
 
-	if (perf_evsel__open_per_thread(evsel, threads) < 0) {
+	if (perf_evsel__open_per_thread(evsel, threads, false, false) < 0) {
 		pr_debug("failed to open counter: %s, "
 			 "tweak /proc/sys/kernel/perf_event_paranoid?\n",
 			 strerror(errno));
@@ -347,9 +348,9 @@ static int test__open_syscall_event_on_all_cpus(void)
 	}
 
 	cpus = cpu_map__new(NULL);
-	if (threads == NULL) {
-		pr_debug("thread_map__new\n");
-		return -1;
+	if (cpus == NULL) {
+		pr_debug("cpu_map__new\n");
+		goto out_thread_map_delete;
 	}
 
 
@@ -364,7 +365,7 @@ static int test__open_syscall_event_on_all_cpus(void)
 		goto out_thread_map_delete;
 	}
 
-	if (perf_evsel__open(evsel, cpus, threads) < 0) {
+	if (perf_evsel__open(evsel, cpus, threads, false, false) < 0) {
 		pr_debug("failed to open counter: %s, "
 			 "tweak /proc/sys/kernel/perf_event_paranoid?\n",
 			 strerror(errno));
@@ -408,6 +409,8 @@ static int test__open_syscall_event_on_all_cpus(void)
 		goto out_close_fd;
 	}
 
+	err = 0;
+
 	for (cpu = 0; cpu < cpus->nr; ++cpu) {
 		unsigned int expected;
 
@@ -416,18 +419,18 @@ static int test__open_syscall_event_on_all_cpus(void)
 
 		if (perf_evsel__read_on_cpu(evsel, cpu, 0) < 0) {
 			pr_debug("perf_evsel__open_read_on_cpu\n");
-			goto out_close_fd;
+			err = -1;
+			break;
 		}
 
 		expected = nr_open_calls + cpu;
 		if (evsel->counts->cpu[cpu].val != expected) {
 			pr_debug("perf_evsel__read_on_cpu: expected to intercept %d calls on cpu %d, got %" PRIu64 "\n",
 				 expected, cpus->map[cpu], evsel->counts->cpu[cpu].val);
-			goto out_close_fd;
+			err = -1;
 		}
 	}
 
-	err = 0;
 out_close_fd:
 	perf_evsel__close_fd(evsel, 1, threads->nr);
 out_evsel_delete:
@@ -437,6 +440,159 @@ out_thread_map_delete:
 	return err;
 }
 
+/*
+ * This test will generate random numbers of calls to some getpid syscalls,
+ * then establish an mmap for a group of events that are created to monitor
+ * the syscalls.
+ *
+ * It will receive the events, using mmap, use its PERF_SAMPLE_ID generated
+ * sample.id field to map back to its respective perf_evsel instance.
+ *
+ * Then it checks if the number of syscalls reported as perf events by
+ * the kernel corresponds to the number of syscalls made.
+ */
+static int test__basic_mmap(void)
+{
+	int err = -1;
+	event_t *event;
+	struct thread_map *threads;
+	struct cpu_map *cpus;
+	struct perf_evlist *evlist;
+	struct perf_event_attr attr = {
+		.type		= PERF_TYPE_TRACEPOINT,
+		.read_format	= PERF_FORMAT_ID,
+		.sample_type	= PERF_SAMPLE_ID,
+		.watermark	= 0,
+	};
+	cpu_set_t cpu_set;
+	const char *syscall_names[] = { "getsid", "getppid", "getpgrp",
+					"getpgid", };
+	pid_t (*syscalls[])(void) = { (void *)getsid, getppid, getpgrp,
+				      (void*)getpgid };
+#define nsyscalls ARRAY_SIZE(syscall_names)
+	int ids[nsyscalls];
+	unsigned int nr_events[nsyscalls],
+		     expected_nr_events[nsyscalls], i, j;
+	struct perf_evsel *evsels[nsyscalls], *evsel;
+
+	for (i = 0; i < nsyscalls; ++i) {
+		char name[64];
+
+		snprintf(name, sizeof(name), "sys_enter_%s", syscall_names[i]);
+		ids[i] = trace_event__id(name);
+		if (ids[i] < 0) {
+			pr_debug("Is debugfs mounted on /sys/kernel/debug?\n");
+			return -1;
+		}
+		nr_events[i] = 0;
+		expected_nr_events[i] = random() % 257;
+	}
+
+	threads = thread_map__new(-1, getpid());
+	if (threads == NULL) {
+		pr_debug("thread_map__new\n");
+		return -1;
+	}
+
+	cpus = cpu_map__new(NULL);
+	if (threads == NULL) {
+		pr_debug("thread_map__new\n");
+		goto out_free_threads;
+	}
+
+	CPU_ZERO(&cpu_set);
+	CPU_SET(cpus->map[0], &cpu_set);
+	sched_setaffinity(0, sizeof(cpu_set), &cpu_set);
+	if (sched_setaffinity(0, sizeof(cpu_set), &cpu_set) < 0) {
+		pr_debug("sched_setaffinity() failed on CPU %d: %s ",
+			 cpus->map[0], strerror(errno));
+		goto out_free_cpus;
+	}
+
+	evlist = perf_evlist__new();
+	if (threads == NULL) {
+		pr_debug("perf_evlist__new\n");
+		goto out_free_cpus;
+	}
+
+	/* anonymous union fields, can't be initialized above */
+	attr.wakeup_events = 1;
+	attr.sample_period = 1;
+
+	for (i = 0; i < nsyscalls; ++i) {
+		attr.config = ids[i];
+		evsels[i] = perf_evsel__new(&attr, i);
+		if (evsels[i] == NULL) {
+			pr_debug("perf_evsel__new\n");
+			goto out_free_evlist;
+		}
+
+		perf_evlist__add(evlist, evsels[i]);
+
+		if (perf_evsel__open(evsels[i], cpus, threads, false, false) < 0) {
+			pr_debug("failed to open counter: %s, "
+				 "tweak /proc/sys/kernel/perf_event_paranoid?\n",
+				 strerror(errno));
+			goto out_close_fd;
+		}
+	}
+
+	if (perf_evlist__mmap(evlist, cpus, threads, 128, true) < 0) {
+		pr_debug("failed to mmap events: %d (%s)\n", errno,
+			 strerror(errno));
+		goto out_close_fd;
+	}
+
+	for (i = 0; i < nsyscalls; ++i)
+		for (j = 0; j < expected_nr_events[i]; ++j) {
+			int foo = syscalls[i]();
+			++foo;
+		}
+
+	while ((event = perf_evlist__read_on_cpu(evlist, 0)) != NULL) {
+		struct sample_data sample;
+
+		if (event->header.type != PERF_RECORD_SAMPLE) {
+			pr_debug("unexpected %s event\n",
+				 event__get_event_name(event->header.type));
+			goto out_munmap;
+		}
+
+		event__parse_sample(event, attr.sample_type, false, &sample);
+		evsel = perf_evlist__id2evsel(evlist, sample.id);
+		if (evsel == NULL) {
+			pr_debug("event with id %" PRIu64
+				 " doesn't map to an evsel\n", sample.id);
+			goto out_munmap;
+		}
+		nr_events[evsel->idx]++;
+	}
+
+	list_for_each_entry(evsel, &evlist->entries, node) {
+		if (nr_events[evsel->idx] != expected_nr_events[evsel->idx]) {
+			pr_debug("expected %d %s events, got %d\n",
+				 expected_nr_events[evsel->idx],
+				 event_name(evsel), nr_events[evsel->idx]);
+			goto out_munmap;
+		}
+	}
+
+	err = 0;
+out_munmap:
+	perf_evlist__munmap(evlist, 1);
+out_close_fd:
+	for (i = 0; i < nsyscalls; ++i)
+		perf_evsel__close_fd(evsels[i], 1, threads->nr);
+out_free_evlist:
+	perf_evlist__delete(evlist);
+out_free_cpus:
+	cpu_map__delete(cpus);
+out_free_threads:
+	thread_map__delete(threads);
+	return err;
+#undef nsyscalls
+}
+
 static struct test {
 	const char *desc;
 	int (*func)(void);
@@ -453,6 +609,10 @@ static struct test {
 		.desc = "detect open syscall event on all cpus",
 		.func = test__open_syscall_event_on_all_cpus,
 	},
+	{
+		.desc = "read samples using the mmap interface",
+		.func = test__basic_mmap,
+	},
 	{
 		.func = NULL,
 	},

tools/perf/builtin-top.c

@@ -21,10 +21,12 @@
 #include "perf.h"
 
 #include "util/color.h"
+#include "util/evlist.h"
 #include "util/evsel.h"
 #include "util/session.h"
 #include "util/symbol.h"
 #include "util/thread.h"
+#include "util/thread_map.h"
 #include "util/util.h"
 #include <linux/rbtree.h>
 #include "util/parse-options.h"
@@ -60,6 +62,8 @@
 
 #define FD(e, x, y) (*(int *)xyarray__entry(e->fd, x, y))
 
+struct perf_evlist		*evsel_list;
+
 static bool			system_wide			=  false;
 
 static int			default_interval		=      0;
@@ -75,7 +79,7 @@ static struct cpu_map		*cpus;
 static int			realtime_prio			=      0;
 static bool			group				=  false;
 static unsigned int		page_size;
-static unsigned int		mmap_pages			=     16;
+static unsigned int		mmap_pages			=    128;
 static int			freq				=   1000; /* 1 KHz */
 
 static int			delay_secs			=      2;
@@ -267,7 +271,7 @@ static void __zero_source_counters(struct sym_entry *syme)
 
 	line = syme->src->lines;
 	while (line) {
-		for (i = 0; i < nr_counters; i++)
+		for (i = 0; i < evsel_list->nr_entries; i++)
 			line->count[i] = 0;
 		line = line->next;
 	}
@@ -414,7 +418,7 @@ static double sym_weight(const struct sym_entry *sym)
 	if (!display_weighted)
 		return weight;
 
-	for (counter = 1; counter < nr_counters-1; counter++)
+	for (counter = 1; counter < evsel_list->nr_entries - 1; counter++)
 		weight *= sym->count[counter];
 
 	weight /= (sym->count[counter] + 1);
@@ -501,7 +505,7 @@ static void print_sym_table(void)
 			rb_insert_active_sym(&tmp, syme);
 			sum_ksamples += syme->snap_count;
 
-			for (j = 0; j < nr_counters; j++)
+			for (j = 0; j < evsel_list->nr_entries; j++)
 				syme->count[j] = zero ? 0 : syme->count[j] * 7 / 8;
 		} else
 			list_remove_active_sym(syme);
@@ -535,9 +539,9 @@ static void print_sym_table(void)
 			esamples_percent);
 	}
 
-	if (nr_counters == 1 || !display_weighted) {
+	if (evsel_list->nr_entries == 1 || !display_weighted) {
 		struct perf_evsel *first;
-		first = list_entry(evsel_list.next, struct perf_evsel, node);
+		first = list_entry(evsel_list->entries.next, struct perf_evsel, node);
 		printf("%" PRIu64, (uint64_t)first->attr.sample_period);
 		if (freq)
 			printf("Hz ");
@@ -547,7 +551,7 @@ static void print_sym_table(void)
 
 	if (!display_weighted)
 		printf("%s", event_name(sym_evsel));
-	else list_for_each_entry(counter, &evsel_list, node) {
+	else list_for_each_entry(counter, &evsel_list->entries, node) {
 		if (counter->idx)
 			printf("/");
 
@@ -606,7 +610,7 @@ static void print_sym_table(void)
 			sym_width = winsize.ws_col - dso_width - 29;
 	}
 	putchar('\n');
-	if (nr_counters == 1)
+	if (evsel_list->nr_entries == 1)
 		printf("             samples  pcnt");
 	else
 		printf("   weight    samples  pcnt");
@@ -615,7 +619,7 @@ static void print_sym_table(void)
 		printf("         RIP       ");
 	printf(" %-*.*s DSO\n", sym_width, sym_width, "function");
 	printf("   %s    _______ _____",
-	       nr_counters == 1 ? "      " : "______");
+	       evsel_list->nr_entries == 1 ? "      " : "______");
 	if (verbose)
 		printf(" ________________");
 	printf(" %-*.*s", sym_width, sym_width, graph_line);
@@ -634,7 +638,7 @@ static void print_sym_table(void)
 		pcnt = 100.0 - (100.0 * ((sum_ksamples - syme->snap_count) /
 					 sum_ksamples));
 
-		if (nr_counters == 1 || !display_weighted)
+		if (evsel_list->nr_entries == 1 || !display_weighted)
 			printf("%20.2f ", syme->weight);
 		else
 			printf("%9.1f %10ld ", syme->weight, syme->snap_count);
@@ -744,7 +748,7 @@ static void print_mapped_keys(void)
 	fprintf(stdout, "\t[d]     display refresh delay.             \t(%d)\n", delay_secs);
 	fprintf(stdout, "\t[e]     display entries (lines).           \t(%d)\n", print_entries);
 
-	if (nr_counters > 1)
+	if (evsel_list->nr_entries > 1)
 		fprintf(stdout, "\t[E]     active event counter.              \t(%s)\n", event_name(sym_evsel));
 
 	fprintf(stdout, "\t[f]     profile display filter (count).    \t(%d)\n", count_filter);
@@ -753,7 +757,7 @@ static void print_mapped_keys(void)
 	fprintf(stdout, "\t[s]     annotate symbol.                   \t(%s)\n", name?: "NULL");
 	fprintf(stdout, "\t[S]     stop annotation.\n");
 
-	if (nr_counters > 1)
+	if (evsel_list->nr_entries > 1)
 		fprintf(stdout, "\t[w]     toggle display weighted/count[E]r. \t(%d)\n", display_weighted ? 1 : 0);
 
 	fprintf(stdout,
@@ -783,7 +787,7 @@ static int key_mapped(int c)
 			return 1;
 		case 'E':
 		case 'w':
-			return nr_counters > 1 ? 1 : 0;
+			return evsel_list->nr_entries > 1 ? 1 : 0;
 		default:
 			break;
 	}
@@ -831,22 +835,22 @@ static void handle_keypress(struct perf_session *session, int c)
 				signal(SIGWINCH, SIG_DFL);
 			break;
 		case 'E':
-			if (nr_counters > 1) {
+			if (evsel_list->nr_entries > 1) {
 				fprintf(stderr, "\nAvailable events:");
 
-				list_for_each_entry(sym_evsel, &evsel_list, node)
+				list_for_each_entry(sym_evsel, &evsel_list->entries, node)
 					fprintf(stderr, "\n\t%d %s", sym_evsel->idx, event_name(sym_evsel));
 
 				prompt_integer(&sym_counter, "Enter details event counter");
 
-				if (sym_counter >= nr_counters) {
-					sym_evsel = list_entry(evsel_list.next, struct perf_evsel, node);
+				if (sym_counter >= evsel_list->nr_entries) {
+					sym_evsel = list_entry(evsel_list->entries.next, struct perf_evsel, node);
 					sym_counter = 0;
 					fprintf(stderr, "Sorry, no such event, using %s.\n", event_name(sym_evsel));
 					sleep(1);
 					break;
 				}
-				list_for_each_entry(sym_evsel, &evsel_list, node)
+				list_for_each_entry(sym_evsel, &evsel_list->entries, node)
 					if (sym_evsel->idx == sym_counter)
 						break;
 			} else sym_counter = 0;
@@ -930,6 +934,7 @@ repeat:
 /* Tag samples to be skipped. */
 static const char *skip_symbols[] = {
 	"default_idle",
+	"native_safe_halt",
 	"cpu_idle",
 	"enter_idle",
 	"exit_idle",
@@ -988,8 +993,7 @@ static int symbol_filter(struct map *map, struct symbol *sym)
 
 static void event__process_sample(const event_t *self,
 				  struct sample_data *sample,
-				  struct perf_session *session,
-				  struct perf_evsel *evsel)
+				  struct perf_session *session)
 {
 	u64 ip = self->ip.ip;
 	struct sym_entry *syme;
@@ -1082,8 +1086,12 @@ static void event__process_sample(const event_t *self,
 
 	syme = symbol__priv(al.sym);
 	if (!syme->skip) {
-		syme->count[evsel->idx]++;
+		struct perf_evsel *evsel;
+
 		syme->origin = origin;
+		evsel = perf_evlist__id2evsel(evsel_list, sample->id);
+		assert(evsel != NULL);
+		syme->count[evsel->idx]++;
 		record_precise_ip(syme, evsel->idx, ip);
 		pthread_mutex_lock(&active_symbols_lock);
 		if (list_empty(&syme->node) || !syme->node.next)
@@ -1092,156 +1100,52 @@ static void event__process_sample(const event_t *self,
 	}
 }
 
-struct mmap_data {
-	void			*base;
-	int			mask;
-	unsigned int		prev;
-};
-
-static int perf_evsel__alloc_mmap_per_thread(struct perf_evsel *evsel,
-					     int ncpus, int nthreads)
-{
-	evsel->priv = xyarray__new(ncpus, nthreads, sizeof(struct mmap_data));
-	return evsel->priv != NULL ? 0 : -ENOMEM;
-}
-
-static void perf_evsel__free_mmap(struct perf_evsel *evsel)
-{
-	xyarray__delete(evsel->priv);
-	evsel->priv = NULL;
-}
-
-static unsigned int mmap_read_head(struct mmap_data *md)
-{
-	struct perf_event_mmap_page *pc = md->base;
-	int head;
-
-	head = pc->data_head;
-	rmb();
-
-	return head;
-}
-
-static void perf_session__mmap_read_counter(struct perf_session *self,
-					    struct perf_evsel *evsel,
-					    int cpu, int thread_idx)
+static void perf_session__mmap_read_cpu(struct perf_session *self, int cpu)
 {
-	struct xyarray *mmap_array = evsel->priv;
-	struct mmap_data *md = xyarray__entry(mmap_array, cpu, thread_idx);
-	unsigned int head = mmap_read_head(md);
-	unsigned int old = md->prev;
-	unsigned char *data = md->base + page_size;
 	struct sample_data sample;
-	int diff;
-
-	/*
-	 * If we're further behind than half the buffer, there's a chance
-	 * the writer will bite our tail and mess up the samples under us.
-	 *
-	 * If we somehow ended up ahead of the head, we got messed up.
-	 *
-	 * In either case, truncate and restart at head.
-	 */
-	diff = head - old;
-	if (diff > md->mask / 2 || diff < 0) {
-		fprintf(stderr, "WARNING: failed to keep up with mmap data.\n");
-
-		/*
-		 * head points to a known good entry, start there.
-		 */
-		old = head;
-	}
-
-	for (; old != head;) {
-		event_t *event = (event_t *)&data[old & md->mask];
+	event_t *event;
 
-		event_t event_copy;
-
-		size_t size = event->header.size;
-
-		/*
-		 * Event straddles the mmap boundary -- header should always
-		 * be inside due to u64 alignment of output.
-		 */
-		if ((old & md->mask) + size != ((old + size) & md->mask)) {
-			unsigned int offset = old;
-			unsigned int len = min(sizeof(*event), size), cpy;
-			void *dst = &event_copy;
-
-			do {
-				cpy = min(md->mask + 1 - (offset & md->mask), len);
-				memcpy(dst, &data[offset & md->mask], cpy);
-				offset += cpy;
-				dst += cpy;
-				len -= cpy;
-			} while (len);
-
-			event = &event_copy;
-		}
+	while ((event = perf_evlist__read_on_cpu(evsel_list, cpu)) != NULL) {
+		perf_session__parse_sample(self, event, &sample);
 
-		event__parse_sample(event, self, &sample);
 		if (event->header.type == PERF_RECORD_SAMPLE)
-			event__process_sample(event, &sample, self, evsel);
+			event__process_sample(event, &sample, self);
 		else
 			event__process(event, &sample, self);
-		old += size;
 	}
-
-	md->prev = old;
 }
 
-static struct pollfd *event_array;
-
 static void perf_session__mmap_read(struct perf_session *self)
 {
-	struct perf_evsel *counter;
-	int i, thread_index;
-
-	for (i = 0; i < cpus->nr; i++) {
-		list_for_each_entry(counter, &evsel_list, node) {
-			for (thread_index = 0;
-				thread_index < threads->nr;
-				thread_index++) {
-				perf_session__mmap_read_counter(self,
-					counter, i, thread_index);
-			}
-		}
-	}
-}
+	int i;
 
-int nr_poll;
-int group_fd;
+	for (i = 0; i < cpus->nr; i++)
+		perf_session__mmap_read_cpu(self, i);
+}
 
-static void start_counter(int i, struct perf_evsel *evsel)
+static void start_counters(struct perf_evlist *evlist)
 {
-	struct xyarray *mmap_array = evsel->priv;
-	struct mmap_data *mm;
-	struct perf_event_attr *attr;
-	int cpu = -1;
-	int thread_index;
-
-	if (target_tid == -1)
-		cpu = cpus->map[i];
+	struct perf_evsel *counter;
 
-	attr = &evsel->attr;
+	list_for_each_entry(counter, &evlist->entries, node) {
+		struct perf_event_attr *attr = &counter->attr;
 
-	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;
-		attr->freq		= 1;
-		attr->sample_freq	= freq;
-	}
+		if (freq) {
+			attr->sample_type |= PERF_SAMPLE_PERIOD;
+			attr->freq	  = 1;
+			attr->sample_freq = freq;
+		}
 
-	attr->inherit		= (cpu < 0) && inherit;
-	attr->mmap		= 1;
+		if (evlist->nr_entries > 1) {
+			attr->sample_type |= PERF_SAMPLE_ID;
+			attr->read_format |= PERF_FORMAT_ID;
+		}
 
-	for (thread_index = 0; thread_index < threads->nr; thread_index++) {
+		attr->mmap = 1;
 try_again:
-		FD(evsel, i, thread_index) = sys_perf_event_open(attr,
-				threads->map[thread_index], cpu, group_fd, 0);
-
-		if (FD(evsel, i, thread_index) < 0) {
+		if (perf_evsel__open(counter, cpus, threads, group, inherit) < 0) {
 			int err = errno;
 
 			if (err == EPERM || err == EACCES)
@@ -1253,8 +1157,8 @@ try_again:
 			 * based cpu-clock-tick sw counter, which
 			 * is always available even if no PMU support:
 			 */
-			if (attr->type == PERF_TYPE_HARDWARE
-					&& attr->config == PERF_COUNT_HW_CPU_CYCLES) {
+			if (attr->type == PERF_TYPE_HARDWARE &&
+			    attr->config == PERF_COUNT_HW_CPU_CYCLES) {
 
 				if (verbose)
 					warning(" ... trying to fall back to cpu-clock-ticks\n");
@@ -1264,39 +1168,23 @@ try_again:
 				goto try_again;
 			}
 			printf("\n");
-			error("sys_perf_event_open() syscall returned with %d (%s).  /bin/dmesg may provide additional information.\n",
-					FD(evsel, i, thread_index), strerror(err));
+			error("sys_perf_event_open() syscall returned with %d "
+			      "(%s).  /bin/dmesg may provide additional information.\n",
+			      err, strerror(err));
 			die("No CONFIG_PERF_EVENTS=y kernel support configured?\n");
 			exit(-1);
 		}
-		assert(FD(evsel, i, thread_index) >= 0);
-		fcntl(FD(evsel, i, thread_index), F_SETFL, O_NONBLOCK);
-
-		/*
-		 * First counter acts as the group leader:
-		 */
-		if (group && group_fd == -1)
-			group_fd = FD(evsel, i, thread_index);
-
-		event_array[nr_poll].fd = FD(evsel, i, thread_index);
-		event_array[nr_poll].events = POLLIN;
-		nr_poll++;
-
-		mm = xyarray__entry(mmap_array, i, thread_index);
-		mm->prev = 0;
-		mm->mask = mmap_pages*page_size - 1;
-		mm->base = mmap(NULL, (mmap_pages+1)*page_size,
-				PROT_READ, MAP_SHARED, FD(evsel, i, thread_index), 0);
-		if (mm->base == MAP_FAILED)
-			die("failed to mmap with %d (%s)\n", errno, strerror(errno));
 	}
+
+	if (perf_evlist__mmap(evlist, cpus, threads, mmap_pages, true) < 0)
+		die("failed to mmap with %d (%s)\n", errno, strerror(errno));
 }
 
 static int __cmd_top(void)
 {
 	pthread_t thread;
-	struct perf_evsel *counter;
-	int i, ret;
+	struct perf_evsel *first;
+	int ret;
 	/*
 	 * FIXME: perf_session__new should allow passing a O_MMAP, so that all this
 	 * mmap reading, etc is encapsulated in it. Use O_WRONLY for now.
@@ -1310,14 +1198,12 @@ static int __cmd_top(void)
 	else
 		event__synthesize_threads(event__process, session);
 
-	for (i = 0; i < cpus->nr; i++) {
-		group_fd = -1;
-		list_for_each_entry(counter, &evsel_list, node)
-			start_counter(i, counter);
-	}
+	start_counters(evsel_list);
+	first = list_entry(evsel_list->entries.next, struct perf_evsel, node);
+	perf_session__set_sample_type(session, first->attr.sample_type);
 
 	/* Wait for a minimal set of events before starting the snapshot */
-	poll(&event_array[0], nr_poll, 100);
+	poll(evsel_list->pollfd, evsel_list->nr_fds, 100);
 
 	perf_session__mmap_read(session);
 
@@ -1342,7 +1228,7 @@ static int __cmd_top(void)
 		perf_session__mmap_read(session);
 
 		if (hits == samples)
-			ret = poll(event_array, nr_poll, 100);
+			ret = poll(evsel_list->pollfd, evsel_list->nr_fds, 100);
 	}
 
 	return 0;
@@ -1354,7 +1240,7 @@ static const char * const top_usage[] = {
 };
 
 static const struct option options[] = {
-	OPT_CALLBACK('e', "event", NULL, "event",
+	OPT_CALLBACK('e', "event", &evsel_list, "event",
 		     "event selector. use 'perf list' to list available events",
 		     parse_events),
 	OPT_INTEGER('c', "count", &default_interval,
@@ -1404,6 +1290,10 @@ int cmd_top(int argc, const char **argv, const char *prefix __used)
 	struct perf_evsel *pos;
 	int status = -ENOMEM;
 
+	evsel_list = perf_evlist__new();
+	if (evsel_list == NULL)
+		return -ENOMEM;
+
 	page_size = sysconf(_SC_PAGE_SIZE);
 
 	argc = parse_options(argc, argv, options, top_usage, 0);
@@ -1419,11 +1309,6 @@ int cmd_top(int argc, const char **argv, const char *prefix __used)
 		usage_with_options(top_usage, options);
 	}
 
-	event_array = malloc((sizeof(struct pollfd) *
-			      MAX_NR_CPUS * MAX_COUNTERS * threads->nr));
-	if (!event_array)
-		return -ENOMEM;
-
 	/* CPU and PID are mutually exclusive */
 	if (target_tid > 0 && cpu_list) {
 		printf("WARNING: PID switch overriding CPU\n");
@@ -1431,7 +1316,8 @@ int cmd_top(int argc, const char **argv, const char *prefix __used)
 		cpu_list = NULL;
 	}
 
-	if (!nr_counters && perf_evsel_list__create_default() < 0) {
+	if (!evsel_list->nr_entries &&
+	    perf_evlist__add_default(evsel_list) < 0) {
 		pr_err("Not enough memory for event selector list\n");
 		return -ENOMEM;
 	}
@@ -1459,9 +1345,8 @@ int cmd_top(int argc, const char **argv, const char *prefix __used)
 	if (cpus == NULL)
 		usage_with_options(top_usage, options);
 
-	list_for_each_entry(pos, &evsel_list, node) {
-		if (perf_evsel__alloc_mmap_per_thread(pos, cpus->nr, threads->nr) < 0 ||
-		    perf_evsel__alloc_fd(pos, cpus->nr, threads->nr) < 0)
+	list_for_each_entry(pos, &evsel_list->entries, node) {
+		if (perf_evsel__alloc_fd(pos, cpus->nr, threads->nr) < 0)
 			goto out_free_fd;
 		/*
 		 * Fill in the ones not specifically initialized via -c:
@@ -1472,10 +1357,14 @@ int cmd_top(int argc, const char **argv, const char *prefix __used)
 		pos->attr.sample_period = default_interval;
 	}
 
-	sym_evsel = list_entry(evsel_list.next, struct perf_evsel, node);
+	if (perf_evlist__alloc_pollfd(evsel_list, cpus->nr, threads->nr) < 0 ||
+	    perf_evlist__alloc_mmap(evsel_list, cpus->nr) < 0)
+		goto out_free_fd;
+
+	sym_evsel = list_entry(evsel_list->entries.next, struct perf_evsel, node);
 
 	symbol_conf.priv_size = (sizeof(struct sym_entry) +
-				 (nr_counters + 1) * sizeof(unsigned long));
+				 (evsel_list->nr_entries + 1) * sizeof(unsigned long));
 
 	symbol_conf.try_vmlinux_path = (symbol_conf.vmlinux_name == NULL);
 	if (symbol__init() < 0)
@@ -1489,9 +1378,7 @@ int cmd_top(int argc, const char **argv, const char *prefix __used)
 
 	status = __cmd_top();
 out_free_fd:
-	list_for_each_entry(pos, &evsel_list, node)
-		perf_evsel__free_mmap(pos);
-	perf_evsel_list__delete();
+	perf_evlist__delete(evsel_list);
 
 	return status;
 }

tools/perf/perf.h

@@ -94,6 +94,32 @@ void get_term_dimensions(struct winsize *ws);
 #include "util/types.h"
 #include <stdbool.h>
 
+struct perf_mmap {
+	void			*base;
+	int			mask;
+	unsigned int		prev;
+};
+
+static inline unsigned int perf_mmap__read_head(struct perf_mmap *mm)
+{
+	struct perf_event_mmap_page *pc = mm->base;
+	int head = pc->data_head;
+	rmb();
+	return head;
+}
+
+static inline void perf_mmap__write_tail(struct perf_mmap *md,
+					 unsigned long tail)
+{
+	struct perf_event_mmap_page *pc = md->base;
+
+	/*
+	 * ensure all reads are done before we write the tail out.
+	 */
+	/* mb(); */
+	pc->data_tail = tail;
+}
+
 /*
  * prctl(PR_TASK_PERF_EVENTS_DISABLE) will (cheaply) disable all
  * counters in the current task.

tools/perf/util/callchain.c

@@ -1,5 +1,5 @@
 /*
- * Copyright (C) 2009-2010, Frederic Weisbecker <fweisbec@gmail.com>
+ * Copyright (C) 2009-2011, Frederic Weisbecker <fweisbec@gmail.com>
  *
  * Handle the callchains from the stream in an ad-hoc radix tree and then
  * sort them in an rbtree.
@@ -26,10 +26,10 @@ bool ip_callchain__valid(struct ip_callchain *chain, const event_t *event)
 }
 
 #define chain_for_each_child(child, parent)	\
-	list_for_each_entry(child, &parent->children, brothers)
+	list_for_each_entry(child, &parent->children, siblings)
 
 #define chain_for_each_child_safe(child, next, parent)	\
-	list_for_each_entry_safe(child, next, &parent->children, brothers)
+	list_for_each_entry_safe(child, next, &parent->children, siblings)
 
 static void
 rb_insert_callchain(struct rb_root *root, struct callchain_node *chain,
@@ -38,14 +38,14 @@ rb_insert_callchain(struct rb_root *root, struct callchain_node *chain,
 	struct rb_node **p = &root->rb_node;
 	struct rb_node *parent = NULL;
 	struct callchain_node *rnode;
-	u64 chain_cumul = cumul_hits(chain);
+	u64 chain_cumul = callchain_cumul_hits(chain);
 
 	while (*p) {
 		u64 rnode_cumul;
 
 		parent = *p;
 		rnode = rb_entry(parent, struct callchain_node, rb_node);
-		rnode_cumul = cumul_hits(rnode);
+		rnode_cumul = callchain_cumul_hits(rnode);
 
 		switch (mode) {
 		case CHAIN_FLAT:
@@ -104,7 +104,7 @@ static void __sort_chain_graph_abs(struct callchain_node *node,
 
 	chain_for_each_child(child, node) {
 		__sort_chain_graph_abs(child, min_hit);
-		if (cumul_hits(child) >= min_hit)
+		if (callchain_cumul_hits(child) >= min_hit)
 			rb_insert_callchain(&node->rb_root, child,
 					    CHAIN_GRAPH_ABS);
 	}
@@ -129,7 +129,7 @@ static void __sort_chain_graph_rel(struct callchain_node *node,
 
 	chain_for_each_child(child, node) {
 		__sort_chain_graph_rel(child, min_percent);
-		if (cumul_hits(child) >= min_hit)
+		if (callchain_cumul_hits(child) >= min_hit)
 			rb_insert_callchain(&node->rb_root, child,
 					    CHAIN_GRAPH_REL);
 	}
@@ -143,7 +143,7 @@ sort_chain_graph_rel(struct rb_root *rb_root, struct callchain_root *chain_root,
 	rb_root->rb_node = chain_root->node.rb_root.rb_node;
 }
 
-int register_callchain_param(struct callchain_param *param)
+int callchain_register_param(struct callchain_param *param)
 {
 	switch (param->mode) {
 	case CHAIN_GRAPH_ABS:
@@ -189,32 +189,27 @@ create_child(struct callchain_node *parent, bool inherit_children)
 		chain_for_each_child(next, new)
 			next->parent = new;
 	}
-	list_add_tail(&new->brothers, &parent->children);
+	list_add_tail(&new->siblings, &parent->children);
 
 	return new;
 }
 
 
-struct resolved_ip {
-	u64		  ip;
-	struct map_symbol ms;
-};
-
-struct resolved_chain {
-	u64			nr;
-	struct resolved_ip	ips[0];
-};
-
-
 /*
  * Fill the node with callchain values
  */
 static void
-fill_node(struct callchain_node *node, struct resolved_chain *chain, int start)
+fill_node(struct callchain_node *node, struct callchain_cursor *cursor)
 {
-	unsigned int i;
+	struct callchain_cursor_node *cursor_node;
+
+	node->val_nr = cursor->nr - cursor->pos;
+	if (!node->val_nr)
+		pr_warning("Warning: empty node in callchain tree\n");
 
-	for (i = start; i < chain->nr; i++) {
+	cursor_node = callchain_cursor_current(cursor);
+
+	while (cursor_node) {
 		struct callchain_list *call;
 
 		call = zalloc(sizeof(*call));
@@ -222,23 +217,25 @@ fill_node(struct callchain_node *node, struct resolved_chain *chain, int start)
 			perror("not enough memory for the code path tree");
 			return;
 		}
-		call->ip = chain->ips[i].ip;
-		call->ms = chain->ips[i].ms;
+		call->ip = cursor_node->ip;
+		call->ms.sym = cursor_node->sym;
+		call->ms.map = cursor_node->map;
 		list_add_tail(&call->list, &node->val);
+
+		callchain_cursor_advance(cursor);
+		cursor_node = callchain_cursor_current(cursor);
 	}
-	node->val_nr = chain->nr - start;
-	if (!node->val_nr)
-		pr_warning("Warning: empty node in callchain tree\n");
 }
 
 static void
-add_child(struct callchain_node *parent, struct resolved_chain *chain,
-	  int start, u64 period)
+add_child(struct callchain_node *parent,
+	  struct callchain_cursor *cursor,
+	  u64 period)
 {
 	struct callchain_node *new;
 
 	new = create_child(parent, false);
-	fill_node(new, chain, start);
+	fill_node(new, cursor);
 
 	new->children_hit = 0;
 	new->hit = period;
@@ -250,9 +247,10 @@ add_child(struct callchain_node *parent, struct resolved_chain *chain,
  * Then create another child to host the given callchain of new branch
  */
 static void
-split_add_child(struct callchain_node *parent, struct resolved_chain *chain,
-		struct callchain_list *to_split, int idx_parents, int idx_local,
-		u64 period)
+split_add_child(struct callchain_node *parent,
+		struct callchain_cursor *cursor,
+		struct callchain_list *to_split,
+		u64 idx_parents, u64 idx_local, u64 period)
 {
 	struct callchain_node *new;
 	struct list_head *old_tail;
@@ -272,14 +270,14 @@ split_add_child(struct callchain_node *parent, struct resolved_chain *chain,
 	/* split the hits */
 	new->hit = parent->hit;
 	new->children_hit = parent->children_hit;
-	parent->children_hit = cumul_hits(new);
+	parent->children_hit = callchain_cumul_hits(new);
 	new->val_nr = parent->val_nr - idx_local;
 	parent->val_nr = idx_local;
 
 	/* create a new child for the new branch if any */
-	if (idx_total < chain->nr) {
+	if (idx_total < cursor->nr) {
 		parent->hit = 0;
-		add_child(parent, chain, idx_total, period);
+		add_child(parent, cursor, period);
 		parent->children_hit += period;
 	} else {
 		parent->hit = period;
@@ -287,36 +285,41 @@ split_add_child(struct callchain_node *parent, struct resolved_chain *chain,
 }
 
 static int
-append_chain(struct callchain_node *root, struct resolved_chain *chain,
-	     unsigned int start, u64 period);
+append_chain(struct callchain_node *root,
+	     struct callchain_cursor *cursor,
+	     u64 period);
 
 static void
-append_chain_children(struct callchain_node *root, struct resolved_chain *chain,
-		      unsigned int start, u64 period)
+append_chain_children(struct callchain_node *root,
+		      struct callchain_cursor *cursor,
+		      u64 period)
 {
 	struct callchain_node *rnode;
 
 	/* lookup in childrens */
 	chain_for_each_child(rnode, root) {
-		unsigned int ret = append_chain(rnode, chain, start, period);
+		unsigned int ret = append_chain(rnode, cursor, period);
 
 		if (!ret)
 			goto inc_children_hit;
 	}
 	/* nothing in children, add to the current node */
-	add_child(root, chain, start, period);
+	add_child(root, cursor, period);
 
 inc_children_hit:
 	root->children_hit += period;
 }
 
 static int
-append_chain(struct callchain_node *root, struct resolved_chain *chain,
-	     unsigned int start, u64 period)
+append_chain(struct callchain_node *root,
+	     struct callchain_cursor *cursor,
+	     u64 period)
 {
+	struct callchain_cursor_node *curr_snap = cursor->curr;
 	struct callchain_list *cnode;
-	unsigned int i = start;
+	u64 start = cursor->pos;
 	bool found = false;
+	u64 matches;
 
 	/*
 	 * Lookup in the current node
@@ -324,141 +327,134 @@ append_chain(struct callchain_node *root, struct resolved_chain *chain,
 	 * anywhere inside a function.
 	 */
 	list_for_each_entry(cnode, &root->val, list) {
+		struct callchain_cursor_node *node;
 		struct symbol *sym;
 
-		if (i == chain->nr)
+		node = callchain_cursor_current(cursor);
+		if (!node)
 			break;
 
-		sym = chain->ips[i].ms.sym;
+		sym = node->sym;
 
 		if (cnode->ms.sym && sym) {
 			if (cnode->ms.sym->start != sym->start)
 				break;
-		} else if (cnode->ip != chain->ips[i].ip)
+		} else if (cnode->ip != node->ip)
 			break;
 
 		if (!found)
 			found = true;
-		i++;
+
+		callchain_cursor_advance(cursor);
 	}
 
 	/* matches not, relay on the parent */
-	if (!found)
+	if (!found) {
+		cursor->curr = curr_snap;
+		cursor->pos = start;
 		return -1;
+	}
+
+	matches = cursor->pos - start;
 
 	/* we match only a part of the node. Split it and add the new chain */
-	if (i - start < root->val_nr) {
-		split_add_child(root, chain, cnode, start, i - start, period);
+	if (matches < root->val_nr) {
+		split_add_child(root, cursor, cnode, start, matches, period);
 		return 0;
 	}
 
 	/* we match 100% of the path, increment the hit */
-	if (i - start == root->val_nr && i == chain->nr) {
+	if (matches == root->val_nr && cursor->pos == cursor->nr) {
 		root->hit += period;
 		return 0;
 	}
 
 	/* We match the node and still have a part remaining */
-	append_chain_children(root, chain, i, period);
+	append_chain_children(root, cursor, period);
 
 	return 0;
 }
 
-static void filter_context(struct ip_callchain *old, struct resolved_chain *new,
-			   struct map_symbol *syms)
-{
-	int i, j = 0;
-
-	for (i = 0; i < (int)old->nr; i++) {
-		if (old->ips[i] >= PERF_CONTEXT_MAX)
-			continue;
-
-		new->ips[j].ip = old->ips[i];
-		new->ips[j].ms = syms[i];
-		j++;
-	}
-
-	new->nr = j;
-}
-
-
-int callchain_append(struct callchain_root *root, struct ip_callchain *chain,
-		     struct map_symbol *syms, u64 period)
+int callchain_append(struct callchain_root *root,
+		     struct callchain_cursor *cursor,
+		     u64 period)
 {
-	struct resolved_chain *filtered;
-
-	if (!chain->nr)
+	if (!cursor->nr)
 		return 0;
 
-	filtered = zalloc(sizeof(*filtered) +
-			  chain->nr * sizeof(struct resolved_ip));
-	if (!filtered)
-		return -ENOMEM;
-
-	filter_context(chain, filtered, syms);
-
-	if (!filtered->nr)
-		goto end;
+	callchain_cursor_commit(cursor);
 
-	append_chain_children(&root->node, filtered, 0, period);
+	append_chain_children(&root->node, cursor, period);
 
-	if (filtered->nr > root->max_depth)
-		root->max_depth = filtered->nr;
-end:
-	free(filtered);
+	if (cursor->nr > root->max_depth)
+		root->max_depth = cursor->nr;
 
 	return 0;
 }
 
 static int
-merge_chain_branch(struct callchain_node *dst, struct callchain_node *src,
-		   struct resolved_chain *chain)
+merge_chain_branch(struct callchain_cursor *cursor,
+		   struct callchain_node *dst, struct callchain_node *src)
 {
+	struct callchain_cursor_node **old_last = cursor->last;
 	struct callchain_node *child, *next_child;
 	struct callchain_list *list, *next_list;
-	int old_pos = chain->nr;
+	int old_pos = cursor->nr;
 	int err = 0;
 
 	list_for_each_entry_safe(list, next_list, &src->val, list) {
-		chain->ips[chain->nr].ip = list->ip;
-		chain->ips[chain->nr].ms = list->ms;
-		chain->nr++;
+		callchain_cursor_append(cursor, list->ip,
+					list->ms.map, list->ms.sym);
 		list_del(&list->list);
 		free(list);
 	}
 
-	if (src->hit)
-		append_chain_children(dst, chain, 0, src->hit);
+	if (src->hit) {
+		callchain_cursor_commit(cursor);
+		append_chain_children(dst, cursor, src->hit);
+	}
 
 	chain_for_each_child_safe(child, next_child, src) {
-		err = merge_chain_branch(dst, child, chain);
+		err = merge_chain_branch(cursor, dst, child);
 		if (err)
 			break;
 
-		list_del(&child->brothers);
+		list_del(&child->siblings);
 		free(child);
 	}
 
-	chain->nr = old_pos;
+	cursor->nr = old_pos;
+	cursor->last = old_last;
 
 	return err;
 }
 
-int callchain_merge(struct callchain_root *dst, struct callchain_root *src)
+int callchain_merge(struct callchain_cursor *cursor,
+		    struct callchain_root *dst, struct callchain_root *src)
+{
+	return merge_chain_branch(cursor, &dst->node, &src->node);
+}
+
+int callchain_cursor_append(struct callchain_cursor *cursor,
+			    u64 ip, struct map *map, struct symbol *sym)
 {
-	struct resolved_chain *chain;
-	int err;
+	struct callchain_cursor_node *node = *cursor->last;
 
-	chain = malloc(sizeof(*chain) +
-		       src->max_depth * sizeof(struct resolved_ip));
-	if (!chain)
-		return -ENOMEM;
+	if (!node) {
+		node = calloc(sizeof(*node), 1);
+		if (!node)
+			return -ENOMEM;
 
-	chain->nr = 0;
+		*cursor->last = node;
+	}
 
-	err = merge_chain_branch(&dst->node, &src->node, chain);
+	node->ip = ip;
+	node->map = map;
+	node->sym = sym;
 
-	free(chain);
+	cursor->nr++;
 
-	return err;
+	cursor->last = &node->next;
+
+	return 0;
 }

tools/perf/util/callchain.h

@@ -16,7 +16,7 @@ enum chain_mode {
 
 struct callchain_node {
 	struct callchain_node	*parent;
-	struct list_head	brothers;
+	struct list_head	siblings;
 	struct list_head	children;
 	struct list_head	val;
 	struct rb_node		rb_node; /* to sort nodes in an rbtree */
@@ -49,9 +49,30 @@ struct callchain_list {
 	struct list_head	list;
 };
 
+/*
+ * A callchain cursor is a single linked list that
+ * let one feed a callchain progressively.
+ * It keeps persitent allocated entries to minimize
+ * allocations.
+ */
+struct callchain_cursor_node {
+	u64				ip;
+	struct map			*map;
+	struct symbol			*sym;
+	struct callchain_cursor_node	*next;
+};
+
+struct callchain_cursor {
+	u64				nr;
+	struct callchain_cursor_node	*first;
+	struct callchain_cursor_node	**last;
+	u64				pos;
+	struct callchain_cursor_node	*curr;
+};
+
 static inline void callchain_init(struct callchain_root *root)
 {
-	INIT_LIST_HEAD(&root->node.brothers);
+	INIT_LIST_HEAD(&root->node.siblings);
 	INIT_LIST_HEAD(&root->node.children);
 	INIT_LIST_HEAD(&root->node.val);
 
@@ -61,15 +82,54 @@ static inline void callchain_init(struct callchain_root *root)
 	root->max_depth = 0;
 }
 
-static inline u64 cumul_hits(struct callchain_node *node)
+static inline u64 callchain_cumul_hits(struct callchain_node *node)
 {
 	return node->hit + node->children_hit;
 }
 
-int register_callchain_param(struct callchain_param *param);
-int callchain_append(struct callchain_root *root, struct ip_callchain *chain,
-		     struct map_symbol *syms, u64 period);
-int callchain_merge(struct callchain_root *dst, struct callchain_root *src);
+int callchain_register_param(struct callchain_param *param);
+int callchain_append(struct callchain_root *root,
+		     struct callchain_cursor *cursor,
+		     u64 period);
+
+int callchain_merge(struct callchain_cursor *cursor,
+		    struct callchain_root *dst, struct callchain_root *src);
 
 bool ip_callchain__valid(struct ip_callchain *chain, const event_t *event);
+
+/*
+ * Initialize a cursor before adding entries inside, but keep
+ * the previously allocated entries as a cache.
+ */
+static inline void callchain_cursor_reset(struct callchain_cursor *cursor)
+{
+	cursor->nr = 0;
+	cursor->last = &cursor->first;
+}
+
+int callchain_cursor_append(struct callchain_cursor *cursor, u64 ip,
+			    struct map *map, struct symbol *sym);
+
+/* Close a cursor writing session. Initialize for the reader */
+static inline void callchain_cursor_commit(struct callchain_cursor *cursor)
+{
+	cursor->curr = cursor->first;
+	cursor->pos = 0;
+}
+
+/* Cursor reading iteration helpers */
+static inline struct callchain_cursor_node *
+callchain_cursor_current(struct callchain_cursor *cursor)
+{
+	if (cursor->pos == cursor->nr)
+		return NULL;
+
+	return cursor->curr;
+}
+
+static inline void callchain_cursor_advance(struct callchain_cursor *cursor)
+{
+	cursor->curr = cursor->curr->next;
+	cursor->pos++;
+}
 #endif	/* __PERF_CALLCHAIN_H */

tools/perf/util/cpumap.c

@@ -177,3 +177,8 @@ struct cpu_map *cpu_map__dummy_new(void)
 
 	return cpus;
 }
+
+void cpu_map__delete(struct cpu_map *map)
+{
+	free(map);
+}

tools/perf/util/cpumap.h

@@ -8,6 +8,6 @@ struct cpu_map {
 
 struct cpu_map *cpu_map__new(const char *cpu_list);
 struct cpu_map *cpu_map__dummy_new(void);
-void *cpu_map__delete(struct cpu_map *map);
+void cpu_map__delete(struct cpu_map *map);
 
 #endif /* __PERF_CPUMAP_H */

tools/perf/util/event.c

@@ -826,128 +826,3 @@ out_filtered:
 	al->filtered = true;
 	return 0;
 }
-
-static int event__parse_id_sample(const event_t *event,
-				  struct perf_session *session,
-				  struct sample_data *sample)
-{
-	const u64 *array;
-	u64 type;
-
-	sample->cpu = sample->pid = sample->tid = -1;
-	sample->stream_id = sample->id = sample->time = -1ULL;
-
-	if (!session->sample_id_all)
-		return 0;
-
-	array = event->sample.array;
-	array += ((event->header.size -
-		   sizeof(event->header)) / sizeof(u64)) - 1;
-	type = session->sample_type;
-
-	if (type & PERF_SAMPLE_CPU) {
-		u32 *p = (u32 *)array;
-		sample->cpu = *p;
-		array--;
-	}
-
-	if (type & PERF_SAMPLE_STREAM_ID) {
-		sample->stream_id = *array;
-		array--;
-	}
-
-	if (type & PERF_SAMPLE_ID) {
-		sample->id = *array;
-		array--;
-	}
-
-	if (type & PERF_SAMPLE_TIME) {
-		sample->time = *array;
-		array--;
-	}
-
-	if (type & PERF_SAMPLE_TID) {
-		u32 *p = (u32 *)array;
-		sample->pid = p[0];
-		sample->tid = p[1];
-	}
-
-	return 0;
-}
-
-int event__parse_sample(const event_t *event, struct perf_session *session,
-			struct sample_data *data)
-{
-	const u64 *array;
-	u64 type;
-
-	if (event->header.type != PERF_RECORD_SAMPLE)
-		return event__parse_id_sample(event, session, data);
-
-	array = event->sample.array;
-	type = session->sample_type;
-
-	if (type & PERF_SAMPLE_IP) {
-		data->ip = event->ip.ip;
-		array++;
-	}
-
-	if (type & PERF_SAMPLE_TID) {
-		u32 *p = (u32 *)array;
-		data->pid = p[0];
-		data->tid = p[1];
-		array++;
-	}
-
-	if (type & PERF_SAMPLE_TIME) {
-		data->time = *array;
-		array++;
-	}
-
-	if (type & PERF_SAMPLE_ADDR) {
-		data->addr = *array;
-		array++;
-	}
-
-	data->id = -1ULL;
-	if (type & PERF_SAMPLE_ID) {
-		data->id = *array;
-		array++;
-	}
-
-	if (type & PERF_SAMPLE_STREAM_ID) {
-		data->stream_id = *array;
-		array++;
-	}
-
-	if (type & PERF_SAMPLE_CPU) {
-		u32 *p = (u32 *)array;
-		data->cpu = *p;
-		array++;
-	} else
-		data->cpu = -1;
-
-	if (type & PERF_SAMPLE_PERIOD) {
-		data->period = *array;
-		array++;
-	}
-
-	if (type & PERF_SAMPLE_READ) {
-		pr_debug("PERF_SAMPLE_READ is unsuported for now\n");
-		return -1;
-	}
-
-	if (type & PERF_SAMPLE_CALLCHAIN) {
-		data->callchain = (struct ip_callchain *)array;
-		array += 1 + data->callchain->nr;
-	}
-
-	if (type & PERF_SAMPLE_RAW) {
-		u32 *p = (u32 *)array;
-		data->raw_size = *p;
-		p++;
-		data->raw_data = p;
-	}
-
-	return 0;
-}

tools/perf/util/event.h

@@ -169,9 +169,10 @@ struct addr_location;
 int event__preprocess_sample(const event_t *self, struct perf_session *session,
 			     struct addr_location *al, struct sample_data *data,
 			     symbol_filter_t filter);
-int event__parse_sample(const event_t *event, struct perf_session *session,
-			struct sample_data *sample);
 
 const char *event__get_event_name(unsigned int id);
 
+int event__parse_sample(const event_t *event, u64 type, bool sample_id_all,
+			struct sample_data *sample);
+
 #endif /* __PERF_RECORD_H */

tools/perf/util/evlist.c

@@ -0,0 +1,170 @@
+#include <poll.h>
+#include "evlist.h"
+#include "evsel.h"
+#include "util.h"
+
+#include <linux/bitops.h>
+#include <linux/hash.h>
+
+void perf_evlist__init(struct perf_evlist *evlist)
+{
+	int i;
+
+	for (i = 0; i < PERF_EVLIST__HLIST_SIZE; ++i)
+		INIT_HLIST_HEAD(&evlist->heads[i]);
+	INIT_LIST_HEAD(&evlist->entries);
+}
+
+struct perf_evlist *perf_evlist__new(void)
+{
+	struct perf_evlist *evlist = zalloc(sizeof(*evlist));
+
+	if (evlist != NULL)
+		perf_evlist__init(evlist);
+
+	return evlist;
+}
+
+static void perf_evlist__purge(struct perf_evlist *evlist)
+{
+	struct perf_evsel *pos, *n;
+
+	list_for_each_entry_safe(pos, n, &evlist->entries, node) {
+		list_del_init(&pos->node);
+		perf_evsel__delete(pos);
+	}
+
+	evlist->nr_entries = 0;
+}
+
+void perf_evlist__exit(struct perf_evlist *evlist)
+{
+	free(evlist->mmap);
+	free(evlist->pollfd);
+	evlist->mmap = NULL;
+	evlist->pollfd = NULL;
+}
+
+void perf_evlist__delete(struct perf_evlist *evlist)
+{
+	perf_evlist__purge(evlist);
+	perf_evlist__exit(evlist);
+	free(evlist);
+}
+
+void perf_evlist__add(struct perf_evlist *evlist, struct perf_evsel *entry)
+{
+	list_add_tail(&entry->node, &evlist->entries);
+	++evlist->nr_entries;
+}
+
+int perf_evlist__add_default(struct perf_evlist *evlist)
+{
+	struct perf_event_attr attr = {
+		.type = PERF_TYPE_HARDWARE,
+		.config = PERF_COUNT_HW_CPU_CYCLES,
+	};
+	struct perf_evsel *evsel = perf_evsel__new(&attr, 0);
+
+	if (evsel == NULL)
+		return -ENOMEM;
+
+	perf_evlist__add(evlist, evsel);
+	return 0;
+}
+
+int perf_evlist__alloc_pollfd(struct perf_evlist *evlist, int ncpus, int nthreads)
+{
+	int nfds = ncpus * nthreads * evlist->nr_entries;
+	evlist->pollfd = malloc(sizeof(struct pollfd) * nfds);
+	return evlist->pollfd != NULL ? 0 : -ENOMEM;
+}
+
+void perf_evlist__add_pollfd(struct perf_evlist *evlist, int fd)
+{
+	fcntl(fd, F_SETFL, O_NONBLOCK);
+	evlist->pollfd[evlist->nr_fds].fd = fd;
+	evlist->pollfd[evlist->nr_fds].events = POLLIN;
+	evlist->nr_fds++;
+}
+
+struct perf_evsel *perf_evlist__id2evsel(struct perf_evlist *evlist, u64 id)
+{
+	struct hlist_head *head;
+	struct hlist_node *pos;
+	struct perf_sample_id *sid;
+	int hash;
+
+	if (evlist->nr_entries == 1)
+		return list_entry(evlist->entries.next, struct perf_evsel, node);
+
+	hash = hash_64(id, PERF_EVLIST__HLIST_BITS);
+	head = &evlist->heads[hash];
+
+	hlist_for_each_entry(sid, pos, head, node)
+		if (sid->id == id)
+			return sid->evsel;
+	return NULL;
+}
+
+event_t *perf_evlist__read_on_cpu(struct perf_evlist *evlist, int cpu)
+{
+	/* XXX Move this to perf.c, making it generally available */
+	unsigned int page_size = sysconf(_SC_PAGE_SIZE);
+	struct perf_mmap *md = &evlist->mmap[cpu];
+	unsigned int head = perf_mmap__read_head(md);
+	unsigned int old = md->prev;
+	unsigned char *data = md->base + page_size;
+	event_t *event = NULL;
+	int diff;
+
+	/*
+	 * If we're further behind than half the buffer, there's a chance
+	 * the writer will bite our tail and mess up the samples under us.
+	 *
+	 * If we somehow ended up ahead of the head, we got messed up.
+	 *
+	 * In either case, truncate and restart at head.
+	 */
+	diff = head - old;
+	if (diff > md->mask / 2 || diff < 0) {
+		fprintf(stderr, "WARNING: failed to keep up with mmap data.\n");
+
+		/*
+		 * head points to a known good entry, start there.
+		 */
+		old = head;
+	}
+
+	if (old != head) {
+		size_t size;
+
+		event = (event_t *)&data[old & md->mask];
+		size = event->header.size;
+
+		/*
+		 * Event straddles the mmap boundary -- header should always
+		 * be inside due to u64 alignment of output.
+		 */
+		if ((old & md->mask) + size != ((old + size) & md->mask)) {
+			unsigned int offset = old;
+			unsigned int len = min(sizeof(*event), size), cpy;
+			void *dst = &evlist->event_copy;
+
+			do {
+				cpy = min(md->mask + 1 - (offset & md->mask), len);
+				memcpy(dst, &data[offset & md->mask], cpy);
+				offset += cpy;
+				dst += cpy;
+				len -= cpy;
+			} while (len);
+
+			event = &evlist->event_copy;
+		}
+
+		old += size;
+	}
+
+	md->prev = old;
+	return event;
+}

tools/perf/util/evlist.h

@@ -0,0 +1,41 @@
+#ifndef __PERF_EVLIST_H
+#define __PERF_EVLIST_H 1
+
+#include <linux/list.h>
+#include "../perf.h"
+#include "event.h"
+
+struct pollfd;
+
+#define PERF_EVLIST__HLIST_BITS 8
+#define PERF_EVLIST__HLIST_SIZE (1 << PERF_EVLIST__HLIST_BITS)
+
+struct perf_evlist {
+	struct list_head entries;
+	struct hlist_head heads[PERF_EVLIST__HLIST_SIZE];
+	int		 nr_entries;
+	int		 nr_fds;
+	int		 mmap_len;
+	event_t		 event_copy;
+	struct perf_mmap *mmap;
+	struct pollfd	 *pollfd;
+};
+
+struct perf_evsel;
+
+struct perf_evlist *perf_evlist__new(void);
+void perf_evlist__init(struct perf_evlist *evlist);
+void perf_evlist__exit(struct perf_evlist *evlist);
+void perf_evlist__delete(struct perf_evlist *evlist);
+
+void perf_evlist__add(struct perf_evlist *evlist, struct perf_evsel *entry);
+int perf_evlist__add_default(struct perf_evlist *evlist);
+
+int perf_evlist__alloc_pollfd(struct perf_evlist *evlist, int ncpus, int nthreads);
+void perf_evlist__add_pollfd(struct perf_evlist *evlist, int fd);
+
+struct perf_evsel *perf_evlist__id2evsel(struct perf_evlist *evlist, u64 id);
+
+event_t *perf_evlist__read_on_cpu(struct perf_evlist *self, int cpu);
+
+#endif /* __PERF_EVLIST_H */

tools/perf/util/evsel.c

@@ -1,20 +1,33 @@
 #include "evsel.h"
+#include "evlist.h"
 #include "../perf.h"
 #include "util.h"
 #include "cpumap.h"
-#include "thread.h"
+#include "thread_map.h"
+
+#include <unistd.h>
+#include <sys/mman.h>
+
+#include <linux/bitops.h>
+#include <linux/hash.h>
 
 #define FD(e, x, y) (*(int *)xyarray__entry(e->fd, x, y))
+#define SID(e, x, y) xyarray__entry(e->id, x, y)
+
+void perf_evsel__init(struct perf_evsel *evsel,
+		      struct perf_event_attr *attr, int idx)
+{
+	evsel->idx	   = idx;
+	evsel->attr	   = *attr;
+	INIT_LIST_HEAD(&evsel->node);
+}
 
 struct perf_evsel *perf_evsel__new(struct perf_event_attr *attr, int idx)
 {
 	struct perf_evsel *evsel = zalloc(sizeof(*evsel));
 
-	if (evsel != NULL) {
-		evsel->idx	   = idx;
-		evsel->attr	   = *attr;
-		INIT_LIST_HEAD(&evsel->node);
-	}
+	if (evsel != NULL)
+		perf_evsel__init(evsel, attr, idx);
 
 	return evsel;
 }
@@ -25,6 +38,12 @@ int perf_evsel__alloc_fd(struct perf_evsel *evsel, int ncpus, int nthreads)
 	return evsel->fd != NULL ? 0 : -ENOMEM;
 }
 
+int perf_evsel__alloc_id(struct perf_evsel *evsel, int ncpus, int nthreads)
+{
+	evsel->id = xyarray__new(ncpus, nthreads, sizeof(struct perf_sample_id));
+	return evsel->id != NULL ? 0 : -ENOMEM;
+}
+
 int perf_evsel__alloc_counts(struct perf_evsel *evsel, int ncpus)
 {
 	evsel->counts = zalloc((sizeof(*evsel->counts) +
@@ -38,6 +57,12 @@ void perf_evsel__free_fd(struct perf_evsel *evsel)
 	evsel->fd = NULL;
 }
 
+void perf_evsel__free_id(struct perf_evsel *evsel)
+{
+	xyarray__delete(evsel->id);
+	evsel->id = NULL;
+}
+
 void perf_evsel__close_fd(struct perf_evsel *evsel, int ncpus, int nthreads)
 {
 	int cpu, thread;
@@ -49,10 +74,34 @@ void perf_evsel__close_fd(struct perf_evsel *evsel, int ncpus, int nthreads)
 		}
 }
 
-void perf_evsel__delete(struct perf_evsel *evsel)
+void perf_evlist__munmap(struct perf_evlist *evlist, int ncpus)
+{
+	int cpu;
+
+	for (cpu = 0; cpu < ncpus; cpu++) {
+		if (evlist->mmap[cpu].base != NULL) {
+			munmap(evlist->mmap[cpu].base, evlist->mmap_len);
+			evlist->mmap[cpu].base = NULL;
+		}
+	}
+}
+
+int perf_evlist__alloc_mmap(struct perf_evlist *evlist, int ncpus)
+{
+	evlist->mmap = zalloc(ncpus * sizeof(struct perf_mmap));
+	return evlist->mmap != NULL ? 0 : -ENOMEM;
+}
+
+void perf_evsel__exit(struct perf_evsel *evsel)
 {
 	assert(list_empty(&evsel->node));
 	xyarray__delete(evsel->fd);
+	xyarray__delete(evsel->id);
+}
+
+void perf_evsel__delete(struct perf_evsel *evsel)
+{
+	perf_evsel__exit(evsel);
 	free(evsel);
 }
 
@@ -128,7 +177,7 @@ int __perf_evsel__read(struct perf_evsel *evsel,
 }
 
 static int __perf_evsel__open(struct perf_evsel *evsel, struct cpu_map *cpus,
-			      struct thread_map *threads)
+			      struct thread_map *threads, bool group, bool inherit)
 {
 	int cpu, thread;
 
@@ -137,12 +186,20 @@ static int __perf_evsel__open(struct perf_evsel *evsel, struct cpu_map *cpus,
 		return -1;
 
 	for (cpu = 0; cpu < cpus->nr; cpu++) {
+		int group_fd = -1;
+
+		evsel->attr.inherit = (cpus->map[cpu] < 0) && inherit;
+
 		for (thread = 0; thread < threads->nr; thread++) {
 			FD(evsel, cpu, thread) = sys_perf_event_open(&evsel->attr,
 								     threads->map[thread],
-								     cpus->map[cpu], -1, 0);
+								     cpus->map[cpu],
+								     group_fd, 0);
 			if (FD(evsel, cpu, thread) < 0)
 				goto out_close;
+
+			if (group && group_fd == -1)
+				group_fd = FD(evsel, cpu, thread);
 		}
 	}
 
@@ -175,10 +232,9 @@ static struct {
 	.threads = { -1, },
 };
 
-int perf_evsel__open(struct perf_evsel *evsel,
-		     struct cpu_map *cpus, struct thread_map *threads)
+int perf_evsel__open(struct perf_evsel *evsel, struct cpu_map *cpus,
+		     struct thread_map *threads, bool group, bool inherit)
 {
-
 	if (cpus == NULL) {
 		/* Work around old compiler warnings about strict aliasing */
 		cpus = &empty_cpu_map.map;
@@ -187,15 +243,243 @@ int perf_evsel__open(struct perf_evsel *evsel,
 	if (threads == NULL)
 		threads = &empty_thread_map.map;
 
-	return __perf_evsel__open(evsel, cpus, threads);
+	return __perf_evsel__open(evsel, cpus, threads, group, inherit);
+}
+
+int perf_evsel__open_per_cpu(struct perf_evsel *evsel,
+			     struct cpu_map *cpus, bool group, bool inherit)
+{
+	return __perf_evsel__open(evsel, cpus, &empty_thread_map.map, group, inherit);
+}
+
+int perf_evsel__open_per_thread(struct perf_evsel *evsel,
+				struct thread_map *threads, bool group, bool inherit)
+{
+	return __perf_evsel__open(evsel, &empty_cpu_map.map, threads, group, inherit);
+}
+
+static int __perf_evlist__mmap(struct perf_evlist *evlist, int cpu, int prot,
+			       int mask, int fd)
+{
+	evlist->mmap[cpu].prev = 0;
+	evlist->mmap[cpu].mask = mask;
+	evlist->mmap[cpu].base = mmap(NULL, evlist->mmap_len, prot,
+				      MAP_SHARED, fd, 0);
+	if (evlist->mmap[cpu].base == MAP_FAILED)
+		return -1;
+
+	perf_evlist__add_pollfd(evlist, fd);
+	return 0;
+}
+
+static int perf_evlist__id_hash(struct perf_evlist *evlist, struct perf_evsel *evsel,
+			       int cpu, int thread, int fd)
+{
+	struct perf_sample_id *sid;
+	u64 read_data[4] = { 0, };
+	int hash, id_idx = 1; /* The first entry is the counter value */
+
+	if (!(evsel->attr.read_format & PERF_FORMAT_ID) ||
+	    read(fd, &read_data, sizeof(read_data)) == -1)
+		return -1;
+
+	if (evsel->attr.read_format & PERF_FORMAT_TOTAL_TIME_ENABLED)
+		++id_idx;
+	if (evsel->attr.read_format & PERF_FORMAT_TOTAL_TIME_RUNNING)
+		++id_idx;
+
+	sid = SID(evsel, cpu, thread);
+	sid->id = read_data[id_idx];
+	sid->evsel = evsel;
+	hash = hash_64(sid->id, PERF_EVLIST__HLIST_BITS);
+	hlist_add_head(&sid->node, &evlist->heads[hash]);
+	return 0;
+}
+
+/** perf_evlist__mmap - Create per cpu maps to receive events
+ *
+ * @evlist - list of events
+ * @cpus - cpu map being monitored
+ * @threads - threads map being monitored
+ * @pages - map length in pages
+ * @overwrite - overwrite older events?
+ *
+ * If overwrite is false the user needs to signal event consuption using:
+ *
+ *	struct perf_mmap *m = &evlist->mmap[cpu];
+ *	unsigned int head = perf_mmap__read_head(m);
+ *
+ *	perf_mmap__write_tail(m, head)
+ */
+int perf_evlist__mmap(struct perf_evlist *evlist, struct cpu_map *cpus,
+		      struct thread_map *threads, int pages, bool overwrite)
+{
+	unsigned int page_size = sysconf(_SC_PAGE_SIZE);
+	int mask = pages * page_size - 1, cpu;
+	struct perf_evsel *first_evsel, *evsel;
+	int thread, prot = PROT_READ | (overwrite ? 0 : PROT_WRITE);
+
+	if (evlist->mmap == NULL &&
+	    perf_evlist__alloc_mmap(evlist, cpus->nr) < 0)
+		return -ENOMEM;
+
+	if (evlist->pollfd == NULL &&
+	    perf_evlist__alloc_pollfd(evlist, cpus->nr, threads->nr) < 0)
+		return -ENOMEM;
+
+	evlist->mmap_len = (pages + 1) * page_size;
+	first_evsel = list_entry(evlist->entries.next, struct perf_evsel, node);
+
+	list_for_each_entry(evsel, &evlist->entries, node) {
+		if ((evsel->attr.read_format & PERF_FORMAT_ID) &&
+		    evsel->id == NULL &&
+		    perf_evsel__alloc_id(evsel, cpus->nr, threads->nr) < 0)
+			return -ENOMEM;
+
+		for (cpu = 0; cpu < cpus->nr; cpu++) {
+			for (thread = 0; thread < threads->nr; thread++) {
+				int fd = FD(evsel, cpu, thread);
+
+				if (evsel->idx || thread) {
+					if (ioctl(fd, PERF_EVENT_IOC_SET_OUTPUT,
+						  FD(first_evsel, cpu, 0)) != 0)
+						goto out_unmap;
+				} else if (__perf_evlist__mmap(evlist, cpu, prot, mask, fd) < 0)
+					goto out_unmap;
+
+				if ((evsel->attr.read_format & PERF_FORMAT_ID) &&
+				    perf_evlist__id_hash(evlist, evsel, cpu, thread, fd) < 0)
+					goto out_unmap;
+			}
+		}
+	}
+
+	return 0;
+
+out_unmap:
+	for (cpu = 0; cpu < cpus->nr; cpu++) {
+		if (evlist->mmap[cpu].base != NULL) {
+			munmap(evlist->mmap[cpu].base, evlist->mmap_len);
+			evlist->mmap[cpu].base = NULL;
+		}
+	}
+	return -1;
 }
 
-int perf_evsel__open_per_cpu(struct perf_evsel *evsel, struct cpu_map *cpus)
+static int event__parse_id_sample(const event_t *event, u64 type,
+				  struct sample_data *sample)
 {
-	return __perf_evsel__open(evsel, cpus, &empty_thread_map.map);
+	const u64 *array = event->sample.array;
+
+	array += ((event->header.size -
+		   sizeof(event->header)) / sizeof(u64)) - 1;
+
+	if (type & PERF_SAMPLE_CPU) {
+		u32 *p = (u32 *)array;
+		sample->cpu = *p;
+		array--;
+	}
+
+	if (type & PERF_SAMPLE_STREAM_ID) {
+		sample->stream_id = *array;
+		array--;
+	}
+
+	if (type & PERF_SAMPLE_ID) {
+		sample->id = *array;
+		array--;
+	}
+
+	if (type & PERF_SAMPLE_TIME) {
+		sample->time = *array;
+		array--;
+	}
+
+	if (type & PERF_SAMPLE_TID) {
+		u32 *p = (u32 *)array;
+		sample->pid = p[0];
+		sample->tid = p[1];
+	}
+
+	return 0;
 }
 
-int perf_evsel__open_per_thread(struct perf_evsel *evsel, struct thread_map *threads)
+int event__parse_sample(const event_t *event, u64 type, bool sample_id_all,
+			struct sample_data *data)
 {
-	return __perf_evsel__open(evsel, &empty_cpu_map.map, threads);
+	const u64 *array;
+
+	data->cpu = data->pid = data->tid = -1;
+	data->stream_id = data->id = data->time = -1ULL;
+
+	if (event->header.type != PERF_RECORD_SAMPLE) {
+		if (!sample_id_all)
+			return 0;
+		return event__parse_id_sample(event, type, data);
+	}
+
+	array = event->sample.array;
+
+	if (type & PERF_SAMPLE_IP) {
+		data->ip = event->ip.ip;
+		array++;
+	}
+
+	if (type & PERF_SAMPLE_TID) {
+		u32 *p = (u32 *)array;
+		data->pid = p[0];
+		data->tid = p[1];
+		array++;
+	}
+
+	if (type & PERF_SAMPLE_TIME) {
+		data->time = *array;
+		array++;
+	}
+
+	if (type & PERF_SAMPLE_ADDR) {
+		data->addr = *array;
+		array++;
+	}
+
+	data->id = -1ULL;
+	if (type & PERF_SAMPLE_ID) {
+		data->id = *array;
+		array++;
+	}
+
+	if (type & PERF_SAMPLE_STREAM_ID) {
+		data->stream_id = *array;
+		array++;
+	}
+
+	if (type & PERF_SAMPLE_CPU) {
+		u32 *p = (u32 *)array;
+		data->cpu = *p;
+		array++;
+	}
+
+	if (type & PERF_SAMPLE_PERIOD) {
+		data->period = *array;
+		array++;
+	}
+
+	if (type & PERF_SAMPLE_READ) {
+		fprintf(stderr, "PERF_SAMPLE_READ is unsuported for now\n");
+		return -1;
+	}
+
+	if (type & PERF_SAMPLE_CALLCHAIN) {
+		data->callchain = (struct ip_callchain *)array;
+		array += 1 + data->callchain->nr;
+	}
+
+	if (type & PERF_SAMPLE_RAW) {
+		u32 *p = (u32 *)array;
+		data->raw_size = *p;
+		p++;
+		data->raw_data = p;
+	}
+
+	return 0;
 }

tools/perf/util/evsel.h

@@ -24,11 +24,24 @@ struct perf_counts {
 	struct perf_counts_values cpu[];
 };
 
+struct perf_evsel;
+
+/*
+ * Per fd, to map back from PERF_SAMPLE_ID to evsel, only used when there are
+ * more than one entry in the evlist.
+ */
+struct perf_sample_id {
+	struct hlist_node 	node;
+	u64		 	id;
+	struct perf_evsel	*evsel;
+};
+
 struct perf_evsel {
 	struct list_head	node;
 	struct perf_event_attr	attr;
 	char			*filter;
 	struct xyarray		*fd;
+	struct xyarray		*id;
 	struct perf_counts	*counts;
 	int			idx;
 	void			*priv;
@@ -36,19 +49,31 @@ struct perf_evsel {
 
 struct cpu_map;
 struct thread_map;
+struct perf_evlist;
 
 struct perf_evsel *perf_evsel__new(struct perf_event_attr *attr, int idx);
+void perf_evsel__init(struct perf_evsel *evsel,
+		      struct perf_event_attr *attr, int idx);
+void perf_evsel__exit(struct perf_evsel *evsel);
 void perf_evsel__delete(struct perf_evsel *evsel);
 
 int perf_evsel__alloc_fd(struct perf_evsel *evsel, int ncpus, int nthreads);
+int perf_evsel__alloc_id(struct perf_evsel *evsel, int ncpus, int nthreads);
 int perf_evsel__alloc_counts(struct perf_evsel *evsel, int ncpus);
+int perf_evlist__alloc_mmap(struct perf_evlist *evlist, int ncpus);
 void perf_evsel__free_fd(struct perf_evsel *evsel);
+void perf_evsel__free_id(struct perf_evsel *evsel);
 void perf_evsel__close_fd(struct perf_evsel *evsel, int ncpus, int nthreads);
 
-int perf_evsel__open_per_cpu(struct perf_evsel *evsel, struct cpu_map *cpus);
-int perf_evsel__open_per_thread(struct perf_evsel *evsel, struct thread_map *threads);
-int perf_evsel__open(struct perf_evsel *evsel, 
-		     struct cpu_map *cpus, struct thread_map *threads);
+int perf_evsel__open_per_cpu(struct perf_evsel *evsel,
+			     struct cpu_map *cpus, bool group, bool inherit);
+int perf_evsel__open_per_thread(struct perf_evsel *evsel,
+				struct thread_map *threads, bool group, bool inherit);
+int perf_evsel__open(struct perf_evsel *evsel, struct cpu_map *cpus,
+		     struct thread_map *threads, bool group, bool inherit);
+int perf_evlist__mmap(struct perf_evlist *evlist, struct cpu_map *cpus,
+		      struct thread_map *threads, int pages, bool overwrite);
+void perf_evlist__munmap(struct perf_evlist *evlist, int ncpus);
 
 #define perf_evsel__match(evsel, t, c)		\
 	(evsel->attr.type == PERF_TYPE_##t &&	\

tools/perf/util/header.c

@@ -8,6 +8,7 @@
 #include <linux/list.h>
 #include <linux/kernel.h>
 
+#include "evlist.h"
 #include "util.h"
 #include "header.h"
 #include "../perf.h"
@@ -428,7 +429,8 @@ static bool perf_session__read_build_ids(struct perf_session *self, bool with_hi
 	return ret;
 }
 
-static int perf_header__adds_write(struct perf_header *self, int fd)
+static int perf_header__adds_write(struct perf_header *self,
+				   struct perf_evlist *evlist, int fd)
 {
 	int nr_sections;
 	struct perf_session *session;
@@ -463,7 +465,7 @@ static int perf_header__adds_write(struct perf_header *self, int fd)
 
 		/* Write trace info */
 		trace_sec->offset = lseek(fd, 0, SEEK_CUR);
-		read_tracing_data(fd, &evsel_list);
+		read_tracing_data(fd, &evlist->entries);
 		trace_sec->size = lseek(fd, 0, SEEK_CUR) - trace_sec->offset;
 	}
 
@@ -513,7 +515,8 @@ int perf_header__write_pipe(int fd)
 	return 0;
 }
 
-int perf_header__write(struct perf_header *self, int fd, bool at_exit)
+int perf_header__write(struct perf_header *self, struct perf_evlist *evlist,
+		       int fd, bool at_exit)
 {
 	struct perf_file_header f_header;
 	struct perf_file_attr   f_attr;
@@ -566,7 +569,7 @@ int perf_header__write(struct perf_header *self, int fd, bool at_exit)
 	self->data_offset = lseek(fd, 0, SEEK_CUR);
 
 	if (at_exit) {
-		err = perf_header__adds_write(self, fd);
+		err = perf_header__adds_write(self, evlist, fd);
 		if (err < 0)
 			return err;
 	}
@@ -1133,7 +1136,7 @@ int event__process_event_type(event_t *self,
 	return 0;
 }
 
-int event__synthesize_tracing_data(int fd, struct list_head *pattrs,
+int event__synthesize_tracing_data(int fd, struct perf_evlist *evlist,
 				   event__handler_t process,
 				   struct perf_session *session __unused)
 {
@@ -1144,7 +1147,7 @@ int event__synthesize_tracing_data(int fd, struct list_head *pattrs,
 	memset(&ev, 0, sizeof(ev));
 
 	ev.tracing_data.header.type = PERF_RECORD_HEADER_TRACING_DATA;
-	size = read_tracing_data_size(fd, pattrs);
+	size = read_tracing_data_size(fd, &evlist->entries);
 	if (size <= 0)
 		return size;
 	aligned_size = ALIGN(size, sizeof(u64));
@@ -1154,7 +1157,7 @@ int event__synthesize_tracing_data(int fd, struct list_head *pattrs,
 
 	process(&ev, NULL, session);
 
-	err = read_tracing_data(fd, pattrs);
+	err = read_tracing_data(fd, &evlist->entries);
 	write_padded(fd, NULL, 0, padding);
 
 	return aligned_size;

tools/perf/util/header.h

@@ -65,8 +65,11 @@ struct perf_header {
 int perf_header__init(struct perf_header *self);
 void perf_header__exit(struct perf_header *self);
 
+struct perf_evlist;
+
 int perf_header__read(struct perf_session *session, int fd);
-int perf_header__write(struct perf_header *self, int fd, bool at_exit);
+int perf_header__write(struct perf_header *self, struct perf_evlist *evlist,
+		       int fd, bool at_exit);
 int perf_header__write_pipe(int fd);
 
 int perf_header__add_attr(struct perf_header *self,
@@ -113,7 +116,7 @@ int event__synthesize_event_types(event__handler_t process,
 int event__process_event_type(event_t *self,
 			      struct perf_session *session);
 
-int event__synthesize_tracing_data(int fd, struct list_head *pattrs,
+int event__synthesize_tracing_data(int fd, struct perf_evlist *evlist,
 				   event__handler_t process,
 				   struct perf_session *session);
 int event__process_tracing_data(event_t *self,

tools/perf/util/hist.c

@@ -211,7 +211,9 @@ void hist_entry__free(struct hist_entry *he)
  * collapse the histogram
  */
 
-static bool collapse__insert_entry(struct rb_root *root, struct hist_entry *he)
+static bool hists__collapse_insert_entry(struct hists *self,
+					 struct rb_root *root,
+					 struct hist_entry *he)
 {
 	struct rb_node **p = &root->rb_node;
 	struct rb_node *parent = NULL;
@@ -226,8 +228,11 @@ static bool collapse__insert_entry(struct rb_root *root, struct hist_entry *he)
 
 		if (!cmp) {
 			iter->period += he->period;
-			if (symbol_conf.use_callchain)
-				callchain_merge(iter->callchain, he->callchain);
+			if (symbol_conf.use_callchain) {
+				callchain_cursor_reset(&self->callchain_cursor);
+				callchain_merge(&self->callchain_cursor, iter->callchain,
+						he->callchain);
+			}
 			hist_entry__free(he);
 			return false;
 		}
@@ -262,7 +267,7 @@ void hists__collapse_resort(struct hists *self)
 		next = rb_next(&n->rb_node);
 
 		rb_erase(&n->rb_node, &self->entries);
-		if (collapse__insert_entry(&tmp, n))
+		if (hists__collapse_insert_entry(self, &tmp, n))
 			hists__inc_nr_entries(self, n);
 	}
 
@@ -425,7 +430,7 @@ static size_t __callchain__fprintf_graph(FILE *fp, struct callchain_node *self,
 		u64 cumul;
 
 		child = rb_entry(node, struct callchain_node, rb_node);
-		cumul = cumul_hits(child);
+		cumul = callchain_cumul_hits(child);
 		remaining -= cumul;
 
 		/*

tools/perf/util/hist.h

@@ -77,6 +77,8 @@ struct hists {
 	u64			event_stream;
 	u32			type;
 	u16			col_len[HISTC_NR_COLS];
+	/* Best would be to reuse the session callchain cursor */
+	struct callchain_cursor	callchain_cursor;
 };
 
 struct hist_entry *__hists__add_entry(struct hists *self,

tools/perf/util/include/linux/list.h

@@ -1,3 +1,4 @@
+#include <linux/kernel.h>
 #include "../../../../include/linux/list.h"
 
 #ifndef PERF_LIST_H

tools/perf/util/parse-events.c

@@ -1,6 +1,7 @@
 #include "../../../include/linux/hw_breakpoint.h"
 #include "util.h"
 #include "../perf.h"
+#include "evlist.h"
 #include "evsel.h"
 #include "parse-options.h"
 #include "parse-events.h"
@@ -11,10 +12,6 @@
 #include "header.h"
 #include "debugfs.h"
 
-int				nr_counters;
-
-LIST_HEAD(evsel_list);
-
 struct event_symbol {
 	u8		type;
 	u64		config;
@@ -449,8 +446,8 @@ parse_single_tracepoint_event(char *sys_name,
 /* sys + ':' + event + ':' + flags*/
 #define MAX_EVOPT_LEN	(MAX_EVENT_LENGTH * 2 + 2 + 128)
 static enum event_result
-parse_multiple_tracepoint_event(char *sys_name, const char *evt_exp,
-				char *flags)
+parse_multiple_tracepoint_event(const struct option *opt, char *sys_name,
+				const char *evt_exp, char *flags)
 {
 	char evt_path[MAXPATHLEN];
 	struct dirent *evt_ent;
@@ -483,15 +480,16 @@ parse_multiple_tracepoint_event(char *sys_name, const char *evt_exp,
 		if (len < 0)
 			return EVT_FAILED;
 
-		if (parse_events(NULL, event_opt, 0))
+		if (parse_events(opt, event_opt, 0))
 			return EVT_FAILED;
 	}
 
 	return EVT_HANDLED_ALL;
 }
 
-static enum event_result parse_tracepoint_event(const char **strp,
-				    struct perf_event_attr *attr)
+static enum event_result
+parse_tracepoint_event(const struct option *opt, const char **strp,
+		       struct perf_event_attr *attr)
 {
 	const char *evt_name;
 	char *flags = NULL, *comma_loc;
@@ -530,7 +528,7 @@ static enum event_result parse_tracepoint_event(const char **strp,
 		return EVT_FAILED;
 	if (strpbrk(evt_name, "*?")) {
 		*strp += strlen(sys_name) + evt_length + 1; /* 1 == the ':' */
-		return parse_multiple_tracepoint_event(sys_name, evt_name,
+		return parse_multiple_tracepoint_event(opt, sys_name, evt_name,
 						       flags);
 	} else {
 		return parse_single_tracepoint_event(sys_name, evt_name,
@@ -740,11 +738,12 @@ parse_event_modifier(const char **strp, struct perf_event_attr *attr)
  * Symbolic names are (almost) exactly matched.
  */
 static enum event_result
-parse_event_symbols(const char **str, struct perf_event_attr *attr)
+parse_event_symbols(const struct option *opt, const char **str,
+		    struct perf_event_attr *attr)
 {
 	enum event_result ret;
 
-	ret = parse_tracepoint_event(str, attr);
+	ret = parse_tracepoint_event(opt, str, attr);
 	if (ret != EVT_FAILED)
 		goto modifier;
 
@@ -778,14 +777,15 @@ modifier:
 	return ret;
 }
 
-int parse_events(const struct option *opt __used, const char *str, int unset __used)
+int parse_events(const struct option *opt, const char *str, int unset __used)
 {
+	struct perf_evlist *evlist = *(struct perf_evlist **)opt->value;
 	struct perf_event_attr attr;
 	enum event_result ret;
 
 	for (;;) {
 		memset(&attr, 0, sizeof(attr));
-		ret = parse_event_symbols(&str, &attr);
+		ret = parse_event_symbols(opt, &str, &attr);
 		if (ret == EVT_FAILED)
 			return -1;
 
@@ -794,12 +794,10 @@ int parse_events(const struct option *opt __used, const char *str, int unset __u
 
 		if (ret != EVT_HANDLED_ALL) {
 			struct perf_evsel *evsel;
-			evsel = perf_evsel__new(&attr,
-						nr_counters);
+			evsel = perf_evsel__new(&attr, evlist->nr_entries);
 			if (evsel == NULL)
 				return -1;
-			list_add_tail(&evsel->node, &evsel_list);
-			++nr_counters;
+			perf_evlist__add(evlist, evsel);
 		}
 
 		if (*str == 0)
@@ -813,13 +811,14 @@ int parse_events(const struct option *opt __used, const char *str, int unset __u
 	return 0;
 }
 
-int parse_filter(const struct option *opt __used, const char *str,
+int parse_filter(const struct option *opt, const char *str,
 		 int unset __used)
 {
+	struct perf_evlist *evlist = *(struct perf_evlist **)opt->value;
 	struct perf_evsel *last = NULL;
 
-	if (!list_empty(&evsel_list))
-		last = list_entry(evsel_list.prev, struct perf_evsel, node);
+	if (evlist->nr_entries > 0)
+		last = list_entry(evlist->entries.prev, struct perf_evsel, node);
 
 	if (last == NULL || last->attr.type != PERF_TYPE_TRACEPOINT) {
 		fprintf(stderr,
@@ -981,33 +980,3 @@ void print_events(void)
 
 	exit(129);
 }
-
-int perf_evsel_list__create_default(void)
-{
-	struct perf_evsel *evsel;
-	struct perf_event_attr attr;
-
-	memset(&attr, 0, sizeof(attr));
-	attr.type = PERF_TYPE_HARDWARE;
-	attr.config = PERF_COUNT_HW_CPU_CYCLES;
-
-	evsel = perf_evsel__new(&attr, 0);
-
-	if (evsel == NULL)
-		return -ENOMEM;
-
-	list_add(&evsel->node, &evsel_list);
-	++nr_counters;
-	return 0;
-}
-
-void perf_evsel_list__delete(void)
-{
-	struct perf_evsel *pos, *n;
-
-	list_for_each_entry_safe(pos, n, &evsel_list, node) {
-		list_del_init(&pos->node);
-		perf_evsel__delete(pos);
-	}
-	nr_counters = 0;
-}

tools/perf/util/parse-events.h

@@ -9,11 +9,6 @@
 struct list_head;
 struct perf_evsel;
 
-extern struct list_head evsel_list;
-
-int perf_evsel_list__create_default(void);
-void perf_evsel_list__delete(void);
-
 struct option;
 
 struct tracepoint_path {
@@ -25,8 +20,6 @@ struct tracepoint_path {
 extern struct tracepoint_path *tracepoint_id_to_path(u64 config);
 extern bool have_tracepoints(struct list_head *evlist);
 
-extern int			nr_counters;
-
 const char *event_name(struct perf_evsel *event);
 extern const char *__event_name(int type, u64 config);
 

tools/perf/util/probe-event.c

@@ -31,6 +31,7 @@
 #include <string.h>
 #include <stdarg.h>
 #include <limits.h>
+#include <elf.h>
 
 #undef _GNU_SOURCE
 #include "util.h"
@@ -111,7 +112,25 @@ static struct symbol *__find_kernel_function_by_name(const char *name,
 						     NULL);
 }
 
-const char *kernel_get_module_path(const char *module)
+static struct map *kernel_get_module_map(const char *module)
+{
+	struct rb_node *nd;
+	struct map_groups *grp = &machine.kmaps;
+
+	if (!module)
+		module = "kernel";
+
+	for (nd = rb_first(&grp->maps[MAP__FUNCTION]); nd; nd = rb_next(nd)) {
+		struct map *pos = rb_entry(nd, struct map, rb_node);
+		if (strncmp(pos->dso->short_name + 1, module,
+			    pos->dso->short_name_len - 2) == 0) {
+			return pos;
+		}
+	}
+	return NULL;
+}
+
+static struct dso *kernel_get_module_dso(const char *module)
 {
 	struct dso *dso;
 	struct map *map;
@@ -141,7 +160,13 @@ const char *kernel_get_module_path(const char *module)
 		}
 	}
 found:
-	return dso->long_name;
+	return dso;
+}
+
+const char *kernel_get_module_path(const char *module)
+{
+	struct dso *dso = kernel_get_module_dso(module);
+	return (dso) ? dso->long_name : NULL;
 }
 
 #ifdef DWARF_SUPPORT
@@ -1913,3 +1938,42 @@ int del_perf_probe_events(struct strlist *dellist)
 	return ret;
 }
 
+/*
+ * If a symbol corresponds to a function with global binding return 0.
+ * For all others return 1.
+ */
+static int filter_non_global_functions(struct map *map __unused,
+					struct symbol *sym)
+{
+	if (sym->binding != STB_GLOBAL)
+		return 1;
+
+	return 0;
+}
+
+int show_available_funcs(const char *module)
+{
+	struct map *map;
+	int ret;
+
+	setup_pager();
+
+	ret = init_vmlinux();
+	if (ret < 0)
+		return ret;
+
+	map = kernel_get_module_map(module);
+	if (!map) {
+		pr_err("Failed to find %s map.\n", (module) ? : "kernel");
+		return -EINVAL;
+	}
+	if (map__load(map, filter_non_global_functions)) {
+		pr_err("Failed to load map.\n");
+		return -EINVAL;
+	}
+	if (!dso__sorted_by_name(map->dso, map->type))
+		dso__sort_by_name(map->dso, map->type);
+
+	dso__fprintf_symbols_by_name(map->dso, map->type, stdout);
+	return 0;
+}

tools/perf/util/probe-event.h

@@ -127,6 +127,7 @@ extern int show_line_range(struct line_range *lr, const char *module);
 extern int show_available_vars(struct perf_probe_event *pevs, int npevs,
 			       int max_probe_points, const char *module,
 			       bool externs);
+extern int show_available_funcs(const char *module);
 
 
 /* Maximum index number of event-name postfix */

tools/perf/util/probe-finder.c

@@ -280,6 +280,19 @@ static bool die_compare_name(Dwarf_Die *dw_die, const char *tname)
 	return name ? (strcmp(tname, name) == 0) : false;
 }
 
+/* Get callsite line number of inline-function instance */
+static int die_get_call_lineno(Dwarf_Die *in_die)
+{
+	Dwarf_Attribute attr;
+	Dwarf_Word ret;
+
+	if (!dwarf_attr(in_die, DW_AT_call_line, &attr))
+		return -ENOENT;
+
+	dwarf_formudata(&attr, &ret);
+	return (int)ret;
+}
+
 /* Get type die */
 static Dwarf_Die *die_get_type(Dwarf_Die *vr_die, Dwarf_Die *die_mem)
 {
@@ -458,6 +471,151 @@ static Dwarf_Die *die_find_inlinefunc(Dwarf_Die *sp_die, Dwarf_Addr addr,
 	return die_find_child(sp_die, __die_find_inline_cb, &addr, die_mem);
 }
 
+/* Walker on lines (Note: line number will not be sorted) */
+typedef int (* line_walk_handler_t) (const char *fname, int lineno,
+				     Dwarf_Addr addr, void *data);
+
+struct __line_walk_param {
+	const char *fname;
+	line_walk_handler_t handler;
+	void *data;
+	int retval;
+};
+
+static int __die_walk_funclines_cb(Dwarf_Die *in_die, void *data)
+{
+	struct __line_walk_param *lw = data;
+	Dwarf_Addr addr;
+	int lineno;
+
+	if (dwarf_tag(in_die) == DW_TAG_inlined_subroutine) {
+		lineno = die_get_call_lineno(in_die);
+		if (lineno > 0 && dwarf_entrypc(in_die, &addr) == 0) {
+			lw->retval = lw->handler(lw->fname, lineno, addr,
+						 lw->data);
+			if (lw->retval != 0)
+				return DIE_FIND_CB_FOUND;
+		}
+	}
+	return DIE_FIND_CB_SIBLING;
+}
+
+/* Walk on lines of blocks included in given DIE */
+static int __die_walk_funclines(Dwarf_Die *sp_die,
+				line_walk_handler_t handler, void *data)
+{
+	struct __line_walk_param lw = {
+		.handler = handler,
+		.data = data,
+		.retval = 0,
+	};
+	Dwarf_Die die_mem;
+	Dwarf_Addr addr;
+	int lineno;
+
+	/* Handle function declaration line */
+	lw.fname = dwarf_decl_file(sp_die);
+	if (lw.fname && dwarf_decl_line(sp_die, &lineno) == 0 &&
+	    dwarf_entrypc(sp_die, &addr) == 0) {
+		lw.retval = handler(lw.fname, lineno, addr, data);
+		if (lw.retval != 0)
+			goto done;
+	}
+	die_find_child(sp_die, __die_walk_funclines_cb, &lw, &die_mem);
+done:
+	return lw.retval;
+}
+
+static int __die_walk_culines_cb(Dwarf_Die *sp_die, void *data)
+{
+	struct __line_walk_param *lw = data;
+
+	lw->retval = __die_walk_funclines(sp_die, lw->handler, lw->data);
+	if (lw->retval != 0)
+		return DWARF_CB_ABORT;
+
+	return DWARF_CB_OK;
+}
+
+/*
+ * Walk on lines inside given PDIE. If the PDIE is subprogram, walk only on
+ * the lines inside the subprogram, otherwise PDIE must be a CU DIE.
+ */
+static int die_walk_lines(Dwarf_Die *pdie, line_walk_handler_t handler,
+			  void *data)
+{
+	Dwarf_Lines *lines;
+	Dwarf_Line *line;
+	Dwarf_Addr addr;
+	const char *fname;
+	int lineno, ret = 0;
+	Dwarf_Die die_mem, *cu_die;
+	size_t nlines, i;
+
+	/* Get the CU die */
+	if (dwarf_tag(pdie) == DW_TAG_subprogram)
+		cu_die = dwarf_diecu(pdie, &die_mem, NULL, NULL);
+	else
+		cu_die = pdie;
+	if (!cu_die) {
+		pr_debug2("Failed to get CU from subprogram\n");
+		return -EINVAL;
+	}
+
+	/* Get lines list in the CU */
+	if (dwarf_getsrclines(cu_die, &lines, &nlines) != 0) {
+		pr_debug2("Failed to get source lines on this CU.\n");
+		return -ENOENT;
+	}
+	pr_debug2("Get %zd lines from this CU\n", nlines);
+
+	/* Walk on the lines on lines list */
+	for (i = 0; i < nlines; i++) {
+		line = dwarf_onesrcline(lines, i);
+		if (line == NULL ||
+		    dwarf_lineno(line, &lineno) != 0 ||
+		    dwarf_lineaddr(line, &addr) != 0) {
+			pr_debug2("Failed to get line info. "
+				  "Possible error in debuginfo.\n");
+			continue;
+		}
+		/* Filter lines based on address */
+		if (pdie != cu_die)
+			/*
+			 * Address filtering
+			 * The line is included in given function, and
+			 * no inline block includes it.
+			 */
+			if (!dwarf_haspc(pdie, addr) ||
+			    die_find_inlinefunc(pdie, addr, &die_mem))
+				continue;
+		/* Get source line */
+		fname = dwarf_linesrc(line, NULL, NULL);
+
+		ret = handler(fname, lineno, addr, data);
+		if (ret != 0)
+			return ret;
+	}
+
+	/*
+	 * Dwarf lines doesn't include function declarations and inlined
+	 * subroutines. We have to check functions list or given function.
+	 */
+	if (pdie != cu_die)
+		ret = __die_walk_funclines(pdie, handler, data);
+	else {
+		struct __line_walk_param param = {
+			.handler = handler,
+			.data = data,
+			.retval = 0,
+		};
+		dwarf_getfuncs(cu_die, __die_walk_culines_cb, &param, 0);
+		ret = param.retval;
+	}
+
+	return ret;
+}
+
 struct __find_variable_param {
 	const char *name;
 	Dwarf_Addr addr;
@@ -1050,43 +1208,26 @@ static int call_probe_finder(Dwarf_Die *sp_die, struct probe_finder *pf)
 	return ret;
 }
 
-/* Find probe point from its line number */
-static int find_probe_point_by_line(struct probe_finder *pf)
+static int probe_point_line_walker(const char *fname, int lineno,
+				   Dwarf_Addr addr, void *data)
 {
-	Dwarf_Lines *lines;
-	Dwarf_Line *line;
-	size_t nlines, i;
-	Dwarf_Addr addr;
-	int lineno;
-	int ret = 0;
-
-	if (dwarf_getsrclines(&pf->cu_die, &lines, &nlines) != 0) {
-		pr_warning("No source lines found.\n");
-		return -ENOENT;
-	}
+	struct probe_finder *pf = data;
+	int ret;
 
-	for (i = 0; i < nlines && ret == 0; i++) {
-		line = dwarf_onesrcline(lines, i);
-		if (dwarf_lineno(line, &lineno) != 0 ||
-		    lineno != pf->lno)
-			continue;
+	if (lineno != pf->lno || strtailcmp(fname, pf->fname) != 0)
+		return 0;
 
-		/* TODO: Get fileno from line, but how? */
-		if (strtailcmp(dwarf_linesrc(line, NULL, NULL), pf->fname) != 0)
-			continue;
+	pf->addr = addr;
+	ret = call_probe_finder(NULL, pf);
 
-		if (dwarf_lineaddr(line, &addr) != 0) {
-			pr_warning("Failed to get the address of the line.\n");
-			return -ENOENT;
-		}
-		pr_debug("Probe line found: line[%d]:%d addr:0x%jx\n",
-			 (int)i, lineno, (uintmax_t)addr);
-		pf->addr = addr;
+	/* Continue if no error, because the line will be in inline function */
+	return ret < 0 ?: 0;
+}
 
-		ret = call_probe_finder(NULL, pf);
-		/* Continuing, because target line might be inlined. */
-	}
-	return ret;
+/* Find probe point from its line number */
+static int find_probe_point_by_line(struct probe_finder *pf)
+{
+	return die_walk_lines(&pf->cu_die, probe_point_line_walker, pf);
 }
 
 /* Find lines which match lazy pattern */
@@ -1140,15 +1281,31 @@ out_close:
 	return nlines;
 }
 
+static int probe_point_lazy_walker(const char *fname, int lineno,
+				   Dwarf_Addr addr, void *data)
+{
+	struct probe_finder *pf = data;
+	int ret;
+
+	if (!line_list__has_line(&pf->lcache, lineno) ||
+	    strtailcmp(fname, pf->fname) != 0)
+		return 0;
+
+	pr_debug("Probe line found: line:%d addr:0x%llx\n",
+		 lineno, (unsigned long long)addr);
+	pf->addr = addr;
+	ret = call_probe_finder(NULL, pf);
+
+	/*
+	 * Continue if no error, because the lazy pattern will match
+	 * to other lines
+	 */
+	return ret < 0 ?: 0;
+}
+
 /* Find probe points from lazy pattern  */
 static int find_probe_point_lazy(Dwarf_Die *sp_die, struct probe_finder *pf)
 {
-	Dwarf_Lines *lines;
-	Dwarf_Line *line;
-	size_t nlines, i;
-	Dwarf_Addr addr;
-	Dwarf_Die die_mem;
-	int lineno;
 	int ret = 0;
 
 	if (list_empty(&pf->lcache)) {
@@ -1162,45 +1319,7 @@ static int find_probe_point_lazy(Dwarf_Die *sp_die, struct probe_finder *pf)
 			return ret;
 	}
 
-	if (dwarf_getsrclines(&pf->cu_die, &lines, &nlines) != 0) {
-		pr_warning("No source lines found.\n");
-		return -ENOENT;
-	}
-
-	for (i = 0; i < nlines && ret >= 0; i++) {
-		line = dwarf_onesrcline(lines, i);
-
-		if (dwarf_lineno(line, &lineno) != 0 ||
-		    !line_list__has_line(&pf->lcache, lineno))
-			continue;
-
-		/* TODO: Get fileno from line, but how? */
-		if (strtailcmp(dwarf_linesrc(line, NULL, NULL), pf->fname) != 0)
-			continue;
-
-		if (dwarf_lineaddr(line, &addr) != 0) {
-			pr_debug("Failed to get the address of line %d.\n",
-				 lineno);
-			continue;
-		}
-		if (sp_die) {
-			/* Address filtering 1: does sp_die include addr? */
-			if (!dwarf_haspc(sp_die, addr))
-				continue;
-			/* Address filtering 2: No child include addr? */
-			if (die_find_inlinefunc(sp_die, addr, &die_mem))
-				continue;
-		}
-
-		pr_debug("Probe line found: line[%d]:%d addr:0x%llx\n",
-			 (int)i, lineno, (unsigned long long)addr);
-		pf->addr = addr;
-
-		ret = call_probe_finder(sp_die, pf);
-		/* Continuing, because target line might be inlined. */
-	}
-	/* TODO: deallocate lines, but how? */
-	return ret;
+	return die_walk_lines(sp_die, probe_point_lazy_walker, pf);
 }
 
 /* Callback parameter with return value */
@@ -1644,91 +1763,28 @@ static int line_range_add_line(const char *src, unsigned int lineno,
 	return line_list__add_line(&lr->line_list, lineno);
 }
 
-/* Search function declaration lines */
-static int line_range_funcdecl_cb(Dwarf_Die *sp_die, void *data)
+static int line_range_walk_cb(const char *fname, int lineno,
+			      Dwarf_Addr addr __used,
+			      void *data)
 {
-	struct dwarf_callback_param *param = data;
-	struct line_finder *lf = param->data;
-	const char *src;
-	int lineno;
+	struct line_finder *lf = data;
 
-	src = dwarf_decl_file(sp_die);
-	if (src && strtailcmp(src, lf->fname) != 0)
-		return DWARF_CB_OK;
-
-	if (dwarf_decl_line(sp_die, &lineno) != 0 ||
+	if ((strtailcmp(fname, lf->fname) != 0) ||
 	    (lf->lno_s > lineno || lf->lno_e < lineno))
-		return DWARF_CB_OK;
+		return 0;
 
-	param->retval = line_range_add_line(src, lineno, lf->lr);
-	if (param->retval < 0)
-		return DWARF_CB_ABORT;
-	return DWARF_CB_OK;
-}
+	if (line_range_add_line(fname, lineno, lf->lr) < 0)
+		return -EINVAL;
 
-static int find_line_range_func_decl_lines(struct line_finder *lf)
-{
-	struct dwarf_callback_param param = {.data = (void *)lf, .retval = 0};
-	dwarf_getfuncs(&lf->cu_die, line_range_funcdecl_cb, &param, 0);
-	return param.retval;
+	return 0;
 }
 
 /* Find line range from its line number */
 static int find_line_range_by_line(Dwarf_Die *sp_die, struct line_finder *lf)
 {
-	Dwarf_Lines *lines;
-	Dwarf_Line *line;
-	size_t nlines, i;
-	Dwarf_Addr addr;
-	int lineno, ret = 0;
-	const char *src;
-	Dwarf_Die die_mem;
-
-	line_list__init(&lf->lr->line_list);
-	if (dwarf_getsrclines(&lf->cu_die, &lines, &nlines) != 0) {
-		pr_warning("No source lines found.\n");
-		return -ENOENT;
-	}
-
-	/* Search probable lines on lines list */
-	for (i = 0; i < nlines; i++) {
-		line = dwarf_onesrcline(lines, i);
-		if (dwarf_lineno(line, &lineno) != 0 ||
-		    (lf->lno_s > lineno || lf->lno_e < lineno))
-			continue;
-
-		if (sp_die) {
-			/* Address filtering 1: does sp_die include addr? */
-			if (dwarf_lineaddr(line, &addr) != 0 ||
-			    !dwarf_haspc(sp_die, addr))
-				continue;
-
-			/* Address filtering 2: No child include addr? */
-			if (die_find_inlinefunc(sp_die, addr, &die_mem))
-				continue;
-		}
-
-		/* TODO: Get fileno from line, but how? */
-		src = dwarf_linesrc(line, NULL, NULL);
-		if (strtailcmp(src, lf->fname) != 0)
-			continue;
-
-		ret = line_range_add_line(src, lineno, lf->lr);
-		if (ret < 0)
-			return ret;
-	}
+	int ret;
 
-	/*
-	 * Dwarf lines doesn't include function declarations. We have to
-	 * check functions list or given function.
-	 */
-	if (sp_die) {
-		src = dwarf_decl_file(sp_die);
-		if (src && dwarf_decl_line(sp_die, &lineno) == 0 &&
-		    (lf->lno_s <= lineno && lf->lno_e >= lineno))
-			ret = line_range_add_line(src, lineno, lf->lr);
-	} else
-		ret = find_line_range_func_decl_lines(lf);
+	ret = die_walk_lines(sp_die ?: &lf->cu_die, line_range_walk_cb, lf);
 
 	/* Update status */
 	if (ret >= 0)
@@ -1758,9 +1814,6 @@ static int line_range_search_cb(Dwarf_Die *sp_die, void *data)
 	struct line_finder *lf = param->data;
 	struct line_range *lr = lf->lr;
 
-	pr_debug("find (%llx) %s\n",
-		 (unsigned long long)dwarf_dieoffset(sp_die),
-		 dwarf_diename(sp_die));
 	if (dwarf_tag(sp_die) == DW_TAG_subprogram &&
 	    die_compare_name(sp_die, lr->function)) {
 		lf->fname = dwarf_decl_file(sp_die);

tools/perf/util/session.c

@@ -242,17 +242,16 @@ static bool symbol__match_parent_regex(struct symbol *sym)
 	return 0;
 }
 
-struct map_symbol *perf_session__resolve_callchain(struct perf_session *self,
-						   struct thread *thread,
-						   struct ip_callchain *chain,
-						   struct symbol **parent)
+int perf_session__resolve_callchain(struct perf_session *self,
+				    struct thread *thread,
+				    struct ip_callchain *chain,
+				    struct symbol **parent)
 {
 	u8 cpumode = PERF_RECORD_MISC_USER;
 	unsigned int i;
-	struct map_symbol *syms = calloc(chain->nr, sizeof(*syms));
+	int err;
 
-	if (!syms)
-		return NULL;
+	callchain_cursor_reset(&self->callchain_cursor);
 
 	for (i = 0; i < chain->nr; i++) {
 		u64 ip = chain->ips[i];
@@ -281,12 +280,15 @@ struct map_symbol *perf_session__resolve_callchain(struct perf_session *self,
 				*parent = al.sym;
 			if (!symbol_conf.use_callchain)
 				break;
-			syms[i].map = al.map;
-			syms[i].sym = al.sym;
 		}
+
+		err = callchain_cursor_append(&self->callchain_cursor,
+					      ip, al.map, al.sym);
+		if (err)
+			return err;
 	}
 
-	return syms;
+	return 0;
 }
 
 static int process_event_synth_stub(event_t *event __used,
@@ -494,7 +496,7 @@ static void flush_sample_queue(struct perf_session *s,
 		if (iter->timestamp > limit)
 			break;
 
-		event__parse_sample(iter->event, s, &sample);
+		perf_session__parse_sample(s, iter->event, &sample);
 		perf_session_deliver_event(s, iter->event, &sample, ops,
 					   iter->file_offset);
 
@@ -804,7 +806,7 @@ static int perf_session__process_event(struct perf_session *session,
 	/*
 	 * For all kernel events we get the sample data
 	 */
-	event__parse_sample(event, session, &sample);
+	perf_session__parse_sample(session, event, &sample);
 
 	/* Preprocess sample records - precheck callchains */
 	if (perf_session__preprocess_sample(session, event, &sample))

tools/perf/util/session.h

@@ -51,7 +51,8 @@ struct perf_session {
 	int			cwdlen;
 	char			*cwd;
 	struct ordered_samples	ordered_samples;
-	char filename[0];
+	struct callchain_cursor	callchain_cursor;
+	char			filename[0];
 };
 
 struct perf_event_ops;
@@ -94,10 +95,10 @@ int __perf_session__process_events(struct perf_session *self,
 int perf_session__process_events(struct perf_session *self,
 				 struct perf_event_ops *event_ops);
 
-struct map_symbol *perf_session__resolve_callchain(struct perf_session *self,
-						   struct thread *thread,
-						   struct ip_callchain *chain,
-						   struct symbol **parent);
+int perf_session__resolve_callchain(struct perf_session *self,
+				    struct thread *thread,
+				    struct ip_callchain *chain,
+				    struct symbol **parent);
 
 bool perf_session__has_traces(struct perf_session *self, const char *msg);
 
@@ -154,4 +155,13 @@ size_t perf_session__fprintf_nr_events(struct perf_session *self, FILE *fp)
 {
 	return hists__fprintf_nr_events(&self->hists, fp);
 }
+
+static inline int perf_session__parse_sample(struct perf_session *session,
+					     const event_t *event,
+					     struct sample_data *sample)
+{
+	return event__parse_sample(event, session->sample_type,
+				   session->sample_id_all, sample);
+}
+
 #endif /* __PERF_SESSION_H */

tools/perf/util/thread.c

@@ -7,61 +7,6 @@
 #include "util.h"
 #include "debug.h"
 
-/* Skip "." and ".." directories */
-static int filter(const struct dirent *dir)
-{
-	if (dir->d_name[0] == '.')
-		return 0;
-	else
-		return 1;
-}
-
-struct thread_map *thread_map__new_by_pid(pid_t pid)
-{
-	struct thread_map *threads;
-	char name[256];
-	int items;
-	struct dirent **namelist = NULL;
-	int i;
-
-	sprintf(name, "/proc/%d/task", pid);
-	items = scandir(name, &namelist, filter, NULL);
-	if (items <= 0)
-                return NULL;
-
-	threads = malloc(sizeof(*threads) + sizeof(pid_t) * items);
-	if (threads != NULL) {
-		for (i = 0; i < items; i++)
-			threads->map[i] = atoi(namelist[i]->d_name);
-		threads->nr = items;
-	}
-
-	for (i=0; i<items; i++)
-		free(namelist[i]);
-	free(namelist);
-
-	return threads;
-}
-
-struct thread_map *thread_map__new_by_tid(pid_t tid)
-{
-	struct thread_map *threads = malloc(sizeof(*threads) + sizeof(pid_t));
-
-	if (threads != NULL) {
-		threads->map[0] = tid;
-		threads->nr	= 1;
-	}
-
-	return threads;
-}
-
-struct thread_map *thread_map__new(pid_t pid, pid_t tid)
-{
-	if (pid != -1)
-		return thread_map__new_by_pid(pid);
-	return thread_map__new_by_tid(tid);
-}
-
 static struct thread *thread__new(pid_t pid)
 {
 	struct thread *self = zalloc(sizeof(*self));

tools/perf/util/thread.h

@@ -18,24 +18,10 @@ struct thread {
 	int			comm_len;
 };
 
-struct thread_map {
-	int nr;
-	int map[];
-};
-
 struct perf_session;
 
 void thread__delete(struct thread *self);
 
-struct thread_map *thread_map__new_by_pid(pid_t pid);
-struct thread_map *thread_map__new_by_tid(pid_t tid);
-struct thread_map *thread_map__new(pid_t pid, pid_t tid);
-
-static inline void thread_map__delete(struct thread_map *threads)
-{
-	free(threads);
-}
-
 int thread__set_comm(struct thread *self, const char *comm);
 int thread__comm_len(struct thread *self);
 struct thread *perf_session__findnew(struct perf_session *self, pid_t pid);

tools/perf/util/thread_map.c

@@ -0,0 +1,64 @@
+#include <dirent.h>
+#include <stdlib.h>
+#include <stdio.h>
+#include "thread_map.h"
+
+/* Skip "." and ".." directories */
+static int filter(const struct dirent *dir)
+{
+	if (dir->d_name[0] == '.')
+		return 0;
+	else
+		return 1;
+}
+
+struct thread_map *thread_map__new_by_pid(pid_t pid)
+{
+	struct thread_map *threads;
+	char name[256];
+	int items;
+	struct dirent **namelist = NULL;
+	int i;
+
+	sprintf(name, "/proc/%d/task", pid);
+	items = scandir(name, &namelist, filter, NULL);
+	if (items <= 0)
+                return NULL;
+
+	threads = malloc(sizeof(*threads) + sizeof(pid_t) * items);
+	if (threads != NULL) {
+		for (i = 0; i < items; i++)
+			threads->map[i] = atoi(namelist[i]->d_name);
+		threads->nr = items;
+	}
+
+	for (i=0; i<items; i++)
+		free(namelist[i]);
+	free(namelist);
+
+	return threads;
+}
+
+struct thread_map *thread_map__new_by_tid(pid_t tid)
+{
+	struct thread_map *threads = malloc(sizeof(*threads) + sizeof(pid_t));
+
+	if (threads != NULL) {
+		threads->map[0] = tid;
+		threads->nr	= 1;
+	}
+
+	return threads;
+}
+
+struct thread_map *thread_map__new(pid_t pid, pid_t tid)
+{
+	if (pid != -1)
+		return thread_map__new_by_pid(pid);
+	return thread_map__new_by_tid(tid);
+}
+
+void thread_map__delete(struct thread_map *threads)
+{
+	free(threads);
+}

tools/perf/util/thread_map.h

@@ -0,0 +1,15 @@
+#ifndef __PERF_THREAD_MAP_H
+#define __PERF_THREAD_MAP_H
+
+#include <sys/types.h>
+
+struct thread_map {
+	int nr;
+	int map[];
+};
+
+struct thread_map *thread_map__new_by_pid(pid_t pid);
+struct thread_map *thread_map__new_by_tid(pid_t tid);
+struct thread_map *thread_map__new(pid_t pid, pid_t tid);
+void thread_map__delete(struct thread_map *threads);
+#endif	/* __PERF_THREAD_MAP_H */

tools/perf/util/ui/browsers/hists.c

@@ -377,7 +377,7 @@ static int hist_browser__show_callchain_node_rb_tree(struct hist_browser *self,
 	while (node) {
 		struct callchain_node *child = rb_entry(node, struct callchain_node, rb_node);
 		struct rb_node *next = rb_next(node);
-		u64 cumul = cumul_hits(child);
+		u64 cumul = callchain_cumul_hits(child);
 		struct callchain_list *chain;
 		char folded_sign = ' ';
 		int first = true;