perf tools: Fix sparse CPU numbering related bugs

At present, the perf subcommands that do system-wide monitoring
(perf stat, perf record and perf top) don't work properly unless
the online cpus are numbered 0, 1, ..., N-1.  These tools ask
for the number of online cpus with sysconf(_SC_NPROCESSORS_ONLN)
and then try to create events for cpus 0, 1, ..., N-1.

This creates problems for systems where the online cpus are
numbered sparsely.  For example, a POWER6 system in
single-threaded mode (i.e. only running 1 hardware thread per
core) will have only even-numbered cpus online.

This fixes the problem by reading the /sys/devices/system/cpu/online
file to find out which cpus are online.  The code that does that is in
tools/perf/util/cpumap.[ch], and consists of a read_cpu_map()
function that sets up a cpumap[] array and returns the number of
online cpus.  If /sys/devices/system/cpu/online can't be read or
can't be parsed successfully, it falls back to using sysconf to
ask how many cpus are online and sets up an identity map in cpumap[].

The perf record, perf stat and perf top code then calls
read_cpu_map() in the system-wide monitoring case (instead of
sysconf) and uses cpumap[] to get the cpu numbers to pass to
perf_event_open.
Signed-off-by: NPaul Mackerras <paulus@samba.org>
Cc: Anton Blanchard <anton@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Arnaldo Carvalho de Melo <acme@infradead.org>
LKML-Reference: <20100310093609.GA3959@brick.ozlabs.ibm.com>
Signed-off-by: NIngo Molnar <mingo@elte.hu>

tools/perf/Makefile

@@ -387,6 +387,7 @@ LIB_H += util/thread.h
 LIB_H += util/trace-event.h
 LIB_H += util/probe-finder.h
 LIB_H += util/probe-event.h
+LIB_H += util/cpumap.h
 
 LIB_OBJS += util/abspath.o
 LIB_OBJS += util/alias.o
@@ -433,6 +434,7 @@ LIB_OBJS += util/sort.o
 LIB_OBJS += util/hist.o
 LIB_OBJS += util/probe-event.o
 LIB_OBJS += util/util.o
+LIB_OBJS += util/cpumap.o
 
 BUILTIN_OBJS += builtin-annotate.o
 

tools/perf/builtin-record.c

@@ -22,6 +22,7 @@
 #include "util/debug.h"
 #include "util/session.h"
 #include "util/symbol.h"
+#include "util/cpumap.h"
 
 #include <unistd.h>
 #include <sched.h>
@@ -421,9 +422,6 @@ static int __cmd_record(int argc, const char **argv)
 	char buf;
 
 	page_size = sysconf(_SC_PAGE_SIZE);
-	nr_cpus = sysconf(_SC_NPROCESSORS_ONLN);
-	assert(nr_cpus <= MAX_NR_CPUS);
-	assert(nr_cpus >= 0);
 
 	atexit(sig_atexit);
 	signal(SIGCHLD, sig_handler);
@@ -547,8 +545,9 @@ static int __cmd_record(int argc, const char **argv)
 	if ((!system_wide && !inherit) || profile_cpu != -1) {
 		open_counters(profile_cpu, target_pid);
 	} else {
+		nr_cpus = read_cpu_map();
 		for (i = 0; i < nr_cpus; i++)
-			open_counters(i, target_pid);
+			open_counters(cpumap[i], target_pid);
 	}
 
 	if (file_new) {

tools/perf/builtin-stat.c

@@ -45,6 +45,7 @@
 #include "util/event.h"
 #include "util/debug.h"
 #include "util/header.h"
+#include "util/cpumap.h"
 
 #include <sys/prctl.h>
 #include <math.h>
@@ -151,7 +152,7 @@ static void create_perf_stat_counter(int counter, int pid)
 		unsigned int cpu;
 
 		for (cpu = 0; cpu < nr_cpus; cpu++) {
-			fd[cpu][counter] = sys_perf_event_open(attr, -1, cpu, -1, 0);
+			fd[cpu][counter] = sys_perf_event_open(attr, -1, cpumap[cpu], -1, 0);
 			if (fd[cpu][counter] < 0 && verbose)
 				fprintf(stderr, ERR_PERF_OPEN, counter,
 					fd[cpu][counter], strerror(errno));
@@ -519,9 +520,10 @@ int cmd_stat(int argc, const char **argv, const char *prefix __used)
 		nr_counters = ARRAY_SIZE(default_attrs);
 	}
 
-	nr_cpus = sysconf(_SC_NPROCESSORS_ONLN);
-	assert(nr_cpus <= MAX_NR_CPUS);
-	assert((int)nr_cpus >= 0);
+	if (system_wide)
+		nr_cpus = read_cpu_map();
+	else
+		nr_cpus = 1;
 
 	/*
 	 * We dont want to block the signals - that would cause

tools/perf/builtin-top.c

@@ -28,6 +28,7 @@
 #include <linux/rbtree.h>
 #include "util/parse-options.h"
 #include "util/parse-events.h"
+#include "util/cpumap.h"
 
 #include "util/debug.h"
 
@@ -1123,7 +1124,7 @@ static void start_counter(int i, int counter)
 
 	cpu = profile_cpu;
 	if (target_pid == -1 && profile_cpu == -1)
-		cpu = i;
+		cpu = cpumap[i];
 
 	attr = attrs + counter;
 
@@ -1347,12 +1348,10 @@ int cmd_top(int argc, const char **argv, const char *prefix __used)
 		attrs[counter].sample_period = default_interval;
 	}
 
-	nr_cpus = sysconf(_SC_NPROCESSORS_ONLN);
-	assert(nr_cpus <= MAX_NR_CPUS);
-	assert(nr_cpus >= 0);
-
 	if (target_pid != -1 || profile_cpu != -1)
 		nr_cpus = 1;
+	else
+		nr_cpus = read_cpu_map();
 
 	get_term_dimensions(&winsize);
 	if (print_entries == 0) {

tools/perf/util/cpumap.c

+#include "util.h"
+#include "../perf.h"
+#include "cpumap.h"
+#include <assert.h>
+#include <stdio.h>
+
+int cpumap[MAX_NR_CPUS];
+
+static int default_cpu_map(void)
+{
+	int nr_cpus, i;
+
+	nr_cpus = sysconf(_SC_NPROCESSORS_ONLN);
+	assert(nr_cpus <= MAX_NR_CPUS);
+	assert((int)nr_cpus >= 0);
+
+	for (i = 0; i < nr_cpus; ++i)
+		cpumap[i] = i;
+
+	return nr_cpus;
+}
+
+int read_cpu_map(void)
+{
+	FILE *onlnf;
+	int nr_cpus = 0;
+	int n, cpu, prev;
+	char sep;
+
+	onlnf = fopen("/sys/devices/system/cpu/online", "r");
+	if (!onlnf)
+		return default_cpu_map();
+
+	sep = 0;
+	prev = -1;
+	for (;;) {
+		n = fscanf(onlnf, "%u%c", &cpu, &sep);
+		if (n <= 0)
+			break;
+		if (prev >= 0) {
+			assert(nr_cpus + cpu - prev - 1 < MAX_NR_CPUS);
+			while (++prev < cpu)
+				cpumap[nr_cpus++] = prev;
+		}
+		assert (nr_cpus < MAX_NR_CPUS);
+		cpumap[nr_cpus++] = cpu;
+		if (n == 2 && sep == '-')
+			prev = cpu;
+		else
+			prev = -1;
+		if (n == 1 || sep == '\n')
+			break;
+	}
+	fclose(onlnf);
+	if (nr_cpus > 0)
+		return nr_cpus;
+
+	return default_cpu_map();
+}

tools/perf/util/cpumap.h

+#ifndef __PERF_CPUMAP_H
+#define __PERF_CPUMAP_H
+
+extern int read_cpu_map(void);
+extern int cpumap[];
+
+#endif /* __PERF_CPUMAP_H */