linux/tools/perf/util/cpumap.c

602 lines
11 KiB
C
Raw Normal View History

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: Paul 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: Ingo Molnar <mingo@elte.hu>
2010-03-10 17:36:09 +08:00
#include "util.h"
#include <api/fs/fs.h>
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: Paul 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: Ingo Molnar <mingo@elte.hu>
2010-03-10 17:36:09 +08:00
#include "../perf.h"
#include "cpumap.h"
#include <assert.h>
#include <stdio.h>
#include <stdlib.h>
#include <linux/bitmap.h>
#include "asm/bug.h"
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: Paul 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: Ingo Molnar <mingo@elte.hu>
2010-03-10 17:36:09 +08:00
static int max_cpu_num;
static int max_node_num;
static int *cpunode_map;
static struct cpu_map *cpu_map__default_new(void)
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: Paul 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: Ingo Molnar <mingo@elte.hu>
2010-03-10 17:36:09 +08:00
{
struct cpu_map *cpus;
int nr_cpus;
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: Paul 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: Ingo Molnar <mingo@elte.hu>
2010-03-10 17:36:09 +08:00
nr_cpus = sysconf(_SC_NPROCESSORS_ONLN);
if (nr_cpus < 0)
return NULL;
cpus = malloc(sizeof(*cpus) + nr_cpus * sizeof(int));
if (cpus != NULL) {
int i;
for (i = 0; i < nr_cpus; ++i)
cpus->map[i] = i;
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: Paul 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: Ingo Molnar <mingo@elte.hu>
2010-03-10 17:36:09 +08:00
cpus->nr = nr_cpus;
atomic_set(&cpus->refcnt, 1);
}
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: Paul 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: Ingo Molnar <mingo@elte.hu>
2010-03-10 17:36:09 +08:00
return cpus;
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: Paul 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: Ingo Molnar <mingo@elte.hu>
2010-03-10 17:36:09 +08:00
}
static struct cpu_map *cpu_map__trim_new(int nr_cpus, int *tmp_cpus)
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: Paul 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: Ingo Molnar <mingo@elte.hu>
2010-03-10 17:36:09 +08:00
{
size_t payload_size = nr_cpus * sizeof(int);
struct cpu_map *cpus = malloc(sizeof(*cpus) + payload_size);
if (cpus != NULL) {
cpus->nr = nr_cpus;
memcpy(cpus->map, tmp_cpus, payload_size);
atomic_set(&cpus->refcnt, 1);
}
return cpus;
}
struct cpu_map *cpu_map__read(FILE *file)
{
struct cpu_map *cpus = NULL;
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: Paul 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: Ingo Molnar <mingo@elte.hu>
2010-03-10 17:36:09 +08:00
int nr_cpus = 0;
int *tmp_cpus = NULL, *tmp;
int max_entries = 0;
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: Paul 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: Ingo Molnar <mingo@elte.hu>
2010-03-10 17:36:09 +08:00
int n, cpu, prev;
char sep;
sep = 0;
prev = -1;
for (;;) {
n = fscanf(file, "%u%c", &cpu, &sep);
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: Paul 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: Ingo Molnar <mingo@elte.hu>
2010-03-10 17:36:09 +08:00
if (n <= 0)
break;
if (prev >= 0) {
int new_max = nr_cpus + cpu - prev - 1;
if (new_max >= max_entries) {
max_entries = new_max + MAX_NR_CPUS / 2;
tmp = realloc(tmp_cpus, max_entries * sizeof(int));
if (tmp == NULL)
goto out_free_tmp;
tmp_cpus = tmp;
}
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: Paul 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: Ingo Molnar <mingo@elte.hu>
2010-03-10 17:36:09 +08:00
while (++prev < cpu)
tmp_cpus[nr_cpus++] = prev;
}
if (nr_cpus == max_entries) {
max_entries += MAX_NR_CPUS;
tmp = realloc(tmp_cpus, max_entries * sizeof(int));
if (tmp == NULL)
goto out_free_tmp;
tmp_cpus = tmp;
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: Paul 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: Ingo Molnar <mingo@elte.hu>
2010-03-10 17:36:09 +08:00
}
tmp_cpus[nr_cpus++] = cpu;
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: Paul 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: Ingo Molnar <mingo@elte.hu>
2010-03-10 17:36:09 +08:00
if (n == 2 && sep == '-')
prev = cpu;
else
prev = -1;
if (n == 1 || sep == '\n')
break;
}
if (nr_cpus > 0)
cpus = cpu_map__trim_new(nr_cpus, tmp_cpus);
else
cpus = cpu_map__default_new();
out_free_tmp:
free(tmp_cpus);
return cpus;
}
static struct cpu_map *cpu_map__read_all_cpu_map(void)
{
struct cpu_map *cpus = NULL;
FILE *onlnf;
onlnf = fopen("/sys/devices/system/cpu/online", "r");
if (!onlnf)
return cpu_map__default_new();
cpus = cpu_map__read(onlnf);
fclose(onlnf);
return cpus;
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: Paul 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: Ingo Molnar <mingo@elte.hu>
2010-03-10 17:36:09 +08:00
}
struct cpu_map *cpu_map__new(const char *cpu_list)
{
struct cpu_map *cpus = NULL;
unsigned long start_cpu, end_cpu = 0;
char *p = NULL;
int i, nr_cpus = 0;
int *tmp_cpus = NULL, *tmp;
int max_entries = 0;
if (!cpu_list)
return cpu_map__read_all_cpu_map();
if (!isdigit(*cpu_list))
goto out;
while (isdigit(*cpu_list)) {
p = NULL;
start_cpu = strtoul(cpu_list, &p, 0);
if (start_cpu >= INT_MAX
|| (*p != '\0' && *p != ',' && *p != '-'))
goto invalid;
if (*p == '-') {
cpu_list = ++p;
p = NULL;
end_cpu = strtoul(cpu_list, &p, 0);
if (end_cpu >= INT_MAX || (*p != '\0' && *p != ','))
goto invalid;
if (end_cpu < start_cpu)
goto invalid;
} else {
end_cpu = start_cpu;
}
for (; start_cpu <= end_cpu; start_cpu++) {
/* check for duplicates */
for (i = 0; i < nr_cpus; i++)
if (tmp_cpus[i] == (int)start_cpu)
goto invalid;
if (nr_cpus == max_entries) {
max_entries += MAX_NR_CPUS;
tmp = realloc(tmp_cpus, max_entries * sizeof(int));
if (tmp == NULL)
goto invalid;
tmp_cpus = tmp;
}
tmp_cpus[nr_cpus++] = (int)start_cpu;
}
if (*p)
++p;
cpu_list = p;
}
if (nr_cpus > 0)
cpus = cpu_map__trim_new(nr_cpus, tmp_cpus);
else
cpus = cpu_map__default_new();
invalid:
free(tmp_cpus);
out:
return cpus;
}
static struct cpu_map *cpu_map__from_entries(struct cpu_map_entries *cpus)
{
struct cpu_map *map;
map = cpu_map__empty_new(cpus->nr);
if (map) {
unsigned i;
for (i = 0; i < cpus->nr; i++) {
/*
* Special treatment for -1, which is not real cpu number,
* and we need to use (int) -1 to initialize map[i],
* otherwise it would become 65535.
*/
if (cpus->cpu[i] == (u16) -1)
map->map[i] = -1;
else
map->map[i] = (int) cpus->cpu[i];
}
}
return map;
}
static struct cpu_map *cpu_map__from_mask(struct cpu_map_mask *mask)
{
struct cpu_map *map;
int nr, nbits = mask->nr * mask->long_size * BITS_PER_BYTE;
nr = bitmap_weight(mask->mask, nbits);
map = cpu_map__empty_new(nr);
if (map) {
int cpu, i = 0;
for_each_set_bit(cpu, mask->mask, nbits)
map->map[i++] = cpu;
}
return map;
}
struct cpu_map *cpu_map__new_data(struct cpu_map_data *data)
{
if (data->type == PERF_CPU_MAP__CPUS)
return cpu_map__from_entries((struct cpu_map_entries *)data->data);
else
return cpu_map__from_mask((struct cpu_map_mask *)data->data);
}
size_t cpu_map__fprintf(struct cpu_map *map, FILE *fp)
{
int i;
size_t printed = fprintf(fp, "%d cpu%s: ",
map->nr, map->nr > 1 ? "s" : "");
for (i = 0; i < map->nr; ++i)
printed += fprintf(fp, "%s%d", i ? ", " : "", map->map[i]);
return printed + fprintf(fp, "\n");
}
struct cpu_map *cpu_map__dummy_new(void)
{
struct cpu_map *cpus = malloc(sizeof(*cpus) + sizeof(int));
if (cpus != NULL) {
cpus->nr = 1;
cpus->map[0] = -1;
atomic_set(&cpus->refcnt, 1);
}
return cpus;
}
struct cpu_map *cpu_map__empty_new(int nr)
{
struct cpu_map *cpus = malloc(sizeof(*cpus) + sizeof(int) * nr);
if (cpus != NULL) {
int i;
cpus->nr = nr;
for (i = 0; i < nr; i++)
cpus->map[i] = -1;
atomic_set(&cpus->refcnt, 1);
}
return cpus;
}
static void cpu_map__delete(struct cpu_map *map)
{
if (map) {
WARN_ONCE(atomic_read(&map->refcnt) != 0,
"cpu_map refcnt unbalanced\n");
free(map);
}
}
struct cpu_map *cpu_map__get(struct cpu_map *map)
{
if (map)
atomic_inc(&map->refcnt);
return map;
}
void cpu_map__put(struct cpu_map *map)
{
if (map && atomic_dec_and_test(&map->refcnt))
cpu_map__delete(map);
}
static int cpu__get_topology_int(int cpu, const char *name, int *value)
{
char path[PATH_MAX];
snprintf(path, PATH_MAX,
"devices/system/cpu/cpu%d/topology/%s", cpu, name);
return sysfs__read_int(path, value);
}
int cpu_map__get_socket_id(int cpu)
{
int value, ret = cpu__get_topology_int(cpu, "physical_package_id", &value);
return ret ?: value;
}
int cpu_map__get_socket(struct cpu_map *map, int idx, void *data __maybe_unused)
{
int cpu;
if (idx > map->nr)
return -1;
cpu = map->map[idx];
return cpu_map__get_socket_id(cpu);
}
static int cmp_ids(const void *a, const void *b)
{
return *(int *)a - *(int *)b;
}
int cpu_map__build_map(struct cpu_map *cpus, struct cpu_map **res,
int (*f)(struct cpu_map *map, int cpu, void *data),
void *data)
{
struct cpu_map *c;
int nr = cpus->nr;
int cpu, s1, s2;
/* allocate as much as possible */
c = calloc(1, sizeof(*c) + nr * sizeof(int));
if (!c)
return -1;
for (cpu = 0; cpu < nr; cpu++) {
s1 = f(cpus, cpu, data);
for (s2 = 0; s2 < c->nr; s2++) {
if (s1 == c->map[s2])
break;
}
if (s2 == c->nr) {
c->map[c->nr] = s1;
c->nr++;
}
}
/* ensure we process id in increasing order */
qsort(c->map, c->nr, sizeof(int), cmp_ids);
atomic_set(&c->refcnt, 1);
*res = c;
return 0;
}
int cpu_map__get_core_id(int cpu)
{
int value, ret = cpu__get_topology_int(cpu, "core_id", &value);
return ret ?: value;
}
int cpu_map__get_core(struct cpu_map *map, int idx, void *data)
{
int cpu, s;
if (idx > map->nr)
return -1;
cpu = map->map[idx];
cpu = cpu_map__get_core_id(cpu);
s = cpu_map__get_socket(map, idx, data);
if (s == -1)
return -1;
/*
* encode socket in upper 16 bits
* core_id is relative to socket, and
* we need a global id. So we combine
* socket+ core id
*/
return (s << 16) | (cpu & 0xffff);
}
int cpu_map__build_socket_map(struct cpu_map *cpus, struct cpu_map **sockp)
{
return cpu_map__build_map(cpus, sockp, cpu_map__get_socket, NULL);
}
int cpu_map__build_core_map(struct cpu_map *cpus, struct cpu_map **corep)
{
return cpu_map__build_map(cpus, corep, cpu_map__get_core, NULL);
}
/* setup simple routines to easily access node numbers given a cpu number */
static int get_max_num(char *path, int *max)
{
size_t num;
char *buf;
int err = 0;
if (filename__read_str(path, &buf, &num))
return -1;
buf[num] = '\0';
/* start on the right, to find highest node num */
while (--num) {
if ((buf[num] == ',') || (buf[num] == '-')) {
num++;
break;
}
}
if (sscanf(&buf[num], "%d", max) < 1) {
err = -1;
goto out;
}
/* convert from 0-based to 1-based */
(*max)++;
out:
free(buf);
return err;
}
/* Determine highest possible cpu in the system for sparse allocation */
static void set_max_cpu_num(void)
{
const char *mnt;
char path[PATH_MAX];
int ret = -1;
/* set up default */
max_cpu_num = 4096;
mnt = sysfs__mountpoint();
if (!mnt)
goto out;
/* get the highest possible cpu number for a sparse allocation */
ret = snprintf(path, PATH_MAX, "%s/devices/system/cpu/possible", mnt);
if (ret == PATH_MAX) {
pr_err("sysfs path crossed PATH_MAX(%d) size\n", PATH_MAX);
goto out;
}
ret = get_max_num(path, &max_cpu_num);
out:
if (ret)
pr_err("Failed to read max cpus, using default of %d\n", max_cpu_num);
}
/* Determine highest possible node in the system for sparse allocation */
static void set_max_node_num(void)
{
const char *mnt;
char path[PATH_MAX];
int ret = -1;
/* set up default */
max_node_num = 8;
mnt = sysfs__mountpoint();
if (!mnt)
goto out;
/* get the highest possible cpu number for a sparse allocation */
ret = snprintf(path, PATH_MAX, "%s/devices/system/node/possible", mnt);
if (ret == PATH_MAX) {
pr_err("sysfs path crossed PATH_MAX(%d) size\n", PATH_MAX);
goto out;
}
ret = get_max_num(path, &max_node_num);
out:
if (ret)
pr_err("Failed to read max nodes, using default of %d\n", max_node_num);
}
int cpu__max_node(void)
{
if (unlikely(!max_node_num))
set_max_node_num();
return max_node_num;
}
int cpu__max_cpu(void)
{
if (unlikely(!max_cpu_num))
set_max_cpu_num();
return max_cpu_num;
}
int cpu__get_node(int cpu)
{
if (unlikely(cpunode_map == NULL)) {
pr_debug("cpu_map not initialized\n");
return -1;
}
return cpunode_map[cpu];
}
static int init_cpunode_map(void)
{
int i;
set_max_cpu_num();
set_max_node_num();
cpunode_map = calloc(max_cpu_num, sizeof(int));
if (!cpunode_map) {
pr_err("%s: calloc failed\n", __func__);
return -1;
}
for (i = 0; i < max_cpu_num; i++)
cpunode_map[i] = -1;
return 0;
}
int cpu__setup_cpunode_map(void)
{
struct dirent *dent1, *dent2;
DIR *dir1, *dir2;
unsigned int cpu, mem;
char buf[PATH_MAX];
char path[PATH_MAX];
const char *mnt;
int n;
/* initialize globals */
if (init_cpunode_map())
return -1;
mnt = sysfs__mountpoint();
if (!mnt)
return 0;
n = snprintf(path, PATH_MAX, "%s/devices/system/node", mnt);
if (n == PATH_MAX) {
pr_err("sysfs path crossed PATH_MAX(%d) size\n", PATH_MAX);
return -1;
}
dir1 = opendir(path);
if (!dir1)
return 0;
/* walk tree and setup map */
while ((dent1 = readdir(dir1)) != NULL) {
if (dent1->d_type != DT_DIR || sscanf(dent1->d_name, "node%u", &mem) < 1)
continue;
n = snprintf(buf, PATH_MAX, "%s/%s", path, dent1->d_name);
if (n == PATH_MAX) {
pr_err("sysfs path crossed PATH_MAX(%d) size\n", PATH_MAX);
continue;
}
dir2 = opendir(buf);
if (!dir2)
continue;
while ((dent2 = readdir(dir2)) != NULL) {
if (dent2->d_type != DT_LNK || sscanf(dent2->d_name, "cpu%u", &cpu) < 1)
continue;
cpunode_map[cpu] = mem;
}
closedir(dir2);
}
closedir(dir1);
return 0;
}
bool cpu_map__has(struct cpu_map *cpus, int cpu)
{
int i;
for (i = 0; i < cpus->nr; ++i) {
if (cpus->map[i] == cpu)
return true;
}
return false;
}