Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]

Groups > linux.kernel > #1584438 > unrolled thread

[PATCH v4 0/3] perf: topology on systems with offline/absent CPUs

Started byJan Stancek <jstancek@redhat.com>
First post2017-02-20 10:40 +0100
Last post2017-02-20 10:40 +0100
Articles 3 — 1 participant

Back to article view | Back to linux.kernel

Contents

  [PATCH v4 0/3] perf: topology on systems with offline/absent CPUs Jan Stancek <jstancek@redhat.com> - 2017-02-20 10:40 +0100
    [PATCH v4 2/3] perf header: Make build_cpu_topology skip offline/absent CPUs Jan Stancek <jstancek@redhat.com> - 2017-02-20 10:40 +0100
    [PATCH v4 1/3] perf cpu_map: Add cpu__max_present_cpu() Jan Stancek <jstancek@redhat.com> - 2017-02-20 10:40 +0100

#1584438 — [PATCH v4 0/3] perf: topology on systems with offline/absent CPUs

FromJan Stancek <jstancek@redhat.com>
Date2017-02-20 10:40 +0100
Subject[PATCH v4 0/3] perf: topology on systems with offline/absent CPUs
Message-ID<tcSBP-2Sv-9@gated-at.bofh.it>
Changes in v4:
- rebase on top of 4.10
- fix style of commit logs
- add comment to process_cpu_topology()

This series aims to address issues on systems with offline/absent CPUs:
1. cpu_topology_map size
   Allocation based on _SC_NPROCESSORS_CONF is too small,
   there can be CPU IDs that are higher, which causes invalid reads. 
2. build_cpu_topology fails for offline/absent CPUs
   It is trying to parse sysfs entries, that do not exist.
3. perf topology test fails

Changes affect HEADER_NRCPUS and HEADER_CPU_TOPOLOGY:

For example on a system like this:
  _SC_NPROCESSORS_CONF == 16
  available: 2 nodes (0-1)
  node 0 cpus: 0 6 8 10 16 22 24 26
  node 0 size: 12004 MB
  node 0 free: 9470 MB
  node 1 cpus: 1 7 9 11 23 25 27
  node 1 size: 12093 MB
  node 1 free: 9406 MB
  node distances:
  node   0   1
    0:  10  20
    1:  20  10

HEADER_NRCPUS before this series:
nr_cpus_online = 15
nr_cpus_available = 16

HEADER_CPU_TOPOLOGY before this series:
core_sib_nr: 2
core_siblings: 0,6,8,10,16,22,24,26
core_siblings: 1,7,9,11,23,25,27
thread_sib_nr: 8
thread_siblings: 0,16
thread_siblings: 1
thread_siblings: 6,22
thread_siblings: 7,23
thread_siblings: 8,24
thread_siblings: 9,25
thread_siblings: 10,26
thread_siblings: 11,27
core_id: 0, socket_id: 0
core_id: 0, socket_id: 1
core_id: -1, socket_id: -1
core_id: -1, socket_id: -1
core_id: -1, socket_id: -1
core_id: -1, socket_id: -1
core_id: 10, socket_id: 0
core_id: 10, socket_id: 1
core_id: 1, socket_id: 0
core_id: 1, socket_id: 1
core_id: 9, socket_id: 0
core_id: 9, socket_id: 1
core_id: -1, socket_id: -1
core_id: -1, socket_id: -1
core_id: -1, socket_id: -1
core_id: -1, socket_id: -1

HEADER_NRCPUS with this series applied:
nr_cpus_online = 15
nr_cpus_available = 28

HEADER_CPU_TOPOLOGY with this series applied:
core_sib_nr: 2
core_siblings: 0,6,8,10,16,22,24,26
core_siblings: 1,7,9,11,23,25,27
thread_sib_nr: 8
thread_siblings: 0,16
thread_siblings: 1
thread_siblings: 6,22
thread_siblings: 7,23
thread_siblings: 8,24
thread_siblings: 9,25
thread_siblings: 10,26
thread_siblings: 11,27
core_id: 0, socket_id: 0
core_id: 0, socket_id: 1
core_id: -1, socket_id: -1
core_id: -1, socket_id: -1
core_id: -1, socket_id: -1
core_id: -1, socket_id: -1
core_id: 10, socket_id: 0
core_id: 10, socket_id: 1
core_id: 1, socket_id: 0
core_id: 1, socket_id: 1
core_id: 9, socket_id: 0
core_id: 9, socket_id: 1
core_id: -1, socket_id: -1
core_id: -1, socket_id: -1
core_id: -1, socket_id: -1
core_id: -1, socket_id: -1
core_id: 0, socket_id: 0
core_id: -1, socket_id: -1
core_id: -1, socket_id: -1
core_id: -1, socket_id: -1
core_id: -1, socket_id: -1
core_id: -1, socket_id: -1
core_id: 10, socket_id: 0
core_id: 10, socket_id: 1
core_id: 1, socket_id: 0
core_id: 1, socket_id: 1
core_id: 9, socket_id: 0
core_id: 9, socket_id: 1

Jan Stancek (3):
  perf cpu_map: Add cpu__max_present_cpu()
  perf header: Make build_cpu_topology skip offline/absent CPUs
  perf: Replace _SC_NPROCESSORS_CONF with max_present_cpu in
    cpu_topology_map

 tools/perf/builtin-stat.c   |  2 +-
 tools/perf/tests/topology.c |  4 +++-
 tools/perf/util/cpumap.c    | 22 ++++++++++++++++++++++
 tools/perf/util/cpumap.h    |  1 +
 tools/perf/util/env.c       |  2 +-
 tools/perf/util/header.c    | 37 ++++++++++++++++++++++++-------------
 6 files changed, 52 insertions(+), 16 deletions(-)

-- 
1.8.3.1

[toc] | [next] | [standalone]

#1584440 — [PATCH v4 2/3] perf header: Make build_cpu_topology skip offline/absent CPUs

FromJan Stancek <jstancek@redhat.com>
Date2017-02-20 10:40 +0100
Subject[PATCH v4 2/3] perf header: Make build_cpu_topology skip offline/absent CPUs
Message-ID<tcSBP-2Sv-17@gated-at.bofh.it>
In reply to#1584438
When build_cpu_topo() encounters offline/absent CPUs,
it fails to find any sysfs entries and returns failure.
This leads to build_cpu_topology() and write_cpu_topology()
failing as well.

Because HEADER_CPU_TOPOLOGY has not been written, read leaves
cpu_topology_map NULL and we get NULL ptr deref at:
 ...
  cmd_test
   __cmd_test
    test_and_print
     run_test
      test_session_topology
       check_cpu_topology

  # perf test 36
  36: Session topology                           :
  --- start ---
  test child forked, pid 14902
  templ file: /tmp/perf-test-4CKocW
  failed to write feature HEADER_CPU_TOPOLOGY
  perf: Segmentation fault
  Obtained 9 stack frames.
  ./perf(sighandler_dump_stack+0x41) [0x5095f1]
  /lib64/libc.so.6(+0x35250) [0x7f4b7c3c9250]
  ./perf(test_session_topology+0x1db) [0x490ceb]
  ./perf() [0x475b68]
  ./perf(cmd_test+0x5b9) [0x4763c9]
  ./perf() [0x4945a3]
  ./perf(main+0x69f) [0x427e8f]
  /lib64/libc.so.6(__libc_start_main+0xf5) [0x7f4b7c3b5b35]
  ./perf() [0x427fb9]
  test child interrupted
  ---- end ----
  Session topology: FAILED!

This patch makes build_cpu_topology() skip offline/absent CPUs,
by checking their presence against cpu_map built from online CPUs.

Signed-off-by: Jan Stancek <jstancek@redhat.com>
---
 tools/perf/util/header.c | 19 ++++++++++++++++---
 1 file changed, 16 insertions(+), 3 deletions(-)

diff --git a/tools/perf/util/header.c b/tools/perf/util/header.c
index d89c9c7ef4e5..ca0f12fb5fd3 100644
--- a/tools/perf/util/header.c
+++ b/tools/perf/util/header.c
@@ -503,24 +503,31 @@ static void free_cpu_topo(struct cpu_topo *tp)
 
 static struct cpu_topo *build_cpu_topology(void)
 {
-	struct cpu_topo *tp;
+	struct cpu_topo *tp = NULL;
 	void *addr;
 	u32 nr, i;
 	size_t sz;
 	long ncpus;
 	int ret = -1;
+	struct cpu_map *map;
 
 	ncpus = sysconf(_SC_NPROCESSORS_CONF);
 	if (ncpus < 0)
 		return NULL;
 
+	/* build online CPU map */
+	map = cpu_map__new(NULL);
+	if (map == NULL) {
+		pr_debug("failed to get system cpumap\n");
+		return NULL;
+	}
+
 	nr = (u32)(ncpus & UINT_MAX);
 
 	sz = nr * sizeof(char *);
-
 	addr = calloc(1, sizeof(*tp) + 2 * sz);
 	if (!addr)
-		return NULL;
+		goto out_free;
 
 	tp = addr;
 	tp->cpu_nr = nr;
@@ -530,10 +537,16 @@ static struct cpu_topo *build_cpu_topology(void)
 	tp->thread_siblings = addr;
 
 	for (i = 0; i < nr; i++) {
+		if (!cpu_map__has(map, i))
+			continue;
+
 		ret = build_cpu_topo(tp, i);
 		if (ret < 0)
 			break;
 	}
+
+out_free:
+	cpu_map__put(map);
 	if (ret) {
 		free_cpu_topo(tp);
 		tp = NULL;
-- 
1.8.3.1

[toc] | [prev] | [next] | [standalone]

#1584441 — [PATCH v4 1/3] perf cpu_map: Add cpu__max_present_cpu()

FromJan Stancek <jstancek@redhat.com>
Date2017-02-20 10:40 +0100
Subject[PATCH v4 1/3] perf cpu_map: Add cpu__max_present_cpu()
Message-ID<tcSBQ-2Sv-33@gated-at.bofh.it>
In reply to#1584438
Similar to cpu__max_cpu() (which returns max possible CPU),
returns max present CPU.

Signed-off-by: Jan Stancek <jstancek@redhat.com>
---
 tools/perf/util/cpumap.c | 22 ++++++++++++++++++++++
 tools/perf/util/cpumap.h |  1 +
 2 files changed, 23 insertions(+)

diff --git a/tools/perf/util/cpumap.c b/tools/perf/util/cpumap.c
index 2c0b52264a46..8c7504939113 100644
--- a/tools/perf/util/cpumap.c
+++ b/tools/perf/util/cpumap.c
@@ -9,6 +9,7 @@
 #include "asm/bug.h"
 
 static int max_cpu_num;
+static int max_present_cpu_num;
 static int max_node_num;
 static int *cpunode_map;
 
@@ -442,6 +443,7 @@ static void set_max_cpu_num(void)
 
 	/* set up default */
 	max_cpu_num = 4096;
+	max_present_cpu_num = 4096;
 
 	mnt = sysfs__mountpoint();
 	if (!mnt)
@@ -455,6 +457,17 @@ static void set_max_cpu_num(void)
 	}
 
 	ret = get_max_num(path, &max_cpu_num);
+	if (ret)
+		goto out;
+
+	/* get the highest present cpu number for a sparse allocation */
+	ret = snprintf(path, PATH_MAX, "%s/devices/system/cpu/present", 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_present_cpu_num);
 
 out:
 	if (ret)
@@ -505,6 +518,15 @@ int cpu__max_cpu(void)
 	return max_cpu_num;
 }
 
+int cpu__max_present_cpu(void)
+{
+	if (unlikely(!max_present_cpu_num))
+		set_max_cpu_num();
+
+	return max_present_cpu_num;
+}
+
+
 int cpu__get_node(int cpu)
 {
 	if (unlikely(cpunode_map == NULL)) {
diff --git a/tools/perf/util/cpumap.h b/tools/perf/util/cpumap.h
index 06bd689f5989..1a0549af8f5c 100644
--- a/tools/perf/util/cpumap.h
+++ b/tools/perf/util/cpumap.h
@@ -62,6 +62,7 @@ static inline bool cpu_map__empty(const struct cpu_map *map)
 
 int cpu__max_node(void);
 int cpu__max_cpu(void);
+int cpu__max_present_cpu(void);
 int cpu__get_node(int cpu);
 
 int cpu_map__build_map(struct cpu_map *cpus, struct cpu_map **res,
-- 
1.8.3.1

[toc] | [prev] | [standalone]

Back to top | Article view | linux.kernel

csiph-web