Skip to contentSkip to conversation

perf/urgent fixes

GIT PULL New Arnaldo Carvalho de Melo 2 participants 17 messages Sep 15, 2015 – Sep 16, 2015 perf events

lore
Share & tools
lore

Arnaldo Carvalho de Melo

author maintainer
#0 lore raw

Hi Ingo,

Please consider pulling, there are quite a few patches, but that is
because Adrian did the right thing and made his patches very granular, thanks Adrian!

- Arnaldo

The following changes since commit 91a4dc9f71e4b8f9b8da547d19f80e7be89d7539:

Merge tag 'perf-urgent-for-mingo' of git://git.kernel.org/pub/scm/linux/kernel/git/acme/linux into perf/urgent (2015-09-14 09:29:22 +0200)

are available in the git repository at:

git://git.kernel.org/pub/scm/linux/kernel/git/acme/linux.git tags/perf-urgent-for-mingo

for you to fetch changes up to c5e6bd2ed3e81df443e4ae11e95ed71ff77bf9e5:

perf tests: Fix software clock events test setting maps (2015-09-15 11:04:49 -0300)

----------------------------------------------------------------
perf/urgent fixes:

User visible:

- Fix segfault pressing -> in 'perf top' with no hist entries (Wang Nan)

E.g:

	perf top -e page-faults --pid 11400 # 11400 generates no page-fault

- Fix propagation of thread and cpu maps, that got broken when doing incomplete changes to better support events with a PMU cpu mask, leading to Intel PT to fail with an error like:

  $ perf record -e intel_pt//u uname
  Error: The sys_perf_event_open() syscall returned with
          22 (Invalid argument) for event (sched:sched_switch).

Because intel_pt adds that sched:sched_switch evsel to the evlist after the thread/cpu maps were propagated to the evsels, fix it (Adrian Hunter)

----------------------------------------------------------------
Adrian Hunter (14):
perf evlist: Simplify propagate_maps() logic
perf evlist: Simplify set_maps() logic
perf evlist: Remove redundant validation from propagate_maps() perf evlist: Add has_user_cpus member
perf evlist: Fix splice_list_tail() not setting evlist
perf evlist: Fix missing thread_map__put in propagate_maps() perf evsel: Add own_cpus member
perf evlist: Make set_maps() more resilient
perf evlist: Make create_maps() use set_maps()
perf evlist: Factor out a function to propagate maps for a single evsel perf evlist: Fix add() not propagating maps
perf evlist: Fix create_syswide_maps() not propagating maps perf tests: Fix task exit test setting maps
perf tests: Fix software clock events test setting maps

Wang Nan (1):

      perf top: Fix segfault pressing -> with no hist entries
 tools/perf/tests/sw-clock.c    |  18 ++++--
 tools/perf/tests/task-exit.c   |  18 ++++--
 tools/perf/ui/browsers/hists.c |  12 +++-
 tools/perf/util/evlist.c       | 138 +++++++++++++++++++++++------------------
 tools/perf/util/evlist.h       |   9 ++-
 tools/perf/util/evsel.c        |   1 +
 tools/perf/util/evsel.h        |   1 +
 tools/perf/util/parse-events.c |   7 +--
 8 files changed, 124 insertions(+), 80 deletions(-)
author maintainer
#1 lore raw

PATCH 1/15 perf top: Fix segfault pressing -> with no hist entries

From: Wang Nan <[email protected]>

'perf top' segfaults with following operation:

 # perf top -e page-faults -p 11400 # 11400 never generates page-fault

Then on the resulting empty interface, press right key:

  # ./perf top -e page-faults -p 11400
  perf: Segmentation fault
  -------- backtrace --------
  ./perf[0x535428]
  /lib64/libc.so.6(+0x3545f)[0x7f0dd360745f]
  ./perf[0x531d46]
  ./perf(perf_evlist__tui_browse_hists+0x96)[0x5340d6]
  ./perf[0x44ba2f]
  /lib64/libpthread.so.0(+0x81d0)[0x7f0dd49dc1d0]
  /lib64/libc.so.6(clone+0x6c)[0x7f0dd36b90dc]

The bug resides in perf_evsel__hists_browse() that, in the above circumstance browser->selection can be NULL, but code after skip_annotation doesn't consider it.

This patch fix it by checking browser->selection before fetching browser->selection->map.

---
 tools/perf/ui/browsers/hists.c | 12 +++++++++++-
 1 file changed, 11 insertions(+), 1 deletion(-)
tools/perf/ui/browsers/hists.c +11 -1
index cf86f2d3a5e7..c04c60d4863c 100644
@@ -1968,7 +1968,8 @@ skip_annotation:
 					  &options[nr_options], dso);
 		nr_options += add_map_opt(browser, &actions[nr_options],
 					  &options[nr_options],
-					  browser->selection->map);
+					  browser->selection ?
+						browser->selection->map : NULL);
 
 		/* perf script support */
 		if (browser->he_selection) {
@@ -1976,6 +1977,15 @@ skip_annotation:
 						     &actions[nr_options],
 						     &options[nr_options],
 						     thread, NULL);
+			/*
+			 * Note that browser->selection != NULL
+			 * when browser->he_selection is not NULL,
+			 * so we don't need to check browser->selection
+			 * before fetching browser->selection->sym like what
+			 * we do before fetching browser->selection->map.
+			 *
+			 * See hist_browser__show_entry.
+			 */
 			nr_options += add_script_opt(browser,
 						     &actions[nr_options],
 						     &options[nr_options],
Signature · 1 line
2.1.0
author maintainer
#2 lore raw

PATCH 2/15 perf evlist: Simplify propagate_maps() logic

In reply to Arnaldo Carvalho de Melo

From: Adrian Hunter <[email protected]>

If evsel->cpus is to be reassigned then the current value must be "put", which works even if it is NULL. Simplify the current logic by moving the "put" next to the assignment.

---
 tools/perf/util/evlist.c | 5 ++---
 1 file changed, 2 insertions(+), 3 deletions(-)
tools/perf/util/evlist.c +2 -3
index d51a5200c8af..95e07ea3904c 100644
@@ -1113,11 +1113,10 @@ static int perf_evlist__propagate_maps(struct perf_evlist *evlist,
 		 * We already have cpus for evsel (via PMU sysfs) so
 		 * keep it, if there's no target cpu list defined.
 		 */
-		if (evsel->cpus && has_user_cpus)
+		if (!evsel->cpus || has_user_cpus) {
 			cpu_map__put(evsel->cpus);
-
-		if (!evsel->cpus || has_user_cpus)
 			evsel->cpus = cpu_map__get(evlist->cpus);
+		}
 
 		evsel->threads = thread_map__get(evlist->threads);
 
Signature · 1 line
2.1.0
author maintainer
#3 lore raw

PATCH 3/15 perf evlist: Simplify set_maps() logic

In reply to Arnaldo Carvalho de Melo

From: Adrian Hunter <[email protected]>

Don't need to check for NULL when "putting" evlist->maps and evlist->threads because the "put" functions already do that.

---
 tools/perf/util/evlist.c | 8 ++------
 1 file changed, 2 insertions(+), 6 deletions(-)
tools/perf/util/evlist.c +2 -6
index 95e07ea3904c..9cb9296cc4f8 100644
@@ -1156,14 +1156,10 @@ int perf_evlist__set_maps(struct perf_evlist *evlist,
 			  struct cpu_map *cpus,
 			  struct thread_map *threads)
 {
-	if (evlist->cpus)
-		cpu_map__put(evlist->cpus);
-
+	cpu_map__put(evlist->cpus);
 	evlist->cpus = cpus;
 
-	if (evlist->threads)
-		thread_map__put(evlist->threads);
-
+	thread_map__put(evlist->threads);
 	evlist->threads = threads;
 
 	return perf_evlist__propagate_maps(evlist, false);
Signature · 1 line
2.1.0
author maintainer
#4 lore raw

PATCH 4/15 perf evlist: Remove redundant validation from propagate_maps()

In reply to Arnaldo Carvalho de Melo

From: Adrian Hunter <[email protected]>

The validation checks that the values that were just assigned, got assigned i.e. the error can't ever happen. Subsequent patches will call this code in places where errors are not being returned. Changing those code paths to return this non-existent error is counter-productive, so just remove it.

That in turn results in perf_evlist__set_maps not needing to return an error, but callers aren't checking it either, so remove that too.

---
 tools/perf/util/evlist.c | 21 ++++++++-------------
 tools/perf/util/evlist.h |  5 ++---
 2 files changed, 10 insertions(+), 16 deletions(-)
tools/perf/util/evlist.c +8 -13
index 9cb9296cc4f8..785bfd392039 100644
@@ -1103,8 +1103,8 @@ int perf_evlist__mmap(struct perf_evlist *evlist, unsigned int pages,
 	return perf_evlist__mmap_ex(evlist, pages, overwrite, 0, false);
 }
 
-static int perf_evlist__propagate_maps(struct perf_evlist *evlist,
-				       bool has_user_cpus)
+static void perf_evlist__propagate_maps(struct perf_evlist *evlist,
+					bool has_user_cpus)
 {
 	struct perf_evsel *evsel;
 
@@ -1119,13 +1119,7 @@ static int perf_evlist__propagate_maps(struct perf_evlist *evlist,
 		}
 
 		evsel->threads = thread_map__get(evlist->threads);
-
-		if ((evlist->cpus && !evsel->cpus) ||
-		    (evlist->threads && !evsel->threads))
-			return -ENOMEM;
 	}
-
-	return 0;
 }
 
 int perf_evlist__create_maps(struct perf_evlist *evlist, struct target *target)
@@ -1144,7 +1138,9 @@ int perf_evlist__create_maps(struct perf_evlist *evlist, struct target *target)
 	if (evlist->cpus == NULL)
 		goto out_delete_threads;
 
-	return perf_evlist__propagate_maps(evlist, !!target->cpu_list);
+	perf_evlist__propagate_maps(evlist, !!target->cpu_list);
+
+	return 0;
 
 out_delete_threads:
 	thread_map__put(evlist->threads);
@@ -1152,9 +1148,8 @@ out_delete_threads:
 	return -1;
 }
 
-int perf_evlist__set_maps(struct perf_evlist *evlist,
-			  struct cpu_map *cpus,
-			  struct thread_map *threads)
+void perf_evlist__set_maps(struct perf_evlist *evlist, struct cpu_map *cpus,
+			   struct thread_map *threads)
 {
 	cpu_map__put(evlist->cpus);
 	evlist->cpus = cpus;
@@ -1162,7 +1157,7 @@ int perf_evlist__set_maps(struct perf_evlist *evlist,
 	thread_map__put(evlist->threads);
 	evlist->threads = threads;
 
-	return perf_evlist__propagate_maps(evlist, false);
+	perf_evlist__propagate_maps(evlist, false);
 }
 
 int perf_evlist__apply_filters(struct perf_evlist *evlist, struct perf_evsel **err_evsel)
tools/perf/util/evlist.h +2 -3
index b39a6198f4ac..da2fa9172bd6 100644
@@ -155,9 +155,8 @@ int perf_evlist__enable_event_idx(struct perf_evlist *evlist,
 void perf_evlist__set_selected(struct perf_evlist *evlist,
 			       struct perf_evsel *evsel);
 
-int perf_evlist__set_maps(struct perf_evlist *evlist,
-			  struct cpu_map *cpus,
-			  struct thread_map *threads);
+void perf_evlist__set_maps(struct perf_evlist *evlist, struct cpu_map *cpus,
+			   struct thread_map *threads);
 int perf_evlist__create_maps(struct perf_evlist *evlist, struct target *target);
 int perf_evlist__apply_filters(struct perf_evlist *evlist, struct perf_evsel **err_evsel);
 
Signature · 1 line
2.1.0
author maintainer
#5 lore raw

PATCH 5/15 perf evlist: Add has_user_cpus member

In reply to Arnaldo Carvalho de Melo

From: Adrian Hunter <[email protected]>

Subsequent patches will need to call perf_evlist__propagate_maps without reference to a "target". Add evlist->has_user_cpus to record whether the user has specified which cpus to target (and therefore whether that list of cpus should override the default settings for a selected event i.e. the cpu maps should be propagated)

---
 tools/perf/util/evlist.c | 11 ++++++-----
 tools/perf/util/evlist.h |  1 +
 2 files changed, 7 insertions(+), 5 deletions(-)
tools/perf/util/evlist.c +6 -5
index 785bfd392039..3a4445f26a2c 100644
@@ -1103,8 +1103,7 @@ int perf_evlist__mmap(struct perf_evlist *evlist, unsigned int pages,
 	return perf_evlist__mmap_ex(evlist, pages, overwrite, 0, false);
 }
 
-static void perf_evlist__propagate_maps(struct perf_evlist *evlist,
-					bool has_user_cpus)
+static void perf_evlist__propagate_maps(struct perf_evlist *evlist)
 {
 	struct perf_evsel *evsel;
 
@@ -1113,7 +1112,7 @@ static void perf_evlist__propagate_maps(struct perf_evlist *evlist,
 		 * We already have cpus for evsel (via PMU sysfs) so
 		 * keep it, if there's no target cpu list defined.
 		 */
-		if (!evsel->cpus || has_user_cpus) {
+		if (!evsel->cpus || evlist->has_user_cpus) {
 			cpu_map__put(evsel->cpus);
 			evsel->cpus = cpu_map__get(evlist->cpus);
 		}
@@ -1138,7 +1137,9 @@ int perf_evlist__create_maps(struct perf_evlist *evlist, struct target *target)
 	if (evlist->cpus == NULL)
 		goto out_delete_threads;
 
-	perf_evlist__propagate_maps(evlist, !!target->cpu_list);
+	evlist->has_user_cpus = !!target->cpu_list;
+
+	perf_evlist__propagate_maps(evlist);
 
 	return 0;
 
@@ -1157,7 +1158,7 @@ void perf_evlist__set_maps(struct perf_evlist *evlist, struct cpu_map *cpus,
 	thread_map__put(evlist->threads);
 	evlist->threads = threads;
 
-	perf_evlist__propagate_maps(evlist, false);
+	perf_evlist__propagate_maps(evlist);
 }
 
 int perf_evlist__apply_filters(struct perf_evlist *evlist, struct perf_evsel **err_evsel)
tools/perf/util/evlist.h +1 -0
index da2fa9172bd6..cfc4df68a9db 100644
@@ -42,6 +42,7 @@ struct perf_evlist {
 	int		 nr_mmaps;
 	bool		 overwrite;
 	bool		 enabled;
+	bool		 has_user_cpus;
 	size_t		 mmap_len;
 	int		 id_pos;
 	int		 is_pos;
Signature · 1 line
2.1.0
author maintainer
#6 lore raw

PATCH 6/15 perf evlist: Fix splice_list_tail() not setting evlist

In reply to Arnaldo Carvalho de Melo

From: Adrian Hunter <[email protected]>

Commit d49e46950772 ("perf evsel: Add a backpointer to the evlist a evsel is in") updated perf_evlist__add() but not perf_evlist__splice_list_tail().

This illustrates that it is better if perf_evlist__splice_list_tail() calls perf_evlist__add() instead of duplicating the logic, so do that. This will also simplify a subsequent fix for propagating maps.

---
 tools/perf/util/evlist.c       | 15 +++++++--------
 tools/perf/util/evlist.h       |  3 +--
 tools/perf/util/parse-events.c |  3 +--
 3 files changed, 9 insertions(+), 12 deletions(-)
tools/perf/util/evlist.c +7 -8
index 3a4445f26a2c..961560b2046b 100644
@@ -136,15 +136,14 @@ void perf_evlist__add(struct perf_evlist *evlist, struct perf_evsel *entry)
 }
 
 void perf_evlist__splice_list_tail(struct perf_evlist *evlist,
-				   struct list_head *list,
-				   int nr_entries)
+				   struct list_head *list)
 {
-	bool set_id_pos = !evlist->nr_entries;
+	struct perf_evsel *evsel, *temp;
 
-	list_splice_tail(list, &evlist->entries);
-	evlist->nr_entries += nr_entries;
-	if (set_id_pos)
-		perf_evlist__set_id_pos(evlist);
+	__evlist__for_each_safe(list, temp, evsel) {
+		list_del_init(&evsel->node);
+		perf_evlist__add(evlist, evsel);
+	}
 }
 
 void __perf_evlist__set_leader(struct list_head *list)
@@ -210,7 +209,7 @@ static int perf_evlist__add_attrs(struct perf_evlist *evlist,
 		list_add_tail(&evsel->node, &head);
 	}
 
-	perf_evlist__splice_list_tail(evlist, &head, nr_attrs);
+	perf_evlist__splice_list_tail(evlist, &head);
 
 	return 0;
 
tools/perf/util/evlist.h +1 -2
index cfc4df68a9db..115d8b53c601 100644
@@ -179,8 +179,7 @@ bool perf_evlist__valid_sample_id_all(struct perf_evlist *evlist);
 bool perf_evlist__valid_read_format(struct perf_evlist *evlist);
 
 void perf_evlist__splice_list_tail(struct perf_evlist *evlist,
-				   struct list_head *list,
-				   int nr_entries);
+				   struct list_head *list);
 
 static inline struct perf_evsel *perf_evlist__first(struct perf_evlist *evlist)
 {
tools/perf/util/parse-events.c +1 -2
index d826e6f515db..7e8ae21906e2 100644
@@ -1140,10 +1140,9 @@ int parse_events(struct perf_evlist *evlist, const char *str,
 	ret = parse_events__scanner(str, &data, PE_START_EVENTS);
 	perf_pmu__parse_cleanup();
 	if (!ret) {
-		int entries = data.idx - evlist->nr_entries;
 		struct perf_evsel *last;
 
-		perf_evlist__splice_list_tail(evlist, &data.list, entries);
+		perf_evlist__splice_list_tail(evlist, &data.list);
 		evlist->nr_groups += data.nr_groups;
 		last = perf_evlist__last(evlist);
 		last->cmdline_group_boundary = true;
Signature · 1 line
2.1.0
author maintainer
#7 lore raw

PATCH 7/15 perf evlist: Fix missing thread_map__put in propagate_maps()

In reply to Arnaldo Carvalho de Melo

From: Adrian Hunter <[email protected]>

perf_evlist__propagate_maps() incorrectly assumes evsel->threads is NULL before reassigning it, but it won't be NULL when perf_evlist__set_maps() is used to set different (or NULL) maps. Thus thread_map__put must be used, which works even if evsel->threads is NULL.

---
 tools/perf/util/evlist.c | 1 +
 1 file changed, 1 insertion(+)
tools/perf/util/evlist.c +1 -0
index 961560b2046b..79056c6ae507 100644
@@ -1116,6 +1116,7 @@ static void perf_evlist__propagate_maps(struct perf_evlist *evlist)
 			evsel->cpus = cpu_map__get(evlist->cpus);
 		}
 
+		thread_map__put(evsel->threads);
 		evsel->threads = thread_map__get(evlist->threads);
 	}
 }
Signature · 1 line
2.1.0
author maintainer
#8 lore raw

PATCH 8/15 perf evsel: Add own_cpus member

In reply to Arnaldo Carvalho de Melo

From: Adrian Hunter <[email protected]>

perf_evlist__propagate_maps() cannot easily tell if an evsel has its own cpu map. To make that simpler, keep a copy of the PMU cpu map and adjust the propagation logic accordingly.

---
 tools/perf/util/evlist.c       | 5 ++++-
 tools/perf/util/evsel.c        | 1 +
 tools/perf/util/evsel.h        | 1 +
 tools/perf/util/parse-events.c | 4 ++--
 4 files changed, 8 insertions(+), 3 deletions(-)
tools/perf/util/evlist.c +4 -1
index 79056c6ae507..5bd3b49452c6 100644
@@ -1111,9 +1111,12 @@ static void perf_evlist__propagate_maps(struct perf_evlist *evlist)
 		 * We already have cpus for evsel (via PMU sysfs) so
 		 * keep it, if there's no target cpu list defined.
 		 */
-		if (!evsel->cpus || evlist->has_user_cpus) {
+		if (!evsel->own_cpus || evlist->has_user_cpus) {
 			cpu_map__put(evsel->cpus);
 			evsel->cpus = cpu_map__get(evlist->cpus);
+		} else if (evsel->cpus != evsel->own_cpus) {
+			cpu_map__put(evsel->cpus);
+			evsel->cpus = cpu_map__get(evsel->own_cpus);
 		}
 
 		thread_map__put(evsel->threads);
tools/perf/util/evsel.c +1 -0
index c53f79123b37..5410483d5219 100644
@@ -1033,6 +1033,7 @@ void perf_evsel__exit(struct perf_evsel *evsel)
 	perf_evsel__free_config_terms(evsel);
 	close_cgroup(evsel->cgrp);
 	cpu_map__put(evsel->cpus);
+	cpu_map__put(evsel->own_cpus);
 	thread_map__put(evsel->threads);
 	zfree(&evsel->group_name);
 	zfree(&evsel->name);
tools/perf/util/evsel.h +1 -0
index 298e6bbca200..ef8925f7211a 100644
@@ -98,6 +98,7 @@ struct perf_evsel {
 	struct cgroup_sel	*cgrp;
 	void			*handler;
 	struct cpu_map		*cpus;
+	struct cpu_map		*own_cpus;
 	struct thread_map	*threads;
 	unsigned int		sample_size;
 	int			id_pos;
tools/perf/util/parse-events.c +2 -2
index 7e8ae21906e2..21ed6ee63da9 100644
@@ -287,8 +287,8 @@ __add_event(struct list_head *list, int *idx,
 	if (!evsel)
 		return NULL;
 
-	if (cpus)
-		evsel->cpus = cpu_map__get(cpus);
+	evsel->cpus     = cpu_map__get(cpus);
+	evsel->own_cpus = cpu_map__get(cpus);
 
 	if (name)
 		evsel->name = strdup(name);
Signature · 1 line
2.1.0
author maintainer
#9 lore raw

PATCH 9/15 perf evlist: Make set_maps() more resilient

In reply to Arnaldo Carvalho de Melo

From: Adrian Hunter <[email protected]>

Make perf_evlist__set_maps() more resilient by allowing for the possibility that one or another of the maps isn't being changed and therefore should not be "put".

---
 tools/perf/util/evlist.c | 19 +++++++++++++++----
 1 file changed, 15 insertions(+), 4 deletions(-)
tools/perf/util/evlist.c +15 -4
index 5bd3b49452c6..78ff52ee8788 100644
@@ -1155,11 +1155,22 @@ out_delete_threads:
 void perf_evlist__set_maps(struct perf_evlist *evlist, struct cpu_map *cpus,
 			   struct thread_map *threads)
 {
-	cpu_map__put(evlist->cpus);
-	evlist->cpus = cpus;
+	/*
+	 * Allow for the possibility that one or another of the maps isn't being
+	 * changed i.e. don't put it.  Note we are assuming the maps that are
+	 * being applied are brand new and evlist is taking ownership of the
+	 * original reference count of 1.  If that is not the case it is up to
+	 * the caller to increase the reference count.
+	 */
+	if (cpus != evlist->cpus) {
+		cpu_map__put(evlist->cpus);
+		evlist->cpus = cpus;
+	}
 
-	thread_map__put(evlist->threads);
-	evlist->threads = threads;
+	if (threads != evlist->threads) {
+		thread_map__put(evlist->threads);
+		evlist->threads = threads;
+	}
 
 	perf_evlist__propagate_maps(evlist);
 }
Signature · 1 line
2.1.0
author maintainer
#10 lore raw

PATCH 10/15 perf evlist: Make create_maps() use set_maps()

In reply to Arnaldo Carvalho de Melo

From: Adrian Hunter <[email protected]>

Since there is a function to set maps, perf_evlist__create_maps() should use it.

---
 tools/perf/util/evlist.c | 19 ++++++++++---------
 1 file changed, 10 insertions(+), 9 deletions(-)
tools/perf/util/evlist.c +10 -9
index 78ff52ee8788..c17f3558a37a 100644
@@ -1126,29 +1126,30 @@ static void perf_evlist__propagate_maps(struct perf_evlist *evlist)
 
 int perf_evlist__create_maps(struct perf_evlist *evlist, struct target *target)
 {
-	evlist->threads = thread_map__new_str(target->pid, target->tid,
-					      target->uid);
+	struct cpu_map *cpus;
+	struct thread_map *threads;
 
-	if (evlist->threads == NULL)
+	threads = thread_map__new_str(target->pid, target->tid, target->uid);
+
+	if (!threads)
 		return -1;
 
 	if (target__uses_dummy_map(target))
-		evlist->cpus = cpu_map__dummy_new();
+		cpus = cpu_map__dummy_new();
 	else
-		evlist->cpus = cpu_map__new(target->cpu_list);
+		cpus = cpu_map__new(target->cpu_list);
 
-	if (evlist->cpus == NULL)
+	if (!cpus)
 		goto out_delete_threads;
 
 	evlist->has_user_cpus = !!target->cpu_list;
 
-	perf_evlist__propagate_maps(evlist);
+	perf_evlist__set_maps(evlist, cpus, threads);
 
 	return 0;
 
 out_delete_threads:
-	thread_map__put(evlist->threads);
-	evlist->threads = NULL;
+	thread_map__put(threads);
 	return -1;
 }
 
Signature · 1 line
2.1.0
author maintainer
#11 lore raw

PATCH 11/15 perf evlist: Factor out a function to propagate maps for a single evsel

In reply to Arnaldo Carvalho de Melo

From: Adrian Hunter <[email protected]>

Subsequent fixes will need a function that just propagates maps for a single evsel so factor it out.

Cc Kan Liang <[email protected]>
[ Moved them to before perf_evlist__add() to avoid having to move it in the next patch ]
---
 tools/perf/util/evlist.c | 49 ++++++++++++++++++++++++++----------------------
 1 file changed, 27 insertions(+), 22 deletions(-)
tools/perf/util/evlist.c +27 -22
index c17f3558a37a..4e840bfc19ef 100644
@@ -124,6 +124,33 @@ void perf_evlist__delete(struct perf_evlist *evlist)
 	free(evlist);
 }
 
+static void __perf_evlist__propagate_maps(struct perf_evlist *evlist,
+					  struct perf_evsel *evsel)
+{
+	/*
+	 * We already have cpus for evsel (via PMU sysfs) so
+	 * keep it, if there's no target cpu list defined.
+	 */
+	if (!evsel->own_cpus || evlist->has_user_cpus) {
+		cpu_map__put(evsel->cpus);
+		evsel->cpus = cpu_map__get(evlist->cpus);
+	} else if (evsel->cpus != evsel->own_cpus) {
+		cpu_map__put(evsel->cpus);
+		evsel->cpus = cpu_map__get(evsel->own_cpus);
+	}
+
+	thread_map__put(evsel->threads);
+	evsel->threads = thread_map__get(evlist->threads);
+}
+
+static void perf_evlist__propagate_maps(struct perf_evlist *evlist)
+{
+	struct perf_evsel *evsel;
+
+	evlist__for_each(evlist, evsel)
+		__perf_evlist__propagate_maps(evlist, evsel);
+}
+
 void perf_evlist__add(struct perf_evlist *evlist, struct perf_evsel *entry)
 {
 	entry->evlist = evlist;
@@ -1102,28 +1129,6 @@ int perf_evlist__mmap(struct perf_evlist *evlist, unsigned int pages,
 	return perf_evlist__mmap_ex(evlist, pages, overwrite, 0, false);
 }
 
-static void perf_evlist__propagate_maps(struct perf_evlist *evlist)
-{
-	struct perf_evsel *evsel;
-
-	evlist__for_each(evlist, evsel) {
-		/*
-		 * We already have cpus for evsel (via PMU sysfs) so
-		 * keep it, if there's no target cpu list defined.
-		 */
-		if (!evsel->own_cpus || evlist->has_user_cpus) {
-			cpu_map__put(evsel->cpus);
-			evsel->cpus = cpu_map__get(evlist->cpus);
-		} else if (evsel->cpus != evsel->own_cpus) {
-			cpu_map__put(evsel->cpus);
-			evsel->cpus = cpu_map__get(evsel->own_cpus);
-		}
-
-		thread_map__put(evsel->threads);
-		evsel->threads = thread_map__get(evlist->threads);
-	}
-}
-
 int perf_evlist__create_maps(struct perf_evlist *evlist, struct target *target)
 {
 	struct cpu_map *cpus;
Signature · 1 line
2.1.0
author maintainer
#12 lore raw

PATCH 12/15 perf evlist: Fix add() not propagating maps

In reply to Arnaldo Carvalho de Melo

From: Adrian Hunter <[email protected]>

If evsels are added after maps are created, then they won't have any maps propagated to them. Fix that.

[ Moved the moving of propagate_maps() to the patch before, so that this
one does _just_ the one lile fix calling in add()]

---
 tools/perf/util/evlist.c | 2 ++
 1 file changed, 2 insertions(+)
tools/perf/util/evlist.c +2 -0
index 4e840bfc19ef..99267ab0d24a 100644
@@ -160,6 +160,8 @@ void perf_evlist__add(struct perf_evlist *evlist, struct perf_evsel *entry)
 
 	if (!evlist->nr_entries++)
 		perf_evlist__set_id_pos(evlist);
+
+	__perf_evlist__propagate_maps(evlist, entry);
 }
 
 void perf_evlist__splice_list_tail(struct perf_evlist *evlist,
Signature · 1 line
2.1.0
author maintainer
#13 lore raw

PATCH 13/15 perf evlist: Fix create_syswide_maps() not propagating maps

In reply to Arnaldo Carvalho de Melo

From: Adrian Hunter <[email protected]>

Fix it by making it call perf_evlist__set_maps() instead of setting the maps itself.

---
 tools/perf/util/evlist.c | 19 ++++++++++---------
 1 file changed, 10 insertions(+), 9 deletions(-)
tools/perf/util/evlist.c +10 -9
index 99267ab0d24a..c8fc8a258f42 100644
@@ -1400,6 +1400,8 @@ void perf_evlist__close(struct perf_evlist *evlist)
 
 static int perf_evlist__create_syswide_maps(struct perf_evlist *evlist)
 {
+	struct cpu_map	  *cpus;
+	struct thread_map *threads;
 	int err = -ENOMEM;
 
 	/*
@@ -1411,20 +1413,19 @@ static int perf_evlist__create_syswide_maps(struct perf_evlist *evlist)
 	 * error, and we may not want to do that fallback to a
 	 * default cpu identity map :-\
 	 */
-	evlist->cpus = cpu_map__new(NULL);
-	if (evlist->cpus == NULL)
+	cpus = cpu_map__new(NULL);
+	if (!cpus)
 		goto out;
 
-	evlist->threads = thread_map__new_dummy();
-	if (evlist->threads == NULL)
-		goto out_free_cpus;
+	threads = thread_map__new_dummy();
+	if (!threads)
+		goto out_put;
 
-	err = 0;
+	perf_evlist__set_maps(evlist, cpus, threads);
 out:
 	return err;
-out_free_cpus:
-	cpu_map__put(evlist->cpus);
-	evlist->cpus = NULL;
+out_put:
+	cpu_map__put(cpus);
 	goto out;
 }
 
Signature · 1 line
2.1.0
author maintainer
#14 lore raw

PATCH 14/15 perf tests: Fix task exit test setting maps

In reply to Arnaldo Carvalho de Melo

From: Adrian Hunter <[email protected]>

The test titled "Test number of exit event of a simple workload" was setting cpu/thread maps directly. Make it use the proper function perf_evlist__set_maps() especially now that it also propagates the maps.

---
 tools/perf/tests/task-exit.c | 18 ++++++++++++++----
 1 file changed, 14 insertions(+), 4 deletions(-)
tools/perf/tests/task-exit.c +14 -4
index 3a8fedef83bc..add16385f13e 100644
@@ -43,6 +43,8 @@ int test__task_exit(void)
 	};
 	const char *argv[] = { "true", NULL };
 	char sbuf[STRERR_BUFSIZE];
+	struct cpu_map *cpus;
+	struct thread_map *threads;
 
 	signal(SIGCHLD, sig_handler);
 
@@ -58,14 +60,19 @@ int test__task_exit(void)
 	 * perf_evlist__prepare_workload we'll fill in the only thread
 	 * we're monitoring, the one forked there.
 	 */
-	evlist->cpus = cpu_map__dummy_new();
-	evlist->threads = thread_map__new_by_tid(-1);
-	if (!evlist->cpus || !evlist->threads) {
+	cpus = cpu_map__dummy_new();
+	threads = thread_map__new_by_tid(-1);
+	if (!cpus || !threads) {
 		err = -ENOMEM;
 		pr_debug("Not enough memory to create thread/cpu maps\n");
-		goto out_delete_evlist;
+		goto out_free_maps;
 	}
 
+	perf_evlist__set_maps(evlist, cpus, threads);
+
+	cpus	= NULL;
+	threads = NULL;
+
 	err = perf_evlist__prepare_workload(evlist, &target, argv, false,
 					    workload_exec_failed_signal);
 	if (err < 0) {
@@ -114,6 +121,9 @@ retry:
 		err = -1;
 	}
 
+out_free_maps:
+	cpu_map__put(cpus);
+	thread_map__put(threads);
 out_delete_evlist:
 	perf_evlist__delete(evlist);
 	return err;
Signature · 1 line
2.1.0
author maintainer
#15 lore raw

PATCH 15/15 perf tests: Fix software clock events test setting maps

In reply to Arnaldo Carvalho de Melo

From: Adrian Hunter <[email protected]>

The test titled "Test software clock events have valid period values" was setting cpu/thread maps directly. Make it use the proper function perf_evlist__set_maps() especially now that it also propagates the maps.

---
 tools/perf/tests/sw-clock.c | 18 ++++++++++++++----
 1 file changed, 14 insertions(+), 4 deletions(-)
tools/perf/tests/sw-clock.c +14 -4
index 1aa21c90731b..5b83f56a3b6f 100644
@@ -34,6 +34,8 @@ static int __test__sw_clock_freq(enum perf_sw_ids clock_id)
 		.disabled = 1,
 		.freq = 1,
 	};
+	struct cpu_map *cpus;
+	struct thread_map *threads;
 
 	attr.sample_freq = 500;
 
@@ -50,14 +52,19 @@ static int __test__sw_clock_freq(enum perf_sw_ids clock_id)
 	}
 	perf_evlist__add(evlist, evsel);
 
-	evlist->cpus = cpu_map__dummy_new();
-	evlist->threads = thread_map__new_by_tid(getpid());
-	if (!evlist->cpus || !evlist->threads) {
+	cpus = cpu_map__dummy_new();
+	threads = thread_map__new_by_tid(getpid());
+	if (!cpus || !threads) {
 		err = -ENOMEM;
 		pr_debug("Not enough memory to create thread/cpu maps\n");
-		goto out_delete_evlist;
+		goto out_free_maps;
 	}
 
+	perf_evlist__set_maps(evlist, cpus, threads);
+
+	cpus	= NULL;
+	threads = NULL;
+
 	if (perf_evlist__open(evlist)) {
 		const char *knob = "/proc/sys/kernel/perf_event_max_sample_rate";
 
@@ -107,6 +114,9 @@ next_event:
 		err = -1;
 	}
 
+out_free_maps:
+	cpu_map__put(cpus);
+	thread_map__put(threads);
 out_delete_evlist:
 	perf_evlist__delete(evlist);
 	return err;
Signature · 1 line
2.1.0
#16 lore raw
In reply to Arnaldo Carvalho de Melo
Quoted 63 lines · Arnaldo Carvalho de Melo

Hi Ingo,

Please consider pulling, there are quite a few patches, but that is
because Adrian did the right thing and made his patches very granular, thanks Adrian!

- Arnaldo

The following changes since commit 91a4dc9f71e4b8f9b8da547d19f80e7be89d7539:

Merge tag 'perf-urgent-for-mingo' of git://git.kernel.org/pub/scm/linux/kernel/git/acme/linux into perf/urgent (2015-09-14 09:29:22 +0200)

are available in the git repository at:

git://git.kernel.org/pub/scm/linux/kernel/git/acme/linux.git tags/perf-urgent-for-mingo

for you to fetch changes up to c5e6bd2ed3e81df443e4ae11e95ed71ff77bf9e5:

perf tests: Fix software clock events test setting maps (2015-09-15 11:04:49 -0300)

----------------------------------------------------------------
perf/urgent fixes:

User visible:

- Fix segfault pressing -> in 'perf top' with no hist entries (Wang Nan)

   E.g:
	perf top -e page-faults --pid 11400 # 11400 generates no page-fault

- Fix propagation of thread and cpu maps, that got broken when doing incomplete changes to better support events with a PMU cpu mask, leading to Intel PT to fail with an error like:

  $ perf record -e intel_pt//u uname
  Error: The sys_perf_event_open() syscall returned with
          22 (Invalid argument) for event (sched:sched_switch).

Because intel_pt adds that sched:sched_switch evsel to the evlist after the thread/cpu maps were propagated to the evsels, fix it (Adrian Hunter)

----------------------------------------------------------------
Adrian Hunter (14):
perf evlist: Simplify propagate_maps() logic
perf evlist: Simplify set_maps() logic
perf evlist: Remove redundant validation from propagate_maps() perf evlist: Add has_user_cpus member
perf evlist: Fix splice_list_tail() not setting evlist
perf evlist: Fix missing thread_map__put in propagate_maps() perf evsel: Add own_cpus member
perf evlist: Make set_maps() more resilient
perf evlist: Make create_maps() use set_maps()
perf evlist: Factor out a function to propagate maps for a single evsel perf evlist: Fix add() not propagating maps
perf evlist: Fix create_syswide_maps() not propagating maps perf tests: Fix task exit test setting maps
perf tests: Fix software clock events test setting maps

Wang Nan (1):

      perf top: Fix segfault pressing -> with no hist entries
Quoted 9 lines
 tools/perf/tests/sw-clock.c    |  18 ++++--
 tools/perf/tests/task-exit.c   |  18 ++++--
 tools/perf/ui/browsers/hists.c |  12 +++-
 tools/perf/util/evlist.c       | 138 +++++++++++++++++++++++------------------
 tools/perf/util/evlist.h       |   9 ++-
 tools/perf/util/evsel.c        |   1 +
 tools/perf/util/evsel.h        |   1 +
 tools/perf/util/parse-events.c |   7 +--
 8 files changed, 124 insertions(+), 80 deletions(-)

Pulled, thanks a lot Arnaldo!

Ingo