Skip to contentSkip to conversation

a perf and some sched patches

PATCH Applied Peter Zijlstra 2 participants 14 messages Dec 17, 2009

Subsystems (2)
lore
Share & tools
lore
#0 lore raw

cover a perf and some sched patches

Fixes an issue with the per-task-per-cpu counters, reported by acme. Fixes an false positive in the set_task_cpu() debug check, reported by Ingo. Fixes up the task->state to char bits, gumbled at by tglx.

#1 lore raw

PATCH 1/6 sched: Move TASK_STATE_TO_CHAR_STR near the TASK_state bits

So that we don't keep forgetting about it

LKML-Reference: <new-submission>

---
 include/linux/sched.h |    4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)
include/linux/sched.h +2 -2
Index: linux-2.6/include/linux/sched.h
===================================================================
@@ -193,6 +193,8 @@ print_cfs_rq(struct seq_file *m, int cpu
 #define TASK_WAKEKILL		128
 #define TASK_WAKING		256
 
+#define TASK_STATE_TO_CHAR_STR "RSDTtZX"
+
 /* Convenience macros for the sake of set_task_state */
 #define TASK_KILLABLE		(TASK_WAKEKILL | TASK_UNINTERRUPTIBLE)
 #define TASK_STOPPED		(TASK_WAKEKILL | __TASK_STOPPED)
@@ -2595,8 +2597,6 @@ static inline void mm_init_owner(struct 
 }
 #endif /* CONFIG_MM_OWNER */
 
-#define TASK_STATE_TO_CHAR_STR "RSDTtZX"
-
 #endif /* __KERNEL__ */
 
 #endif
#2 lore raw

[tip:sched/urgent] sched: Move TASK_STATE_TO_CHAR_STR near the TASK_state bits

Commit-ID:  733421516b42c44b9e21f1793c430cc801ef8324
Gitweb:     http://git.kernel.org/tip/733421516b42c44b9e21f1793c430cc801ef8324
Author:     Peter Zijlstra <[email protected]>
AuthorDate: Thu, 17 Dec 2009 13:16:27 +0100
Committer:  Ingo Molnar <[email protected]>
CommitDate: Thu, 17 Dec 2009 13:22:43 +0100

sched: Move TASK_STATE_TO_CHAR_STR near the TASK_state bits

So that we don't keep forgetting about it.

LKML-Reference: <[email protected]>

---
 include/linux/sched.h |    4 ++--
 1 files changed, 2 insertions(+), 2 deletions(-)
include/linux/sched.h +2 -2
index 973b2b8..c28ed1b 100644
@@ -193,6 +193,8 @@ print_cfs_rq(struct seq_file *m, int cpu, struct cfs_rq *cfs_rq)
 #define TASK_WAKEKILL		128
 #define TASK_WAKING		256
 
+#define TASK_STATE_TO_CHAR_STR "RSDTtZX"
+
 /* Convenience macros for the sake of set_task_state */
 #define TASK_KILLABLE		(TASK_WAKEKILL | TASK_UNINTERRUPTIBLE)
 #define TASK_STOPPED		(TASK_WAKEKILL | __TASK_STOPPED)
@@ -2595,8 +2597,6 @@ static inline void mm_init_owner(struct mm_struct *mm, struct task_struct *p)
 }
 #endif /* CONFIG_MM_OWNER */
 
-#define TASK_STATE_TO_CHAR_STR "RSDTtZX"
-
 #endif /* __KERNEL__ */
 
 #endif
#3 lore raw

PATCH 2/6 sched: Add missing state chars to TASK_STATE_TO_CHAR_STR

In reply to Peter Zijlstra

We grew 3 new task states since the last time someone touched it.

LKML-Reference: <new-submission>

---
 include/linux/sched.h |    2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
include/linux/sched.h +1 -1
Index: linux-2.6/include/linux/sched.h
===================================================================
@@ -193,7 +193,7 @@ print_cfs_rq(struct seq_file *m, int cpu
 #define TASK_WAKEKILL		128
 #define TASK_WAKING		256
 
-#define TASK_STATE_TO_CHAR_STR "RSDTtZX"
+#define TASK_STATE_TO_CHAR_STR "RSDTtZXxKW"
 
 /* Convenience macros for the sake of set_task_state */
 #define TASK_KILLABLE		(TASK_WAKEKILL | TASK_UNINTERRUPTIBLE)
#4 lore raw

[tip:sched/urgent] sched: Add missing state chars to TASK_STATE_TO_CHAR_STR

Commit-ID:  44d90df6b757c59651ddd55f1a84f28132b50d29
Gitweb:     http://git.kernel.org/tip/44d90df6b757c59651ddd55f1a84f28132b50d29
Author:     Peter Zijlstra <[email protected]>
AuthorDate: Thu, 17 Dec 2009 13:16:28 +0100
Committer:  Ingo Molnar <[email protected]>
CommitDate: Thu, 17 Dec 2009 13:22:44 +0100

sched: Add missing state chars to TASK_STATE_TO_CHAR_STR

We grew 3 new task states since the last time someone touched it.

LKML-Reference: <[email protected]>

---
 include/linux/sched.h |    2 +-
 1 files changed, 1 insertions(+), 1 deletions(-)
include/linux/sched.h +1 -1
index c28ed1b..94858df 100644
@@ -193,7 +193,7 @@ print_cfs_rq(struct seq_file *m, int cpu, struct cfs_rq *cfs_rq)
 #define TASK_WAKEKILL		128
 #define TASK_WAKING		256
 
-#define TASK_STATE_TO_CHAR_STR "RSDTtZX"
+#define TASK_STATE_TO_CHAR_STR "RSDTtZXxKW"
 
 /* Convenience macros for the sake of set_task_state */
 #define TASK_KILLABLE		(TASK_WAKEKILL | TASK_UNINTERRUPTIBLE)
#5 lore raw

PATCH 3/6 sched: Update task_state_arraypwith new states

In reply to Peter Zijlstra

Neglected because its hidden... (who reads comments anyway)

LKML-Reference: <new-submission>

---
 fs/proc/array.c |    7 +++++--
 1 file changed, 5 insertions(+), 2 deletions(-)
fs/proc/array.c +5 -2
Index: linux-2.6/fs/proc/array.c
===================================================================
@@ -138,9 +138,12 @@ static const char *task_state_array[] = 
 	"S (sleeping)",		/*  1 */
 	"D (disk sleep)",	/*  2 */
 	"T (stopped)",		/*  4 */
-	"T (tracing stop)",	/*  8 */
+	"t (tracing stop)",	/*  8 */
 	"Z (zombie)",		/* 16 */
-	"X (dead)"		/* 32 */
+	"X (dead)",		/* 32 */
+	"x (dead)",		/* 64 */
+	"K (wakekill)",		/* 128 */
+	"W (waking)",		/* 256 */
 };
 
 static inline const char *get_task_state(struct task_struct *tsk)
#6 lore raw

[tip:sched/urgent] sched: Update task_state_arraypwith new states

Commit-ID:  464763cf1c6df632dccc8f2f4c7e50163154a2c0
Gitweb:     http://git.kernel.org/tip/464763cf1c6df632dccc8f2f4c7e50163154a2c0
Author:     Peter Zijlstra <[email protected]>
AuthorDate: Thu, 17 Dec 2009 13:16:29 +0100
Committer:  Ingo Molnar <[email protected]>
CommitDate: Thu, 17 Dec 2009 13:22:45 +0100

sched: Update task_state_arraypwith new states

Neglected because its hidden... (who reads comments anyway)

LKML-Reference: <[email protected]>

---
 fs/proc/array.c |    7 +++++--
 1 files changed, 5 insertions(+), 2 deletions(-)
fs/proc/array.c +5 -2
index 4badde1..96361e8 100644
@@ -138,9 +138,12 @@ static const char *task_state_array[] = {
 	"S (sleeping)",		/*  1 */
 	"D (disk sleep)",	/*  2 */
 	"T (stopped)",		/*  4 */
-	"T (tracing stop)",	/*  8 */
+	"t (tracing stop)",	/*  8 */
 	"Z (zombie)",		/* 16 */
-	"X (dead)"		/* 32 */
+	"X (dead)",		/* 32 */
+	"x (dead)",		/* 64 */
+	"K (wakekill)",		/* 128 */
+	"W (waking)",		/* 256 */
 };
 
 static inline const char *get_task_state(struct task_struct *tsk)
#7 lore raw

PATCH 4/6 sched: Assert task state bits at build time

In reply to Peter Zijlstra

Since everybody is lazy and prone to forgetting things, make the compiler help us a bit.

LKML-Reference: <new-submission>

---
 fs/proc/array.c       |   18 ++++++++++--------
 include/linux/sched.h |    4 ++++
 2 files changed, 14 insertions(+), 8 deletions(-)
fs/proc/array.c +10 -8
Index: linux-2.6/fs/proc/array.c
===================================================================
@@ -134,14 +134,14 @@ static inline void task_name(struct seq_
  * simple bit tests.
  */
 static const char *task_state_array[] = {
-	"R (running)",		/*  0 */
-	"S (sleeping)",		/*  1 */
-	"D (disk sleep)",	/*  2 */
-	"T (stopped)",		/*  4 */
-	"t (tracing stop)",	/*  8 */
-	"Z (zombie)",		/* 16 */
-	"X (dead)",		/* 32 */
-	"x (dead)",		/* 64 */
+	"R (running)",		/*   0 */
+	"S (sleeping)",		/*   1 */
+	"D (disk sleep)",	/*   2 */
+	"T (stopped)",		/*   4 */
+	"t (tracing stop)",	/*   8 */
+	"Z (zombie)",		/*  16 */
+	"X (dead)",		/*  32 */
+	"x (dead)",		/*  64 */
 	"K (wakekill)",		/* 128 */
 	"W (waking)",		/* 256 */
 };
@@ -151,6 +151,8 @@ static inline const char *get_task_state
 	unsigned int state = (tsk->state & TASK_REPORT) | tsk->exit_state;
 	const char **p = &task_state_array[0];
 
+	BUILD_BUG_ON(1 + ilog2(TASK_STATE_MAX) != ARRAY_SIZE(task_state_array));
+
 	while (state) {
 		p++;
 		state >>= 1;
include/linux/sched.h +4 -0
Index: linux-2.6/include/linux/sched.h
===================================================================
@@ -192,9 +192,13 @@ print_cfs_rq(struct seq_file *m, int cpu
 #define TASK_DEAD		64
 #define TASK_WAKEKILL		128
 #define TASK_WAKING		256
+#define TASK_STATE_MAX		512
 
 #define TASK_STATE_TO_CHAR_STR "RSDTtZXxKW"
 
+extern char ___assert_task_state[1 - 2*!!(
+		sizeof(TASK_STATE_TO_CHAR_STR)-1 != ilog2(TASK_STATE_MAX)+1)];
+
 /* Convenience macros for the sake of set_task_state */
 #define TASK_KILLABLE		(TASK_WAKEKILL | TASK_UNINTERRUPTIBLE)
 #define TASK_STOPPED		(TASK_WAKEKILL | __TASK_STOPPED)
#8 lore raw

[tip:sched/urgent] sched: Assert task state bits at build time

Commit-ID:  e1781538cf5c870ab696e9b8f0a5c498d3900f2f
Gitweb:     http://git.kernel.org/tip/e1781538cf5c870ab696e9b8f0a5c498d3900f2f
Author:     Peter Zijlstra <[email protected]>
AuthorDate: Thu, 17 Dec 2009 13:16:30 +0100
Committer:  Ingo Molnar <[email protected]>
CommitDate: Thu, 17 Dec 2009 13:22:45 +0100

sched: Assert task state bits at build time

Since everybody is lazy and prone to forgetting things, make the compiler help us a bit.

LKML-Reference: <[email protected]>

---
 fs/proc/array.c       |   18 ++++++++++--------
 include/linux/sched.h |    4 ++++
 2 files changed, 14 insertions(+), 8 deletions(-)
fs/proc/array.c +10 -8
index 96361e8..f560325 100644
@@ -134,14 +134,14 @@ static inline void task_name(struct seq_file *m, struct task_struct *p)
  * simple bit tests.
  */
 static const char *task_state_array[] = {
-	"R (running)",		/*  0 */
-	"S (sleeping)",		/*  1 */
-	"D (disk sleep)",	/*  2 */
-	"T (stopped)",		/*  4 */
-	"t (tracing stop)",	/*  8 */
-	"Z (zombie)",		/* 16 */
-	"X (dead)",		/* 32 */
-	"x (dead)",		/* 64 */
+	"R (running)",		/*   0 */
+	"S (sleeping)",		/*   1 */
+	"D (disk sleep)",	/*   2 */
+	"T (stopped)",		/*   4 */
+	"t (tracing stop)",	/*   8 */
+	"Z (zombie)",		/*  16 */
+	"X (dead)",		/*  32 */
+	"x (dead)",		/*  64 */
 	"K (wakekill)",		/* 128 */
 	"W (waking)",		/* 256 */
 };
@@ -151,6 +151,8 @@ static inline const char *get_task_state(struct task_struct *tsk)
 	unsigned int state = (tsk->state & TASK_REPORT) | tsk->exit_state;
 	const char **p = &task_state_array[0];
 
+	BUILD_BUG_ON(1 + ilog2(TASK_STATE_MAX) != ARRAY_SIZE(task_state_array));
+
 	while (state) {
 		p++;
 		state >>= 1;
include/linux/sched.h +4 -0
index 94858df..3754387 100644
@@ -192,9 +192,13 @@ print_cfs_rq(struct seq_file *m, int cpu, struct cfs_rq *cfs_rq)
 #define TASK_DEAD		64
 #define TASK_WAKEKILL		128
 #define TASK_WAKING		256
+#define TASK_STATE_MAX		512
 
 #define TASK_STATE_TO_CHAR_STR "RSDTtZXxKW"
 
+extern char ___assert_task_state[1 - 2*!!(
+		sizeof(TASK_STATE_TO_CHAR_STR)-1 != ilog2(TASK_STATE_MAX)+1)];
+
 /* Convenience macros for the sake of set_task_state */
 #define TASK_KILLABLE		(TASK_WAKEKILL | TASK_UNINTERRUPTIBLE)
 #define TASK_STOPPED		(TASK_WAKEKILL | __TASK_STOPPED)
#9 lore raw

PATCH 5/6 sched: Fix broken assertion

In reply to Peter Zijlstra

There's a preemption race in the set_task_cpu() debug check in that when we get preempted after setting task->state we'd still be on the rq proper, but fail the test.

Check for preempted tasks, since those are always on the RQ.

LKML-Reference: <new-submission>

---
 kernel/sched.c |    3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)
kernel/sched.c +2 -1
Index: linux-2.6/kernel/sched.c
===================================================================
@@ -2041,7 +2041,8 @@ void set_task_cpu(struct task_struct *p,
 	 * We should never call set_task_cpu() on a blocked task,
 	 * ttwu() will sort out the placement.
 	 */
-	WARN_ON_ONCE(p->state != TASK_RUNNING && p->state != TASK_WAKING);
+	WARN_ON_ONCE(p->state != TASK_RUNNING && p->state != TASK_WAKING &&
+			!(task_thread_info(p)->preempt_count & PREEMPT_ACTIVE));
 #endif
 
 	trace_sched_migrate_task(p, new_cpu);
#10 lore raw

[tip:sched/urgent] sched: Fix broken assertion

Commit-ID:  077614ee1e93245a3b9a4e1213659405dbeb0ba6
Gitweb:     http://git.kernel.org/tip/077614ee1e93245a3b9a4e1213659405dbeb0ba6
Author:     Peter Zijlstra <[email protected]>
AuthorDate: Thu, 17 Dec 2009 13:16:31 +0100
Committer:  Ingo Molnar <[email protected]>
CommitDate: Thu, 17 Dec 2009 13:22:46 +0100

sched: Fix broken assertion

There's a preemption race in the set_task_cpu() debug check in that when we get preempted after setting task->state we'd still be on the rq proper, but fail the test.

Check for preempted tasks, since those are always on the RQ.

LKML-Reference: <[email protected]>

---
 kernel/sched.c |    3 ++-
 1 files changed, 2 insertions(+), 1 deletions(-)
kernel/sched.c +2 -1
index 7be88a7..720df10 100644
@@ -2041,7 +2041,8 @@ void set_task_cpu(struct task_struct *p, unsigned int new_cpu)
 	 * We should never call set_task_cpu() on a blocked task,
 	 * ttwu() will sort out the placement.
 	 */
-	WARN_ON_ONCE(p->state != TASK_RUNNING && p->state != TASK_WAKING);
+	WARN_ON_ONCE(p->state != TASK_RUNNING && p->state != TASK_WAKING &&
+			!(task_thread_info(p)->preempt_count & PREEMPT_ACTIVE));
 #endif
 
 	trace_sched_migrate_task(p, new_cpu);
#11 lore raw

PATCH 6/6 perf: Dont report side-band events on each cpu for per-task-per-cpu events

In reply to Peter Zijlstra

Acme noticed that his FORK/MMAP numbers were inflated by about the same factor as his cpu-count.

This led to the discovery of a few more sites that need to respect the event->cpu filter.

---
 kernel/perf_event.c |   32 ++++++++++++++++++--------------
 1 file changed, 18 insertions(+), 14 deletions(-)
kernel/perf_event.c +18 -14
Index: linux-2.6/kernel/perf_event.c
===================================================================
@@ -1381,6 +1381,9 @@ static void perf_ctx_adjust_freq(struct 
 		if (event->state != PERF_EVENT_STATE_ACTIVE)
 			continue;
 
+		if (event->cpu != -1 && event->cpu != smp_processor_id())
+			continue;
+
 		hwc = &event->hw;
 
 		interrupts = hwc->interrupts;
@@ -3265,6 +3268,9 @@ static void perf_event_task_output(struc
 
 static int perf_event_task_match(struct perf_event *event)
 {
+	if (event->cpu != -1 && event->cpu != smp_processor_id())
+		return 0;
+
 	if (event->attr.comm || event->attr.mmap || event->attr.task)
 		return 1;
 
@@ -3290,12 +3296,11 @@ static void perf_event_task_event(struct
 	rcu_read_lock();
 	cpuctx = &get_cpu_var(perf_cpu_context);
 	perf_event_task_ctx(&cpuctx->ctx, task_event);
-	put_cpu_var(perf_cpu_context);
-
 	if (!ctx)
 		ctx = rcu_dereference(task_event->task->perf_event_ctxp);
 	if (ctx)
 		perf_event_task_ctx(ctx, task_event);
+	put_cpu_var(perf_cpu_context);
 	rcu_read_unlock();
 }
 
@@ -3372,6 +3377,9 @@ static void perf_event_comm_output(struc
 
 static int perf_event_comm_match(struct perf_event *event)
 {
+	if (event->cpu != -1 && event->cpu != smp_processor_id())
+		return 0;
+
 	if (event->attr.comm)
 		return 1;
 
@@ -3408,15 +3416,10 @@ static void perf_event_comm_event(struct
 	rcu_read_lock();
 	cpuctx = &get_cpu_var(perf_cpu_context);
 	perf_event_comm_ctx(&cpuctx->ctx, comm_event);
-	put_cpu_var(perf_cpu_context);
-
-	/*
-	 * doesn't really matter which of the child contexts the
-	 * events ends up in.
-	 */
 	ctx = rcu_dereference(current->perf_event_ctxp);
 	if (ctx)
 		perf_event_comm_ctx(ctx, comm_event);
+	put_cpu_var(perf_cpu_context);
 	rcu_read_unlock();
 }
 
@@ -3491,6 +3494,9 @@ static void perf_event_mmap_output(struc
 static int perf_event_mmap_match(struct perf_event *event,
 				   struct perf_mmap_event *mmap_event)
 {
+	if (event->cpu != -1 && event->cpu != smp_processor_id())
+		return 0;
+
 	if (event->attr.mmap)
 		return 1;
 
@@ -3564,15 +3570,10 @@ got_name:
 	rcu_read_lock();
 	cpuctx = &get_cpu_var(perf_cpu_context);
 	perf_event_mmap_ctx(&cpuctx->ctx, mmap_event);
-	put_cpu_var(perf_cpu_context);
-
-	/*
-	 * doesn't really matter which of the child contexts the
-	 * events ends up in.
-	 */
 	ctx = rcu_dereference(current->perf_event_ctxp);
 	if (ctx)
 		perf_event_mmap_ctx(ctx, mmap_event);
+	put_cpu_var(perf_cpu_context);
 	rcu_read_unlock();
 
 	kfree(buf);
@@ -3863,6 +3864,9 @@ static int perf_swevent_match(struct per
 				struct perf_sample_data *data,
 				struct pt_regs *regs)
 {
+	if (event->cpu != -1 && event->cpu != smp_processor_id())
+		return 0;
+
 	if (!perf_swevent_is_counting(event))
 		return 0;
 
#12 lore raw

[tip:perf/urgent] perf events: Dont report side-band events on each cpu for per-task-per-cpu events

Commit-ID:  5d27c23df09b702868d9a3bff86ec6abd22963ac
Gitweb:     http://git.kernel.org/tip/5d27c23df09b702868d9a3bff86ec6abd22963ac
Author:     Peter Zijlstra <[email protected]>
AuthorDate: Thu, 17 Dec 2009 13:16:32 +0100
Committer:  Ingo Molnar <[email protected]>
CommitDate: Thu, 17 Dec 2009 13:21:36 +0100

perf events: Dont report side-band events on each cpu for per-task-per-cpu events

Acme noticed that his FORK/MMAP numbers were inflated by about the same factor as his cpu-count.

This led to the discovery of a few more sites that need to respect the event->cpu filter.

LKML-Reference: <[email protected]>

---
 kernel/perf_event.c |   32 ++++++++++++++++++--------------
 1 files changed, 18 insertions(+), 14 deletions(-)
kernel/perf_event.c +18 -14
index 8ab8698..03cc061 100644
@@ -1381,6 +1381,9 @@ static void perf_ctx_adjust_freq(struct perf_event_context *ctx)
 		if (event->state != PERF_EVENT_STATE_ACTIVE)
 			continue;
 
+		if (event->cpu != -1 && event->cpu != smp_processor_id())
+			continue;
+
 		hwc = &event->hw;
 
 		interrupts = hwc->interrupts;
@@ -3265,6 +3268,9 @@ static void perf_event_task_output(struct perf_event *event,
 
 static int perf_event_task_match(struct perf_event *event)
 {
+	if (event->cpu != -1 && event->cpu != smp_processor_id())
+		return 0;
+
 	if (event->attr.comm || event->attr.mmap || event->attr.task)
 		return 1;
 
@@ -3290,12 +3296,11 @@ static void perf_event_task_event(struct perf_task_event *task_event)
 	rcu_read_lock();
 	cpuctx = &get_cpu_var(perf_cpu_context);
 	perf_event_task_ctx(&cpuctx->ctx, task_event);
-	put_cpu_var(perf_cpu_context);
-
 	if (!ctx)
 		ctx = rcu_dereference(task_event->task->perf_event_ctxp);
 	if (ctx)
 		perf_event_task_ctx(ctx, task_event);
+	put_cpu_var(perf_cpu_context);
 	rcu_read_unlock();
 }
 
@@ -3372,6 +3377,9 @@ static void perf_event_comm_output(struct perf_event *event,
 
 static int perf_event_comm_match(struct perf_event *event)
 {
+	if (event->cpu != -1 && event->cpu != smp_processor_id())
+		return 0;
+
 	if (event->attr.comm)
 		return 1;
 
@@ -3408,15 +3416,10 @@ static void perf_event_comm_event(struct perf_comm_event *comm_event)
 	rcu_read_lock();
 	cpuctx = &get_cpu_var(perf_cpu_context);
 	perf_event_comm_ctx(&cpuctx->ctx, comm_event);
-	put_cpu_var(perf_cpu_context);
-
-	/*
-	 * doesn't really matter which of the child contexts the
-	 * events ends up in.
-	 */
 	ctx = rcu_dereference(current->perf_event_ctxp);
 	if (ctx)
 		perf_event_comm_ctx(ctx, comm_event);
+	put_cpu_var(perf_cpu_context);
 	rcu_read_unlock();
 }
 
@@ -3491,6 +3494,9 @@ static void perf_event_mmap_output(struct perf_event *event,
 static int perf_event_mmap_match(struct perf_event *event,
 				   struct perf_mmap_event *mmap_event)
 {
+	if (event->cpu != -1 && event->cpu != smp_processor_id())
+		return 0;
+
 	if (event->attr.mmap)
 		return 1;
 
@@ -3564,15 +3570,10 @@ got_name:
 	rcu_read_lock();
 	cpuctx = &get_cpu_var(perf_cpu_context);
 	perf_event_mmap_ctx(&cpuctx->ctx, mmap_event);
-	put_cpu_var(perf_cpu_context);
-
-	/*
-	 * doesn't really matter which of the child contexts the
-	 * events ends up in.
-	 */
 	ctx = rcu_dereference(current->perf_event_ctxp);
 	if (ctx)
 		perf_event_mmap_ctx(ctx, mmap_event);
+	put_cpu_var(perf_cpu_context);
 	rcu_read_unlock();
 
 	kfree(buf);
@@ -3863,6 +3864,9 @@ static int perf_swevent_match(struct perf_event *event,
 				struct perf_sample_data *data,
 				struct pt_regs *regs)
 {
+	if (event->cpu != -1 && event->cpu != smp_processor_id())
+		return 0;
+
 	if (!perf_swevent_is_counting(event))
 		return 0;
 
Applied
#13 lore raw
In reply to Peter Zijlstra
Quoted 3 lines · Peter Zijlstra

Fixes an issue with the per-task-per-cpu counters, reported by acme. Fixes an false positive in the set_task_cpu() debug check, reported by Ingo. Fixes up the task->state to char bits, gumbled at by tglx.

Applied, thanks Peter!

Ingo