Merge remote-tracking branch 'tip/perf/core' into kvm-updates/3.3

* tip/perf/core: (66 commits)
  perf, x86: Expose perf capability to other modules
  perf, x86: Implement arch event mask as quirk
  x86, perf: Disable non available architectural events
  jump_label: Provide jump_label_key initializers
  jump_label, x86: Fix section mismatch
  perf, core: Rate limit perf_sched_events jump_label patching
  perf: Fix enable_on_exec for sibling events
  perf: Remove superfluous arguments
  perf, x86: Prefer fixed-purpose counters when scheduling
  perf, x86: Fix event scheduler for constraints with overlapping counters
  perf, x86: Implement event scheduler helper functions
  perf: Avoid a useless pmu_disable() in the perf-tick
  x86/tools: Add decoded instruction dump mode
  x86: Update instruction decoder to support new AVX formats
  x86/tools: Fix insn_sanity message outputs
  x86/tools: Fix instruction decoder message output
  x86: Fix instruction decoder to handle grouped AVX instructions
  x86/tools: Fix Makefile to build all test tools
  perf test: Soft errors shouldn't stop the "Validate PERF_RECORD_" test
  perf test: Validate PERF_RECORD_ events and perf_sample fields
  ...

Signed-off-by: Avi Kivity <avi@redhat.com>

* commit 'b3d9468a8bd218a695e3a0ff112cd4efd27b670a': (66 commits)
  perf, x86: Expose perf capability to other modules
  perf, x86: Implement arch event mask as quirk
  x86, perf: Disable non available architectural events
  jump_label: Provide jump_label_key initializers
  jump_label, x86: Fix section mismatch
  perf, core: Rate limit perf_sched_events jump_label patching
  perf: Fix enable_on_exec for sibling events
  perf: Remove superfluous arguments
  perf, x86: Prefer fixed-purpose counters when scheduling
  perf, x86: Fix event scheduler for constraints with overlapping counters
  perf, x86: Implement event scheduler helper functions
  perf: Avoid a useless pmu_disable() in the perf-tick
  x86/tools: Add decoded instruction dump mode
  x86: Update instruction decoder to support new AVX formats
  x86/tools: Fix insn_sanity message outputs
  x86/tools: Fix instruction decoder message output
  x86: Fix instruction decoder to handle grouped AVX instructions
  x86/tools: Fix Makefile to build all test tools
  perf test: Soft errors shouldn't stop the "Validate PERF_RECORD_" test
  perf test: Validate PERF_RECORD_ events and perf_sample fields
  ...

4 files changed, 209 insertions, 33 deletions

diff --git a/drivers/oprofile/nmi_timer_int.c b/drivers/oprofile/nmi_timer_int.c
new file mode 100644
index 000000000000..76f1c9357f39
--- /dev/null
+++ b/drivers/oprofile/nmi_timer_int.c
@@ -0,0 +1,173 @@
+/**
+ * @file nmi_timer_int.c
+ *
+ * @remark Copyright 2011 Advanced Micro Devices, Inc.
+ *
+ * @author Robert Richter <robert.richter@amd.com>
+ */
+
+#include <linux/init.h>
+#include <linux/smp.h>
+#include <linux/errno.h>
+#include <linux/oprofile.h>
+#include <linux/perf_event.h>
+
+#ifdef CONFIG_OPROFILE_NMI_TIMER
+
+static DEFINE_PER_CPU(struct perf_event *, nmi_timer_events);
+static int ctr_running;
+
+static struct perf_event_attr nmi_timer_attr = {
+	.type           = PERF_TYPE_HARDWARE,
+	.config         = PERF_COUNT_HW_CPU_CYCLES,
+	.size           = sizeof(struct perf_event_attr),
+	.pinned         = 1,
+	.disabled       = 1,
+};
+
+static void nmi_timer_callback(struct perf_event *event,
+			       struct perf_sample_data *data,
+			       struct pt_regs *regs)
+{
+	event->hw.interrupts = 0;       /* don't throttle interrupts */
+	oprofile_add_sample(regs, 0);
+}
+
+static int nmi_timer_start_cpu(int cpu)
+{
+	struct perf_event *event = per_cpu(nmi_timer_events, cpu);
+
+	if (!event) {
+		event = perf_event_create_kernel_counter(&nmi_timer_attr, cpu, NULL,
+							 nmi_timer_callback, NULL);
+		if (IS_ERR(event))
+			return PTR_ERR(event);
+		per_cpu(nmi_timer_events, cpu) = event;
+	}
+
+	if (event && ctr_running)
+		perf_event_enable(event);
+
+	return 0;
+}
+
+static void nmi_timer_stop_cpu(int cpu)
+{
+	struct perf_event *event = per_cpu(nmi_timer_events, cpu);
+
+	if (event && ctr_running)
+		perf_event_disable(event);
+}
+
+static int nmi_timer_cpu_notifier(struct notifier_block *b, unsigned long action,
+				  void *data)
+{
+	int cpu = (unsigned long)data;
+	switch (action) {
+	case CPU_DOWN_FAILED:
+	case CPU_ONLINE:
+		nmi_timer_start_cpu(cpu);
+		break;
+	case CPU_DOWN_PREPARE:
+		nmi_timer_stop_cpu(cpu);
+		break;
+	}
+	return NOTIFY_DONE;
+}
+
+static struct notifier_block nmi_timer_cpu_nb = {
+	.notifier_call = nmi_timer_cpu_notifier
+};
+
+static int nmi_timer_start(void)
+{
+	int cpu;
+
+	get_online_cpus();
+	ctr_running = 1;
+	for_each_online_cpu(cpu)
+		nmi_timer_start_cpu(cpu);
+	put_online_cpus();
+
+	return 0;
+}
+
+static void nmi_timer_stop(void)
+{
+	int cpu;
+
+	get_online_cpus();
+	for_each_online_cpu(cpu)
+		nmi_timer_stop_cpu(cpu);
+	ctr_running = 0;
+	put_online_cpus();
+}
+
+static void nmi_timer_shutdown(void)
+{
+	struct perf_event *event;
+	int cpu;
+
+	get_online_cpus();
+	unregister_cpu_notifier(&nmi_timer_cpu_nb);
+	for_each_possible_cpu(cpu) {
+		event = per_cpu(nmi_timer_events, cpu);
+		if (!event)
+			continue;
+		perf_event_disable(event);
+		per_cpu(nmi_timer_events, cpu) = NULL;
+		perf_event_release_kernel(event);
+	}
+
+	put_online_cpus();
+}
+
+static int nmi_timer_setup(void)
+{
+	int cpu, err;
+	u64 period;
+
+	/* clock cycles per tick: */
+	period = (u64)cpu_khz * 1000;
+	do_div(period, HZ);
+	nmi_timer_attr.sample_period = period;
+
+	get_online_cpus();
+	err = register_cpu_notifier(&nmi_timer_cpu_nb);
+	if (err)
+		goto out;
+	/* can't attach events to offline cpus: */
+	for_each_online_cpu(cpu) {
+		err = nmi_timer_start_cpu(cpu);
+		if (err)
+			break;
+	}
+	if (err)
+		nmi_timer_shutdown();
+out:
+	put_online_cpus();
+	return err;
+}
+
+int __init op_nmi_timer_init(struct oprofile_operations *ops)
+{
+	int err = 0;
+
+	err = nmi_timer_setup();
+	if (err)
+		return err;
+	nmi_timer_shutdown();		/* only check, don't alloc */
+
+	ops->create_files	= NULL;
+	ops->setup		= nmi_timer_setup;
+	ops->shutdown		= nmi_timer_shutdown;
+	ops->start		= nmi_timer_start;
+	ops->stop		= nmi_timer_stop;
+	ops->cpu_type		= "timer";
+
+	printk(KERN_INFO "oprofile: using NMI timer interrupt.\n");
+
+	return 0;
+}
+
+#endif
diff --git a/drivers/oprofile/oprof.c b/drivers/oprofile/oprof.c
index f8c752e408a6..ed2c3ec07024 100644
--- a/drivers/oprofile/oprof.c
+++ b/drivers/oprofile/oprof.c
@@ -246,37 +246,31 @@ static int __init oprofile_init(void)
 	int err;
 
 	/* always init architecture to setup backtrace support */
+	timer_mode = 0;
 	err = oprofile_arch_init(&oprofile_ops);
+	if (!err) {
+		if (!timer && !oprofilefs_register())
+			return 0;
+		oprofile_arch_exit();
+	}
 
-	timer_mode = err || timer;	/* fall back to timer mode on errors */
-	if (timer_mode) {
-		if (!err)
-			oprofile_arch_exit();
+	/* setup timer mode: */
+	timer_mode = 1;
+	/* no nmi timer mode if oprofile.timer is set */
+	if (timer || op_nmi_timer_init(&oprofile_ops)) {
 		err = oprofile_timer_init(&oprofile_ops);
 		if (err)
 			return err;
 	}
 
-	err = oprofilefs_register();
-	if (!err)
-		return 0;
-
-	/* failed */
-	if (timer_mode)
-		oprofile_timer_exit();
-	else
-		oprofile_arch_exit();
-
-	return err;
+	return oprofilefs_register();
 }
 
 
 static void __exit oprofile_exit(void)
 {
 	oprofilefs_unregister();
-	if (timer_mode)
-		oprofile_timer_exit();
-	else
+	if (!timer_mode)
 		oprofile_arch_exit();
 }
 
diff --git a/drivers/oprofile/oprof.h b/drivers/oprofile/oprof.h
index 177b73de5e5f..769fb0fcac44 100644
--- a/drivers/oprofile/oprof.h
+++ b/drivers/oprofile/oprof.h
@@ -36,6 +36,15 @@ struct dentry;
 void oprofile_create_files(struct super_block *sb, struct dentry *root);
 int oprofile_timer_init(struct oprofile_operations *ops);
 void oprofile_timer_exit(void);
+#ifdef CONFIG_OPROFILE_NMI_TIMER
+int op_nmi_timer_init(struct oprofile_operations *ops);
+#else
+static inline int op_nmi_timer_init(struct oprofile_operations *ops)
+{
+	return -ENODEV;
+}
+#endif
+
 
 int oprofile_set_ulong(unsigned long *addr, unsigned long val);
 int oprofile_set_timeout(unsigned long time);
diff --git a/drivers/oprofile/timer_int.c b/drivers/oprofile/timer_int.c
index 878fba126582..93404f72dfa8 100644
--- a/drivers/oprofile/timer_int.c
+++ b/drivers/oprofile/timer_int.c
@@ -97,24 +97,24 @@ static struct notifier_block __refdata oprofile_cpu_notifier = {
 	.notifier_call = oprofile_cpu_notify,
 };
 
-int oprofile_timer_init(struct oprofile_operations *ops)
+static int oprofile_hrtimer_setup(void)
 {
-	int rc;
-
-	rc = register_hotcpu_notifier(&oprofile_cpu_notifier);
-	if (rc)
-		return rc;
-	ops->create_files = NULL;
-	ops->setup = NULL;
-	ops->shutdown = NULL;
-	ops->start = oprofile_hrtimer_start;
-	ops->stop = oprofile_hrtimer_stop;
-	ops->cpu_type = "timer";
-	printk(KERN_INFO "oprofile: using timer interrupt.\n");
-	return 0;
+	return register_hotcpu_notifier(&oprofile_cpu_notifier);
 }
 
-void oprofile_timer_exit(void)
+static void oprofile_hrtimer_shutdown(void)
 {
 	unregister_hotcpu_notifier(&oprofile_cpu_notifier);
 }
+
+int oprofile_timer_init(struct oprofile_operations *ops)
+{
+	ops->create_files	= NULL;
+	ops->setup		= oprofile_hrtimer_setup;
+	ops->shutdown		= oprofile_hrtimer_shutdown;
+	ops->start		= oprofile_hrtimer_start;
+	ops->stop		= oprofile_hrtimer_stop;
+	ops->cpu_type		= "timer";
+	printk(KERN_INFO "oprofile: using timer interrupt.\n");
+	return 0;
+}