11 files changed, 252 insertions, 94 deletions

diff --git a/Documentation/kernel-parameters.txt b/Documentation/kernel-parameters.txt
index 854ed5ca7e3f..f7735a125f4b 100644
--- a/Documentation/kernel-parameters.txt
+++ b/Documentation/kernel-parameters.txt
@@ -1848,6 +1848,9 @@ bytes respectively. Such letter suffixes can also be entirely omitted.
                         arch_perfmon: [X86] Force use of architectural
                                 perfmon on Intel CPUs instead of the
                                 CPU specific event set.
+                        timer: [X86] Force use of architectural NMI
+                                timer mode (see also oprofile.timer
+                                for generic hr timer mode)
 
         oops=panic      Always panic on oopses. Default is to just kill the
                         process, but there is a small probability of
diff --git a/arch/Kconfig b/arch/Kconfig
index 4b0669cbb3b0..2505740b81d2 100644
--- a/arch/Kconfig
+++ b/arch/Kconfig
@@ -30,6 +30,10 @@ config OPROFILE_EVENT_MULTIPLEX
 config HAVE_OPROFILE
         bool
 
+config OPROFILE_NMI_TIMER
+        def_bool y
+        depends on PERF_EVENTS && HAVE_PERF_EVENTS_NMI
+
 config KPROBES
         bool "Kprobes"
         depends on MODULES
diff --git a/arch/x86/oprofile/Makefile b/arch/x86/oprofile/Makefile
index 446902b2a6b6..1599f568f0e2 100644
--- a/arch/x86/oprofile/Makefile
+++ b/arch/x86/oprofile/Makefile
@@ -4,9 +4,8 @@ DRIVER_OBJS = $(addprefix ../../../drivers/oprofile/, \
                 oprof.o cpu_buffer.o buffer_sync.o \
                 event_buffer.o oprofile_files.o \
                 oprofilefs.o oprofile_stats.o  \
-                timer_int.o )
+                timer_int.o nmi_timer_int.o )
 
 oprofile-y                              := $(DRIVER_OBJS) init.o backtrace.o
 oprofile-$(CONFIG_X86_LOCAL_APIC)       += nmi_int.o op_model_amd.o \
                                            op_model_ppro.o op_model_p4.o
-oprofile-$(CONFIG_X86_IO_APIC)          += nmi_timer_int.o
diff --git a/arch/x86/oprofile/init.c b/arch/x86/oprofile/init.c
index cdfe4c54deca..9e138d00ad36 100644
--- a/arch/x86/oprofile/init.c
+++ b/arch/x86/oprofile/init.c
@@ -16,34 +16,23 @@
  * with the NMI mode driver.
  */
 
+#ifdef CONFIG_X86_LOCAL_APIC
 extern int op_nmi_init(struct oprofile_operations *ops);
-extern int op_nmi_timer_init(struct oprofile_operations *ops);
 extern void op_nmi_exit(void);
-extern void x86_backtrace(struct pt_regs * const regs, unsigned int depth);
+#else
+static int op_nmi_init(struct oprofile_operations *ops) { return -ENODEV; }
+static void op_nmi_exit(void) { }
+#endif
 
+extern void x86_backtrace(struct pt_regs * const regs, unsigned int depth);
 
 int __init oprofile_arch_init(struct oprofile_operations *ops)
 {
-        int ret;
-
-        ret = -ENODEV;
-
-#ifdef CONFIG_X86_LOCAL_APIC
-        ret = op_nmi_init(ops);
-#endif
-#ifdef CONFIG_X86_IO_APIC
-        if (ret < 0)
-                ret = op_nmi_timer_init(ops);
-#endif
         ops->backtrace = x86_backtrace;
-
-        return ret;
+        return op_nmi_init(ops);
 }
 
-
 void oprofile_arch_exit(void)
 {
-#ifdef CONFIG_X86_LOCAL_APIC
         op_nmi_exit();
-#endif
 }
diff --git a/arch/x86/oprofile/nmi_int.c b/arch/x86/oprofile/nmi_int.c
index adf8fb31aa7b..990c35bfa88f 100644
--- a/arch/x86/oprofile/nmi_int.c
+++ b/arch/x86/oprofile/nmi_int.c
@@ -597,24 +597,36 @@ static int __init p4_init(char **cpu_type)
         return 0;
 }
 
-static int force_arch_perfmon;
-static int force_cpu_type(const char *str, struct kernel_param *kp)
+enum __force_cpu_type {
+        reserved = 0,           /* do not force */
+        timer,
+        arch_perfmon,
+};
+
+static int force_cpu_type;
+
+static int set_cpu_type(const char *str, struct kernel_param *kp)
 {
-        if (!strcmp(str, "arch_perfmon")) {
-                force_arch_perfmon = 1;
+        if (!strcmp(str, "timer")) {
+                force_cpu_type = timer;
+                printk(KERN_INFO "oprofile: forcing NMI timer mode\n");
+        } else if (!strcmp(str, "arch_perfmon")) {
+                force_cpu_type = arch_perfmon;
                 printk(KERN_INFO "oprofile: forcing architectural perfmon\n");
+        } else {
+                force_cpu_type = 0;
         }
 
         return 0;
 }
-module_param_call(cpu_type, force_cpu_type, NULL, NULL, 0);
+module_param_call(cpu_type, set_cpu_type, NULL, NULL, 0);
 
 static int __init ppro_init(char **cpu_type)
 {
         __u8 cpu_model = boot_cpu_data.x86_model;
         struct op_x86_model_spec *spec = &op_ppro_spec; /* default */
 
-        if (force_arch_perfmon && cpu_has_arch_perfmon)
+        if (force_cpu_type == arch_perfmon && cpu_has_arch_perfmon)
                 return 0;
 
         /*
@@ -681,6 +693,9 @@ int __init op_nmi_init(struct oprofile_operations *ops)
         if (!cpu_has_apic)
                 return -ENODEV;
 
+        if (force_cpu_type == timer)
+                return -ENODEV;
+
         switch (vendor) {
         case X86_VENDOR_AMD:
                 /* Needs to be at least an Athlon (or hammer in 32bit mode) */
diff --git a/arch/x86/oprofile/nmi_timer_int.c b/arch/x86/oprofile/nmi_timer_int.c
deleted file mode 100644
index 7f8052cd6620..000000000000
--- a/arch/x86/oprofile/nmi_timer_int.c
+++ /dev/null
@@ -1,50 +0,0 @@
-/**
- * @file nmi_timer_int.c
- *
- * @remark Copyright 2003 OProfile authors
- * @remark Read the file COPYING
- *
- * @author Zwane Mwaikambo <zwane@linuxpower.ca>
- */
-
-#include <linux/init.h>
-#include <linux/smp.h>
-#include <linux/errno.h>
-#include <linux/oprofile.h>
-#include <linux/rcupdate.h>
-#include <linux/kdebug.h>
-
-#include <asm/nmi.h>
-#include <asm/apic.h>
-#include <asm/ptrace.h>
-
-static int profile_timer_exceptions_notify(unsigned int val, struct pt_regs *regs)
-{
-        oprofile_add_sample(regs, 0);
-        return NMI_HANDLED;
-}
-
-static int timer_start(void)
-{
-        if (register_nmi_handler(NMI_LOCAL, profile_timer_exceptions_notify,
-                                        0, "oprofile-timer"))
-                return 1;
-        return 0;
-}
-
-
-static void timer_stop(void)
-{
-        unregister_nmi_handler(NMI_LOCAL, "oprofile-timer");
-        synchronize_sched();  /* Allow already-started NMIs to complete. */
-}
-
-
-int __init op_nmi_timer_init(struct oprofile_operations *ops)
-{
-        ops->start = timer_start;
-        ops->stop = timer_stop;
-        ops->cpu_type = "timer";
-        printk(KERN_INFO "oprofile: using NMI timer interrupt.\n");
-        return 0;
-}
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 dccd8636095c..ed2c3ec07024 100644
--- a/drivers/oprofile/oprof.c
+++ b/drivers/oprofile/oprof.c
@@ -239,26 +239,39 @@ int oprofile_set_ulong(unsigned long *addr, unsigned long val)
         return err;
 }
 
+static int timer_mode;
+
 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 < 0 || timer) {
-                printk(KERN_INFO "oprofile: using timer interrupt.\n");
+        if (!err) {
+                if (!timer && !oprofilefs_register())
+                        return 0;
+                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;
         }
+
         return oprofilefs_register();
 }
 
 
 static void __exit oprofile_exit(void)
 {
-        oprofile_timer_exit();
         oprofilefs_unregister();
-        oprofile_arch_exit();
+        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 3ef44624f510..93404f72dfa8 100644
--- a/drivers/oprofile/timer_int.c
+++ b/drivers/oprofile/timer_int.c
@@ -97,23 +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";
-        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;
+}
diff --git a/kernel/events/core.c b/kernel/events/core.c
index d1a1bee35228..d2e28bdd523a 100644
--- a/kernel/events/core.c
+++ b/kernel/events/core.c
@@ -1322,6 +1322,7 @@ retry:
         }
         raw_spin_unlock_irq(&ctx->lock);
 }
+EXPORT_SYMBOL_GPL(perf_event_disable);
 
 static void perf_set_shadow_time(struct perf_event *event,
                                  struct perf_event_context *ctx,
@@ -1806,6 +1807,7 @@ retry:
 out:
         raw_spin_unlock_irq(&ctx->lock);
 }
+EXPORT_SYMBOL_GPL(perf_event_enable);
 
 int perf_event_refresh(struct perf_event *event, int refresh)
 {