#include <linux/mm.h>
#include <linux/cpu.h>
#include <linux/smp.h>
+#include <linux/idr.h>
#include <linux/file.h>
#include <linux/poll.h>
#include <linux/slab.h>
#include <linux/dcache.h>
#include <linux/percpu.h>
#include <linux/ptrace.h>
+#include <linux/reboot.h>
#include <linux/vmstat.h>
+#include <linux/device.h>
#include <linux/vmalloc.h>
#include <linux/hardirq.h>
#include <linux/rculist.h>
#include <asm/irq_regs.h>
+enum event_type_t {
+ EVENT_FLEXIBLE = 0x1,
+ EVENT_PINNED = 0x2,
+ EVENT_ALL = EVENT_FLEXIBLE | EVENT_PINNED,
+};
+
atomic_t perf_task_events __read_mostly;
static atomic_t nr_mmap_events __read_mostly;
static atomic_t nr_comm_events __read_mostly;
static atomic64_t perf_event_id;
+static void cpu_ctx_sched_out(struct perf_cpu_context *cpuctx,
+ enum event_type_t event_type);
+
+static void cpu_ctx_sched_in(struct perf_cpu_context *cpuctx,
+ enum event_type_t event_type);
+
void __weak perf_event_print_debug(void) { }
extern __weak const char *perf_pmu_name(void)
return "pmu";
}
+static inline u64 perf_clock(void)
+{
+ return local_clock();
+}
+
void perf_pmu_disable(struct pmu *pmu)
{
int *count = this_cpu_ptr(pmu->pmu_disable_count);
put_ctx(ctx);
}
-static inline u64 perf_clock(void)
-{
- return local_clock();
-}
-
/*
* Update the record of the current time in a context.
*/
ctx->timestamp = now;
}
+static u64 perf_event_time(struct perf_event *event)
+{
+ struct perf_event_context *ctx = event->ctx;
+ return ctx ? ctx->time : 0;
+}
+
/*
* Update the total_time_enabled and total_time_running fields for a event.
*/
return;
if (ctx->is_active)
- run_end = ctx->time;
+ run_end = perf_event_time(event);
else
run_end = event->tstamp_stopped;
if (event->state == PERF_EVENT_STATE_INACTIVE)
run_end = event->tstamp_stopped;
else
- run_end = ctx->time;
+ run_end = perf_event_time(event);
event->total_time_running = run_end - event->tstamp_running;
}
struct perf_cpu_context *cpuctx,
struct perf_event_context *ctx)
{
+ u64 tstamp = perf_event_time(event);
u64 delta;
/*
* An event which could not be activated because of
&& !event_filter_match(event)) {
delta = ctx->time - event->tstamp_stopped;
event->tstamp_running += delta;
- event->tstamp_stopped = ctx->time;
+ event->tstamp_stopped = tstamp;
}
if (event->state != PERF_EVENT_STATE_ACTIVE)
event->pending_disable = 0;
event->state = PERF_EVENT_STATE_OFF;
}
- event->tstamp_stopped = ctx->time;
+ event->tstamp_stopped = tstamp;
event->pmu->del(event, 0);
event->oncpu = -1;
struct perf_cpu_context *cpuctx,
struct perf_event_context *ctx)
{
+ u64 tstamp = perf_event_time(event);
+
if (event->state <= PERF_EVENT_STATE_OFF)
return 0;
return -EAGAIN;
}
- event->tstamp_running += ctx->time - event->tstamp_stopped;
+ event->tstamp_running += tstamp - event->tstamp_stopped;
- event->shadow_ctx_time = ctx->time - ctx->timestamp;
+ event->shadow_ctx_time = tstamp - ctx->timestamp;
if (!is_software_event(event))
cpuctx->active_oncpu++;
static void add_event_to_ctx(struct perf_event *event,
struct perf_event_context *ctx)
{
+ u64 tstamp = perf_event_time(event);
+
list_add_event(event, ctx);
perf_group_attach(event);
- event->tstamp_enabled = ctx->time;
- event->tstamp_running = ctx->time;
- event->tstamp_stopped = ctx->time;
+ event->tstamp_enabled = tstamp;
+ event->tstamp_running = tstamp;
+ event->tstamp_stopped = tstamp;
}
/*
add_event_to_ctx(event, ctx);
- if (event->cpu != -1 && event->cpu != smp_processor_id())
+ if (!event_filter_match(event))
goto unlock;
/*
struct perf_event_context *ctx)
{
struct perf_event *sub;
+ u64 tstamp = perf_event_time(event);
event->state = PERF_EVENT_STATE_INACTIVE;
- event->tstamp_enabled = ctx->time - event->total_time_enabled;
+ event->tstamp_enabled = tstamp - event->total_time_enabled;
list_for_each_entry(sub, &event->sibling_list, group_entry) {
- if (sub->state >= PERF_EVENT_STATE_INACTIVE) {
- sub->tstamp_enabled =
- ctx->time - sub->total_time_enabled;
- }
+ if (sub->state >= PERF_EVENT_STATE_INACTIVE)
+ sub->tstamp_enabled = tstamp - sub->total_time_enabled;
}
}
goto unlock;
__perf_event_mark_enabled(event, ctx);
- if (event->cpu != -1 && event->cpu != smp_processor_id())
+ if (!event_filter_match(event))
goto unlock;
/*
return 0;
}
-enum event_type_t {
- EVENT_FLEXIBLE = 0x1,
- EVENT_PINNED = 0x2,
- EVENT_ALL = EVENT_FLEXIBLE | EVENT_PINNED,
-};
-
static void ctx_sched_out(struct perf_event_context *ctx,
struct perf_cpu_context *cpuctx,
enum event_type_t event_type)
list_for_each_entry(event, &ctx->pinned_groups, group_entry) {
if (event->state <= PERF_EVENT_STATE_OFF)
continue;
- if (event->cpu != -1 && event->cpu != smp_processor_id())
+ if (!event_filter_match(event))
continue;
if (group_can_go_on(event, cpuctx, 1))
* Listen to the 'cpu' scheduling filter constraint
* of events:
*/
- if (event->cpu != -1 && event->cpu != smp_processor_id())
+ if (!event_filter_match(event))
continue;
if (group_can_go_on(event, cpuctx, can_add_hw)) {
if (event->state != PERF_EVENT_STATE_ACTIVE)
continue;
- if (event->cpu != -1 && event->cpu != smp_processor_id())
+ if (!event_filter_match(event))
continue;
hwc = &event->hw;
unsigned long flags;
int ctxn, err;
- if (!task && cpu != -1) {
+ if (!task) {
/* Must be root to operate on a CPU event: */
if (perf_paranoid_cpu() && !capable(CAP_SYS_ADMIN))
return ERR_PTR(-EACCES);
- if (cpu < 0 || cpu >= nr_cpumask_bits)
- return ERR_PTR(-EINVAL);
-
/*
* We could be clever and allow to attach a event to an
* offline CPU and activate it when the CPU comes up, but
if (event->state < PERF_EVENT_STATE_INACTIVE)
return 0;
- if (event->cpu != -1 && event->cpu != smp_processor_id())
+ if (!event_filter_match(event))
return 0;
if (event->attr.comm || event->attr.mmap ||
rcu_read_lock();
list_for_each_entry_rcu(pmu, &pmus, entry) {
cpuctx = get_cpu_ptr(pmu->pmu_cpu_context);
+ if (cpuctx->active_pmu != pmu)
+ goto next;
perf_event_task_ctx(&cpuctx->ctx, task_event);
ctx = task_event->task_ctx;
if (event->state < PERF_EVENT_STATE_INACTIVE)
return 0;
- if (event->cpu != -1 && event->cpu != smp_processor_id())
+ if (!event_filter_match(event))
return 0;
if (event->attr.comm)
rcu_read_lock();
list_for_each_entry_rcu(pmu, &pmus, entry) {
cpuctx = get_cpu_ptr(pmu->pmu_cpu_context);
+ if (cpuctx->active_pmu != pmu)
+ goto next;
perf_event_comm_ctx(&cpuctx->ctx, comm_event);
ctxn = pmu->task_ctx_nr;
if (event->state < PERF_EVENT_STATE_INACTIVE)
return 0;
- if (event->cpu != -1 && event->cpu != smp_processor_id())
+ if (!event_filter_match(event))
return 0;
if ((!executable && event->attr.mmap_data) ||
rcu_read_lock();
list_for_each_entry_rcu(pmu, &pmus, entry) {
cpuctx = get_cpu_ptr(pmu->pmu_cpu_context);
+ if (cpuctx->active_pmu != pmu)
+ goto next;
perf_event_mmap_ctx(&cpuctx->ctx, mmap_event,
vma->vm_flags & VM_EXEC);
break;
}
- if (event_id > PERF_COUNT_SW_MAX)
+ if (event_id >= PERF_COUNT_SW_MAX)
return -ENOENT;
if (!event->parent) {
static inline void perf_tp_register(void)
{
- perf_pmu_register(&perf_tracepoint);
+ perf_pmu_register(&perf_tracepoint, "tracepoint", PERF_TYPE_TRACEPOINT);
}
static int perf_event_set_filter(struct perf_event *event, void __user *arg)
return NULL;
}
-static void free_pmu_context(void * __percpu cpu_context)
+static void update_pmu_context(struct pmu *pmu, struct pmu *old_pmu)
{
- struct pmu *pmu;
+ int cpu;
+
+ for_each_possible_cpu(cpu) {
+ struct perf_cpu_context *cpuctx;
+
+ cpuctx = per_cpu_ptr(pmu->pmu_cpu_context, cpu);
+
+ if (cpuctx->active_pmu == old_pmu)
+ cpuctx->active_pmu = pmu;
+ }
+}
+
+static void free_pmu_context(struct pmu *pmu)
+{
+ struct pmu *i;
mutex_lock(&pmus_lock);
/*
* Like a real lame refcount.
*/
- list_for_each_entry(pmu, &pmus, entry) {
- if (pmu->pmu_cpu_context == cpu_context)
+ list_for_each_entry(i, &pmus, entry) {
+ if (i->pmu_cpu_context == pmu->pmu_cpu_context) {
+ update_pmu_context(i, pmu);
goto out;
+ }
}
- free_percpu(cpu_context);
+ free_percpu(pmu->pmu_cpu_context);
out:
mutex_unlock(&pmus_lock);
}
+static struct idr pmu_idr;
+
+static ssize_t
+type_show(struct device *dev, struct device_attribute *attr, char *page)
+{
+ struct pmu *pmu = dev_get_drvdata(dev);
+
+ return snprintf(page, PAGE_SIZE-1, "%d\n", pmu->type);
+}
+
+static struct device_attribute pmu_dev_attrs[] = {
+ __ATTR_RO(type),
+ __ATTR_NULL,
+};
+
+static int pmu_bus_running;
+static struct bus_type pmu_bus = {
+ .name = "event_source",
+ .dev_attrs = pmu_dev_attrs,
+};
+
+static void pmu_dev_release(struct device *dev)
+{
+ kfree(dev);
+}
+
+static int pmu_dev_alloc(struct pmu *pmu)
+{
+ int ret = -ENOMEM;
+
+ pmu->dev = kzalloc(sizeof(struct device), GFP_KERNEL);
+ if (!pmu->dev)
+ goto out;
+
+ device_initialize(pmu->dev);
+ ret = dev_set_name(pmu->dev, "%s", pmu->name);
+ if (ret)
+ goto free_dev;
+
+ dev_set_drvdata(pmu->dev, pmu);
+ pmu->dev->bus = &pmu_bus;
+ pmu->dev->release = pmu_dev_release;
+ ret = device_add(pmu->dev);
+ if (ret)
+ goto free_dev;
+
+out:
+ return ret;
+
+free_dev:
+ put_device(pmu->dev);
+ goto out;
+}
-int perf_pmu_register(struct pmu *pmu)
+int perf_pmu_register(struct pmu *pmu, char *name, int type)
{
int cpu, ret;
if (!pmu->pmu_disable_count)
goto unlock;
+ pmu->type = -1;
+ if (!name)
+ goto skip_type;
+ pmu->name = name;
+
+ if (type < 0) {
+ int err = idr_pre_get(&pmu_idr, GFP_KERNEL);
+ if (!err)
+ goto free_pdc;
+
+ err = idr_get_new_above(&pmu_idr, pmu, PERF_TYPE_MAX, &type);
+ if (err) {
+ ret = err;
+ goto free_pdc;
+ }
+ }
+ pmu->type = type;
+
+ if (pmu_bus_running) {
+ ret = pmu_dev_alloc(pmu);
+ if (ret)
+ goto free_idr;
+ }
+
+skip_type:
pmu->pmu_cpu_context = find_pmu_context(pmu->task_ctx_nr);
if (pmu->pmu_cpu_context)
goto got_cpu_context;
pmu->pmu_cpu_context = alloc_percpu(struct perf_cpu_context);
if (!pmu->pmu_cpu_context)
- goto free_pdc;
+ goto free_dev;
for_each_possible_cpu(cpu) {
struct perf_cpu_context *cpuctx;
cpuctx->ctx.pmu = pmu;
cpuctx->jiffies_interval = 1;
INIT_LIST_HEAD(&cpuctx->rotation_list);
+ cpuctx->active_pmu = pmu;
}
got_cpu_context:
return ret;
+free_dev:
+ device_del(pmu->dev);
+ put_device(pmu->dev);
+
+free_idr:
+ if (pmu->type >= PERF_TYPE_MAX)
+ idr_remove(&pmu_idr, pmu->type);
+
free_pdc:
free_percpu(pmu->pmu_disable_count);
goto unlock;
synchronize_rcu();
free_percpu(pmu->pmu_disable_count);
- free_pmu_context(pmu->pmu_cpu_context);
+ if (pmu->type >= PERF_TYPE_MAX)
+ idr_remove(&pmu_idr, pmu->type);
+ device_del(pmu->dev);
+ put_device(pmu->dev);
+ free_pmu_context(pmu);
}
struct pmu *perf_init_event(struct perf_event *event)
int idx;
idx = srcu_read_lock(&pmus_srcu);
+
+ rcu_read_lock();
+ pmu = idr_find(&pmu_idr, event->attr.type);
+ rcu_read_unlock();
+ if (pmu)
+ goto unlock;
+
list_for_each_entry_rcu(pmu, &pmus, entry) {
int ret = pmu->event_init(event);
if (!ret)
struct hw_perf_event *hwc;
long err;
+ if ((unsigned)cpu >= nr_cpu_ids) {
+ if (!task || cpu != -1)
+ return ERR_PTR(-EINVAL);
+ }
+
event = kzalloc(sizeof(*event), GFP_KERNEL);
if (!event)
return ERR_PTR(-ENOMEM);
if (!overflow_handler && parent_event)
overflow_handler = parent_event->overflow_handler;
-
+
event->overflow_handler = overflow_handler;
if (attr->disabled)
raw_spin_lock_irqsave(&parent_ctx->lock, flags);
parent_ctx->rotate_disable = 0;
- raw_spin_unlock_irqrestore(&parent_ctx->lock, flags);
child_ctx = child->perf_event_ctxp[ctxn];
/*
* Mark the child context as a clone of the parent
* context, or of whatever the parent is a clone of.
- * Note that if the parent is a clone, it could get
- * uncloned at any point, but that doesn't matter
- * because the list of events and the generation
- * count can't have changed since we took the mutex.
+ *
+ * Note that if the parent is a clone, the holding of
+ * parent_ctx->lock avoids it from being uncloned.
*/
- cloned_ctx = rcu_dereference(parent_ctx->parent_ctx);
+ cloned_ctx = parent_ctx->parent_ctx;
if (cloned_ctx) {
child_ctx->parent_ctx = cloned_ctx;
child_ctx->parent_gen = parent_ctx->parent_gen;
get_ctx(child_ctx->parent_ctx);
}
+ raw_spin_unlock_irqrestore(&parent_ctx->lock, flags);
mutex_unlock(&parent_ctx->mutex);
perf_unpin_context(parent_ctx);
mutex_unlock(&swhash->hlist_mutex);
}
-#ifdef CONFIG_HOTPLUG_CPU
+#if defined CONFIG_HOTPLUG_CPU || defined CONFIG_KEXEC
static void perf_pmu_rotate_stop(struct pmu *pmu)
{
struct perf_cpu_context *cpuctx = this_cpu_ptr(pmu->pmu_cpu_context);
static inline void perf_event_exit_cpu(int cpu) { }
#endif
+static int
+perf_reboot(struct notifier_block *notifier, unsigned long val, void *v)
+{
+ int cpu;
+
+ for_each_online_cpu(cpu)
+ perf_event_exit_cpu(cpu);
+
+ return NOTIFY_OK;
+}
+
+/*
+ * Run the perf reboot notifier at the very last possible moment so that
+ * the generic watchdog code runs as long as possible.
+ */
+static struct notifier_block perf_reboot_notifier = {
+ .notifier_call = perf_reboot,
+ .priority = INT_MIN,
+};
+
static int __cpuinit
perf_cpu_notify(struct notifier_block *self, unsigned long action, void *hcpu)
{
{
int ret;
+ idr_init(&pmu_idr);
+
perf_event_init_all_cpus();
init_srcu_struct(&pmus_srcu);
- perf_pmu_register(&perf_swevent);
- perf_pmu_register(&perf_cpu_clock);
- perf_pmu_register(&perf_task_clock);
+ perf_pmu_register(&perf_swevent, "software", PERF_TYPE_SOFTWARE);
+ perf_pmu_register(&perf_cpu_clock, NULL, -1);
+ perf_pmu_register(&perf_task_clock, NULL, -1);
perf_tp_register();
perf_cpu_notifier(perf_cpu_notify);
+ register_reboot_notifier(&perf_reboot_notifier);
ret = init_hw_breakpoint();
WARN(ret, "hw_breakpoint initialization failed with: %d", ret);
}
+
+static int __init perf_event_sysfs_init(void)
+{
+ struct pmu *pmu;
+ int ret;
+
+ mutex_lock(&pmus_lock);
+
+ ret = bus_register(&pmu_bus);
+ if (ret)
+ goto unlock;
+
+ list_for_each_entry(pmu, &pmus, entry) {
+ if (!pmu->name || pmu->type < 0)
+ continue;
+
+ ret = pmu_dev_alloc(pmu);
+ WARN(ret, "Failed to register pmu: %s, reason %d\n", pmu->name, ret);
+ }
+ pmu_bus_running = 1;
+ ret = 0;
+
+unlock:
+ mutex_unlock(&pmus_lock);
+
+ return ret;
+}
+device_initcall(perf_event_sysfs_init);