summaryrefslogtreecommitdiff
diff options
context:
space:
mode:
-rw-r--r--arch/x86/kernel/dumpstack_64.c2
-rw-r--r--include/linux/perf_event.h1
-rw-r--r--kernel/events/core.c104
-rw-r--r--tools/perf/builtin-top.c13
4 files changed, 78 insertions, 42 deletions
diff --git a/arch/x86/kernel/dumpstack_64.c b/arch/x86/kernel/dumpstack_64.c
index 6d728d9..af7785f 100644
--- a/arch/x86/kernel/dumpstack_64.c
+++ b/arch/x86/kernel/dumpstack_64.c
@@ -129,7 +129,7 @@ void dump_trace(struct task_struct *task, struct pt_regs *regs,
if (!stack) {
if (regs)
stack = (unsigned long *)regs->sp;
- else if (task && task != current)
+ else if (task != current)
stack = (unsigned long *)task->thread.sp;
else
stack = &dummy;
diff --git a/include/linux/perf_event.h b/include/linux/perf_event.h
index 0b91db2..412b790 100644
--- a/include/linux/perf_event.h
+++ b/include/linux/perf_event.h
@@ -589,6 +589,7 @@ struct hw_perf_event {
u64 sample_period;
u64 last_period;
local64_t period_left;
+ u64 interrupts_seq;
u64 interrupts;
u64 freq_time_stamp;
diff --git a/kernel/events/core.c b/kernel/events/core.c
index de859fb..7c3b9de 100644
--- a/kernel/events/core.c
+++ b/kernel/events/core.c
@@ -2300,6 +2300,9 @@ do { \
return div64_u64(dividend, divisor);
}
+static DEFINE_PER_CPU(int, perf_throttled_count);
+static DEFINE_PER_CPU(u64, perf_throttled_seq);
+
static void perf_adjust_period(struct perf_event *event, u64 nsec, u64 count)
{
struct hw_perf_event *hwc = &event->hw;
@@ -2325,16 +2328,29 @@ static void perf_adjust_period(struct perf_event *event, u64 nsec, u64 count)
}
}
-static void perf_ctx_adjust_freq(struct perf_event_context *ctx, u64 period)
+/*
+ * combine freq adjustment with unthrottling to avoid two passes over the
+ * events. At the same time, make sure, having freq events does not change
+ * the rate of unthrottling as that would introduce bias.
+ */
+static void perf_adjust_freq_unthr_context(struct perf_event_context *ctx,
+ int needs_unthr)
{
struct perf_event *event;
struct hw_perf_event *hwc;
- u64 interrupts, now;
+ u64 now, period = TICK_NSEC;
s64 delta;
- if (!ctx->nr_freq)
+ /*
+ * only need to iterate over all events iff:
+ * - context have events in frequency mode (needs freq adjust)
+ * - there are events to unthrottle on this cpu
+ */
+ if (!(ctx->nr_freq || needs_unthr))
return;
+ raw_spin_lock(&ctx->lock);
+
list_for_each_entry_rcu(event, &ctx->event_list, event_entry) {
if (event->state != PERF_EVENT_STATE_ACTIVE)
continue;
@@ -2344,13 +2360,8 @@ static void perf_ctx_adjust_freq(struct perf_event_context *ctx, u64 period)
hwc = &event->hw;
- interrupts = hwc->interrupts;
- hwc->interrupts = 0;
-
- /*
- * unthrottle events on the tick
- */
- if (interrupts == MAX_INTERRUPTS) {
+ if (needs_unthr && hwc->interrupts == MAX_INTERRUPTS) {
+ hwc->interrupts = 0;
perf_log_throttle(event, 1);
event->pmu->start(event, 0);
}
@@ -2358,14 +2369,26 @@ static void perf_ctx_adjust_freq(struct perf_event_context *ctx, u64 period)
if (!event->attr.freq || !event->attr.sample_freq)
continue;
- event->pmu->read(event);
+ /*
+ * stop the event and update event->count
+ */
+ event->pmu->stop(event, PERF_EF_UPDATE);
+
now = local64_read(&event->count);
delta = now - hwc->freq_count_stamp;
hwc->freq_count_stamp = now;
+ /*
+ * restart the event
+ * reload only if value has changed
+ */
if (delta > 0)
perf_adjust_period(event, period, delta);
+
+ event->pmu->start(event, delta > 0 ? PERF_EF_RELOAD : 0);
}
+
+ raw_spin_unlock(&ctx->lock);
}
/*
@@ -2388,16 +2411,13 @@ static void rotate_ctx(struct perf_event_context *ctx)
*/
static void perf_rotate_context(struct perf_cpu_context *cpuctx)
{
- u64 interval = (u64)cpuctx->jiffies_interval * TICK_NSEC;
struct perf_event_context *ctx = NULL;
- int rotate = 0, remove = 1, freq = 0;
+ int rotate = 0, remove = 1;
if (cpuctx->ctx.nr_events) {
remove = 0;
if (cpuctx->ctx.nr_events != cpuctx->ctx.nr_active)
rotate = 1;
- if (cpuctx->ctx.nr_freq)
- freq = 1;
}
ctx = cpuctx->task_ctx;
@@ -2405,37 +2425,26 @@ static void perf_rotate_context(struct perf_cpu_context *cpuctx)
remove = 0;
if (ctx->nr_events != ctx->nr_active)
rotate = 1;
- if (ctx->nr_freq)
- freq = 1;
}
- if (!rotate && !freq)
+ if (!rotate)
goto done;
perf_ctx_lock(cpuctx, cpuctx->task_ctx);
perf_pmu_disable(cpuctx->ctx.pmu);
- if (freq) {
- perf_ctx_adjust_freq(&cpuctx->ctx, interval);
- if (ctx)
- perf_ctx_adjust_freq(ctx, interval);
- }
-
- if (rotate) {
- cpu_ctx_sched_out(cpuctx, EVENT_FLEXIBLE);
- if (ctx)
- ctx_sched_out(ctx, cpuctx, EVENT_FLEXIBLE);
+ cpu_ctx_sched_out(cpuctx, EVENT_FLEXIBLE);
+ if (ctx)
+ ctx_sched_out(ctx, cpuctx, EVENT_FLEXIBLE);
- rotate_ctx(&cpuctx->ctx);
- if (ctx)
- rotate_ctx(ctx);
+ rotate_ctx(&cpuctx->ctx);
+ if (ctx)
+ rotate_ctx(ctx);
- perf_event_sched_in(cpuctx, ctx, current);
- }
+ perf_event_sched_in(cpuctx, ctx, current);
perf_pmu_enable(cpuctx->ctx.pmu);
perf_ctx_unlock(cpuctx, cpuctx->task_ctx);
-
done:
if (remove)
list_del_init(&cpuctx->rotation_list);
@@ -2445,10 +2454,22 @@ void perf_event_task_tick(void)
{
struct list_head *head = &__get_cpu_var(rotation_list);
struct perf_cpu_context *cpuctx, *tmp;
+ struct perf_event_context *ctx;
+ int throttled;
WARN_ON(!irqs_disabled());
+ __this_cpu_inc(perf_throttled_seq);
+ throttled = __this_cpu_xchg(perf_throttled_count, 0);
+
list_for_each_entry_safe(cpuctx, tmp, head, rotation_list) {
+ ctx = &cpuctx->ctx;
+ perf_adjust_freq_unthr_context(ctx, throttled);
+
+ ctx = cpuctx->task_ctx;
+ if (ctx)
+ perf_adjust_freq_unthr_context(ctx, throttled);
+
if (cpuctx->jiffies_interval == 1 ||
!(jiffies % cpuctx->jiffies_interval))
perf_rotate_context(cpuctx);
@@ -4514,6 +4535,7 @@ static int __perf_event_overflow(struct perf_event *event,
{
int events = atomic_read(&event->event_limit);
struct hw_perf_event *hwc = &event->hw;
+ u64 seq;
int ret = 0;
/*
@@ -4523,14 +4545,20 @@ static int __perf_event_overflow(struct perf_event *event,
if (unlikely(!is_sampling_event(event)))
return 0;
- if (unlikely(hwc->interrupts >= max_samples_per_tick)) {
- if (throttle) {
+ seq = __this_cpu_read(perf_throttled_seq);
+ if (seq != hwc->interrupts_seq) {
+ hwc->interrupts_seq = seq;
+ hwc->interrupts = 1;
+ } else {
+ hwc->interrupts++;
+ if (unlikely(throttle
+ && hwc->interrupts >= max_samples_per_tick)) {
+ __this_cpu_inc(perf_throttled_count);
hwc->interrupts = MAX_INTERRUPTS;
perf_log_throttle(event, 0);
ret = 1;
}
- } else
- hwc->interrupts++;
+ }
if (event->attr.freq) {
u64 now = perf_clock();
diff --git a/tools/perf/builtin-top.c b/tools/perf/builtin-top.c
index e8b033c..d869b21 100644
--- a/tools/perf/builtin-top.c
+++ b/tools/perf/builtin-top.c
@@ -88,8 +88,6 @@ void get_term_dimensions(struct winsize *ws)
static void perf_top__update_print_entries(struct perf_top *top)
{
- top->print_entries = top->winsize.ws_row;
-
if (top->print_entries > 9)
top->print_entries -= 9;
}
@@ -99,6 +97,13 @@ static void perf_top__sig_winch(int sig __used, siginfo_t *info __used, void *ar
struct perf_top *top = arg;
get_term_dimensions(&top->winsize);
+ if (!top->print_entries
+ || (top->print_entries+4) > top->winsize.ws_row) {
+ top->print_entries = top->winsize.ws_row;
+ } else {
+ top->print_entries += 4;
+ top->winsize.ws_row = top->print_entries;
+ }
perf_top__update_print_entries(top);
}
@@ -452,8 +457,10 @@ static void perf_top__handle_keypress(struct perf_top *top, int c)
};
perf_top__sig_winch(SIGWINCH, NULL, top);
sigaction(SIGWINCH, &act, NULL);
- } else
+ } else {
+ perf_top__sig_winch(SIGWINCH, NULL, top);
signal(SIGWINCH, SIG_DFL);
+ }
break;
case 'E':
if (top->evlist->nr_entries > 1) {