1 files changed, 124 insertions, 50 deletions

diff --git a/arch/x86/kernel/cpu/perf_event.c b/arch/x86/kernel/cpu/perf_event.c
index 87848ebe2bb7..5801a14f7524 100644
--- a/arch/x86/kernel/cpu/perf_event.c
+++ b/arch/x86/kernel/cpu/perf_event.c
@@ -135,6 +135,7 @@ static int x86_pmu_extra_regs(u64 config, struct perf_event *event)
 }
 
 static atomic_t active_events;
+static atomic_t pmc_refcount;
 static DEFINE_MUTEX(pmc_reserve_mutex);
 
 #ifdef CONFIG_X86_LOCAL_APIC
@@ -190,6 +191,7 @@ static bool check_hw_exists(void)
 	u64 val, val_fail, val_new= ~0;
 	int i, reg, reg_fail, ret = 0;
 	int bios_fail = 0;
+	int reg_safe = -1;
 
 	/*
 	 * Check to see if the BIOS enabled any of the counters, if so
@@ -204,6 +206,8 @@ static bool check_hw_exists(void)
 			bios_fail = 1;
 			val_fail = val;
 			reg_fail = reg;
+		} else {
+			reg_safe = i;
 		}
 	}
 
@@ -222,11 +226,22 @@ static bool check_hw_exists(void)
 	}
 
 	/*
+	 * If all the counters are enabled, the below test will always
+	 * fail.  The tools will also become useless in this scenario.
+	 * Just fail and disable the hardware counters.
+	 */
+
+	if (reg_safe == -1) {
+		reg = reg_safe;
+		goto msr_fail;
+	}
+
+	/*
 	 * Read the current value, change it and read it back to see if it
 	 * matches, this is needed to detect certain hardware emulators
 	 * (qemu/kvm) that don't trap on the MSR access and always return 0s.
 	 */
-	reg = x86_pmu_event_addr(0);
+	reg = x86_pmu_event_addr(reg_safe);
 	if (rdmsrl_safe(reg, &val))
 		goto msr_fail;
 	val ^= 0xffffUL;
@@ -256,11 +271,8 @@ msr_fail:
 
 static void hw_perf_event_destroy(struct perf_event *event)
 {
-	if (atomic_dec_and_mutex_lock(&active_events, &pmc_reserve_mutex)) {
-		release_pmc_hardware();
-		release_ds_buffers();
-		mutex_unlock(&pmc_reserve_mutex);
-	}
+	x86_release_hardware();
+	atomic_dec(&active_events);
 }
 
 void hw_perf_lbr_event_destroy(struct perf_event *event)
@@ -310,6 +322,35 @@ set_ext_hw_attr(struct hw_perf_event *hwc, struct perf_event *event)
 	return x86_pmu_extra_regs(val, event);
 }
 
+int x86_reserve_hardware(void)
+{
+	int err = 0;
+
+	if (!atomic_inc_not_zero(&pmc_refcount)) {
+		mutex_lock(&pmc_reserve_mutex);
+		if (atomic_read(&pmc_refcount) == 0) {
+			if (!reserve_pmc_hardware())
+				err = -EBUSY;
+			else
+				reserve_ds_buffers();
+		}
+		if (!err)
+			atomic_inc(&pmc_refcount);
+		mutex_unlock(&pmc_reserve_mutex);
+	}
+
+	return err;
+}
+
+void x86_release_hardware(void)
+{
+	if (atomic_dec_and_mutex_lock(&pmc_refcount, &pmc_reserve_mutex)) {
+		release_pmc_hardware();
+		release_ds_buffers();
+		mutex_unlock(&pmc_reserve_mutex);
+	}
+}
+
 /*
  * Check if we can create event of a certain type (that no conflicting events
  * are present).
@@ -322,21 +363,34 @@ int x86_add_exclusive(unsigned int what)
 		return 0;
 
 	mutex_lock(&pmc_reserve_mutex);
-	for (i = 0; i < ARRAY_SIZE(x86_pmu.lbr_exclusive); i++)
+	for (i = 0; i < ARRAY_SIZE(x86_pmu.lbr_exclusive); i++) {
 		if (i != what && atomic_read(&x86_pmu.lbr_exclusive[i]))
 			goto out;
+	}
 
 	atomic_inc(&x86_pmu.lbr_exclusive[what]);
 	ret = 0;
 
 out:
 	mutex_unlock(&pmc_reserve_mutex);
+
+	/*
+	 * Assuming that all exclusive events will share the PMI handler
+	 * (which checks active_events for whether there is work to do),
+	 * we can bump active_events counter right here, except for
+	 * x86_lbr_exclusive_lbr events that go through x86_pmu_event_init()
+	 * path, which already bumps active_events for them.
+	 */
+	if (!ret && what != x86_lbr_exclusive_lbr)
+		atomic_inc(&active_events);
+
 	return ret;
 }
 
 void x86_del_exclusive(unsigned int what)
 {
 	atomic_dec(&x86_pmu.lbr_exclusive[what]);
+	atomic_dec(&active_events);
 }
 
 int x86_setup_perfctr(struct perf_event *event)
@@ -513,22 +567,11 @@ static int __x86_pmu_event_init(struct perf_event *event)
 	if (!x86_pmu_initialized())
 		return -ENODEV;
 
-	err = 0;
-	if (!atomic_inc_not_zero(&active_events)) {
-		mutex_lock(&pmc_reserve_mutex);
-		if (atomic_read(&active_events) == 0) {
-			if (!reserve_pmc_hardware())
-				err = -EBUSY;
-			else
-				reserve_ds_buffers();
-		}
-		if (!err)
-			atomic_inc(&active_events);
-		mutex_unlock(&pmc_reserve_mutex);
-	}
+	err = x86_reserve_hardware();
 	if (err)
 		return err;
 
+	atomic_inc(&active_events);
 	event->destroy = hw_perf_event_destroy;
 
 	event->hw.idx = -1;
@@ -611,6 +654,7 @@ struct sched_state {
 	int	event;		/* event index */
 	int	counter;	/* counter index */
 	int	unassigned;	/* number of events to be assigned left */
+	int	nr_gp;		/* number of GP counters used */
 	unsigned long used[BITS_TO_LONGS(X86_PMC_IDX_MAX)];
 };
 
@@ -620,27 +664,29 @@ struct sched_state {
 struct perf_sched {
 	int			max_weight;
 	int			max_events;
-	struct perf_event	**events;
-	struct sched_state	state;
+	int			max_gp;
 	int			saved_states;
+	struct event_constraint	**constraints;
+	struct sched_state	state;
 	struct sched_state	saved[SCHED_STATES_MAX];
 };
 
 /*
  * Initialize interator that runs through all events and counters.
  */
-static void perf_sched_init(struct perf_sched *sched, struct perf_event **events,
-			    int num, int wmin, int wmax)
+static void perf_sched_init(struct perf_sched *sched, struct event_constraint **constraints,
+			    int num, int wmin, int wmax, int gpmax)
 {
 	int idx;
 
 	memset(sched, 0, sizeof(*sched));
 	sched->max_events	= num;
 	sched->max_weight	= wmax;
-	sched->events		= events;
+	sched->max_gp		= gpmax;
+	sched->constraints	= constraints;
 
 	for (idx = 0; idx < num; idx++) {
-		if (events[idx]->hw.constraint->weight == wmin)
+		if (constraints[idx]->weight == wmin)
 			break;
 	}
 
@@ -687,7 +733,7 @@ static bool __perf_sched_find_counter(struct perf_sched *sched)
 	if (sched->state.event >= sched->max_events)
 		return false;
 
-	c = sched->events[sched->state.event]->hw.constraint;
+	c = sched->constraints[sched->state.event];
 	/* Prefer fixed purpose counters */
 	if (c->idxmsk64 & (~0ULL << INTEL_PMC_IDX_FIXED)) {
 		idx = INTEL_PMC_IDX_FIXED;
@@ -696,11 +742,16 @@ static bool __perf_sched_find_counter(struct perf_sched *sched)
 				goto done;
 		}
 	}
+
 	/* Grab the first unused counter starting with idx */
 	idx = sched->state.counter;
 	for_each_set_bit_from(idx, c->idxmsk, INTEL_PMC_IDX_FIXED) {
-		if (!__test_and_set_bit(idx, sched->state.used))
+		if (!__test_and_set_bit(idx, sched->state.used)) {
+			if (sched->state.nr_gp++ >= sched->max_gp)
+				return false;
+
 			goto done;
+		}
 	}
 
 	return false;
@@ -745,7 +796,7 @@ static bool perf_sched_next_event(struct perf_sched *sched)
 			if (sched->state.weight > sched->max_weight)
 				return false;
 		}
-		c = sched->events[sched->state.event]->hw.constraint;
+		c = sched->constraints[sched->state.event];
 	} while (c->weight != sched->state.weight);
 
 	sched->state.counter = 0;	/* start with first counter */
@@ -756,12 +807,12 @@ static bool perf_sched_next_event(struct perf_sched *sched)
 /*
  * Assign a counter for each event.
  */
-int perf_assign_events(struct perf_event **events, int n,
-			int wmin, int wmax, int *assign)
+int perf_assign_events(struct event_constraint **constraints, int n,
+			int wmin, int wmax, int gpmax, int *assign)
 {
 	struct perf_sched sched;
 
-	perf_sched_init(&sched, events, n, wmin, wmax);
+	perf_sched_init(&sched, constraints, n, wmin, wmax, gpmax);
 
 	do {
 		if (!perf_sched_find_counter(&sched))
@@ -788,9 +839,9 @@ int x86_schedule_events(struct cpu_hw_events *cpuc, int n, int *assign)
 		x86_pmu.start_scheduling(cpuc);
 
 	for (i = 0, wmin = X86_PMC_IDX_MAX, wmax = 0; i < n; i++) {
-		hwc = &cpuc->event_list[i]->hw;
+		cpuc->event_constraint[i] = NULL;
 		c = x86_pmu.get_event_constraints(cpuc, i, cpuc->event_list[i]);
-		hwc->constraint = c;
+		cpuc->event_constraint[i] = c;
 
 		wmin = min(wmin, c->weight);
 		wmax = max(wmax, c->weight);
@@ -801,7 +852,7 @@ int x86_schedule_events(struct cpu_hw_events *cpuc, int n, int *assign)
 	 */
 	for (i = 0; i < n; i++) {
 		hwc = &cpuc->event_list[i]->hw;
-		c = hwc->constraint;
+		c = cpuc->event_constraint[i];
 
 		/* never assigned */
 		if (hwc->idx == -1)
@@ -821,9 +872,26 @@ int x86_schedule_events(struct cpu_hw_events *cpuc, int n, int *assign)
 	}
 
 	/* slow path */
-	if (i != n)
-		unsched = perf_assign_events(cpuc->event_list, n, wmin,
-					     wmax, assign);
+	if (i != n) {
+		int gpmax = x86_pmu.num_counters;
+
+		/*
+		 * Do not allow scheduling of more than half the available
+		 * generic counters.
+		 *
+		 * This helps avoid counter starvation of sibling thread by
+		 * ensuring at most half the counters cannot be in exclusive
+		 * mode. There is no designated counters for the limits. Any
+		 * N/2 counters can be used. This helps with events with
+		 * specific counter constraints.
+		 */
+		if (is_ht_workaround_enabled() && !cpuc->is_fake &&
+		    READ_ONCE(cpuc->excl_cntrs->exclusive_present))
+			gpmax /= 2;
+
+		unsched = perf_assign_events(cpuc->event_constraint, n, wmin,
+					     wmax, gpmax, assign);
+	}
 
 	/*
 	 * In case of success (unsched = 0), mark events as committed,
@@ -840,12 +908,9 @@ int x86_schedule_events(struct cpu_hw_events *cpuc, int n, int *assign)
 			e = cpuc->event_list[i];
 			e->hw.flags |= PERF_X86_EVENT_COMMITTED;
 			if (x86_pmu.commit_scheduling)
-				x86_pmu.commit_scheduling(cpuc, e, assign[i]);
+				x86_pmu.commit_scheduling(cpuc, i, assign[i]);
 		}
-	}
-
-	if (!assign || unsched) {
-
+	} else {
 		for (i = 0; i < n; i++) {
 			e = cpuc->event_list[i];
 			/*
@@ -1058,13 +1123,16 @@ int x86_perf_event_set_period(struct perf_event *event)
 
 	per_cpu(pmc_prev_left[idx], smp_processor_id()) = left;
 
-	/*
-	 * The hw event starts counting from this event offset,
-	 * mark it to be able to extra future deltas:
-	 */
-	local64_set(&hwc->prev_count, (u64)-left);
+	if (!(hwc->flags & PERF_X86_EVENT_AUTO_RELOAD) ||
+	    local64_read(&hwc->prev_count) != (u64)-left) {
+		/*
+		 * The hw event starts counting from this event offset,
+		 * mark it to be able to extra future deltas:
+		 */
+		local64_set(&hwc->prev_count, (u64)-left);
 
-	wrmsrl(hwc->event_base, (u64)(-left) & x86_pmu.cntval_mask);
+		wrmsrl(hwc->event_base, (u64)(-left) & x86_pmu.cntval_mask);
+	}
 
 	/*
 	 * Due to erratum on certan cpu we need
@@ -1292,8 +1360,10 @@ static void x86_pmu_del(struct perf_event *event, int flags)
 		x86_pmu.put_event_constraints(cpuc, event);
 
 	/* Delete the array entry. */
-	while (++i < cpuc->n_events)
+	while (++i < cpuc->n_events) {
 		cpuc->event_list[i-1] = cpuc->event_list[i];
+		cpuc->event_constraint[i-1] = cpuc->event_constraint[i];
+	}
 	--cpuc->n_events;
 
 	perf_event_update_userpage(event);
@@ -1374,6 +1444,10 @@ perf_event_nmi_handler(unsigned int cmd, struct pt_regs *regs)
 	u64 finish_clock;
 	int ret;
 
+	/*
+	 * All PMUs/events that share this PMI handler should make sure to
+	 * increment active_events for their events.
+	 */
 	if (!atomic_read(&active_events))
 		return NMI_DONE;