x86/perf/intel/uncore|rapl: Fix error handling and sanitize pmu management

@@ -774,9 +774,9 @@ static void __uncore_exit_boxes(struct i
 
 static void uncore_exit_boxes(void *dummy)
 {
-	struct intel_uncore_type **types = uncore_msr_uncores;
+	struct intel_uncore_type **types;
 
-	while (*types)
+	for (types = uncore_msr_uncores; *types; types++)
 		__uncore_exit_boxes(*types++, smp_processor_id());
 }
 

@@ -313,7 +313,7 @@ static int snb_uncore_imc_event_init(str
 		return -EINVAL;
 
 	event->cpu = box->cpu;
-	event->pmu_private = pmu;
+	event->pmu_private = box;
 
 	event->hw.idx = -1;
 	event->hw.last_tag = ~0ULL;

@@ -12,7 +12,6 @@ static void show_cpuinfo_core(struct seq
 {
 #ifdef CONFIG_SMP
 	seq_printf(m, "physical id\t: %d\n", c->phys_proc_id);
-	seq_printf(m, "logical id\t: %d\n", c->logical_proc_id);
 	seq_printf(m, "siblings\t: %d\n",
 		   cpumask_weight(topology_core_cpumask(cpu)));
 	seq_printf(m, "core id\t\t: %d\n", c->cpu_core_id);

@@ -9335,7 +9335,7 @@ ssize_t perf_event_sysfs_show(struct dev
 
 	return 0;
 }
-EXPORT_SYMBOL(perf_event_sysfs_show);
+EXPORT_SYMBOL_GPL(perf_event_sysfs_show);
 
 static int __init perf_event_sysfs_init(void)
 {

@@ -1342,12 +1342,6 @@ static void __init uncore_cpumask_init(v
 {
 	int cpu;
 
-	/*
-	 * ony invoke once from msr or pci init code
-	 */
-	if (!cpumask_empty(&uncore_cpu_mask))
-		return;
-
 	cpu_notifier_register_begin();
 
 	for_each_online_cpu(cpu) {

@@ -767,10 +767,12 @@ static void __init uncore_type_exit(stru
 {
 	int i;
 
-	for (i = 0; i < type->num_boxes; i++)
-		free_percpu(type->pmus[i].box);
-	kfree(type->pmus);
-	type->pmus = NULL;
+	if (type->pmus) {
+		for (i = 0; i < type->num_boxes; i++)
+			free_percpu(type->pmus[i].box);
+		kfree(type->pmus);
+		type->pmus = NULL;
+	}
 	kfree(type->events_group);
 	type->events_group = NULL;
 }

@@ -778,6 +780,7 @@ static void __init uncore_type_exit(stru
 static void __init uncore_types_exit(struct intel_uncore_type **types)
 {
 	int i;
+
 	for (i = 0; types[i]; i++)
 		uncore_type_exit(types[i]);
 }

@@ -806,7 +809,7 @@ static int __init uncore_type_init(struc
 		INIT_LIST_HEAD(&pmus[i].box_list);
 		pmus[i].box = alloc_percpu(struct intel_uncore_box *);
 		if (!pmus[i].box)
-			goto fail;
+			return -ENOMEM;
 	}
 
 	if (type->event_descs) {

@@ -817,7 +820,7 @@ static int __init uncore_type_init(struc
 		attr_group = kzalloc(sizeof(struct attribute *) * (i + 1) +
 					sizeof(*attr_group), GFP_KERNEL);
 		if (!attr_group)
-			goto fail;
+			return -ENOMEM;
 
 		attrs = (struct attribute **)(attr_group + 1);
 		attr_group->name = "events";

@@ -831,9 +834,6 @@ static int __init uncore_type_init(struc
 
 	type->pmu_group = &uncore_pmu_attr_group;
 	return 0;
-fail:
-	uncore_type_exit(type);
-	return -ENOMEM;
 }
 
 static int __init uncore_types_init(struct intel_uncore_type **types)

@@ -843,13 +843,9 @@ static int __init uncore_types_init(stru
 	for (i = 0; types[i]; i++) {
 		ret = uncore_type_init(types[i]);
 		if (ret)
-			goto fail;
+			return ret;
 	}
 	return 0;
-fail:
-	while (--i >= 0)
-		uncore_type_exit(types[i]);
-	return ret;
 }
 
 /*

@@ -1007,17 +1003,21 @@ static int __init uncore_pci_init(void)
 
 	ret = uncore_types_init(uncore_pci_uncores);
 	if (ret)
-		return ret;
+		goto err;
 
 	uncore_pci_driver->probe = uncore_pci_probe;
 	uncore_pci_driver->remove = uncore_pci_remove;
 
 	ret = pci_register_driver(uncore_pci_driver);
-	if (ret == 0)
-		pcidrv_registered = true;
-	else
-		uncore_types_exit(uncore_pci_uncores);
+	if (ret)
+		goto err;
+
+	pcidrv_registered = true;
+	return 0;
 
+err:
+	uncore_types_exit(uncore_pci_uncores);
+	uncore_pci_uncores = empty_uncore;
 	return ret;
 }
 

@@ -1316,9 +1316,12 @@ static int __init uncore_cpu_init(void)
 
 	ret = uncore_types_init(uncore_msr_uncores);
 	if (ret)
-		return ret;
-
+		goto err;
 	return 0;
+err:
+	uncore_types_exit(uncore_msr_uncores);
+	uncore_msr_uncores = empty_uncore;
+	return ret;
 }
 
 static int __init uncore_pmus_register(void)

Index: b/arch/x86/kernel/cpu/perf_event_intel_uncore.c
===================================================================
@@ -38,6 +38,16 @@ int uncore_pcibus_to_physid(struct pci_b
 	return phys_id;
 }
 
+static void uncore_free_pcibus_map(void)
+{
+	struct pci2phy_map *map, *tmp;
+
+	list_for_each_entry_safe(map, tmp, &pci2phy_map_head, list) {
+		list_del(&map->list);
+		kfree(map);
+	}
+}
+
 struct pci2phy_map *__find_pci2phy_map(int segment)
 {
 	struct pci2phy_map *map, *alloc = NULL;

@@ -760,16 +770,28 @@ static int uncore_pmu_register(struct in
 	}
 
 	ret = perf_pmu_register(&pmu->pmu, pmu->name, -1);
+	if (!ret)
+		pmu->registered = true;
 	return ret;
 }
 
+static void uncore_pmu_unregister(struct intel_uncore_pmu *pmu)
+{
+	if (!pmu->registered)
+		return;
+	perf_pmu_unregister(&pmu->pmu);
+	pmu->registered = false;
+}
+
 static void __init uncore_type_exit(struct intel_uncore_type *type)
 {
 	int i;
 
 	if (type->pmus) {
-		for (i = 0; i < type->num_boxes; i++)
+		for (i = 0; i < type->num_boxes; i++) {
+			uncore_pmu_unregister(&type->pmus[i]);
 			free_percpu(type->pmus[i].box);
+		}
 		kfree(type->pmus);
 		type->pmus = NULL;
 	}

@@ -856,8 +878,8 @@ static int uncore_pci_probe(struct pci_d
 	struct intel_uncore_pmu *pmu;
 	struct intel_uncore_box *box;
 	struct intel_uncore_type *type;
-	int phys_id;
 	bool first_box = false;
+	int phys_id, ret;
 
 	phys_id = uncore_pcibus_to_physid(pdev->bus);
 	if (phys_id < 0)

@@ -906,9 +928,18 @@ static int uncore_pci_probe(struct pci_d
 	list_add_tail(&box->list, &pmu->box_list);
 	raw_spin_unlock(&uncore_box_lock);
 
-	if (first_box)
-		uncore_pmu_register(pmu);
-	return 0;
+	if (!first_box)
+		return 0;
+
+	ret = uncore_pmu_register(pmu);
+	if (ret) {
+		pci_set_drvdata(pdev, NULL);
+		raw_spin_lock(&uncore_box_lock);
+		list_del(&box->list);
+		raw_spin_unlock(&uncore_box_lock);
+		kfree(box);
+	}
+	return ret;
 }
 
 static void uncore_pci_remove(struct pci_dev *pdev)

@@ -954,7 +985,7 @@ static void uncore_pci_remove(struct pci
 	kfree(box);
 
 	if (last_box)
-		perf_pmu_unregister(&pmu->pmu);
+		uncore_pmu_unregister(pmu);
 }
 
 static int __init uncore_pci_init(void)

@@ -1018,6 +1049,7 @@ static int __init uncore_pci_init(void)
 err:
 	uncore_types_exit(uncore_pci_uncores);
 	uncore_pci_uncores = empty_uncore;
+	uncore_free_pcibus_map();
 	return ret;
 }
 

@@ -1027,6 +1059,7 @@ static void __init uncore_pci_exit(void)
 		pcidrv_registered = false;
 		pci_unregister_driver(uncore_pci_driver);
 		uncore_types_exit(uncore_pci_uncores);
+		uncore_free_pcibus_map();
 	}
 }
 

@@ -1223,8 +1256,7 @@ static int uncore_cpu_notifier(struct no
 	/* allocate/free data structure for uncore box */
 	switch (action & ~CPU_TASKS_FROZEN) {
 	case CPU_UP_PREPARE:
-		uncore_cpu_prepare(cpu, -1);
-		break;
+		return notifier_from_errno(uncore_cpu_prepare(cpu, -1));
 	case CPU_STARTING:
 		uncore_cpu_starting(cpu);
 		break;

@@ -1265,9 +1297,29 @@ static struct notifier_block uncore_cpu_
 	.priority	= CPU_PRI_PERF + 1,
 };
 
-static void __init uncore_cpu_setup(void *dummy)
+static int __init type_pmu_register(struct intel_uncore_type *type)
 {
-	uncore_cpu_starting(smp_processor_id());
+	int i, ret;
+
+	for (i = 0; i < type->num_boxes; i++) {
+		ret = uncore_pmu_register(&type->pmus[i]);
+		if (ret)
+			return ret;
+	}
+	return 0;
+}
+
+static int __init uncore_msr_pmus_register(void)
+{
+	struct intel_uncore_type **types = uncore_msr_uncores;
+	int ret;
+
+	while (*types) {
+		ret = type_pmu_register(*types++);
+		if (ret)
+			return ret;
+	}
+	return 0;
 }
 
 static int __init uncore_cpu_init(void)

@@ -1317,6 +1369,10 @@ static int __init uncore_cpu_init(void)
 	ret = uncore_types_init(uncore_msr_uncores);
 	if (ret)
 		goto err;
+
+	ret = uncore_msr_pmus_register();
+	if (ret)
+		goto err;
 	return 0;
 err:
 	uncore_types_exit(uncore_msr_uncores);

@@ -1324,26 +1380,14 @@ static int __init uncore_cpu_init(void)
 	return ret;
 }
 
-static int __init uncore_pmus_register(void)
+static void __init uncore_cpu_setup(void *dummy)
 {
-	struct intel_uncore_pmu *pmu;
-	struct intel_uncore_type *type;
-	int i, j;
-
-	for (i = 0; uncore_msr_uncores[i]; i++) {
-		type = uncore_msr_uncores[i];
-		for (j = 0; j < type->num_boxes; j++) {
-			pmu = &type->pmus[j];
-			uncore_pmu_register(pmu);
-		}
-	}
-
-	return 0;
+	uncore_cpu_starting(smp_processor_id());
 }
 
-static void __init uncore_cpumask_init(void)
+static int __init uncore_cpumask_init(void)
 {
-	int cpu;
+	int cpu, ret = 0;
 
 	cpu_notifier_register_begin();
 

@@ -1359,17 +1403,20 @@ static void __init uncore_cpumask_init(v
 		if (phys_id < 0)
 			continue;
 
-		uncore_cpu_prepare(cpu, phys_id);
+		ret = uncore_cpu_prepare(cpu, phys_id);
+		if (ret)
+			goto out;
 		uncore_event_init_cpu(cpu);
 	}
 	on_each_cpu(uncore_cpu_setup, NULL, 1);
 
 	__register_cpu_notifier(&uncore_cpu_nb);
 
+out:
 	cpu_notifier_register_done();
+	return ret;
 }
 
-
 static int __init intel_uncore_init(void)
 {
 	int ret;

@@ -1382,17 +1429,20 @@ static int __init intel_uncore_init(void
 
 	ret = uncore_pci_init();
 	if (ret)
-		goto fail;
+		return ret;
 	ret = uncore_cpu_init();
-	if (ret) {
-		uncore_pci_exit();
-		goto fail;
-	}
-	uncore_cpumask_init();
+	if (ret)
+		goto errpci;
+	ret = uncore_cpumask_init();
+	if (ret)
+		goto errcpu;
 
-	uncore_pmus_register();
 	return 0;
-fail:
+
+errcpu:
+	uncore_types_exit(uncore_msr_uncores);
+errpci:
+	uncore_pci_exit();
 	return ret;
 }
 device_initcall(intel_uncore_init);

Index: b/arch/x86/kernel/cpu/perf_event_intel_uncore.h
===================================================================
@@ -73,13 +73,14 @@ struct intel_uncore_ops {
 };
 
 struct intel_uncore_pmu {
-	struct pmu pmu;
-	char name[UNCORE_PMU_NAME_LEN];
-	int pmu_idx;
-	int func_id;
-	struct intel_uncore_type *type;
-	struct intel_uncore_box ** __percpu box;
-	struct list_head box_list;
+	struct pmu			pmu;
+	char				name[UNCORE_PMU_NAME_LEN];
+	int				pmu_idx;
+	int				func_id;
+	bool				registered;
+	struct intel_uncore_type	*type;
+	struct intel_uncore_box		** __percpu box;
+	struct list_head		box_list;
 };
 
 struct intel_uncore_extra_reg {

Index: b/arch/x86/kernel/cpu/perf_event_intel_uncore.c
===================================================================
@@ -882,7 +882,7 @@ static int uncore_pci_probe(struct pci_d
 	int phys_id, ret;
 
 	phys_id = uncore_pcibus_to_physid(pdev->bus);
-	if (phys_id < 0)
+	if (phys_id < 0 || phys_id >= UNCORE_SOCKET_MAX)
 		return -ENODEV;
 
 	if (UNCORE_PCI_DEV_TYPE(id->driver_data) == UNCORE_EXTRA_PCI_DEV) {

@@ -937,6 +937,7 @@ static int uncore_pci_probe(struct pci_d
 		raw_spin_lock(&uncore_box_lock);
 		list_del(&box->list);
 		raw_spin_unlock(&uncore_box_lock);
+		uncore_box_exit(box);
 		kfree(box);
 	}
 	return ret;

@@ -982,6 +983,7 @@ static void uncore_pci_remove(struct pci
 	}
 
 	WARN_ON_ONCE(atomic_read(&box->refcnt) != 1);
+	uncore_box_exit(box);
 	kfree(box);
 
 	if (last_box)

@@ -1091,8 +1093,10 @@ static void uncore_cpu_dying(int cpu)
 			pmu = &type->pmus[j];
 			box = *per_cpu_ptr(pmu->box, cpu);
 			*per_cpu_ptr(pmu->box, cpu) = NULL;
-			if (box && atomic_dec_and_test(&box->refcnt))
+			if (box && atomic_dec_and_test(&box->refcnt)) {
 				list_add(&box->list, &boxes_to_free);
+				uncore_box_exit(box);
+			}
 		}
 	}
 }

@@ -61,6 +61,7 @@ struct intel_uncore_type {
 
 struct intel_uncore_ops {
 	void (*init_box)(struct intel_uncore_box *);
+	void (*exit_box)(struct intel_uncore_box *);
 	void (*disable_box)(struct intel_uncore_box *);
 	void (*enable_box)(struct intel_uncore_box *);
 	void (*disable_event)(struct intel_uncore_box *, struct perf_event *);

@@ -306,6 +307,14 @@ static inline void uncore_box_init(struc
 	}
 }
 
+static inline void uncore_box_exit(struct intel_uncore_box *box)
+{
+	if (test_and_clear_bit(UNCORE_BOX_FLAG_INITIATED, &box->flags)) {
+		if (box->pmu->type->ops->exit_box)
+			box->pmu->type->ops->exit_box(box);
+	}
+}
+
 static inline bool uncore_box_is_fake(struct intel_uncore_box *box)
 {
 	return (box->phys_id < 0);

@@ -201,6 +201,11 @@ static void nhmex_uncore_msr_init_box(st
 	wrmsrl(NHMEX_U_MSR_PMON_GLOBAL_CTL, NHMEX_U_PMON_GLOBAL_EN_ALL);
 }
 
+static void nhmex_uncore_msr_exit_box(struct intel_uncore_box *box)
+{
+	wrmsrl(NHMEX_U_MSR_PMON_GLOBAL_CTL, 0);
+}
+
 static void nhmex_uncore_msr_disable_box(struct intel_uncore_box *box)
 {
 	unsigned msr = uncore_msr_box_ctl(box);

@@ -250,6 +255,7 @@ static void nhmex_uncore_msr_enable_even
 
 #define NHMEX_UNCORE_OPS_COMMON_INIT()				\
 	.init_box	= nhmex_uncore_msr_init_box,		\
+	.exit_box	= nhmex_uncore_msr_exit_box,		\
 	.disable_box	= nhmex_uncore_msr_disable_box,		\
 	.enable_box	= nhmex_uncore_msr_enable_box,		\
 	.disable_event	= nhmex_uncore_msr_disable_event,	\

@@ -95,6 +95,12 @@ static void snb_uncore_msr_init_box(stru
 	}
 }
 
+static void snb_uncore_msr_exit_box(struct intel_uncore_box *box)
+{
+	if (box->pmu->pmu_idx == 0)
+		wrmsrl(SNB_UNC_PERF_GLOBAL_CTL, 0);
+}
+
 static struct uncore_event_desc snb_uncore_events[] = {
 	INTEL_UNCORE_EVENT_DESC(clockticks, "event=0xff,umask=0x00"),
 	{ /* end: all zeroes */ },

@@ -116,6 +122,7 @@ static struct attribute_group snb_uncore
 
 static struct intel_uncore_ops snb_uncore_msr_ops = {
 	.init_box	= snb_uncore_msr_init_box,
+	.exit_box	= snb_uncore_msr_exit_box,
 	.disable_event	= snb_uncore_msr_disable_event,
 	.enable_event	= snb_uncore_msr_enable_event,
 	.read_counter	= uncore_msr_read_counter,

@@ -231,6 +238,11 @@ static void snb_uncore_imc_init_box(stru
 	box->hrtimer_duration = UNCORE_SNB_IMC_HRTIMER_INTERVAL;
 }
 
+static void snb_uncore_imc_exit_box(struct intel_uncore_box *box)
+{
+	iounmap(box->io_addr);
+}
+
 static void snb_uncore_imc_enable_box(struct intel_uncore_box *box)
 {}
 

@@ -458,6 +470,7 @@ static struct pmu snb_uncore_imc_pmu = {
 
 static struct intel_uncore_ops snb_uncore_imc_ops = {
 	.init_box	= snb_uncore_imc_init_box,
+	.exit_box	= snb_uncore_imc_exit_box,
 	.enable_box	= snb_uncore_imc_enable_box,
 	.disable_box	= snb_uncore_imc_disable_box,
 	.disable_event	= snb_uncore_imc_disable_event,

@@ -1294,11 +1294,6 @@ static int uncore_cpu_notifier(struct no
 
 static struct notifier_block uncore_cpu_nb = {
 	.notifier_call	= uncore_cpu_notifier,
-	/*
-	 * to migrate uncore events, our notifier should be executed
-	 * before perf core's notifier.
-	 */
-	.priority	= CPU_PRI_PERF + 1,
 };
 
 static int __init type_pmu_register(struct intel_uncore_type *type)

Index: b/arch/x86/kernel/cpu/perf_event_intel_uncore.c
===================================================================
@@ -217,7 +217,8 @@ u64 uncore_shared_reg_config(struct inte
 	return config;
 }
 
-static void uncore_assign_hw_event(struct intel_uncore_box *box, struct perf_event *event, int idx)
+static void uncore_assign_hw_event(struct intel_uncore_box *box,
+				   struct perf_event *event, int idx)
 {
 	struct hw_perf_event *hwc = &event->hw;
 

@@ -312,18 +313,19 @@ static void uncore_pmu_init_hrtimer(stru
 	box->hrtimer.function = uncore_pmu_hrtimer;
 }
 
-static struct intel_uncore_box *uncore_alloc_box(struct intel_uncore_type *type, int node)
+static struct intel_uncore_box *uncore_alloc_box(struct intel_uncore_type *type,
+						 int node)
 {
+	int i, size, numshared = type->num_shared_regs ;
 	struct intel_uncore_box *box;
-	int i, size;
 
-	size = sizeof(*box) + type->num_shared_regs * sizeof(struct intel_uncore_extra_reg);
+	size = sizeof(*box) + numshared * sizeof(struct intel_uncore_extra_reg);
 
 	box = kzalloc_node(size, GFP_KERNEL, node);
 	if (!box)
 		return NULL;
 
-	for (i = 0; i < type->num_shared_regs; i++)
+	for (i = 0; i < numshared; i++)
 		raw_spin_lock_init(&box->shared_regs[i].lock);
 
 	uncore_pmu_init_hrtimer(box);

@@ -351,7 +353,8 @@ static bool is_uncore_event(struct perf_
 }
 
 static int
-uncore_collect_events(struct intel_uncore_box *box, struct perf_event *leader, bool dogrp)
+uncore_collect_events(struct intel_uncore_box *box, struct perf_event *leader,
+		      bool dogrp)
 {
 	struct perf_event *event;
 	int n, max_count;

@@ -412,7 +415,8 @@ uncore_get_event_constraint(struct intel
 	return &type->unconstrainted;
 }
 
-static void uncore_put_event_constraint(struct intel_uncore_box *box, struct perf_event *event)
+static void uncore_put_event_constraint(struct intel_uncore_box *box,
+					struct perf_event *event)
 {
 	if (box->pmu->type->ops->put_constraint)
 		box->pmu->type->ops->put_constraint(box, event);

@@ -592,7 +596,7 @@ static void uncore_pmu_event_del(struct
 		if (event == box->event_list[i]) {
 			uncore_put_event_constraint(box, event);
 
-			while (++i < box->n_events)
+			for (++i; i < box->n_events; i++)
 				box->event_list[i - 1] = box->event_list[i];
 
 			--box->n_events;

@@ -801,10 +805,8 @@ static void __init uncore_type_exit(stru
 
 static void __init uncore_types_exit(struct intel_uncore_type **types)
 {
-	int i;
-
-	for (i = 0; types[i]; i++)
-		uncore_type_exit(types[i]);
+	for (; *types; types++)
+		uncore_type_exit(*types);
 }
 
 static int __init uncore_type_init(struct intel_uncore_type *type)

@@ -908,9 +910,11 @@ static int uncore_pci_probe(struct pci_d
 	 * some device types. Hence PCI device idx would be 0 for all devices.
 	 * So increment pmu pointer to point to an unused array element.
 	 */
-	if (boot_cpu_data.x86_model == 87)
+	if (boot_cpu_data.x86_model == 87) {
 		while (pmu->func_id >= 0)
 			pmu++;
+	}
+
 	if (pmu->func_id < 0)
 		pmu->func_id = pdev->devfn;
 	else

@@ -1170,44 +1174,45 @@ static int uncore_cpu_prepare(int cpu, i
 	return 0;
 }
 
-static void
-uncore_change_context(struct intel_uncore_type **uncores, int old_cpu, int new_cpu)
+static void uncore_change_type_ctx(struct intel_uncore_type *type, int old_cpu,
+				   int new_cpu)
 {
-	struct intel_uncore_type *type;
-	struct intel_uncore_pmu *pmu;
+	struct intel_uncore_pmu *pmu = type->pmus;
 	struct intel_uncore_box *box;
-	int i, j;
-
-	for (i = 0; uncores[i]; i++) {
-		type = uncores[i];
-		for (j = 0; j < type->num_boxes; j++) {
-			pmu = &type->pmus[j];
-			if (old_cpu < 0)
-				box = uncore_pmu_to_box(pmu, new_cpu);
-			else
-				box = uncore_pmu_to_box(pmu, old_cpu);
-			if (!box)
-				continue;
+	int i;
 
-			if (old_cpu < 0) {
-				WARN_ON_ONCE(box->cpu != -1);
-				box->cpu = new_cpu;
-				continue;
-			}
+	for (i = 0; i < type->num_boxes; i++, pmu++) {
+		if (old_cpu < 0)
+			box = uncore_pmu_to_box(pmu, new_cpu);
+		else
+			box = uncore_pmu_to_box(pmu, old_cpu);
+		if (!box)
+			continue;
 
-			WARN_ON_ONCE(box->cpu != old_cpu);
-			if (new_cpu >= 0) {
-				uncore_pmu_cancel_hrtimer(box);
-				perf_pmu_migrate_context(&pmu->pmu,
-						old_cpu, new_cpu);
-				box->cpu = new_cpu;
-			} else {
-				box->cpu = -1;
-			}
+		if (old_cpu < 0) {
+			WARN_ON_ONCE(box->cpu != -1);
+			box->cpu = new_cpu;
+			continue;
 		}
+
+		WARN_ON_ONCE(box->cpu != old_cpu);
+		box->cpu = -1;
+		if (new_cpu < 0)
+			continue;
+
+		uncore_pmu_cancel_hrtimer(box);
+		perf_pmu_migrate_context(&pmu->pmu, old_cpu, new_cpu);
+		box->cpu = new_cpu;
 	}
 }
 
+static void uncore_change_context(struct intel_uncore_type **uncores,
+				  int old_cpu, int new_cpu)
+{
+	for (; *uncores; uncores++)
+		uncore_change_type_ctx(*uncores, old_cpu, new_cpu);
+}
+
 static void uncore_event_exit_cpu(int cpu)
 {
 	int i, phys_id, target;

@@ -1313,8 +1318,8 @@ static int __init uncore_msr_pmus_regist
 	struct intel_uncore_type **types = uncore_msr_uncores;
 	int ret;
 
-	while (*types) {
-		ret = type_pmu_register(*types++);
+	for (; *types; types++) {
+		ret = type_pmu_register(*types);
 		if (ret)
 			return ret;
 	}

Index: b/arch/x86/kernel/cpu/perf_event_intel_uncore.c
===================================================================
@@ -21,7 +21,7 @@ static struct event_constraint uncore_co
 struct event_constraint uncore_constraint_empty =
 	EVENT_CONSTRAINT(0, 0, 0);
 
-int uncore_pcibus_to_physid(struct pci_bus *bus)
+static int uncore_pcibus_to_physid(struct pci_bus *bus)
 {
 	struct pci2phy_map *map;
 	int phys_id = -1;

Index: b/arch/x86/kernel/cpu/perf_event_intel_uncore.h
===================================================================
@@ -125,7 +125,6 @@ struct pci2phy_map {
 	int pbus_to_physid[256];
 };
 
-int uncore_pcibus_to_physid(struct pci_bus *bus);
 struct pci2phy_map *__find_pci2phy_map(int segment);
 
 ssize_t uncore_event_show(struct kobject *kobj,

@@ -467,6 +467,7 @@ struct perf_event {
 	int				group_flags;
 	struct perf_event		*group_leader;
 	struct pmu			*pmu;
+	void				*pmu_private;
 
 	enum perf_event_active_state	state;
 	unsigned int			attach_state;

@@ -92,11 +92,6 @@ ssize_t uncore_event_show(struct kobject
 	return sprintf(buf, "%s", event->config);
 }
 
-struct intel_uncore_pmu *uncore_event_to_pmu(struct perf_event *event)
-{
-	return container_of(event->pmu, struct intel_uncore_pmu, pmu);
-}
-
 struct intel_uncore_box *uncore_pmu_to_box(struct intel_uncore_pmu *pmu, int cpu)
 {
 	struct intel_uncore_box *box;

@@ -122,15 +117,6 @@ struct intel_uncore_box *uncore_pmu_to_b
 	return *per_cpu_ptr(pmu->box, cpu);
 }
 
-struct intel_uncore_box *uncore_event_to_box(struct perf_event *event)
-{
-	/*
-	 * perf core schedules event on the basis of cpu, uncore events are
-	 * collected by one of the cpus inside a physical package.
-	 */
-	return uncore_pmu_to_box(uncore_event_to_pmu(event), smp_processor_id());
-}
-
 u64 uncore_msr_read_counter(struct intel_uncore_box *box, struct perf_event *event)
 {
 	u64 count;

@@ -690,6 +676,7 @@ static int uncore_pmu_event_init(struct
 	if (!box || box->cpu < 0)
 		return -EINVAL;
 	event->cpu = box->cpu;
+	event->pmu_private = box;
 
 	event->hw.idx = -1;
 	event->hw.last_tag = ~0ULL;

@@ -319,9 +319,17 @@ static inline bool uncore_box_is_fake(st
 	return (box->phys_id < 0);
 }
 
-struct intel_uncore_pmu *uncore_event_to_pmu(struct perf_event *event);
+static inline struct intel_uncore_pmu *uncore_event_to_pmu(struct perf_event *event)
+{
+	return container_of(event->pmu, struct intel_uncore_pmu, pmu);
+}
+
+static inline struct intel_uncore_box *uncore_event_to_box(struct perf_event *event)
+{
+	return event->pmu_private;
+}
+
 struct intel_uncore_box *uncore_pmu_to_box(struct intel_uncore_pmu *pmu, int cpu);
-struct intel_uncore_box *uncore_event_to_box(struct perf_event *event);
 u64 uncore_msr_read_counter(struct intel_uncore_box *box, struct perf_event *event);
 void uncore_pmu_start_hrtimer(struct intel_uncore_box *box);
 void uncore_pmu_cancel_hrtimer(struct intel_uncore_box *box);

@@ -313,6 +313,7 @@ static int snb_uncore_imc_event_init(str
 		return -EINVAL;
 
 	event->cpu = box->cpu;
+	event->pmu_private = box;
 
 	event->hw.idx = -1;
 	event->hw.last_tag = ~0ULL;

@@ -129,6 +129,8 @@ struct cpuinfo_x86 {
 	u16			booted_cores;
 	/* Physical processor id: */
 	u16			phys_proc_id;
+	/* Logical processor id: */
+	u16			logical_proc_id;
 	/* Core id: */
 	u16			cpu_core_id;
 	/* Compute unit id */

@@ -119,12 +119,22 @@ static inline void setup_node_to_cpumask
 
 extern const struct cpumask *cpu_coregroup_mask(int cpu);
 
+#define topology_logical_package_id(cpu)	(cpu_data(cpu).logical_proc_id)
 #define topology_physical_package_id(cpu)	(cpu_data(cpu).phys_proc_id)
 #define topology_core_id(cpu)			(cpu_data(cpu).cpu_core_id)
 
 #ifdef ENABLE_TOPO_DEFINES
 #define topology_core_cpumask(cpu)		(per_cpu(cpu_core_map, cpu))
 #define topology_sibling_cpumask(cpu)		(per_cpu(cpu_sibling_map, cpu))
+
+extern unsigned int __max_logical_packages;
+#define topology_max_packages()			(__max_logical_packages)
+int topology_update_package_map(unsigned int apicid, unsigned int cpu);
+extern int topology_phys_to_logical_pkg(unsigned int pkg);
+#else
+#define topology_max_packages()			(1)
+static inline int
+topology_update_package_map(unsigned int apicid, unsigned int cpu) { return 0; }
 #endif
 
 static inline void arch_fix_phys_package_id(int num, u32 slot)

@@ -2078,6 +2078,20 @@ int generic_processor_info(int apicid, i
 		cpu = cpumask_next_zero(-1, cpu_present_mask);
 
 	/*
+	 * This can happen on physical hotplug. The sanity check at boot time
+	 * is done from native_smp_prepare_cpus() after num_possible_cpus() is
+	 * established.
+	 */
+	if (topology_update_package_map(apicid, cpu) < 0) {
+		int thiscpu = max + disabled_cpus;
+
+		pr_warning("ACPI: Package limit reached. Processor %d/0x%x ignored.\n",
+			   thiscpu, apicid);
+		disabled_cpus++;
+		return -ENOSPC;
+	}
+
+	/*
 	 * Validate version
 	 */
 	if (version == 0x0) {

@@ -977,6 +977,8 @@ static void identify_cpu(struct cpuinfo_
 #ifdef CONFIG_NUMA
 	numa_add_cpu(smp_processor_id());
 #endif
+	/* The boot/hotplug time assigment got cleared, restore it */
+	c->logical_proc_id = topology_phys_to_logical_pkg(c->phys_proc_id);
 }
 
 /*

@@ -160,6 +160,19 @@ static void early_init_intel(struct cpui
 		pr_info("Disabling PGE capability bit\n");
 		setup_clear_cpu_cap(X86_FEATURE_PGE);
 	}
+
+	if (c->cpuid_level >= 0x00000001) {
+		u32 eax, ebx, ecx, edx;
+
+		cpuid(0x00000001, &eax, &ebx, &ecx, &edx);
+		/*
+		 * If HTT (EDX[28]) is set EBX[16:23] contain the number of
+		 * apicids which are reserved per package. Store the resulting
+		 * shift value for the package management code.
+		 */
+		if (edx & (1U << 28))
+			c->x86_coreid_bits = get_count_order((ebx >> 16) & 0xff);
+	}
 }
 
 #ifdef CONFIG_X86_32

118 lines not on this page · raw message

@@ -9,9 +9,9 @@ struct pci_driver *uncore_pci_driver;
 /* pci bus to socket mapping */
 DEFINE_RAW_SPINLOCK(pci2phy_map_lock);
 struct list_head pci2phy_map_head = LIST_HEAD_INIT(pci2phy_map_head);
-struct pci_dev *uncore_extra_pci_dev[UNCORE_SOCKET_MAX][UNCORE_EXTRA_PCI_DEV_MAX];
+struct pci_extra_dev *uncore_extra_pci_dev;
+static int max_packages;
 
-static DEFINE_RAW_SPINLOCK(uncore_box_lock);
 /* mask of cpus that collect uncore events */
 static cpumask_t uncore_cpu_mask;
 

@@ -94,27 +94,7 @@ ssize_t uncore_event_show(struct kobject
 
 struct intel_uncore_box *uncore_pmu_to_box(struct intel_uncore_pmu *pmu, int cpu)
 {
-	struct intel_uncore_box *box;
-
-	box = *per_cpu_ptr(pmu->box, cpu);
-	if (box)
-		return box;
-
-	raw_spin_lock(&uncore_box_lock);
-	/* Recheck in lock to handle races. */
-	if (*per_cpu_ptr(pmu->box, cpu))
-		goto out;
-	list_for_each_entry(box, &pmu->box_list, list) {
-		if (box->phys_id == topology_physical_package_id(cpu)) {
-			atomic_inc(&box->refcnt);
-			*per_cpu_ptr(pmu->box, cpu) = box;
-			break;
-		}
-	}
-out:
-	raw_spin_unlock(&uncore_box_lock);
-
-	return *per_cpu_ptr(pmu->box, cpu);
+	return pmu->boxes[topology_logical_package_id(cpu)];
 }
 
 u64 uncore_msr_read_counter(struct intel_uncore_box *box, struct perf_event *event)

@@ -315,9 +295,9 @@ static struct intel_uncore_box *uncore_a
 		raw_spin_lock_init(&box->shared_regs[i].lock);
 
 	uncore_pmu_init_hrtimer(box);
-	atomic_set(&box->refcnt, 1);
 	box->cpu = -1;
-	box->phys_id = -1;
+	box->pci_phys_id = -1;
+	box->pkgid = -1;
 
 	/* set default hrtimer timeout */
 	box->hrtimer_duration = UNCORE_PMU_HRTIMER_INTERVAL;

@@ -774,14 +754,24 @@ static void uncore_pmu_unregister(struct
 	pmu->registered = false;
 }
 
+static void uncore_free_boxes(struct intel_uncore_pmu *pmu)
+{
+	int pkg;
+
+	for (pkg = 0; pkg < max_packages; pkg++)
+		kfree(pmu->boxes[pkg]);
+	kfree(pmu->boxes);
+}
+
 static void __init uncore_type_exit(struct intel_uncore_type *type)
 {
+	struct intel_uncore_pmu *pmu = type->pmus;
 	int i;
 
-	if (type->pmus) {
-		for (i = 0; i < type->num_boxes; i++) {
-			uncore_pmu_unregister(&type->pmus[i]);
-			free_percpu(type->pmus[i].box);
+	if (pmu) {
+		for (i = 0; i < type->num_boxes; i++, pmu++) {
+			uncore_pmu_unregister(pmu);
+			uncore_free_boxes(pmu);
 		}
 		kfree(type->pmus);
 		type->pmus = NULL;

@@ -796,37 +786,36 @@ static void __init uncore_types_exit(str
 		uncore_type_exit(*types);
 }
 
-static int __init uncore_type_init(struct intel_uncore_type *type)
+static int __init uncore_type_init(struct intel_uncore_type *type, bool setid)
 {
 	struct intel_uncore_pmu *pmus;
 	struct attribute_group *attr_group;
 	struct attribute **attrs;
+	size_t size;
 	int i, j;
 
 	pmus = kzalloc(sizeof(*pmus) * type->num_boxes, GFP_KERNEL);
 	if (!pmus)
 		return -ENOMEM;
 
-	type->pmus = pmus;
-
-	type->unconstrainted = (struct event_constraint)
-		__EVENT_CONSTRAINT(0, (1ULL << type->num_counters) - 1,
-				0, type->num_counters, 0, 0);
+	size = max_packages * sizeof(struct intel_uncore_box *);
 
 	for (i = 0; i < type->num_boxes; i++) {
-		pmus[i].func_id = -1;
-		pmus[i].pmu_idx = i;
-		pmus[i].type = type;
-		INIT_LIST_HEAD(&pmus[i].box_list);
-		pmus[i].box = alloc_percpu(struct intel_uncore_box *);
-		if (!pmus[i].box)
+		pmus[i].func_id	= setid ? i : -1;
+		pmus[i].pmu_idx	= i;
+		pmus[i].type	= type;
+		pmus[i].boxes	= kzalloc(size, GFP_KERNEL);
+		if (!pmus[i].boxes)
 			return -ENOMEM;
 	}
 
+	type->pmus = pmus;
+	type->unconstrainted = (struct event_constraint)
+		__EVENT_CONSTRAINT(0, (1ULL << type->num_counters) - 1,
+				0, type->num_counters, 0, 0);
+
 	if (type->event_descs) {
-		i = 0;
-		while (type->event_descs[i].attr.attr.name)
-			i++;
+		for (i = 0; type->event_descs[i].attr.attr.name; i++);
 
 		attr_group = kzalloc(sizeof(struct attribute *) * (i + 1) +
 					sizeof(*attr_group), GFP_KERNEL);

@@ -847,12 +836,13 @@ static int __init uncore_type_init(struc
 	return 0;
 }
 
-static int __init uncore_types_init(struct intel_uncore_type **types)
+static int __init
+uncore_types_init(struct intel_uncore_type **types, bool setid)
 {
-	int i, ret;
+	int ret;
 
-	for (i = 0; types[i]; i++) {
-		ret = uncore_type_init(types[i]);
+	for (; *types; types++) {
+		ret = uncore_type_init(*types, setid);
 		if (ret)
 			return ret;
 	}

@@ -864,28 +854,28 @@ static int __init uncore_types_init(stru
  */
 static int uncore_pci_probe(struct pci_dev *pdev, const struct pci_device_id *id)
 {
+	struct intel_uncore_type *type;
 	struct intel_uncore_pmu *pmu;
 	struct intel_uncore_box *box;
-	struct intel_uncore_type *type;
-	bool first_box = false;
-	int phys_id, ret;
+	int phys_id, pkg, ret;
 
 	phys_id = uncore_pcibus_to_physid(pdev->bus);
-	if (phys_id < 0 || phys_id >= UNCORE_SOCKET_MAX)
+	if (phys_id < 0)
 		return -ENODEV;
 
+	pkg = topology_phys_to_logical_pkg(phys_id);
+	if (WARN_ON_ONCE(pkg < 0))
+		return -EINVAL;
+
 	if (UNCORE_PCI_DEV_TYPE(id->driver_data) == UNCORE_EXTRA_PCI_DEV) {
 		int idx = UNCORE_PCI_DEV_IDX(id->driver_data);
-		uncore_extra_pci_dev[phys_id][idx] = pdev;
+
+		uncore_extra_pci_dev[pkg].dev[idx] = pdev;
 		pci_set_drvdata(pdev, NULL);
 		return 0;
 	}
 
 	type = uncore_pci_uncores[UNCORE_PCI_DEV_TYPE(id->driver_data)];
-	box = uncore_alloc_box(type, NUMA_NO_NODE);
-	if (!box)
-		return -ENOMEM;
-
 	/*
 	 * for performance monitoring unit with multiple boxes,
 	 * each box has a different function id.

@@ -902,32 +892,35 @@ static int uncore_pci_probe(struct pci_d
 			pmu++;
 	}
 
+	if (WARN_ON_ONCE(pmu->boxes[pkg] != NULL))
+		return -EINVAL;
+
+	box = uncore_alloc_box(type, NUMA_NO_NODE);
+	if (!box)
+		return -ENOMEM;
+
 	if (pmu->func_id < 0)
 		pmu->func_id = pdev->devfn;
 	else
 		WARN_ON_ONCE(pmu->func_id != pdev->devfn);
 
-	box->phys_id = phys_id;
+	atomic_inc(&box->refcnt);
+	box->pci_phys_id = phys_id;
+	box->pkgid = pkg;
 	box->pci_dev = pdev;
 	box->pmu = pmu;
 	uncore_box_init(box);
 	pci_set_drvdata(pdev, box);
 
-	raw_spin_lock(&uncore_box_lock);
-	if (list_empty(&pmu->box_list))
-		first_box = true;
-	list_add_tail(&box->list, &pmu->box_list);
-	raw_spin_unlock(&uncore_box_lock);
-
-	if (!first_box)
+	pmu->boxes[pkg] = box;
+	if (atomic_inc_return(&pmu->activeboxes) > 1)
 		return 0;
 
+	/* First active box registers the pmu */
 	ret = uncore_pmu_register(pmu);
 	if (ret) {
 		pci_set_drvdata(pdev, NULL);
-		raw_spin_lock(&uncore_box_lock);
-		list_del(&box->list);
-		raw_spin_unlock(&uncore_box_lock);
+		pmu->boxes[pkg] = NULL;
 		uncore_box_exit(box);
 		kfree(box);
 	}

@@ -938,15 +931,16 @@ static void uncore_pci_remove(struct pci
 {
 	struct intel_uncore_box *box = pci_get_drvdata(pdev);
 	struct intel_uncore_pmu *pmu;
-	int i, cpu, phys_id;
-	bool last_box = false;
+	int i, phys_id, pkg;
 
 	phys_id = uncore_pcibus_to_physid(pdev->bus);
+	pkg = topology_phys_to_logical_pkg(phys_id);
+
 	box = pci_get_drvdata(pdev);
 	if (!box) {
 		for (i = 0; i < UNCORE_EXTRA_PCI_DEV_MAX; i++) {
-			if (uncore_extra_pci_dev[phys_id][i] == pdev) {
-				uncore_extra_pci_dev[phys_id][i] = NULL;
+			if (uncore_extra_pci_dev[pkg].dev[i] == pdev) {
+				uncore_extra_pci_dev[pkg].dev[i] = NULL;
 				break;
 			}
 		}

@@ -955,34 +949,20 @@ static void uncore_pci_remove(struct pci
 	}
 
 	pmu = box->pmu;
-	if (WARN_ON_ONCE(phys_id != box->phys_id))
+	if (WARN_ON_ONCE(phys_id != box->pci_phys_id))
 		return;
 
 	pci_set_drvdata(pdev, NULL);
-
-	raw_spin_lock(&uncore_box_lock);
-	list_del(&box->list);
-	if (list_empty(&pmu->box_list))
-		last_box = true;
-	raw_spin_unlock(&uncore_box_lock);
-
-	for_each_possible_cpu(cpu) {
-		if (*per_cpu_ptr(pmu->box, cpu) == box) {
-			*per_cpu_ptr(pmu->box, cpu) = NULL;
-			atomic_dec(&box->refcnt);
-		}
-	}
-
-	WARN_ON_ONCE(atomic_read(&box->refcnt) != 1);
+	pmu->boxes[pkg] = NULL;
+	if (atomic_dec_return(&pmu->activeboxes) == 0)
+		uncore_pmu_unregister(pmu);
 	uncore_box_exit(box);
 	kfree(box);
-
-	if (last_box)
-		uncore_pmu_unregister(pmu);
 }
 
 static int __init uncore_pci_init(void)
 {
+	size_t size;
 	int ret;
 
 	switch (boot_cpu_data.x86_model) {

@@ -1025,24 +1005,34 @@ static int __init uncore_pci_init(void)
 	if (ret)
 		return ret;
 
-	ret = uncore_types_init(uncore_pci_uncores);
-	if (ret)
+	size = max_packages * sizeof(struct pci_extra_dev);
+	uncore_extra_pci_dev = kzalloc(size, GFP_KERNEL);
+	if (!uncore_extra_pci_dev) {
+		ret = -ENOMEM;
 		goto err;
+	}
+
+	ret = uncore_types_init(uncore_pci_uncores, false);
+	if (ret)
+		goto errtype;
 
 	uncore_pci_driver->probe = uncore_pci_probe;
 	uncore_pci_driver->remove = uncore_pci_remove;
 
 	ret = pci_register_driver(uncore_pci_driver);
 	if (ret)
-		goto err;
+		goto errtype;
 
 	pcidrv_registered = true;
 	return 0;
 
-err:
+errtype:
 	uncore_types_exit(uncore_pci_uncores);
-	uncore_pci_uncores = empty_uncore;
+	kfree(uncore_extra_pci_dev);
+	uncore_extra_pci_dev = NULL;
 	uncore_free_pcibus_map();
+err:
+	uncore_pci_uncores = empty_uncore;
 	return ret;
 }
 

@@ -1052,110 +1042,81 @@ static void __init uncore_pci_exit(void)
 		pcidrv_registered = false;
 		pci_unregister_driver(uncore_pci_driver);
 		uncore_types_exit(uncore_pci_uncores);
+		kfree(uncore_extra_pci_dev);
 		uncore_free_pcibus_map();
 	}
 }
 
-/* CPU hot plug/unplug are serialized by cpu_add_remove_lock mutex */
-static LIST_HEAD(boxes_to_free);
-
-static void uncore_kfree_boxes(void)
-{
-	struct intel_uncore_box *box;
-
-	while (!list_empty(&boxes_to_free)) {
-		box = list_entry(boxes_to_free.next,
-				 struct intel_uncore_box, list);
-		list_del(&box->list);
-		kfree(box);
-	}
-}
-
 static void uncore_cpu_dying(int cpu)
 {
-	struct intel_uncore_type *type;
+	struct intel_uncore_type *type, **types = uncore_msr_uncores;
 	struct intel_uncore_pmu *pmu;
 	struct intel_uncore_box *box;
-	int i, j;
+	int i, pkg;
 
-	for (i = 0; uncore_msr_uncores[i]; i++) {
-		type = uncore_msr_uncores[i];
-		for (j = 0; j < type->num_boxes; j++) {
-			pmu = &type->pmus[j];
-			box = *per_cpu_ptr(pmu->box, cpu);
-			*per_cpu_ptr(pmu->box, cpu) = NULL;
-			if (box && atomic_dec_and_test(&box->refcnt)) {
-				list_add(&box->list, &boxes_to_free);
+	pkg = topology_logical_package_id(cpu);
+	for (; *types; types++) {
+		type = *types;
+		pmu = type->pmus;
+		for (i = 0; i < type->num_boxes; i++, pmu++) {
+			box = pmu->boxes[pkg];
+			if (box && atomic_dec_return(&box->refcnt) == 0)
 				uncore_box_exit(box);
-			}
 		}
 	}
 }
 
-static int uncore_cpu_starting(int cpu)
+static void uncore_cpu_starting(int cpu, bool init)
 {
-	struct intel_uncore_type *type;
+	struct intel_uncore_type *type, **types = uncore_msr_uncores;
 	struct intel_uncore_pmu *pmu;
-	struct intel_uncore_box *box, *exist;
-	int i, j, k, phys_id;
+	struct intel_uncore_box *box;
+	int i, pkg, ncpus = 1;
 
-	phys_id = topology_physical_package_id(cpu);
+	if (init) {
+		/*
+		 * On init we get the number of online cpus in the package
+		 * and set refcount for all of them.
+		 */
+		ncpus = cpumask_weight(topology_core_cpumask(cpu));
+	}
 
-	for (i = 0; uncore_msr_uncores[i]; i++) {
-		type = uncore_msr_uncores[i];
-		for (j = 0; j < type->num_boxes; j++) {
-			pmu = &type->pmus[j];
-			box = *per_cpu_ptr(pmu->box, cpu);
-			/* called by uncore_cpu_init? */
-			if (box && box->phys_id >= 0) {
-				uncore_box_init(box);
+	pkg = topology_logical_package_id(cpu);
+	for (; *types; types++) {
+		type = *types;
+		pmu = type->pmus;
+		for (i = 0; i < type->num_boxes; i++, pmu++) {
+			box = pmu->boxes[pkg];
+			if (!box)
 				continue;
-			}
-
-			for_each_online_cpu(k) {
-				exist = *per_cpu_ptr(pmu->box, k);
-				if (exist && exist->phys_id == phys_id) {
-					atomic_inc(&exist->refcnt);
-					*per_cpu_ptr(pmu->box, cpu) = exist;
-					if (box) {
-						list_add(&box->list,
-							 &boxes_to_free);
-						box = NULL;
-					}
-					break;
-				}
-			}
-
-			if (box) {
-				box->phys_id = phys_id;
+			/* The first cpu on a package activates the box */
+			if (atomic_add_return(ncpus, &box->refcnt) == ncpus)
 				uncore_box_init(box);
-			}
 		}
 	}
-	return 0;
 }
 
-static int uncore_cpu_prepare(int cpu, int phys_id)
+static int uncore_cpu_prepare(int cpu)
 {
-	struct intel_uncore_type *type;
+	struct intel_uncore_type *type, **types = uncore_msr_uncores;
 	struct intel_uncore_pmu *pmu;
 	struct intel_uncore_box *box;
-	int i, j;
-
-	for (i = 0; uncore_msr_uncores[i]; i++) {
-		type = uncore_msr_uncores[i];
-		for (j = 0; j < type->num_boxes; j++) {
-			pmu = &type->pmus[j];
-			if (pmu->func_id < 0)
-				pmu->func_id = j;
+	int i, pkg;
 
+	pkg = topology_logical_package_id(cpu);
+	for (; *types; types++) {
+		type = *types;
+		pmu = type->pmus;
+		for (i = 0; i < type->num_boxes; i++, pmu++) {
+			if (pmu->boxes[pkg])
+				continue;
+			/* First cpu of a package allocates the box */
 			box = uncore_alloc_box(type, cpu_to_node(cpu));
 			if (!box)
 				return -ENOMEM;
-
 			box->pmu = pmu;
-			box->phys_id = phys_id;
-			*per_cpu_ptr(pmu->box, cpu) = box;
+			box->pkgid = pkg;
+			pmu->boxes[pkg] = box;
 		}
 	}
 	return 0;

@@ -1166,13 +1127,11 @@ static void uncore_change_type_ctx(struc
 {
 	struct intel_uncore_pmu *pmu = type->pmus;
 	struct intel_uncore_box *box;
-	int i;
+	int i, pkg;
 
+	pkg = topology_logical_package_id(old_cpu < 0 ? new_cpu : old_cpu);
 	for (i = 0; i < type->num_boxes; i++, pmu++) {
-		if (old_cpu < 0)
-			box = uncore_pmu_to_box(pmu, new_cpu);
-		else
-			box = uncore_pmu_to_box(pmu, old_cpu);
+		box = pmu->boxes[pkg];
 		if (!box)
 			continue;
 

@@ -1202,27 +1161,20 @@ static void uncore_change_context(struct
 
 static void uncore_event_exit_cpu(int cpu)
 {
-	int i, phys_id, target;
+	int target;
 
-	/* if exiting cpu is used for collecting uncore events */
+	/* Check if exiting cpu is used for collecting uncore events */
 	if (!cpumask_test_and_clear_cpu(cpu, &uncore_cpu_mask))
 		return;
 
-	/* find a new cpu to collect uncore events */
-	phys_id = topology_physical_package_id(cpu);
-	target = -1;
-	for_each_online_cpu(i) {
-		if (i == cpu)
-			continue;
-		if (phys_id == topology_physical_package_id(i)) {
-			target = i;
-			break;
-		}
-	}
+	/* Find a new cpu to collect uncore events */
+	target = cpumask_any_but(topology_core_cpumask(cpu), cpu);
 
-	/* migrate uncore events to the new cpu */
-	if (target >= 0)
+	/* Migrate uncore events to the new target */
+	if (target < nr_cpu_ids)
 		cpumask_set_cpu(target, &uncore_cpu_mask);
+	else
+		target = -1;
 
 	uncore_change_context(uncore_msr_uncores, cpu, target);
 	uncore_change_context(uncore_pci_uncores, cpu, target);

@@ -1230,13 +1182,15 @@ static void uncore_event_exit_cpu(int cp
 
 static void uncore_event_init_cpu(int cpu)
 {
-	int i, phys_id;
+	int target;
 
-	phys_id = topology_physical_package_id(cpu);
-	for_each_cpu(i, &uncore_cpu_mask) {
-		if (phys_id == topology_physical_package_id(i))
-			return;
-	}
+	/*
+	 * Check if there is an online cpu in the package
+	 * which collects uncore events already.
+	 */
+	target = cpumask_any_and(&uncore_cpu_mask, topology_core_cpumask(cpu));
+	if (target < nr_cpu_ids)
+		return;
 
 	cpumask_set_cpu(cpu, &uncore_cpu_mask);
 

@@ -1249,38 +1203,25 @@ static int uncore_cpu_notifier(struct no
 {
 	unsigned int cpu = (long)hcpu;
 
-	/* allocate/free data structure for uncore box */
 	switch (action & ~CPU_TASKS_FROZEN) {
 	case CPU_UP_PREPARE:
-		return notifier_from_errno(uncore_cpu_prepare(cpu, -1));
+		return notifier_from_errno(uncore_cpu_prepare(cpu));
+
 	case CPU_STARTING:
-		uncore_cpu_starting(cpu);
+		uncore_cpu_starting(cpu, false);
+	case CPU_DOWN_FAILED:
+		uncore_event_init_cpu(cpu);
 		break;
+
 	case CPU_UP_CANCELED:
 	case CPU_DYING:
 		uncore_cpu_dying(cpu);
 		break;
-	case CPU_ONLINE:
-	case CPU_DEAD:
-		uncore_kfree_boxes();
-		break;
-	default:
-		break;
-	}
 
-	/* select the cpu that collects uncore events */
-	switch (action & ~CPU_TASKS_FROZEN) {
-	case CPU_DOWN_FAILED:
-	case CPU_STARTING:
-		uncore_event_init_cpu(cpu);
-		break;
 	case CPU_DOWN_PREPARE:
 		uncore_event_exit_cpu(cpu);
 		break;
-	default:
-		break;
 	}
-
 	return NOTIFY_OK;
 }
 

@@ -1357,7 +1298,7 @@ static int __init uncore_cpu_init(void)
 		return 0;
 	}
 
-	ret = uncore_types_init(uncore_msr_uncores);
+	ret = uncore_types_init(uncore_msr_uncores, true);
 	if (ret)
 		goto err;
 

@@ -1373,39 +1314,34 @@ static int __init uncore_cpu_init(void)
 
 static void __init uncore_cpu_setup(void *dummy)
 {
-	uncore_cpu_starting(smp_processor_id());
+	uncore_cpu_starting(smp_processor_id(), true);
 }
 
+/* Lazy to avoid allocation of a few bytes for the normal case */
+static __initdata DECLARE_BITMAP(packages, MAX_LOCAL_APIC);
+
 static int __init uncore_cpumask_init(void)
 {
-	int cpu, ret = 0;
-
-	cpu_notifier_register_begin();
+	unsigned int cpu;
 
 	for_each_online_cpu(cpu) {
-		int i, phys_id = topology_physical_package_id(cpu);
+		unsigned int pkg = topology_logical_package_id(cpu);
+		int ret;
 
-		for_each_cpu(i, &uncore_cpu_mask) {
-			if (phys_id == topology_physical_package_id(i)) {
-				phys_id = -1;
-				break;
-			}
-		}
-		if (phys_id < 0)
+		if (test_and_set_bit(pkg, packages))
 			continue;
-
-		ret = uncore_cpu_prepare(cpu, phys_id);
+		/*
+		 * The first online cpu of each package takes the refcounts
+		 * for all other online cpus in that package.
+		 */
+		ret = uncore_cpu_prepare(cpu);
 		if (ret)
-			goto out;
+			return ret;
 		uncore_event_init_cpu(cpu);
+		smp_call_function_single(cpu, uncore_cpu_setup, NULL, 1);
 	}
-	on_each_cpu(uncore_cpu_setup, NULL, 1);
-
 	__register_cpu_notifier(&uncore_cpu_nb);
-
-out:
-	cpu_notifier_register_done();
-	return ret;
+	return 0;
 }
 
 static int __init intel_uncore_init(void)

@@ -1418,22 +1354,26 @@ static int __init intel_uncore_init(void
 	if (cpu_has_hypervisor)
 		return -ENODEV;
 
+	max_packages = topology_max_packages();
+
 	ret = uncore_pci_init();
 	if (ret)
 		return ret;
 	ret = uncore_cpu_init();
 	if (ret)
-		goto errpci;
+		goto err;
+
+	cpu_notifier_register_begin();
 	ret = uncore_cpumask_init();
 	if (ret)
-		goto errcpu;
-
+		goto err;
+	cpu_notifier_register_done();
 	return 0;
 
-errcpu:
+err:
 	uncore_types_exit(uncore_msr_uncores);
-errpci:
 	uncore_pci_exit();
+	cpu_notifier_register_done();
 	return ret;
 }
 device_initcall(intel_uncore_init);

@@ -19,11 +19,12 @@
 #define UNCORE_EXTRA_PCI_DEV		0xff
 #define UNCORE_EXTRA_PCI_DEV_MAX	3
 
-/* support up to 8 sockets */
-#define UNCORE_SOCKET_MAX		8
-
 #define UNCORE_EVENT_CONSTRAINT(c, n) EVENT_CONSTRAINT(c, n, 0xff)
 
+struct pci_extra_dev {
+	struct pci_dev *dev[UNCORE_EXTRA_PCI_DEV_MAX];
+};
+
 struct intel_uncore_ops;
 struct intel_uncore_pmu;
 struct intel_uncore_box;

@@ -79,9 +80,9 @@ struct intel_uncore_pmu {
 	int				pmu_idx;
 	int				func_id;
 	bool				registered;
+	atomic_t			activeboxes;
 	struct intel_uncore_type	*type;
-	struct intel_uncore_box		** __percpu box;
-	struct list_head		box_list;
+	struct intel_uncore_box		**boxes;
 };
 
 struct intel_uncore_extra_reg {

@@ -91,7 +92,8 @@ struct intel_uncore_extra_reg {
 };
 
 struct intel_uncore_box {
-	int phys_id;
+	int pci_phys_id;
+	int pkgid;
 	int n_active;	/* number of active events */
 	int n_events;
 	int cpu;	/* cpu to collect events */

@@ -316,7 +318,7 @@ static inline void uncore_box_exit(struc
 
 static inline bool uncore_box_is_fake(struct intel_uncore_box *box)
 {
-	return (box->phys_id < 0);
+	return (box->pkgid < 0);
 }
 
 static inline struct intel_uncore_pmu *uncore_event_to_pmu(struct perf_event *event)

@@ -345,7 +347,7 @@ extern struct intel_uncore_type **uncore
 extern struct pci_driver *uncore_pci_driver;
 extern raw_spinlock_t pci2phy_map_lock;
 extern struct list_head pci2phy_map_head;
-extern struct pci_dev *uncore_extra_pci_dev[UNCORE_SOCKET_MAX][UNCORE_EXTRA_PCI_DEV_MAX];
+extern struct pci_extra_dev *uncore_extra_pci_dev;
 extern struct event_constraint uncore_constraint_empty;
 
 /* perf_event_intel_uncore_snb.c */

@@ -987,7 +987,9 @@ static void snbep_qpi_enable_event(struc
 
 	if (reg1->idx != EXTRA_REG_NONE) {
 		int idx = box->pmu->pmu_idx + SNBEP_PCI_QPI_PORT0_FILTER;
-		struct pci_dev *filter_pdev = uncore_extra_pci_dev[box->phys_id][idx];
+		int pkg = topology_phys_to_logical_pkg(box->pci_phys_id);
+		struct pci_dev *filter_pdev = uncore_extra_pci_dev[pkg].dev[idx];
+
 		if (filter_pdev) {
 			pci_write_config_dword(filter_pdev, reg1->reg,
 						(u32)reg1->config);

@@ -2521,14 +2523,16 @@ static struct intel_uncore_type *hswep_m
 
 void hswep_uncore_cpu_init(void)
 {
+	int pkg = topology_phys_to_logical_pkg(0);
+
 	if (hswep_uncore_cbox.num_boxes > boot_cpu_data.x86_max_cores)
 		hswep_uncore_cbox.num_boxes = boot_cpu_data.x86_max_cores;
 
 	/* Detect 6-8 core systems with only two SBOXes */
-	if (uncore_extra_pci_dev[0][HSWEP_PCI_PCU_3]) {
+	if (uncore_extra_pci_dev[pkg].dev[HSWEP_PCI_PCU_3]) {
 		u32 capid4;
 
-		pci_read_config_dword(uncore_extra_pci_dev[0][HSWEP_PCI_PCU_3],
+		pci_read_config_dword(uncore_extra_pci_dev[pkg].dev[HSWEP_PCI_PCU_3],
 				      0x94, &capid4);
 		if (((capid4 >> 6) & 0x3) == 0)
 			hswep_uncore_sbox.num_boxes = 2;