8 files changed, 627 insertions, 105 deletions

diff --git a/arch/x86/events/amd/uncore.c b/arch/x86/events/amd/uncore.c
index a6ea07f2aa84..46018e515fe2 100644
--- a/arch/x86/events/amd/uncore.c
+++ b/arch/x86/events/amd/uncore.c
@@ -180,6 +180,31 @@ static void amd_uncore_del(struct perf_event *event, int flags)
 	hwc->idx = -1;
 }
 
+/*
+ * Convert logical CPU number to L3 PMC Config ThreadMask format
+ */
+static u64 l3_thread_slice_mask(int cpu)
+{
+	u64 thread_mask, core = topology_core_id(cpu);
+	unsigned int shift, thread = 0;
+
+	if (topology_smt_supported() && !topology_is_primary_thread(cpu))
+		thread = 1;
+
+	if (boot_cpu_data.x86 <= 0x18) {
+		shift = AMD64_L3_THREAD_SHIFT + 2 * (core % 4) + thread;
+		thread_mask = BIT_ULL(shift);
+
+		return AMD64_L3_SLICE_MASK | thread_mask;
+	}
+
+	core = (core << AMD64_L3_COREID_SHIFT) & AMD64_L3_COREID_MASK;
+	shift = AMD64_L3_THREAD_SHIFT + thread;
+	thread_mask = BIT_ULL(shift);
+
+	return AMD64_L3_EN_ALL_SLICES | core | thread_mask;
+}
+
 static int amd_uncore_event_init(struct perf_event *event)
 {
 	struct amd_uncore *uncore;
@@ -206,18 +231,11 @@ static int amd_uncore_event_init(struct perf_event *event)
 		return -EINVAL;
 
 	/*
-	 * SliceMask and ThreadMask need to be set for certain L3 events in
-	 * Family 17h. For other events, the two fields do not affect the count.
+	 * SliceMask and ThreadMask need to be set for certain L3 events.
+	 * For other events, the two fields do not affect the count.
 	 */
-	if (l3_mask && is_llc_event(event)) {
-		int thread = 2 * (cpu_data(event->cpu).cpu_core_id % 4);
-
-		if (smp_num_siblings > 1)
-			thread += cpu_data(event->cpu).apicid & 1;
-
-		hwc->config |= (1ULL << (AMD64_L3_THREAD_SHIFT + thread) &
-				AMD64_L3_THREAD_MASK) | AMD64_L3_SLICE_MASK;
-	}
+	if (l3_mask && is_llc_event(event))
+		hwc->config |= l3_thread_slice_mask(event->cpu);
 
 	uncore = event_to_amd_uncore(event);
 	if (!uncore)
@@ -523,9 +541,9 @@ static int __init amd_uncore_init(void)
 	if (!boot_cpu_has(X86_FEATURE_TOPOEXT))
 		return -ENODEV;
 
-	if (boot_cpu_data.x86 == 0x17 || boot_cpu_data.x86 == 0x18) {
+	if (boot_cpu_data.x86 >= 0x17) {
 		/*
-		 * For F17h or F18h, the Northbridge counters are
+		 * For F17h and above, the Northbridge counters are
 		 * repurposed as Data Fabric counters. Also, L3
 		 * counters are supported too. The PMUs are exported
 		 * based on family as either L2 or L3 and NB or DF.
diff --git a/arch/x86/include/asm/perf_event.h b/arch/x86/include/asm/perf_event.h
index 29964b0e1075..e855e9cf2c37 100644
--- a/arch/x86/include/asm/perf_event.h
+++ b/arch/x86/include/asm/perf_event.h
@@ -50,11 +50,22 @@
 
 #define AMD64_L3_SLICE_SHIFT				48
 #define AMD64_L3_SLICE_MASK				\
-	((0xFULL) << AMD64_L3_SLICE_SHIFT)
+	(0xFULL << AMD64_L3_SLICE_SHIFT)
+#define AMD64_L3_SLICEID_MASK				\
+	(0x7ULL << AMD64_L3_SLICE_SHIFT)
 
 #define AMD64_L3_THREAD_SHIFT				56
 #define AMD64_L3_THREAD_MASK				\
-	((0xFFULL) << AMD64_L3_THREAD_SHIFT)
+	(0xFFULL << AMD64_L3_THREAD_SHIFT)
+#define AMD64_L3_F19H_THREAD_MASK			\
+	(0x3ULL << AMD64_L3_THREAD_SHIFT)
+
+#define AMD64_L3_EN_ALL_CORES				BIT_ULL(47)
+#define AMD64_L3_EN_ALL_SLICES				BIT_ULL(46)
+
+#define AMD64_L3_COREID_SHIFT				42
+#define AMD64_L3_COREID_MASK				\
+	(0x7ULL << AMD64_L3_COREID_SHIFT)
 
 #define X86_RAW_EVENT_MASK		\
 	(ARCH_PERFMON_EVENTSEL_EVENT |	\
diff --git a/include/linux/min_heap.h b/include/linux/min_heap.h
new file mode 100644
index 000000000000..44077837385f
--- /dev/null
+++ b/include/linux/min_heap.h
@@ -0,0 +1,134 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+#ifndef _LINUX_MIN_HEAP_H
+#define _LINUX_MIN_HEAP_H
+
+#include <linux/bug.h>
+#include <linux/string.h>
+#include <linux/types.h>
+
+/**
+ * struct min_heap - Data structure to hold a min-heap.
+ * @data: Start of array holding the heap elements.
+ * @nr: Number of elements currently in the heap.
+ * @size: Maximum number of elements that can be held in current storage.
+ */
+struct min_heap {
+	void *data;
+	int nr;
+	int size;
+};
+
+/**
+ * struct min_heap_callbacks - Data/functions to customise the min_heap.
+ * @elem_size: The nr of each element in bytes.
+ * @less: Partial order function for this heap.
+ * @swp: Swap elements function.
+ */
+struct min_heap_callbacks {
+	int elem_size;
+	bool (*less)(const void *lhs, const void *rhs);
+	void (*swp)(void *lhs, void *rhs);
+};
+
+/* Sift the element at pos down the heap. */
+static __always_inline
+void min_heapify(struct min_heap *heap, int pos,
+		const struct min_heap_callbacks *func)
+{
+	void *left, *right, *parent, *smallest;
+	void *data = heap->data;
+
+	for (;;) {
+		if (pos * 2 + 1 >= heap->nr)
+			break;
+
+		left = data + ((pos * 2 + 1) * func->elem_size);
+		parent = data + (pos * func->elem_size);
+		smallest = parent;
+		if (func->less(left, smallest))
+			smallest = left;
+
+		if (pos * 2 + 2 < heap->nr) {
+			right = data + ((pos * 2 + 2) * func->elem_size);
+			if (func->less(right, smallest))
+				smallest = right;
+		}
+		if (smallest == parent)
+			break;
+		func->swp(smallest, parent);
+		if (smallest == left)
+			pos = (pos * 2) + 1;
+		else
+			pos = (pos * 2) + 2;
+	}
+}
+
+/* Floyd's approach to heapification that is O(nr). */
+static __always_inline
+void min_heapify_all(struct min_heap *heap,
+		const struct min_heap_callbacks *func)
+{
+	int i;
+
+	for (i = heap->nr / 2; i >= 0; i--)
+		min_heapify(heap, i, func);
+}
+
+/* Remove minimum element from the heap, O(log2(nr)). */
+static __always_inline
+void min_heap_pop(struct min_heap *heap,
+		const struct min_heap_callbacks *func)
+{
+	void *data = heap->data;
+
+	if (WARN_ONCE(heap->nr <= 0, "Popping an empty heap"))
+		return;
+
+	/* Place last element at the root (position 0) and then sift down. */
+	heap->nr--;
+	memcpy(data, data + (heap->nr * func->elem_size), func->elem_size);
+	min_heapify(heap, 0, func);
+}
+
+/*
+ * Remove the minimum element and then push the given element. The
+ * implementation performs 1 sift (O(log2(nr))) and is therefore more
+ * efficient than a pop followed by a push that does 2.
+ */
+static __always_inline
+void min_heap_pop_push(struct min_heap *heap,
+		const void *element,
+		const struct min_heap_callbacks *func)
+{
+	memcpy(heap->data, element, func->elem_size);
+	min_heapify(heap, 0, func);
+}
+
+/* Push an element on to the heap, O(log2(nr)). */
+static __always_inline
+void min_heap_push(struct min_heap *heap, const void *element,
+		const struct min_heap_callbacks *func)
+{
+	void *data = heap->data;
+	void *child, *parent;
+	int pos;
+
+	if (WARN_ONCE(heap->nr >= heap->size, "Pushing on a full heap"))
+		return;
+
+	/* Place at the end of data. */
+	pos = heap->nr;
+	memcpy(data + (pos * func->elem_size), element, func->elem_size);
+	heap->nr++;
+
+	/* Sift child at pos up. */
+	for (; pos > 0; pos = (pos - 1) / 2) {
+		child = data + (pos * func->elem_size);
+		parent = data + ((pos - 1) / 2) * func->elem_size;
+		if (func->less(parent, child))
+			break;
+		func->swp(parent, child);
+	}
+}
+
+#endif /* _LINUX_MIN_HEAP_H */
diff --git a/include/linux/perf_event.h b/include/linux/perf_event.h
index 68e21e828893..8768a39b5258 100644
--- a/include/linux/perf_event.h
+++ b/include/linux/perf_event.h
@@ -862,6 +862,13 @@ struct perf_cpu_context {
 	int				sched_cb_usage;
 
 	int				online;
+	/*
+	 * Per-CPU storage for iterators used in visit_groups_merge. The default
+	 * storage is of size 2 to hold the CPU and any CPU event iterators.
+	 */
+	int				heap_size;
+	struct perf_event		**heap;
+	struct perf_event		*heap_default[2];
 };
 
 struct perf_output_handle {
diff --git a/kernel/events/core.c b/kernel/events/core.c
index 3f1f77de7247..ccf8d4fc6374 100644
--- a/kernel/events/core.c
+++ b/kernel/events/core.c
@@ -49,6 +49,7 @@
 #include <linux/sched/mm.h>
 #include <linux/proc_ns.h>
 #include <linux/mount.h>
+#include <linux/min_heap.h>
 
 #include "internal.h"
 
@@ -891,6 +892,47 @@ static inline void perf_cgroup_sched_in(struct task_struct *prev,
 	rcu_read_unlock();
 }
 
+static int perf_cgroup_ensure_storage(struct perf_event *event,
+				struct cgroup_subsys_state *css)
+{
+	struct perf_cpu_context *cpuctx;
+	struct perf_event **storage;
+	int cpu, heap_size, ret = 0;
+
+	/*
+	 * Allow storage to have sufficent space for an iterator for each
+	 * possibly nested cgroup plus an iterator for events with no cgroup.
+	 */
+	for (heap_size = 1; css; css = css->parent)
+		heap_size++;
+
+	for_each_possible_cpu(cpu) {
+		cpuctx = per_cpu_ptr(event->pmu->pmu_cpu_context, cpu);
+		if (heap_size <= cpuctx->heap_size)
+			continue;
+
+		storage = kmalloc_node(heap_size * sizeof(struct perf_event *),
+				       GFP_KERNEL, cpu_to_node(cpu));
+		if (!storage) {
+			ret = -ENOMEM;
+			break;
+		}
+
+		raw_spin_lock_irq(&cpuctx->ctx.lock);
+		if (cpuctx->heap_size < heap_size) {
+			swap(cpuctx->heap, storage);
+			if (storage == cpuctx->heap_default)
+				storage = NULL;
+			cpuctx->heap_size = heap_size;
+		}
+		raw_spin_unlock_irq(&cpuctx->ctx.lock);
+
+		kfree(storage);
+	}
+
+	return ret;
+}
+
 static inline int perf_cgroup_connect(int fd, struct perf_event *event,
 				      struct perf_event_attr *attr,
 				      struct perf_event *group_leader)
@@ -910,6 +952,10 @@ static inline int perf_cgroup_connect(int fd, struct perf_event *event,
 		goto out;
 	}
 
+	ret = perf_cgroup_ensure_storage(event, css);
+	if (ret)
+		goto out;
+
 	cgrp = container_of(css, struct perf_cgroup, css);
 	event->cgrp = cgrp;
 
@@ -1531,6 +1577,30 @@ perf_event_groups_less(struct perf_event *left, struct perf_event *right)
 	if (left->cpu > right->cpu)
 		return false;
 
+#ifdef CONFIG_CGROUP_PERF
+	if (left->cgrp != right->cgrp) {
+		if (!left->cgrp || !left->cgrp->css.cgroup) {
+			/*
+			 * Left has no cgroup but right does, no cgroups come
+			 * first.
+			 */
+			return true;
+		}
+		if (!right->cgrp || right->cgrp->css.cgroup) {
+			/*
+			 * Right has no cgroup but left does, no cgroups come
+			 * first.
+			 */
+			return false;
+		}
+		/* Two dissimilar cgroups, order by id. */
+		if (left->cgrp->css.cgroup->kn->id < right->cgrp->css.cgroup->kn->id)
+			return true;
+
+		return false;
+	}
+#endif
+
 	if (left->group_index < right->group_index)
 		return true;
 	if (left->group_index > right->group_index)
@@ -1610,25 +1680,48 @@ del_event_from_groups(struct perf_event *event, struct perf_event_context *ctx)
 }
 
 /*
- * Get the leftmost event in the @cpu subtree.
+ * Get the leftmost event in the cpu/cgroup subtree.
  */
 static struct perf_event *
-perf_event_groups_first(struct perf_event_groups *groups, int cpu)
+perf_event_groups_first(struct perf_event_groups *groups, int cpu,
+			struct cgroup *cgrp)
 {
 	struct perf_event *node_event = NULL, *match = NULL;
 	struct rb_node *node = groups->tree.rb_node;
+#ifdef CONFIG_CGROUP_PERF
+	u64 node_cgrp_id, cgrp_id = 0;
+
+	if (cgrp)
+		cgrp_id = cgrp->kn->id;
+#endif
 
 	while (node) {
 		node_event = container_of(node, struct perf_event, group_node);
 
 		if (cpu < node_event->cpu) {
 			node = node->rb_left;
-		} else if (cpu > node_event->cpu) {
+			continue;
+		}
+		if (cpu > node_event->cpu) {
 			node = node->rb_right;
-		} else {
-			match = node_event;
+			continue;
+		}
+#ifdef CONFIG_CGROUP_PERF
+		node_cgrp_id = 0;
+		if (node_event->cgrp && node_event->cgrp->css.cgroup)
+			node_cgrp_id = node_event->cgrp->css.cgroup->kn->id;
+
+		if (cgrp_id < node_cgrp_id) {
 			node = node->rb_left;
+			continue;
+		}
+		if (cgrp_id > node_cgrp_id) {
+			node = node->rb_right;
+			continue;
 		}
+#endif
+		match = node_event;
+		node = node->rb_left;
 	}
 
 	return match;
@@ -1641,12 +1734,26 @@ static struct perf_event *
 perf_event_groups_next(struct perf_event *event)
 {
 	struct perf_event *next;
+#ifdef CONFIG_CGROUP_PERF
+	u64 curr_cgrp_id = 0;
+	u64 next_cgrp_id = 0;
+#endif
 
 	next = rb_entry_safe(rb_next(&event->group_node), typeof(*event), group_node);
-	if (next && next->cpu == event->cpu)
-		return next;
+	if (next == NULL || next->cpu != event->cpu)
+		return NULL;
 
-	return NULL;
+#ifdef CONFIG_CGROUP_PERF
+	if (event->cgrp && event->cgrp->css.cgroup)
+		curr_cgrp_id = event->cgrp->css.cgroup->kn->id;
+
+	if (next->cgrp && next->cgrp->css.cgroup)
+		next_cgrp_id = next->cgrp->css.cgroup->kn->id;
+
+	if (curr_cgrp_id != next_cgrp_id)
+		return NULL;
+#endif
+	return next;
 }
 
 /*
@@ -1986,6 +2093,12 @@ static int perf_get_aux_event(struct perf_event *event,
 	return 1;
 }
 
+static inline struct list_head *get_event_list(struct perf_event *event)
+{
+	struct perf_event_context *ctx = event->ctx;
+	return event->attr.pinned ? &ctx->pinned_active : &ctx->flexible_active;
+}
+
 static void perf_group_detach(struct perf_event *event)
 {
 	struct perf_event *sibling, *tmp;
@@ -2028,12 +2141,8 @@ static void perf_group_detach(struct perf_event *event)
 		if (!RB_EMPTY_NODE(&event->group_node)) {
 			add_event_to_groups(sibling, event->ctx);
 
-			if (sibling->state == PERF_EVENT_STATE_ACTIVE) {
-				struct list_head *list = sibling->attr.pinned ?
-					&ctx->pinned_active : &ctx->flexible_active;
-
-				list_add_tail(&sibling->active_list, list);
-			}
+			if (sibling->state == PERF_EVENT_STATE_ACTIVE)
+				list_add_tail(&sibling->active_list, get_event_list(sibling));
 		}
 
 		WARN_ON_ONCE(sibling->ctx != event->ctx);
@@ -2350,6 +2459,8 @@ event_sched_in(struct perf_event *event,
 {
 	int ret = 0;
 
+	WARN_ON_ONCE(event->ctx != ctx);
+
 	lockdep_assert_held(&ctx->lock);
 
 	if (event->state <= PERF_EVENT_STATE_OFF)
@@ -3388,71 +3499,103 @@ static void cpu_ctx_sched_out(struct perf_cpu_context *cpuctx,
 	ctx_sched_out(&cpuctx->ctx, cpuctx, event_type);
 }
 
-static int visit_groups_merge(struct perf_event_groups *groups, int cpu,
-			      int (*func)(struct perf_event *, void *), void *data)
+static bool perf_less_group_idx(const void *l, const void *r)
 {
-	struct perf_event **evt, *evt1, *evt2;
-	int ret;
-
-	evt1 = perf_event_groups_first(groups, -1);
-	evt2 = perf_event_groups_first(groups, cpu);
+	const struct perf_event *le = l, *re = r;
 
-	while (evt1 || evt2) {
-		if (evt1 && evt2) {
-			if (evt1->group_index < evt2->group_index)
-				evt = &evt1;
-			else
-				evt = &evt2;
-		} else if (evt1) {
-			evt = &evt1;
-		} else {
-			evt = &evt2;
-		}
-
-		ret = func(*evt, data);
-		if (ret)
-			return ret;
+	return le->group_index < re->group_index;
+}
 
-		*evt = perf_event_groups_next(*evt);
-	}
+static void swap_ptr(void *l, void *r)
+{
+	void **lp = l, **rp = r;
 
-	return 0;
+	swap(*lp, *rp);
 }
 
-struct sched_in_data {
-	struct perf_event_context *ctx;
-	struct perf_cpu_context *cpuctx;
-	int can_add_hw;
+static const struct min_heap_callbacks perf_min_heap = {
+	.elem_size = sizeof(struct perf_event *),
+	.less = perf_less_group_idx,
+	.swp = swap_ptr,
 };
 
-static int pinned_sched_in(struct perf_event *event, void *data)
+static void __heap_add(struct min_heap *heap, struct perf_event *event)
 {
-	struct sched_in_data *sid = data;
+	struct perf_event **itrs = heap->data;
 
-	if (event->state <= PERF_EVENT_STATE_OFF)
-		return 0;
+	if (event) {
+		itrs[heap->nr] = event;
+		heap->nr++;
+	}
+}
 
-	if (!event_filter_match(event))
-		return 0;
+static noinline int visit_groups_merge(struct perf_cpu_context *cpuctx,
+				struct perf_event_groups *groups, int cpu,
+				int (*func)(struct perf_event *, void *),
+				void *data)
+{
+#ifdef CONFIG_CGROUP_PERF
+	struct cgroup_subsys_state *css = NULL;
+#endif
+	/* Space for per CPU and/or any CPU event iterators. */
+	struct perf_event *itrs[2];
+	struct min_heap event_heap;
+	struct perf_event **evt;
+	int ret;
 
-	if (group_can_go_on(event, sid->cpuctx, sid->can_add_hw)) {
-		if (!group_sched_in(event, sid->cpuctx, sid->ctx))
-			list_add_tail(&event->active_list, &sid->ctx->pinned_active);
+	if (cpuctx) {
+		event_heap = (struct min_heap){
+			.data = cpuctx->heap,
+			.nr = 0,
+			.size = cpuctx->heap_size,
+		};
+
+		lockdep_assert_held(&cpuctx->ctx.lock);
+
+#ifdef CONFIG_CGROUP_PERF
+		if (cpuctx->cgrp)
+			css = &cpuctx->cgrp->css;
+#endif
+	} else {
+		event_heap = (struct min_heap){
+			.data = itrs,
+			.nr = 0,
+			.size = ARRAY_SIZE(itrs),
+		};
+		/* Events not within a CPU context may be on any CPU. */
+		__heap_add(&event_heap, perf_event_groups_first(groups, -1, NULL));
 	}
+	evt = event_heap.data;
 
-	/*
-	 * If this pinned group hasn't been scheduled,
-	 * put it in error state.
-	 */
-	if (event->state == PERF_EVENT_STATE_INACTIVE)
-		perf_event_set_state(event, PERF_EVENT_STATE_ERROR);
+	__heap_add(&event_heap, perf_event_groups_first(groups, cpu, NULL));
+
+#ifdef CONFIG_CGROUP_PERF
+	for (; css; css = css->parent)
+		__heap_add(&event_heap, perf_event_groups_first(groups, cpu, css->cgroup));
+#endif
+
+	min_heapify_all(&event_heap, &perf_min_heap);
+
+	while (event_heap.nr) {
+		ret = func(*evt, data);
+		if (ret)
+			return ret;
+
+		*evt = perf_event_groups_next(*evt);
+		if (*evt)
+			min_heapify(&event_heap, 0, &perf_min_heap);
+		else
+			min_heap_pop(&event_heap, &perf_min_heap);
+	}
 
 	return 0;
 }
 
-static int flexible_sched_in(struct perf_event *event, void *data)
+static int merge_sched_in(struct perf_event *event, void *data)
 {
-	struct sched_in_data *sid = data;
+	struct perf_event_context *ctx = event->ctx;
+	struct perf_cpu_context *cpuctx = __get_cpu_context(ctx);
+	int *can_add_hw = data;
 
 	if (event->state <= PERF_EVENT_STATE_OFF)
 		return 0;
@@ -3460,14 +3603,17 @@ static int flexible_sched_in(struct perf_event *event, void *data)
 	if (!event_filter_match(event))
 		return 0;
 
-	if (group_can_go_on(event, sid->cpuctx, sid->can_add_hw)) {
-		int ret = group_sched_in(event, sid->cpuctx, sid->ctx);
-		if (ret) {
-			sid->can_add_hw = 0;
-			sid->ctx->rotate_necessary = 1;
-			return 0;
-		}
-		list_add_tail(&event->active_list, &sid->ctx->flexible_active);
+	if (group_can_go_on(event, cpuctx, *can_add_hw)) {
+		if (!group_sched_in(event, cpuctx, ctx))
+			list_add_tail(&event->active_list, get_event_list(event));
+	}
+
+	if (event->state == PERF_EVENT_STATE_INACTIVE) {
+		if (event->attr.pinned)
+			perf_event_set_state(event, PERF_EVENT_STATE_ERROR);
+
+		*can_add_hw = 0;
+		ctx->rotate_necessary = 1;
 	}
 
 	return 0;
@@ -3477,30 +3623,28 @@ static void
 ctx_pinned_sched_in(struct perf_event_context *ctx,
 		    struct perf_cpu_context *cpuctx)
 {
-	struct sched_in_data sid = {
-		.ctx = ctx,
-		.cpuctx = cpuctx,
-		.can_add_hw = 1,
-	};
+	int can_add_hw = 1;
+
+	if (ctx != &cpuctx->ctx)
+		cpuctx = NULL;
 
-	visit_groups_merge(&ctx->pinned_groups,
+	visit_groups_merge(cpuctx, &ctx->pinned_groups,
 			   smp_processor_id(),
-			   pinned_sched_in, &sid);
+			   merge_sched_in, &can_add_hw);
 }
 
 static void
 ctx_flexible_sched_in(struct perf_event_context *ctx,
 		      struct perf_cpu_context *cpuctx)
 {
-	struct sched_in_data sid = {
-		.ctx = ctx,
-		.cpuctx = cpuctx,
-		.can_add_hw = 1,
-	};
+	int can_add_hw = 1;
+
+	if (ctx != &cpuctx->ctx)
+		cpuctx = NULL;
 
-	visit_groups_merge(&ctx->flexible_groups,
+	visit_groups_merge(cpuctx, &ctx->flexible_groups,
 			   smp_processor_id(),
-			   flexible_sched_in, &sid);
+			   merge_sched_in, &can_add_hw);
 }
 
 static void
@@ -10359,6 +10503,9 @@ skip_type:
 		cpuctx->online = cpumask_test_cpu(cpu, perf_online_mask);
 
 		__perf_mux_hrtimer_init(cpuctx, cpu);
+
+		cpuctx->heap_size = ARRAY_SIZE(cpuctx->heap_default);
+		cpuctx->heap = cpuctx->heap_default;
 	}
 
 got_cpu_context:
@@ -10804,12 +10951,6 @@ perf_event_alloc(struct perf_event_attr *attr, int cpu,
 	if (!has_branch_stack(event))
 		event->attr.branch_sample_type = 0;
 
-	if (cgroup_fd != -1) {
-		err = perf_cgroup_connect(cgroup_fd, event, attr, group_leader);
-		if (err)
-			goto err_ns;
-	}
-
 	pmu = perf_init_event(event);
 	if (IS_ERR(pmu)) {
 		err = PTR_ERR(pmu);
@@ -10831,6 +10972,12 @@ perf_event_alloc(struct perf_event_attr *attr, int cpu,
 		goto err_pmu;
 	}
 
+	if (cgroup_fd != -1) {
+		err = perf_cgroup_connect(cgroup_fd, event, attr, group_leader);
+		if (err)
+			goto err_pmu;
+	}
+
 	err = exclusive_event_init(event);
 	if (err)
 		goto err_pmu;
@@ -10891,12 +11038,12 @@ err_per_task:
 	exclusive_event_destroy(event);
 
 err_pmu:
+	if (is_cgroup_event(event))
+		perf_detach_cgroup(event);
 	if (event->destroy)
 		event->destroy(event);
 	module_put(pmu->module);
 err_ns:
-	if (is_cgroup_event(event))
-		perf_detach_cgroup(event);
 	if (event->ns)
 		put_pid_ns(event->ns);
 	if (event->hw.target)
diff --git a/lib/Kconfig.debug b/lib/Kconfig.debug
index 69def4a9df00..f04b61c1a1cc 100644
--- a/lib/Kconfig.debug
+++ b/lib/Kconfig.debug
@@ -1769,6 +1769,16 @@ config TEST_LIST_SORT
 
 	  If unsure, say N.
 
+config TEST_MIN_HEAP
+	tristate "Min heap test"
+	depends on DEBUG_KERNEL || m
+	help
+	  Enable this to turn on min heap function tests. This test is
+	  executed only once during system boot (so affects only boot time),
+	  or at module load time.
+
+	  If unsure, say N.
+
 config TEST_SORT
 	tristate "Array-based sort test"
 	depends on DEBUG_KERNEL || m
diff --git a/lib/Makefile b/lib/Makefile
index 611872c06926..09a8acb0cf92 100644
--- a/lib/Makefile
+++ b/lib/Makefile
@@ -67,6 +67,7 @@ CFLAGS_test_ubsan.o += $(call cc-disable-warning, vla)
 UBSAN_SANITIZE_test_ubsan.o := y
 obj-$(CONFIG_TEST_KSTRTOX) += test-kstrtox.o
 obj-$(CONFIG_TEST_LIST_SORT) += test_list_sort.o
+obj-$(CONFIG_TEST_MIN_HEAP) += test_min_heap.o
 obj-$(CONFIG_TEST_LKM) += test_module.o
 obj-$(CONFIG_TEST_VMALLOC) += test_vmalloc.o
 obj-$(CONFIG_TEST_OVERFLOW) += test_overflow.o
diff --git a/lib/test_min_heap.c b/lib/test_min_heap.c
new file mode 100644
index 000000000000..d19c8080fd4d
--- /dev/null
+++ b/lib/test_min_heap.c
@@ -0,0 +1,194 @@
+// SPDX-License-Identifier: GPL-2.0-only
+#define pr_fmt(fmt) "min_heap_test: " fmt
+
+/*
+ * Test cases for the min max heap.
+ */
+
+#include <linux/log2.h>
+#include <linux/min_heap.h>
+#include <linux/module.h>
+#include <linux/printk.h>
+#include <linux/random.h>
+
+static __init bool less_than(const void *lhs, const void *rhs)
+{
+	return *(int *)lhs < *(int *)rhs;
+}
+
+static __init bool greater_than(const void *lhs, const void *rhs)
+{
+	return *(int *)lhs > *(int *)rhs;
+}
+
+static __init void swap_ints(void *lhs, void *rhs)
+{
+	int temp = *(int *)lhs;
+
+	*(int *)lhs = *(int *)rhs;
+	*(int *)rhs = temp;
+}
+
+static __init int pop_verify_heap(bool min_heap,
+				struct min_heap *heap,
+				const struct min_heap_callbacks *funcs)
+{
+	int *values = heap->data;
+	int err = 0;
+	int last;
+
+	last = values[0];
+	min_heap_pop(heap, funcs);
+	while (heap->nr > 0) {
+		if (min_heap) {
+			if (last > values[0]) {
+				pr_err("error: expected %d <= %d\n", last,
+					values[0]);
+				err++;
+			}
+		} else {
+			if (last < values[0]) {
+				pr_err("error: expected %d >= %d\n", last,
+					values[0]);
+				err++;
+			}
+		}
+		last = values[0];
+		min_heap_pop(heap, funcs);
+	}
+	return err;
+}
+
+static __init int test_heapify_all(bool min_heap)
+{
+	int values[] = { 3, 1, 2, 4, 0x8000000, 0x7FFFFFF, 0,
+			 -3, -1, -2, -4, 0x8000000, 0x7FFFFFF };
+	struct min_heap heap = {
+		.data = values,
+		.nr = ARRAY_SIZE(values),
+		.size =  ARRAY_SIZE(values),
+	};
+	struct min_heap_callbacks funcs = {
+		.elem_size = sizeof(int),
+		.less = min_heap ? less_than : greater_than,
+		.swp = swap_ints,
+	};
+	int i, err;
+
+	/* Test with known set of values. */
+	min_heapify_all(&heap, &funcs);
+	err = pop_verify_heap(min_heap, &heap, &funcs);
+
+
+	/* Test with randomly generated values. */
+	heap.nr = ARRAY_SIZE(values);
+	for (i = 0; i < heap.nr; i++)
+		values[i] = get_random_int();
+
+	min_heapify_all(&heap, &funcs);
+	err += pop_verify_heap(min_heap, &heap, &funcs);
+
+	return err;
+}
+
+static __init int test_heap_push(bool min_heap)
+{
+	const int data[] = { 3, 1, 2, 4, 0x80000000, 0x7FFFFFFF, 0,
+			     -3, -1, -2, -4, 0x80000000, 0x7FFFFFFF };
+	int values[ARRAY_SIZE(data)];
+	struct min_heap heap = {
+		.data = values,
+		.nr = 0,
+		.size =  ARRAY_SIZE(values),
+	};
+	struct min_heap_callbacks funcs = {
+		.elem_size = sizeof(int),
+		.less = min_heap ? less_than : greater_than,
+		.swp = swap_ints,
+	};
+	int i, temp, err;
+
+	/* Test with known set of values copied from data. */
+	for (i = 0; i < ARRAY_SIZE(data); i++)
+		min_heap_push(&heap, &data[i], &funcs);
+
+	err = pop_verify_heap(min_heap, &heap, &funcs);
+
+	/* Test with randomly generated values. */
+	while (heap.nr < heap.size) {
+		temp = get_random_int();
+		min_heap_push(&heap, &temp, &funcs);
+	}
+	err += pop_verify_heap(min_heap, &heap, &funcs);
+
+	return err;
+}
+
+static __init int test_heap_pop_push(bool min_heap)
+{
+	const int data[] = { 3, 1, 2, 4, 0x80000000, 0x7FFFFFFF, 0,
+			     -3, -1, -2, -4, 0x80000000, 0x7FFFFFFF };
+	int values[ARRAY_SIZE(data)];
+	struct min_heap heap = {
+		.data = values,
+		.nr = 0,
+		.size =  ARRAY_SIZE(values),
+	};
+	struct min_heap_callbacks funcs = {
+		.elem_size = sizeof(int),
+		.less = min_heap ? less_than : greater_than,
+		.swp = swap_ints,
+	};
+	int i, temp, err;
+
+	/* Fill values with data to pop and replace. */
+	temp = min_heap ? 0x80000000 : 0x7FFFFFFF;
+	for (i = 0; i < ARRAY_SIZE(data); i++)
+		min_heap_push(&heap, &temp, &funcs);
+
+	/* Test with known set of values copied from data. */
+	for (i = 0; i < ARRAY_SIZE(data); i++)
+		min_heap_pop_push(&heap, &data[i], &funcs);
+
+	err = pop_verify_heap(min_heap, &heap, &funcs);
+
+	heap.nr = 0;
+	for (i = 0; i < ARRAY_SIZE(data); i++)
+		min_heap_push(&heap, &temp, &funcs);
+
+	/* Test with randomly generated values. */
+	for (i = 0; i < ARRAY_SIZE(data); i++) {
+		temp = get_random_int();
+		min_heap_pop_push(&heap, &temp, &funcs);
+	}
+	err += pop_verify_heap(min_heap, &heap, &funcs);
+
+	return err;
+}
+
+static int __init test_min_heap_init(void)
+{
+	int err = 0;
+
+	err += test_heapify_all(true);
+	err += test_heapify_all(false);
+	err += test_heap_push(true);
+	err += test_heap_push(false);
+	err += test_heap_pop_push(true);
+	err += test_heap_pop_push(false);
+	if (err) {
+		pr_err("test failed with %d errors\n", err);
+		return -EINVAL;
+	}
+	pr_info("test passed\n");
+	return 0;
+}
+module_init(test_min_heap_init);
+
+static void __exit test_min_heap_exit(void)
+{
+	/* do nothing */
+}
+module_exit(test_min_heap_exit);
+
+MODULE_LICENSE("GPL");