processor.c   [plain text]

/*
 * Copyright (c) 2000-2019 Apple Inc. All rights reserved.
 *
 * @APPLE_OSREFERENCE_LICENSE_HEADER_START@
 *
 * This file contains Original Code and/or Modifications of Original Code
 * as defined in and that are subject to the Apple Public Source License
 * Version 2.0 (the 'License'). You may not use this file except in
 * compliance with the License. The rights granted to you under the License
 * may not be used to create, or enable the creation or redistribution of,
 * unlawful or unlicensed copies of an Apple operating system, or to
 * circumvent, violate, or enable the circumvention or violation of, any
 * terms of an Apple operating system software license agreement.
 *
 * Please obtain a copy of the License at
 * http://www.opensource.apple.com/apsl/ and read it before using this file.
 *
 * The Original Code and all software distributed under the License are
 * distributed on an 'AS IS' basis, WITHOUT WARRANTY OF ANY KIND, EITHER
 * EXPRESS OR IMPLIED, AND APPLE HEREBY DISCLAIMS ALL SUCH WARRANTIES,
 * INCLUDING WITHOUT LIMITATION, ANY WARRANTIES OF MERCHANTABILITY,
 * FITNESS FOR A PARTICULAR PURPOSE, QUIET ENJOYMENT OR NON-INFRINGEMENT.
 * Please see the License for the specific language governing rights and
 * limitations under the License.
 *
 * @APPLE_OSREFERENCE_LICENSE_HEADER_END@
 */
/*
 * @OSF_COPYRIGHT@
 */
/*
 * Mach Operating System
 * Copyright (c) 1991,1990,1989,1988 Carnegie Mellon University
 * All Rights Reserved.
 *
 * Permission to use, copy, modify and distribute this software and its
 * documentation is hereby granted, provided that both the copyright
 * notice and this permission notice appear in all copies of the
 * software, derivative works or modified versions, and any portions
 * thereof, and that both notices appear in supporting documentation.
 *
 * CARNEGIE MELLON ALLOWS FREE USE OF THIS SOFTWARE IN ITS "AS IS"
 * CONDITION.  CARNEGIE MELLON DISCLAIMS ANY LIABILITY OF ANY KIND FOR
 * ANY DAMAGES WHATSOEVER RESULTING FROM THE USE OF THIS SOFTWARE.
 *
 * Carnegie Mellon requests users of this software to return to
 *
 *  Software Distribution Coordinator  or  Software.Distribution@CS.CMU.EDU
 *  School of Computer Science
 *  Carnegie Mellon University
 *  Pittsburgh PA 15213-3890
 *
 * any improvements or extensions that they make and grant Carnegie Mellon
 * the rights to redistribute these changes.
 */
/*
 */

/*
 *	processor.c: processor and processor_set manipulation routines.
 */

#include <mach/boolean.h>
#include <mach/policy.h>
#include <mach/processor.h>
#include <mach/processor_info.h>
#include <mach/vm_param.h>
#include <kern/cpu_number.h>
#include <kern/host.h>
#include <kern/ipc_host.h>
#include <kern/ipc_tt.h>
#include <kern/kalloc.h>
#include <kern/machine.h>
#include <kern/misc_protos.h>
#include <kern/processor.h>
#include <kern/sched.h>
#include <kern/task.h>
#include <kern/thread.h>
#include <kern/timer.h>
#if KPERF
#include <kperf/kperf.h>
#endif /* KPERF */
#include <ipc/ipc_port.h>

#include <security/mac_mach_internal.h>

#if defined(CONFIG_XNUPOST)

#include <tests/xnupost.h>

#endif /* CONFIG_XNUPOST */

/*
 * Exported interface
 */
#include <mach/mach_host_server.h>
#include <mach/processor_set_server.h>

struct processor_set    pset0;
struct pset_node        pset_node0;

static SIMPLE_LOCK_DECLARE(pset_node_lock, 0);
LCK_GRP_DECLARE(pset_lck_grp, "pset");

queue_head_t            tasks;
queue_head_t            terminated_tasks;       /* To be used ONLY for stackshot. */
queue_head_t            corpse_tasks;
int                     tasks_count;
int                     terminated_tasks_count;
queue_head_t            threads;
int                     threads_count;
LCK_GRP_DECLARE(task_lck_grp, "task");
LCK_ATTR_DECLARE(task_lck_attr, 0, 0);
LCK_MTX_DECLARE_ATTR(tasks_threads_lock, &task_lck_grp, &task_lck_attr);
LCK_MTX_DECLARE_ATTR(tasks_corpse_lock, &task_lck_grp, &task_lck_attr);

processor_t             processor_list;
unsigned int            processor_count;
static processor_t      processor_list_tail;
SIMPLE_LOCK_DECLARE(processor_list_lock, 0);

uint32_t                processor_avail_count;
uint32_t                processor_avail_count_user;
uint32_t                primary_processor_avail_count;
uint32_t                primary_processor_avail_count_user;

int                     master_cpu = 0;

struct processor        PERCPU_DATA(processor);
processor_t             processor_array[MAX_SCHED_CPUS] = { 0 };
processor_set_t         pset_array[MAX_PSETS] = { 0 };

static timer_call_func_t running_timer_funcs[] = {
	[RUNNING_TIMER_QUANTUM] = thread_quantum_expire,
	[RUNNING_TIMER_KPERF] = kperf_timer_expire,
};
static_assert(sizeof(running_timer_funcs) / sizeof(running_timer_funcs[0])
    == RUNNING_TIMER_MAX, "missing running timer function");

#if defined(CONFIG_XNUPOST)
kern_return_t ipi_test(void);
extern void arm64_ipi_test(void);

kern_return_t
ipi_test()
{
#if __arm64__
	processor_t p;

	for (p = processor_list; p != NULL; p = p->processor_list) {
		thread_bind(p);
		thread_block(THREAD_CONTINUE_NULL);
		kprintf("Running IPI test on cpu %d\n", p->cpu_id);
		arm64_ipi_test();
	}

	/* unbind thread from specific cpu */
	thread_bind(PROCESSOR_NULL);
	thread_block(THREAD_CONTINUE_NULL);

	T_PASS("Done running IPI tests");
#else
	T_PASS("Unsupported platform. Not running IPI tests");

#endif /* __arm64__ */

	return KERN_SUCCESS;
}
#endif /* defined(CONFIG_XNUPOST) */

int sched_enable_smt = 1;

void
processor_bootstrap(void)
{
	pset_node0.psets = &pset0;
	pset_init(&pset0, &pset_node0);

	queue_init(&tasks);
	queue_init(&terminated_tasks);
	queue_init(&threads);
	queue_init(&corpse_tasks);

	processor_init(master_processor, master_cpu, &pset0);
}

/*
 *	Initialize the given processor for the cpu
 *	indicated by cpu_id, and assign to the
 *	specified processor set.
 */
void
processor_init(
	processor_t            processor,
	int                    cpu_id,
	processor_set_t        pset)
{
	spl_t           s;

	assert(cpu_id < MAX_SCHED_CPUS);
	processor->cpu_id = cpu_id;

	if (processor != master_processor) {
		/* Scheduler state for master_processor initialized in sched_init() */
		SCHED(processor_init)(processor);
	}

	processor->state = PROCESSOR_OFF_LINE;
	processor->active_thread = processor->startup_thread = processor->idle_thread = THREAD_NULL;
	processor->processor_set = pset;
	processor_state_update_idle(processor);
	processor->starting_pri = MINPRI;
	processor->quantum_end = UINT64_MAX;
	processor->deadline = UINT64_MAX;
	processor->first_timeslice = FALSE;
	processor->processor_offlined = false;
	processor->processor_primary = processor; /* no SMT relationship known at this point */
	processor->processor_secondary = NULL;
	processor->is_SMT = false;
	processor->is_recommended = true;
	processor->processor_self = IP_NULL;
	processor->processor_list = NULL;
	processor->must_idle = false;
	processor->running_timers_active = false;
	for (int i = 0; i < RUNNING_TIMER_MAX; i++) {
		timer_call_setup(&processor->running_timers[i],
		    running_timer_funcs[i], processor);
		running_timer_clear(processor, i);
	}

	timer_init(&processor->idle_state);
	timer_init(&processor->system_state);
	timer_init(&processor->user_state);

	s = splsched();
	pset_lock(pset);
	bit_set(pset->cpu_bitmask, cpu_id);
	bit_set(pset->recommended_bitmask, cpu_id);
	bit_set(pset->primary_map, cpu_id);
	bit_set(pset->cpu_state_map[PROCESSOR_OFF_LINE], cpu_id);
	if (pset->cpu_set_count++ == 0) {
		pset->cpu_set_low = pset->cpu_set_hi = cpu_id;
	} else {
		pset->cpu_set_low = (cpu_id < pset->cpu_set_low)? cpu_id: pset->cpu_set_low;
		pset->cpu_set_hi = (cpu_id > pset->cpu_set_hi)? cpu_id: pset->cpu_set_hi;
	}
	pset_unlock(pset);
	splx(s);

	simple_lock(&processor_list_lock, LCK_GRP_NULL);
	if (processor_list == NULL) {
		processor_list = processor;
	} else {
		processor_list_tail->processor_list = processor;
	}
	processor_list_tail = processor;
	processor_count++;
	processor_array[cpu_id] = processor;
	simple_unlock(&processor_list_lock);
}

bool system_is_SMT = false;

void
processor_set_primary(
	processor_t             processor,
	processor_t             primary)
{
	assert(processor->processor_primary == primary || processor->processor_primary == processor);
	/* Re-adjust primary point for this (possibly) secondary processor */
	processor->processor_primary = primary;

	assert(primary->processor_secondary == NULL || primary->processor_secondary == processor);
	if (primary != processor) {
		/* Link primary to secondary, assumes a 2-way SMT model
		 * We'll need to move to a queue if any future architecture
		 * requires otherwise.
		 */
		assert(processor->processor_secondary == NULL);
		primary->processor_secondary = processor;
		/* Mark both processors as SMT siblings */
		primary->is_SMT = TRUE;
		processor->is_SMT = TRUE;

		if (!system_is_SMT) {
			system_is_SMT = true;
		}

		processor_set_t pset = processor->processor_set;
		spl_t s = splsched();
		pset_lock(pset);
		if (!pset->is_SMT) {
			pset->is_SMT = true;
		}
		bit_clear(pset->primary_map, processor->cpu_id);
		pset_unlock(pset);
		splx(s);
	}
}

processor_set_t
processor_pset(
	processor_t     processor)
{
	return processor->processor_set;
}

#if CONFIG_SCHED_EDGE

cluster_type_t
pset_type_for_id(uint32_t cluster_id)
{
	return pset_array[cluster_id]->pset_type;
}

/*
 * Processor foreign threads
 *
 * With the Edge scheduler, each pset maintains a bitmap of processors running threads
 * which are foreign to the pset/cluster. A thread is defined as foreign for a cluster
 * if its of a different type than its preferred cluster type (E/P). The bitmap should
 * be updated every time a new thread is assigned to run on a processor.
 *
 * This bitmap allows the Edge scheduler to quickly find CPUs running foreign threads
 * for rebalancing.
 */
static void
processor_state_update_running_foreign(processor_t processor, thread_t thread)
{
	cluster_type_t current_processor_type = pset_type_for_id(processor->processor_set->pset_cluster_id);
	cluster_type_t thread_type = pset_type_for_id(sched_edge_thread_preferred_cluster(thread));

	/* Update the bitmap for the pset only for unbounded non-RT threads. */
	if ((processor->current_pri < BASEPRI_RTQUEUES) && (thread->bound_processor == PROCESSOR_NULL) && (current_processor_type != thread_type)) {
		bit_set(processor->processor_set->cpu_running_foreign, processor->cpu_id);
	} else {
		bit_clear(processor->processor_set->cpu_running_foreign, processor->cpu_id);
	}
}
#else /* CONFIG_SCHED_EDGE */
static void
processor_state_update_running_foreign(__unused processor_t processor, __unused thread_t thread)
{
}
#endif /* CONFIG_SCHED_EDGE */

void
processor_state_update_idle(processor_t processor)
{
	processor->current_pri = IDLEPRI;
	processor->current_sfi_class = SFI_CLASS_KERNEL;
	processor->current_recommended_pset_type = PSET_SMP;
#if CONFIG_THREAD_GROUPS
	processor->current_thread_group = NULL;
#endif
	processor->current_perfctl_class = PERFCONTROL_CLASS_IDLE;
	processor->current_urgency = THREAD_URGENCY_NONE;
	processor->current_is_NO_SMT = false;
	processor->current_is_bound = false;
	processor->current_is_eagerpreempt = false;
	os_atomic_store(&processor->processor_set->cpu_running_buckets[processor->cpu_id], TH_BUCKET_SCHED_MAX, relaxed);
}

void
processor_state_update_from_thread(processor_t processor, thread_t thread)
{
	processor->current_pri = thread->sched_pri;
	processor->current_sfi_class = thread->sfi_class;
	processor->current_recommended_pset_type = recommended_pset_type(thread);
	processor_state_update_running_foreign(processor, thread);
	/* Since idle and bound threads are not tracked by the edge scheduler, ignore when those threads go on-core */
	sched_bucket_t bucket = ((thread->state & TH_IDLE) || (thread->bound_processor != PROCESSOR_NULL)) ? TH_BUCKET_SCHED_MAX : thread->th_sched_bucket;
	os_atomic_store(&processor->processor_set->cpu_running_buckets[processor->cpu_id], bucket, relaxed);

#if CONFIG_THREAD_GROUPS
	processor->current_thread_group = thread_group_get(thread);
#endif
	processor->current_perfctl_class = thread_get_perfcontrol_class(thread);
	processor->current_urgency = thread_get_urgency(thread, NULL, NULL);
	processor->current_is_NO_SMT = thread_no_smt(thread);
	processor->current_is_bound = thread->bound_processor != PROCESSOR_NULL;
	processor->current_is_eagerpreempt = thread_is_eager_preempt(thread);
}

void
processor_state_update_explicit(processor_t processor, int pri, sfi_class_id_t sfi_class,
    pset_cluster_type_t pset_type, perfcontrol_class_t perfctl_class, thread_urgency_t urgency, sched_bucket_t bucket)
{
	processor->current_pri = pri;
	processor->current_sfi_class = sfi_class;
	processor->current_recommended_pset_type = pset_type;
	processor->current_perfctl_class = perfctl_class;
	processor->current_urgency = urgency;
	os_atomic_store(&processor->processor_set->cpu_running_buckets[processor->cpu_id], bucket, relaxed);
}

pset_node_t
pset_node_root(void)
{
	return &pset_node0;
}

processor_set_t
pset_create(
	pset_node_t                     node)
{
	/* some schedulers do not support multiple psets */
	if (SCHED(multiple_psets_enabled) == FALSE) {
		return processor_pset(master_processor);
	}

	processor_set_t *prev, pset = zalloc_permanent_type(struct processor_set);

	if (pset != PROCESSOR_SET_NULL) {
		pset_init(pset, node);

		simple_lock(&pset_node_lock, LCK_GRP_NULL);

		prev = &node->psets;
		while (*prev != PROCESSOR_SET_NULL) {
			prev = &(*prev)->pset_list;
		}

		*prev = pset;

		simple_unlock(&pset_node_lock);
	}

	return pset;
}

/*
 *	Find processor set with specified cluster_id.
 *	Returns default_pset if not found.
 */
processor_set_t
pset_find(
	uint32_t cluster_id,
	processor_set_t default_pset)
{
	simple_lock(&pset_node_lock, LCK_GRP_NULL);
	pset_node_t node = &pset_node0;
	processor_set_t pset = NULL;

	do {
		pset = node->psets;
		while (pset != NULL) {
			if (pset->pset_cluster_id == cluster_id) {
				break;
			}
			pset = pset->pset_list;
		}
	} while (pset == NULL && (node = node->node_list) != NULL);
	simple_unlock(&pset_node_lock);
	if (pset == NULL) {
		return default_pset;
	}
	return pset;
}

#if !defined(RC_HIDE_XNU_FIRESTORM) && (MAX_CPU_CLUSTERS > 2)

/*
 * Find the first processor_set for the given pset_cluster_type.
 * Should be removed with rdar://57340304, as it's only
 * useful for the workaround described in rdar://57306691.
 */

processor_set_t
pset_find_first_by_cluster_type(
	pset_cluster_type_t pset_cluster_type)
{
	simple_lock(&pset_node_lock, LCK_GRP_NULL);
	pset_node_t node = &pset_node0;
	processor_set_t pset = NULL;

	do {
		pset = node->psets;
		while (pset != NULL) {
			if (pset->pset_cluster_type == pset_cluster_type) {
				break;
			}
			pset = pset->pset_list;
		}
	} while (pset == NULL && (node = node->node_list) != NULL);
	simple_unlock(&pset_node_lock);
	return pset;
}

#endif /* !defined(RC_HIDE_XNU_FIRESTORM) && (MAX_CPU_CLUSTERS > 2) */

/*
 *	Initialize the given processor_set structure.
 */
void
pset_init(
	processor_set_t         pset,
	pset_node_t                     node)
{
	static uint32_t pset_count = 0;

	if (pset != &pset0) {
		/*
		 * Scheduler runqueue initialization for non-boot psets.
		 * This initialization for pset0 happens in sched_init().
		 */
		SCHED(pset_init)(pset);
		SCHED(rt_init)(pset);
	}

	pset->online_processor_count = 0;
	pset->load_average = 0;
	bzero(&pset->pset_load_average, sizeof(pset->pset_load_average));
#if CONFIG_SCHED_EDGE
	bzero(&pset->pset_execution_time, sizeof(pset->pset_execution_time));
#endif /* CONFIG_SCHED_EDGE */
	pset->cpu_set_low = pset->cpu_set_hi = 0;
	pset->cpu_set_count = 0;
	pset->last_chosen = -1;
	pset->cpu_bitmask = 0;
	pset->recommended_bitmask = 0;
	pset->primary_map = 0;
	pset->realtime_map = 0;
	pset->cpu_running_foreign = 0;

	for (uint i = 0; i < PROCESSOR_STATE_LEN; i++) {
		pset->cpu_state_map[i] = 0;
	}
	pset->pending_AST_URGENT_cpu_mask = 0;
	pset->pending_AST_PREEMPT_cpu_mask = 0;
#if defined(CONFIG_SCHED_DEFERRED_AST)
	pset->pending_deferred_AST_cpu_mask = 0;
#endif
	pset->pending_spill_cpu_mask = 0;
	pset_lock_init(pset);
	pset->pset_self = IP_NULL;
	pset->pset_name_self = IP_NULL;
	pset->pset_list = PROCESSOR_SET_NULL;
	pset->node = node;

	/*
	 * The pset_cluster_type & pset_cluster_id for all psets
	 * on the platform are initialized as part of the SCHED(init).
	 * That works well for small cluster platforms; for large cluster
	 * count systems, it might be cleaner to do all the setup
	 * dynamically in SCHED(pset_init).
	 *
	 * <Edge Multi-cluster Support Needed>
	 */
	pset->is_SMT = false;

	simple_lock(&pset_node_lock, LCK_GRP_NULL);
	pset->pset_id = pset_count++;
	bit_set(node->pset_map, pset->pset_id);
	simple_unlock(&pset_node_lock);

	pset_array[pset->pset_id] = pset;
}

kern_return_t
processor_info_count(
	processor_flavor_t              flavor,
	mach_msg_type_number_t  *count)
{
	switch (flavor) {
	case PROCESSOR_BASIC_INFO:
		*count = PROCESSOR_BASIC_INFO_COUNT;
		break;

	case PROCESSOR_CPU_LOAD_INFO:
		*count = PROCESSOR_CPU_LOAD_INFO_COUNT;
		break;

	default:
		return cpu_info_count(flavor, count);
	}

	return KERN_SUCCESS;
}


kern_return_t
processor_info(
	processor_t     processor,
	processor_flavor_t              flavor,
	host_t                                  *host,
	processor_info_t                info,
	mach_msg_type_number_t  *count)
{
	int     cpu_id, state;
	kern_return_t   result;

	if (processor == PROCESSOR_NULL) {
		return KERN_INVALID_ARGUMENT;
	}

	cpu_id = processor->cpu_id;

	switch (flavor) {
	case PROCESSOR_BASIC_INFO:
	{
		processor_basic_info_t          basic_info;

		if (*count < PROCESSOR_BASIC_INFO_COUNT) {
			return KERN_FAILURE;
		}

		basic_info = (processor_basic_info_t) info;
		basic_info->cpu_type = slot_type(cpu_id);
		basic_info->cpu_subtype = slot_subtype(cpu_id);
		state = processor->state;
		if (state == PROCESSOR_OFF_LINE
#if defined(__x86_64__)
		    || !processor->is_recommended
#endif
		    ) {
			basic_info->running = FALSE;
		} else {
			basic_info->running = TRUE;
		}
		basic_info->slot_num = cpu_id;
		if (processor == master_processor) {
			basic_info->is_master = TRUE;
		} else {
			basic_info->is_master = FALSE;
		}

		*count = PROCESSOR_BASIC_INFO_COUNT;
		*host = &realhost;

		return KERN_SUCCESS;
	}

	case PROCESSOR_CPU_LOAD_INFO:
	{
		processor_cpu_load_info_t       cpu_load_info;
		timer_t         idle_state;
		uint64_t        idle_time_snapshot1, idle_time_snapshot2;
		uint64_t        idle_time_tstamp1, idle_time_tstamp2;

		/*
		 * We capture the accumulated idle time twice over
		 * the course of this function, as well as the timestamps
		 * when each were last updated. Since these are
		 * all done using non-atomic racy mechanisms, the
		 * most we can infer is whether values are stable.
		 * timer_grab() is the only function that can be
		 * used reliably on another processor's per-processor
		 * data.
		 */

		if (*count < PROCESSOR_CPU_LOAD_INFO_COUNT) {
			return KERN_FAILURE;
		}

		cpu_load_info = (processor_cpu_load_info_t) info;
		if (precise_user_kernel_time) {
			cpu_load_info->cpu_ticks[CPU_STATE_USER] =
			    (uint32_t)(timer_grab(&processor->user_state) / hz_tick_interval);
			cpu_load_info->cpu_ticks[CPU_STATE_SYSTEM] =
			    (uint32_t)(timer_grab(&processor->system_state) / hz_tick_interval);
		} else {
			uint64_t tval = timer_grab(&processor->user_state) +
			    timer_grab(&processor->system_state);

			cpu_load_info->cpu_ticks[CPU_STATE_USER] = (uint32_t)(tval / hz_tick_interval);
			cpu_load_info->cpu_ticks[CPU_STATE_SYSTEM] = 0;
		}

		idle_state = &processor->idle_state;
		idle_time_snapshot1 = timer_grab(idle_state);
		idle_time_tstamp1 = idle_state->tstamp;

		/*
		 * Idle processors are not continually updating their
		 * per-processor idle timer, so it may be extremely
		 * out of date, resulting in an over-representation
		 * of non-idle time between two measurement
		 * intervals by e.g. top(1). If we are non-idle, or
		 * have evidence that the timer is being updated
		 * concurrently, we consider its value up-to-date.
		 */
		if (processor->current_state != idle_state) {
			cpu_load_info->cpu_ticks[CPU_STATE_IDLE] =
			    (uint32_t)(idle_time_snapshot1 / hz_tick_interval);
		} else if ((idle_time_snapshot1 != (idle_time_snapshot2 = timer_grab(idle_state))) ||
		    (idle_time_tstamp1 != (idle_time_tstamp2 = idle_state->tstamp))) {
			/* Idle timer is being updated concurrently, second stamp is good enough */
			cpu_load_info->cpu_ticks[CPU_STATE_IDLE] =
			    (uint32_t)(idle_time_snapshot2 / hz_tick_interval);
		} else {
			/*
			 * Idle timer may be very stale. Fortunately we have established
			 * that idle_time_snapshot1 and idle_time_tstamp1 are unchanging
			 */
			idle_time_snapshot1 += mach_absolute_time() - idle_time_tstamp1;

			cpu_load_info->cpu_ticks[CPU_STATE_IDLE] =
			    (uint32_t)(idle_time_snapshot1 / hz_tick_interval);
		}

		cpu_load_info->cpu_ticks[CPU_STATE_NICE] = 0;

		*count = PROCESSOR_CPU_LOAD_INFO_COUNT;
		*host = &realhost;

		return KERN_SUCCESS;
	}

	default:
		result = cpu_info(flavor, cpu_id, info, count);
		if (result == KERN_SUCCESS) {
			*host = &realhost;
		}

		return result;
	}
}

kern_return_t
processor_start(
	processor_t                     processor)
{
	processor_set_t         pset;
	thread_t                        thread;
	kern_return_t           result;
	spl_t                           s;

	if (processor == PROCESSOR_NULL || processor->processor_set == PROCESSOR_SET_NULL) {
		return KERN_INVALID_ARGUMENT;
	}

	if (processor == master_processor) {
		processor_t             prev;

		prev = thread_bind(processor);
		thread_block(THREAD_CONTINUE_NULL);

		result = cpu_start(processor->cpu_id);

		thread_bind(prev);

		return result;
	}

	bool scheduler_disable = false;

	if ((processor->processor_primary != processor) && (sched_enable_smt == 0)) {
		if (cpu_can_exit(processor->cpu_id)) {
			return KERN_SUCCESS;
		}
		/*
		 * This secondary SMT processor must start in order to service interrupts,
		 * so instead it will be disabled at the scheduler level.
		 */
		scheduler_disable = true;
	}

	ml_cpu_begin_state_transition(processor->cpu_id);
	s = splsched();
	pset = processor->processor_set;
	pset_lock(pset);
	if (processor->state != PROCESSOR_OFF_LINE) {
		pset_unlock(pset);
		splx(s);
		ml_cpu_end_state_transition(processor->cpu_id);

		return KERN_FAILURE;
	}

	pset_update_processor_state(pset, processor, PROCESSOR_START);
	pset_unlock(pset);
	splx(s);

	/*
	 *	Create the idle processor thread.
	 */
	if (processor->idle_thread == THREAD_NULL) {
		result = idle_thread_create(processor);
		if (result != KERN_SUCCESS) {
			s = splsched();
			pset_lock(pset);
			pset_update_processor_state(pset, processor, PROCESSOR_OFF_LINE);
			pset_unlock(pset);
			splx(s);
			ml_cpu_end_state_transition(processor->cpu_id);

			return result;
		}
	}

	/*
	 *	If there is no active thread, the processor
	 *	has never been started.  Create a dedicated
	 *	start up thread.
	 */
	if (processor->active_thread == THREAD_NULL &&
	    processor->startup_thread == THREAD_NULL) {
		result = kernel_thread_create(processor_start_thread, NULL, MAXPRI_KERNEL, &thread);
		if (result != KERN_SUCCESS) {
			s = splsched();
			pset_lock(pset);
			pset_update_processor_state(pset, processor, PROCESSOR_OFF_LINE);
			pset_unlock(pset);
			splx(s);
			ml_cpu_end_state_transition(processor->cpu_id);

			return result;
		}

		s = splsched();
		thread_lock(thread);
		thread->bound_processor = processor;
		processor->startup_thread = thread;
		thread->state = TH_RUN;
		thread->last_made_runnable_time = mach_absolute_time();
		thread_unlock(thread);
		splx(s);

		thread_deallocate(thread);
	}

	if (processor->processor_self == IP_NULL) {
		ipc_processor_init(processor);
	}

	ml_broadcast_cpu_event(CPU_BOOT_REQUESTED, processor->cpu_id);
	result = cpu_start(processor->cpu_id);
	if (result != KERN_SUCCESS) {
		s = splsched();
		pset_lock(pset);
		pset_update_processor_state(pset, processor, PROCESSOR_OFF_LINE);
		pset_unlock(pset);
		splx(s);
		ml_cpu_end_state_transition(processor->cpu_id);

		return result;
	}
	if (scheduler_disable) {
		assert(processor->processor_primary != processor);
		sched_processor_enable(processor, FALSE);
	}

	ipc_processor_enable(processor);
	ml_cpu_end_state_transition(processor->cpu_id);
	ml_broadcast_cpu_event(CPU_ACTIVE, processor->cpu_id);

	return KERN_SUCCESS;
}


kern_return_t
processor_exit(
	processor_t     processor)
{
	if (processor == PROCESSOR_NULL) {
		return KERN_INVALID_ARGUMENT;
	}

	return processor_shutdown(processor);
}


kern_return_t
processor_start_from_user(
	processor_t                     processor)
{
	kern_return_t ret;

	if (processor == PROCESSOR_NULL) {
		return KERN_INVALID_ARGUMENT;
	}

	if (!cpu_can_exit(processor->cpu_id)) {
		ret = sched_processor_enable(processor, TRUE);
	} else {
		ret = processor_start(processor);
	}

	return ret;
}

kern_return_t
processor_exit_from_user(
	processor_t     processor)
{
	kern_return_t ret;

	if (processor == PROCESSOR_NULL) {
		return KERN_INVALID_ARGUMENT;
	}

	if (!cpu_can_exit(processor->cpu_id)) {
		ret = sched_processor_enable(processor, FALSE);
	} else {
		ret = processor_shutdown(processor);
	}

	return ret;
}

kern_return_t
enable_smt_processors(bool enable)
{
	if (machine_info.logical_cpu_max == machine_info.physical_cpu_max) {
		/* Not an SMT system */
		return KERN_INVALID_ARGUMENT;
	}

	int ncpus = machine_info.logical_cpu_max;

	for (int i = 1; i < ncpus; i++) {
		processor_t processor = processor_array[i];

		if (processor->processor_primary != processor) {
			if (enable) {
				processor_start_from_user(processor);
			} else { /* Disable */
				processor_exit_from_user(processor);
			}
		}
	}

#define BSD_HOST 1
	host_basic_info_data_t hinfo;
	mach_msg_type_number_t count = HOST_BASIC_INFO_COUNT;
	kern_return_t kret = host_info((host_t)BSD_HOST, HOST_BASIC_INFO, (host_info_t)&hinfo, &count);
	if (kret != KERN_SUCCESS) {
		return kret;
	}

	if (enable && (hinfo.logical_cpu != hinfo.logical_cpu_max)) {
		return KERN_FAILURE;
	}

	if (!enable && (hinfo.logical_cpu != hinfo.physical_cpu)) {
		return KERN_FAILURE;
	}

	return KERN_SUCCESS;
}

kern_return_t
processor_control(
	processor_t             processor,
	processor_info_t        info,
	mach_msg_type_number_t  count)
{
	if (processor == PROCESSOR_NULL) {
		return KERN_INVALID_ARGUMENT;
	}

	return cpu_control(processor->cpu_id, info, count);
}

kern_return_t
processor_set_create(
	__unused host_t         host,
	__unused processor_set_t        *new_set,
	__unused processor_set_t        *new_name)
{
	return KERN_FAILURE;
}

kern_return_t
processor_set_destroy(
	__unused processor_set_t        pset)
{
	return KERN_FAILURE;
}

kern_return_t
processor_get_assignment(
	processor_t     processor,
	processor_set_t *pset)
{
	int state;

	if (processor == PROCESSOR_NULL) {
		return KERN_INVALID_ARGUMENT;
	}

	state = processor->state;
	if (state == PROCESSOR_SHUTDOWN || state == PROCESSOR_OFF_LINE) {
		return KERN_FAILURE;
	}

	*pset = &pset0;

	return KERN_SUCCESS;
}

kern_return_t
processor_set_info(
	processor_set_t         pset,
	int                     flavor,
	host_t                  *host,
	processor_set_info_t    info,
	mach_msg_type_number_t  *count)
{
	if (pset == PROCESSOR_SET_NULL) {
		return KERN_INVALID_ARGUMENT;
	}

	if (flavor == PROCESSOR_SET_BASIC_INFO) {
		processor_set_basic_info_t      basic_info;

		if (*count < PROCESSOR_SET_BASIC_INFO_COUNT) {
			return KERN_FAILURE;
		}

		basic_info = (processor_set_basic_info_t) info;
#if defined(__x86_64__)
		basic_info->processor_count = processor_avail_count_user;
#else
		basic_info->processor_count = processor_avail_count;
#endif
		basic_info->default_policy = POLICY_TIMESHARE;

		*count = PROCESSOR_SET_BASIC_INFO_COUNT;
		*host = &realhost;
		return KERN_SUCCESS;
	} else if (flavor == PROCESSOR_SET_TIMESHARE_DEFAULT) {
		policy_timeshare_base_t ts_base;

		if (*count < POLICY_TIMESHARE_BASE_COUNT) {
			return KERN_FAILURE;
		}

		ts_base = (policy_timeshare_base_t) info;
		ts_base->base_priority = BASEPRI_DEFAULT;

		*count = POLICY_TIMESHARE_BASE_COUNT;
		*host = &realhost;
		return KERN_SUCCESS;
	} else if (flavor == PROCESSOR_SET_FIFO_DEFAULT) {
		policy_fifo_base_t              fifo_base;

		if (*count < POLICY_FIFO_BASE_COUNT) {
			return KERN_FAILURE;
		}

		fifo_base = (policy_fifo_base_t) info;
		fifo_base->base_priority = BASEPRI_DEFAULT;

		*count = POLICY_FIFO_BASE_COUNT;
		*host = &realhost;
		return KERN_SUCCESS;
	} else if (flavor == PROCESSOR_SET_RR_DEFAULT) {
		policy_rr_base_t                rr_base;

		if (*count < POLICY_RR_BASE_COUNT) {
			return KERN_FAILURE;
		}

		rr_base = (policy_rr_base_t) info;
		rr_base->base_priority = BASEPRI_DEFAULT;
		rr_base->quantum = 1;

		*count = POLICY_RR_BASE_COUNT;
		*host = &realhost;
		return KERN_SUCCESS;
	} else if (flavor == PROCESSOR_SET_TIMESHARE_LIMITS) {
		policy_timeshare_limit_t        ts_limit;

		if (*count < POLICY_TIMESHARE_LIMIT_COUNT) {
			return KERN_FAILURE;
		}

		ts_limit = (policy_timeshare_limit_t) info;
		ts_limit->max_priority = MAXPRI_KERNEL;

		*count = POLICY_TIMESHARE_LIMIT_COUNT;
		*host = &realhost;
		return KERN_SUCCESS;
	} else if (flavor == PROCESSOR_SET_FIFO_LIMITS) {
		policy_fifo_limit_t             fifo_limit;

		if (*count < POLICY_FIFO_LIMIT_COUNT) {
			return KERN_FAILURE;
		}

		fifo_limit = (policy_fifo_limit_t) info;
		fifo_limit->max_priority = MAXPRI_KERNEL;

		*count = POLICY_FIFO_LIMIT_COUNT;
		*host = &realhost;
		return KERN_SUCCESS;
	} else if (flavor == PROCESSOR_SET_RR_LIMITS) {
		policy_rr_limit_t               rr_limit;

		if (*count < POLICY_RR_LIMIT_COUNT) {
			return KERN_FAILURE;
		}

		rr_limit = (policy_rr_limit_t) info;
		rr_limit->max_priority = MAXPRI_KERNEL;

		*count = POLICY_RR_LIMIT_COUNT;
		*host = &realhost;
		return KERN_SUCCESS;
	} else if (flavor == PROCESSOR_SET_ENABLED_POLICIES) {
		int                             *enabled;

		if (*count < (sizeof(*enabled) / sizeof(int))) {
			return KERN_FAILURE;
		}

		enabled = (int *) info;
		*enabled = POLICY_TIMESHARE | POLICY_RR | POLICY_FIFO;

		*count = sizeof(*enabled) / sizeof(int);
		*host = &realhost;
		return KERN_SUCCESS;
	}


	*host = HOST_NULL;
	return KERN_INVALID_ARGUMENT;
}

/*
 *	processor_set_statistics
 *
 *	Returns scheduling statistics for a processor set.
 */
kern_return_t
processor_set_statistics(
	processor_set_t         pset,
	int                     flavor,
	processor_set_info_t    info,
	mach_msg_type_number_t  *count)
{
	if (pset == PROCESSOR_SET_NULL || pset != &pset0) {
		return KERN_INVALID_PROCESSOR_SET;
	}

	if (flavor == PROCESSOR_SET_LOAD_INFO) {
		processor_set_load_info_t     load_info;

		if (*count < PROCESSOR_SET_LOAD_INFO_COUNT) {
			return KERN_FAILURE;
		}

		load_info = (processor_set_load_info_t) info;

		load_info->mach_factor = sched_mach_factor;
		load_info->load_average = sched_load_average;

		load_info->task_count = tasks_count;
		load_info->thread_count = threads_count;

		*count = PROCESSOR_SET_LOAD_INFO_COUNT;
		return KERN_SUCCESS;
	}

	return KERN_INVALID_ARGUMENT;
}

/*
 *	processor_set_max_priority:
 *
 *	Specify max priority permitted on processor set.  This affects
 *	newly created and assigned threads.  Optionally change existing
 *      ones.
 */
kern_return_t
processor_set_max_priority(
	__unused processor_set_t        pset,
	__unused int                    max_priority,
	__unused boolean_t              change_threads)
{
	return KERN_INVALID_ARGUMENT;
}

/*
 *	processor_set_policy_enable:
 *
 *	Allow indicated policy on processor set.
 */

kern_return_t
processor_set_policy_enable(
	__unused processor_set_t        pset,
	__unused int                    policy)
{
	return KERN_INVALID_ARGUMENT;
}

/*
 *	processor_set_policy_disable:
 *
 *	Forbid indicated policy on processor set.  Time sharing cannot
 *	be forbidden.
 */
kern_return_t
processor_set_policy_disable(
	__unused processor_set_t        pset,
	__unused int                    policy,
	__unused boolean_t              change_threads)
{
	return KERN_INVALID_ARGUMENT;
}

/*
 *	processor_set_things:
 *
 *	Common internals for processor_set_{threads,tasks}
 */
static kern_return_t
processor_set_things(
	processor_set_t pset,
	void **thing_list,
	mach_msg_type_number_t *count,
	int type)
{
	unsigned int i;
	task_t task;
	thread_t thread;

	task_t *task_list;
	unsigned int actual_tasks;
	vm_size_t task_size, task_size_needed;

	thread_t *thread_list;
	unsigned int actual_threads;
	vm_size_t thread_size, thread_size_needed;

	void *addr, *newaddr;
	vm_size_t size, size_needed;

	if (pset == PROCESSOR_SET_NULL || pset != &pset0) {
		return KERN_INVALID_ARGUMENT;
	}

	task_size = 0;
	task_size_needed = 0;
	task_list = NULL;
	actual_tasks = 0;

	thread_size = 0;
	thread_size_needed = 0;
	thread_list = NULL;
	actual_threads = 0;

	for (;;) {
		lck_mtx_lock(&tasks_threads_lock);

		/* do we have the memory we need? */
		if (type == PSET_THING_THREAD) {
			thread_size_needed = threads_count * sizeof(void *);
		}
#if !CONFIG_MACF
		else
#endif
		task_size_needed = tasks_count * sizeof(void *);

		if (task_size_needed <= task_size &&
		    thread_size_needed <= thread_size) {
			break;
		}

		/* unlock and allocate more memory */
		lck_mtx_unlock(&tasks_threads_lock);

		/* grow task array */
		if (task_size_needed > task_size) {
			if (task_size != 0) {
				kfree(task_list, task_size);
			}

			assert(task_size_needed > 0);
			task_size = task_size_needed;

			task_list = (task_t *)kalloc(task_size);
			if (task_list == NULL) {
				if (thread_size != 0) {
					kfree(thread_list, thread_size);
				}
				return KERN_RESOURCE_SHORTAGE;
			}
		}

		/* grow thread array */
		if (thread_size_needed > thread_size) {
			if (thread_size != 0) {
				kfree(thread_list, thread_size);
			}

			assert(thread_size_needed > 0);
			thread_size = thread_size_needed;

			thread_list = (thread_t *)kalloc(thread_size);
			if (thread_list == 0) {
				if (task_size != 0) {
					kfree(task_list, task_size);
				}
				return KERN_RESOURCE_SHORTAGE;
			}
		}
	}

	/* OK, have memory and the list locked */

	/* If we need it, get the thread list */
	if (type == PSET_THING_THREAD) {
		for (thread = (thread_t)queue_first(&threads);
		    !queue_end(&threads, (queue_entry_t)thread);
		    thread = (thread_t)queue_next(&thread->threads)) {
#if defined(SECURE_KERNEL)
			if (thread->task != kernel_task) {
#endif
			thread_reference_internal(thread);
			thread_list[actual_threads++] = thread;
#if defined(SECURE_KERNEL)
		}
#endif
		}
	}
#if !CONFIG_MACF
	else {
#endif
	/* get a list of the tasks */
	for (task = (task_t)queue_first(&tasks);
	    !queue_end(&tasks, (queue_entry_t)task);
	    task = (task_t)queue_next(&task->tasks)) {
#if defined(SECURE_KERNEL)
		if (task != kernel_task) {
#endif
		task_reference_internal(task);
		task_list[actual_tasks++] = task;
#if defined(SECURE_KERNEL)
	}
#endif
	}
#if !CONFIG_MACF
}
#endif

	lck_mtx_unlock(&tasks_threads_lock);

#if CONFIG_MACF
	unsigned int j, used;

	/* for each task, make sure we are allowed to examine it */
	for (i = used = 0; i < actual_tasks; i++) {
		if (mac_task_check_expose_task(task_list[i])) {
			task_deallocate(task_list[i]);
			continue;
		}
		task_list[used++] = task_list[i];
	}
	actual_tasks = used;
	task_size_needed = actual_tasks * sizeof(void *);

	if (type == PSET_THING_THREAD) {
		/* for each thread (if any), make sure it's task is in the allowed list */
		for (i = used = 0; i < actual_threads; i++) {
			boolean_t found_task = FALSE;

			task = thread_list[i]->task;
			for (j = 0; j < actual_tasks; j++) {
				if (task_list[j] == task) {
					found_task = TRUE;
					break;
				}
			}
			if (found_task) {
				thread_list[used++] = thread_list[i];
			} else {
				thread_deallocate(thread_list[i]);
			}
		}
		actual_threads = used;
		thread_size_needed = actual_threads * sizeof(void *);

		/* done with the task list */
		for (i = 0; i < actual_tasks; i++) {
			task_deallocate(task_list[i]);
		}
		kfree(task_list, task_size);
		task_size = 0;
		actual_tasks = 0;
		task_list = NULL;
	}
#endif

	if (type == PSET_THING_THREAD) {
		if (actual_threads == 0) {
			/* no threads available to return */
			assert(task_size == 0);
			if (thread_size != 0) {
				kfree(thread_list, thread_size);
			}
			*thing_list = NULL;
			*count = 0;
			return KERN_SUCCESS;
		}
		size_needed = actual_threads * sizeof(void *);
		size = thread_size;
		addr = thread_list;
	} else {
		if (actual_tasks == 0) {
			/* no tasks available to return */
			assert(thread_size == 0);
			if (task_size != 0) {
				kfree(task_list, task_size);
			}
			*thing_list = NULL;
			*count = 0;
			return KERN_SUCCESS;
		}
		size_needed = actual_tasks * sizeof(void *);
		size = task_size;
		addr = task_list;
	}

	/* if we allocated too much, must copy */
	if (size_needed < size) {
		newaddr = kalloc(size_needed);
		if (newaddr == 0) {
			for (i = 0; i < actual_tasks; i++) {
				if (type == PSET_THING_THREAD) {
					thread_deallocate(thread_list[i]);
				} else {
					task_deallocate(task_list[i]);
				}
			}
			if (size) {
				kfree(addr, size);
			}
			return KERN_RESOURCE_SHORTAGE;
		}

		bcopy((void *) addr, (void *) newaddr, size_needed);
		kfree(addr, size);

		addr = newaddr;
		size = size_needed;
	}

	*thing_list = (void **)addr;
	*count = (unsigned int)size / sizeof(void *);

	return KERN_SUCCESS;
}

/*
 *	processor_set_tasks:
 *
 *	List all tasks in the processor set.
 */
static kern_return_t
processor_set_tasks_internal(
	processor_set_t         pset,
	task_array_t            *task_list,
	mach_msg_type_number_t  *count,
	int                     flavor)
{
	kern_return_t ret;
	mach_msg_type_number_t i;

	ret = processor_set_things(pset, (void **)task_list, count, PSET_THING_TASK);
	if (ret != KERN_SUCCESS) {
		return ret;
	}

	/* do the conversion that Mig should handle */
	switch (flavor) {
	case TASK_FLAVOR_CONTROL:
		for (i = 0; i < *count; i++) {
			(*task_list)[i] = (task_t)convert_task_to_port((*task_list)[i]);
		}
		break;
	case TASK_FLAVOR_READ:
		for (i = 0; i < *count; i++) {
			(*task_list)[i] = (task_t)convert_task_read_to_port((*task_list)[i]);
		}
		break;
	case TASK_FLAVOR_INSPECT:
		for (i = 0; i < *count; i++) {
			(*task_list)[i] = (task_t)convert_task_inspect_to_port((*task_list)[i]);
		}
		break;
	case TASK_FLAVOR_NAME:
		for (i = 0; i < *count; i++) {
			(*task_list)[i] = (task_t)convert_task_name_to_port((*task_list)[i]);
		}
		break;
	default:
		return KERN_INVALID_ARGUMENT;
	}

	return KERN_SUCCESS;
}

kern_return_t
processor_set_tasks(
	processor_set_t         pset,
	task_array_t            *task_list,
	mach_msg_type_number_t  *count)
{
	return processor_set_tasks_internal(pset, task_list, count, TASK_FLAVOR_CONTROL);
}

/*
 *	processor_set_tasks_with_flavor:
 *
 *	Based on flavor, return task/inspect/read port to all tasks in the processor set.
 */
kern_return_t
processor_set_tasks_with_flavor(
	processor_set_t         pset,
	mach_task_flavor_t      flavor,
	task_array_t            *task_list,
	mach_msg_type_number_t  *count)
{
	switch (flavor) {
	case TASK_FLAVOR_CONTROL:
	case TASK_FLAVOR_READ:
	case TASK_FLAVOR_INSPECT:
	case TASK_FLAVOR_NAME:
		return processor_set_tasks_internal(pset, task_list, count, flavor);
	default:
		return KERN_INVALID_ARGUMENT;
	}
}

/*
 *	processor_set_threads:
 *
 *	List all threads in the processor set.
 */
#if defined(SECURE_KERNEL)
kern_return_t
processor_set_threads(
	__unused processor_set_t                pset,
	__unused thread_array_t         *thread_list,
	__unused mach_msg_type_number_t *count)
{
	return KERN_FAILURE;
}
#elif !defined(XNU_TARGET_OS_OSX)
kern_return_t
processor_set_threads(
	__unused processor_set_t                pset,
	__unused thread_array_t         *thread_list,
	__unused mach_msg_type_number_t *count)
{
	return KERN_NOT_SUPPORTED;
}
#else
kern_return_t
processor_set_threads(
	processor_set_t         pset,
	thread_array_t          *thread_list,
	mach_msg_type_number_t  *count)
{
	kern_return_t ret;
	mach_msg_type_number_t i;

	ret = processor_set_things(pset, (void **)thread_list, count, PSET_THING_THREAD);
	if (ret != KERN_SUCCESS) {
		return ret;
	}

	/* do the conversion that Mig should handle */
	for (i = 0; i < *count; i++) {
		(*thread_list)[i] = (thread_t)convert_thread_to_port((*thread_list)[i]);
	}
	return KERN_SUCCESS;
}
#endif

/*
 *	processor_set_policy_control
 *
 *	Controls the scheduling attributes governing the processor set.
 *	Allows control of enabled policies, and per-policy base and limit
 *	priorities.
 */
kern_return_t
processor_set_policy_control(
	__unused processor_set_t                pset,
	__unused int                            flavor,
	__unused processor_set_info_t   policy_info,
	__unused mach_msg_type_number_t count,
	__unused boolean_t                      change)
{
	return KERN_INVALID_ARGUMENT;
}

#undef pset_deallocate
void pset_deallocate(processor_set_t pset);
void
pset_deallocate(
	__unused processor_set_t        pset)
{
	return;
}

#undef pset_reference
void pset_reference(processor_set_t pset);
void
pset_reference(
	__unused processor_set_t        pset)
{
	return;
}

#if CONFIG_THREAD_GROUPS

pset_cluster_type_t
thread_group_pset_recommendation(__unused struct thread_group *tg, __unused cluster_type_t recommendation)
{
#if __AMP__
	switch (recommendation) {
	case CLUSTER_TYPE_SMP:
	default:
		/*
		 * In case of SMP recommendations, check if the thread
		 * group has special flags which restrict it to the E
		 * cluster.
		 */
		if (thread_group_smp_restricted(tg)) {
			return PSET_AMP_E;
		}
		return PSET_AMP_P;
	case CLUSTER_TYPE_E:
		return PSET_AMP_E;
	case CLUSTER_TYPE_P:
		return PSET_AMP_P;
	}
#else /* __AMP__ */
	return PSET_SMP;
#endif /* __AMP__ */
}

#endif

pset_cluster_type_t
recommended_pset_type(thread_t thread)
{
#if CONFIG_THREAD_GROUPS && __AMP__
	if (thread == THREAD_NULL) {
		return PSET_AMP_E;
	}

	if (thread->sched_flags & TH_SFLAG_ECORE_ONLY) {
		return PSET_AMP_E;
	} else if (thread->sched_flags & TH_SFLAG_PCORE_ONLY) {
		return PSET_AMP_P;
	}

	if (thread->base_pri <= MAXPRI_THROTTLE) {
		if (os_atomic_load(&sched_perfctl_policy_bg, relaxed) != SCHED_PERFCTL_POLICY_FOLLOW_GROUP) {
			return PSET_AMP_E;
		}
	} else if (thread->base_pri <= BASEPRI_UTILITY) {
		if (os_atomic_load(&sched_perfctl_policy_util, relaxed) != SCHED_PERFCTL_POLICY_FOLLOW_GROUP) {
			return PSET_AMP_E;
		}
	}

#if DEVELOPMENT || DEBUG
	extern bool system_ecore_only;
	extern processor_set_t pcore_set;
	if (system_ecore_only) {
		if (thread->task->pset_hint == pcore_set) {
			return PSET_AMP_P;
		}
		return PSET_AMP_E;
	}
#endif

	struct thread_group *tg = thread_group_get(thread);
	cluster_type_t recommendation = thread_group_recommendation(tg);
	switch (recommendation) {
	case CLUSTER_TYPE_SMP:
	default:
		if (thread->task == kernel_task) {
			return PSET_AMP_E;
		}
		return PSET_AMP_P;
	case CLUSTER_TYPE_E:
		return PSET_AMP_E;
	case CLUSTER_TYPE_P:
		return PSET_AMP_P;
	}
#else
	(void)thread;
	return PSET_SMP;
#endif
}

#if CONFIG_THREAD_GROUPS && __AMP__

void
sched_perfcontrol_inherit_recommendation_from_tg(perfcontrol_class_t perfctl_class, boolean_t inherit)
{
	sched_perfctl_class_policy_t sched_policy = inherit ? SCHED_PERFCTL_POLICY_FOLLOW_GROUP : SCHED_PERFCTL_POLICY_RESTRICT_E;

	KDBG(MACHDBG_CODE(DBG_MACH_SCHED, MACH_AMP_PERFCTL_POLICY_CHANGE) | DBG_FUNC_NONE, perfctl_class, sched_policy, 0, 0);

	switch (perfctl_class) {
	case PERFCONTROL_CLASS_UTILITY:
		os_atomic_store(&sched_perfctl_policy_util, sched_policy, relaxed);
		break;
	case PERFCONTROL_CLASS_BACKGROUND:
		os_atomic_store(&sched_perfctl_policy_bg, sched_policy, relaxed);
		break;
	default:
		panic("perfctl_class invalid");
		break;
	}
}

#elif defined(__arm64__)

/* Define a stub routine since this symbol is exported on all arm64 platforms */
void
sched_perfcontrol_inherit_recommendation_from_tg(__unused perfcontrol_class_t perfctl_class, __unused boolean_t inherit)
{
}

#endif /* defined(__arm64__) */