]> git.saurik.com Git - apple/xnu.git/blobdiff - osfmk/kern/processor.c
xnu-4903.270.47.tar.gz
[apple/xnu.git] / osfmk / kern / processor.c
index 6a1b9e45998aae7b59f67e4aec5b561c843b138b..486efc100f6589b8ec95415243c3611879b34361 100644 (file)
@@ -1,49 +1,55 @@
 /*
- * Copyright (c) 2000 Apple Computer, Inc. All rights reserved.
+ * Copyright (c) 2000-2009 Apple Inc. All rights reserved.
  *
- * @APPLE_LICENSE_HEADER_START@
- * 
- * The contents of this file constitute Original Code as defined in and
- * are subject to the Apple Public Source License Version 1.1 (the
- * "License").  You may not use this file except in compliance with the
- * License.  Please obtain a copy of the License at
- * http://www.apple.com/publicsource and read it before using this file.
- * 
- * This Original Code and all software distributed under the License are
- * distributed on an "AS IS" basis, WITHOUT WARRANTY OF ANY KIND, EITHER
+ * @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 OR NON-INFRINGEMENT.  Please see the
- * License for the specific language governing rights and limitations
- * under the License.
- * 
- * @APPLE_LICENSE_HEADER_END@
+ * 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 <cpus.h>
-
 #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 <ipc/ipc_port.h>
 #include <kern/kalloc.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>
 
-/*
- *     Exported variables.
- */
-struct processor_set default_pset;
-struct processor processor_array[NCPUS];
-
-int            master_cpu = 0;
+struct processor_set    pset0;
+struct pset_node                pset_node0;
+decl_simple_lock_data(static, pset_node_lock)
 
-processor_t    master_processor;
-processor_t    processor_ptr[NCPUS];
+lck_grp_t pset_lck_grp;
 
-/* Forwards */
-void   pset_init(
-               processor_set_t 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;
+decl_lck_mtx_data(, tasks_threads_lock)
+decl_lck_mtx_data(, tasks_corpse_lock)
 
-void   processor_init(
-               register processor_t    pr,
-               int                     slot_num);
+processor_t                             processor_list;
+unsigned int                    processor_count;
+static processor_t              processor_list_tail;
+decl_simple_lock_data(, processor_list_lock)
 
-void   pset_quanta_setup(
-               processor_set_t         pset);
+uint32_t                                processor_avail_count;
+uint32_t                                processor_avail_count_user;
 
-kern_return_t  processor_set_base(
-               processor_set_t         pset,
-               policy_t                policy,
-               policy_base_t           base,
-               boolean_t               change);
+processor_t             master_processor;
+int                     master_cpu = 0;
+boolean_t               sched_stats_active = FALSE;
 
-kern_return_t  processor_set_limit(
-               processor_set_t         pset,
-               policy_t                policy,
-               policy_limit_t          limit,
-               boolean_t               change);
+processor_t             processor_array[MAX_SCHED_CPUS] = { 0 };
 
-kern_return_t  processor_set_things(
-               processor_set_t         pset,
-               mach_port_t             **thing_list,
-               mach_msg_type_number_t  *count,
-               int                     type);
+#if defined(CONFIG_XNUPOST)
+kern_return_t ipi_test(void);
+extern void arm64_ipi_test(void);
 
-
-/*
- *     Bootstrap the processor/pset system so the scheduler can run.
- */
-void
-pset_sys_bootstrap(void)
+kern_return_t
+ipi_test()
 {
-       register int    i;
-
-       pset_init(&default_pset);
+#if __arm64__
+       processor_t p;
 
-       for (i = 0; i < NCPUS; i++) {
-               /*
-                *      Initialize processor data structures.
-                *      Note that cpu_to_processor(i) is processor_ptr[i].
-                */
-               processor_ptr[i] = &processor_array[i];
-               processor_init(processor_ptr[i], i);
+       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();
        }
 
-       master_processor = cpu_to_processor(master_cpu);
+       /* 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__ */
 
-       default_pset.active = TRUE;
+       return KERN_SUCCESS;
 }
+#endif /* defined(CONFIG_XNUPOST) */
 
-/*
- *     Initialize the given processor_set structure.
- */
+int sched_enable_smt = 1;
 
-void pset_init(
-       register processor_set_t        pset)
+void
+processor_bootstrap(void)
 {
-       register int    i;
-
-       /* setup run queue */
-       pset->runq.highq = IDLEPRI;
-       for (i = 0; i < NRQBM; i++)
-           pset->runq.bitmap[i] = 0;
-       setbit(MAXPRI - IDLEPRI, pset->runq.bitmap); 
-       pset->runq.urgency = pset->runq.count = 0;
-       for (i = 0; i < NRQS; i++)
-           queue_init(&pset->runq.queues[i]);
-
-       queue_init(&pset->idle_queue);
-       pset->idle_count = 0;
-       queue_init(&pset->active_queue);
-       simple_lock_init(&pset->sched_lock, ETAP_THREAD_PSET_IDLE);
-       pset->run_count = pset->share_count = 0;
-       pset->mach_factor = pset->load_average = 0;
-       pset->sched_load = 0;
-       queue_init(&pset->processors);
-       pset->processor_count = 0;
-       queue_init(&pset->tasks);
-       pset->task_count = 0;
-       queue_init(&pset->threads);
-       pset->thread_count = 0;
-       pset->ref_count = 1;
-       pset->active = FALSE;
-       mutex_init(&pset->lock, ETAP_THREAD_PSET);
-       pset->pset_self = IP_NULL;
-       pset->pset_name_self = IP_NULL;
-       pset->timeshare_quanta = 1;
+       lck_grp_init(&pset_lck_grp, "pset", LCK_GRP_ATTR_NULL);
 
-       for (i = 0; i <= NCPUS; i++)
-           pset->quantum_factors[i] = 1;
+       simple_lock_init(&pset_node_lock, 0);
+
+       pset_node0.psets = &pset0;
+       pset_init(&pset0, &pset_node0);
+
+       queue_init(&tasks);
+       queue_init(&terminated_tasks);
+       queue_init(&threads);
+       queue_init(&corpse_tasks);
+
+       simple_lock_init(&processor_list_lock, 0);
+
+       master_processor = cpu_to_processor(master_cpu);
+
+       processor_init(master_processor, master_cpu, &pset0);
 }
 
 /*
- *     Initialize the given processor structure for the processor in
- *     the slot specified by slot_num.
+ *     Initialize the given processor for the cpu
+ *     indicated by cpu_id, and assign to the
+ *     specified processor set.
  */
 void
 processor_init(
-       register processor_t    p,
-       int                                             slot_num)
+       processor_t                     processor,
+       int                                     cpu_id,
+       processor_set_t         pset)
 {
-       register int    i;
-
-       /* setup run queue */
-       p->runq.highq = IDLEPRI;
-       for (i = 0; i < NRQBM; i++)
-           p->runq.bitmap[i] = 0;
-       setbit(MAXPRI - IDLEPRI, p->runq.bitmap); 
-       p->runq.urgency = p->runq.count = 0;
-       for (i = 0; i < NRQS; i++)
-           queue_init(&p->runq.queues[i]);
-
-       p->state = PROCESSOR_OFF_LINE;
-       p->active_thread = p->next_thread = p->idle_thread = THREAD_NULL;
-       p->processor_set = PROCESSOR_SET_NULL;
-       p->current_pri = MINPRI;
-       timer_call_setup(&p->quantum_timer, thread_quantum_expire, p);
-       p->timeslice = 0;
-       p->deadline = UINT64_MAX;
-       simple_lock_init(&p->lock, ETAP_THREAD_PROC);
-       p->processor_self = IP_NULL;
-       p->slot_num = slot_num;
-}
+       spl_t           s;
 
-/*
- *     pset_deallocate:
- *
- *     Remove one reference to the processor set.  Destroy processor_set
- *     if this was the last reference.
- */
-void
-pset_deallocate(
-       processor_set_t pset)
-{
-       if (pset == PROCESSOR_SET_NULL)
-               return;
+       if (processor != master_processor) {
+               /* Scheduler state for master_processor initialized in sched_init() */
+               SCHED(processor_init)(processor);
+       }
 
-       assert(pset == &default_pset);
-       return;
+       assert(cpu_id < MAX_SCHED_CPUS);
+
+       processor->state = PROCESSOR_OFF_LINE;
+       processor->active_thread = processor->next_thread = processor->idle_thread = THREAD_NULL;
+       processor->processor_set = pset;
+       processor_state_update_idle(processor);
+       processor->starting_pri = MINPRI;
+       processor->cpu_id = cpu_id;
+       timer_call_setup(&processor->quantum_timer, thread_quantum_expire, processor);
+       processor->quantum_end = UINT64_MAX;
+       processor->deadline = UINT64_MAX;
+       processor->first_timeslice = FALSE;
+       processor->processor_primary = processor; /* no SMT relationship known at this point */
+       processor->processor_secondary = NULL;
+       processor->is_SMT = FALSE;
+       processor->is_recommended = (pset->recommended_bitmask & (1ULL << cpu_id)) ? TRUE : FALSE;
+       processor->processor_self = IP_NULL;
+       processor_data_init(processor);
+       processor->processor_list = NULL;
+       processor->cpu_quiesce_state = CPU_QUIESCE_COUNTER_NONE;
+       processor->cpu_quiesce_last_checkin = 0;
+       processor->must_idle = false;
+
+       s = splsched();
+       pset_lock(pset);
+       bit_set(pset->cpu_bitmask, 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);
 }
 
-/*
- *     pset_reference:
- *
- *     Add one reference to the processor set.
- */
 void
-pset_reference(
-       processor_set_t pset)
+processor_set_primary(
+       processor_t             processor,
+       processor_t             primary)
 {
-       assert(pset == &default_pset);
+       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;
+
+               processor_set_t pset = processor->processor_set;
+               spl_t s = splsched();
+               pset_lock(pset);
+               bit_clear(pset->primary_map, processor->cpu_id);
+               pset_unlock(pset);
+               splx(s);
+       }
 }
 
-#define pset_reference_locked(pset) assert(pset == &default_pset)
+processor_set_t
+processor_pset(
+       processor_t     processor)
+{
+       return processor->processor_set;
+}
 
-/*
- *     pset_remove_processor() removes a processor from a processor_set.
- *     It can only be called on the current processor.  Caller must
- *     hold lock on current processor and processor set.
- */
 void
-pset_remove_processor(
-       processor_set_t pset,
-       processor_t     processor)
+processor_state_update_idle(processor_t processor)
 {
-       if (pset != processor->processor_set)
-               panic("pset_remove_processor: wrong pset");
-
-       queue_remove(&pset->processors, processor, processor_t, processors);
-       processor->processor_set = PROCESSOR_SET_NULL;
-       pset->processor_count--;
-       pset_quanta_setup(pset);
+       processor->current_pri = IDLEPRI;
+       processor->current_sfi_class = SFI_CLASS_KERNEL;
+       processor->current_recommended_pset_type = PSET_SMP;
+       processor->current_perfctl_class = PERFCONTROL_CLASS_IDLE;
+       processor->current_urgency = THREAD_URGENCY_NONE;
+       processor->current_is_NO_SMT = false;
+       processor->current_is_bound = false;
 }
 
-/*
- *     pset_add_processor() adds a  processor to a processor_set.
- *     It can only be called on the current processor.  Caller must
- *     hold lock on curent processor and on pset.  No reference counting on
- *     processors.  Processor reference to pset is implicit.
- */
 void
-pset_add_processor(
-       processor_set_t pset,
-       processor_t     processor)
+processor_state_update_from_thread(processor_t processor, thread_t thread)
 {
-       queue_enter(&pset->processors, processor, processor_t, processors);
-       processor->processor_set = pset;
-       pset->processor_count++;
-       pset_quanta_setup(pset);
+       processor->current_pri = thread->sched_pri;
+       processor->current_sfi_class = thread->sfi_class;
+       processor->current_recommended_pset_type = recommended_pset_type(thread);
+       processor->current_perfctl_class = thread_get_perfcontrol_class(thread);
+       processor->current_urgency = thread_get_urgency(thread, NULL, NULL);
+#if DEBUG || DEVELOPMENT
+       processor->current_is_NO_SMT = (thread->sched_flags & TH_SFLAG_NO_SMT) || (thread->task->t_flags & TF_NO_SMT);
+#else
+       processor->current_is_NO_SMT = (thread->sched_flags & TH_SFLAG_NO_SMT);
+#endif
+       processor->current_is_bound = thread->bound_processor != PROCESSOR_NULL;
 }
 
-/*
- *     pset_remove_task() removes a task from a processor_set.
- *     Caller must hold locks on pset and task (unless task has
- *     no references left, in which case just the pset lock is
- *     needed).  Pset reference count is not decremented;
- *     caller must explicitly pset_deallocate.
- */
 void
-pset_remove_task(
-       processor_set_t pset,
-       task_t          task)
+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)
 {
-       if (pset != task->processor_set)
-               return;
-
-       queue_remove(&pset->tasks, task, task_t, pset_tasks);
-       task->processor_set = PROCESSOR_SET_NULL;
-       pset->task_count--;
+       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;
 }
 
-/*
- *     pset_add_task() adds a  task to a processor_set.
- *     Caller must hold locks on pset and task.  Pset references to
- *     tasks are implicit.
- */
-void
-pset_add_task(
-       processor_set_t pset,
-       task_t          task)
+pset_node_t
+pset_node_root(void)
 {
-       queue_enter(&pset->tasks, task, task_t, pset_tasks);
-       task->processor_set = pset;
-       pset->task_count++;
-       pset_reference_locked(pset);
+       return &pset_node0;
 }
 
-/*
- *     pset_remove_thread() removes a thread from a processor_set.
- *     Caller must hold locks on pset and thread (but only if thread
- *  has outstanding references that could be used to lookup the pset).
- *  The pset reference count is not decremented; caller must explicitly
- *  pset_deallocate.
- */
-void
-pset_remove_thread(
-       processor_set_t pset,
-       thread_t        thread)
+processor_set_t
+pset_create(
+       pset_node_t                     node)
 {
-       queue_remove(&pset->threads, thread, thread_t, pset_threads);
-       thread->processor_set = PROCESSOR_SET_NULL;
-       pset->thread_count--;
+       /* some schedulers do not support multiple psets */
+       if (SCHED(multiple_psets_enabled) == FALSE) {
+               return processor_pset(master_processor);
+       }
+
+       processor_set_t         *prev, pset = kalloc(sizeof(*pset));
+
+       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;
 }
 
 /*
- *     pset_add_thread() adds a  thread to a processor_set.
- *     Caller must hold locks on pset and thread.  Pset references to
- *     threads are implicit.
+ *     Find processor set in specified node with specified cluster_id.
+ *     Returns default_pset if not found.
  */
-void
-pset_add_thread(
-       processor_set_t pset,
-       thread_t        thread)
+processor_set_t
+pset_find(
+       uint32_t cluster_id,
+       processor_set_t default_pset)
 {
-       queue_enter(&pset->threads, thread, thread_t, pset_threads);
-       thread->processor_set = pset;
-       pset->thread_count++;
-       pset_reference_locked(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 ((node = node->node_list) != NULL);
+       simple_unlock(&pset_node_lock);
+       if (pset == NULL) {
+               return default_pset;
+       }
+       return pset;
 }
 
 /*
- *     thread_change_psets() changes the pset of a thread.  Caller must
- *     hold locks on both psets and thread.  The old pset must be
- *     explicitly pset_deallocat()'ed by caller.
+ *     Initialize the given processor_set structure.
  */
 void
-thread_change_psets(
-       thread_t        thread,
-       processor_set_t old_pset,
-       processor_set_t new_pset)
+pset_init(
+       processor_set_t         pset,
+       pset_node_t                     node)
 {
-       queue_remove(&old_pset->threads, thread, thread_t, pset_threads);
-       old_pset->thread_count--;
-       queue_enter(&new_pset->threads, thread, thread_t, pset_threads);
-       thread->processor_set = new_pset;
-       new_pset->thread_count++;
-       pset_reference_locked(new_pset);
-}      
+       if (pset != &pset0) {
+               /* Scheduler state for pset0 initialized in sched_init() */
+               SCHED(pset_init)(pset);
+               SCHED(rt_init)(pset);
+       }
 
+       pset->online_processor_count = 0;
+       pset->load_average = 0;
+       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 = ~0ULL;
+       pset->primary_map = ~0ULL;
+       pset->cpu_state_map[PROCESSOR_OFF_LINE] = ~0ULL;
+       for (uint i = PROCESSOR_SHUTDOWN; 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;
+       pset->pset_cluster_type = PSET_SMP;
+       pset->pset_cluster_id = 0;
+
+       simple_lock(&pset_node_lock, LCK_GRP_NULL);
+       node->pset_count++;
+       simple_unlock(&pset_node_lock);
+}
 
 kern_return_t
 processor_info_count(
-       processor_flavor_t      flavor,
-       mach_msg_type_number_t  *count)
+       processor_flavor_t              flavor,
+       mach_msg_type_number_t  *count)
 {
-       kern_return_t           kr;
-
        switch (flavor) {
        case PROCESSOR_BASIC_INFO:
                *count = PROCESSOR_BASIC_INFO_COUNT;
-               return KERN_SUCCESS;
+               break;
+
        case PROCESSOR_CPU_LOAD_INFO:
                *count = PROCESSOR_CPU_LOAD_INFO_COUNT;
-               return KERN_SUCCESS;
+               break;
+
        default:
-               kr = cpu_info_count(flavor, count);
-               return kr;
+               return cpu_info_count(flavor, count);
        }
+
+       return KERN_SUCCESS;
 }
 
 
 kern_return_t
 processor_info(
-       register processor_t    processor,
-       processor_flavor_t      flavor,
-       host_t                  *host,
-       processor_info_t        info,
-       mach_msg_type_number_t  *count)
+       processor_t     processor,
+       processor_flavor_t              flavor,
+       host_t                                  *host,
+       processor_info_t                info,
+       mach_msg_type_number_t  *count)
 {
-       register int    i, slot_num, state;
-       register processor_basic_info_t         basic_info;
-       register processor_cpu_load_info_t      cpu_load_info;
-       kern_return_t   kr;
+       int     cpu_id, state;
+       kern_return_t   result;
 
-       if (processor == PROCESSOR_NULL)
-               return(KERN_INVALID_ARGUMENT);
+       if (processor == PROCESSOR_NULL) {
+               return KERN_INVALID_ARGUMENT;
+       }
 
-       slot_num = processor->slot_num;
+       cpu_id = processor->cpu_id;
 
        switch (flavor) {
-
        case PROCESSOR_BASIC_INFO:
-         {
-           if (*count < PROCESSOR_BASIC_INFO_COUNT)
-             return(KERN_FAILURE);
-
-           basic_info = (processor_basic_info_t) info;
-           basic_info->cpu_type = machine_slot[slot_num].cpu_type;
-           basic_info->cpu_subtype = machine_slot[slot_num].cpu_subtype;
-           state = processor->state;
-           if (state == PROCESSOR_OFF_LINE)
-             basic_info->running = FALSE;
-           else
-             basic_info->running = TRUE;
-           basic_info->slot_num = slot_num;
-           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);
-         }
+       {
+               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:
-         {
-           if (*count < PROCESSOR_CPU_LOAD_INFO_COUNT)
-             return(KERN_FAILURE);
-
-           cpu_load_info = (processor_cpu_load_info_t) info;
-           for (i=0;i<CPU_STATE_MAX;i++)
-             cpu_load_info->cpu_ticks[i] = machine_slot[slot_num].cpu_ticks[i];
-
-           *count = PROCESSOR_CPU_LOAD_INFO_COUNT;
-           *host = &realhost;
-           return(KERN_SUCCESS);
-         }
+       {
+               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_DATA(processor, user_state)) / hz_tick_interval);
+                       cpu_load_info->cpu_ticks[CPU_STATE_SYSTEM] =
+                           (uint32_t)(timer_grab(&PROCESSOR_DATA(processor, system_state)) / hz_tick_interval);
+               } else {
+                       uint64_t tval = timer_grab(&PROCESSOR_DATA(processor, user_state)) +
+                           timer_grab(&PROCESSOR_DATA(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_DATA(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_DATA(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:
-         {
-           kr=cpu_info(flavor, slot_num, info, count);
-           if (kr == KERN_SUCCESS)
-               *host = &realhost;                 
-           return(kr);
-         }
+               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_t                     processor)
 {
-       kern_return_t   result;
-       spl_t                   s;
+       processor_set_t         pset;
+       thread_t                        thread;
+       kern_return_t           result;
+       spl_t                           s;
 
-       if (processor == PROCESSOR_NULL)
-               return(KERN_INVALID_ARGUMENT);
+       if (processor == PROCESSOR_NULL || processor->processor_set == PROCESSOR_SET_NULL) {
+               return KERN_INVALID_ARGUMENT;
+       }
 
        if (processor == master_processor) {
-               processor_t             prev;
+               processor_t             prev;
 
-               prev = thread_bind(current_thread(), processor);
+               prev = thread_bind(processor);
                thread_block(THREAD_CONTINUE_NULL);
 
-               result = cpu_start(processor->slot_num);
+               result = cpu_start(processor->cpu_id);
+
+               thread_bind(prev);
+
+               return result;
+       }
 
-               thread_bind(current_thread(), prev);
+       bool scheduler_disable = false;
 
-               return (result);
+       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;
        }
 
        s = splsched();
-       processor_lock(processor);
+       pset = processor->processor_set;
+       pset_lock(pset);
        if (processor->state != PROCESSOR_OFF_LINE) {
-               processor_unlock(processor);
+               pset_unlock(pset);
                splx(s);
 
-               return (KERN_FAILURE);
+               return KERN_FAILURE;
        }
 
-       processor->state = PROCESSOR_START;
-       processor_unlock(processor);
+       pset_update_processor_state(pset, processor, PROCESSOR_START);
+       pset_unlock(pset);
        splx(s);
 
-       if (processor->next_thread == THREAD_NULL) {
-               thread_t                thread;   
-               extern void             start_cpu_thread(void);
-       
-               thread = kernel_thread_create(start_cpu_thread, MAXPRI_KERNEL);
+       /*
+        *      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);
+
+                       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->next_thread == THREAD_NULL) {
+               result = kernel_thread_create((thread_continue_t)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);
+
+                       return result;
+               }
 
                s = splsched();
                thread_lock(thread);
                thread->bound_processor = processor;
                processor->next_thread = thread;
                thread->state = TH_RUN;
-               pset_run_incr(thread->processor_set);
+               thread->last_made_runnable_time = mach_absolute_time();
                thread_unlock(thread);
                splx(s);
+
+               thread_deallocate(thread);
        }
 
-       if (processor->processor_self == IP_NULL)
+       if (processor->processor_self == IP_NULL) {
                ipc_processor_init(processor);
+       }
 
-       result = cpu_start(processor->slot_num);
+       result = cpu_start(processor->cpu_id);
        if (result != KERN_SUCCESS) {
                s = splsched();
-               processor_lock(processor);
-               processor->state = PROCESSOR_OFF_LINE;
-               processor_unlock(processor);
+               pset_lock(pset);
+               pset_update_processor_state(pset, processor, PROCESSOR_OFF_LINE);
+               pset_unlock(pset);
                splx(s);
 
-               return (result);
+               return result;
+       }
+       if (scheduler_disable) {
+               assert(processor->processor_primary != processor);
+               sched_processor_enable(processor, FALSE);
        }
 
        ipc_processor_enable(processor);
 
-       return (KERN_SUCCESS);
+       return KERN_SUCCESS;
 }
 
+
 kern_return_t
 processor_exit(
-       processor_t     processor)
+       processor_t     processor)
 {
-       if (processor == PROCESSOR_NULL)
-               return(KERN_INVALID_ARGUMENT);
+       if (processor == PROCESSOR_NULL) {
+               return KERN_INVALID_ARGUMENT;
+       }
 
-       return(processor_shutdown(processor));
+       return processor_shutdown(processor);
 }
 
+
 kern_return_t
-processor_control(
-       processor_t             processor,
-       processor_info_t        info,
-       mach_msg_type_number_t  count)
+processor_start_from_user(
+       processor_t                     processor)
 {
-       if (processor == PROCESSOR_NULL)
-               return(KERN_INVALID_ARGUMENT);
+       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(cpu_control(processor->slot_num, info, count));
+       return ret;
 }
 
-/*
- *     Precalculate the appropriate timesharing quanta based on load.  The
- *     index into quantum_factors[] is the number of threads on the
- *     processor set queue.  It is limited to the number of processors in
- *     the set.
- */
+kern_return_t
+processor_exit_from_user(
+       processor_t     processor)
+{
+       kern_return_t ret;
 
-void
-pset_quanta_setup(
-       processor_set_t         pset)
+       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)
 {
-       register int    i, count = pset->processor_count;
+       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;
+       }
 
-       for (i = 1; i <= count; i++)
-               pset->quantum_factors[i] = (count + (i / 2)) / i;
+       if (!enable && (hinfo.logical_cpu != hinfo.physical_cpu)) {
+               return KERN_FAILURE;
+       }
+
+       return KERN_SUCCESS;
+}
 
-       pset->quantum_factors[0] = pset->quantum_factors[1];
+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;
+       }
 
-       timeshare_quanta_update(pset);
+       return cpu_control(processor->cpu_id, info, count);
 }
-           
+
 kern_return_t
 processor_set_create(
-       host_t  host,
-       processor_set_t *new_set,
-       processor_set_t *new_name)
+       __unused host_t         host,
+       __unused processor_set_t        *new_set,
+       __unused processor_set_t        *new_name)
 {
-#ifdef lint
-       host++; new_set++; new_name++;
-#endif /* lint */
-       return(KERN_FAILURE);
+       return KERN_FAILURE;
 }
 
 kern_return_t
 processor_set_destroy(
-       processor_set_t pset)
+       __unused processor_set_t        pset)
 {
-#ifdef lint
-       pset++;
-#endif /* lint */
-       return(KERN_FAILURE);
+       return KERN_FAILURE;
 }
 
 kern_return_t
 processor_get_assignment(
-       processor_t     processor,
-       processor_set_t *pset)
+       processor_t     processor,
+       processor_set_t *pset)
 {
-       int state;
+       int state;
+
+       if (processor == PROCESSOR_NULL) {
+               return KERN_INVALID_ARGUMENT;
+       }
 
        state = processor->state;
-       if (state == PROCESSOR_SHUTDOWN || state == PROCESSOR_OFF_LINE)
-               return(KERN_FAILURE);
+       if (state == PROCESSOR_SHUTDOWN || state == PROCESSOR_OFF_LINE) {
+               return KERN_FAILURE;
+       }
+
+       *pset = &pset0;
 
-       *pset = processor->processor_set;
-       pset_reference(*pset);
-       return(KERN_SUCCESS);
+       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)
+       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 (pset == PROCESSOR_SET_NULL) {
+               return KERN_INVALID_ARGUMENT;
+       }
 
        if (flavor == PROCESSOR_SET_BASIC_INFO) {
-               register processor_set_basic_info_t     basic_info;
+               processor_set_basic_info_t      basic_info;
 
-               if (*count < PROCESSOR_SET_BASIC_INFO_COUNT)
-                       return(KERN_FAILURE);
+               if (*count < PROCESSOR_SET_BASIC_INFO_COUNT) {
+                       return KERN_FAILURE;
+               }
 
                basic_info = (processor_set_basic_info_t) info;
-               basic_info->processor_count = pset->processor_count;
+#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) {
-               register policy_timeshare_base_t        ts_base;
+               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);
+               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) {
-               register policy_fifo_base_t             fifo_base;
+               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);
+               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) {
-               register policy_rr_base_t               rr_base;
+               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);
+               if (*count < POLICY_RR_BASE_COUNT) {
+                       return KERN_FAILURE;
+               }
 
                rr_base = (policy_rr_base_t) info;
                rr_base->base_priority = BASEPRI_DEFAULT;
@@ -675,101 +921,102 @@ processor_set_info(
 
                *count = POLICY_RR_BASE_COUNT;
                *host = &realhost;
-               return(KERN_SUCCESS);
-       }
-       else if (flavor == PROCESSOR_SET_TIMESHARE_LIMITS) {
-               register policy_timeshare_limit_t       ts_limit;
+               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);
+               if (*count < POLICY_TIMESHARE_LIMIT_COUNT) {
+                       return KERN_FAILURE;
+               }
 
                ts_limit = (policy_timeshare_limit_t) info;
-               ts_limit->max_priority = MAXPRI_STANDARD;
+               ts_limit->max_priority = MAXPRI_KERNEL;
 
                *count = POLICY_TIMESHARE_LIMIT_COUNT;
                *host = &realhost;
-               return(KERN_SUCCESS);
-       }
-       else if (flavor == PROCESSOR_SET_FIFO_LIMITS) {
-               register policy_fifo_limit_t            fifo_limit;
+               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);
+               if (*count < POLICY_FIFO_LIMIT_COUNT) {
+                       return KERN_FAILURE;
+               }
 
                fifo_limit = (policy_fifo_limit_t) info;
-               fifo_limit->max_priority = MAXPRI_STANDARD;
+               fifo_limit->max_priority = MAXPRI_KERNEL;
 
                *count = POLICY_FIFO_LIMIT_COUNT;
                *host = &realhost;
-               return(KERN_SUCCESS);
-       }
-       else if (flavor == PROCESSOR_SET_RR_LIMITS) {
-               register policy_rr_limit_t              rr_limit;
+               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);
+               if (*count < POLICY_RR_LIMIT_COUNT) {
+                       return KERN_FAILURE;
+               }
 
                rr_limit = (policy_rr_limit_t) info;
-               rr_limit->max_priority = MAXPRI_STANDARD;
+               rr_limit->max_priority = MAXPRI_KERNEL;
 
                *count = POLICY_RR_LIMIT_COUNT;
                *host = &realhost;
-               return(KERN_SUCCESS);
-       }
-       else if (flavor == PROCESSOR_SET_ENABLED_POLICIES) {
-               register int                            *enabled;
+               return KERN_SUCCESS;
+       } else if (flavor == PROCESSOR_SET_ENABLED_POLICIES) {
+               int                             *enabled;
 
-               if (*count < (sizeof(*enabled)/sizeof(int)))
-                       return(KERN_FAILURE);
+               if (*count < (sizeof(*enabled) / sizeof(int))) {
+                       return KERN_FAILURE;
+               }
 
                enabled = (int *) info;
                *enabled = POLICY_TIMESHARE | POLICY_RR | POLICY_FIFO;
 
-               *count = sizeof(*enabled)/sizeof(int);
+               *count = sizeof(*enabled) / sizeof(int);
                *host = &realhost;
-               return(KERN_SUCCESS);
+               return KERN_SUCCESS;
        }
 
 
        *host = HOST_NULL;
-       return(KERN_INVALID_ARGUMENT);
+       return KERN_INVALID_ARGUMENT;
 }
 
 /*
  *     processor_set_statistics
  *
- *     Returns scheduling statistics for a processor set. 
+ *     Returns scheduling statistics for a processor set.
  */
-kern_return_t 
+kern_return_t
 processor_set_statistics(
        processor_set_t         pset,
        int                     flavor,
        processor_set_info_t    info,
-       mach_msg_type_number_t  *count)
+       mach_msg_type_number_t  *count)
 {
-        if (pset == PROCESSOR_SET_NULL)
-                return (KERN_INVALID_PROCESSOR_SET);
+       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 (flavor == PROCESSOR_SET_LOAD_INFO) {
-                register processor_set_load_info_t     load_info;
+               if (*count < PROCESSOR_SET_LOAD_INFO_COUNT) {
+                       return KERN_FAILURE;
+               }
 
-                if (*count < PROCESSOR_SET_LOAD_INFO_COUNT)
-                        return(KERN_FAILURE);
+               load_info = (processor_set_load_info_t) info;
 
-                load_info = (processor_set_load_info_t) info;
+               load_info->mach_factor = sched_mach_factor;
+               load_info->load_average = sched_load_average;
 
-                pset_lock(pset);
-                load_info->task_count = pset->task_count;
-                load_info->thread_count = pset->thread_count;
-                load_info->mach_factor = pset->mach_factor;
-                load_info->load_average = pset->load_average;
-                pset_unlock(pset);
+               load_info->task_count = tasks_count;
+               load_info->thread_count = threads_count;
 
-                *count = PROCESSOR_SET_LOAD_INFO_COUNT;
-                return(KERN_SUCCESS);
-        }
+               *count = PROCESSOR_SET_LOAD_INFO_COUNT;
+               return KERN_SUCCESS;
+       }
 
-        return(KERN_INVALID_ARGUMENT);
+       return KERN_INVALID_ARGUMENT;
 }
 
 /*
@@ -777,15 +1024,15 @@ processor_set_statistics(
  *
  *     Specify max priority permitted on processor set.  This affects
  *     newly created and assigned threads.  Optionally change existing
- *     ones.
+ *      ones.
  */
 kern_return_t
 processor_set_max_priority(
-       processor_set_t pset,
-       int             max_priority,
-       boolean_t       change_threads)
+       __unused processor_set_t        pset,
+       __unused int                    max_priority,
+       __unused boolean_t              change_threads)
 {
-       return (KERN_INVALID_ARGUMENT);
+       return KERN_INVALID_ARGUMENT;
 }
 
 /*
@@ -796,10 +1043,10 @@ processor_set_max_priority(
 
 kern_return_t
 processor_set_policy_enable(
-       processor_set_t pset,
-       int             policy)
+       __unused processor_set_t        pset,
+       __unused int                    policy)
 {
-       return (KERN_INVALID_ARGUMENT);
+       return KERN_INVALID_ARGUMENT;
 }
 
 /*
@@ -810,16 +1057,13 @@ processor_set_policy_enable(
  */
 kern_return_t
 processor_set_policy_disable(
-       processor_set_t pset,
-       int             policy,
-       boolean_t       change_threads)
+       __unused processor_set_t        pset,
+       __unused int                    policy,
+       __unused boolean_t              change_threads)
 {
-       return (KERN_INVALID_ARGUMENT);
+       return KERN_INVALID_ARGUMENT;
 }
 
-#define THING_TASK     0
-#define THING_THREAD   1
-
 /*
  *     processor_set_things:
  *
@@ -827,172 +1071,240 @@ processor_set_policy_disable(
  */
 kern_return_t
 processor_set_things(
-       processor_set_t         pset,
-       mach_port_t             **thing_list,
-       mach_msg_type_number_t  *count,
-       int                     type)
+       processor_set_t pset,
+       void **thing_list,
+       mach_msg_type_number_t *count,
+       int type)
 {
-       unsigned int actual;    /* this many things */
-       int i;
+       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;
-       vm_offset_t addr;
 
-       if (pset == PROCESSOR_SET_NULL)
+       if (pset == PROCESSOR_SET_NULL || pset != &pset0) {
                return KERN_INVALID_ARGUMENT;
+       }
 
-       size = 0; addr = 0;
+       task_size = 0;
+       task_size_needed = 0;
+       task_list = NULL;
+       actual_tasks = 0;
 
-       for (;;) {
-               pset_lock(pset);
-               if (!pset->active) {
-                       pset_unlock(pset);
-                       return KERN_FAILURE;
-               }
+       thread_size = 0;
+       thread_size_needed = 0;
+       thread_list = NULL;
+       actual_threads = 0;
 
-               if (type == THING_TASK)
-                       actual = pset->task_count;
-               else
-                       actual = pset->thread_count;
+       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 *);
 
-               size_needed = actual * sizeof(mach_port_t);
-               if (size_needed <= size)
+               if (task_size_needed <= task_size &&
+                   thread_size_needed <= thread_size) {
                        break;
+               }
 
-               /* unlock the pset and allocate more memory */
-               pset_unlock(pset);
-
-               if (size != 0)
-                       kfree(addr, size);
+               /* unlock and allocate more memory */
+               lck_mtx_unlock(&tasks_threads_lock);
 
-               assert(size_needed > 0);
-               size = size_needed;
+               /* grow task array */
+               if (task_size_needed > task_size) {
+                       if (task_size != 0) {
+                               kfree(task_list, task_size);
+                       }
 
-               addr = kalloc(size);
-               if (addr == 0)
-                       return KERN_RESOURCE_SHORTAGE;
-       }
+                       assert(task_size_needed > 0);
+                       task_size = task_size_needed;
 
-       /* OK, have memory and the processor_set is locked & active */
-
-       switch (type) {
-           case THING_TASK: {
-               task_t *tasks = (task_t *) addr;
-               task_t task;
-
-               for (i = 0, task = (task_t) queue_first(&pset->tasks);
-                    !queue_end(&pset->tasks, (queue_entry_t) task);
-                        task = (task_t) queue_next(&task->pset_tasks)) {
-                       
-                       task_lock(task);
-                       if (task->ref_count > 0) {
-                               /* take ref for convert_task_to_port */
-                               task_reference_locked(task);
-                               tasks[i++] = task;
+                       task_list = (task_t *)kalloc(task_size);
+                       if (task_list == NULL) {
+                               if (thread_size != 0) {
+                                       kfree(thread_list, thread_size);
+                               }
+                               return KERN_RESOURCE_SHORTAGE;
                        }
-                       task_unlock(task);
                }
-               break;
-           }
-
-           case THING_THREAD: {
-               thread_act_t *thr_acts = (thread_act_t *) addr;
-               thread_t thread;
-               thread_act_t thr_act;
-
-               for (i = 0, thread = (thread_t) queue_first(&pset->threads);
-                        !queue_end(&pset->threads, (queue_entry_t)thread);
-                thread = (thread_t) queue_next(&thread->pset_threads)) {
-
-                       thr_act = thread_lock_act(thread);
-                       if (thr_act && thr_act->act_ref_count > 0) {
-                               /* take ref for convert_act_to_port */
-                               act_reference_locked(thr_act);
-                               thr_acts[i++] = thr_act;
+
+               /* 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;
                        }
-                       thread_unlock_act(thread);
-               }
-               break;
                }
        }
-               
-       /* can unlock processor set now that we have the task/thread refs */
-       pset_unlock(pset);
 
-       if (i < actual) {
-                       actual = i;
-                       size_needed = actual * sizeof(mach_port_t);
+       /* 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
+               }
        }
-       assert(i == actual);
-
-       if (actual == 0) {
-               /* no things, so return null pointer and deallocate memory */
-               *thing_list = 0;
-               *count = 0;
-
-               if (size != 0)
-                       kfree(addr, size);
-       } else {
-               /* if we allocated too much, must copy */
+#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
 
-               if (size_needed < size) {
-                       vm_offset_t newaddr;
+       lck_mtx_unlock(&tasks_threads_lock);
 
-                       newaddr = kalloc(size_needed);
-                       if (newaddr == 0) {
-                               switch (type) {
-                                   case THING_TASK: {
-                                       task_t *tasks = (task_t *) addr;
+#if CONFIG_MACF
+       unsigned int j, used;
 
-                                       for (i = 0; i < actual; i++)
-                                               task_deallocate(tasks[i]);
+       /* 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 *);
 
-                                   case THING_THREAD: {
-                                       thread_act_t *acts = (thread_act_t *) addr;
+               /* 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;
+       }
 
-                                       for (i = 0; i < actual; i++)
-                                               act_deallocate(acts[i]);
-                                       break;
-                                   }
+       /* 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((char *) addr, (char *) newaddr, size_needed);
-                       kfree(addr, size);
-                       addr = newaddr;
+                       return KERN_RESOURCE_SHORTAGE;
                }
 
-               *thing_list = (mach_port_t *) addr;
-               *count = actual;
-
-               /* do the conversion that Mig should handle */
-
-               switch (type) {
-                   case THING_TASK: {
-                       task_t *tasks = (task_t *) addr;
-
-                       for (i = 0; i < actual; i++)
-                               (*thing_list)[i] = convert_task_to_port(tasks[i]);
-                       break;
-                   }
-
-                   case THING_THREAD: {
-                       thread_act_t *thr_acts = (thread_act_t *) addr;
+               bcopy((void *) addr, (void *) newaddr, size_needed);
+               kfree(addr, size);
 
-                       for (i = 0; i < actual; i++)
-                               (*thing_list)[i] = convert_act_to_port(thr_acts[i]);
-                       break;
-                   }
-               }
+               addr = newaddr;
+               size = size_needed;
        }
 
-       return(KERN_SUCCESS);
+       *thing_list = (void **)addr;
+       *count = (unsigned int)size / sizeof(void *);
+
+       return KERN_SUCCESS;
 }
 
 
@@ -1003,11 +1315,23 @@ processor_set_things(
  */
 kern_return_t
 processor_set_tasks(
-       processor_set_t         pset,
-       task_array_t            *task_list,
-       mach_msg_type_number_t  *count)
+       processor_set_t         pset,
+       task_array_t            *task_list,
+       mach_msg_type_number_t  *count)
 {
-    return(processor_set_things(pset, (mach_port_t **)task_list, count, THING_TASK));
+       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 */
+       for (i = 0; i < *count; i++) {
+               (*task_list)[i] = (task_t)convert_task_to_port((*task_list)[i]);
+       }
+       return KERN_SUCCESS;
 }
 
 /*
@@ -1015,48 +1339,46 @@ processor_set_tasks(
  *
  *     List all threads in the processor set.
  */
+#if defined(SECURE_KERNEL)
 kern_return_t
 processor_set_threads(
-       processor_set_t         pset,
-       thread_array_t          *thread_list,
-       mach_msg_type_number_t  *count)
+       __unused processor_set_t                pset,
+       __unused thread_array_t         *thread_list,
+       __unused mach_msg_type_number_t *count)
 {
-    return(processor_set_things(pset, (mach_port_t **)thread_list, count, THING_THREAD));
+       return KERN_FAILURE;
 }
-
-/*
- *      processor_set_base:
- *
- *      Specify per-policy base priority for a processor set.  Set processor
- *     set default policy to the given policy. This affects newly created
- *      and assigned threads.  Optionally change existing ones.
- */
+#elif defined(CONFIG_EMBEDDED)
 kern_return_t
-processor_set_base(
-       processor_set_t         pset,
-       policy_t                policy,
-        policy_base_t           base,
-       boolean_t               change)
+processor_set_threads(
+       __unused processor_set_t                pset,
+       __unused thread_array_t         *thread_list,
+       __unused mach_msg_type_number_t *count)
 {
-       return (KERN_INVALID_ARGUMENT);
+       return KERN_NOT_SUPPORTED;
 }
-
-/*
- *      processor_set_limit:
- *
- *      Specify per-policy limits for a processor set.  This affects
- *      newly created and assigned threads.  Optionally change existing
- *      ones.
- */
+#else
 kern_return_t
-processor_set_limit(
-       processor_set_t         pset,
-       policy_t                policy,
-        policy_limit_t         limit,
-       boolean_t               change)
+processor_set_threads(
+       processor_set_t         pset,
+       thread_array_t          *thread_list,
+       mach_msg_type_number_t  *count)
 {
-       return (KERN_POLICY_LIMIT);
+       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
@@ -1067,11 +1389,36 @@ processor_set_limit(
  */
 kern_return_t
 processor_set_policy_control(
-       processor_set_t         pset,
-       int                     flavor,
-       processor_set_info_t    policy_info,
-       mach_msg_type_number_t  count,
-       boolean_t               change)
+       __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;
+}
+
+pset_cluster_type_t
+recommended_pset_type(thread_t thread)
 {
-       return (KERN_INVALID_ARGUMENT);
+       (void)thread;
+       return PSET_SMP;
 }