]> git.saurik.com Git - apple/xnu.git/blobdiff - osfmk/kern/priority.c
xnu-6153.81.5.tar.gz
[apple/xnu.git] / osfmk / kern / priority.c
index 40eb17242b70832b44113719822e03184b7b652d..5ac1ce75682771a2cdec2bc0ae407397e584ead3 100644 (file)
@@ -1,8 +1,8 @@
 /*
- * Copyright (c) 2000-2010 Apple Inc. All rights reserved.
+ * 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
  * 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,
  * 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) 1991,1990,1989,1988,1987 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.
  */
@@ -76,6 +76,8 @@
 #include <kern/ledger.h>
 #include <machine/machparam.h>
 #include <kern/machine.h>
+#include <kern/policy_internal.h>
+#include <kern/sched_clutch.h>
 
 #ifdef CONFIG_MACH_APPROXIMATE_TIME
 #include <machine/commpage.h>  /* for commpage_update_mach_approximate_time */
@@ -85,8 +87,6 @@
 #include <kern/monotonic.h>
 #endif /* MONOTONIC */
 
-static void sched_update_thread_bucket(thread_t thread);
-
 /*
  *     thread_quantum_expire:
  *
@@ -97,15 +97,13 @@ static void sched_update_thread_bucket(thread_t thread);
 
 void
 thread_quantum_expire(
-       timer_call_param_t      p0,
-       timer_call_param_t      p1)
+       timer_call_param_t      p0,
+       timer_call_param_t      p1)
 {
-       processor_t                     processor = p0;
-       thread_t                        thread = p1;
-       ast_t                           preempt;
-       uint64_t                        ctime;
-       int                                     urgency;
-       uint64_t                        ignore1, ignore2;
+       processor_t                     processor = p0;
+       thread_t                        thread = p1;
+       ast_t                           preempt;
+       uint64_t                        ctime;
 
        assert(processor == current_processor());
        assert(thread == current_thread());
@@ -130,7 +128,7 @@ thread_quantum_expire(
        ledger_credit(thread->t_threadledger, thread_ledgers.cpu_time, thread->quantum_remaining);
        if (thread->t_bankledger) {
                ledger_credit(thread->t_bankledger, bank_ledgers.cpu_time,
-                               (thread->quantum_remaining - thread->t_deduct_bank_ledger_time));
+                   (thread->quantum_remaining - thread->t_deduct_bank_ledger_time));
        }
        thread->t_deduct_bank_ledger_time = 0;
 
@@ -158,6 +156,7 @@ thread_quantum_expire(
         */
        if ((thread->sched_mode == TH_MODE_REALTIME || thread->sched_mode == TH_MODE_FIXED) &&
            !(thread->sched_flags & TH_SFLAG_PROMOTED) &&
+           !(thread->kern_promotion_schedpri != 0) &&
            !(thread->sched_flags & TH_SFLAG_PROMOTE_REASON_MASK) &&
            !(thread->options & TH_OPT_SYSTEM_CRITICAL)) {
                uint64_t new_computation;
@@ -165,8 +164,8 @@ thread_quantum_expire(
                new_computation = ctime - thread->computation_epoch;
                new_computation += thread->computation_metered;
                if (new_computation > max_unsafe_computation) {
-                       KERNEL_DEBUG_CONSTANT(MACHDBG_CODE(DBG_MACH_SCHED, MACH_FAILSAFE)|DBG_FUNC_NONE,
-                                       (uintptr_t)thread->sched_pri, (uintptr_t)thread->sched_mode, 0, 0, 0);
+                       KERNEL_DEBUG_CONSTANT(MACHDBG_CODE(DBG_MACH_SCHED, MACH_FAILSAFE) | DBG_FUNC_NONE,
+                           (uintptr_t)thread->sched_pri, (uintptr_t)thread->sched_mode, 0, 0, 0);
 
                        thread->safe_release = ctime + sched_safe_duration;
 
@@ -177,15 +176,15 @@ thread_quantum_expire(
        /*
         *      Recompute scheduled priority if appropriate.
         */
-       if (SCHED(can_update_priority)(thread))
+       if (SCHED(can_update_priority)(thread)) {
                SCHED(update_priority)(thread);
-       else
+       } else {
                SCHED(lightweight_update_priority)(thread);
+       }
 
-       if (thread->sched_mode != TH_MODE_REALTIME)
+       if (thread->sched_mode != TH_MODE_REALTIME) {
                SCHED(quantum_expire)(thread);
-
-       processor_state_update_from_thread(processor, thread);
+       }
 
        /*
         *      This quantum is up, give this thread another.
@@ -218,11 +217,13 @@ thread_quantum_expire(
         */
 
        ast_t check_reason = AST_QUANTUM;
-       if (thread->task == kernel_task)
+       if (thread->task == kernel_task) {
                check_reason |= AST_URGENT;
+       }
 
-       if ((preempt = csw_check(processor, check_reason)) != AST_NONE)
+       if ((preempt = csw_check(thread, processor, check_reason)) != AST_NONE) {
                ast_on(preempt);
+       }
 
        /*
         * AST_KEVENT does not send an IPI when setting the AST,
@@ -234,21 +235,22 @@ thread_quantum_expire(
        thread_unlock(thread);
 
        timer_call_quantum_timer_enter(&processor->quantum_timer, thread,
-               processor->quantum_end, ctime);
+           processor->quantum_end, ctime);
 
        /* Tell platform layer that we are still running this thread */
-       urgency = thread_get_urgency(thread, &ignore1, &ignore2);
+       thread_urgency_t urgency = thread_get_urgency(thread, NULL, NULL);
        machine_thread_going_on_core(thread, urgency, 0, 0, ctime);
        machine_switch_perfcontrol_state_update(QUANTUM_EXPIRY, ctime,
-               0, thread);
+           0, thread);
 
 #if defined(CONFIG_SCHED_TIMESHARE_CORE)
        sched_timeshare_consider_maintenance(ctime);
 #endif /* CONFIG_SCHED_TIMESHARE_CORE */
 
 #if __arm__ || __arm64__
-       if (thread->sched_mode == TH_MODE_REALTIME)
+       if (thread->sched_mode == TH_MODE_REALTIME) {
                sched_consider_recommended_cores(ctime, thread);
+       }
 #endif /* __arm__ || __arm64__ */
 
        KERNEL_DEBUG_CONSTANT(MACHDBG_CODE(DBG_MACH_SCHED, MACH_SCHED_QUANTUM_EXPIRED) | DBG_FUNC_END, preempt, 0, 0, 0, 0);
@@ -270,12 +272,17 @@ sched_set_thread_base_priority(thread_t thread, int priority)
        assert(priority >= MINPRI);
        uint64_t ctime = 0;
 
-       if (thread->sched_mode == TH_MODE_REALTIME)
+       if (thread->sched_mode == TH_MODE_REALTIME) {
                assert(priority <= BASEPRI_RTQUEUES);
-       else
+       } else {
                assert(priority < BASEPRI_RTQUEUES);
+       }
 
        int old_base_pri = thread->base_pri;
+       thread->req_base_pri = priority;
+       if (thread->sched_flags & TH_SFLAG_BASE_PRI_FROZEN) {
+               priority = MAX(priority, old_base_pri);
+       }
        thread->base_pri = priority;
 
        if ((thread->state & TH_RUN) == TH_RUN) {
@@ -287,23 +294,61 @@ sched_set_thread_base_priority(thread_t thread, int priority)
                assert(thread->last_made_runnable_time == THREAD_NOT_RUNNABLE);
        }
 
-       /* 
-        * Currently the perfcontrol_attr depends on the base pri of the 
-        * thread. Therefore, we use this function as the hook for the 
-        * perfcontrol callout. 
+       /*
+        * Currently the perfcontrol_attr depends on the base pri of the
+        * thread. Therefore, we use this function as the hook for the
+        * perfcontrol callout.
         */
        if (thread == current_thread() && old_base_pri != priority) {
                if (!ctime) {
-                   ctime = mach_approximate_time();
+                       ctime = mach_approximate_time();
                }
                machine_switch_perfcontrol_state_update(PERFCONTROL_ATTR_UPDATE,
-                       ctime, PERFCONTROL_CALLOUT_WAKE_UNSAFE, thread);
+                   ctime, PERFCONTROL_CALLOUT_WAKE_UNSAFE, thread);
        }
-       sched_update_thread_bucket(thread);
+#if !CONFIG_SCHED_CLUTCH
+       /* For the clutch scheduler, this operation is done in set_sched_pri() */
+       SCHED(update_thread_bucket)(thread);
+#endif /* !CONFIG_SCHED_CLUTCH */
 
        thread_recompute_sched_pri(thread, SETPRI_DEFAULT);
 }
 
+/*
+ *     sched_set_kernel_thread_priority:
+ *
+ *     Set the absolute base priority of the thread
+ *     and reset its scheduled priority.
+ *
+ *     Called with the thread unlocked.
+ */
+void
+sched_set_kernel_thread_priority(thread_t thread, int new_priority)
+{
+       spl_t s = splsched();
+
+       thread_lock(thread);
+
+       assert(thread->sched_mode != TH_MODE_REALTIME);
+       assert(thread->effective_policy.thep_qos == THREAD_QOS_UNSPECIFIED);
+
+       if (new_priority > thread->max_priority) {
+               new_priority = thread->max_priority;
+       }
+#if CONFIG_EMBEDDED
+       if (new_priority < MAXPRI_THROTTLE) {
+               new_priority = MAXPRI_THROTTLE;
+       }
+#endif /* CONFIG_EMBEDDED */
+
+       thread->importance = new_priority - thread->task_priority;
+
+       sched_set_thread_base_priority(thread, new_priority);
+
+       thread_unlock(thread);
+       splx(s);
+}
+
 /*
  *     thread_recompute_sched_pri:
  *
@@ -327,8 +372,9 @@ thread_recompute_sched_pri(thread_t thread, set_sched_pri_options_t options)
 
        int priority = thread->base_pri;
 
-       if (sched_mode == TH_MODE_TIMESHARE)
+       if (sched_mode == TH_MODE_TIMESHARE) {
                priority = SCHED(compute_timeshare_priority)(thread);
+       }
 
        if (sched_flags & TH_SFLAG_DEPRESS) {
                /* thread_yield_internal overrides kernel mutex promotion */
@@ -339,22 +385,34 @@ thread_recompute_sched_pri(thread_t thread, set_sched_pri_options_t options)
                        priority = DEPRESSPRI;
                }
 
+               if (thread->kern_promotion_schedpri > 0) {
+                       priority = MAX(priority, thread->kern_promotion_schedpri);
+
+                       if (sched_mode != TH_MODE_REALTIME) {
+                               priority = MIN(priority, MAXPRI_PROMOTE);
+                       }
+               }
+
                if (sched_flags & TH_SFLAG_PROMOTED) {
                        priority = MAX(priority, thread->promotion_priority);
 
-                       if (sched_mode != TH_MODE_REALTIME)
+                       if (sched_mode != TH_MODE_REALTIME) {
                                priority = MIN(priority, MAXPRI_PROMOTE);
+                       }
                }
 
                if (sched_flags & TH_SFLAG_PROMOTE_REASON_MASK) {
-                       if (sched_flags & TH_SFLAG_RW_PROMOTED)
+                       if (sched_flags & TH_SFLAG_RW_PROMOTED) {
                                priority = MAX(priority, MINPRI_RWLOCK);
+                       }
 
-                       if (sched_flags & TH_SFLAG_WAITQ_PROMOTED)
+                       if (sched_flags & TH_SFLAG_WAITQ_PROMOTED) {
                                priority = MAX(priority, MINPRI_WAITQ);
+                       }
 
-                       if (sched_flags & TH_SFLAG_EXEC_PROMOTED)
+                       if (sched_flags & TH_SFLAG_EXEC_PROMOTED) {
                                priority = MAX(priority, MINPRI_EXEC);
+                       }
                }
        }
 
@@ -364,10 +422,10 @@ thread_recompute_sched_pri(thread_t thread, set_sched_pri_options_t options)
 void
 sched_default_quantum_expire(thread_t thread __unused)
 {
-      /*
-       * No special behavior when a timeshare, fixed, or realtime thread
-       * uses up its entire quantum
-       */
+       /*
+        * No special behavior when a timeshare, fixed, or realtime thread
+        * uses up its entire quantum
+        */
 }
 
 #if defined(CONFIG_SCHED_TIMESHARE_CORE)
@@ -399,15 +457,26 @@ lightweight_update_priority(thread_t thread)
                 *      during contention for processor
                 *      resources.
                 */
-               if (thread->pri_shift < INT8_MAX)
+               if (thread->pri_shift < INT8_MAX) {
                        thread->sched_usage += delta;
+               }
 
                thread->cpu_delta += delta;
 
+#if CONFIG_SCHED_CLUTCH
+               /*
+                * Update the CPU usage for the thread group to which the thread belongs.
+                * The implementation assumes that the thread ran for the entire delta
+                * as part of the same thread group.
+                */
+               sched_clutch_cpu_usage_update(thread, delta);
+#endif /* CONFIG_SCHED_CLUTCH */
+
                priority = sched_compute_timeshare_priority(thread);
 
-               if (priority != thread->sched_pri)
+               if (priority != thread->sched_pri) {
                        thread_recompute_sched_pri(thread, SETPRI_LAZY);
+               }
        }
 }
 
@@ -418,17 +487,40 @@ lightweight_update_priority(thread_t thread)
  *     is  usage = (usage >> shift1) +/- (usage >> abs(shift2))  where the
  *     +/- is determined by the sign of shift 2.
  */
-struct shift_data {
-       int     shift1;
-       int     shift2;
-};
 
-#define SCHED_DECAY_TICKS      32
-static struct shift_data       sched_decay_shifts[SCHED_DECAY_TICKS] = {
-       {1,1},{1,3},{1,-3},{2,-7},{3,5},{3,-5},{4,-8},{5,7},
-       {5,-7},{6,-10},{7,10},{7,-9},{8,-11},{9,12},{9,-11},{10,-13},
-       {11,14},{11,-13},{12,-15},{13,17},{13,-15},{14,-17},{15,19},{16,18},
-       {16,-19},{17,22},{18,20},{18,-20},{19,26},{20,22},{20,-22},{21,-27}
+const struct shift_data        sched_decay_shifts[SCHED_DECAY_TICKS] = {
+       { .shift1 = 1, .shift2 = 1 },
+       { .shift1 = 1, .shift2 = 3 },
+       { .shift1 = 1, .shift2 = -3 },
+       { .shift1 = 2, .shift2 = -7 },
+       { .shift1 = 3, .shift2 = 5 },
+       { .shift1 = 3, .shift2 = -5 },
+       { .shift1 = 4, .shift2 = -8 },
+       { .shift1 = 5, .shift2 = 7 },
+       { .shift1 = 5, .shift2 = -7 },
+       { .shift1 = 6, .shift2 = -10 },
+       { .shift1 = 7, .shift2 = 10 },
+       { .shift1 = 7, .shift2 = -9 },
+       { .shift1 = 8, .shift2 = -11 },
+       { .shift1 = 9, .shift2 = 12 },
+       { .shift1 = 9, .shift2 = -11 },
+       { .shift1 = 10, .shift2 = -13 },
+       { .shift1 = 11, .shift2 = 14 },
+       { .shift1 = 11, .shift2 = -13 },
+       { .shift1 = 12, .shift2 = -15 },
+       { .shift1 = 13, .shift2 = 17 },
+       { .shift1 = 13, .shift2 = -15 },
+       { .shift1 = 14, .shift2 = -17 },
+       { .shift1 = 15, .shift2 = 19 },
+       { .shift1 = 16, .shift2 = 18 },
+       { .shift1 = 16, .shift2 = -19 },
+       { .shift1 = 17, .shift2 = 22 },
+       { .shift1 = 18, .shift2 = 20 },
+       { .shift1 = 18, .shift2 = -20 },
+       { .shift1 = 19, .shift2 = 26 },
+       { .shift1 = 20, .shift2 = 22 },
+       { .shift1 = 20, .shift2 = -22 },
+       { .shift1 = 21, .shift2 = -27 }
 };
 
 /*
@@ -438,7 +530,9 @@ static struct shift_data    sched_decay_shifts[SCHED_DECAY_TICKS] = {
  */
 extern int sched_pri_decay_band_limit;
 
-#ifdef CONFIG_EMBEDDED
+
+/* Only use the decay floor logic on embedded non-clutch schedulers */
+#if CONFIG_EMBEDDED && !CONFIG_SCHED_CLUTCH
 
 int
 sched_compute_timeshare_priority(thread_t thread)
@@ -470,7 +564,7 @@ sched_compute_timeshare_priority(thread_t thread)
        return priority;
 }
 
-#else /* CONFIG_EMBEDDED */
+#else /* CONFIG_EMBEDDED && !CONFIG_SCHED_CLUTCH */
 
 int
 sched_compute_timeshare_priority(thread_t thread)
@@ -478,15 +572,16 @@ sched_compute_timeshare_priority(thread_t thread)
        /* start with base priority */
        int priority = thread->base_pri - (thread->sched_usage >> thread->pri_shift);
 
-       if (priority < MINPRI_USER)
+       if (priority < MINPRI_USER) {
                priority = MINPRI_USER;
-       else if (priority > MAXPRI_KERNEL)
+       } else if (priority > MAXPRI_KERNEL) {
                priority = MAXPRI_KERNEL;
+       }
 
        return priority;
 }
 
-#endif /* CONFIG_EMBEDDED */
+#endif /* CONFIG_EMBEDDED && !CONFIG_SCHED_CLUTCH */
 
 /*
  *     can_update_priority
@@ -497,12 +592,13 @@ sched_compute_timeshare_priority(thread_t thread)
  */
 boolean_t
 can_update_priority(
-                                       thread_t        thread)
+       thread_t        thread)
 {
-       if (sched_tick == thread->sched_stamp)
-               return (FALSE);
-       else
-               return (TRUE);
+       if (sched_tick == thread->sched_stamp) {
+               return FALSE;
+       } else {
+               return TRUE;
+       }
 }
 
 /*
@@ -514,7 +610,7 @@ can_update_priority(
  */
 void
 update_priority(
-       thread_t        thread)
+       thread_t        thread)
 {
        uint32_t ticks, delta;
 
@@ -524,8 +620,9 @@ update_priority(
        thread->sched_stamp += ticks;
 
        /* If requested, accelerate aging of sched_usage */
-       if (sched_decay_usage_age_factor > 1)
+       if (sched_decay_usage_age_factor > 1) {
                ticks *= sched_decay_usage_age_factor;
+       }
 
        /*
         *      Gather cpu usage data.
@@ -534,27 +631,37 @@ update_priority(
        if (ticks < SCHED_DECAY_TICKS) {
                /*
                 *      Accumulate timesharing usage only during contention for processor
-                *      resources. Use the pri_shift from the previous tick window to 
+                *      resources. Use the pri_shift from the previous tick window to
                 *      determine if the system was in a contended state.
                 */
-               if (thread->pri_shift < INT8_MAX)
+               if (thread->pri_shift < INT8_MAX) {
                        thread->sched_usage += delta;
+               }
 
                thread->cpu_usage += delta + thread->cpu_delta;
                thread->cpu_delta = 0;
 
-               struct shift_data *shiftp = &sched_decay_shifts[ticks];
+#if CONFIG_SCHED_CLUTCH
+               /*
+                * Update the CPU usage for the thread group to which the thread belongs.
+                * The implementation assumes that the thread ran for the entire delta
+                * as part of the same thread group.
+                */
+               sched_clutch_cpu_usage_update(thread, delta);
+#endif /* CONFIG_SCHED_CLUTCH */
+
+               const struct shift_data *shiftp = &sched_decay_shifts[ticks];
 
                if (shiftp->shift2 > 0) {
                        thread->cpu_usage =   (thread->cpu_usage >> shiftp->shift1) +
-                                             (thread->cpu_usage >> shiftp->shift2);
+                           (thread->cpu_usage >> shiftp->shift2);
                        thread->sched_usage = (thread->sched_usage >> shiftp->shift1) +
-                                             (thread->sched_usage >> shiftp->shift2);
+                           (thread->sched_usage >> shiftp->shift2);
                } else {
                        thread->cpu_usage =   (thread->cpu_usage >>   shiftp->shift1) -
-                                             (thread->cpu_usage >> -(shiftp->shift2));
+                           (thread->cpu_usage >> -(shiftp->shift2));
                        thread->sched_usage = (thread->sched_usage >>   shiftp->shift1) -
-                                             (thread->sched_usage >> -(shiftp->shift2));
+                           (thread->sched_usage >> -(shiftp->shift2));
                }
        } else {
                thread->cpu_usage = thread->cpu_delta = 0;
@@ -576,11 +683,16 @@ update_priority(
         * values. The updated pri_shift would be used to calculate the
         * new priority of the thread.
         */
+#if CONFIG_SCHED_CLUTCH
+       thread->pri_shift = sched_clutch_thread_pri_shift(thread, thread->th_sched_bucket);
+#else /* CONFIG_SCHED_CLUTCH */
        thread->pri_shift = sched_pri_shifts[thread->th_sched_bucket];
+#endif /* CONFIG_SCHED_CLUTCH */
 
        /* Recompute scheduled priority if appropriate. */
-       if (thread->sched_mode == TH_MODE_TIMESHARE)
+       if (thread->sched_mode == TH_MODE_TIMESHARE) {
                thread_recompute_sched_pri(thread, SETPRI_LAZY);
+       }
 }
 
 #endif /* CONFIG_SCHED_TIMESHARE_CORE */
@@ -589,38 +701,40 @@ update_priority(
 /*
  * TH_BUCKET_RUN is a count of *all* runnable non-idle threads.
  * Each other bucket is a count of the runnable non-idle threads
- * with that property.
+ * with that property. All updates to these counts should be
+ * performed with os_atomic_* operations.
+ *
+ * For the clutch scheduler, this global bucket is used only for
+ * keeping the total global run count.
  */
-volatile uint32_t       sched_run_buckets[TH_BUCKET_MAX];
+uint32_t       sched_run_buckets[TH_BUCKET_MAX];
 
 static void
 sched_incr_bucket(sched_bucket_t bucket)
 {
        assert(bucket >= TH_BUCKET_FIXPRI &&
-              bucket <= TH_BUCKET_SHARE_BG);
+           bucket <= TH_BUCKET_SHARE_BG);
 
-       hw_atomic_add(&sched_run_buckets[bucket], 1);
+       os_atomic_inc(&sched_run_buckets[bucket], relaxed);
 }
 
 static void
 sched_decr_bucket(sched_bucket_t bucket)
 {
        assert(bucket >= TH_BUCKET_FIXPRI &&
-              bucket <= TH_BUCKET_SHARE_BG);
+           bucket <= TH_BUCKET_SHARE_BG);
 
-       assert(sched_run_buckets[bucket] > 0);
+       assert(os_atomic_load(&sched_run_buckets[bucket], relaxed) > 0);
 
-       hw_atomic_sub(&sched_run_buckets[bucket], 1);
+       os_atomic_dec(&sched_run_buckets[bucket], relaxed);
 }
 
-/* TH_RUN & !TH_IDLE controls whether a thread has a run count */
-
 uint32_t
 sched_run_incr(thread_t thread)
 {
-       assert((thread->state & (TH_RUN|TH_IDLE)) == TH_RUN);
+       assert((thread->state & (TH_RUN | TH_IDLE)) == TH_RUN);
 
-       uint32_t new_count = hw_atomic_add(&sched_run_buckets[TH_BUCKET_RUN], 1);
+       uint32_t new_count = os_atomic_inc(&sched_run_buckets[TH_BUCKET_RUN], relaxed);
 
        sched_incr_bucket(thread->th_sched_bucket);
 
@@ -630,16 +744,16 @@ sched_run_incr(thread_t thread)
 uint32_t
 sched_run_decr(thread_t thread)
 {
-       assert((thread->state & (TH_RUN|TH_IDLE)) != TH_RUN);
+       assert((thread->state & (TH_RUN | TH_IDLE)) != TH_RUN);
 
        sched_decr_bucket(thread->th_sched_bucket);
 
-       uint32_t new_count = hw_atomic_sub(&sched_run_buckets[TH_BUCKET_RUN], 1);
+       uint32_t new_count = os_atomic_dec(&sched_run_buckets[TH_BUCKET_RUN], relaxed);
 
        return new_count;
 }
 
-static void
+void
 sched_update_thread_bucket(thread_t thread)
 {
        sched_bucket_t old_bucket = thread->th_sched_bucket;
@@ -652,14 +766,15 @@ sched_update_thread_bucket(thread_t thread)
                break;
 
        case TH_MODE_TIMESHARE:
-               if (thread->base_pri > BASEPRI_DEFAULT)
+               if (thread->base_pri > BASEPRI_DEFAULT) {
                        new_bucket = TH_BUCKET_SHARE_FG;
-               else if (thread->base_pri > BASEPRI_UTILITY)
+               } else if (thread->base_pri > BASEPRI_UTILITY) {
                        new_bucket = TH_BUCKET_SHARE_DF;
-               else if (thread->base_pri > MAXPRI_THROTTLE)
+               } else if (thread->base_pri > MAXPRI_THROTTLE) {
                        new_bucket = TH_BUCKET_SHARE_UT;
-               else
+               } else {
                        new_bucket = TH_BUCKET_SHARE_BG;
+               }
                break;
 
        default:
@@ -671,7 +786,7 @@ sched_update_thread_bucket(thread_t thread)
                thread->th_sched_bucket = new_bucket;
                thread->pri_shift = sched_pri_shifts[new_bucket];
 
-               if ((thread->state & (TH_RUN|TH_IDLE)) == TH_RUN) {
+               if ((thread->state & (TH_RUN | TH_IDLE)) == TH_RUN) {
                        sched_decr_bucket(old_bucket);
                        sched_incr_bucket(new_bucket);
                }
@@ -703,7 +818,7 @@ sched_set_thread_mode(thread_t thread, sched_mode_t new_mode)
 
        thread->sched_mode = new_mode;
 
-       sched_update_thread_bucket(thread);
+       SCHED(update_thread_bucket)(thread);
 }
 
 /*
@@ -715,8 +830,9 @@ sched_thread_mode_demote(thread_t thread, uint32_t reason)
        assert(reason & TH_SFLAG_DEMOTED_MASK);
        assert((thread->sched_flags & reason) != reason);
 
-       if (thread->policy_reset)
+       if (thread->policy_reset) {
                return;
+       }
 
        if (thread->sched_flags & TH_SFLAG_DEMOTED_MASK) {
                /* Another demotion reason is already active */
@@ -736,8 +852,9 @@ sched_thread_mode_demote(thread_t thread, uint32_t reason)
 
        thread_recompute_priority(thread);
 
-       if (removed)
+       if (removed) {
                thread_run_queue_reinsert(thread, SCHED_TAILQ);
+       }
 }
 
 /*
@@ -767,97 +884,11 @@ sched_thread_mode_undemote(thread_t thread, uint32_t reason)
 
        thread_recompute_priority(thread);
 
-       if (removed)
+       if (removed) {
                thread_run_queue_reinsert(thread, SCHED_TAILQ);
-}
-
-/*
- * Promote thread to a specific priority
- *
- * Promotion must not last past syscall boundary
- * Clients must always pair promote and unpromote 1:1
- *
- * Called at splsched with thread locked
- */
-void
-sched_thread_promote_to_pri(thread_t    thread,
-                            int         priority,
-              __kdebug_only uintptr_t   trace_obj /* already unslid */)
-{
-       assert((thread->sched_flags & TH_SFLAG_PROMOTED) != TH_SFLAG_PROMOTED);
-       assert(thread->promotion_priority == 0);
-       assert(priority <= MAXPRI_PROMOTE);
-       assert(priority > 0);
-
-       KDBG(MACHDBG_CODE(DBG_MACH_SCHED, MACH_PROMOTED),
-            thread_tid(thread), trace_obj, priority);
-
-       thread->sched_flags |= TH_SFLAG_PROMOTED;
-       thread->promotion_priority = priority;
-
-       thread_recompute_sched_pri(thread, SETPRI_DEFAULT);
-}
-
-
-/*
- * Update a pre-existing priority promotion to have a higher priority floor
- * Priority can only go up from the previous value
- * Update must occur while a promotion is active
- *
- * Called at splsched with thread locked
- */
-void
-sched_thread_update_promotion_to_pri(thread_t   thread,
-                                     int        priority,
-                       __kdebug_only uintptr_t  trace_obj /* already unslid */)
-{
-       assert(thread->promotions > 0);
-       assert((thread->sched_flags & TH_SFLAG_PROMOTED) == TH_SFLAG_PROMOTED);
-       assert(thread->promotion_priority > 0);
-       assert(priority <= MAXPRI_PROMOTE);
-
-       if (thread->promotion_priority < priority) {
-               KDBG(MACHDBG_CODE(DBG_MACH_SCHED, MACH_PROMOTED_UPDATE),
-                    thread_tid(thread), trace_obj, priority);
-
-               thread->promotion_priority = priority;
-               thread_recompute_sched_pri(thread, SETPRI_DEFAULT);
        }
 }
 
-/*
- * End a priority promotion
- * Demotes a thread back to its expected priority without the promotion in place
- *
- * Called at splsched with thread locked
- */
-void
-sched_thread_unpromote(thread_t     thread,
-         __kdebug_only uintptr_t    trace_obj /* already unslid */)
-{
-       assert((thread->sched_flags & TH_SFLAG_PROMOTED) == TH_SFLAG_PROMOTED);
-       assert(thread->promotion_priority > 0);
-
-       KDBG(MACHDBG_CODE(DBG_MACH_SCHED, MACH_UNPROMOTED),
-            thread_tid(thread), trace_obj, 0);
-
-       thread->sched_flags &= ~TH_SFLAG_PROMOTED;
-       thread->promotion_priority = 0;
-
-       thread_recompute_sched_pri(thread, SETPRI_DEFAULT);
-}
-
-/* called with thread locked */
-void
-assert_promotions_invariant(thread_t thread)
-{
-       if (thread->promotions > 0)
-               assert((thread->sched_flags & TH_SFLAG_PROMOTED) == TH_SFLAG_PROMOTED);
-
-       if (thread->promotions == 0)
-               assert((thread->sched_flags & TH_SFLAG_PROMOTED) != TH_SFLAG_PROMOTED);
-}
-
 /*
  * Promote thread to have a sched pri floor for a specific reason
  *
@@ -869,8 +900,8 @@ assert_promotions_invariant(thread_t thread)
  */
 void
 sched_thread_promote_reason(thread_t    thread,
-                            uint32_t    reason,
-              __kdebug_only uintptr_t   trace_obj /* already unslid */)
+    uint32_t    reason,
+    __kdebug_only uintptr_t   trace_obj /* already unslid */)
 {
        assert(reason & TH_SFLAG_PROMOTE_REASON_MASK);
        assert((thread->sched_flags & reason) != reason);
@@ -878,18 +909,18 @@ sched_thread_promote_reason(thread_t    thread,
        switch (reason) {
        case TH_SFLAG_RW_PROMOTED:
                KDBG(MACHDBG_CODE(DBG_MACH_SCHED, MACH_RW_PROMOTE),
-                    thread_tid(thread), thread->sched_pri,
-                    thread->base_pri, trace_obj);
+                   thread_tid(thread), thread->sched_pri,
+                   thread->base_pri, trace_obj);
                break;
        case TH_SFLAG_WAITQ_PROMOTED:
                KDBG(MACHDBG_CODE(DBG_MACH_SCHED, MACH_WAITQ_PROMOTE),
-                    thread_tid(thread), thread->sched_pri,
-                    thread->base_pri, trace_obj);
+                   thread_tid(thread), thread->sched_pri,
+                   thread->base_pri, trace_obj);
                break;
        case TH_SFLAG_EXEC_PROMOTED:
                KDBG(MACHDBG_CODE(DBG_MACH_SCHED, MACH_EXEC_PROMOTE),
-                    thread_tid(thread), thread->sched_pri,
-                    thread->base_pri, trace_obj);
+                   thread_tid(thread), thread->sched_pri,
+                   thread->base_pri, trace_obj);
                break;
        }
 
@@ -906,8 +937,8 @@ sched_thread_promote_reason(thread_t    thread,
  */
 void
 sched_thread_unpromote_reason(thread_t  thread,
-                              uint32_t  reason,
-                __kdebug_only uintptr_t trace_obj /* already unslid */)
+    uint32_t  reason,
+    __kdebug_only uintptr_t trace_obj /* already unslid */)
 {
        assert(reason & TH_SFLAG_PROMOTE_REASON_MASK);
        assert((thread->sched_flags & reason) == reason);
@@ -915,18 +946,18 @@ sched_thread_unpromote_reason(thread_t  thread,
        switch (reason) {
        case TH_SFLAG_RW_PROMOTED:
                KDBG(MACHDBG_CODE(DBG_MACH_SCHED, MACH_RW_DEMOTE),
-                    thread_tid(thread), thread->sched_pri,
-                    thread->base_pri, trace_obj);
+                   thread_tid(thread), thread->sched_pri,
+                   thread->base_pri, trace_obj);
                break;
        case TH_SFLAG_WAITQ_PROMOTED:
                KDBG(MACHDBG_CODE(DBG_MACH_SCHED, MACH_WAITQ_DEMOTE),
-                    thread_tid(thread), thread->sched_pri,
-                    thread->base_pri, trace_obj);
+                   thread_tid(thread), thread->sched_pri,
+                   thread->base_pri, trace_obj);
                break;
        case TH_SFLAG_EXEC_PROMOTED:
                KDBG(MACHDBG_CODE(DBG_MACH_SCHED, MACH_EXEC_DEMOTE),
-                    thread_tid(thread), thread->sched_pri,
-                    thread->base_pri, trace_obj);
+                   thread_tid(thread), thread->sched_pri,
+                   thread->base_pri, trace_obj);
                break;
        }
 
@@ -934,5 +965,3 @@ sched_thread_unpromote_reason(thread_t  thread,
 
        thread_recompute_sched_pri(thread, SETPRI_DEFAULT);
 }
-
-