xnu-6153.61.1.tar.gz

[apple/xnu.git] / osfmk / kern / syscall_subr.c

diff --git a/osfmk/kern/syscall_subr.c b/osfmk/kern/syscall_subr.c
index a0c1d38e32911ceea1f56c7b665bcb56cb1c20c3..ef23168835391a01e8c081191870736faf2bac36 100644 (file)
--- a/osfmk/kern/syscall_subr.c
+++ b/osfmk/kern/syscall_subr.c
@@ -1,8 +1,8 @@
 /*
 /*

- * Copyright (c) 2000-2009 Apple Inc. All rights reserved.
+ * Copyright (c) 2000-2017 Apple Inc. All rights reserved.

  *
  * @APPLE_OSREFERENCE_LICENSE_HEADER_START@
  *
  * @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
  * 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


@@ -11,10 +11,10 @@
  * 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.
  * 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.
  * 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,
  * 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,


@@ -22,39 +22,37 @@
  * 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.
  * 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@
  */
  * @APPLE_OSREFERENCE_LICENSE_HEADER_END@
  */
 /*
  * @OSF_COPYRIGHT@
  */

-/* 
+/*

  * Mach Operating System
  * Copyright (c) 1991,1990,1989,1988,1987 Carnegie Mellon University
  * All Rights Reserved.
  * 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.
  * 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 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
  * 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
  *  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.
  */
  * any improvements or extensions that they make and grant Carnegie Mellon
  * the rights to redistribute these changes.
  */

-/*
- */

 
 #include <mach/boolean.h>
 #include <mach/thread_switch.h>
 
 #include <mach/boolean.h>
 #include <mach/thread_switch.h>


@@ -78,23 +76,19 @@
 #include <sys/kdebug.h>
 #include <kern/ast.h>
 
 #include <sys/kdebug.h>
 #include <kern/ast.h>
 

-#ifdef MACH_BSD
-extern void workqueue_thread_yielded(void);
-extern sched_call_t workqueue_get_sched_callback(void);
-#endif /* MACH_BSD */
-
-extern wait_result_t thread_handoff_reason(thread_t thread, ast_t reason);
+static void thread_depress_abstime(uint64_t interval);
+static void thread_depress_ms(mach_msg_timeout_t interval);

 
 /* Called from commpage to take a delayed preemption when exiting
  * the "Preemption Free Zone" (PFZ).
  */
 kern_return_t
 pfz_exit(
 
 /* Called from commpage to take a delayed preemption when exiting
  * the "Preemption Free Zone" (PFZ).
  */
 kern_return_t
 pfz_exit(

-__unused       struct pfz_exit_args *args)
+       __unused        struct pfz_exit_args *args)

 {
        /* For now, nothing special to do.  We'll pick up the ASTs on kernel exit. */
 
 {
        /* For now, nothing special to do.  We'll pick up the ASTs on kernel exit. */
 

-       return (KERN_SUCCESS);
+       return KERN_SUCCESS;

 }
 
 
 }
 
 


@@ -108,14 +102,16 @@ __unused  struct pfz_exit_args *args)
 static void
 swtch_continue(void)
 {
 static void
 swtch_continue(void)
 {

-       processor_t     myprocessor;
-    boolean_t                          result;
+       processor_t     myprocessor;
+       boolean_t       result;

 
 

-    disable_preemption();
+       disable_preemption();

        myprocessor = current_processor();
        myprocessor = current_processor();

-       result = !SCHED(processor_queue_empty)(myprocessor) || rt_runq.count > 0;
+       result = SCHED(thread_should_yield)(myprocessor, current_thread());

        enable_preemption();
 
        enable_preemption();
 

+       ml_delay_on_yield();
+

        thread_syscall_return(result);
        /*NOTREACHED*/
 }
        thread_syscall_return(result);
        /*NOTREACHED*/
 }


@@ -124,60 +120,53 @@ boolean_t
 swtch(
        __unused struct swtch_args *args)
 {
 swtch(
        __unused struct swtch_args *args)
 {

-       processor_t     myprocessor;
-       boolean_t                               result;
+       processor_t     myprocessor;

 
        disable_preemption();
        myprocessor = current_processor();
 
        disable_preemption();
        myprocessor = current_processor();

-       if (SCHED(processor_queue_empty)(myprocessor) &&        rt_runq.count == 0) {
+       if (!SCHED(thread_should_yield)(myprocessor, current_thread())) {

                mp_enable_preemption();
 
                mp_enable_preemption();
 

-               return (FALSE);
+               return FALSE;

        }
        enable_preemption();
 
        counter(c_swtch_block++);
 
        }
        enable_preemption();
 
        counter(c_swtch_block++);
 

-       thread_block_reason((thread_continue_t)swtch_continue, NULL, AST_YIELD);
-
-       disable_preemption();
-       myprocessor = current_processor();
-       result = !SCHED(processor_queue_empty)(myprocessor) || rt_runq.count > 0;
-       enable_preemption();
-
-       return (result);
+       thread_yield_with_continuation((thread_continue_t)swtch_continue, NULL);

 }
 
 static void
 swtch_pri_continue(void)
 {
 }
 
 static void
 swtch_pri_continue(void)
 {

-       processor_t     myprocessor;
-    boolean_t                          result;
+       processor_t     myprocessor;
+       boolean_t       result;

 
 

-       thread_depress_abort_internal(current_thread());
+       thread_depress_abort(current_thread());

 
 

-    disable_preemption();
+       disable_preemption();

        myprocessor = current_processor();
        myprocessor = current_processor();

-       result = !SCHED(processor_queue_empty)(myprocessor) || rt_runq.count > 0;
+       result = SCHED(thread_should_yield)(myprocessor, current_thread());

        mp_enable_preemption();
 
        mp_enable_preemption();
 

+       ml_delay_on_yield();
+

        thread_syscall_return(result);
        /*NOTREACHED*/
 }
 
 boolean_t
 swtch_pri(
        thread_syscall_return(result);
        /*NOTREACHED*/
 }
 
 boolean_t
 swtch_pri(

-__unused       struct swtch_pri_args *args)
+       __unused        struct swtch_pri_args *args)

 {
 {

-       processor_t     myprocessor;
-       boolean_t                               result;
+       processor_t     myprocessor;

 
        disable_preemption();
        myprocessor = current_processor();
 
        disable_preemption();
        myprocessor = current_processor();

-       if (SCHED(processor_queue_empty)(myprocessor) && rt_runq.count == 0) {
+       if (!SCHED(thread_should_yield)(myprocessor, current_thread())) {

                mp_enable_preemption();
 
                mp_enable_preemption();
 

-               return (FALSE);
+               return FALSE;

        }
        enable_preemption();
 
        }
        enable_preemption();
 


@@ -185,45 +174,20 @@ __unused  struct swtch_pri_args *args)
 
        thread_depress_abstime(thread_depress_time);
 
 
        thread_depress_abstime(thread_depress_time);
 

-       thread_block_reason((thread_continue_t)swtch_pri_continue, NULL, AST_YIELD);
-
-       thread_depress_abort_internal(current_thread());
-
-       disable_preemption();
-       myprocessor = current_processor();
-       result = !SCHED(processor_queue_empty)(myprocessor) || rt_runq.count > 0;
-       enable_preemption();
-
-       return (result);
-}
-
-static boolean_t
-thread_switch_disable_workqueue_sched_callback(void)
-{
-       sched_call_t callback = workqueue_get_sched_callback();
-       return thread_disable_sched_call(current_thread(), callback) != NULL;
+       thread_yield_with_continuation((thread_continue_t)swtch_pri_continue, NULL);

 }
 
 static void
 }
 
 static void

-thread_switch_enable_workqueue_sched_callback(void)
+thread_switch_continue(void *parameter, __unused int ret)

 {
 {

-       sched_call_t callback = workqueue_get_sched_callback();
-       thread_reenable_sched_call(current_thread(), callback);
-}
-
-static void
-thread_switch_continue(void)
-{
-       thread_t        self = current_thread();
-       int                                     option = self->saved.swtch.option;
-       boolean_t                       reenable_workq_callback = self->saved.swtch.reenable_workq_callback;
-
+       thread_t self = current_thread();
+       int option = (int)(intptr_t)parameter;

 
 

-       if (option == SWITCH_OPTION_DEPRESS || option == SWITCH_OPTION_OSLOCK_DEPRESS)
-               thread_depress_abort_internal(self);
+       if (option == SWITCH_OPTION_DEPRESS || option == SWITCH_OPTION_OSLOCK_DEPRESS) {
+               thread_depress_abort(self);
+       }

 
 

-       if (reenable_workq_callback)
-               thread_switch_enable_workqueue_sched_callback();
+       ml_delay_on_yield();

 
        thread_syscall_return(KERN_SUCCESS);
        /*NOTREACHED*/
 
        thread_syscall_return(KERN_SUCCESS);
        /*NOTREACHED*/


@@ -238,97 +202,71 @@ kern_return_t
 thread_switch(
        struct thread_switch_args *args)
 {
 thread_switch(
        struct thread_switch_args *args)
 {

-       thread_t                        thread = THREAD_NULL;
-       thread_t                        self = current_thread();
-       mach_port_name_t                thread_name = args->thread_name;
-       int                                             option = args->option;
-       mach_msg_timeout_t              option_time = args->option_time;
-       uint32_t                                scale_factor = NSEC_PER_MSEC;
-       boolean_t                               reenable_workq_callback = FALSE;
-       boolean_t                               depress_option = FALSE;
-       boolean_t                               wait_option = FALSE;
-
-    /*
-     * Validate and process option.
-     */
-    switch (option) {
+       thread_t                        thread = THREAD_NULL;
+       thread_t                        self = current_thread();
+       mach_port_name_t                thread_name = args->thread_name;
+       int                             option = args->option;
+       mach_msg_timeout_t              option_time = args->option_time;
+       uint32_t                        scale_factor = NSEC_PER_MSEC;
+       boolean_t                       depress_option = FALSE;
+       boolean_t                       wait_option = FALSE;
+       wait_interrupt_t                interruptible = THREAD_ABORTSAFE;
+       port_to_thread_options_t        ptt_options = PORT_TO_THREAD_NOT_CURRENT_THREAD;

 
 

+       /*
+        *      Validate and process option.
+        *
+        * OSLock boosting only applies to other threads
+        * in your same task (even if you have a port for
+        * a thread in another task)
+        */
+       switch (option) {

        case SWITCH_OPTION_NONE:
        case SWITCH_OPTION_NONE:

-               workqueue_thread_yielded();

                break;
        case SWITCH_OPTION_WAIT:
                wait_option = TRUE;
                break;
        case SWITCH_OPTION_WAIT:
                wait_option = TRUE;

-               workqueue_thread_yielded();

                break;
        case SWITCH_OPTION_DEPRESS:
                depress_option = TRUE;
                break;
        case SWITCH_OPTION_DEPRESS:
                depress_option = TRUE;

-               workqueue_thread_yielded();

                break;
        case SWITCH_OPTION_DISPATCH_CONTENTION:
                scale_factor = NSEC_PER_USEC;
                wait_option = TRUE;
                break;
        case SWITCH_OPTION_DISPATCH_CONTENTION:
                scale_factor = NSEC_PER_USEC;
                wait_option = TRUE;

-               if (thread_switch_disable_workqueue_sched_callback())
-                       reenable_workq_callback = TRUE;
+               interruptible |= THREAD_WAIT_NOREPORT;

                break;
        case SWITCH_OPTION_OSLOCK_DEPRESS:
                depress_option = TRUE;
                break;
        case SWITCH_OPTION_OSLOCK_DEPRESS:
                depress_option = TRUE;

-               if (thread_switch_disable_workqueue_sched_callback())
-                       reenable_workq_callback = TRUE;
+               interruptible |= THREAD_WAIT_NOREPORT;
+               ptt_options |= PORT_TO_THREAD_IN_CURRENT_TASK;

                break;
        case SWITCH_OPTION_OSLOCK_WAIT:
                wait_option = TRUE;
                break;
        case SWITCH_OPTION_OSLOCK_WAIT:
                wait_option = TRUE;

-               if (thread_switch_disable_workqueue_sched_callback())
-                       reenable_workq_callback = TRUE;
+               interruptible |= THREAD_WAIT_NOREPORT;
+               ptt_options |= PORT_TO_THREAD_IN_CURRENT_TASK;

                break;
        default:
                break;
        default:

-           return (KERN_INVALID_ARGUMENT);
-    }
+               return KERN_INVALID_ARGUMENT;
+       }

 
        /*
         * Translate the port name if supplied.
         */
        if (thread_name != MACH_PORT_NULL) {
 
        /*
         * Translate the port name if supplied.
         */
        if (thread_name != MACH_PORT_NULL) {

-               ipc_port_t port;
-
-               if (ipc_port_translate_send(self->task->itk_space,
-                                           thread_name, &port) == KERN_SUCCESS) {
-                       ip_reference(port);
-                       ip_unlock(port);
-
-                       thread = convert_port_to_thread(port);
-                       ip_release(port);
-
-                       if (thread == self) {
-                               thread_deallocate(thread);
-                               thread = THREAD_NULL;
-                       }
-               }
+               thread = port_name_to_thread(thread_name, ptt_options);

        }
 
        if (option == SWITCH_OPTION_OSLOCK_DEPRESS || option == SWITCH_OPTION_OSLOCK_WAIT) {
                if (thread != THREAD_NULL) {
        }
 
        if (option == SWITCH_OPTION_OSLOCK_DEPRESS || option == SWITCH_OPTION_OSLOCK_WAIT) {
                if (thread != THREAD_NULL) {

-
-                       if (thread->task != self->task) {
-                               /*
-                                * OSLock boosting only applies to other threads
-                                * in your same task (even if you have a port for
-                                * a thread in another task)
-                                */
-
-                               thread_deallocate(thread);
-                               thread = THREAD_NULL;
-                       } else {
-                               /*
-                                * Attempt to kick the lock owner up to our same IO throttling tier.
-                                * If the thread is currently blocked in throttle_lowpri_io(),
-                                * it will immediately break out.
-                                *
-                                * TODO: SFI break out?
-                                */
-                               int new_policy = proc_get_effective_thread_policy(self, TASK_POLICY_IO);
-
-                               set_thread_iotier_override(thread, new_policy);
-                       }
+                       /*
+                        * Attempt to kick the lock owner up to our same IO throttling tier.
+                        * If the thread is currently blocked in throttle_lowpri_io(),
+                        * it will immediately break out.
+                        *
+                        * TODO: SFI break out?
+                        */
+                       int new_policy = proc_get_effective_thread_policy(self, TASK_POLICY_IO);
+
+                       set_thread_iotier_override(thread, new_policy);

                }
        }
 
                }
        }
 


@@ -341,26 +279,23 @@ thread_switch(
                /* This may return a different thread if the target is pushing on something */
                thread_t pulled_thread = thread_run_queue_remove_for_handoff(thread);
 
                /* This may return a different thread if the target is pushing on something */
                thread_t pulled_thread = thread_run_queue_remove_for_handoff(thread);
 

-               KERNEL_DEBUG_CONSTANT(MACHDBG_CODE(DBG_MACH_SCHED,MACH_SCHED_THREAD_SWITCH)|DBG_FUNC_NONE,
-                                     thread_tid(thread), thread->state,
-                                     pulled_thread ? TRUE : FALSE, 0, 0);
+               KERNEL_DEBUG_CONSTANT(MACHDBG_CODE(DBG_MACH_SCHED, MACH_SCHED_THREAD_SWITCH) | DBG_FUNC_NONE,
+                   thread_tid(thread), thread->state,
+                   pulled_thread ? TRUE : FALSE, 0, 0);

 
                if (pulled_thread != THREAD_NULL) {
                        /* We can't be dropping the last ref here */
                        thread_deallocate_safe(thread);
 
 
                if (pulled_thread != THREAD_NULL) {
                        /* We can't be dropping the last ref here */
                        thread_deallocate_safe(thread);
 

-                       if (wait_option)
-                               assert_wait_timeout((event_t)assert_wait_timeout, THREAD_ABORTSAFE,
-                                                   option_time, scale_factor);
-                       else if (depress_option)
+                       if (wait_option) {
+                               assert_wait_timeout((event_t)assert_wait_timeout, interruptible,
+                                   option_time, scale_factor);
+                       } else if (depress_option) {

                                thread_depress_ms(option_time);
                                thread_depress_ms(option_time);

+                       }

 
 

-                       self->saved.swtch.option = option;
-                       self->saved.swtch.reenable_workq_callback = reenable_workq_callback;
-
-                       thread_run(self, (thread_continue_t)thread_switch_continue, NULL, pulled_thread);
-                       /* NOTREACHED */
-                       panic("returned from thread_run!");
+                       thread_run(self, thread_switch_continue, (void *)(intptr_t)option, pulled_thread);
+                       __builtin_unreachable();

                }
 
                splx(s);
                }
 
                splx(s);


@@ -368,67 +303,50 @@ thread_switch(
                thread_deallocate(thread);
        }
 
                thread_deallocate(thread);
        }
 

-       if (wait_option)
-               assert_wait_timeout((event_t)assert_wait_timeout, THREAD_ABORTSAFE, option_time, scale_factor);
-       else if (depress_option)
-               thread_depress_ms(option_time);
-
-       self->saved.swtch.option = option;
-       self->saved.swtch.reenable_workq_callback = reenable_workq_callback;
+       if (wait_option) {
+               assert_wait_timeout((event_t)assert_wait_timeout, interruptible, option_time, scale_factor);
+       } else {
+               disable_preemption();
+               bool should_yield = SCHED(thread_should_yield)(current_processor(), current_thread());
+               enable_preemption();

 
 

-       thread_block_reason((thread_continue_t)thread_switch_continue, NULL, AST_YIELD);
-
-       if (depress_option)
-               thread_depress_abort_internal(self);
+               if (should_yield == false) {
+                       /* Early-return if yielding to the scheduler will not be beneficial */
+                       return KERN_SUCCESS;
+               }

 
 

-       if (reenable_workq_callback)
-               thread_switch_enable_workqueue_sched_callback();
+               if (depress_option) {
+                       thread_depress_ms(option_time);
+               }
+       }

 
 

-    return (KERN_SUCCESS);
+       thread_yield_with_continuation(thread_switch_continue, (void *)(intptr_t)option);
+       __builtin_unreachable();

 }
 
 }
 

-/* Returns a +1 thread reference */
-thread_t
-port_name_to_thread_for_ulock(mach_port_name_t thread_name)
+void
+thread_yield_with_continuation(
+       thread_continue_t       continuation,
+       void                            *parameter)

 {
 {

-       thread_t thread = THREAD_NULL;
-       thread_t self = current_thread();
-
-       /*
-        * Translate the port name if supplied.
-        */
-       if (thread_name != MACH_PORT_NULL) {
-               ipc_port_t port;
-
-               if (ipc_port_translate_send(self->task->itk_space,
-                                           thread_name, &port) == KERN_SUCCESS) {
-                       ip_reference(port);
-                       ip_unlock(port);
-
-                       thread = convert_port_to_thread(port);
-                       ip_release(port);
-
-                       if (thread == THREAD_NULL) {
-                               return thread;
-                       }
-
-                       if ((thread == self) || (thread->task != self->task)) {
-                               thread_deallocate(thread);
-                               thread = THREAD_NULL;
-                       }
-               }
-       }
-
-       return thread;
+       assert(continuation);
+       thread_block_reason(continuation, parameter, AST_YIELD);
+       __builtin_unreachable();

 }
 
 }
 

+

 /* This function is called after an assert_wait(), therefore it must not
  * cause another wait until after the thread_run() or thread_block()
  *
 /* This function is called after an assert_wait(), therefore it must not
  * cause another wait until after the thread_run() or thread_block()
  *

- * Consumes a ref on thread
+ *
+ * When called with a NULL continuation, the thread ref is consumed
+ * (thread_handoff_deallocate calling convention) else it is up to the
+ * continuation to do the cleanup (thread_handoff_parameter calling convention)
+ * and it instead doesn't return.

  */
  */

-wait_result_t
-thread_handoff(thread_t thread)
+static wait_result_t
+thread_handoff_internal(thread_t thread, thread_continue_t continuation,
+    void *parameter)

 {
        thread_t deallocate_thread = THREAD_NULL;
        thread_t self = current_thread();
 {
        thread_t deallocate_thread = THREAD_NULL;
        thread_t self = current_thread();


@@ -441,15 +359,17 @@ thread_handoff(thread_t thread)
 
                thread_t pulled_thread = thread_run_queue_remove_for_handoff(thread);
 
 
                thread_t pulled_thread = thread_run_queue_remove_for_handoff(thread);
 

-               KERNEL_DEBUG_CONSTANT(MACHDBG_CODE(DBG_MACH_SCHED,MACH_SCHED_THREAD_SWITCH)|DBG_FUNC_NONE,
-                                     thread_tid(thread), thread->state,
-                                     pulled_thread ? TRUE : FALSE, 0, 0);
+               KERNEL_DEBUG_CONSTANT(MACHDBG_CODE(DBG_MACH_SCHED, MACH_SCHED_THREAD_SWITCH) | DBG_FUNC_NONE,
+                   thread_tid(thread), thread->state,
+                   pulled_thread ? TRUE : FALSE, 0, 0);

 
                if (pulled_thread != THREAD_NULL) {
 
                if (pulled_thread != THREAD_NULL) {

-                       /* We can't be dropping the last ref here */
-                       thread_deallocate_safe(thread);
+                       if (continuation == NULL) {
+                               /* We can't be dropping the last ref here */
+                               thread_deallocate_safe(thread);
+                       }

 
 

-                       int result = thread_run(self, THREAD_CONTINUE_NULL, NULL, pulled_thread);
+                       int result = thread_run(self, continuation, parameter, pulled_thread);

 
                        splx(s);
                        return result;
 
                        splx(s);
                        return result;


@@ -461,7 +381,7 @@ thread_handoff(thread_t thread)
                thread = THREAD_NULL;
        }
 
                thread = THREAD_NULL;
        }
 

-       int result = thread_block(THREAD_CONTINUE_NULL);
+       int result = thread_block_parameter(continuation, parameter);

        if (deallocate_thread != THREAD_NULL) {
                thread_deallocate(deallocate_thread);
        }
        if (deallocate_thread != THREAD_NULL) {
                thread_deallocate(deallocate_thread);
        }


@@ -469,53 +389,76 @@ thread_handoff(thread_t thread)
        return result;
 }
 
        return result;
 }
 

+void
+thread_handoff_parameter(thread_t thread, thread_continue_t continuation,
+    void *parameter)
+{
+       thread_handoff_internal(thread, continuation, parameter);
+       panic("NULL continuation passed to %s", __func__);
+       __builtin_unreachable();
+}
+
+wait_result_t
+thread_handoff_deallocate(thread_t thread)
+{
+       return thread_handoff_internal(thread, NULL, NULL);
+}
+
+/*
+ * Thread depression
+ *
+ * This mechanism drops a thread to priority 0 in order for it to yield to
+ * all other runnnable threads on the system.  It can be canceled or timed out,
+ * whereupon the thread goes back to where it was.
+ *
+ * Note that TH_SFLAG_DEPRESS and TH_SFLAG_POLLDEPRESS are never set at the
+ * same time.  DEPRESS always defers to POLLDEPRESS.
+ *
+ * DEPRESS only lasts across a single thread_block call, and never returns
+ * to userspace.
+ * POLLDEPRESS can be active anywhere up until thread termination.
+ */
+

 /*
  * Depress thread's priority to lowest possible for the specified interval,
 /*
  * Depress thread's priority to lowest possible for the specified interval,

- * with a value of zero resulting in no timeout being scheduled.
+ * with an interval of zero resulting in no timeout being scheduled.
+ *
+ * Must block with AST_YIELD afterwards to take effect

  */
 void
  */
 void

-thread_depress_abstime(
-       uint64_t                                interval)
+thread_depress_abstime(uint64_t interval)

 {
 {

-       thread_t                self = current_thread();
-       uint64_t                                deadline;
-    spl_t                                      s;
-
-    s = splsched();
-    thread_lock(self);
-       if (!(self->sched_flags & TH_SFLAG_DEPRESSED_MASK)) {
-               processor_t             myprocessor = self->last_processor;
+       thread_t self = current_thread();

 
 

-               self->sched_pri = DEPRESSPRI;
+       spl_t s = splsched();
+       thread_lock(self);

 
 

-               KERNEL_DEBUG_CONSTANT(MACHDBG_CODE(DBG_MACH_SCHED, MACH_SCHED_CHANGE_PRIORITY),
-                                     (uintptr_t)thread_tid(self),
-                                     self->base_pri,
-                                     self->sched_pri,
-                                     0, /* eventually, 'reason' */
-                                     0);
+       assert((self->sched_flags & TH_SFLAG_DEPRESS) == 0);

 
 

-               myprocessor->current_pri = self->sched_pri;
+       if ((self->sched_flags & TH_SFLAG_POLLDEPRESS) == 0) {

                self->sched_flags |= TH_SFLAG_DEPRESS;
                self->sched_flags |= TH_SFLAG_DEPRESS;

+               thread_recompute_sched_pri(self, SETPRI_LAZY);

 
                if (interval != 0) {
 
                if (interval != 0) {

+                       uint64_t deadline;
+

                        clock_absolutetime_interval_to_deadline(interval, &deadline);
                        clock_absolutetime_interval_to_deadline(interval, &deadline);

-                       if (!timer_call_enter(&self->depress_timer, deadline, TIMER_CALL_USER_CRITICAL))
+                       if (!timer_call_enter(&self->depress_timer, deadline, TIMER_CALL_USER_CRITICAL)) {

                                self->depress_timer_active++;
                                self->depress_timer_active++;

+                       }

                }
        }
                }
        }

+

        thread_unlock(self);
        thread_unlock(self);

-    splx(s);
+       splx(s);

 }
 
 void
 }
 
 void

-thread_depress_ms(
-       mach_msg_timeout_t              interval)
+thread_depress_ms(mach_msg_timeout_t interval)

 {
 {

-       uint64_t                abstime;
+       uint64_t abstime;

 
 

-       clock_interval_to_absolutetime_interval(
-                                                       interval, NSEC_PER_MSEC, &abstime);
+       clock_interval_to_absolutetime_interval(interval, NSEC_PER_MSEC, &abstime);

        thread_depress_abstime(abstime);
 }
 
        thread_depress_abstime(abstime);
 }
 


@@ -523,110 +466,142 @@ thread_depress_ms(
  *     Priority depression expiration.
  */
 void
  *     Priority depression expiration.
  */
 void

-thread_depress_expire(
-       void                    *p0,
-       __unused void   *p1)
+thread_depress_expire(void      *p0,
+    __unused void      *p1)

 {
 {

-       thread_t                thread = p0;
-    spl_t                      s;
+       thread_t thread = (thread_t)p0;
+
+       spl_t s = splsched();
+       thread_lock(thread);
+
+       assert((thread->sched_flags & TH_SFLAG_DEPRESSED_MASK) != TH_SFLAG_DEPRESSED_MASK);

 
 

-    s = splsched();
-    thread_lock(thread);

        if (--thread->depress_timer_active == 0) {
                thread->sched_flags &= ~TH_SFLAG_DEPRESSED_MASK;
        if (--thread->depress_timer_active == 0) {
                thread->sched_flags &= ~TH_SFLAG_DEPRESSED_MASK;

-               thread_recompute_sched_pri(thread, FALSE);
+               if ((thread->state & TH_RUN) == TH_RUN) {
+                       thread->last_basepri_change_time = mach_absolute_time();
+               }
+               thread_recompute_sched_pri(thread, SETPRI_DEFAULT);

        }
        }

-    thread_unlock(thread);
-    splx(s);
+
+       thread_unlock(thread);
+       splx(s);

 }
 
 /*
 }
 
 /*

- *     Prematurely abort priority depression if there is one.
+ * Prematurely abort priority depression if there is one.

  */
 kern_return_t
  */
 kern_return_t

-thread_depress_abort_internal(
-       thread_t                                thread)
+thread_depress_abort(thread_t thread)

 {
 {

-    kern_return_t                      result = KERN_NOT_DEPRESSED;
-    spl_t                                      s;
-
-    s = splsched();
-    thread_lock(thread);
-       if (!(thread->sched_flags & TH_SFLAG_POLLDEPRESS)) {
-               if (thread->sched_flags & TH_SFLAG_DEPRESSED_MASK) {
-                       thread->sched_flags &= ~TH_SFLAG_DEPRESSED_MASK;
-                       thread_recompute_sched_pri(thread, FALSE);
-                       result = KERN_SUCCESS;
-               }
+       kern_return_t result = KERN_NOT_DEPRESSED;
+
+       spl_t s = splsched();
+       thread_lock(thread);
+
+       assert((thread->sched_flags & TH_SFLAG_DEPRESSED_MASK) != TH_SFLAG_DEPRESSED_MASK);

 
 

-               if (timer_call_cancel(&thread->depress_timer))
-                       thread->depress_timer_active--;
+       /*
+        * User-triggered depress-aborts should not get out
+        * of the poll-depress, but they should cancel a regular depress.
+        */
+       if ((thread->sched_flags & TH_SFLAG_POLLDEPRESS) == 0) {
+               result = thread_depress_abort_locked(thread);

        }
        }

+

        thread_unlock(thread);
        thread_unlock(thread);

-    splx(s);
+       splx(s);

 
 

-    return (result);
+       return result;

 }
 
 }
 

-void
-thread_poll_yield(
-       thread_t                self)
+/*
+ * Prematurely abort priority depression or poll depression if one is active.
+ * Called with the thread locked.
+ */
+kern_return_t
+thread_depress_abort_locked(thread_t thread)

 {
 {

-       spl_t                   s;
+       if ((thread->sched_flags & TH_SFLAG_DEPRESSED_MASK) == 0) {
+               return KERN_NOT_DEPRESSED;
+       }
+
+       assert((thread->sched_flags & TH_SFLAG_DEPRESSED_MASK) != TH_SFLAG_DEPRESSED_MASK);
+
+       thread->sched_flags &= ~TH_SFLAG_DEPRESSED_MASK;
+       if ((thread->state & TH_RUN) == TH_RUN) {
+               thread->last_basepri_change_time = mach_absolute_time();
+       }
+
+       thread_recompute_sched_pri(thread, SETPRI_LAZY);
+
+       if (timer_call_cancel(&thread->depress_timer)) {
+               thread->depress_timer_active--;
+       }

 
 

+       return KERN_SUCCESS;
+}
+
+/*
+ * Invoked as part of a polling operation like a no-timeout port receive
+ *
+ * Forces a fixpri thread to yield if it is detected polling without blocking for too long.
+ */
+void
+thread_poll_yield(thread_t self)
+{

        assert(self == current_thread());
        assert(self == current_thread());

+       assert((self->sched_flags & TH_SFLAG_DEPRESS) == 0);

 
 

-       s = splsched();
-       if (self->sched_mode == TH_MODE_FIXED) {
-               uint64_t                        total_computation, abstime;
-
-               abstime = mach_absolute_time();
-               total_computation = abstime - self->computation_epoch;
-               total_computation += self->computation_metered;
-               if (total_computation >= max_poll_computation) {
-                       processor_t             myprocessor = current_processor();
-                       ast_t                   preempt;
-
-                       thread_lock(self);
-                       if (!(self->sched_flags & TH_SFLAG_DEPRESSED_MASK)) {
-                               self->sched_pri = DEPRESSPRI;
-
-                               KERNEL_DEBUG_CONSTANT(MACHDBG_CODE(DBG_MACH_SCHED, MACH_SCHED_CHANGE_PRIORITY),
-                                                     (uintptr_t)thread_tid(self),
-                                                     self->base_pri,
-                                                     self->sched_pri,
-                                                     0, /* eventually, 'reason' */
-                                                     0);
-
-                               myprocessor->current_pri = self->sched_pri;
-                       }
-                       self->computation_epoch = abstime;
-                       self->computation_metered = 0;
-                       self->sched_flags |= TH_SFLAG_POLLDEPRESS;
+       if (self->sched_mode != TH_MODE_FIXED) {
+               return;
+       }

 
 

-                       abstime += (total_computation >> sched_poll_yield_shift);
-                       if (!timer_call_enter(&self->depress_timer, abstime, TIMER_CALL_USER_CRITICAL))
-                               self->depress_timer_active++;
+       spl_t s = splsched();
+
+       uint64_t abstime = mach_absolute_time();
+       uint64_t total_computation = abstime -
+           self->computation_epoch + self->computation_metered;
+
+       if (total_computation >= max_poll_computation) {
+               thread_lock(self);

 
 

-                       if ((preempt = csw_check(myprocessor, AST_NONE)) != AST_NONE)
-                               ast_on(preempt);
+               self->computation_epoch   = abstime;
+               self->computation_metered = 0;

 
 

-                       thread_unlock(self);
+               uint64_t yield_expiration = abstime +
+                   (total_computation >> sched_poll_yield_shift);
+
+               if (!timer_call_enter(&self->depress_timer, yield_expiration,
+                   TIMER_CALL_USER_CRITICAL)) {
+                       self->depress_timer_active++;

                }
                }

+
+               self->sched_flags |= TH_SFLAG_POLLDEPRESS;
+               thread_recompute_sched_pri(self, SETPRI_DEFAULT);
+
+               thread_unlock(self);

        }
        splx(s);
 }
 
        }
        splx(s);
 }
 

-
+/*
+ * Kernel-internal interface to yield for a specified period
+ *
+ * WARNING: Will still yield to priority 0 even if the thread is holding a contended lock!
+ */

 void
 void

-thread_yield_internal(
-       mach_msg_timeout_t      ms)
+thread_yield_internal(mach_msg_timeout_t ms)

 {
 {

-       processor_t     myprocessor;
+       thread_t self = current_thread();
+
+       assert((self->sched_flags & TH_SFLAG_DEPRESSED_MASK) != TH_SFLAG_DEPRESSED_MASK);
+
+       processor_t     myprocessor;

 
        disable_preemption();
        myprocessor = current_processor();
 
        disable_preemption();
        myprocessor = current_processor();

-       if (SCHED(processor_queue_empty)(myprocessor) && rt_runq.count == 0) {
+       if (!SCHED(thread_should_yield)(myprocessor, self)) {

                mp_enable_preemption();
 
                return;
                mp_enable_preemption();
 
                return;


@@ -637,7 +612,7 @@ thread_yield_internal(
 
        thread_block_reason(THREAD_CONTINUE_NULL, NULL, AST_YIELD);
 
 
        thread_block_reason(THREAD_CONTINUE_NULL, NULL, AST_YIELD);
 

-       thread_depress_abort_internal(current_thread());
+       thread_depress_abort(self);

 }
 
 /*
 }
 
 /*


@@ -650,8 +625,8 @@ thread_yield_internal(
 void
 thread_yield_to_preemption()
 {
 void
 thread_yield_to_preemption()
 {

-       /* 
-        * ast_pending() should ideally be called with interrupts disabled, but 
+       /*
+        * ast_pending() should ideally be called with interrupts disabled, but

         * the check here is fine because csw_check() will do the right thing.
         */
        ast_t *pending_ast = ast_pending();
         * the check here is fine because csw_check() will do the right thing.
         */
        ast_t *pending_ast = ast_pending();


@@ -665,15 +640,14 @@ thread_yield_to_preemption()
 
                p = current_processor();
                thread_lock(self);
 
                p = current_processor();
                thread_lock(self);

-               ast = csw_check(p, AST_YIELD);
+               ast = csw_check(self, p, AST_YIELD);

                ast_on(ast);
                thread_unlock(self);
 
                if (ast != AST_NONE) {
                ast_on(ast);
                thread_unlock(self);
 
                if (ast != AST_NONE) {

-                       (void)thread_block_reason(THREAD_CONTINUE_NULL, NULL, ast);
+                       (void)thread_block_reason(THREAD_CONTINUE_NULL, NULL, ast);

                }
 
                splx(s);
        }
 }
                }
 
                splx(s);
        }
 }

-