- Change mi_switch() and sched_switch() to accept an optional thread to

  switch to.  If a non-NULL thread pointer is passed in, then the CPU will
  switch to that thread directly rather than calling choosethread() to pick
  a thread to choose to.
- Make sched_switch() aware of idle threads and know to do
  TD_SET_CAN_RUN() instead of sticking them on the run queue rather than
  requiring all callers of mi_switch() to know to do this if they can be
  called from an idlethread.
- Move constants for arguments to mi_switch() and thread_single() out of
  the middle of the function prototypes and up above into their own
  section.

16 files changed, 51 insertions, 38 deletions

diff --git a/sys/kern/kern_idle.c b/sys/kern/kern_idle.c
index 62d9704..5644d8f 100644
--- a/sys/kern/kern_idle.c
+++ b/sys/kern/kern_idle.c
@@ -58,7 +58,7 @@ idle_setup(void *dummy)
 		p->p_flag |= P_NOLOAD;
 		mtx_lock_spin(&sched_lock);
 		td = FIRST_THREAD_IN_PROC(p);
-		td->td_state = TDS_CAN_RUN;
+		TD_SET_CAN_RUN(td);
 		td->td_flags |= TDF_IDLETD;
 		td->td_priority = PRI_MAX_IDLE;
 		mtx_unlock_spin(&sched_lock);
@@ -86,8 +86,7 @@ idle_proc(void *dummy)
 			cpu_idle();
 
 		mtx_lock_spin(&sched_lock);
-		td->td_state = TDS_CAN_RUN;
-		mi_switch(SW_VOL);
+		mi_switch(SW_VOL, NULL);
 		mtx_unlock_spin(&sched_lock);
 	}
 }
diff --git a/sys/kern/kern_intr.c b/sys/kern/kern_intr.c
index 920b9da..d11e9d2 100644
--- a/sys/kern/kern_intr.c
+++ b/sys/kern/kern_intr.c
@@ -416,7 +416,7 @@ ithread_schedule(struct ithd *ithread, int do_switch)
 			    ("ithread_schedule: Bad state for curthread."));
 			if (ctd->td_flags & TDF_IDLETD)
 				ctd->td_state = TDS_CAN_RUN; /* XXXKSE */
-			mi_switch(SW_INVOL);
+			mi_switch(SW_INVOL, NULL);
 		} else {
 			curthread->td_flags |= TDF_NEEDRESCHED;
 		}
@@ -618,7 +618,7 @@ restart:
 		if (!ithd->it_need) {
 			TD_SET_IWAIT(td);
 			CTR2(KTR_INTR, "%s: pid %d: done", __func__, p->p_pid);
-			mi_switch(SW_VOL);
+			mi_switch(SW_VOL, NULL);
 			CTR2(KTR_INTR, "%s: pid %d: resumed", __func__, p->p_pid);
 		}
 		mtx_unlock_spin(&sched_lock);
diff --git a/sys/kern/kern_mutex.c b/sys/kern/kern_mutex.c
index a4f0446..2388983 100644
--- a/sys/kern/kern_mutex.c
+++ b/sys/kern/kern_mutex.c
@@ -695,7 +695,7 @@ _mtx_unlock_sleep(struct mtx *m, int opts, const char *file, int line)
 			    "_mtx_unlock_sleep: %p switching out lock=%p", m,
 			    (void *)m->mtx_lock);
 
-		mi_switch(SW_INVOL);
+		mi_switch(SW_INVOL, NULL);
 		if (LOCK_LOG_TEST(&m->mtx_object, opts))
 			CTR2(KTR_LOCK, "_mtx_unlock_sleep: %p resuming lock=%p",
 			    m, (void *)m->mtx_lock);
diff --git a/sys/kern/kern_shutdown.c b/sys/kern/kern_shutdown.c
index 491bc9e..b9bfd39 100644
--- a/sys/kern/kern_shutdown.c
+++ b/sys/kern/kern_shutdown.c
@@ -307,7 +307,7 @@ boot(int howto)
 					/*
 					 * Allow interrupt threads to run
 					 */
-     					mi_switch(SW_VOL);
+     					mi_switch(SW_VOL, NULL);
      					mtx_unlock_spin(&sched_lock);
      					DELAY(1000);
    				}
diff --git a/sys/kern/kern_sig.c b/sys/kern/kern_sig.c
index e3466a9..d17cbe2 100644
--- a/sys/kern/kern_sig.c
+++ b/sys/kern/kern_sig.c
@@ -2019,7 +2019,7 @@ ptracestop(struct thread *td, int sig)
 	thread_suspend_one(td);
 	PROC_UNLOCK(p);
 	DROP_GIANT();
-	mi_switch(SW_INVOL);
+	mi_switch(SW_INVOL, NULL);
 	mtx_unlock_spin(&sched_lock);
 	PICKUP_GIANT();
 }
@@ -2168,7 +2168,7 @@ issignal(td)
 				thread_suspend_one(td);
 				PROC_UNLOCK(p);
 				DROP_GIANT();
-				mi_switch(SW_INVOL);
+				mi_switch(SW_INVOL, NULL);
 				mtx_unlock_spin(&sched_lock);
 				PICKUP_GIANT();
 				PROC_LOCK(p);
diff --git a/sys/kern/kern_subr.c b/sys/kern/kern_subr.c
index d662f5e..7ccccdd 100644
--- a/sys/kern/kern_subr.c
+++ b/sys/kern/kern_subr.c
@@ -430,7 +430,7 @@ uio_yield(void)
 	mtx_lock_spin(&sched_lock);
 	DROP_GIANT();
 	sched_prio(td, td->td_ksegrp->kg_user_pri); /* XXXKSE */
-	mi_switch(SW_INVOL);
+	mi_switch(SW_INVOL, NULL);
 	mtx_unlock_spin(&sched_lock);
 	PICKUP_GIANT();
 }
diff --git a/sys/kern/kern_synch.c b/sys/kern/kern_synch.c
index 54f7cdf..5c211d1 100644
--- a/sys/kern/kern_synch.c
+++ b/sys/kern/kern_synch.c
@@ -285,7 +285,7 @@ wakeup_one(ident)
  * The machine independent parts of context switching.
  */
 void
-mi_switch(int flags)
+mi_switch(int flags, struct thread *newtd)
 {
 	struct bintime new_switchtime;
 	struct thread *td;
@@ -349,7 +349,7 @@ mi_switch(int flags)
 	    (void *)td, (long)p->p_pid, p->p_comm);
 	if (td->td_proc->p_flag & P_SA)
 		thread_switchout(td);
-	sched_switch(td);
+	sched_switch(td, newtd);
 
 	CTR3(KTR_PROC, "mi_switch: new thread %p (pid %ld, %s)",
 	    (void *)td, (long)p->p_pid, p->p_comm);
@@ -468,7 +468,7 @@ yield(struct thread *td, struct yield_args *uap)
 	mtx_assert(&Giant, MA_NOTOWNED);
 	mtx_lock_spin(&sched_lock);
 	sched_prio(td, PRI_MAX_TIMESHARE);
-	mi_switch(SW_VOL);
+	mi_switch(SW_VOL, NULL);
 	mtx_unlock_spin(&sched_lock);
 	td->td_retval[0] = 0;
 	return (0);
diff --git a/sys/kern/kern_thread.c b/sys/kern/kern_thread.c
index 8554b48..4b7bd47 100644
--- a/sys/kern/kern_thread.c
+++ b/sys/kern/kern_thread.c
@@ -908,7 +908,7 @@ thread_single(int force_exit)
 		 */
 		thread_suspend_one(td);
 		PROC_UNLOCK(p);
-		mi_switch(SW_VOL);
+		mi_switch(SW_VOL, NULL);
 		mtx_unlock_spin(&sched_lock);
 		PROC_LOCK(p);
 		mtx_lock_spin(&sched_lock);
@@ -1011,7 +1011,7 @@ thread_suspend_check(int return_instead)
 			}
 		}
 		PROC_UNLOCK(p);
-		mi_switch(SW_INVOL);
+		mi_switch(SW_INVOL, NULL);
 		mtx_unlock_spin(&sched_lock);
 		PROC_LOCK(p);
 	}
diff --git a/sys/kern/sched_4bsd.c b/sys/kern/sched_4bsd.c
index ae8492f..5d8961e 100644
--- a/sys/kern/sched_4bsd.c
+++ b/sys/kern/sched_4bsd.c
@@ -637,9 +637,8 @@ sched_sleep(struct thread *td)
 }
 
 void
-sched_switch(struct thread *td)
+sched_switch(struct thread *td, struct thread *newtd)
 {
-	struct thread *newtd;
 	struct kse *ke;
 	struct proc *p;
 
@@ -651,6 +650,8 @@ sched_switch(struct thread *td)
 
 	if ((p->p_flag & P_NOLOAD) == 0)
 		sched_tdcnt--;
+	if (newtd != NULL && (newtd->td_proc->p_flag & P_NOLOAD) == 0)
+		sched_tdcnt++;
 	td->td_lastcpu = td->td_oncpu;
 	td->td_last_kse = ke;
 	td->td_flags &= ~TDF_NEEDRESCHED;
@@ -658,9 +659,12 @@ sched_switch(struct thread *td)
 	/*
 	 * At the last moment, if this thread is still marked RUNNING,
 	 * then put it back on the run queue as it has not been suspended
-	 * or stopped or any thing else similar.
+	 * or stopped or any thing else similar.  We never put the idle
+	 * threads on the run queue, however.
 	 */
-	if (TD_IS_RUNNING(td)) {
+	if (td == PCPU_GET(idlethread))
+		TD_SET_CAN_RUN(td);
+	else if (TD_IS_RUNNING(td)) {
 		/* Put us back on the run queue (kse and all). */
 		setrunqueue(td);
 	} else if (p->p_flag & P_SA) {
@@ -671,7 +675,8 @@ sched_switch(struct thread *td)
 		 */
 		kse_reassign(ke);
 	}
-	newtd = choosethread();
+	if (newtd == NULL)
+		newtd = choosethread();
 	if (td != newtd)
 		cpu_switch(td, newtd);
 	sched_lock.mtx_lock = (uintptr_t)td;
@@ -830,7 +835,7 @@ sched_bind(struct thread *td, int cpu)
 
 	ke->ke_state = KES_THREAD;
 
-	mi_switch(SW_VOL);
+	mi_switch(SW_VOL, NULL);
 #endif
 }
 
diff --git a/sys/kern/sched_ule.c b/sys/kern/sched_ule.c
index 874dbd1..6340b71 100644
--- a/sys/kern/sched_ule.c
+++ b/sys/kern/sched_ule.c
@@ -788,7 +788,7 @@ kseq_choose(struct kseq *kseq)
 		ke = runq_choose(kseq->ksq_curr);
 		if (ke == NULL) {
 			/*
-			 * We already swaped once and didn't get anywhere.
+			 * We already swapped once and didn't get anywhere.
 			 */
 			if (swap)
 				break;
@@ -1128,9 +1128,8 @@ sched_prio(struct thread *td, u_char prio)
 }
 
 void
-sched_switch(struct thread *td)
+sched_switch(struct thread *td, struct thread *newtd)
 {
-	struct thread *newtd;
 	struct kse *ke;
 
 	mtx_assert(&sched_lock, MA_OWNED);
@@ -1147,7 +1146,9 @@ sched_switch(struct thread *td)
 	 * to the new cpu.  This is the case in sched_bind().
 	 */
 	if ((ke->ke_flags & KEF_ASSIGNED) == 0) {
-		if (TD_IS_RUNNING(td)) {
+		if (td == PCPU_GET(idlethread))
+			TD_SET_CAN_RUN(td);
+		else if (TD_IS_RUNNING(td)) {
 			kseq_load_rem(KSEQ_CPU(ke->ke_cpu), ke);
 			setrunqueue(td);
 		} else {
@@ -1163,7 +1164,10 @@ sched_switch(struct thread *td)
 				kse_reassign(ke);
 		}
 	}
-	newtd = choosethread();
+	if (newtd == NULL)
+		newtd = choosethread();
+	else
+		kseq_load_add(KSEQ_SELF(), newtd->td_kse);
 	if (td != newtd)
 		cpu_switch(td, newtd);
 	sched_lock.mtx_lock = (uintptr_t)td;
diff --git a/sys/kern/subr_sleepqueue.c b/sys/kern/subr_sleepqueue.c
index 79bbf2c..a7a07d5 100644
--- a/sys/kern/subr_sleepqueue.c
+++ b/sys/kern/subr_sleepqueue.c
@@ -409,7 +409,7 @@ sleepq_switch(void *wchan)
 
 	sched_sleep(td);
 	TD_SET_SLEEPING(td);
-	mi_switch(SW_VOL);
+	mi_switch(SW_VOL, NULL);
 	KASSERT(TD_IS_RUNNING(td), ("running but not TDS_RUNNING"));
 	CTR3(KTR_PROC, "sleepq resume: thread %p (pid %ld, %s)",
 	    (void *)td, (long)td->td_proc->p_pid, (void *)td->td_proc->p_comm);
@@ -449,7 +449,7 @@ sleepq_check_timeout(void)
 	else if (callout_stop(&td->td_slpcallout) == 0) {
 		td->td_flags |= TDF_TIMEOUT;
 		TD_SET_SLEEPING(td);
-		mi_switch(SW_INVOL);
+		mi_switch(SW_INVOL, NULL);
 	}
 	return (0);
 }
diff --git a/sys/kern/subr_trap.c b/sys/kern/subr_trap.c
index 11550c0..b7b461e 100644
--- a/sys/kern/subr_trap.c
+++ b/sys/kern/subr_trap.c
@@ -242,7 +242,7 @@ ast(struct trapframe *framep)
 #endif
 		mtx_lock_spin(&sched_lock);
 		sched_prio(td, kg->kg_user_pri);
-		mi_switch(SW_INVOL);
+		mi_switch(SW_INVOL, NULL);
 		mtx_unlock_spin(&sched_lock);
 #ifdef KTRACE
 		if (KTRPOINT(td, KTR_CSW))
diff --git a/sys/kern/subr_turnstile.c b/sys/kern/subr_turnstile.c
index cae7e9a..c983379 100644
--- a/sys/kern/subr_turnstile.c
+++ b/sys/kern/subr_turnstile.c
@@ -559,7 +559,7 @@ turnstile_wait(struct turnstile *ts, struct lock_object *lock,
 		CTR4(KTR_LOCK, "%s: td %p blocked on [%p] %s", __func__, td,
 		    lock, lock->lo_name);
 
-	mi_switch(SW_VOL);
+	mi_switch(SW_VOL, NULL);
 
 	if (LOCK_LOG_TEST(lock, 0))
 		CTR4(KTR_LOCK, "%s: td %p free from blocked on [%p] %s",
diff --git a/sys/sys/proc.h b/sys/sys/proc.h
index 46b5f5d..6e268b2 100644
--- a/sys/sys/proc.h
+++ b/sys/sys/proc.h
@@ -675,6 +675,16 @@ struct proc {
 
 #ifdef _KERNEL
 
+/* Flags for mi_switch(). */
+#define	SW_VOL		0x0001		/* Voluntary switch. */
+#define	SW_INVOL	0x0002		/* Involuntary switch. */
+
+/* How values for thread_single(). */
+#define	SINGLE_NO_EXIT	0
+#define	SINGLE_EXIT	1
+
+/* XXXKSE: Missing values for thread_signal_check(). */
+
 #ifdef MALLOC_DECLARE
 MALLOC_DECLARE(M_PARGS);
 MALLOC_DECLARE(M_PGRP);
@@ -840,10 +850,7 @@ void	fork_exit(void (*)(void *, struct trapframe *), void *,
 void	fork_return(struct thread *, struct trapframe *);
 int	inferior(struct proc *p);
 int	leavepgrp(struct proc *p);
-void	mi_switch(int flags);
-/* Flags for mi_switch(). */
-#define	SW_VOL		0x0001		/* Voluntary switch. */
-#define	SW_INVOL	0x0002		/* Involuntary switch. */
+void	mi_switch(int flags, struct thread *newtd);
 int	p_candebug(struct thread *td, struct proc *p);
 int	p_cansee(struct thread *td, struct proc *p);
 int	p_cansched(struct thread *td, struct proc *p);
@@ -906,8 +913,6 @@ void	thread_link(struct thread *td, struct ksegrp *kg);
 void	thread_reap(void);
 struct thread *thread_schedule_upcall(struct thread *td, struct kse_upcall *ku);
 int	thread_single(int how);
-#define	SINGLE_NO_EXIT 0			/* values for 'how' */
-#define	SINGLE_EXIT 1
 void	thread_single_end(void);
 void	thread_stash(struct thread *td);
 int	thread_suspend_check(int how);
diff --git a/sys/sys/sched.h b/sys/sys/sched.h
index 8a7e84e..461842a 100644
--- a/sys/sys/sched.h
+++ b/sys/sys/sched.h
@@ -66,7 +66,7 @@ void	sched_fork_thread(struct thread *td, struct thread *child);
 fixpt_t	sched_pctcpu(struct thread *td);
 void	sched_prio(struct thread *td, u_char prio);
 void	sched_sleep(struct thread *td);
-void	sched_switch(struct thread *td);
+void	sched_switch(struct thread *td, struct thread *newtd);
 void	sched_userret(struct thread *td);
 void	sched_wakeup(struct thread *td);
 
diff --git a/sys/vm/vm_zeroidle.c b/sys/vm/vm_zeroidle.c
index 06338ff..71b7a71 100644
--- a/sys/vm/vm_zeroidle.c
+++ b/sys/vm/vm_zeroidle.c
@@ -153,7 +153,7 @@ vm_pagezero(void __unused *arg)
 			pages += vm_page_zero_idle();
 			if (pages > idlezero_maxrun || sched_runnable()) {
 				mtx_lock_spin(&sched_lock);
-				mi_switch(SW_VOL);
+				mi_switch(SW_VOL, NULL);
 				mtx_unlock_spin(&sched_lock);
 				pages = 0;
 			}