19 files changed, 275 insertions, 259 deletions

diff --git a/sys/alpha/osf1/osf1_misc.c b/sys/alpha/osf1/osf1_misc.c
index 9cebf524dbfb..8f3a7b26a42a 100644
--- a/sys/alpha/osf1/osf1_misc.c
+++ b/sys/alpha/osf1/osf1_misc.c
@@ -1335,29 +1335,16 @@ osf1_getrusage(td, uap)
 	struct thread *td;
 	struct osf1_getrusage_args *uap;
 {
-	struct proc *p;
-	struct rusage *rup;
 	struct osf1_rusage oru;
+	struct rusage ru;
+	int error;
 
-	p = td->td_proc;
-	switch (uap->who) {
-	case RUSAGE_SELF:
-		rup = &p->p_stats->p_ru;
-		mtx_lock_spin(&sched_lock);
-		calcru(p, &rup->ru_utime, &rup->ru_stime, NULL);
-		mtx_unlock_spin(&sched_lock);
-		break;
-
-	case RUSAGE_CHILDREN:
-		rup = &p->p_stats->p_cru;
-		break;
-
-	default:
-		return (EINVAL);
-	}
-	TV_CP(rup->ru_utime, oru.ru_utime);
-	TV_CP(rup->ru_stime, oru.ru_stime);
-	bcopy(&(rup->ru_first), &(oru.ru_first),
+	error = kern_getrusage(td, uap->who, &ru);
+	if (error)
+		return (error);
+	TV_CP(ru.ru_utime, oru.ru_utime);
+	TV_CP(ru.ru_stime, oru.ru_stime);
+	bcopy(&(ru.ru_first), &(oru.ru_first),
 	    (&(oru.ru_last) - &(oru.ru_first)));
 
 	return (copyout((caddr_t)&oru, (caddr_t)uap->rusage,
@@ -1372,9 +1359,13 @@ osf1_wait4(td, uap)
 {
 	int error, status;
 	struct osf1_rusage oru;
-	struct rusage ru;
+	struct rusage ru, *rup;
 
-	error = kern_wait(td, uap->pid, &status, uap->options, &ru);
+	if (uap->rusage != NULL)
+		rup = &ru;
+	else
+		rup = NULL;
+	error = kern_wait(td, uap->pid, &status, uap->options, rup);
 	if (error)
 		return (error);
 	if (uap->status != NULL)
diff --git a/sys/amd64/linux32/linux32_machdep.c b/sys/amd64/linux32/linux32_machdep.c
index 731754cd04c7..0f0a733a9ba0 100644
--- a/sys/amd64/linux32/linux32_machdep.c
+++ b/sys/amd64/linux32/linux32_machdep.c
@@ -940,24 +940,14 @@ linux_nanosleep(struct thread *td, struct linux_nanosleep_args *uap)
 int
 linux_getrusage(struct thread *td, struct linux_getrusage_args *uap)
 {
+	struct l_rusage s32;
+	struct rusage s;
 	int error;
-	caddr_t sg;
-	struct l_rusage *p32, s32;
-	struct rusage *p = NULL, s;
-
-	p32 = uap->rusage;
-	if (p32 != NULL) {
-		sg = stackgap_init();
-		p = stackgap_alloc(&sg, sizeof(struct rusage));
-		uap->rusage = (struct l_rusage *)p;
-	}
-	error = getrusage(td, (struct getrusage_args *) uap);
+
+	error = kern_getrusage(td, uap->who, &s);
 	if (error != 0)
 		return (error);
-	if (p32 != NULL) {
-		error = copyin(p, &s, sizeof(s));
-		if (error != 0)
-			return (error);
+	if (uap->rusage != NULL) {
 		s32.ru_utime.tv_sec = s.ru_utime.tv_sec;
 		s32.ru_utime.tv_usec = s.ru_utime.tv_usec;
 		s32.ru_stime.tv_sec = s.ru_stime.tv_sec;
@@ -976,7 +966,7 @@ linux_getrusage(struct thread *td, struct linux_getrusage_args *uap)
 		s32.ru_nsignals = s.ru_nsignals;
 		s32.ru_nvcsw = s.ru_nvcsw;
 		s32.ru_nivcsw = s.ru_nivcsw;
-		error = copyout(&s32, p32, sizeof(s32));
+		error = copyout(&s32, uap->rusage, sizeof(s32));
 	}
 	return (error);
 }
diff --git a/sys/compat/freebsd32/freebsd32_misc.c b/sys/compat/freebsd32/freebsd32_misc.c
index 6ff5910c7ebf..b78f552dbff5 100644
--- a/sys/compat/freebsd32/freebsd32_misc.c
+++ b/sys/compat/freebsd32/freebsd32_misc.c
@@ -91,9 +91,13 @@ freebsd32_wait4(struct thread *td, struct freebsd32_wait4_args *uap)
 {
 	int error, status;
 	struct rusage32 ru32;
-	struct rusage ru;
+	struct rusage ru, *rup;
 
-	error = kern_wait(td, uap->pid, &status, uap->options, &ru);
+	if (uap->rusage != NULL)
+		rup = &ru;
+	else
+		rup = NULL;
+	error = kern_wait(td, uap->pid, &status, uap->options, rup);
 	if (error)
 		return (error);
 	if (uap->status != NULL)
@@ -639,24 +643,14 @@ freebsd32_gettimeofday(struct thread *td,
 int
 freebsd32_getrusage(struct thread *td, struct freebsd32_getrusage_args *uap)
 {
+	struct rusage32 s32;
+	struct rusage s;
 	int error;
-	caddr_t sg;
-	struct rusage32 *p32, s32;
-	struct rusage *p = NULL, s;
 
-	p32 = uap->rusage;
-	if (p32) {
-		sg = stackgap_init();
-		p = stackgap_alloc(&sg, sizeof(struct rusage));
-		uap->rusage = (struct rusage32 *)p;
-	}
-	error = getrusage(td, (struct getrusage_args *) uap);
+	error = kern_getrusage(td, uap->who, &s);
 	if (error)
 		return (error);
-	if (p32) {
-		error = copyin(p, &s, sizeof(s));
-		if (error)
-			return (error);
+	if (uap->rusage != NULL) {
 		TV_CP(s, s32, ru_utime);
 		TV_CP(s, s32, ru_stime);
 		CP(s, s32, ru_maxrss);
@@ -673,7 +667,7 @@ freebsd32_getrusage(struct thread *td, struct freebsd32_getrusage_args *uap)
 		CP(s, s32, ru_nsignals);
 		CP(s, s32, ru_nvcsw);
 		CP(s, s32, ru_nivcsw);
-		error = copyout(&s32, p32, sizeof(s32));
+		error = copyout(&s32, uap->rusage, sizeof(s32));
 	}
 	return (error);
 }
diff --git a/sys/compat/linux/linux_misc.c b/sys/compat/linux/linux_misc.c
index 3f080280bdcb..f17e0adf3772 100644
--- a/sys/compat/linux/linux_misc.c
+++ b/sys/compat/linux/linux_misc.c
@@ -689,9 +689,9 @@ struct l_times_argv {
 int
 linux_times(struct thread *td, struct linux_times_args *args)
 {
-	struct timeval tv;
+	struct timeval tv, utime, stime, cutime, cstime;
 	struct l_times_argv tms;
-	struct rusage ru;
+	struct proc *p;
 	int error;
 
 #ifdef DEBUG
@@ -699,15 +699,17 @@ linux_times(struct thread *td, struct linux_times_args *args)
 		printf(ARGS(times, "*"));
 #endif
 
-	mtx_lock_spin(&sched_lock);
-	calcru(td->td_proc, &ru.ru_utime, &ru.ru_stime, NULL);
-	mtx_unlock_spin(&sched_lock);
+	p = td->td_proc;
+	PROC_LOCK(p);
+	calcru(p, &utime, &stime);
+	calccru(p, &cutime, &cstime);
+	PROC_UNLOCK(p);
 
-	tms.tms_utime = CONVTCK(ru.ru_utime);
-	tms.tms_stime = CONVTCK(ru.ru_stime);
+	tms.tms_utime = CONVTCK(utime);
+	tms.tms_stime = CONVTCK(stime);
 
-	tms.tms_cutime = CONVTCK(td->td_proc->p_stats->p_cru.ru_utime);
-	tms.tms_cstime = CONVTCK(td->td_proc->p_stats->p_cru.ru_stime);
+	tms.tms_cutime = CONVTCK(cutime);
+	tms.tms_cstime = CONVTCK(cstime);
 
 	if ((error = copyout(&tms, args->buf, sizeof(tms))))
 		return error;
@@ -851,7 +853,7 @@ int
 linux_wait4(struct thread *td, struct linux_wait4_args *args)
 {
 	int error, options, tmpstat;
-	struct rusage ru;
+	struct rusage ru, *rup;
 	struct proc *p;
 
 #ifdef DEBUG
@@ -866,7 +868,11 @@ linux_wait4(struct thread *td, struct linux_wait4_args *args)
 	if (args->options & __WCLONE)
 		options |= WLINUXCLONE;
 
-	error = kern_wait(td, args->pid, &tmpstat, options, &ru);
+	if (args->rusage != NULL)
+		rup = &ru;
+	else
+		rup = NULL;
+	error = kern_wait(td, args->pid, &tmpstat, options, rup);
 	if (error)
 		return error;
 
diff --git a/sys/compat/svr4/svr4_misc.c b/sys/compat/svr4/svr4_misc.c
index c0109d794888..349487bca870 100644
--- a/sys/compat/svr4/svr4_misc.c
+++ b/sys/compat/svr4/svr4_misc.c
@@ -861,44 +861,30 @@ svr4_sys_times(td, uap)
 	struct thread *td;
 	struct svr4_sys_times_args *uap;
 {
-	int			 error, *retval = td->td_retval;
-	struct tms		 tms;
-	struct timeval		 t;
-	struct rusage		*ru;
-	struct rusage		 r;
-	struct getrusage_args 	 ga;
-
-	caddr_t sg = stackgap_init();
-	ru = stackgap_alloc(&sg, sizeof(struct rusage));
-
-	ga.who = RUSAGE_SELF;
-	ga.rusage = ru;
+	struct timeval tv, utime, stime, cutime, cstime;
+	struct tms tms;
+	struct proc *p;
+	int error;
 
-	error = getrusage(td, &ga);
-	if (error)
-		return error;
+	p = td->td_proc;
+	PROC_LOCK(p);
+	calcru(p, &utime, &stime);
+	calccru(p, &cutime, &cstime);
+	PROC_UNLOCK(p);
 
-	if ((error = copyin(ru, &r, sizeof r)) != 0)
-		return error;
+	tms.tms_utime = timeval_to_clock_t(&utime);
+	tms.tms_stime = timeval_to_clock_t(&stime);
 
-	tms.tms_utime = timeval_to_clock_t(&r.ru_utime);
-	tms.tms_stime = timeval_to_clock_t(&r.ru_stime);
+	tms.tms_cutime = timeval_to_clock_t(&cutime);
+	tms.tms_cstime = timeval_to_clock_t(&cstime);
 
-	ga.who = RUSAGE_CHILDREN;
-	error = getrusage(td, &ga);
+	error = copyout(&tms, uap->tp, sizeof(tms));
 	if (error)
-		return error;
-
-	if ((error = copyin(ru, &r, sizeof r)) != 0)
-		return error;
-
-	tms.tms_cutime = timeval_to_clock_t(&r.ru_utime);
-	tms.tms_cstime = timeval_to_clock_t(&r.ru_stime);
-
-	microtime(&t);
-	*retval = timeval_to_clock_t(&t);
+		return (error);
 
-	return copyout(&tms, uap->tp, sizeof(tms));
+	microtime(&tv);
+	td->td_retval[0] = (int)timeval_to_clock_t(&tv);
+	return (0);
 }
 
 
@@ -1149,6 +1135,7 @@ svr4_setinfo(p, st, s)
 	int st;
 	svr4_siginfo_t *s;
 {
+	struct timeval utime, stime;
 	svr4_siginfo_t i;
 	int sig;
 
@@ -1159,16 +1146,11 @@ svr4_setinfo(p, st, s)
 
 	if (p) {
 		i.si_pid = p->p_pid;
-		mtx_lock_spin(&sched_lock);
-		if (p->p_state == PRS_ZOMBIE) {
-			i.si_stime = p->p_ru->ru_stime.tv_sec;
-			i.si_utime = p->p_ru->ru_utime.tv_sec;
-		}
-		else {
-			i.si_stime = p->p_stats->p_ru.ru_stime.tv_sec;
-			i.si_utime = p->p_stats->p_ru.ru_utime.tv_sec;
-		}
-		mtx_unlock_spin(&sched_lock);
+		PROC_LOCK(p);
+		calcru(p, &utime, &stime);
+		PROC_UNLOCK(p);
+		i.si_stime = stime.tv_sec;
+		i.si_utime = utime.tv_sec;
 	}
 
 	if (WIFEXITED(st)) {
@@ -1296,7 +1278,8 @@ loop:
 			PROC_UNLOCK(q);
 			sx_xunlock(&proctree_lock);
 			q->p_xstat = 0;
-			ruadd(&p->p_stats->p_cru, q->p_ru);
+			ruadd(&p->p_stats->p_cru, &p->p_crux, q->p_ru,
+			    &q->p_rux);
 			FREE(q->p_ru, M_ZOMBIE);
 			q->p_ru = NULL;
 
diff --git a/sys/fs/procfs/procfs_status.c b/sys/fs/procfs/procfs_status.c
index 87d3c0b51fda..b832f42289e8 100644
--- a/sys/fs/procfs/procfs_status.c
+++ b/sys/fs/procfs/procfs_status.c
@@ -124,12 +124,12 @@ procfs_doprocstatus(PFS_FILL_ARGS)
 		} else
 			wmesg = "nochan";
 	}
+	mtx_unlock_spin(&sched_lock);
 
 	if (p->p_sflag & PS_INMEM) {
 		struct timeval start, ut, st;
 
-		calcru(p, &ut, &st, (struct timeval *) NULL);
-		mtx_unlock_spin(&sched_lock);
+		calcru(p, &ut, &st);
 		start = p->p_stats->p_start;
 		timevaladd(&start, &boottime);
 		sbuf_printf(sb, " %ld,%ld %ld,%ld %ld,%ld",
@@ -137,7 +137,6 @@ procfs_doprocstatus(PFS_FILL_ARGS)
 		    ut.tv_sec, ut.tv_usec,
 		    st.tv_sec, st.tv_usec);
 	} else {
-		mtx_unlock_spin(&sched_lock);
 		sbuf_printf(sb, " -1,-1 -1,-1 -1,-1");
 	}
 
diff --git a/sys/kern/init_main.c b/sys/kern/init_main.c
index 6acb6b01252f..e23e44608108 100644
--- a/sys/kern/init_main.c
+++ b/sys/kern/init_main.c
@@ -468,8 +468,8 @@ proc0_post(void *dummy __unused)
 	sx_slock(&allproc_lock);
 	LIST_FOREACH(p, &allproc, p_list) {
 		microuptime(&p->p_stats->p_start);
-		p->p_runtime.sec = 0;
-		p->p_runtime.frac = 0;
+		p->p_rux.rux_runtime.sec = 0;
+		p->p_rux.rux_runtime.frac = 0;
 	}
 	sx_sunlock(&allproc_lock);
 	binuptime(PCPU_PTR(switchtime));
diff --git a/sys/kern/kern_acct.c b/sys/kern/kern_acct.c
index 5dca0f60dd13..2b810edc47ca 100644
--- a/sys/kern/kern_acct.c
+++ b/sys/kern/kern_acct.c
@@ -249,9 +249,7 @@ acct_process(td)
 	bcopy(p->p_comm, acct.ac_comm, sizeof acct.ac_comm);
 
 	/* (2) The amount of user and system time that was used */
-	mtx_lock_spin(&sched_lock);
-	calcru(p, &ut, &st, NULL);
-	mtx_unlock_spin(&sched_lock);
+	calcru(p, &ut, &st);
 	acct.ac_utime = encode_comp_t(ut.tv_sec, ut.tv_usec);
 	acct.ac_stime = encode_comp_t(st.tv_sec, st.tv_usec);
 
diff --git a/sys/kern/kern_clock.c b/sys/kern/kern_clock.c
index 21d627551c31..ab2f12957ec1 100644
--- a/sys/kern/kern_clock.c
+++ b/sys/kern/kern_clock.c
@@ -363,7 +363,7 @@ stopprofclock(p)
  * Statistics clock.  Grab profile sample, and if divider reaches 0,
  * do process and kernel statistics.  Most of the statistics are only
  * used by user-level statistics programs.  The main exceptions are
- * ke->ke_uticks, p->p_sticks, p->p_iticks, and p->p_estcpu.
+ * ke->ke_uticks, p->p_rux.rux_sticks, p->p_rux.rux_iticks, and p->p_estcpu.
  * This should be called by all active processors.
  */
 void
@@ -386,7 +386,7 @@ statclock(frame)
 		 */
 		if (p->p_flag & P_SA)
 			thread_statclock(1);
-		p->p_uticks++;
+		p->p_rux.rux_uticks++;
 		if (p->p_nice > NZERO)
 			cp_time[CP_NICE]++;
 		else
@@ -405,13 +405,13 @@ statclock(frame)
 		 * in ``non-process'' (i.e., interrupt) work.
 		 */
 		if ((td->td_ithd != NULL) || td->td_intr_nesting_level >= 2) {
-			p->p_iticks++;
+			p->p_rux.rux_iticks++;
 			cp_time[CP_INTR]++;
 		} else {
 			if (p->p_flag & P_SA)
 				thread_statclock(0);
 			td->td_sticks++;
-			p->p_sticks++;
+			p->p_rux.rux_sticks++;
 			if (p != PCPU_GET(idlethread)->td_proc)
 				cp_time[CP_SYS]++;
 			else
diff --git a/sys/kern/kern_exit.c b/sys/kern/kern_exit.c
index cd7e2a46f46f..ae44fc348b14 100644
--- a/sys/kern/kern_exit.c
+++ b/sys/kern/kern_exit.c
@@ -410,20 +410,16 @@ retry:
 	}
 
 	/*
-	 * Save exit status and final rusage info, adding in child rusage
-	 * info and self times.
+	 * Save exit status and finalize rusage info except for times,
+	 * adding in child rusage info.
 	 */
-	mtx_lock(&Giant);	
 	PROC_LOCK(p);
 	p->p_xstat = rv;
 	p->p_xthread = td;
+	p->p_stats->p_ru.ru_nvcsw++;
 	*p->p_ru = p->p_stats->p_ru;
-	mtx_lock_spin(&sched_lock);
-	calcru(p, &p->p_ru->ru_utime, &p->p_ru->ru_stime, NULL);
-	mtx_unlock_spin(&sched_lock);
-	ruadd(p->p_ru, &p->p_stats->p_cru);
+	ruadd(p->p_ru, &p->p_rux, &p->p_stats->p_cru, &p->p_crux);
 
-	mtx_unlock(&Giant);	
 	/*
 	 * Notify interested parties of our demise.
 	 */
@@ -487,9 +483,6 @@ retry:
 	PROC_LOCK(p->p_pptr);
 	sx_xunlock(&proctree_lock);
 
-	while (mtx_owned(&Giant))
-		mtx_unlock(&Giant);
-
 	/*
 	 * We have to wait until after acquiring all locks before
 	 * changing p_state.  We need to avoid all possible context
@@ -508,8 +501,8 @@ retry:
 
 	/* Do the same timestamp bookkeeping that mi_switch() would do. */
 	binuptime(&new_switchtime);
-	bintime_add(&p->p_runtime, &new_switchtime);
-	bintime_sub(&p->p_runtime, PCPU_PTR(switchtime));
+	bintime_add(&p->p_rux.rux_runtime, &new_switchtime);
+	bintime_sub(&p->p_rux.rux_runtime, PCPU_PTR(switchtime));
 	PCPU_SET(switchtime, new_switchtime);
 	PCPU_SET(switchticks, ticks);
 	cnt.v_swtch++;
@@ -556,10 +549,14 @@ owait(struct thread *td, struct owait_args *uap __unused)
 int
 wait4(struct thread *td, struct wait_args *uap)
 {
-	struct rusage ru;
+	struct rusage ru, *rup;
 	int error, status;
 
-	error = kern_wait(td, uap->pid, &status, uap->options, &ru);
+	if (uap->rusage != NULL)
+		rup = &ru;
+	else
+		rup = NULL;
+	error = kern_wait(td, uap->pid, &status, uap->options, rup);
 	if (uap->status != NULL && error == 0)
 		error = copyout(&status, uap->status, sizeof(status));
 	if (uap->rusage != NULL && error == 0)
@@ -612,8 +609,10 @@ loop:
 			td->td_retval[0] = p->p_pid;
 			if (status)
 				*status = p->p_xstat;	/* convert to int */
-			if (rusage)
+			if (rusage) {
 				bcopy(p->p_ru, rusage, sizeof(struct rusage));
+				calcru(p, &rusage->ru_utime, &rusage->ru_stime);
+			}
 
 			/*
 			 * If we got the child via a ptrace 'attach',
@@ -648,16 +647,15 @@ loop:
 			 * all other writes to this proc are visible now, so
 			 * no more locking is needed for p.
 			 */
-			mtx_lock(&Giant);
 			PROC_LOCK(p);
 			p->p_xstat = 0;		/* XXX: why? */
 			PROC_UNLOCK(p);
 			PROC_LOCK(q);
-			ruadd(&q->p_stats->p_cru, p->p_ru);
+			ruadd(&q->p_stats->p_cru, &q->p_crux, p->p_ru,
+			    &p->p_rux);
 			PROC_UNLOCK(q);
 			FREE(p->p_ru, M_ZOMBIE);
 			p->p_ru = NULL;
-			mtx_unlock(&Giant);
 
 			/*
 			 * Decrement the count of procs running with this uid.
diff --git a/sys/kern/kern_proc.c b/sys/kern/kern_proc.c
index 2f8b4df28ba9..f4f564847877 100644
--- a/sys/kern/kern_proc.c
+++ b/sys/kern/kern_proc.c
@@ -674,23 +674,11 @@ fill_kinfo_thread(struct thread *td, struct kinfo_proc *kp)
 		kp->ki_dsize = vm->vm_dsize;
 		kp->ki_ssize = vm->vm_ssize;
 	}
-	if ((p->p_sflag & PS_INMEM) && p->p_stats) {
-		kp->ki_start = p->p_stats->p_start;
-		timevaladd(&kp->ki_start, &boottime);
-		kp->ki_rusage = p->p_stats->p_ru;
-		calcru(p, &kp->ki_rusage.ru_utime, &kp->ki_rusage.ru_stime,
-		    NULL);
-		kp->ki_childstime = p->p_stats->p_cru.ru_stime;
-		kp->ki_childutime = p->p_stats->p_cru.ru_utime;
-		/* Some callers want child-times in a single value */
-		kp->ki_childtime = kp->ki_childstime;
-		timevaladd(&kp->ki_childtime, &kp->ki_childutime);
-	}
 	kp->ki_sflag = p->p_sflag;
 	kp->ki_swtime = p->p_swtime;
 	kp->ki_pid = p->p_pid;
 	kp->ki_nice = p->p_nice;
-	bintime2timeval(&p->p_runtime, &tv);
+	bintime2timeval(&p->p_rux.rux_runtime, &tv);
 	kp->ki_runtime = tv.tv_sec * (u_int64_t)1000000 + tv.tv_usec;
 	if (p->p_state != PRS_ZOMBIE) {
 #if 0
@@ -758,6 +746,17 @@ fill_kinfo_thread(struct thread *td, struct kinfo_proc *kp)
 		kp->ki_stat = SZOMB;
 	}
 	mtx_unlock_spin(&sched_lock);
+	if ((p->p_sflag & PS_INMEM) && p->p_stats != NULL) {
+		kp->ki_start = p->p_stats->p_start;
+		timevaladd(&kp->ki_start, &boottime);
+		kp->ki_rusage = p->p_stats->p_ru;
+		calcru(p, &kp->ki_rusage.ru_utime, &kp->ki_rusage.ru_stime);
+		calccru(p, &kp->ki_childutime, &kp->ki_childstime);
+
+		/* Some callers want child-times in a single value */
+		kp->ki_childtime = kp->ki_childstime;
+		timevaladd(&kp->ki_childtime, &kp->ki_childutime);
+	}
 	sp = NULL;
 	tp = NULL;
 	if (p->p_pgrp) {
diff --git a/sys/kern/kern_resource.c b/sys/kern/kern_resource.c
index e0a6655c6997..d9e25c81d6f5 100644
--- a/sys/kern/kern_resource.c
+++ b/sys/kern/kern_resource.c
@@ -51,6 +51,7 @@ __FBSDID("$FreeBSD$");
 #include <sys/resourcevar.h>
 #include <sys/sched.h>
 #include <sys/sx.h>
+#include <sys/syscallsubr.h>
 #include <sys/sysent.h>
 #include <sys/time.h>
 
@@ -67,6 +68,8 @@ static struct mtx uihashtbl_mtx;
 static LIST_HEAD(uihashhead, uidinfo) *uihashtbl;
 static u_long uihash;		/* size of hash table - 1 */
 
+static void	calcru1(struct proc *p, struct rusage_ext *ruxp,
+		    struct timeval *up, struct timeval *sp);
 static int	donice(struct thread *td, struct proc *chgp, int n);
 static struct uidinfo *uilookup(uid_t uid);
 
@@ -691,65 +694,84 @@ getrlimit(td, uap)
  * system, and interrupt time usage.
  */
 void
-calcru(p, up, sp, ip)
+calcru(p, up, sp)
 	struct proc *p;
 	struct timeval *up;
 	struct timeval *sp;
-	struct timeval *ip;
 {
-	struct bintime bt, rt;
-	struct timeval tv;
+	struct bintime bt;
+	struct rusage_ext rux;
 	struct thread *td;
-	/* {user, system, interrupt, total} {ticks, usec}; previous tu: */
-	u_int64_t ut, uu, st, su, it, iu, tt, tu, ptu;
-	int problemcase;
+	int bt_valid;
 
-	mtx_assert(&sched_lock, MA_OWNED);
-	/* XXX: why spl-protect ?  worst case is an off-by-one report */
+	PROC_LOCK_ASSERT(p, MA_OWNED);
+	mtx_assert(&sched_lock, MA_NOTOWNED);
+	bt_valid = 0;
+	mtx_lock_spin(&sched_lock);
+	rux = p->p_rux;
+	FOREACH_THREAD_IN_PROC(p, td) {
+		if (TD_IS_RUNNING(td)) {
+			/*
+			 * Adjust for the current time slice.  This is
+			 * actually fairly important since the error here is
+			 * on the order of a time quantum which is much
+			 * greater than the precision of binuptime().
+			 */
+			KASSERT(td->td_oncpu != NOCPU,
+			    ("%s: running thread has no CPU", __func__));
+			if (!bt_valid) {
+				binuptime(&bt);
+				bt_valid = 1;
+			}
+			bintime_add(&rux.rux_runtime, &bt);
+			bintime_sub(&rux.rux_runtime,
+			    &pcpu_find(td->td_oncpu)->pc_switchtime);
+		}
+	}
+	mtx_unlock_spin(&sched_lock);
+	calcru1(p, &rux, up, sp);
+	p->p_rux.rux_uu = rux.rux_uu;
+	p->p_rux.rux_su = rux.rux_su;
+	p->p_rux.rux_iu = rux.rux_iu;
+}
 
-	ut = p->p_uticks;
-	st = p->p_sticks;
-	it = p->p_iticks;
+void
+calccru(p, up, sp)
+	struct proc *p;
+	struct timeval *up;
+	struct timeval *sp;
+{
+
+	PROC_LOCK_ASSERT(p, MA_OWNED);
+	calcru1(p, &p->p_crux, up, sp);
+}
+
+static void
+calcru1(p, ruxp, up, sp)
+	struct proc *p;
+	struct rusage_ext *ruxp;
+	struct timeval *up;
+	struct timeval *sp;
+{
+	struct timeval tv;
+	/* {user, system, interrupt, total} {ticks, usec}; previous tu: */
+	u_int64_t ut, uu, st, su, it, iu, tt, tu, ptu;
 
+	ut = ruxp->rux_uticks;
+	st = ruxp->rux_sticks;
+	it = ruxp->rux_iticks;
 	tt = ut + st + it;
 	if (tt == 0) {
 		st = 1;
 		tt = 1;
 	}
-	rt = p->p_runtime;
-	problemcase = 0;
-	FOREACH_THREAD_IN_PROC(p, td) {
-		/*
-		 * Adjust for the current time slice.  This is actually fairly
-		 * important since the error here is on the order of a time
-		 * quantum, which is much greater than the sampling error.
-		 */
-		if (td == curthread) {
-			binuptime(&bt);
-			bintime_sub(&bt, PCPU_PTR(switchtime));
-			bintime_add(&rt, &bt);
-		} else if (TD_IS_RUNNING(td)) {
-			/*
-			 * XXX: this case should add the difference between
-			 * the current time and the switch time as above,
-			 * but the switch time is inaccessible, so we can't
-			 * do the adjustment and will end up with a wrong
-			 * runtime.  A previous call with a different
-			 * curthread may have obtained a (right or wrong)
-			 * runtime that is in advance of ours.  Just set a
-			 * flag to avoid warning about this known problem.
-			 */
-			problemcase = 1;
-		}
-	}
-	bintime2timeval(&rt, &tv);
+	bintime2timeval(&ruxp->rux_runtime, &tv);
 	tu = (u_int64_t)tv.tv_sec * 1000000 + tv.tv_usec;
-	ptu = p->p_uu + p->p_su + p->p_iu;
+	ptu = ruxp->rux_uu + ruxp->rux_su + ruxp->rux_iu;
 	if (tu < ptu) {
-		if (!problemcase)
-			printf(
+		printf(
 "calcru: runtime went backwards from %ju usec to %ju usec for pid %d (%s)\n",
-			    (uintmax_t)ptu, (uintmax_t)tu, p->p_pid, p->p_comm);
+		    (uintmax_t)ptu, (uintmax_t)tu, p->p_pid, p->p_comm);
 		tu = ptu;
 	}
 	if ((int64_t)tu < 0) {
@@ -764,38 +786,34 @@ calcru(p, up, sp, ip)
 	iu = tu - uu - su;
 
 	/* Enforce monotonicity. */
-	if (uu < p->p_uu || su < p->p_su || iu < p->p_iu) {
-		if (uu < p->p_uu)
-			uu = p->p_uu;
-		else if (uu + p->p_su + p->p_iu > tu)
-			uu = tu - p->p_su - p->p_iu;
+	if (uu < ruxp->rux_uu || su < ruxp->rux_su || iu < ruxp->rux_iu) {
+		if (uu < ruxp->rux_uu)
+			uu = ruxp->rux_uu;
+		else if (uu + ruxp->rux_su + ruxp->rux_iu > tu)
+			uu = tu - ruxp->rux_su - ruxp->rux_iu;
 		if (st == 0)
-			su = p->p_su;
+			su = ruxp->rux_su;
 		else {
 			su = ((tu - uu) * st) / (st + it);
-			if (su < p->p_su)
-				su = p->p_su;
-			else if (uu + su + p->p_iu > tu)
-				su = tu - uu - p->p_iu;
+			if (su < ruxp->rux_su)
+				su = ruxp->rux_su;
+			else if (uu + su + ruxp->rux_iu > tu)
+				su = tu - uu - ruxp->rux_iu;
 		}
-		KASSERT(uu + su + p->p_iu <= tu,
+		KASSERT(uu + su + ruxp->rux_iu <= tu,
 		    ("calcru: monotonisation botch 1"));
 		iu = tu - uu - su;
-		KASSERT(iu >= p->p_iu,
+		KASSERT(iu >= ruxp->rux_iu,
 		    ("calcru: monotonisation botch 2"));
 	}
-	p->p_uu = uu;
-	p->p_su = su;
-	p->p_iu = iu;
+	ruxp->rux_uu = uu;
+	ruxp->rux_su = su;
+	ruxp->rux_iu = iu;
 
 	up->tv_sec = uu / 1000000;
 	up->tv_usec = uu % 1000000;
 	sp->tv_sec = su / 1000000;
 	sp->tv_usec = su % 1000000;
-	if (ip != NULL) {
-		ip->tv_sec = iu / 1000000;
-		ip->tv_usec = iu % 1000000;
-	}
 }
 
 #ifndef _SYS_SYSPROTO_H_
@@ -807,51 +825,67 @@ struct getrusage_args {
 /*
  * MPSAFE
  */
-/* ARGSUSED */
 int
 getrusage(td, uap)
 	register struct thread *td;
 	register struct getrusage_args *uap;
 {
 	struct rusage ru;
+	int error;
+
+	error = kern_getrusage(td, uap->who, &ru);
+	if (error == 0)
+		error = copyout(&ru, uap->rusage, sizeof(struct rusage));
+	return (error);
+}
+
+int
+kern_getrusage(td, who, rup)
+	struct thread *td;
+	int who;
+	struct rusage *rup;
+{
 	struct proc *p;
 
 	p = td->td_proc;
-	switch (uap->who) {
+	PROC_LOCK(p);
+	switch (who) {
 
 	case RUSAGE_SELF:
-		mtx_lock(&Giant);
-		mtx_lock_spin(&sched_lock);
-		calcru(p, &p->p_stats->p_ru.ru_utime, &p->p_stats->p_ru.ru_stime,
-		    NULL);
-		mtx_unlock_spin(&sched_lock);
-		ru = p->p_stats->p_ru;
-		mtx_unlock(&Giant);
+		*rup = p->p_stats->p_ru;
+		calcru(p, &rup->ru_utime, &rup->ru_stime);
 		break;
 
 	case RUSAGE_CHILDREN:
-		mtx_lock(&Giant);
-		ru = p->p_stats->p_cru;
-		mtx_unlock(&Giant);
+		*rup = p->p_stats->p_cru;
+		calccru(p, &rup->ru_utime, &rup->ru_stime);
 		break;
 
 	default:
+		PROC_UNLOCK(p);
 		return (EINVAL);
-		break;
 	}
-	return (copyout(&ru, uap->rusage, sizeof(struct rusage)));
+	PROC_UNLOCK(p);
+	return (0);
 }
 
 void
-ruadd(ru, ru2)
+ruadd(ru, rux, ru2, rux2)
 	struct rusage *ru;
+	struct rusage_ext *rux;
 	struct rusage *ru2;
+	struct rusage_ext *rux2;
 {
 	register long *ip, *ip2;
 	register int i;
 
-	timevaladd(&ru->ru_utime, &ru2->ru_utime);
-	timevaladd(&ru->ru_stime, &ru2->ru_stime);
+	bintime_add(&rux->rux_runtime, &rux2->rux_runtime);
+	rux->rux_uticks += rux2->rux_uticks;
+	rux->rux_sticks += rux2->rux_sticks;
+	rux->rux_iticks += rux2->rux_iticks;
+	rux->rux_uu += rux2->rux_uu;
+	rux->rux_su += rux2->rux_su;
+	rux->rux_iu += rux2->rux_iu;
 	if (ru->ru_maxrss < ru2->ru_maxrss)
 		ru->ru_maxrss = ru2->ru_maxrss;
 	ip = &ru->ru_first;
diff --git a/sys/kern/kern_synch.c b/sys/kern/kern_synch.c
index 13749e460f58..52863e39b966 100644
--- a/sys/kern/kern_synch.c
+++ b/sys/kern/kern_synch.c
@@ -302,8 +302,8 @@ mi_switch(int flags, struct thread *newtd)
 	 * process was running, and add that to its total so far.
 	 */
 	binuptime(&new_switchtime);
-	bintime_add(&p->p_runtime, &new_switchtime);
-	bintime_sub(&p->p_runtime, PCPU_PTR(switchtime));
+	bintime_add(&p->p_rux.rux_runtime, &new_switchtime);
+	bintime_sub(&p->p_rux.rux_runtime, PCPU_PTR(switchtime));
 
 	td->td_generation++;	/* bump preempt-detect counter */
 
@@ -322,7 +322,7 @@ mi_switch(int flags, struct thread *newtd)
 	 * over max, arrange to kill the process in ast().
 	 */
 	if (p->p_cpulimit != RLIM_INFINITY &&
-	    p->p_runtime.sec > p->p_cpulimit) {
+	    p->p_rux.rux_runtime.sec > p->p_cpulimit) {
 		p->p_sflag |= PS_XCPU;
 		td->td_flags |= TDF_ASTPENDING;
 	}
diff --git a/sys/kern/kern_time.c b/sys/kern/kern_time.c
index 1a6819835a44..3b62532d9a1f 100644
--- a/sys/kern/kern_time.c
+++ b/sys/kern/kern_time.c
@@ -157,21 +157,23 @@ clock_gettime(struct thread *td, struct clock_gettime_args *uap)
 {
 	struct timespec ats;
 	struct timeval sys, user;
+	struct proc *p;
 
+	p = td->td_proc;
 	switch (uap->clock_id) {
 	case CLOCK_REALTIME:
 		nanotime(&ats);
 		break;
 	case CLOCK_VIRTUAL:
-		mtx_lock_spin(&sched_lock);
-		calcru(td->td_proc, &user, &sys, NULL);
-		mtx_unlock_spin(&sched_lock);
+		PROC_LOCK(p);
+		calcru(p, &user, &sys);
+		PROC_UNLOCK(p);
 		TIMEVAL_TO_TIMESPEC(&user, &ats);
 		break;
 	case CLOCK_PROF:
-		mtx_lock_spin(&sched_lock);
-		calcru(td->td_proc, &user, &sys, NULL);
-		mtx_unlock_spin(&sched_lock);
+		PROC_LOCK(p);
+		calcru(p, &user, &sys);
+		PROC_UNLOCK(p);
 		timevaladd(&user, &sys);
 		TIMEVAL_TO_TIMESPEC(&user, &ats);
 		break;
diff --git a/sys/kern/subr_trap.c b/sys/kern/subr_trap.c
index 7753823eb008..86026a41b23a 100644
--- a/sys/kern/subr_trap.c
+++ b/sys/kern/subr_trap.c
@@ -221,7 +221,7 @@ ast(struct trapframe *framep)
 		PROC_LOCK(p);
 		lim_rlimit(p, RLIMIT_CPU, &rlim);
 		mtx_lock_spin(&sched_lock);
-		if (p->p_runtime.sec >= rlim.rlim_max) {
+		if (p->p_rux.rux_runtime.sec >= rlim.rlim_max) {
 			mtx_unlock_spin(&sched_lock);
 			killproc(p, "exceeded maximum CPU limit");
 		} else {
diff --git a/sys/kern/tty.c b/sys/kern/tty.c
index bcb3f9e8e7e0..b1956598a29c 100644
--- a/sys/kern/tty.c
+++ b/sys/kern/tty.c
@@ -2550,9 +2550,9 @@ ttyinfo(struct tty *tp)
 		tp->t_rocount = 0;
 		return;
 	}
-	PGRP_LOCK(tp->t_pgrp);
-	if ((p = LIST_FIRST(&tp->t_pgrp->pg_members)) == 0) {
-		PGRP_UNLOCK(tp->t_pgrp);
+	sx_slock(&proctree_lock);
+	if ((p = LIST_FIRST(&tp->t_pgrp->pg_members)) == NULL) {
+		sx_sunlock(&proctree_lock);
 		ttyprintf(tp, "empty foreground process group\n");
 		tp->t_rocount = 0;
 		return;
@@ -2567,10 +2567,9 @@ ttyinfo(struct tty *tp)
 	 * too much.
 	 */
 	mtx_lock_spin(&sched_lock);
-	for (pick = NULL; p != 0; p = LIST_NEXT(p, p_pglist))
+	for (pick = NULL; p != NULL; p = LIST_NEXT(p, p_pglist))
 		if (proc_compare(pick, p))
 			pick = p;
-	PGRP_UNLOCK(tp->t_pgrp);
 
 	td = FIRST_THREAD_IN_PROC(pick);	/* XXXKSE */
 #if 0
@@ -2578,6 +2577,7 @@ ttyinfo(struct tty *tp)
 #else
 	if (td == NULL) {
 		mtx_unlock_spin(&sched_lock);
+		sx_sunlock(&proctree_lock);
 		ttyprintf(tp, "foreground process without thread\n");
 		tp->t_rocount = 0;
 		return;
@@ -2603,13 +2603,15 @@ ttyinfo(struct tty *tp)
 		state = "intrwait";
 	else
 		state = "unknown";
-	calcru(pick, &utime, &stime, NULL);
 	pctcpu = (sched_pctcpu(td) * 10000 + FSCALE / 2) >> FSHIFT;
 	if (pick->p_state == PRS_NEW || pick->p_state == PRS_ZOMBIE)
 		rss = 0;
 	else
 		rss = pgtok(vmspace_resident_count(pick->p_vmspace));
 	mtx_unlock_spin(&sched_lock);
+	PROC_LOCK(pick);
+	calcru(pick, &utime, &stime);
+	PROC_UNLOCK(pick);
 
 	/* Print command, pid, state, utime, stime, %cpu, and rss. */
 	ttyprintf(tp,
@@ -2619,6 +2621,7 @@ ttyinfo(struct tty *tp)
 	    (long)stime.tv_sec, stime.tv_usec / 10000,
 	    pctcpu / 100, rss);
 	tp->t_rocount = 0;
+	sx_sunlock(&proctree_lock);
 }
 
 /*
diff --git a/sys/sys/proc.h b/sys/sys/proc.h
index f37856fabbe4..9d75c927dd40 100644
--- a/sys/sys/proc.h
+++ b/sys/sys/proc.h
@@ -487,6 +487,26 @@ struct ksegrp {
 };
 
 /*
+ * XXX: Does this belong in resource.h or resourcevar.h instead?
+ * Resource usage extension.  The times in rusage structs in the kernel are
+ * never up to date.  The actual times are kept as runtimes and tick counts
+ * (with control info in the "previous" times), and are converted when
+ * userland asks for rusage info.  Backwards compatibility prevents putting
+ * this directly in the user-visible rusage struct.
+ *
+ * Locking: (cj) means (j) for p_rux and (c) for p_crux.
+ */
+struct rusage_ext {
+	struct bintime	rux_runtime;    /* (cj) Real time. */
+	u_int64_t	rux_uticks;     /* (cj) Statclock hits in user mode. */
+	u_int64_t	rux_sticks;     /* (cj) Statclock hits in sys mode. */
+	u_int64_t	rux_iticks;     /* (cj) Statclock hits in intr mode. */
+	u_int64_t	rux_uu;         /* (c) Previous user time in usec. */
+	u_int64_t	rux_su;         /* (c) Previous sys time in usec. */
+	u_int64_t	rux_iu;         /* (c) Previous intr time in usec. */
+};
+
+/*
  * The old fashionned process. May have multiple threads, KSEGRPs
  * and KSEs. Starts off with a single embedded KSEGRP and THREAD.
  */
@@ -530,13 +550,8 @@ struct proc {
 	struct vmspace	*p_vmspace;	/* (b) Address space. */
 	u_int		p_swtime;	/* (j) Time swapped in or out. */
 	struct itimerval p_realtimer;	/* (c) Alarm timer. */
-	struct bintime	p_runtime;	/* (j) Real time. */
-	u_int64_t	p_uu;		/* (j) Previous user time in usec. */
-	u_int64_t	p_su;		/* (j) Previous system time in usec. */
-	u_int64_t	p_iu;		/* (j) Previous intr time in usec. */
-	u_int64_t	p_uticks;	/* (j) Statclock hits in user mode. */
-	u_int64_t	p_sticks;	/* (j) Statclock hits in system mode. */
-	u_int64_t	p_iticks;	/* (j) Statclock hits in intr. */
+	struct rusage_ext p_rux;	/* (cj) Internal resource usage. */
+	struct rusage_ext p_crux;	/* (c) Internal child resource usage. */
 	int		p_profthreads;	/* (c) Num threads in addupc_task. */
 	int		p_maxthrwaits;	/* (c) Max threads num waiters */
 	int		p_traceflag;	/* (o) Kernel trace points. */
diff --git a/sys/sys/resourcevar.h b/sys/sys/resourcevar.h
index 9fb4aa4b97b1..3a03b0c4b1ad 100644
--- a/sys/sys/resourcevar.h
+++ b/sys/sys/resourcevar.h
@@ -106,12 +106,13 @@ struct uidinfo {
 #define	UIDINFO_UNLOCK(ui)	mtx_unlock((ui)->ui_mtxp)
 
 struct proc;
+struct rusage_ext;
 struct thread;
 
 void	 addupc_intr(struct thread *td, uintptr_t pc, u_int ticks);
 void	 addupc_task(struct thread *td, uintptr_t pc, u_int ticks);
-void	 calcru(struct proc *p, struct timeval *up, struct timeval *sp,
-	    struct timeval *ip);
+void	 calccru(struct proc *p, struct timeval *up, struct timeval *sp);
+void	 calcru(struct proc *p, struct timeval *up, struct timeval *sp);
 int	 chgproccnt(struct uidinfo *uip, int diff, int maxval);
 int	 chgsbsize(struct uidinfo *uip, u_int *hiwat, u_int to,
 	    rlim_t maxval);
@@ -125,7 +126,8 @@ struct plimit
 	*lim_hold(struct plimit *limp);
 rlim_t	 lim_max(struct proc *p, int which);
 void	 lim_rlimit(struct proc *p, int which, struct rlimit *rlp);
-void	 ruadd(struct rusage *ru, struct rusage *ru2);
+void	 ruadd(struct rusage *ru, struct rusage_ext *rux, struct rusage *ru2,
+	    struct rusage_ext *rux2);
 int	 suswintr(void *base, int word);
 struct uidinfo
 	*uifind(uid_t uid);
diff --git a/sys/sys/syscallsubr.h b/sys/sys/syscallsubr.h
index aa697e1975ff..02fc20a5c89b 100644
--- a/sys/sys/syscallsubr.h
+++ b/sys/sys/syscallsubr.h
@@ -34,6 +34,7 @@
 
 struct mbuf;
 struct msghdr;
+struct rusage;
 struct sockaddr;
 
 int	kern___getcwd(struct thread *td, u_char *buf, enum uio_seg bufseg,
@@ -50,6 +51,7 @@ int	kern_connect(struct thread *td, int fd, struct sockaddr *sa);
 int	kern_fcntl(struct thread *td, int fd, int cmd, intptr_t arg);
 int	kern_futimes(struct thread *td, int fd, struct timeval *tptr,
 	    enum uio_seg tptrseg);
+int	kern_getrusage(struct thread *td, int who, struct rusage *rup);
 int	kern_getsockopt(struct thread *td, int s, int level, int name,
 	    void *optval, enum uio_seg valseg, socklen_t *valsize);
 int	kern_lchown(struct thread *td, char *path, enum uio_seg pathseg,