freebsd-nq/sys/kern/kern_condvar.c
John Baldwin c86b6ff551 Change the preemption code for software interrupt thread schedules and
mutex releases to not require flags for the cases when preemption is
not allowed:

The purpose of the MTX_NOSWITCH and SWI_NOSWITCH flags is to prevent
switching to a higher priority thread on mutex releease and swi schedule,
respectively when that switch is not safe.  Now that the critical section
API maintains a per-thread nesting count, the kernel can easily check
whether or not it should switch without relying on flags from the
programmer.  This fixes a few bugs in that all current callers of
swi_sched() used SWI_NOSWITCH, when in fact, only the ones called from
fast interrupt handlers and the swi_sched of softclock needed this flag.
Note that to ensure that swi_sched()'s in clock and fast interrupt
handlers do not switch, these handlers have to be explicitly wrapped
in critical_enter/exit pairs.  Presently, just wrapping the handlers is
sufficient, but in the future with the fully preemptive kernel, the
interrupt must be EOI'd before critical_exit() is called.  (critical_exit()
can switch due to a deferred preemption in a fully preemptive kernel.)

I've tested the changes to the interrupt code on i386 and alpha.  I have
not tested ia64, but the interrupt code is almost identical to the alpha
code, so I expect it will work fine.  PowerPC and ARM do not yet have
interrupt code in the tree so they shouldn't be broken.  Sparc64 is
broken, but that's been ok'd by jake and tmm who will be fixing the
interrupt code for sparc64 shortly.

Reviewed by:	peter
Tested on:	i386, alpha
2002-01-05 08:47:13 +00:00

579 lines
14 KiB
C

/*-
* Copyright (c) 2000 Jake Burkholder <jake@freebsd.org>.
* All rights reserved.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
* 1. Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer.
* 2. Redistributions in binary form must reproduce the above copyright
* notice, this list of conditions and the following disclaimer in the
* documentation and/or other materials provided with the distribution.
*
* THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
* ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
* ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
* FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
* DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
* OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
* HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
* LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
* OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
* SUCH DAMAGE.
*
* $FreeBSD$
*/
#include "opt_ktrace.h"
#include <sys/param.h>
#include <sys/systm.h>
#include <sys/lock.h>
#include <sys/mutex.h>
#include <sys/proc.h>
#include <sys/kernel.h>
#include <sys/ktr.h>
#include <sys/condvar.h>
#include <sys/signalvar.h>
#include <sys/resourcevar.h>
#ifdef KTRACE
#include <sys/uio.h>
#include <sys/ktrace.h>
#endif
/*
* Common sanity checks for cv_wait* functions.
*/
#define CV_ASSERT(cvp, mp, td) do { \
KASSERT((td) != NULL, ("%s: curthread NULL", __func__)); \
KASSERT((td)->td_proc->p_stat == SRUN, ("%s: not SRUN", __func__)); \
KASSERT((cvp) != NULL, ("%s: cvp NULL", __func__)); \
KASSERT((mp) != NULL, ("%s: mp NULL", __func__)); \
mtx_assert((mp), MA_OWNED | MA_NOTRECURSED); \
} while (0)
#ifdef CV_DEBUG
#define CV_WAIT_VALIDATE(cvp, mp) do { \
if (TAILQ_EMPTY(&(cvp)->cv_waitq)) { \
/* Only waiter. */ \
(cvp)->cv_mtx = (mp); \
} else { \
/* \
* Other waiter; assert that we're using the \
* same mutex. \
*/ \
KASSERT((cvp)->cv_mtx == (mp), \
("%s: Multiple mutexes", __func__)); \
} \
} while (0)
#define CV_SIGNAL_VALIDATE(cvp) do { \
if (!TAILQ_EMPTY(&(cvp)->cv_waitq)) { \
KASSERT(mtx_owned((cvp)->cv_mtx), \
("%s: Mutex not owned", __func__)); \
} \
} while (0)
#else
#define CV_WAIT_VALIDATE(cvp, mp)
#define CV_SIGNAL_VALIDATE(cvp)
#endif
static void cv_timedwait_end(void *arg);
/*
* Initialize a condition variable. Must be called before use.
*/
void
cv_init(struct cv *cvp, const char *desc)
{
TAILQ_INIT(&cvp->cv_waitq);
cvp->cv_mtx = NULL;
cvp->cv_description = desc;
}
/*
* Destroy a condition variable. The condition variable must be re-initialized
* in order to be re-used.
*/
void
cv_destroy(struct cv *cvp)
{
KASSERT(cv_waitq_empty(cvp), ("%s: cv_waitq non-empty", __func__));
}
/*
* Common code for cv_wait* functions. All require sched_lock.
*/
/*
* Switch context.
*/
static __inline void
cv_switch(struct thread *td)
{
td->td_proc->p_stat = SSLEEP;
td->td_proc->p_stats->p_ru.ru_nvcsw++;
mi_switch();
CTR3(KTR_PROC, "cv_switch: resume thread %p (pid %d, %s)", td,
td->td_proc->p_pid, td->td_proc->p_comm);
}
/*
* Switch context, catching signals.
*/
static __inline int
cv_switch_catch(struct thread *td)
{
struct proc *p;
int sig;
/*
* We put ourselves on the sleep queue and start our timeout before
* calling CURSIG, as we could stop there, and a wakeup or a SIGCONT (or
* both) could occur while we were stopped. A SIGCONT would cause us to
* be marked as SSLEEP without resuming us, thus we must be ready for
* sleep when CURSIG is called. If the wakeup happens while we're
* stopped, td->td_wchan will be 0 upon return from CURSIG.
*/
td->td_flags |= TDF_SINTR;
mtx_unlock_spin(&sched_lock);
p = td->td_proc;
PROC_LOCK(p);
sig = CURSIG(p); /* XXXKSE */
mtx_lock_spin(&sched_lock);
PROC_UNLOCK(p);
if (sig != 0) {
if (td->td_wchan != NULL)
cv_waitq_remove(td);
td->td_proc->p_stat = SRUN;
} else if (td->td_wchan != NULL) {
cv_switch(td);
}
td->td_flags &= ~TDF_SINTR;
return sig;
}
/*
* Add a thread to the wait queue of a condition variable.
*/
static __inline void
cv_waitq_add(struct cv *cvp, struct thread *td)
{
/*
* Process may be sitting on a slpque if asleep() was called, remove it
* before re-adding.
*/
if (td->td_wchan != NULL)
unsleep(td);
td->td_flags |= TDF_CVWAITQ;
td->td_wchan = cvp;
td->td_wmesg = cvp->cv_description;
td->td_kse->ke_slptime = 0; /* XXXKSE */
td->td_ksegrp->kg_slptime = 0; /* XXXKSE */
td->td_ksegrp->kg_pri.pri_native = td->td_ksegrp->kg_pri.pri_level;
CTR3(KTR_PROC, "cv_waitq_add: thread %p (pid %d, %s)", td,
td->td_proc->p_pid, td->td_proc->p_comm);
TAILQ_INSERT_TAIL(&cvp->cv_waitq, td, td_slpq);
}
/*
* Wait on a condition variable. The current thread is placed on the condition
* variable's wait queue and suspended. A cv_signal or cv_broadcast on the same
* condition variable will resume the thread. The mutex is released before
* sleeping and will be held on return. It is recommended that the mutex be
* held when cv_signal or cv_broadcast are called.
*/
void
cv_wait(struct cv *cvp, struct mtx *mp)
{
struct thread *td;
WITNESS_SAVE_DECL(mp);
td = curthread;
#ifdef KTRACE
if (td->td_proc && KTRPOINT(td->td_proc, KTR_CSW))
ktrcsw(td->td_proc->p_tracep, 1, 0);
#endif
CV_ASSERT(cvp, mp, td);
WITNESS_SLEEP(0, &mp->mtx_object);
WITNESS_SAVE(&mp->mtx_object, mp);
mtx_lock_spin(&sched_lock);
if (cold || panicstr) {
/*
* After a panic, or during autoconfiguration, just give
* interrupts a chance, then just return; don't run any other
* thread or panic below, in case this is the idle process and
* already asleep.
*/
mtx_unlock_spin(&sched_lock);
return;
}
CV_WAIT_VALIDATE(cvp, mp);
DROP_GIANT();
mtx_unlock(mp);
cv_waitq_add(cvp, td);
cv_switch(td);
mtx_unlock_spin(&sched_lock);
#ifdef KTRACE
if (KTRPOINT(td->td_proc, KTR_CSW))
ktrcsw(td->td_proc->p_tracep, 0, 0);
#endif
PICKUP_GIANT();
mtx_lock(mp);
WITNESS_RESTORE(&mp->mtx_object, mp);
}
/*
* Wait on a condition variable, allowing interruption by signals. Return 0 if
* the thread was resumed with cv_signal or cv_broadcast, EINTR or ERESTART if
* a signal was caught. If ERESTART is returned the system call should be
* restarted if possible.
*/
int
cv_wait_sig(struct cv *cvp, struct mtx *mp)
{
struct thread *td;
struct proc *p;
int rval;
int sig;
WITNESS_SAVE_DECL(mp);
td = curthread;
p = td->td_proc;
rval = 0;
#ifdef KTRACE
if (td->td_proc && KTRPOINT(td->td_proc, KTR_CSW))
ktrcsw(td->td_proc->p_tracep, 1, 0);
#endif
CV_ASSERT(cvp, mp, td);
WITNESS_SLEEP(0, &mp->mtx_object);
WITNESS_SAVE(&mp->mtx_object, mp);
mtx_lock_spin(&sched_lock);
if (cold || panicstr) {
/*
* After a panic, or during autoconfiguration, just give
* interrupts a chance, then just return; don't run any other
* procs or panic below, in case this is the idle process and
* already asleep.
*/
mtx_unlock_spin(&sched_lock);
return 0;
}
CV_WAIT_VALIDATE(cvp, mp);
DROP_GIANT();
mtx_unlock(mp);
cv_waitq_add(cvp, td);
sig = cv_switch_catch(td);
mtx_unlock_spin(&sched_lock);
PICKUP_GIANT();
PROC_LOCK(p);
if (sig == 0)
sig = CURSIG(p); /* XXXKSE */
if (sig != 0) {
if (SIGISMEMBER(p->p_sigacts->ps_sigintr, sig))
rval = EINTR;
else
rval = ERESTART;
}
PROC_UNLOCK(p);
#ifdef KTRACE
mtx_lock(&Giant);
if (KTRPOINT(td->td_proc, KTR_CSW))
ktrcsw(td->td_proc->p_tracep, 0, 0);
mtx_unlock(&Giant);
#endif
mtx_lock(mp);
WITNESS_RESTORE(&mp->mtx_object, mp);
return (rval);
}
/*
* Wait on a condition variable for at most timo/hz seconds. Returns 0 if the
* process was resumed by cv_signal or cv_broadcast, EWOULDBLOCK if the timeout
* expires.
*/
int
cv_timedwait(struct cv *cvp, struct mtx *mp, int timo)
{
struct thread *td;
int rval;
WITNESS_SAVE_DECL(mp);
td = curthread;
rval = 0;
#ifdef KTRACE
ktrcsw(td->td_proc->p_tracep, 1, 0);
#endif
CV_ASSERT(cvp, mp, td);
WITNESS_SLEEP(0, &mp->mtx_object);
WITNESS_SAVE(&mp->mtx_object, mp);
mtx_lock_spin(&sched_lock);
if (cold || panicstr) {
/*
* After a panic, or during autoconfiguration, just give
* interrupts a chance, then just return; don't run any other
* thread or panic below, in case this is the idle process and
* already asleep.
*/
mtx_unlock_spin(&sched_lock);
return 0;
}
CV_WAIT_VALIDATE(cvp, mp);
DROP_GIANT();
mtx_unlock(mp);
cv_waitq_add(cvp, td);
callout_reset(&td->td_slpcallout, timo, cv_timedwait_end, td);
cv_switch(td);
if (td->td_flags & TDF_TIMEOUT) {
td->td_flags &= ~TDF_TIMEOUT;
rval = EWOULDBLOCK;
} else if (td->td_flags & TDF_TIMOFAIL)
td->td_flags &= ~TDF_TIMOFAIL;
else if (callout_stop(&td->td_slpcallout) == 0) {
/*
* Work around race with cv_timedwait_end similar to that
* between msleep and endtsleep.
*/
td->td_flags |= TDF_TIMEOUT;
td->td_proc->p_stats->p_ru.ru_nivcsw++;
mi_switch();
}
mtx_unlock_spin(&sched_lock);
#ifdef KTRACE
if (KTRPOINT(td->td_proc, KTR_CSW))
ktrcsw(td->td_proc->p_tracep, 0, 0);
#endif
PICKUP_GIANT();
mtx_lock(mp);
WITNESS_RESTORE(&mp->mtx_object, mp);
return (rval);
}
/*
* Wait on a condition variable for at most timo/hz seconds, allowing
* interruption by signals. Returns 0 if the thread was resumed by cv_signal
* or cv_broadcast, EWOULDBLOCK if the timeout expires, and EINTR or ERESTART if
* a signal was caught.
*/
int
cv_timedwait_sig(struct cv *cvp, struct mtx *mp, int timo)
{
struct thread *td;
struct proc *p;
int rval;
int sig;
WITNESS_SAVE_DECL(mp);
td = curthread;
p = td->td_proc;
rval = 0;
#ifdef KTRACE
if (td->td_proc && KTRPOINT(td->td_proc, KTR_CSW))
ktrcsw(td->td_proc->p_tracep, 1, 0);
#endif
CV_ASSERT(cvp, mp, td);
WITNESS_SLEEP(0, &mp->mtx_object);
WITNESS_SAVE(&mp->mtx_object, mp);
mtx_lock_spin(&sched_lock);
if (cold || panicstr) {
/*
* After a panic, or during autoconfiguration, just give
* interrupts a chance, then just return; don't run any other
* thread or panic below, in case this is the idle process and
* already asleep.
*/
mtx_unlock_spin(&sched_lock);
return 0;
}
CV_WAIT_VALIDATE(cvp, mp);
DROP_GIANT();
mtx_unlock(mp);
cv_waitq_add(cvp, td);
callout_reset(&td->td_slpcallout, timo, cv_timedwait_end, td);
sig = cv_switch_catch(td);
if (td->td_flags & TDF_TIMEOUT) {
td->td_flags &= ~TDF_TIMEOUT;
rval = EWOULDBLOCK;
} else if (td->td_flags & TDF_TIMOFAIL)
td->td_flags &= ~TDF_TIMOFAIL;
else if (callout_stop(&td->td_slpcallout) == 0) {
/*
* Work around race with cv_timedwait_end similar to that
* between msleep and endtsleep.
*/
td->td_flags |= TDF_TIMEOUT;
td->td_proc->p_stats->p_ru.ru_nivcsw++;
mi_switch();
}
mtx_unlock_spin(&sched_lock);
PICKUP_GIANT();
PROC_LOCK(p);
if (sig == 0)
sig = CURSIG(p);
if (sig != 0) {
if (SIGISMEMBER(p->p_sigacts->ps_sigintr, sig))
rval = EINTR;
else
rval = ERESTART;
}
PROC_UNLOCK(p);
#ifdef KTRACE
mtx_lock(&Giant);
if (KTRPOINT(td->td_proc, KTR_CSW))
ktrcsw(td->td_proc->p_tracep, 0, 0);
mtx_unlock(&Giant);
#endif
mtx_lock(mp);
WITNESS_RESTORE(&mp->mtx_object, mp);
return (rval);
}
/*
* Common code for signal and broadcast. Assumes waitq is not empty. Must be
* called with sched_lock held.
*/
static __inline void
cv_wakeup(struct cv *cvp)
{
struct thread *td;
mtx_assert(&sched_lock, MA_OWNED);
td = TAILQ_FIRST(&cvp->cv_waitq);
KASSERT(td->td_wchan == cvp, ("%s: bogus wchan", __func__));
KASSERT(td->td_flags & TDF_CVWAITQ, ("%s: not on waitq", __func__));
TAILQ_REMOVE(&cvp->cv_waitq, td, td_slpq);
td->td_flags &= ~TDF_CVWAITQ;
td->td_wchan = 0;
if (td->td_proc->p_stat == SSLEEP) {
/* OPTIMIZED EXPANSION OF setrunnable(td); */
CTR3(KTR_PROC, "cv_signal: thread %p (pid %d, %s)",
td, td->td_proc->p_pid, td->td_proc->p_comm);
if (td->td_ksegrp->kg_slptime > 1) /* XXXKSE */
updatepri(td);
td->td_kse->ke_slptime = 0;
td->td_ksegrp->kg_slptime = 0;
td->td_proc->p_stat = SRUN;
if (td->td_proc->p_sflag & PS_INMEM) {
setrunqueue(td);
maybe_resched(td->td_ksegrp);
} else {
td->td_proc->p_sflag |= PS_SWAPINREQ;
wakeup(&proc0); /* XXXKSE */
}
/* END INLINE EXPANSION */
}
}
/*
* Signal a condition variable, wakes up one waiting thread. Will also wakeup
* the swapper if the process is not in memory, so that it can bring the
* sleeping process in. Note that this may also result in additional threads
* being made runnable. Should be called with the same mutex as was passed to
* cv_wait held.
*/
void
cv_signal(struct cv *cvp)
{
KASSERT(cvp != NULL, ("%s: cvp NULL", __func__));
mtx_lock_spin(&sched_lock);
if (!TAILQ_EMPTY(&cvp->cv_waitq)) {
CV_SIGNAL_VALIDATE(cvp);
cv_wakeup(cvp);
}
mtx_unlock_spin(&sched_lock);
}
/*
* Broadcast a signal to a condition variable. Wakes up all waiting threads.
* Should be called with the same mutex as was passed to cv_wait held.
*/
void
cv_broadcast(struct cv *cvp)
{
KASSERT(cvp != NULL, ("%s: cvp NULL", __func__));
mtx_lock_spin(&sched_lock);
CV_SIGNAL_VALIDATE(cvp);
while (!TAILQ_EMPTY(&cvp->cv_waitq))
cv_wakeup(cvp);
mtx_unlock_spin(&sched_lock);
}
/*
* Remove a thread from the wait queue of its condition variable. This may be
* called externally.
*/
void
cv_waitq_remove(struct thread *td)
{
struct cv *cvp;
mtx_lock_spin(&sched_lock);
if ((cvp = td->td_wchan) != NULL && td->td_flags & TDF_CVWAITQ) {
TAILQ_REMOVE(&cvp->cv_waitq, td, td_slpq);
td->td_flags &= ~TDF_CVWAITQ;
td->td_wchan = NULL;
}
mtx_unlock_spin(&sched_lock);
}
/*
* Timeout function for cv_timedwait. Put the thread on the runqueue and set
* its timeout flag.
*/
static void
cv_timedwait_end(void *arg)
{
struct thread *td;
td = arg;
CTR3(KTR_PROC, "cv_timedwait_end: thread %p (pid %d, %s)", td, td->td_proc->p_pid,
td->td_proc->p_comm);
mtx_lock_spin(&sched_lock);
if (td->td_flags & TDF_TIMEOUT) {
td->td_flags &= ~TDF_TIMEOUT;
setrunqueue(td);
} else if (td->td_wchan != NULL) {
if (td->td_proc->p_stat == SSLEEP) /* XXXKSE */
setrunnable(td);
else
cv_waitq_remove(td);
td->td_flags |= TDF_TIMEOUT;
} else
td->td_flags |= TDF_TIMOFAIL;
mtx_unlock_spin(&sched_lock);
}