freebsd-dev/sys/kern/kern_intr.c

809 lines
20 KiB
C
Raw Normal View History

/*-
* Copyright (c) 1997, Stefan Esser <se@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 unmodified, 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 ``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 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.
*/
2003-06-11 00:56:59 +00:00
#include <sys/cdefs.h>
__FBSDID("$FreeBSD$");
#include "opt_ddb.h"
#include <sys/param.h>
#include <sys/bus.h>
#include <sys/conf.h>
#include <sys/rtprio.h>
#include <sys/systm.h>
#include <sys/interrupt.h>
#include <sys/kernel.h>
#include <sys/kthread.h>
#include <sys/ktr.h>
#include <sys/limits.h>
#include <sys/lock.h>
#include <sys/malloc.h>
#include <sys/mutex.h>
#include <sys/proc.h>
#include <sys/random.h>
#include <sys/resourcevar.h>
#include <sys/sched.h>
#include <sys/sysctl.h>
#include <sys/unistd.h>
#include <sys/vmmeter.h>
#include <machine/atomic.h>
#include <machine/cpu.h>
#include <machine/md_var.h>
#include <machine/stdarg.h>
#ifdef DDB
#include <ddb/ddb.h>
#include <ddb/db_sym.h>
#endif
struct int_entropy {
struct proc *proc;
uintptr_t vector;
};
struct ithd *clk_ithd;
struct ithd *tty_ithd;
void *softclock_ih;
void *vm_ih;
static MALLOC_DEFINE(M_ITHREAD, "ithread", "Interrupt Threads");
- Enable (unmask) interrupt sources earlier in the ithread loop. Specifically, we used to enable the source after locking sched_lock and just before we had already decided to do a context switch. This meant that an ithread could never process more than one interrupt per context switch. Enabling earlier in the loop before sched_lock is acquired allows an ithread to handle multiple interrupts per context switch if interrupts fire very rapidly. For the case of heavy interrupt load this can reduce the number of context switches (and thus overhead) as well as reduce interrupt latency. - Now that we can handle multiple interrupts per context switch, add simple interrupt storm protection to threaded interrupts. If X number of consecutive interrupts are triggered before the itherad voluntarily yields to another thread, then the interrupt thread will sleep with the associated interrupt source disabled (masked) for 1/10th of a second. The default value of X is 500, but it can be tweaked via the tunable/ sysctl hw.intr_storm_threshold. If an interrupt storm is detected, then a message is output to the kernel console on the first occurrence per interrupt thread. Interrupt storm protection can be disabled completely by setting this value to 0. There is no scientific reasoning for the 1/10th of a second or 500 interrupts values, so they may require tweaking at some point in the future. Tested by: rwatson (an earlier version w/o the storm protection) Tested by: mux (reportedly made a machine with two PCI interrupts storming usable rather than hard locked) Reviewed by: imp
2004-04-16 20:25:40 +00:00
static int intr_storm_threshold = 500;
TUNABLE_INT("hw.intr_storm_threshold", &intr_storm_threshold);
SYSCTL_INT(_hw, OID_AUTO, intr_storm_threshold, CTLFLAG_RW,
&intr_storm_threshold, 0,
"Number of consecutive interrupts before storm protection is enabled");
static void ithread_loop(void *);
static void ithread_update(struct ithd *);
static void start_softintr(void *);
- Enable (unmask) interrupt sources earlier in the ithread loop. Specifically, we used to enable the source after locking sched_lock and just before we had already decided to do a context switch. This meant that an ithread could never process more than one interrupt per context switch. Enabling earlier in the loop before sched_lock is acquired allows an ithread to handle multiple interrupts per context switch if interrupts fire very rapidly. For the case of heavy interrupt load this can reduce the number of context switches (and thus overhead) as well as reduce interrupt latency. - Now that we can handle multiple interrupts per context switch, add simple interrupt storm protection to threaded interrupts. If X number of consecutive interrupts are triggered before the itherad voluntarily yields to another thread, then the interrupt thread will sleep with the associated interrupt source disabled (masked) for 1/10th of a second. The default value of X is 500, but it can be tweaked via the tunable/ sysctl hw.intr_storm_threshold. If an interrupt storm is detected, then a message is output to the kernel console on the first occurrence per interrupt thread. Interrupt storm protection can be disabled completely by setting this value to 0. There is no scientific reasoning for the 1/10th of a second or 500 interrupts values, so they may require tweaking at some point in the future. Tested by: rwatson (an earlier version w/o the storm protection) Tested by: mux (reportedly made a machine with two PCI interrupts storming usable rather than hard locked) Reviewed by: imp
2004-04-16 20:25:40 +00:00
u_char
ithread_priority(enum intr_type flags)
{
u_char pri;
flags &= (INTR_TYPE_TTY | INTR_TYPE_BIO | INTR_TYPE_NET |
INTR_TYPE_CAM | INTR_TYPE_MISC | INTR_TYPE_CLK | INTR_TYPE_AV);
switch (flags) {
case INTR_TYPE_TTY:
pri = PI_TTYLOW;
break;
case INTR_TYPE_BIO:
/*
* XXX We need to refine this. BSD/OS distinguishes
* between tape and disk priorities.
*/
pri = PI_DISK;
break;
case INTR_TYPE_NET:
pri = PI_NET;
break;
case INTR_TYPE_CAM:
pri = PI_DISK; /* XXX or PI_CAM? */
break;
case INTR_TYPE_AV: /* Audio/video */
pri = PI_AV;
break;
case INTR_TYPE_CLK:
pri = PI_REALTIME;
break;
case INTR_TYPE_MISC:
pri = PI_DULL; /* don't care */
break;
default:
/* We didn't specify an interrupt level. */
panic("ithread_priority: no interrupt type in flags");
}
return pri;
}
/*
* Regenerate the name (p_comm) and priority for a threaded interrupt thread.
*/
static void
ithread_update(struct ithd *ithd)
{
struct intrhand *ih;
struct thread *td;
struct proc *p;
int missed;
mtx_assert(&ithd->it_lock, MA_OWNED);
td = ithd->it_td;
if (td == NULL)
return;
p = td->td_proc;
strlcpy(p->p_comm, ithd->it_name, sizeof(p->p_comm));
ithd->it_flags &= ~IT_ENTROPY;
ih = TAILQ_FIRST(&ithd->it_handlers);
if (ih == NULL) {
mtx_lock_spin(&sched_lock);
sched_prio(td, PRI_MAX_ITHD);
mtx_unlock_spin(&sched_lock);
return;
}
mtx_lock_spin(&sched_lock);
sched_prio(td, ih->ih_pri);
mtx_unlock_spin(&sched_lock);
missed = 0;
TAILQ_FOREACH(ih, &ithd->it_handlers, ih_next) {
if (strlen(p->p_comm) + strlen(ih->ih_name) + 1 <
sizeof(p->p_comm)) {
strcat(p->p_comm, " ");
strcat(p->p_comm, ih->ih_name);
} else
missed++;
if (ih->ih_flags & IH_ENTROPY)
ithd->it_flags |= IT_ENTROPY;
}
while (missed-- > 0) {
if (strlen(p->p_comm) + 1 == sizeof(p->p_comm)) {
if (p->p_comm[sizeof(p->p_comm) - 2] == '+')
p->p_comm[sizeof(p->p_comm) - 2] = '*';
else
p->p_comm[sizeof(p->p_comm) - 2] = '+';
} else
strcat(p->p_comm, "+");
}
2002-12-28 23:22:22 +00:00
CTR2(KTR_INTR, "%s: updated %s", __func__, p->p_comm);
}
int
ithread_create(struct ithd **ithread, uintptr_t vector, int flags,
void (*disable)(uintptr_t), void (*enable)(uintptr_t), const char *fmt, ...)
{
struct ithd *ithd;
struct thread *td;
struct proc *p;
int error;
va_list ap;
/* The only valid flag during creation is IT_SOFT. */
if ((flags & ~IT_SOFT) != 0)
return (EINVAL);
ithd = malloc(sizeof(struct ithd), M_ITHREAD, M_WAITOK | M_ZERO);
ithd->it_vector = vector;
ithd->it_disable = disable;
ithd->it_enable = enable;
ithd->it_flags = flags;
TAILQ_INIT(&ithd->it_handlers);
mtx_init(&ithd->it_lock, "ithread", NULL, MTX_DEF);
va_start(ap, fmt);
vsnprintf(ithd->it_name, sizeof(ithd->it_name), fmt, ap);
va_end(ap);
error = kthread_create(ithread_loop, ithd, &p, RFSTOPPED | RFHIGHPID,
0, "%s", ithd->it_name);
if (error) {
mtx_destroy(&ithd->it_lock);
free(ithd, M_ITHREAD);
return (error);
}
td = FIRST_THREAD_IN_PROC(p); /* XXXKSE */
mtx_lock_spin(&sched_lock);
td->td_ksegrp->kg_pri_class = PRI_ITHD;
td->td_priority = PRI_MAX_ITHD;
TD_SET_IWAIT(td);
mtx_unlock_spin(&sched_lock);
ithd->it_td = td;
td->td_ithd = ithd;
if (ithread != NULL)
*ithread = ithd;
CTR2(KTR_INTR, "%s: created %s", __func__, ithd->it_name);
return (0);
}
int
ithread_destroy(struct ithd *ithread)
{
struct thread *td;
if (ithread == NULL)
return (EINVAL);
td = ithread->it_td;
mtx_lock(&ithread->it_lock);
if (!TAILQ_EMPTY(&ithread->it_handlers)) {
mtx_unlock(&ithread->it_lock);
return (EINVAL);
}
ithread->it_flags |= IT_DEAD;
mtx_lock_spin(&sched_lock);
if (TD_AWAITING_INTR(td)) {
TD_CLR_IWAIT(td);
setrunqueue(td, SRQ_INTR);
}
mtx_unlock_spin(&sched_lock);
mtx_unlock(&ithread->it_lock);
CTR2(KTR_INTR, "%s: killing %s", __func__, ithread->it_name);
return (0);
}
int
ithread_add_handler(struct ithd* ithread, const char *name,
driver_intr_t handler, void *arg, u_char pri, enum intr_type flags,
void **cookiep)
{
struct intrhand *ih, *temp_ih;
if (ithread == NULL || name == NULL || handler == NULL)
return (EINVAL);
ih = malloc(sizeof(struct intrhand), M_ITHREAD, M_WAITOK | M_ZERO);
ih->ih_handler = handler;
ih->ih_argument = arg;
ih->ih_name = name;
ih->ih_ithread = ithread;
ih->ih_pri = pri;
if (flags & INTR_FAST)
ih->ih_flags = IH_FAST;
else if (flags & INTR_EXCL)
ih->ih_flags = IH_EXCLUSIVE;
if (flags & INTR_MPSAFE)
ih->ih_flags |= IH_MPSAFE;
if (flags & INTR_ENTROPY)
ih->ih_flags |= IH_ENTROPY;
mtx_lock(&ithread->it_lock);
if ((flags & INTR_EXCL) != 0 && !TAILQ_EMPTY(&ithread->it_handlers))
goto fail;
if (!TAILQ_EMPTY(&ithread->it_handlers)) {
temp_ih = TAILQ_FIRST(&ithread->it_handlers);
if (temp_ih->ih_flags & IH_EXCLUSIVE)
goto fail;
if ((ih->ih_flags & IH_FAST) && !(temp_ih->ih_flags & IH_FAST))
goto fail;
if (!(ih->ih_flags & IH_FAST) && (temp_ih->ih_flags & IH_FAST))
goto fail;
}
TAILQ_FOREACH(temp_ih, &ithread->it_handlers, ih_next)
if (temp_ih->ih_pri > ih->ih_pri)
break;
if (temp_ih == NULL)
TAILQ_INSERT_TAIL(&ithread->it_handlers, ih, ih_next);
else
TAILQ_INSERT_BEFORE(temp_ih, ih, ih_next);
ithread_update(ithread);
mtx_unlock(&ithread->it_lock);
if (cookiep != NULL)
*cookiep = ih;
CTR3(KTR_INTR, "%s: added %s to %s", __func__, ih->ih_name,
ithread->it_name);
return (0);
fail:
mtx_unlock(&ithread->it_lock);
free(ih, M_ITHREAD);
return (EINVAL);
}
int
ithread_remove_handler(void *cookie)
{
struct intrhand *handler = (struct intrhand *)cookie;
struct ithd *ithread;
#ifdef INVARIANTS
struct intrhand *ih;
#endif
if (handler == NULL)
return (EINVAL);
ithread = handler->ih_ithread;
KASSERT(ithread != NULL,
("interrupt handler \"%s\" has a NULL interrupt thread",
handler->ih_name));
CTR3(KTR_INTR, "%s: removing %s from %s", __func__, handler->ih_name,
ithread->it_name);
mtx_lock(&ithread->it_lock);
#ifdef INVARIANTS
TAILQ_FOREACH(ih, &ithread->it_handlers, ih_next)
if (ih == handler)
goto ok;
mtx_unlock(&ithread->it_lock);
panic("interrupt handler \"%s\" not found in interrupt thread \"%s\"",
ih->ih_name, ithread->it_name);
ok:
#endif
/*
* If the interrupt thread is already running, then just mark this
* handler as being dead and let the ithread do the actual removal.
*
* During a cold boot while cold is set, msleep() does not sleep,
* so we have to remove the handler here rather than letting the
* thread do it.
*/
mtx_lock_spin(&sched_lock);
if (!TD_AWAITING_INTR(ithread->it_td) && !cold) {
handler->ih_flags |= IH_DEAD;
/*
* Ensure that the thread will process the handler list
* again and remove this handler if it has already passed
* it on the list.
*/
ithread->it_need = 1;
} else
TAILQ_REMOVE(&ithread->it_handlers, handler, ih_next);
mtx_unlock_spin(&sched_lock);
if ((handler->ih_flags & IH_DEAD) != 0)
msleep(handler, &ithread->it_lock, PUSER, "itrmh", 0);
ithread_update(ithread);
mtx_unlock(&ithread->it_lock);
free(handler, M_ITHREAD);
return (0);
}
int
Implement preemption of kernel threads natively in the scheduler rather than as one-off hacks in various other parts of the kernel: - Add a function maybe_preempt() that is called from sched_add() to determine if a thread about to be added to a run queue should be preempted to directly. If it is not safe to preempt or if the new thread does not have a high enough priority, then the function returns false and sched_add() adds the thread to the run queue. If the thread should be preempted to but the current thread is in a nested critical section, then the flag TDF_OWEPREEMPT is set and the thread is added to the run queue. Otherwise, mi_switch() is called immediately and the thread is never added to the run queue since it is switch to directly. When exiting an outermost critical section, if TDF_OWEPREEMPT is set, then clear it and call mi_switch() to perform the deferred preemption. - Remove explicit preemption from ithread_schedule() as calling setrunqueue() now does all the correct work. This also removes the do_switch argument from ithread_schedule(). - Do not use the manual preemption code in mtx_unlock if the architecture supports native preemption. - Don't call mi_switch() in a loop during shutdown to give ithreads a chance to run if the architecture supports native preemption since the ithreads will just preempt DELAY(). - Don't call mi_switch() from the page zeroing idle thread for architectures that support native preemption as it is unnecessary. - Native preemption is enabled on the same archs that supported ithread preemption, namely alpha, i386, and amd64. This change should largely be a NOP for the default case as committed except that we will do fewer context switches in a few cases and will avoid the run queues completely when preempting. Approved by: scottl (with his re@ hat)
2004-07-02 20:21:44 +00:00
ithread_schedule(struct ithd *ithread)
{
struct int_entropy entropy;
struct thread *td;
struct thread *ctd;
struct proc *p;
/*
* If no ithread or no handlers, then we have a stray interrupt.
*/
if ((ithread == NULL) || TAILQ_EMPTY(&ithread->it_handlers))
return (EINVAL);
ctd = curthread;
td = ithread->it_td;
p = td->td_proc;
/*
* If any of the handlers for this ithread claim to be good
* sources of entropy, then gather some.
*/
if (harvest.interrupt && ithread->it_flags & IT_ENTROPY) {
CTR3(KTR_INTR, "%s: pid %d (%s) gathering entropy", __func__,
p->p_pid, p->p_comm);
entropy.vector = ithread->it_vector;
2002-09-06 00:18:52 +00:00
entropy.proc = ctd->td_proc;
random_harvest(&entropy, sizeof(entropy), 2, 0,
RANDOM_INTERRUPT);
}
KASSERT(p != NULL, ("ithread %s has no process", ithread->it_name));
CTR4(KTR_INTR, "%s: pid %d: (%s) need = %d",
__func__, p->p_pid, p->p_comm, ithread->it_need);
/*
* Set it_need to tell the thread to keep running if it is already
* running. Then, grab sched_lock and see if we actually need to
Implement preemption of kernel threads natively in the scheduler rather than as one-off hacks in various other parts of the kernel: - Add a function maybe_preempt() that is called from sched_add() to determine if a thread about to be added to a run queue should be preempted to directly. If it is not safe to preempt or if the new thread does not have a high enough priority, then the function returns false and sched_add() adds the thread to the run queue. If the thread should be preempted to but the current thread is in a nested critical section, then the flag TDF_OWEPREEMPT is set and the thread is added to the run queue. Otherwise, mi_switch() is called immediately and the thread is never added to the run queue since it is switch to directly. When exiting an outermost critical section, if TDF_OWEPREEMPT is set, then clear it and call mi_switch() to perform the deferred preemption. - Remove explicit preemption from ithread_schedule() as calling setrunqueue() now does all the correct work. This also removes the do_switch argument from ithread_schedule(). - Do not use the manual preemption code in mtx_unlock if the architecture supports native preemption. - Don't call mi_switch() in a loop during shutdown to give ithreads a chance to run if the architecture supports native preemption since the ithreads will just preempt DELAY(). - Don't call mi_switch() from the page zeroing idle thread for architectures that support native preemption as it is unnecessary. - Native preemption is enabled on the same archs that supported ithread preemption, namely alpha, i386, and amd64. This change should largely be a NOP for the default case as committed except that we will do fewer context switches in a few cases and will avoid the run queues completely when preempting. Approved by: scottl (with his re@ hat)
2004-07-02 20:21:44 +00:00
* put this thread on the runqueue.
*/
ithread->it_need = 1;
mtx_lock_spin(&sched_lock);
if (TD_AWAITING_INTR(td)) {
CTR2(KTR_INTR, "%s: setrunqueue %d", __func__, p->p_pid);
TD_CLR_IWAIT(td);
setrunqueue(td, SRQ_INTR);
} else {
CTR4(KTR_INTR, "%s: pid %d: it_need %d, state %d",
__func__, p->p_pid, ithread->it_need, td->td_state);
}
mtx_unlock_spin(&sched_lock);
return (0);
}
int
swi_add(struct ithd **ithdp, const char *name, driver_intr_t handler,
void *arg, int pri, enum intr_type flags, void **cookiep)
{
struct ithd *ithd;
int error;
if (flags & (INTR_FAST | INTR_ENTROPY))
return (EINVAL);
ithd = (ithdp != NULL) ? *ithdp : NULL;
if (ithd != NULL) {
if ((ithd->it_flags & IT_SOFT) == 0)
return(EINVAL);
} else {
error = ithread_create(&ithd, pri, IT_SOFT, NULL, NULL,
"swi%d:", pri);
if (error)
return (error);
if (ithdp != NULL)
*ithdp = ithd;
}
Implement a unified run queue and adjust priority levels accordingly. - All processes go into the same array of queues, with different scheduling classes using different portions of the array. This allows user processes to have their priorities propogated up into interrupt thread range if need be. - I chose 64 run queues as an arbitrary number that is greater than 32. We used to have 4 separate arrays of 32 queues each, so this may not be optimal. The new run queue code was written with this in mind; changing the number of run queues only requires changing constants in runq.h and adjusting the priority levels. - The new run queue code takes the run queue as a parameter. This is intended to be used to create per-cpu run queues. Implement wrappers for compatibility with the old interface which pass in the global run queue structure. - Group the priority level, user priority, native priority (before propogation) and the scheduling class into a struct priority. - Change any hard coded priority levels that I found to use symbolic constants (TTIPRI and TTOPRI). - Remove the curpriority global variable and use that of curproc. This was used to detect when a process' priority had lowered and it should yield. We now effectively yield on every interrupt. - Activate propogate_priority(). It should now have the desired effect without needing to also propogate the scheduling class. - Temporarily comment out the call to vm_page_zero_idle() in the idle loop. It interfered with propogate_priority() because the idle process needed to do a non-blocking acquire of Giant and then other processes would try to propogate their priority onto it. The idle process should not do anything except idle. vm_page_zero_idle() will return in the form of an idle priority kernel thread which is woken up at apprioriate times by the vm system. - Update struct kinfo_proc to the new priority interface. Deliberately change its size by adjusting the spare fields. It remained the same size, but the layout has changed, so userland processes that use it would parse the data incorrectly. The size constraint should really be changed to an arbitrary version number. Also add a debug.sizeof sysctl node for struct kinfo_proc.
2001-02-12 00:20:08 +00:00
return (ithread_add_handler(ithd, name, handler, arg,
(pri * RQ_PPQ) + PI_SOFT, flags, cookiep));
Refactor a bunch of scheduler code to give basically the same behaviour but with slightly cleaned up interfaces. The KSE structure has become the same as the "per thread scheduler private data" structure. In order to not make the diffs too great one is #defined as the other at this time. The KSE (or td_sched) structure is now allocated per thread and has no allocation code of its own. Concurrency for a KSEGRP is now kept track of via a simple pair of counters rather than using KSE structures as tokens. Since the KSE structure is different in each scheduler, kern_switch.c is now included at the end of each scheduler. Nothing outside the scheduler knows the contents of the KSE (aka td_sched) structure. The fields in the ksegrp structure that are to do with the scheduler's queueing mechanisms are now moved to the kg_sched structure. (per ksegrp scheduler private data structure). In other words how the scheduler queues and keeps track of threads is no-one's business except the scheduler's. This should allow people to write experimental schedulers with completely different internal structuring. A scheduler call sched_set_concurrency(kg, N) has been added that notifies teh scheduler that no more than N threads from that ksegrp should be allowed to be on concurrently scheduled. This is also used to enforce 'fainess' at this time so that a ksegrp with 10000 threads can not swamp a the run queue and force out a process with 1 thread, since the current code will not set the concurrency above NCPU, and both schedulers will not allow more than that many onto the system run queue at a time. Each scheduler should eventualy develop their own methods to do this now that they are effectively separated. Rejig libthr's kernel interface to follow the same code paths as linkse for scope system threads. This has slightly hurt libthr's performance but I will work to recover as much of it as I can. Thread exit code has been cleaned up greatly. exit and exec code now transitions a process back to 'standard non-threaded mode' before taking the next step. Reviewed by: scottl, peter MFC after: 1 week
2004-09-05 02:09:54 +00:00
/* XXKSE.. think of a better way to get separate queues */
}
/*
* Schedule a heavyweight software interrupt process.
*/
void
swi_sched(void *cookie, int flags)
{
struct intrhand *ih = (struct intrhand *)cookie;
struct ithd *it = ih->ih_ithread;
int error;
atomic_add_int(&cnt.v_intr, 1); /* one more global interrupt */
CTR3(KTR_INTR, "swi_sched pid %d(%s) need=%d",
it->it_td->td_proc->p_pid, it->it_td->td_proc->p_comm, it->it_need);
/*
* Set ih_need for this handler so that if the ithread is already
* running it will execute this handler on the next pass. Otherwise,
* it will execute it the next time it runs.
*/
atomic_store_rel_int(&ih->ih_need, 1);
if (!(flags & SWI_DELAY)) {
Implement preemption of kernel threads natively in the scheduler rather than as one-off hacks in various other parts of the kernel: - Add a function maybe_preempt() that is called from sched_add() to determine if a thread about to be added to a run queue should be preempted to directly. If it is not safe to preempt or if the new thread does not have a high enough priority, then the function returns false and sched_add() adds the thread to the run queue. If the thread should be preempted to but the current thread is in a nested critical section, then the flag TDF_OWEPREEMPT is set and the thread is added to the run queue. Otherwise, mi_switch() is called immediately and the thread is never added to the run queue since it is switch to directly. When exiting an outermost critical section, if TDF_OWEPREEMPT is set, then clear it and call mi_switch() to perform the deferred preemption. - Remove explicit preemption from ithread_schedule() as calling setrunqueue() now does all the correct work. This also removes the do_switch argument from ithread_schedule(). - Do not use the manual preemption code in mtx_unlock if the architecture supports native preemption. - Don't call mi_switch() in a loop during shutdown to give ithreads a chance to run if the architecture supports native preemption since the ithreads will just preempt DELAY(). - Don't call mi_switch() from the page zeroing idle thread for architectures that support native preemption as it is unnecessary. - Native preemption is enabled on the same archs that supported ithread preemption, namely alpha, i386, and amd64. This change should largely be a NOP for the default case as committed except that we will do fewer context switches in a few cases and will avoid the run queues completely when preempting. Approved by: scottl (with his re@ hat)
2004-07-02 20:21:44 +00:00
error = ithread_schedule(it);
KASSERT(error == 0, ("stray software interrupt"));
}
}
/*
* This is the main code for interrupt threads.
*/
static void
ithread_loop(void *arg)
{
struct ithd *ithd; /* our thread context */
struct intrhand *ih; /* and our interrupt handler chain */
struct thread *td;
struct proc *p;
int count, warned, storming;
td = curthread;
p = td->td_proc;
ithd = (struct ithd *)arg; /* point to myself */
KASSERT(ithd->it_td == td && td->td_ithd == ithd,
("%s: ithread and proc linkage out of sync", __func__));
count = 0;
- Enable (unmask) interrupt sources earlier in the ithread loop. Specifically, we used to enable the source after locking sched_lock and just before we had already decided to do a context switch. This meant that an ithread could never process more than one interrupt per context switch. Enabling earlier in the loop before sched_lock is acquired allows an ithread to handle multiple interrupts per context switch if interrupts fire very rapidly. For the case of heavy interrupt load this can reduce the number of context switches (and thus overhead) as well as reduce interrupt latency. - Now that we can handle multiple interrupts per context switch, add simple interrupt storm protection to threaded interrupts. If X number of consecutive interrupts are triggered before the itherad voluntarily yields to another thread, then the interrupt thread will sleep with the associated interrupt source disabled (masked) for 1/10th of a second. The default value of X is 500, but it can be tweaked via the tunable/ sysctl hw.intr_storm_threshold. If an interrupt storm is detected, then a message is output to the kernel console on the first occurrence per interrupt thread. Interrupt storm protection can be disabled completely by setting this value to 0. There is no scientific reasoning for the 1/10th of a second or 500 interrupts values, so they may require tweaking at some point in the future. Tested by: rwatson (an earlier version w/o the storm protection) Tested by: mux (reportedly made a machine with two PCI interrupts storming usable rather than hard locked) Reviewed by: imp
2004-04-16 20:25:40 +00:00
warned = 0;
storming = 0;
/*
* As long as we have interrupts outstanding, go through the
* list of handlers, giving each one a go at it.
*/
for (;;) {
/*
* If we are an orphaned thread, then just die.
*/
if (ithd->it_flags & IT_DEAD) {
CTR3(KTR_INTR, "%s: pid %d: (%s) exiting", __func__,
p->p_pid, p->p_comm);
td->td_ithd = NULL;
mtx_destroy(&ithd->it_lock);
free(ithd, M_ITHREAD);
kthread_exit(0);
}
CTR4(KTR_INTR, "%s: pid %d: (%s) need=%d", __func__,
p->p_pid, p->p_comm, ithd->it_need);
while (ithd->it_need) {
/*
* Service interrupts. If another interrupt
* arrives while we are running, they will set
* it_need to denote that we should make
* another pass.
*/
atomic_store_rel_int(&ithd->it_need, 0);
restart:
TAILQ_FOREACH(ih, &ithd->it_handlers, ih_next) {
if (ithd->it_flags & IT_SOFT && !ih->ih_need)
continue;
atomic_store_rel_int(&ih->ih_need, 0);
CTR6(KTR_INTR,
"%s: pid %d ih=%p: %p(%p) flg=%x", __func__,
p->p_pid, (void *)ih,
(void *)ih->ih_handler, ih->ih_argument,
ih->ih_flags);
if ((ih->ih_flags & IH_DEAD) != 0) {
mtx_lock(&ithd->it_lock);
TAILQ_REMOVE(&ithd->it_handlers, ih,
ih_next);
wakeup(ih);
mtx_unlock(&ithd->it_lock);
goto restart;
}
if ((ih->ih_flags & IH_MPSAFE) == 0)
mtx_lock(&Giant);
ih->ih_handler(ih->ih_argument);
if ((ih->ih_flags & IH_MPSAFE) == 0)
Change and clean the mutex lock interface. mtx_enter(lock, type) becomes: mtx_lock(lock) for sleep locks (MTX_DEF-initialized locks) mtx_lock_spin(lock) for spin locks (MTX_SPIN-initialized) similarily, for releasing a lock, we now have: mtx_unlock(lock) for MTX_DEF and mtx_unlock_spin(lock) for MTX_SPIN. We change the caller interface for the two different types of locks because the semantics are entirely different for each case, and this makes it explicitly clear and, at the same time, it rids us of the extra `type' argument. The enter->lock and exit->unlock change has been made with the idea that we're "locking data" and not "entering locked code" in mind. Further, remove all additional "flags" previously passed to the lock acquire/release routines with the exception of two: MTX_QUIET and MTX_NOSWITCH The functionality of these flags is preserved and they can be passed to the lock/unlock routines by calling the corresponding wrappers: mtx_{lock, unlock}_flags(lock, flag(s)) and mtx_{lock, unlock}_spin_flags(lock, flag(s)) for MTX_DEF and MTX_SPIN locks, respectively. Re-inline some lock acq/rel code; in the sleep lock case, we only inline the _obtain_lock()s in order to ensure that the inlined code fits into a cache line. In the spin lock case, we inline recursion and actually only perform a function call if we need to spin. This change has been made with the idea that we generally tend to avoid spin locks and that also the spin locks that we do have and are heavily used (i.e. sched_lock) do recurse, and therefore in an effort to reduce function call overhead for some architectures (such as alpha), we inline recursion for this case. Create a new malloc type for the witness code and retire from using the M_DEV type. The new type is called M_WITNESS and is only declared if WITNESS is enabled. Begin cleaning up some machdep/mutex.h code - specifically updated the "optimized" inlined code in alpha/mutex.h and wrote MTX_LOCK_SPIN and MTX_UNLOCK_SPIN asm macros for the i386/mutex.h as we presently need those. Finally, caught up to the interface changes in all sys code. Contributors: jake, jhb, jasone (in no particular order)
2001-02-09 06:11:45 +00:00
mtx_unlock(&Giant);
}
/*
* Interrupt storm handling:
*
* If this interrupt source is currently storming,
* then throttle it to only fire the handler once
* per clock tick.
*
* If this interrupt source is not currently
* storming, but the number of back to back
* interrupts exceeds the storm threshold, then
* enter storming mode.
*/
if (!storming && intr_storm_threshold != 0 &&
count >= intr_storm_threshold) {
if (!warned) {
printf(
"Interrupt storm detected on \"%s\"; throttling interrupt source\n",
p->p_comm);
warned = 1;
}
storming = 1;
}
if (storming)
tsleep(&count, td->td_priority, "istorm", 1);
else
count++;
if (ithd->it_enable != NULL)
ithd->it_enable(ithd->it_vector);
}
- Enable (unmask) interrupt sources earlier in the ithread loop. Specifically, we used to enable the source after locking sched_lock and just before we had already decided to do a context switch. This meant that an ithread could never process more than one interrupt per context switch. Enabling earlier in the loop before sched_lock is acquired allows an ithread to handle multiple interrupts per context switch if interrupts fire very rapidly. For the case of heavy interrupt load this can reduce the number of context switches (and thus overhead) as well as reduce interrupt latency. - Now that we can handle multiple interrupts per context switch, add simple interrupt storm protection to threaded interrupts. If X number of consecutive interrupts are triggered before the itherad voluntarily yields to another thread, then the interrupt thread will sleep with the associated interrupt source disabled (masked) for 1/10th of a second. The default value of X is 500, but it can be tweaked via the tunable/ sysctl hw.intr_storm_threshold. If an interrupt storm is detected, then a message is output to the kernel console on the first occurrence per interrupt thread. Interrupt storm protection can be disabled completely by setting this value to 0. There is no scientific reasoning for the 1/10th of a second or 500 interrupts values, so they may require tweaking at some point in the future. Tested by: rwatson (an earlier version w/o the storm protection) Tested by: mux (reportedly made a machine with two PCI interrupts storming usable rather than hard locked) Reviewed by: imp
2004-04-16 20:25:40 +00:00
WITNESS_WARN(WARN_PANIC, NULL, "suspending ithread");
mtx_assert(&Giant, MA_NOTOWNED);
/*
* Processed all our interrupts. Now get the sched
* lock. This may take a while and it_need may get
* set again, so we have to check it again.
*/
Change and clean the mutex lock interface. mtx_enter(lock, type) becomes: mtx_lock(lock) for sleep locks (MTX_DEF-initialized locks) mtx_lock_spin(lock) for spin locks (MTX_SPIN-initialized) similarily, for releasing a lock, we now have: mtx_unlock(lock) for MTX_DEF and mtx_unlock_spin(lock) for MTX_SPIN. We change the caller interface for the two different types of locks because the semantics are entirely different for each case, and this makes it explicitly clear and, at the same time, it rids us of the extra `type' argument. The enter->lock and exit->unlock change has been made with the idea that we're "locking data" and not "entering locked code" in mind. Further, remove all additional "flags" previously passed to the lock acquire/release routines with the exception of two: MTX_QUIET and MTX_NOSWITCH The functionality of these flags is preserved and they can be passed to the lock/unlock routines by calling the corresponding wrappers: mtx_{lock, unlock}_flags(lock, flag(s)) and mtx_{lock, unlock}_spin_flags(lock, flag(s)) for MTX_DEF and MTX_SPIN locks, respectively. Re-inline some lock acq/rel code; in the sleep lock case, we only inline the _obtain_lock()s in order to ensure that the inlined code fits into a cache line. In the spin lock case, we inline recursion and actually only perform a function call if we need to spin. This change has been made with the idea that we generally tend to avoid spin locks and that also the spin locks that we do have and are heavily used (i.e. sched_lock) do recurse, and therefore in an effort to reduce function call overhead for some architectures (such as alpha), we inline recursion for this case. Create a new malloc type for the witness code and retire from using the M_DEV type. The new type is called M_WITNESS and is only declared if WITNESS is enabled. Begin cleaning up some machdep/mutex.h code - specifically updated the "optimized" inlined code in alpha/mutex.h and wrote MTX_LOCK_SPIN and MTX_UNLOCK_SPIN asm macros for the i386/mutex.h as we presently need those. Finally, caught up to the interface changes in all sys code. Contributors: jake, jhb, jasone (in no particular order)
2001-02-09 06:11:45 +00:00
mtx_lock_spin(&sched_lock);
if (!ithd->it_need) {
- Enable (unmask) interrupt sources earlier in the ithread loop. Specifically, we used to enable the source after locking sched_lock and just before we had already decided to do a context switch. This meant that an ithread could never process more than one interrupt per context switch. Enabling earlier in the loop before sched_lock is acquired allows an ithread to handle multiple interrupts per context switch if interrupts fire very rapidly. For the case of heavy interrupt load this can reduce the number of context switches (and thus overhead) as well as reduce interrupt latency. - Now that we can handle multiple interrupts per context switch, add simple interrupt storm protection to threaded interrupts. If X number of consecutive interrupts are triggered before the itherad voluntarily yields to another thread, then the interrupt thread will sleep with the associated interrupt source disabled (masked) for 1/10th of a second. The default value of X is 500, but it can be tweaked via the tunable/ sysctl hw.intr_storm_threshold. If an interrupt storm is detected, then a message is output to the kernel console on the first occurrence per interrupt thread. Interrupt storm protection can be disabled completely by setting this value to 0. There is no scientific reasoning for the 1/10th of a second or 500 interrupts values, so they may require tweaking at some point in the future. Tested by: rwatson (an earlier version w/o the storm protection) Tested by: mux (reportedly made a machine with two PCI interrupts storming usable rather than hard locked) Reviewed by: imp
2004-04-16 20:25:40 +00:00
TD_SET_IWAIT(td);
count = 0;
storming = 0;
CTR2(KTR_INTR, "%s: pid %d: done", __func__, p->p_pid);
mi_switch(SW_VOL, NULL);
CTR2(KTR_INTR, "%s: pid %d: resumed", __func__, p->p_pid);
}
Change and clean the mutex lock interface. mtx_enter(lock, type) becomes: mtx_lock(lock) for sleep locks (MTX_DEF-initialized locks) mtx_lock_spin(lock) for spin locks (MTX_SPIN-initialized) similarily, for releasing a lock, we now have: mtx_unlock(lock) for MTX_DEF and mtx_unlock_spin(lock) for MTX_SPIN. We change the caller interface for the two different types of locks because the semantics are entirely different for each case, and this makes it explicitly clear and, at the same time, it rids us of the extra `type' argument. The enter->lock and exit->unlock change has been made with the idea that we're "locking data" and not "entering locked code" in mind. Further, remove all additional "flags" previously passed to the lock acquire/release routines with the exception of two: MTX_QUIET and MTX_NOSWITCH The functionality of these flags is preserved and they can be passed to the lock/unlock routines by calling the corresponding wrappers: mtx_{lock, unlock}_flags(lock, flag(s)) and mtx_{lock, unlock}_spin_flags(lock, flag(s)) for MTX_DEF and MTX_SPIN locks, respectively. Re-inline some lock acq/rel code; in the sleep lock case, we only inline the _obtain_lock()s in order to ensure that the inlined code fits into a cache line. In the spin lock case, we inline recursion and actually only perform a function call if we need to spin. This change has been made with the idea that we generally tend to avoid spin locks and that also the spin locks that we do have and are heavily used (i.e. sched_lock) do recurse, and therefore in an effort to reduce function call overhead for some architectures (such as alpha), we inline recursion for this case. Create a new malloc type for the witness code and retire from using the M_DEV type. The new type is called M_WITNESS and is only declared if WITNESS is enabled. Begin cleaning up some machdep/mutex.h code - specifically updated the "optimized" inlined code in alpha/mutex.h and wrote MTX_LOCK_SPIN and MTX_UNLOCK_SPIN asm macros for the i386/mutex.h as we presently need those. Finally, caught up to the interface changes in all sys code. Contributors: jake, jhb, jasone (in no particular order)
2001-02-09 06:11:45 +00:00
mtx_unlock_spin(&sched_lock);
}
}
#ifdef DDB
/*
* Dump details about an interrupt handler
*/
static void
db_dump_intrhand(struct intrhand *ih)
{
int comma;
db_printf("\t%-10s ", ih->ih_name);
switch (ih->ih_pri) {
case PI_REALTIME:
db_printf("CLK ");
break;
case PI_AV:
db_printf("AV ");
break;
case PI_TTYHIGH:
case PI_TTYLOW:
db_printf("TTY ");
break;
case PI_TAPE:
db_printf("TAPE");
break;
case PI_NET:
db_printf("NET ");
break;
case PI_DISK:
case PI_DISKLOW:
db_printf("DISK");
break;
case PI_DULL:
db_printf("DULL");
break;
default:
if (ih->ih_pri >= PI_SOFT)
db_printf("SWI ");
else
db_printf("%4u", ih->ih_pri);
break;
}
db_printf(" ");
db_printsym((uintptr_t)ih->ih_handler, DB_STGY_PROC);
db_printf("(%p)", ih->ih_argument);
if (ih->ih_need ||
(ih->ih_flags & (IH_FAST | IH_EXCLUSIVE | IH_ENTROPY | IH_DEAD |
IH_MPSAFE)) != 0) {
db_printf(" {");
comma = 0;
if (ih->ih_flags & IH_FAST) {
db_printf("FAST");
comma = 1;
}
if (ih->ih_flags & IH_EXCLUSIVE) {
if (comma)
db_printf(", ");
db_printf("EXCL");
comma = 1;
}
if (ih->ih_flags & IH_ENTROPY) {
if (comma)
db_printf(", ");
db_printf("ENTROPY");
comma = 1;
}
if (ih->ih_flags & IH_DEAD) {
if (comma)
db_printf(", ");
db_printf("DEAD");
comma = 1;
}
if (ih->ih_flags & IH_MPSAFE) {
if (comma)
db_printf(", ");
db_printf("MPSAFE");
comma = 1;
}
if (ih->ih_need) {
if (comma)
db_printf(", ");
db_printf("NEED");
}
db_printf("}");
}
db_printf("\n");
}
/*
* Dump details about an ithread
*/
void
db_dump_ithread(struct ithd *ithd, int handlers)
{
struct proc *p;
struct intrhand *ih;
int comma;
if (ithd->it_td != NULL) {
p = ithd->it_td->td_proc;
db_printf("%s (pid %d)", p->p_comm, p->p_pid);
} else
db_printf("%s: (no thread)", ithd->it_name);
if ((ithd->it_flags & (IT_SOFT | IT_ENTROPY | IT_DEAD)) != 0 ||
ithd->it_need) {
db_printf(" {");
comma = 0;
if (ithd->it_flags & IT_SOFT) {
db_printf("SOFT");
comma = 1;
}
if (ithd->it_flags & IT_ENTROPY) {
if (comma)
db_printf(", ");
db_printf("ENTROPY");
comma = 1;
}
if (ithd->it_flags & IT_DEAD) {
if (comma)
db_printf(", ");
db_printf("DEAD");
comma = 1;
}
if (ithd->it_need) {
if (comma)
db_printf(", ");
db_printf("NEED");
}
db_printf("}");
}
db_printf("\n");
if (handlers)
TAILQ_FOREACH(ih, &ithd->it_handlers, ih_next)
db_dump_intrhand(ih);
}
#endif /* DDB */
/*
* Start standard software interrupt threads
*/
static void
start_softintr(void *dummy)
{
struct proc *p;
if (swi_add(&clk_ithd, "clock", softclock, NULL, SWI_CLOCK,
INTR_MPSAFE, &softclock_ih) ||
swi_add(NULL, "vm", swi_vm, NULL, SWI_VM, INTR_MPSAFE, &vm_ih))
panic("died while creating standard software ithreads");
p = clk_ithd->it_td->td_proc;
PROC_LOCK(p);
p->p_flag |= P_NOLOAD;
PROC_UNLOCK(p);
}
SYSINIT(start_softintr, SI_SUB_SOFTINTR, SI_ORDER_FIRST, start_softintr, NULL)
/*
* Sysctls used by systat and others: hw.intrnames and hw.intrcnt.
* The data for this machine dependent, and the declarations are in machine
* dependent code. The layout of intrnames and intrcnt however is machine
* independent.
*
* We do not know the length of intrcnt and intrnames at compile time, so
* calculate things at run time.
*/
static int
sysctl_intrnames(SYSCTL_HANDLER_ARGS)
{
return (sysctl_handle_opaque(oidp, intrnames, eintrnames - intrnames,
req));
}
SYSCTL_PROC(_hw, OID_AUTO, intrnames, CTLTYPE_OPAQUE | CTLFLAG_RD,
NULL, 0, sysctl_intrnames, "", "Interrupt Names");
static int
sysctl_intrcnt(SYSCTL_HANDLER_ARGS)
{
return (sysctl_handle_opaque(oidp, intrcnt,
(char *)eintrcnt - (char *)intrcnt, req));
}
SYSCTL_PROC(_hw, OID_AUTO, intrcnt, CTLTYPE_OPAQUE | CTLFLAG_RD,
NULL, 0, sysctl_intrcnt, "", "Interrupt Counts");
#ifdef DDB
/*
* DDB command to dump the interrupt statistics.
*/
DB_SHOW_COMMAND(intrcnt, db_show_intrcnt)
{
u_long *i;
char *cp;
int quit;
cp = intrnames;
db_setup_paging(db_simple_pager, &quit, db_lines_per_page);
for (i = intrcnt, quit = 0; i != eintrcnt && !quit; i++) {
if (*cp == '\0')
break;
if (*i != 0)
db_printf("%s\t%lu\n", cp, *i);
cp += strlen(cp) + 1;
}
}
#endif