freebsd-skq/sys/kern/sched_4bsd.c

1688 lines
42 KiB
C
Raw Normal View History

/*-
* Copyright (c) 1982, 1986, 1990, 1991, 1993
* The Regents of the University of California. All rights reserved.
* (c) UNIX System Laboratories, Inc.
* All or some portions of this file are derived from material licensed
* to the University of California by American Telephone and Telegraph
* Co. or Unix System Laboratories, Inc. and are reproduced herein with
* the permission of UNIX System Laboratories, Inc.
*
* 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.
* 4. Neither the name of the University nor the names of its contributors
* may be used to endorse or promote products derived from this software
* without specific prior written permission.
*
* THIS SOFTWARE IS PROVIDED BY THE REGENTS 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 REGENTS 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.
*/
2003-06-11 00:56:59 +00:00
#include <sys/cdefs.h>
__FBSDID("$FreeBSD$");
#include "opt_hwpmc_hooks.h"
#include "opt_sched.h"
#include "opt_kdtrace.h"
#include <sys/param.h>
#include <sys/systm.h>
#include <sys/cpuset.h>
#include <sys/kernel.h>
#include <sys/ktr.h>
#include <sys/lock.h>
#include <sys/kthread.h>
#include <sys/mutex.h>
#include <sys/proc.h>
#include <sys/resourcevar.h>
#include <sys/sched.h>
#include <sys/smp.h>
#include <sys/sysctl.h>
#include <sys/sx.h>
Rework the interface between priority propagation (lending) and the schedulers a bit to ensure more correct handling of priorities and fewer priority inversions: - Add two functions to the sched(9) API to handle priority lending: sched_lend_prio() and sched_unlend_prio(). The turnstile code uses these functions to ask the scheduler to lend a thread a set priority and to tell the scheduler when it thinks it is ok for a thread to stop borrowing priority. The unlend case is slightly complex in that the turnstile code tells the scheduler what the minimum priority of the thread needs to be to satisfy the requirements of any other threads blocked on locks owned by the thread in question. The scheduler then decides where the thread can go back to normal mode (if it's normal priority is high enough to satisfy the pending lock requests) or it it should continue to use the priority specified to the sched_unlend_prio() call. This involves adding a new per-thread flag TDF_BORROWING that replaces the ULE-only kse flag for priority elevation. - Schedulers now refuse to lower the priority of a thread that is currently borrowing another therad's priority. - If a scheduler changes the priority of a thread that is currently sitting on a turnstile, it will call a new function turnstile_adjust() to inform the turnstile code of the change. This function resorts the thread on the priority list of the turnstile if needed, and if the thread ends up at the head of the list (due to having the highest priority) and its priority was raised, then it will propagate that new priority to the owner of the lock it is blocked on. Some additional fixes specific to the 4BSD scheduler include: - Common code for updating the priority of a thread when the user priority of its associated kse group has been consolidated in a new static function resetpriority_thread(). One change to this function is that it will now only adjust the priority of a thread if it already has a time sharing priority, thus preserving any boosts from a tsleep() until the thread returns to userland. Also, resetpriority() no longer calls maybe_resched() on each thread in the group. Instead, the code calling resetpriority() is responsible for calling resetpriority_thread() on any threads that need to be updated. - schedcpu() now uses resetpriority_thread() instead of just calling sched_prio() directly after it updates a kse group's user priority. - sched_clock() now uses resetpriority_thread() rather than writing directly to td_priority. - sched_nice() now updates all the priorities of the threads after the group priority has been adjusted. Discussed with: bde Reviewed by: ups, jeffr Tested on: 4bsd, ule Tested on: i386, alpha, sparc64
2004-12-30 20:52:44 +00:00
#include <sys/turnstile.h>
#include <sys/umtx.h>
#include <machine/pcb.h>
#include <machine/smp.h>
#ifdef HWPMC_HOOKS
#include <sys/pmckern.h>
#endif
#ifdef KDTRACE_HOOKS
#include <sys/dtrace_bsd.h>
int dtrace_vtime_active;
dtrace_vtime_switch_func_t dtrace_vtime_switch_func;
#endif
/*
* INVERSE_ESTCPU_WEIGHT is only suitable for statclock() frequencies in
* the range 100-256 Hz (approximately).
*/
#define ESTCPULIM(e) \
min((e), INVERSE_ESTCPU_WEIGHT * (NICE_WEIGHT * (PRIO_MAX - PRIO_MIN) - \
RQ_PPQ) + INVERSE_ESTCPU_WEIGHT - 1)
Quick fix for scaling of statclock ticks in the SMP case. As explained in the log message for kern_sched.c 1.83 (which should have been repo-copied to preserve history for this file), the (4BSD) scheduler algorithm only works right if stathz is nearly 128 Hz. The old commit lock said 64 Hz; the scheduler actually wants nearly 16 Hz but there was a scale factor of 4 to give the requirement of 64 Hz, and rev.1.83 changed the scale factor so that the requirement became 128 Hz. The change of the scale factor was incomplete in the SMP case. Then scheduling ticks are provided by smp_ncpu CPUs, and the scheduler cannot tell the difference between this and 1 CPU providing scheduling ticks smp_ncpu times faster, so we need another scale factor of smp_ncp or an algorithm change. This quick fix uses the scale factor without even trying to optimize the runtime divisions required for this as is done for the other scale factor. The main algorithmic problem is the clamp on the scheduling tick counts. This was 295; it is now approximately 295 * smp_ncpu. When the limit is reached, threads get free timeslices and scheduling becomes very unfair to the threads that don't hit the limit. The limit can be reached and maintained in the worst case if the load average is larger than (limit / effective_stathz - 1) / 2 = 0.65 now (was just 0.08 with 2 CPUs before this change), so there are algorithmic problems even for a load average of 1. Fortunately, the worst case isn't common enough for the problem to be very noticeable (it is mainly for niced CPU hogs competing with less nice CPU hogs).
2003-11-09 13:45:54 +00:00
#ifdef SMP
#define INVERSE_ESTCPU_WEIGHT (8 * smp_cpus)
#else
#define INVERSE_ESTCPU_WEIGHT 8 /* 1 / (priorities per estcpu level). */
Quick fix for scaling of statclock ticks in the SMP case. As explained in the log message for kern_sched.c 1.83 (which should have been repo-copied to preserve history for this file), the (4BSD) scheduler algorithm only works right if stathz is nearly 128 Hz. The old commit lock said 64 Hz; the scheduler actually wants nearly 16 Hz but there was a scale factor of 4 to give the requirement of 64 Hz, and rev.1.83 changed the scale factor so that the requirement became 128 Hz. The change of the scale factor was incomplete in the SMP case. Then scheduling ticks are provided by smp_ncpu CPUs, and the scheduler cannot tell the difference between this and 1 CPU providing scheduling ticks smp_ncpu times faster, so we need another scale factor of smp_ncp or an algorithm change. This quick fix uses the scale factor without even trying to optimize the runtime divisions required for this as is done for the other scale factor. The main algorithmic problem is the clamp on the scheduling tick counts. This was 295; it is now approximately 295 * smp_ncpu. When the limit is reached, threads get free timeslices and scheduling becomes very unfair to the threads that don't hit the limit. The limit can be reached and maintained in the worst case if the load average is larger than (limit / effective_stathz - 1) / 2 = 0.65 now (was just 0.08 with 2 CPUs before this change), so there are algorithmic problems even for a load average of 1. Fortunately, the worst case isn't common enough for the problem to be very noticeable (it is mainly for niced CPU hogs competing with less nice CPU hogs).
2003-11-09 13:45:54 +00:00
#endif
#define NICE_WEIGHT 1 /* Priorities per nice level. */
#define TS_NAME_LEN (MAXCOMLEN + sizeof(" td ") + sizeof(__XSTRING(UINT_MAX)))
/*
* The schedulable entity that runs a context.
* This is an extension to the thread structure and is tailored to
* the requirements of this scheduler
*/
struct td_sched {
fixpt_t ts_pctcpu; /* (j) %cpu during p_swtime. */
int ts_cpticks; /* (j) Ticks of cpu time. */
int ts_slptime; /* (j) Seconds !RUNNING. */
int ts_flags;
struct runq *ts_runq; /* runq the thread is currently on */
#ifdef KTR
char ts_name[TS_NAME_LEN];
#endif
};
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
/* flags kept in td_flags */
#define TDF_DIDRUN TDF_SCHED0 /* thread actually ran. */
#define TDF_BOUND TDF_SCHED1 /* Bound to one CPU. */
/* flags kept in ts_flags */
#define TSF_AFFINITY 0x0001 /* Has a non-"full" CPU set. */
#define SKE_RUNQ_PCPU(ts) \
((ts)->ts_runq != 0 && (ts)->ts_runq != &runq)
#define THREAD_CAN_SCHED(td, cpu) \
CPU_ISSET((cpu), &(td)->td_cpuset->cs_mask)
static struct td_sched td_sched0;
struct mtx sched_lock;
static int sched_tdcnt; /* Total runnable threads in the system. */
static int sched_quantum; /* Roundrobin scheduling quantum in ticks. */
#define SCHED_QUANTUM (hz / 10) /* Default sched quantum */
static void setup_runqs(void);
static void schedcpu(void);
static void schedcpu_thread(void);
Rework the interface between priority propagation (lending) and the schedulers a bit to ensure more correct handling of priorities and fewer priority inversions: - Add two functions to the sched(9) API to handle priority lending: sched_lend_prio() and sched_unlend_prio(). The turnstile code uses these functions to ask the scheduler to lend a thread a set priority and to tell the scheduler when it thinks it is ok for a thread to stop borrowing priority. The unlend case is slightly complex in that the turnstile code tells the scheduler what the minimum priority of the thread needs to be to satisfy the requirements of any other threads blocked on locks owned by the thread in question. The scheduler then decides where the thread can go back to normal mode (if it's normal priority is high enough to satisfy the pending lock requests) or it it should continue to use the priority specified to the sched_unlend_prio() call. This involves adding a new per-thread flag TDF_BORROWING that replaces the ULE-only kse flag for priority elevation. - Schedulers now refuse to lower the priority of a thread that is currently borrowing another therad's priority. - If a scheduler changes the priority of a thread that is currently sitting on a turnstile, it will call a new function turnstile_adjust() to inform the turnstile code of the change. This function resorts the thread on the priority list of the turnstile if needed, and if the thread ends up at the head of the list (due to having the highest priority) and its priority was raised, then it will propagate that new priority to the owner of the lock it is blocked on. Some additional fixes specific to the 4BSD scheduler include: - Common code for updating the priority of a thread when the user priority of its associated kse group has been consolidated in a new static function resetpriority_thread(). One change to this function is that it will now only adjust the priority of a thread if it already has a time sharing priority, thus preserving any boosts from a tsleep() until the thread returns to userland. Also, resetpriority() no longer calls maybe_resched() on each thread in the group. Instead, the code calling resetpriority() is responsible for calling resetpriority_thread() on any threads that need to be updated. - schedcpu() now uses resetpriority_thread() instead of just calling sched_prio() directly after it updates a kse group's user priority. - sched_clock() now uses resetpriority_thread() rather than writing directly to td_priority. - sched_nice() now updates all the priorities of the threads after the group priority has been adjusted. Discussed with: bde Reviewed by: ups, jeffr Tested on: 4bsd, ule Tested on: i386, alpha, sparc64
2004-12-30 20:52:44 +00:00
static void sched_priority(struct thread *td, u_char prio);
static void sched_setup(void *dummy);
static void maybe_resched(struct thread *td);
static void updatepri(struct thread *td);
static void resetpriority(struct thread *td);
static void resetpriority_thread(struct thread *td);
#ifdef SMP
static int sched_pickcpu(struct thread *td);
static int forward_wakeup(int cpunum);
static void kick_other_cpu(int pri, int cpuid);
#endif
static struct kproc_desc sched_kp = {
"schedcpu",
schedcpu_thread,
NULL
};
SYSINIT(schedcpu, SI_SUB_RUN_SCHEDULER, SI_ORDER_FIRST, kproc_start,
&sched_kp);
SYSINIT(sched_setup, SI_SUB_RUN_QUEUE, SI_ORDER_FIRST, sched_setup, NULL);
/*
* Global run queue.
*/
static struct runq runq;
#ifdef SMP
/*
* Per-CPU run queues
*/
static struct runq runq_pcpu[MAXCPU];
long runq_length[MAXCPU];
Commit the support for removing cpumask_t and replacing it directly with cpuset_t objects. That is going to offer the underlying support for a simple bump of MAXCPU and then support for number of cpus > 32 (as it is today). Right now, cpumask_t is an int, 32 bits on all our supported architecture. cpumask_t on the other side is implemented as an array of longs, and easilly extendible by definition. The architectures touched by this commit are the following: - amd64 - i386 - pc98 - arm - ia64 - XEN while the others are still missing. Userland is believed to be fully converted with the changes contained here. Some technical notes: - This commit may be considered an ABI nop for all the architectures different from amd64 and ia64 (and sparc64 in the future) - per-cpu members, which are now converted to cpuset_t, needs to be accessed avoiding migration, because the size of cpuset_t should be considered unknown - size of cpuset_t objects is different from kernel and userland (this is primirally done in order to leave some more space in userland to cope with KBI extensions). If you need to access kernel cpuset_t from the userland please refer to example in this patch on how to do that correctly (kgdb may be a good source, for example). - Support for other architectures is going to be added soon - Only MAXCPU for amd64 is bumped now The patch has been tested by sbruno and Nicholas Esborn on opteron 4 x 12 pack CPUs. More testing on big SMP is expected to came soon. pluknet tested the patch with his 8-ways on both amd64 and i386. Tested by: pluknet, sbruno, gianni, Nicholas Esborn Reviewed by: jeff, jhb, sbruno
2011-05-05 14:39:14 +00:00
static cpuset_t idle_cpus_mask;
#endif
struct pcpuidlestat {
u_int idlecalls;
u_int oldidlecalls;
};
static DPCPU_DEFINE(struct pcpuidlestat, idlestat);
static void
setup_runqs(void)
{
#ifdef SMP
int i;
for (i = 0; i < MAXCPU; ++i)
runq_init(&runq_pcpu[i]);
#endif
runq_init(&runq);
}
static int
sysctl_kern_quantum(SYSCTL_HANDLER_ARGS)
{
int error, new_val;
new_val = sched_quantum * tick;
error = sysctl_handle_int(oidp, &new_val, 0, req);
if (error != 0 || req->newptr == NULL)
return (error);
if (new_val < tick)
return (EINVAL);
sched_quantum = new_val / tick;
hogticks = 2 * sched_quantum;
return (0);
}
SYSCTL_NODE(_kern, OID_AUTO, sched, CTLFLAG_RD, 0, "Scheduler");
SYSCTL_STRING(_kern_sched, OID_AUTO, name, CTLFLAG_RD, "4BSD", 0,
"Scheduler name");
SYSCTL_PROC(_kern_sched, OID_AUTO, quantum, CTLTYPE_INT | CTLFLAG_RW,
0, sizeof sched_quantum, sysctl_kern_quantum, "I",
"Roundrobin scheduling quantum in microseconds");
2004-09-03 09:15:10 +00:00
#ifdef SMP
/* Enable forwarding of wakeups to all other cpus */
SYSCTL_NODE(_kern_sched, OID_AUTO, ipiwakeup, CTLFLAG_RD, NULL, "Kernel SMP");
static int runq_fuzz = 1;
SYSCTL_INT(_kern_sched, OID_AUTO, runq_fuzz, CTLFLAG_RW, &runq_fuzz, 0, "");
static int forward_wakeup_enabled = 1;
SYSCTL_INT(_kern_sched_ipiwakeup, OID_AUTO, enabled, CTLFLAG_RW,
&forward_wakeup_enabled, 0,
"Forwarding of wakeup to idle CPUs");
static int forward_wakeups_requested = 0;
SYSCTL_INT(_kern_sched_ipiwakeup, OID_AUTO, requested, CTLFLAG_RD,
&forward_wakeups_requested, 0,
"Requests for Forwarding of wakeup to idle CPUs");
static int forward_wakeups_delivered = 0;
SYSCTL_INT(_kern_sched_ipiwakeup, OID_AUTO, delivered, CTLFLAG_RD,
&forward_wakeups_delivered, 0,
"Completed Forwarding of wakeup to idle CPUs");
static int forward_wakeup_use_mask = 1;
SYSCTL_INT(_kern_sched_ipiwakeup, OID_AUTO, usemask, CTLFLAG_RW,
&forward_wakeup_use_mask, 0,
"Use the mask of idle cpus");
static int forward_wakeup_use_loop = 0;
SYSCTL_INT(_kern_sched_ipiwakeup, OID_AUTO, useloop, CTLFLAG_RW,
&forward_wakeup_use_loop, 0,
"Use a loop to find idle cpus");
2004-09-03 09:15:10 +00:00
#endif
#if 0
static int sched_followon = 0;
SYSCTL_INT(_kern_sched, OID_AUTO, followon, CTLFLAG_RW,
&sched_followon, 0,
"allow threads to share a quantum");
#endif
static __inline void
sched_load_add(void)
{
sched_tdcnt++;
KTR_COUNTER0(KTR_SCHED, "load", "global load", sched_tdcnt);
}
static __inline void
sched_load_rem(void)
{
sched_tdcnt--;
KTR_COUNTER0(KTR_SCHED, "load", "global load", sched_tdcnt);
}
/*
* Arrange to reschedule if necessary, taking the priorities and
* schedulers into account.
*/
static void
maybe_resched(struct thread *td)
{
THREAD_LOCK_ASSERT(td, MA_OWNED);
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
if (td->td_priority < curthread->td_priority)
curthread->td_flags |= TDF_NEEDRESCHED;
}
/*
* This function is called when a thread is about to be put on run queue
* because it has been made runnable or its priority has been adjusted. It
* determines if the new thread should be immediately preempted to. If so,
* it switches to it and eventually returns true. If not, it returns false
* so that the caller may place the thread on an appropriate run queue.
*/
int
maybe_preempt(struct thread *td)
{
#ifdef PREEMPTION
struct thread *ctd;
int cpri, pri;
/*
* The new thread should not preempt the current thread if any of the
* following conditions are true:
*
* - The kernel is in the throes of crashing (panicstr).
* - The current thread has a higher (numerically lower) or
* equivalent priority. Note that this prevents curthread from
* trying to preempt to itself.
* - It is too early in the boot for context switches (cold is set).
* - The current thread has an inhibitor set or is in the process of
* exiting. In this case, the current thread is about to switch
* out anyways, so there's no point in preempting. If we did,
* the current thread would not be properly resumed as well, so
* just avoid that whole landmine.
* - If the new thread's priority is not a realtime priority and
* the current thread's priority is not an idle priority and
* FULL_PREEMPTION is disabled.
*
* If all of these conditions are false, but the current thread is in
* a nested critical section, then we have to defer the preemption
* until we exit the critical section. Otherwise, switch immediately
* to the new thread.
*/
ctd = curthread;
THREAD_LOCK_ASSERT(td, MA_OWNED);
KASSERT((td->td_inhibitors == 0),
("maybe_preempt: trying to run inhibited thread"));
pri = td->td_priority;
cpri = ctd->td_priority;
if (panicstr != NULL || pri >= cpri || cold /* || dumping */ ||
TD_IS_INHIBITED(ctd))
return (0);
#ifndef FULL_PREEMPTION
if (pri > PRI_MAX_ITHD && cpri < PRI_MIN_IDLE)
return (0);
#endif
if (ctd->td_critnest > 1) {
CTR1(KTR_PROC, "maybe_preempt: in critical section %d",
ctd->td_critnest);
ctd->td_owepreempt = 1;
return (0);
}
/*
* Thread is runnable but not yet put on system run queue.
*/
MPASS(ctd->td_lock == td->td_lock);
MPASS(TD_ON_RUNQ(td));
TD_SET_RUNNING(td);
CTR3(KTR_PROC, "preempting to thread %p (pid %d, %s)\n", td,
td->td_proc->p_pid, td->td_name);
mi_switch(SW_INVOL | SW_PREEMPT | SWT_PREEMPT, td);
/*
* td's lock pointer may have changed. We have to return with it
* locked.
*/
spinlock_enter();
thread_unlock(ctd);
thread_lock(td);
spinlock_exit();
return (1);
#else
return (0);
#endif
}
/*
* Constants for digital decay and forget:
* 90% of (td_estcpu) usage in 5 * loadav time
* 95% of (ts_pctcpu) usage in 60 seconds (load insensitive)
* Note that, as ps(1) mentions, this can let percentages
* total over 100% (I've seen 137.9% for 3 processes).
*
* Note that schedclock() updates td_estcpu and p_cpticks asynchronously.
*
* We wish to decay away 90% of td_estcpu in (5 * loadavg) seconds.
* That is, the system wants to compute a value of decay such
* that the following for loop:
* for (i = 0; i < (5 * loadavg); i++)
* td_estcpu *= decay;
* will compute
* td_estcpu *= 0.1;
* for all values of loadavg:
*
* Mathematically this loop can be expressed by saying:
* decay ** (5 * loadavg) ~= .1
*
* The system computes decay as:
* decay = (2 * loadavg) / (2 * loadavg + 1)
*
* We wish to prove that the system's computation of decay
* will always fulfill the equation:
* decay ** (5 * loadavg) ~= .1
*
* If we compute b as:
* b = 2 * loadavg
* then
* decay = b / (b + 1)
*
* We now need to prove two things:
* 1) Given factor ** (5 * loadavg) ~= .1, prove factor == b/(b+1)
* 2) Given b/(b+1) ** power ~= .1, prove power == (5 * loadavg)
*
* Facts:
* For x close to zero, exp(x) =~ 1 + x, since
* exp(x) = 0! + x**1/1! + x**2/2! + ... .
* therefore exp(-1/b) =~ 1 - (1/b) = (b-1)/b.
* For x close to zero, ln(1+x) =~ x, since
* ln(1+x) = x - x**2/2 + x**3/3 - ... -1 < x < 1
* therefore ln(b/(b+1)) = ln(1 - 1/(b+1)) =~ -1/(b+1).
* ln(.1) =~ -2.30
*
* Proof of (1):
* Solve (factor)**(power) =~ .1 given power (5*loadav):
* solving for factor,
* ln(factor) =~ (-2.30/5*loadav), or
* factor =~ exp(-1/((5/2.30)*loadav)) =~ exp(-1/(2*loadav)) =
* exp(-1/b) =~ (b-1)/b =~ b/(b+1). QED
*
* Proof of (2):
* Solve (factor)**(power) =~ .1 given factor == (b/(b+1)):
* solving for power,
* power*ln(b/(b+1)) =~ -2.30, or
* power =~ 2.3 * (b + 1) = 4.6*loadav + 2.3 =~ 5*loadav. QED
*
* Actual power values for the implemented algorithm are as follows:
* loadav: 1 2 3 4
* power: 5.68 10.32 14.94 19.55
*/
/* calculations for digital decay to forget 90% of usage in 5*loadav sec */
#define loadfactor(loadav) (2 * (loadav))
#define decay_cpu(loadfac, cpu) (((loadfac) * (cpu)) / ((loadfac) + FSCALE))
/* decay 95% of `ts_pctcpu' in 60 seconds; see CCPU_SHIFT before changing */
static fixpt_t ccpu = 0.95122942450071400909 * FSCALE; /* exp(-1/20) */
SYSCTL_UINT(_kern, OID_AUTO, ccpu, CTLFLAG_RD, &ccpu, 0, "");
/*
* If `ccpu' is not equal to `exp(-1/20)' and you still want to use the
* faster/more-accurate formula, you'll have to estimate CCPU_SHIFT below
* and possibly adjust FSHIFT in "param.h" so that (FSHIFT >= CCPU_SHIFT).
*
* To estimate CCPU_SHIFT for exp(-1/20), the following formula was used:
* 1 - exp(-1/20) ~= 0.0487 ~= 0.0488 == 1 (fixed pt, *11* bits).
*
* If you don't want to bother with the faster/more-accurate formula, you
* can set CCPU_SHIFT to (FSHIFT + 1) which will use a slower/less-accurate
* (more general) method of calculating the %age of CPU used by a process.
*/
#define CCPU_SHIFT 11
/*
* Recompute process priorities, every hz ticks.
* MP-safe, called without the Giant mutex.
*/
/* ARGSUSED */
static void
schedcpu(void)
{
register fixpt_t loadfac = loadfactor(averunnable.ldavg[0]);
struct thread *td;
struct proc *p;
struct td_sched *ts;
int awake, realstathz;
realstathz = stathz ? stathz : hz;
sx_slock(&allproc_lock);
FOREACH_PROC_IN_SYSTEM(p) {
PROC_LOCK(p);
if (p->p_state == PRS_NEW) {
PROC_UNLOCK(p);
continue;
}
FOREACH_THREAD_IN_PROC(p, td) {
awake = 0;
thread_lock(td);
ts = td->td_sched;
/*
* Increment sleep time (if sleeping). We
* ignore overflow, as above.
*/
/*
* The td_sched slptimes are not touched in wakeup
* because the thread may not HAVE everything in
* memory? XXX I think this is out of date.
*/
if (TD_ON_RUNQ(td)) {
awake = 1;
td->td_flags &= ~TDF_DIDRUN;
} else if (TD_IS_RUNNING(td)) {
awake = 1;
/* Do not clear TDF_DIDRUN */
} else if (td->td_flags & TDF_DIDRUN) {
awake = 1;
td->td_flags &= ~TDF_DIDRUN;
}
/*
* ts_pctcpu is only for ps and ttyinfo().
*/
ts->ts_pctcpu = (ts->ts_pctcpu * ccpu) >> FSHIFT;
/*
* If the td_sched has been idle the entire second,
* stop recalculating its priority until
* it wakes up.
*/
if (ts->ts_cpticks != 0) {
#if (FSHIFT >= CCPU_SHIFT)
ts->ts_pctcpu += (realstathz == 100)
? ((fixpt_t) ts->ts_cpticks) <<
(FSHIFT - CCPU_SHIFT) :
100 * (((fixpt_t) ts->ts_cpticks)
<< (FSHIFT - CCPU_SHIFT)) / realstathz;
#else
ts->ts_pctcpu += ((FSCALE - ccpu) *
(ts->ts_cpticks *
FSCALE / realstathz)) >> FSHIFT;
#endif
ts->ts_cpticks = 0;
2006-11-14 05:48:27 +00:00
}
/*
* If there are ANY running threads in this process,
* then don't count it as sleeping.
* XXX: this is broken.
*/
if (awake) {
if (ts->ts_slptime > 1) {
/*
* In an ideal world, this should not
* happen, because whoever woke us
* up from the long sleep should have
* unwound the slptime and reset our
* priority before we run at the stale
* priority. Should KASSERT at some
* point when all the cases are fixed.
*/
updatepri(td);
}
ts->ts_slptime = 0;
} else
ts->ts_slptime++;
if (ts->ts_slptime > 1) {
thread_unlock(td);
continue;
}
td->td_estcpu = decay_cpu(loadfac, td->td_estcpu);
resetpriority(td);
resetpriority_thread(td);
thread_unlock(td);
}
PROC_UNLOCK(p);
}
sx_sunlock(&allproc_lock);
}
/*
* Main loop for a kthread that executes schedcpu once a second.
*/
static void
schedcpu_thread(void)
{
for (;;) {
schedcpu();
pause("-", hz);
}
}
/*
* Recalculate the priority of a process after it has slept for a while.
* For all load averages >= 1 and max td_estcpu of 255, sleeping for at
* least six times the loadfactor will decay td_estcpu to zero.
*/
static void
updatepri(struct thread *td)
{
struct td_sched *ts;
fixpt_t loadfac;
unsigned int newcpu;
ts = td->td_sched;
loadfac = loadfactor(averunnable.ldavg[0]);
if (ts->ts_slptime > 5 * loadfac)
td->td_estcpu = 0;
else {
newcpu = td->td_estcpu;
ts->ts_slptime--; /* was incremented in schedcpu() */
while (newcpu && --ts->ts_slptime)
newcpu = decay_cpu(loadfac, newcpu);
td->td_estcpu = newcpu;
}
}
/*
* Compute the priority of a process when running in user mode.
* Arrange to reschedule if the resulting priority is better
* than that of the current process.
*/
static void
resetpriority(struct thread *td)
{
register unsigned int newpriority;
if (td->td_pri_class == PRI_TIMESHARE) {
newpriority = PUSER + td->td_estcpu / INVERSE_ESTCPU_WEIGHT +
NICE_WEIGHT * (td->td_proc->p_nice - PRIO_MIN);
newpriority = min(max(newpriority, PRI_MIN_TIMESHARE),
PRI_MAX_TIMESHARE);
sched_user_prio(td, newpriority);
}
Rework the interface between priority propagation (lending) and the schedulers a bit to ensure more correct handling of priorities and fewer priority inversions: - Add two functions to the sched(9) API to handle priority lending: sched_lend_prio() and sched_unlend_prio(). The turnstile code uses these functions to ask the scheduler to lend a thread a set priority and to tell the scheduler when it thinks it is ok for a thread to stop borrowing priority. The unlend case is slightly complex in that the turnstile code tells the scheduler what the minimum priority of the thread needs to be to satisfy the requirements of any other threads blocked on locks owned by the thread in question. The scheduler then decides where the thread can go back to normal mode (if it's normal priority is high enough to satisfy the pending lock requests) or it it should continue to use the priority specified to the sched_unlend_prio() call. This involves adding a new per-thread flag TDF_BORROWING that replaces the ULE-only kse flag for priority elevation. - Schedulers now refuse to lower the priority of a thread that is currently borrowing another therad's priority. - If a scheduler changes the priority of a thread that is currently sitting on a turnstile, it will call a new function turnstile_adjust() to inform the turnstile code of the change. This function resorts the thread on the priority list of the turnstile if needed, and if the thread ends up at the head of the list (due to having the highest priority) and its priority was raised, then it will propagate that new priority to the owner of the lock it is blocked on. Some additional fixes specific to the 4BSD scheduler include: - Common code for updating the priority of a thread when the user priority of its associated kse group has been consolidated in a new static function resetpriority_thread(). One change to this function is that it will now only adjust the priority of a thread if it already has a time sharing priority, thus preserving any boosts from a tsleep() until the thread returns to userland. Also, resetpriority() no longer calls maybe_resched() on each thread in the group. Instead, the code calling resetpriority() is responsible for calling resetpriority_thread() on any threads that need to be updated. - schedcpu() now uses resetpriority_thread() instead of just calling sched_prio() directly after it updates a kse group's user priority. - sched_clock() now uses resetpriority_thread() rather than writing directly to td_priority. - sched_nice() now updates all the priorities of the threads after the group priority has been adjusted. Discussed with: bde Reviewed by: ups, jeffr Tested on: 4bsd, ule Tested on: i386, alpha, sparc64
2004-12-30 20:52:44 +00:00
}
/*
* Update the thread's priority when the associated process's user
Rework the interface between priority propagation (lending) and the schedulers a bit to ensure more correct handling of priorities and fewer priority inversions: - Add two functions to the sched(9) API to handle priority lending: sched_lend_prio() and sched_unlend_prio(). The turnstile code uses these functions to ask the scheduler to lend a thread a set priority and to tell the scheduler when it thinks it is ok for a thread to stop borrowing priority. The unlend case is slightly complex in that the turnstile code tells the scheduler what the minimum priority of the thread needs to be to satisfy the requirements of any other threads blocked on locks owned by the thread in question. The scheduler then decides where the thread can go back to normal mode (if it's normal priority is high enough to satisfy the pending lock requests) or it it should continue to use the priority specified to the sched_unlend_prio() call. This involves adding a new per-thread flag TDF_BORROWING that replaces the ULE-only kse flag for priority elevation. - Schedulers now refuse to lower the priority of a thread that is currently borrowing another therad's priority. - If a scheduler changes the priority of a thread that is currently sitting on a turnstile, it will call a new function turnstile_adjust() to inform the turnstile code of the change. This function resorts the thread on the priority list of the turnstile if needed, and if the thread ends up at the head of the list (due to having the highest priority) and its priority was raised, then it will propagate that new priority to the owner of the lock it is blocked on. Some additional fixes specific to the 4BSD scheduler include: - Common code for updating the priority of a thread when the user priority of its associated kse group has been consolidated in a new static function resetpriority_thread(). One change to this function is that it will now only adjust the priority of a thread if it already has a time sharing priority, thus preserving any boosts from a tsleep() until the thread returns to userland. Also, resetpriority() no longer calls maybe_resched() on each thread in the group. Instead, the code calling resetpriority() is responsible for calling resetpriority_thread() on any threads that need to be updated. - schedcpu() now uses resetpriority_thread() instead of just calling sched_prio() directly after it updates a kse group's user priority. - sched_clock() now uses resetpriority_thread() rather than writing directly to td_priority. - sched_nice() now updates all the priorities of the threads after the group priority has been adjusted. Discussed with: bde Reviewed by: ups, jeffr Tested on: 4bsd, ule Tested on: i386, alpha, sparc64
2004-12-30 20:52:44 +00:00
* priority changes.
*/
static void
resetpriority_thread(struct thread *td)
Rework the interface between priority propagation (lending) and the schedulers a bit to ensure more correct handling of priorities and fewer priority inversions: - Add two functions to the sched(9) API to handle priority lending: sched_lend_prio() and sched_unlend_prio(). The turnstile code uses these functions to ask the scheduler to lend a thread a set priority and to tell the scheduler when it thinks it is ok for a thread to stop borrowing priority. The unlend case is slightly complex in that the turnstile code tells the scheduler what the minimum priority of the thread needs to be to satisfy the requirements of any other threads blocked on locks owned by the thread in question. The scheduler then decides where the thread can go back to normal mode (if it's normal priority is high enough to satisfy the pending lock requests) or it it should continue to use the priority specified to the sched_unlend_prio() call. This involves adding a new per-thread flag TDF_BORROWING that replaces the ULE-only kse flag for priority elevation. - Schedulers now refuse to lower the priority of a thread that is currently borrowing another therad's priority. - If a scheduler changes the priority of a thread that is currently sitting on a turnstile, it will call a new function turnstile_adjust() to inform the turnstile code of the change. This function resorts the thread on the priority list of the turnstile if needed, and if the thread ends up at the head of the list (due to having the highest priority) and its priority was raised, then it will propagate that new priority to the owner of the lock it is blocked on. Some additional fixes specific to the 4BSD scheduler include: - Common code for updating the priority of a thread when the user priority of its associated kse group has been consolidated in a new static function resetpriority_thread(). One change to this function is that it will now only adjust the priority of a thread if it already has a time sharing priority, thus preserving any boosts from a tsleep() until the thread returns to userland. Also, resetpriority() no longer calls maybe_resched() on each thread in the group. Instead, the code calling resetpriority() is responsible for calling resetpriority_thread() on any threads that need to be updated. - schedcpu() now uses resetpriority_thread() instead of just calling sched_prio() directly after it updates a kse group's user priority. - sched_clock() now uses resetpriority_thread() rather than writing directly to td_priority. - sched_nice() now updates all the priorities of the threads after the group priority has been adjusted. Discussed with: bde Reviewed by: ups, jeffr Tested on: 4bsd, ule Tested on: i386, alpha, sparc64
2004-12-30 20:52:44 +00:00
{
/* Only change threads with a time sharing user priority. */
if (td->td_priority < PRI_MIN_TIMESHARE ||
td->td_priority > PRI_MAX_TIMESHARE)
return;
/* XXX the whole needresched thing is broken, but not silly. */
maybe_resched(td);
sched_prio(td, td->td_user_pri);
}
/* ARGSUSED */
static void
sched_setup(void *dummy)
{
setup_runqs();
if (sched_quantum == 0)
sched_quantum = SCHED_QUANTUM;
hogticks = 2 * sched_quantum;
/* Account for thread0. */
sched_load_add();
}
/* External interfaces start here */
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
/*
* Very early in the boot some setup of scheduler-specific
* parts of proc0 and of some scheduler resources needs to be done.
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
* Called from:
* proc0_init()
*/
void
schedinit(void)
{
/*
* Set up the scheduler specific parts of proc0.
*/
proc0.p_sched = NULL; /* XXX */
thread0.td_sched = &td_sched0;
thread0.td_lock = &sched_lock;
mtx_init(&sched_lock, "sched lock", NULL, MTX_SPIN | MTX_RECURSE);
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
}
int
sched_runnable(void)
{
#ifdef SMP
return runq_check(&runq) + runq_check(&runq_pcpu[PCPU_GET(cpuid)]);
#else
return runq_check(&runq);
#endif
}
int
sched_rr_interval(void)
{
if (sched_quantum == 0)
sched_quantum = SCHED_QUANTUM;
return (sched_quantum);
}
/*
* We adjust the priority of the current process. The priority of
* a process gets worse as it accumulates CPU time. The cpu usage
* estimator (td_estcpu) is increased here. resetpriority() will
* compute a different priority each time td_estcpu increases by
* INVERSE_ESTCPU_WEIGHT
* (until MAXPRI is reached). The cpu usage estimator ramps up
* quite quickly when the process is running (linearly), and decays
* away exponentially, at a rate which is proportionally slower when
* the system is busy. The basic principle is that the system will
* 90% forget that the process used a lot of CPU time in 5 * loadav
* seconds. This causes the system to favor processes which haven't
* run much recently, and to round-robin among other processes.
*/
void
sched_clock(struct thread *td)
{
struct pcpuidlestat *stat;
struct td_sched *ts;
2003-04-11 03:39:48 +00:00
THREAD_LOCK_ASSERT(td, MA_OWNED);
ts = td->td_sched;
ts->ts_cpticks++;
td->td_estcpu = ESTCPULIM(td->td_estcpu + 1);
if ((td->td_estcpu % INVERSE_ESTCPU_WEIGHT) == 0) {
resetpriority(td);
resetpriority_thread(td);
}
/*
* Force a context switch if the current thread has used up a full
* quantum (default quantum is 100ms).
*/
if (!TD_IS_IDLETHREAD(td) &&
ticks - PCPU_GET(switchticks) >= sched_quantum)
td->td_flags |= TDF_NEEDRESCHED;
stat = DPCPU_PTR(idlestat);
stat->oldidlecalls = stat->idlecalls;
stat->idlecalls = 0;
}
/*
* Charge child's scheduling CPU usage to parent.
*/
void
sched_exit(struct proc *p, struct thread *td)
2003-04-11 03:39:48 +00:00
{
KTR_STATE1(KTR_SCHED, "thread", sched_tdname(td), "proc exit",
"prio:td", td->td_priority);
PROC_LOCK_ASSERT(p, MA_OWNED);
sched_exit_thread(FIRST_THREAD_IN_PROC(p), td);
}
void
2003-04-11 03:39:48 +00:00
sched_exit_thread(struct thread *td, struct thread *child)
{
KTR_STATE1(KTR_SCHED, "thread", sched_tdname(child), "exit",
"prio:td", child->td_priority);
thread_lock(td);
td->td_estcpu = ESTCPULIM(td->td_estcpu + child->td_estcpu);
thread_unlock(td);
thread_lock(child);
if ((child->td_flags & TDF_NOLOAD) == 0)
sched_load_rem();
thread_unlock(child);
2003-04-11 03:39:48 +00:00
}
2003-04-11 03:39:48 +00:00
void
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
sched_fork(struct thread *td, struct thread *childtd)
2003-04-11 03:39:48 +00:00
{
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
sched_fork_thread(td, childtd);
2003-04-11 03:39:48 +00:00
}
2003-04-11 03:39:48 +00:00
void
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
sched_fork_thread(struct thread *td, struct thread *childtd)
2003-04-11 03:39:48 +00:00
{
struct td_sched *ts;
childtd->td_estcpu = td->td_estcpu;
childtd->td_lock = &sched_lock;
childtd->td_cpuset = cpuset_ref(td->td_cpuset);
childtd->td_priority = childtd->td_base_pri;
ts = childtd->td_sched;
bzero(ts, sizeof(*ts));
ts->ts_flags |= (td->td_sched->ts_flags & TSF_AFFINITY);
}
void
sched_nice(struct proc *p, int nice)
{
Rework the interface between priority propagation (lending) and the schedulers a bit to ensure more correct handling of priorities and fewer priority inversions: - Add two functions to the sched(9) API to handle priority lending: sched_lend_prio() and sched_unlend_prio(). The turnstile code uses these functions to ask the scheduler to lend a thread a set priority and to tell the scheduler when it thinks it is ok for a thread to stop borrowing priority. The unlend case is slightly complex in that the turnstile code tells the scheduler what the minimum priority of the thread needs to be to satisfy the requirements of any other threads blocked on locks owned by the thread in question. The scheduler then decides where the thread can go back to normal mode (if it's normal priority is high enough to satisfy the pending lock requests) or it it should continue to use the priority specified to the sched_unlend_prio() call. This involves adding a new per-thread flag TDF_BORROWING that replaces the ULE-only kse flag for priority elevation. - Schedulers now refuse to lower the priority of a thread that is currently borrowing another therad's priority. - If a scheduler changes the priority of a thread that is currently sitting on a turnstile, it will call a new function turnstile_adjust() to inform the turnstile code of the change. This function resorts the thread on the priority list of the turnstile if needed, and if the thread ends up at the head of the list (due to having the highest priority) and its priority was raised, then it will propagate that new priority to the owner of the lock it is blocked on. Some additional fixes specific to the 4BSD scheduler include: - Common code for updating the priority of a thread when the user priority of its associated kse group has been consolidated in a new static function resetpriority_thread(). One change to this function is that it will now only adjust the priority of a thread if it already has a time sharing priority, thus preserving any boosts from a tsleep() until the thread returns to userland. Also, resetpriority() no longer calls maybe_resched() on each thread in the group. Instead, the code calling resetpriority() is responsible for calling resetpriority_thread() on any threads that need to be updated. - schedcpu() now uses resetpriority_thread() instead of just calling sched_prio() directly after it updates a kse group's user priority. - sched_clock() now uses resetpriority_thread() rather than writing directly to td_priority. - sched_nice() now updates all the priorities of the threads after the group priority has been adjusted. Discussed with: bde Reviewed by: ups, jeffr Tested on: 4bsd, ule Tested on: i386, alpha, sparc64
2004-12-30 20:52:44 +00:00
struct thread *td;
PROC_LOCK_ASSERT(p, MA_OWNED);
p->p_nice = nice;
FOREACH_THREAD_IN_PROC(p, td) {
thread_lock(td);
resetpriority(td);
resetpriority_thread(td);
thread_unlock(td);
}
}
2003-04-11 03:39:48 +00:00
void
sched_class(struct thread *td, int class)
2003-04-11 03:39:48 +00:00
{
THREAD_LOCK_ASSERT(td, MA_OWNED);
td->td_pri_class = class;
2003-04-11 03:39:48 +00:00
}
/*
* Adjust the priority of a thread.
*/
Rework the interface between priority propagation (lending) and the schedulers a bit to ensure more correct handling of priorities and fewer priority inversions: - Add two functions to the sched(9) API to handle priority lending: sched_lend_prio() and sched_unlend_prio(). The turnstile code uses these functions to ask the scheduler to lend a thread a set priority and to tell the scheduler when it thinks it is ok for a thread to stop borrowing priority. The unlend case is slightly complex in that the turnstile code tells the scheduler what the minimum priority of the thread needs to be to satisfy the requirements of any other threads blocked on locks owned by the thread in question. The scheduler then decides where the thread can go back to normal mode (if it's normal priority is high enough to satisfy the pending lock requests) or it it should continue to use the priority specified to the sched_unlend_prio() call. This involves adding a new per-thread flag TDF_BORROWING that replaces the ULE-only kse flag for priority elevation. - Schedulers now refuse to lower the priority of a thread that is currently borrowing another therad's priority. - If a scheduler changes the priority of a thread that is currently sitting on a turnstile, it will call a new function turnstile_adjust() to inform the turnstile code of the change. This function resorts the thread on the priority list of the turnstile if needed, and if the thread ends up at the head of the list (due to having the highest priority) and its priority was raised, then it will propagate that new priority to the owner of the lock it is blocked on. Some additional fixes specific to the 4BSD scheduler include: - Common code for updating the priority of a thread when the user priority of its associated kse group has been consolidated in a new static function resetpriority_thread(). One change to this function is that it will now only adjust the priority of a thread if it already has a time sharing priority, thus preserving any boosts from a tsleep() until the thread returns to userland. Also, resetpriority() no longer calls maybe_resched() on each thread in the group. Instead, the code calling resetpriority() is responsible for calling resetpriority_thread() on any threads that need to be updated. - schedcpu() now uses resetpriority_thread() instead of just calling sched_prio() directly after it updates a kse group's user priority. - sched_clock() now uses resetpriority_thread() rather than writing directly to td_priority. - sched_nice() now updates all the priorities of the threads after the group priority has been adjusted. Discussed with: bde Reviewed by: ups, jeffr Tested on: 4bsd, ule Tested on: i386, alpha, sparc64
2004-12-30 20:52:44 +00:00
static void
sched_priority(struct thread *td, u_char prio)
{
KTR_POINT3(KTR_SCHED, "thread", sched_tdname(td), "priority change",
"prio:%d", td->td_priority, "new prio:%d", prio, KTR_ATTR_LINKED,
sched_tdname(curthread));
if (td != curthread && prio > td->td_priority) {
KTR_POINT3(KTR_SCHED, "thread", sched_tdname(curthread),
"lend prio", "prio:%d", td->td_priority, "new prio:%d",
prio, KTR_ATTR_LINKED, sched_tdname(td));
}
THREAD_LOCK_ASSERT(td, MA_OWNED);
Rework the interface between priority propagation (lending) and the schedulers a bit to ensure more correct handling of priorities and fewer priority inversions: - Add two functions to the sched(9) API to handle priority lending: sched_lend_prio() and sched_unlend_prio(). The turnstile code uses these functions to ask the scheduler to lend a thread a set priority and to tell the scheduler when it thinks it is ok for a thread to stop borrowing priority. The unlend case is slightly complex in that the turnstile code tells the scheduler what the minimum priority of the thread needs to be to satisfy the requirements of any other threads blocked on locks owned by the thread in question. The scheduler then decides where the thread can go back to normal mode (if it's normal priority is high enough to satisfy the pending lock requests) or it it should continue to use the priority specified to the sched_unlend_prio() call. This involves adding a new per-thread flag TDF_BORROWING that replaces the ULE-only kse flag for priority elevation. - Schedulers now refuse to lower the priority of a thread that is currently borrowing another therad's priority. - If a scheduler changes the priority of a thread that is currently sitting on a turnstile, it will call a new function turnstile_adjust() to inform the turnstile code of the change. This function resorts the thread on the priority list of the turnstile if needed, and if the thread ends up at the head of the list (due to having the highest priority) and its priority was raised, then it will propagate that new priority to the owner of the lock it is blocked on. Some additional fixes specific to the 4BSD scheduler include: - Common code for updating the priority of a thread when the user priority of its associated kse group has been consolidated in a new static function resetpriority_thread(). One change to this function is that it will now only adjust the priority of a thread if it already has a time sharing priority, thus preserving any boosts from a tsleep() until the thread returns to userland. Also, resetpriority() no longer calls maybe_resched() on each thread in the group. Instead, the code calling resetpriority() is responsible for calling resetpriority_thread() on any threads that need to be updated. - schedcpu() now uses resetpriority_thread() instead of just calling sched_prio() directly after it updates a kse group's user priority. - sched_clock() now uses resetpriority_thread() rather than writing directly to td_priority. - sched_nice() now updates all the priorities of the threads after the group priority has been adjusted. Discussed with: bde Reviewed by: ups, jeffr Tested on: 4bsd, ule Tested on: i386, alpha, sparc64
2004-12-30 20:52:44 +00:00
if (td->td_priority == prio)
return;
td->td_priority = prio;
if (TD_ON_RUNQ(td) && td->td_rqindex != (prio / RQ_PPQ)) {
sched_rem(td);
sched_add(td, SRQ_BORING);
}
}
Rework the interface between priority propagation (lending) and the schedulers a bit to ensure more correct handling of priorities and fewer priority inversions: - Add two functions to the sched(9) API to handle priority lending: sched_lend_prio() and sched_unlend_prio(). The turnstile code uses these functions to ask the scheduler to lend a thread a set priority and to tell the scheduler when it thinks it is ok for a thread to stop borrowing priority. The unlend case is slightly complex in that the turnstile code tells the scheduler what the minimum priority of the thread needs to be to satisfy the requirements of any other threads blocked on locks owned by the thread in question. The scheduler then decides where the thread can go back to normal mode (if it's normal priority is high enough to satisfy the pending lock requests) or it it should continue to use the priority specified to the sched_unlend_prio() call. This involves adding a new per-thread flag TDF_BORROWING that replaces the ULE-only kse flag for priority elevation. - Schedulers now refuse to lower the priority of a thread that is currently borrowing another therad's priority. - If a scheduler changes the priority of a thread that is currently sitting on a turnstile, it will call a new function turnstile_adjust() to inform the turnstile code of the change. This function resorts the thread on the priority list of the turnstile if needed, and if the thread ends up at the head of the list (due to having the highest priority) and its priority was raised, then it will propagate that new priority to the owner of the lock it is blocked on. Some additional fixes specific to the 4BSD scheduler include: - Common code for updating the priority of a thread when the user priority of its associated kse group has been consolidated in a new static function resetpriority_thread(). One change to this function is that it will now only adjust the priority of a thread if it already has a time sharing priority, thus preserving any boosts from a tsleep() until the thread returns to userland. Also, resetpriority() no longer calls maybe_resched() on each thread in the group. Instead, the code calling resetpriority() is responsible for calling resetpriority_thread() on any threads that need to be updated. - schedcpu() now uses resetpriority_thread() instead of just calling sched_prio() directly after it updates a kse group's user priority. - sched_clock() now uses resetpriority_thread() rather than writing directly to td_priority. - sched_nice() now updates all the priorities of the threads after the group priority has been adjusted. Discussed with: bde Reviewed by: ups, jeffr Tested on: 4bsd, ule Tested on: i386, alpha, sparc64
2004-12-30 20:52:44 +00:00
/*
* Update a thread's priority when it is lent another thread's
* priority.
*/
void
sched_lend_prio(struct thread *td, u_char prio)
{
td->td_flags |= TDF_BORROWING;
sched_priority(td, prio);
}
/*
* Restore a thread's priority when priority propagation is
* over. The prio argument is the minimum priority the thread
* needs to have to satisfy other possible priority lending
* requests. If the thread's regulary priority is less
* important than prio the thread will keep a priority boost
* of prio.
*/
void
sched_unlend_prio(struct thread *td, u_char prio)
{
u_char base_pri;
if (td->td_base_pri >= PRI_MIN_TIMESHARE &&
td->td_base_pri <= PRI_MAX_TIMESHARE)
base_pri = td->td_user_pri;
Rework the interface between priority propagation (lending) and the schedulers a bit to ensure more correct handling of priorities and fewer priority inversions: - Add two functions to the sched(9) API to handle priority lending: sched_lend_prio() and sched_unlend_prio(). The turnstile code uses these functions to ask the scheduler to lend a thread a set priority and to tell the scheduler when it thinks it is ok for a thread to stop borrowing priority. The unlend case is slightly complex in that the turnstile code tells the scheduler what the minimum priority of the thread needs to be to satisfy the requirements of any other threads blocked on locks owned by the thread in question. The scheduler then decides where the thread can go back to normal mode (if it's normal priority is high enough to satisfy the pending lock requests) or it it should continue to use the priority specified to the sched_unlend_prio() call. This involves adding a new per-thread flag TDF_BORROWING that replaces the ULE-only kse flag for priority elevation. - Schedulers now refuse to lower the priority of a thread that is currently borrowing another therad's priority. - If a scheduler changes the priority of a thread that is currently sitting on a turnstile, it will call a new function turnstile_adjust() to inform the turnstile code of the change. This function resorts the thread on the priority list of the turnstile if needed, and if the thread ends up at the head of the list (due to having the highest priority) and its priority was raised, then it will propagate that new priority to the owner of the lock it is blocked on. Some additional fixes specific to the 4BSD scheduler include: - Common code for updating the priority of a thread when the user priority of its associated kse group has been consolidated in a new static function resetpriority_thread(). One change to this function is that it will now only adjust the priority of a thread if it already has a time sharing priority, thus preserving any boosts from a tsleep() until the thread returns to userland. Also, resetpriority() no longer calls maybe_resched() on each thread in the group. Instead, the code calling resetpriority() is responsible for calling resetpriority_thread() on any threads that need to be updated. - schedcpu() now uses resetpriority_thread() instead of just calling sched_prio() directly after it updates a kse group's user priority. - sched_clock() now uses resetpriority_thread() rather than writing directly to td_priority. - sched_nice() now updates all the priorities of the threads after the group priority has been adjusted. Discussed with: bde Reviewed by: ups, jeffr Tested on: 4bsd, ule Tested on: i386, alpha, sparc64
2004-12-30 20:52:44 +00:00
else
base_pri = td->td_base_pri;
if (prio >= base_pri) {
td->td_flags &= ~TDF_BORROWING;
sched_prio(td, base_pri);
} else
sched_lend_prio(td, prio);
}
void
sched_prio(struct thread *td, u_char prio)
{
u_char oldprio;
/* First, update the base priority. */
td->td_base_pri = prio;
/*
* If the thread is borrowing another thread's priority, don't ever
* lower the priority.
*/
if (td->td_flags & TDF_BORROWING && td->td_priority < prio)
return;
/* Change the real priority. */
oldprio = td->td_priority;
sched_priority(td, prio);
/*
* If the thread is on a turnstile, then let the turnstile update
* its state.
*/
if (TD_ON_LOCK(td) && oldprio != prio)
turnstile_adjust(td, oldprio);
}
void
sched_user_prio(struct thread *td, u_char prio)
{
THREAD_LOCK_ASSERT(td, MA_OWNED);
td->td_base_user_pri = prio;
if (td->td_lend_user_pri <= prio)
return;
td->td_user_pri = prio;
}
void
sched_lend_user_prio(struct thread *td, u_char prio)
{
THREAD_LOCK_ASSERT(td, MA_OWNED);
td->td_lend_user_pri = prio;
td->td_user_pri = min(prio, td->td_base_user_pri);
if (td->td_priority > td->td_user_pri)
sched_prio(td, td->td_user_pri);
else if (td->td_priority != td->td_user_pri)
td->td_flags |= TDF_NEEDRESCHED;
}
void
sched_sleep(struct thread *td, int pri)
{
THREAD_LOCK_ASSERT(td, MA_OWNED);
td->td_slptick = ticks;
td->td_sched->ts_slptime = 0;
if (pri != 0 && PRI_BASE(td->td_pri_class) == PRI_TIMESHARE)
sched_prio(td, pri);
if (TD_IS_SUSPENDED(td) || pri >= PSOCK)
td->td_flags |= TDF_CANSWAP;
}
void
sched_switch(struct thread *td, struct thread *newtd, int flags)
{
struct mtx *tmtx;
struct td_sched *ts;
struct proc *p;
tmtx = NULL;
ts = td->td_sched;
p = td->td_proc;
THREAD_LOCK_ASSERT(td, MA_OWNED);
/*
* Switch to the sched lock to fix things up and pick
* a new thread.
* Block the td_lock in order to avoid breaking the critical path.
*/
if (td->td_lock != &sched_lock) {
mtx_lock_spin(&sched_lock);
tmtx = thread_lock_block(td);
}
if ((td->td_flags & TDF_NOLOAD) == 0)
sched_load_rem();
td->td_lastcpu = td->td_oncpu;
if (!(flags & SW_PREEMPT))
td->td_flags &= ~TDF_NEEDRESCHED;
td->td_owepreempt = 0;
td->td_oncpu = NOCPU;
/*
* 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. We never put the idle
* threads on the run queue, however.
*/
if (td->td_flags & TDF_IDLETD) {
TD_SET_CAN_RUN(td);
#ifdef SMP
Commit the support for removing cpumask_t and replacing it directly with cpuset_t objects. That is going to offer the underlying support for a simple bump of MAXCPU and then support for number of cpus > 32 (as it is today). Right now, cpumask_t is an int, 32 bits on all our supported architecture. cpumask_t on the other side is implemented as an array of longs, and easilly extendible by definition. The architectures touched by this commit are the following: - amd64 - i386 - pc98 - arm - ia64 - XEN while the others are still missing. Userland is believed to be fully converted with the changes contained here. Some technical notes: - This commit may be considered an ABI nop for all the architectures different from amd64 and ia64 (and sparc64 in the future) - per-cpu members, which are now converted to cpuset_t, needs to be accessed avoiding migration, because the size of cpuset_t should be considered unknown - size of cpuset_t objects is different from kernel and userland (this is primirally done in order to leave some more space in userland to cope with KBI extensions). If you need to access kernel cpuset_t from the userland please refer to example in this patch on how to do that correctly (kgdb may be a good source, for example). - Support for other architectures is going to be added soon - Only MAXCPU for amd64 is bumped now The patch has been tested by sbruno and Nicholas Esborn on opteron 4 x 12 pack CPUs. More testing on big SMP is expected to came soon. pluknet tested the patch with his 8-ways on both amd64 and i386. Tested by: pluknet, sbruno, gianni, Nicholas Esborn Reviewed by: jeff, jhb, sbruno
2011-05-05 14:39:14 +00:00
/* Spinlock held here, assume no migration. */
CPU_NAND(&idle_cpus_mask, PCPU_PTR(cpumask));
#endif
} else {
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
if (TD_IS_RUNNING(td)) {
/* Put us back on the run queue. */
sched_add(td, (flags & SW_PREEMPT) ?
SRQ_OURSELF|SRQ_YIELDING|SRQ_PREEMPTED :
SRQ_OURSELF|SRQ_YIELDING);
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
}
}
if (newtd) {
/*
* The thread we are about to run needs to be counted
* as if it had been added to the run queue and selected.
* It came from:
* * A preemption
* * An upcall
* * A followon
*/
KASSERT((newtd->td_inhibitors == 0),
("trying to run inhibited thread"));
newtd->td_flags |= TDF_DIDRUN;
TD_SET_RUNNING(newtd);
if ((newtd->td_flags & TDF_NOLOAD) == 0)
sched_load_add();
} else {
newtd = choosethread();
MPASS(newtd->td_lock == &sched_lock);
}
if (td != newtd) {
#ifdef HWPMC_HOOKS
if (PMC_PROC_IS_USING_PMCS(td->td_proc))
PMC_SWITCH_CONTEXT(td, PMC_FN_CSW_OUT);
#endif
/* I feel sleepy */
lock_profile_release_lock(&sched_lock.lock_object);
#ifdef KDTRACE_HOOKS
/*
* If DTrace has set the active vtime enum to anything
* other than INACTIVE (0), then it should have set the
* function to call.
*/
if (dtrace_vtime_active)
(*dtrace_vtime_switch_func)(newtd);
#endif
cpu_switch(td, newtd, tmtx != NULL ? tmtx : td->td_lock);
lock_profile_obtain_lock_success(&sched_lock.lock_object,
0, 0, __FILE__, __LINE__);
/*
* Where am I? What year is it?
* We are in the same thread that went to sleep above,
* but any amount of time may have passed. All our context
* will still be available as will local variables.
* PCPU values however may have changed as we may have
* changed CPU so don't trust cached values of them.
* New threads will go to fork_exit() instead of here
* so if you change things here you may need to change
* things there too.
*
* If the thread above was exiting it will never wake
* up again here, so either it has saved everything it
* needed to, or the thread_wait() or wait() will
* need to reap it.
*/
#ifdef HWPMC_HOOKS
if (PMC_PROC_IS_USING_PMCS(td->td_proc))
PMC_SWITCH_CONTEXT(td, PMC_FN_CSW_IN);
#endif
}
#ifdef SMP
if (td->td_flags & TDF_IDLETD)
Commit the support for removing cpumask_t and replacing it directly with cpuset_t objects. That is going to offer the underlying support for a simple bump of MAXCPU and then support for number of cpus > 32 (as it is today). Right now, cpumask_t is an int, 32 bits on all our supported architecture. cpumask_t on the other side is implemented as an array of longs, and easilly extendible by definition. The architectures touched by this commit are the following: - amd64 - i386 - pc98 - arm - ia64 - XEN while the others are still missing. Userland is believed to be fully converted with the changes contained here. Some technical notes: - This commit may be considered an ABI nop for all the architectures different from amd64 and ia64 (and sparc64 in the future) - per-cpu members, which are now converted to cpuset_t, needs to be accessed avoiding migration, because the size of cpuset_t should be considered unknown - size of cpuset_t objects is different from kernel and userland (this is primirally done in order to leave some more space in userland to cope with KBI extensions). If you need to access kernel cpuset_t from the userland please refer to example in this patch on how to do that correctly (kgdb may be a good source, for example). - Support for other architectures is going to be added soon - Only MAXCPU for amd64 is bumped now The patch has been tested by sbruno and Nicholas Esborn on opteron 4 x 12 pack CPUs. More testing on big SMP is expected to came soon. pluknet tested the patch with his 8-ways on both amd64 and i386. Tested by: pluknet, sbruno, gianni, Nicholas Esborn Reviewed by: jeff, jhb, sbruno
2011-05-05 14:39:14 +00:00
CPU_OR(&idle_cpus_mask, PCPU_PTR(cpumask));
#endif
sched_lock.mtx_lock = (uintptr_t)td;
td->td_oncpu = PCPU_GET(cpuid);
MPASS(td->td_lock == &sched_lock);
}
void
sched_wakeup(struct thread *td)
{
struct td_sched *ts;
THREAD_LOCK_ASSERT(td, MA_OWNED);
ts = td->td_sched;
td->td_flags &= ~TDF_CANSWAP;
if (ts->ts_slptime > 1) {
updatepri(td);
resetpriority(td);
}
td->td_slptick = 0;
ts->ts_slptime = 0;
sched_add(td, SRQ_BORING);
}
2004-09-03 09:15:10 +00:00
#ifdef SMP
static int
forward_wakeup(int cpunum)
{
struct pcpu *pc;
Commit the support for removing cpumask_t and replacing it directly with cpuset_t objects. That is going to offer the underlying support for a simple bump of MAXCPU and then support for number of cpus > 32 (as it is today). Right now, cpumask_t is an int, 32 bits on all our supported architecture. cpumask_t on the other side is implemented as an array of longs, and easilly extendible by definition. The architectures touched by this commit are the following: - amd64 - i386 - pc98 - arm - ia64 - XEN while the others are still missing. Userland is believed to be fully converted with the changes contained here. Some technical notes: - This commit may be considered an ABI nop for all the architectures different from amd64 and ia64 (and sparc64 in the future) - per-cpu members, which are now converted to cpuset_t, needs to be accessed avoiding migration, because the size of cpuset_t should be considered unknown - size of cpuset_t objects is different from kernel and userland (this is primirally done in order to leave some more space in userland to cope with KBI extensions). If you need to access kernel cpuset_t from the userland please refer to example in this patch on how to do that correctly (kgdb may be a good source, for example). - Support for other architectures is going to be added soon - Only MAXCPU for amd64 is bumped now The patch has been tested by sbruno and Nicholas Esborn on opteron 4 x 12 pack CPUs. More testing on big SMP is expected to came soon. pluknet tested the patch with his 8-ways on both amd64 and i386. Tested by: pluknet, sbruno, gianni, Nicholas Esborn Reviewed by: jeff, jhb, sbruno
2011-05-05 14:39:14 +00:00
cpuset_t dontuse, id, map, map2, me;
int iscpuset;
mtx_assert(&sched_lock, MA_OWNED);
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
CTR0(KTR_RUNQ, "forward_wakeup()");
if ((!forward_wakeup_enabled) ||
(forward_wakeup_use_mask == 0 && forward_wakeup_use_loop == 0))
return (0);
if (!smp_started || cold || panicstr)
return (0);
forward_wakeups_requested++;
/*
* Check the idle mask we received against what we calculated
* before in the old version.
Commit the support for removing cpumask_t and replacing it directly with cpuset_t objects. That is going to offer the underlying support for a simple bump of MAXCPU and then support for number of cpus > 32 (as it is today). Right now, cpumask_t is an int, 32 bits on all our supported architecture. cpumask_t on the other side is implemented as an array of longs, and easilly extendible by definition. The architectures touched by this commit are the following: - amd64 - i386 - pc98 - arm - ia64 - XEN while the others are still missing. Userland is believed to be fully converted with the changes contained here. Some technical notes: - This commit may be considered an ABI nop for all the architectures different from amd64 and ia64 (and sparc64 in the future) - per-cpu members, which are now converted to cpuset_t, needs to be accessed avoiding migration, because the size of cpuset_t should be considered unknown - size of cpuset_t objects is different from kernel and userland (this is primirally done in order to leave some more space in userland to cope with KBI extensions). If you need to access kernel cpuset_t from the userland please refer to example in this patch on how to do that correctly (kgdb may be a good source, for example). - Support for other architectures is going to be added soon - Only MAXCPU for amd64 is bumped now The patch has been tested by sbruno and Nicholas Esborn on opteron 4 x 12 pack CPUs. More testing on big SMP is expected to came soon. pluknet tested the patch with his 8-ways on both amd64 and i386. Tested by: pluknet, sbruno, gianni, Nicholas Esborn Reviewed by: jeff, jhb, sbruno
2011-05-05 14:39:14 +00:00
*
* Also note that sched_lock is held now, thus no migration is
* expected.
*/
me = PCPU_GET(cpumask);
/* Don't bother if we should be doing it ourself. */
Commit the support for removing cpumask_t and replacing it directly with cpuset_t objects. That is going to offer the underlying support for a simple bump of MAXCPU and then support for number of cpus > 32 (as it is today). Right now, cpumask_t is an int, 32 bits on all our supported architecture. cpumask_t on the other side is implemented as an array of longs, and easilly extendible by definition. The architectures touched by this commit are the following: - amd64 - i386 - pc98 - arm - ia64 - XEN while the others are still missing. Userland is believed to be fully converted with the changes contained here. Some technical notes: - This commit may be considered an ABI nop for all the architectures different from amd64 and ia64 (and sparc64 in the future) - per-cpu members, which are now converted to cpuset_t, needs to be accessed avoiding migration, because the size of cpuset_t should be considered unknown - size of cpuset_t objects is different from kernel and userland (this is primirally done in order to leave some more space in userland to cope with KBI extensions). If you need to access kernel cpuset_t from the userland please refer to example in this patch on how to do that correctly (kgdb may be a good source, for example). - Support for other architectures is going to be added soon - Only MAXCPU for amd64 is bumped now The patch has been tested by sbruno and Nicholas Esborn on opteron 4 x 12 pack CPUs. More testing on big SMP is expected to came soon. pluknet tested the patch with his 8-ways on both amd64 and i386. Tested by: pluknet, sbruno, gianni, Nicholas Esborn Reviewed by: jeff, jhb, sbruno
2011-05-05 14:39:14 +00:00
if (CPU_OVERLAP(&me, &idle_cpus_mask) &&
(cpunum == NOCPU || CPU_ISSET(cpunum, &me)))
return (0);
Commit the support for removing cpumask_t and replacing it directly with cpuset_t objects. That is going to offer the underlying support for a simple bump of MAXCPU and then support for number of cpus > 32 (as it is today). Right now, cpumask_t is an int, 32 bits on all our supported architecture. cpumask_t on the other side is implemented as an array of longs, and easilly extendible by definition. The architectures touched by this commit are the following: - amd64 - i386 - pc98 - arm - ia64 - XEN while the others are still missing. Userland is believed to be fully converted with the changes contained here. Some technical notes: - This commit may be considered an ABI nop for all the architectures different from amd64 and ia64 (and sparc64 in the future) - per-cpu members, which are now converted to cpuset_t, needs to be accessed avoiding migration, because the size of cpuset_t should be considered unknown - size of cpuset_t objects is different from kernel and userland (this is primirally done in order to leave some more space in userland to cope with KBI extensions). If you need to access kernel cpuset_t from the userland please refer to example in this patch on how to do that correctly (kgdb may be a good source, for example). - Support for other architectures is going to be added soon - Only MAXCPU for amd64 is bumped now The patch has been tested by sbruno and Nicholas Esborn on opteron 4 x 12 pack CPUs. More testing on big SMP is expected to came soon. pluknet tested the patch with his 8-ways on both amd64 and i386. Tested by: pluknet, sbruno, gianni, Nicholas Esborn Reviewed by: jeff, jhb, sbruno
2011-05-05 14:39:14 +00:00
dontuse = me;
CPU_OR(&dontuse, &stopped_cpus);
CPU_OR(&dontuse, &hlt_cpus_mask);
CPU_ZERO(&map2);
if (forward_wakeup_use_loop) {
SLIST_FOREACH(pc, &cpuhead, pc_allcpu) {
id = pc->pc_cpumask;
Commit the support for removing cpumask_t and replacing it directly with cpuset_t objects. That is going to offer the underlying support for a simple bump of MAXCPU and then support for number of cpus > 32 (as it is today). Right now, cpumask_t is an int, 32 bits on all our supported architecture. cpumask_t on the other side is implemented as an array of longs, and easilly extendible by definition. The architectures touched by this commit are the following: - amd64 - i386 - pc98 - arm - ia64 - XEN while the others are still missing. Userland is believed to be fully converted with the changes contained here. Some technical notes: - This commit may be considered an ABI nop for all the architectures different from amd64 and ia64 (and sparc64 in the future) - per-cpu members, which are now converted to cpuset_t, needs to be accessed avoiding migration, because the size of cpuset_t should be considered unknown - size of cpuset_t objects is different from kernel and userland (this is primirally done in order to leave some more space in userland to cope with KBI extensions). If you need to access kernel cpuset_t from the userland please refer to example in this patch on how to do that correctly (kgdb may be a good source, for example). - Support for other architectures is going to be added soon - Only MAXCPU for amd64 is bumped now The patch has been tested by sbruno and Nicholas Esborn on opteron 4 x 12 pack CPUs. More testing on big SMP is expected to came soon. pluknet tested the patch with his 8-ways on both amd64 and i386. Tested by: pluknet, sbruno, gianni, Nicholas Esborn Reviewed by: jeff, jhb, sbruno
2011-05-05 14:39:14 +00:00
if (!CPU_OVERLAP(&id, &dontuse) &&
pc->pc_curthread == pc->pc_idlethread) {
Commit the support for removing cpumask_t and replacing it directly with cpuset_t objects. That is going to offer the underlying support for a simple bump of MAXCPU and then support for number of cpus > 32 (as it is today). Right now, cpumask_t is an int, 32 bits on all our supported architecture. cpumask_t on the other side is implemented as an array of longs, and easilly extendible by definition. The architectures touched by this commit are the following: - amd64 - i386 - pc98 - arm - ia64 - XEN while the others are still missing. Userland is believed to be fully converted with the changes contained here. Some technical notes: - This commit may be considered an ABI nop for all the architectures different from amd64 and ia64 (and sparc64 in the future) - per-cpu members, which are now converted to cpuset_t, needs to be accessed avoiding migration, because the size of cpuset_t should be considered unknown - size of cpuset_t objects is different from kernel and userland (this is primirally done in order to leave some more space in userland to cope with KBI extensions). If you need to access kernel cpuset_t from the userland please refer to example in this patch on how to do that correctly (kgdb may be a good source, for example). - Support for other architectures is going to be added soon - Only MAXCPU for amd64 is bumped now The patch has been tested by sbruno and Nicholas Esborn on opteron 4 x 12 pack CPUs. More testing on big SMP is expected to came soon. pluknet tested the patch with his 8-ways on both amd64 and i386. Tested by: pluknet, sbruno, gianni, Nicholas Esborn Reviewed by: jeff, jhb, sbruno
2011-05-05 14:39:14 +00:00
CPU_OR(&map2, &id);
}
}
}
if (forward_wakeup_use_mask) {
Commit the support for removing cpumask_t and replacing it directly with cpuset_t objects. That is going to offer the underlying support for a simple bump of MAXCPU and then support for number of cpus > 32 (as it is today). Right now, cpumask_t is an int, 32 bits on all our supported architecture. cpumask_t on the other side is implemented as an array of longs, and easilly extendible by definition. The architectures touched by this commit are the following: - amd64 - i386 - pc98 - arm - ia64 - XEN while the others are still missing. Userland is believed to be fully converted with the changes contained here. Some technical notes: - This commit may be considered an ABI nop for all the architectures different from amd64 and ia64 (and sparc64 in the future) - per-cpu members, which are now converted to cpuset_t, needs to be accessed avoiding migration, because the size of cpuset_t should be considered unknown - size of cpuset_t objects is different from kernel and userland (this is primirally done in order to leave some more space in userland to cope with KBI extensions). If you need to access kernel cpuset_t from the userland please refer to example in this patch on how to do that correctly (kgdb may be a good source, for example). - Support for other architectures is going to be added soon - Only MAXCPU for amd64 is bumped now The patch has been tested by sbruno and Nicholas Esborn on opteron 4 x 12 pack CPUs. More testing on big SMP is expected to came soon. pluknet tested the patch with his 8-ways on both amd64 and i386. Tested by: pluknet, sbruno, gianni, Nicholas Esborn Reviewed by: jeff, jhb, sbruno
2011-05-05 14:39:14 +00:00
map = idle_cpus_mask;
CPU_NAND(&map, &dontuse);
/* If they are both on, compare and use loop if different. */
if (forward_wakeup_use_loop) {
Commit the support for removing cpumask_t and replacing it directly with cpuset_t objects. That is going to offer the underlying support for a simple bump of MAXCPU and then support for number of cpus > 32 (as it is today). Right now, cpumask_t is an int, 32 bits on all our supported architecture. cpumask_t on the other side is implemented as an array of longs, and easilly extendible by definition. The architectures touched by this commit are the following: - amd64 - i386 - pc98 - arm - ia64 - XEN while the others are still missing. Userland is believed to be fully converted with the changes contained here. Some technical notes: - This commit may be considered an ABI nop for all the architectures different from amd64 and ia64 (and sparc64 in the future) - per-cpu members, which are now converted to cpuset_t, needs to be accessed avoiding migration, because the size of cpuset_t should be considered unknown - size of cpuset_t objects is different from kernel and userland (this is primirally done in order to leave some more space in userland to cope with KBI extensions). If you need to access kernel cpuset_t from the userland please refer to example in this patch on how to do that correctly (kgdb may be a good source, for example). - Support for other architectures is going to be added soon - Only MAXCPU for amd64 is bumped now The patch has been tested by sbruno and Nicholas Esborn on opteron 4 x 12 pack CPUs. More testing on big SMP is expected to came soon. pluknet tested the patch with his 8-ways on both amd64 and i386. Tested by: pluknet, sbruno, gianni, Nicholas Esborn Reviewed by: jeff, jhb, sbruno
2011-05-05 14:39:14 +00:00
if (CPU_CMP(&map, &map2)) {
printf("map != map2, loop method preferred\n");
map = map2;
}
}
} else {
map = map2;
}
/* If we only allow a specific CPU, then mask off all the others. */
if (cpunum != NOCPU) {
KASSERT((cpunum <= mp_maxcpus),("forward_wakeup: bad cpunum."));
Commit the support for removing cpumask_t and replacing it directly with cpuset_t objects. That is going to offer the underlying support for a simple bump of MAXCPU and then support for number of cpus > 32 (as it is today). Right now, cpumask_t is an int, 32 bits on all our supported architecture. cpumask_t on the other side is implemented as an array of longs, and easilly extendible by definition. The architectures touched by this commit are the following: - amd64 - i386 - pc98 - arm - ia64 - XEN while the others are still missing. Userland is believed to be fully converted with the changes contained here. Some technical notes: - This commit may be considered an ABI nop for all the architectures different from amd64 and ia64 (and sparc64 in the future) - per-cpu members, which are now converted to cpuset_t, needs to be accessed avoiding migration, because the size of cpuset_t should be considered unknown - size of cpuset_t objects is different from kernel and userland (this is primirally done in order to leave some more space in userland to cope with KBI extensions). If you need to access kernel cpuset_t from the userland please refer to example in this patch on how to do that correctly (kgdb may be a good source, for example). - Support for other architectures is going to be added soon - Only MAXCPU for amd64 is bumped now The patch has been tested by sbruno and Nicholas Esborn on opteron 4 x 12 pack CPUs. More testing on big SMP is expected to came soon. pluknet tested the patch with his 8-ways on both amd64 and i386. Tested by: pluknet, sbruno, gianni, Nicholas Esborn Reviewed by: jeff, jhb, sbruno
2011-05-05 14:39:14 +00:00
iscpuset = CPU_ISSET(cpunum, &map);
if (iscpuset == 0)
CPU_ZERO(&map);
else
CPU_SETOF(cpunum, &map);
}
Commit the support for removing cpumask_t and replacing it directly with cpuset_t objects. That is going to offer the underlying support for a simple bump of MAXCPU and then support for number of cpus > 32 (as it is today). Right now, cpumask_t is an int, 32 bits on all our supported architecture. cpumask_t on the other side is implemented as an array of longs, and easilly extendible by definition. The architectures touched by this commit are the following: - amd64 - i386 - pc98 - arm - ia64 - XEN while the others are still missing. Userland is believed to be fully converted with the changes contained here. Some technical notes: - This commit may be considered an ABI nop for all the architectures different from amd64 and ia64 (and sparc64 in the future) - per-cpu members, which are now converted to cpuset_t, needs to be accessed avoiding migration, because the size of cpuset_t should be considered unknown - size of cpuset_t objects is different from kernel and userland (this is primirally done in order to leave some more space in userland to cope with KBI extensions). If you need to access kernel cpuset_t from the userland please refer to example in this patch on how to do that correctly (kgdb may be a good source, for example). - Support for other architectures is going to be added soon - Only MAXCPU for amd64 is bumped now The patch has been tested by sbruno and Nicholas Esborn on opteron 4 x 12 pack CPUs. More testing on big SMP is expected to came soon. pluknet tested the patch with his 8-ways on both amd64 and i386. Tested by: pluknet, sbruno, gianni, Nicholas Esborn Reviewed by: jeff, jhb, sbruno
2011-05-05 14:39:14 +00:00
if (!CPU_EMPTY(&map)) {
forward_wakeups_delivered++;
SLIST_FOREACH(pc, &cpuhead, pc_allcpu) {
id = pc->pc_cpumask;
Commit the support for removing cpumask_t and replacing it directly with cpuset_t objects. That is going to offer the underlying support for a simple bump of MAXCPU and then support for number of cpus > 32 (as it is today). Right now, cpumask_t is an int, 32 bits on all our supported architecture. cpumask_t on the other side is implemented as an array of longs, and easilly extendible by definition. The architectures touched by this commit are the following: - amd64 - i386 - pc98 - arm - ia64 - XEN while the others are still missing. Userland is believed to be fully converted with the changes contained here. Some technical notes: - This commit may be considered an ABI nop for all the architectures different from amd64 and ia64 (and sparc64 in the future) - per-cpu members, which are now converted to cpuset_t, needs to be accessed avoiding migration, because the size of cpuset_t should be considered unknown - size of cpuset_t objects is different from kernel and userland (this is primirally done in order to leave some more space in userland to cope with KBI extensions). If you need to access kernel cpuset_t from the userland please refer to example in this patch on how to do that correctly (kgdb may be a good source, for example). - Support for other architectures is going to be added soon - Only MAXCPU for amd64 is bumped now The patch has been tested by sbruno and Nicholas Esborn on opteron 4 x 12 pack CPUs. More testing on big SMP is expected to came soon. pluknet tested the patch with his 8-ways on both amd64 and i386. Tested by: pluknet, sbruno, gianni, Nicholas Esborn Reviewed by: jeff, jhb, sbruno
2011-05-05 14:39:14 +00:00
if (!CPU_OVERLAP(&map, &id))
continue;
if (cpu_idle_wakeup(pc->pc_cpuid))
Commit the support for removing cpumask_t and replacing it directly with cpuset_t objects. That is going to offer the underlying support for a simple bump of MAXCPU and then support for number of cpus > 32 (as it is today). Right now, cpumask_t is an int, 32 bits on all our supported architecture. cpumask_t on the other side is implemented as an array of longs, and easilly extendible by definition. The architectures touched by this commit are the following: - amd64 - i386 - pc98 - arm - ia64 - XEN while the others are still missing. Userland is believed to be fully converted with the changes contained here. Some technical notes: - This commit may be considered an ABI nop for all the architectures different from amd64 and ia64 (and sparc64 in the future) - per-cpu members, which are now converted to cpuset_t, needs to be accessed avoiding migration, because the size of cpuset_t should be considered unknown - size of cpuset_t objects is different from kernel and userland (this is primirally done in order to leave some more space in userland to cope with KBI extensions). If you need to access kernel cpuset_t from the userland please refer to example in this patch on how to do that correctly (kgdb may be a good source, for example). - Support for other architectures is going to be added soon - Only MAXCPU for amd64 is bumped now The patch has been tested by sbruno and Nicholas Esborn on opteron 4 x 12 pack CPUs. More testing on big SMP is expected to came soon. pluknet tested the patch with his 8-ways on both amd64 and i386. Tested by: pluknet, sbruno, gianni, Nicholas Esborn Reviewed by: jeff, jhb, sbruno
2011-05-05 14:39:14 +00:00
CPU_NAND(&map, &id);
}
Commit the support for removing cpumask_t and replacing it directly with cpuset_t objects. That is going to offer the underlying support for a simple bump of MAXCPU and then support for number of cpus > 32 (as it is today). Right now, cpumask_t is an int, 32 bits on all our supported architecture. cpumask_t on the other side is implemented as an array of longs, and easilly extendible by definition. The architectures touched by this commit are the following: - amd64 - i386 - pc98 - arm - ia64 - XEN while the others are still missing. Userland is believed to be fully converted with the changes contained here. Some technical notes: - This commit may be considered an ABI nop for all the architectures different from amd64 and ia64 (and sparc64 in the future) - per-cpu members, which are now converted to cpuset_t, needs to be accessed avoiding migration, because the size of cpuset_t should be considered unknown - size of cpuset_t objects is different from kernel and userland (this is primirally done in order to leave some more space in userland to cope with KBI extensions). If you need to access kernel cpuset_t from the userland please refer to example in this patch on how to do that correctly (kgdb may be a good source, for example). - Support for other architectures is going to be added soon - Only MAXCPU for amd64 is bumped now The patch has been tested by sbruno and Nicholas Esborn on opteron 4 x 12 pack CPUs. More testing on big SMP is expected to came soon. pluknet tested the patch with his 8-ways on both amd64 and i386. Tested by: pluknet, sbruno, gianni, Nicholas Esborn Reviewed by: jeff, jhb, sbruno
2011-05-05 14:39:14 +00:00
if (!CPU_EMPTY(&map))
ipi_selected(map, IPI_AST);
return (1);
}
if (cpunum == NOCPU)
printf("forward_wakeup: Idle processor not found\n");
return (0);
}
static void
kick_other_cpu(int pri, int cpuid)
{
struct pcpu *pcpu;
int cpri;
pcpu = pcpu_find(cpuid);
Commit the support for removing cpumask_t and replacing it directly with cpuset_t objects. That is going to offer the underlying support for a simple bump of MAXCPU and then support for number of cpus > 32 (as it is today). Right now, cpumask_t is an int, 32 bits on all our supported architecture. cpumask_t on the other side is implemented as an array of longs, and easilly extendible by definition. The architectures touched by this commit are the following: - amd64 - i386 - pc98 - arm - ia64 - XEN while the others are still missing. Userland is believed to be fully converted with the changes contained here. Some technical notes: - This commit may be considered an ABI nop for all the architectures different from amd64 and ia64 (and sparc64 in the future) - per-cpu members, which are now converted to cpuset_t, needs to be accessed avoiding migration, because the size of cpuset_t should be considered unknown - size of cpuset_t objects is different from kernel and userland (this is primirally done in order to leave some more space in userland to cope with KBI extensions). If you need to access kernel cpuset_t from the userland please refer to example in this patch on how to do that correctly (kgdb may be a good source, for example). - Support for other architectures is going to be added soon - Only MAXCPU for amd64 is bumped now The patch has been tested by sbruno and Nicholas Esborn on opteron 4 x 12 pack CPUs. More testing on big SMP is expected to came soon. pluknet tested the patch with his 8-ways on both amd64 and i386. Tested by: pluknet, sbruno, gianni, Nicholas Esborn Reviewed by: jeff, jhb, sbruno
2011-05-05 14:39:14 +00:00
if (CPU_OVERLAP(&idle_cpus_mask, &pcpu->pc_cpumask)) {
forward_wakeups_delivered++;
if (!cpu_idle_wakeup(cpuid))
ipi_cpu(cpuid, IPI_AST);
return;
}
cpri = pcpu->pc_curthread->td_priority;
if (pri >= cpri)
return;
#if defined(IPI_PREEMPTION) && defined(PREEMPTION)
#if !defined(FULL_PREEMPTION)
if (pri <= PRI_MAX_ITHD)
#endif /* ! FULL_PREEMPTION */
{
ipi_cpu(cpuid, IPI_PREEMPT);
return;
}
#endif /* defined(IPI_PREEMPTION) && defined(PREEMPTION) */
pcpu->pc_curthread->td_flags |= TDF_NEEDRESCHED;
ipi_cpu(cpuid, IPI_AST);
return;
}
#endif /* SMP */
#ifdef SMP
static int
sched_pickcpu(struct thread *td)
{
int best, cpu;
mtx_assert(&sched_lock, MA_OWNED);
if (THREAD_CAN_SCHED(td, td->td_lastcpu))
best = td->td_lastcpu;
else
best = NOCPU;
CPU_FOREACH(cpu) {
if (!THREAD_CAN_SCHED(td, cpu))
continue;
if (best == NOCPU)
best = cpu;
else if (runq_length[cpu] < runq_length[best])
best = cpu;
}
KASSERT(best != NOCPU, ("no valid CPUs"));
return (best);
}
#endif
void
sched_add(struct thread *td, int flags)
#ifdef SMP
{
Commit the support for removing cpumask_t and replacing it directly with cpuset_t objects. That is going to offer the underlying support for a simple bump of MAXCPU and then support for number of cpus > 32 (as it is today). Right now, cpumask_t is an int, 32 bits on all our supported architecture. cpumask_t on the other side is implemented as an array of longs, and easilly extendible by definition. The architectures touched by this commit are the following: - amd64 - i386 - pc98 - arm - ia64 - XEN while the others are still missing. Userland is believed to be fully converted with the changes contained here. Some technical notes: - This commit may be considered an ABI nop for all the architectures different from amd64 and ia64 (and sparc64 in the future) - per-cpu members, which are now converted to cpuset_t, needs to be accessed avoiding migration, because the size of cpuset_t should be considered unknown - size of cpuset_t objects is different from kernel and userland (this is primirally done in order to leave some more space in userland to cope with KBI extensions). If you need to access kernel cpuset_t from the userland please refer to example in this patch on how to do that correctly (kgdb may be a good source, for example). - Support for other architectures is going to be added soon - Only MAXCPU for amd64 is bumped now The patch has been tested by sbruno and Nicholas Esborn on opteron 4 x 12 pack CPUs. More testing on big SMP is expected to came soon. pluknet tested the patch with his 8-ways on both amd64 and i386. Tested by: pluknet, sbruno, gianni, Nicholas Esborn Reviewed by: jeff, jhb, sbruno
2011-05-05 14:39:14 +00:00
cpuset_t idle, me, tidlemsk;
struct td_sched *ts;
int forwarded = 0;
int cpu;
int single_cpu = 0;
ts = td->td_sched;
THREAD_LOCK_ASSERT(td, MA_OWNED);
KASSERT((td->td_inhibitors == 0),
("sched_add: trying to run inhibited thread"));
KASSERT((TD_CAN_RUN(td) || TD_IS_RUNNING(td)),
("sched_add: bad thread state"));
KASSERT(td->td_flags & TDF_INMEM,
("sched_add: thread swapped out"));
KTR_STATE2(KTR_SCHED, "thread", sched_tdname(td), "runq add",
"prio:%d", td->td_priority, KTR_ATTR_LINKED,
sched_tdname(curthread));
KTR_POINT1(KTR_SCHED, "thread", sched_tdname(curthread), "wokeup",
KTR_ATTR_LINKED, sched_tdname(td));
/*
* Now that the thread is moving to the run-queue, set the lock
* to the scheduler's lock.
*/
if (td->td_lock != &sched_lock) {
mtx_lock_spin(&sched_lock);
thread_lock_set(td, &sched_lock);
}
TD_SET_RUNQ(td);
/*
* If SMP is started and the thread is pinned or otherwise limited to
* a specific set of CPUs, queue the thread to a per-CPU run queue.
* Otherwise, queue the thread to the global run queue.
*
* If SMP has not yet been started we must use the global run queue
* as per-CPU state may not be initialized yet and we may crash if we
* try to access the per-CPU run queues.
*/
if (smp_started && (td->td_pinned != 0 || td->td_flags & TDF_BOUND ||
ts->ts_flags & TSF_AFFINITY)) {
if (td->td_pinned != 0)
cpu = td->td_lastcpu;
else if (td->td_flags & TDF_BOUND) {
/* Find CPU from bound runq. */
KASSERT(SKE_RUNQ_PCPU(ts),
("sched_add: bound td_sched not on cpu runq"));
cpu = ts->ts_runq - &runq_pcpu[0];
} else
/* Find a valid CPU for our cpuset */
cpu = sched_pickcpu(td);
ts->ts_runq = &runq_pcpu[cpu];
single_cpu = 1;
CTR3(KTR_RUNQ,
"sched_add: Put td_sched:%p(td:%p) on cpu%d runq", ts, td,
cpu);
} else {
CTR2(KTR_RUNQ,
"sched_add: adding td_sched:%p (td:%p) to gbl runq", ts,
td);
cpu = NOCPU;
ts->ts_runq = &runq;
}
if (single_cpu && (cpu != PCPU_GET(cpuid))) {
kick_other_cpu(td->td_priority, cpu);
} else {
if (!single_cpu) {
Commit the support for removing cpumask_t and replacing it directly with cpuset_t objects. That is going to offer the underlying support for a simple bump of MAXCPU and then support for number of cpus > 32 (as it is today). Right now, cpumask_t is an int, 32 bits on all our supported architecture. cpumask_t on the other side is implemented as an array of longs, and easilly extendible by definition. The architectures touched by this commit are the following: - amd64 - i386 - pc98 - arm - ia64 - XEN while the others are still missing. Userland is believed to be fully converted with the changes contained here. Some technical notes: - This commit may be considered an ABI nop for all the architectures different from amd64 and ia64 (and sparc64 in the future) - per-cpu members, which are now converted to cpuset_t, needs to be accessed avoiding migration, because the size of cpuset_t should be considered unknown - size of cpuset_t objects is different from kernel and userland (this is primirally done in order to leave some more space in userland to cope with KBI extensions). If you need to access kernel cpuset_t from the userland please refer to example in this patch on how to do that correctly (kgdb may be a good source, for example). - Support for other architectures is going to be added soon - Only MAXCPU for amd64 is bumped now The patch has been tested by sbruno and Nicholas Esborn on opteron 4 x 12 pack CPUs. More testing on big SMP is expected to came soon. pluknet tested the patch with his 8-ways on both amd64 and i386. Tested by: pluknet, sbruno, gianni, Nicholas Esborn Reviewed by: jeff, jhb, sbruno
2011-05-05 14:39:14 +00:00
/*
* Thread spinlock is held here, assume no
* migration is possible.
*/
me = PCPU_GET(cpumask);
idle = idle_cpus_mask;
tidlemsk = idle;
CPU_AND(&idle, &me);
CPU_OR(&me, &hlt_cpus_mask);
CPU_NAND(&tidlemsk, &me);
if (CPU_EMPTY(&idle) && ((flags & SRQ_INTR) == 0) &&
!CPU_EMPTY(&tidlemsk))
forwarded = forward_wakeup(cpu);
}
if (!forwarded) {
if ((flags & SRQ_YIELDING) == 0 && maybe_preempt(td))
return;
else
maybe_resched(td);
}
}
if ((td->td_flags & TDF_NOLOAD) == 0)
sched_load_add();
runq_add(ts->ts_runq, td, flags);
if (cpu != NOCPU)
runq_length[cpu]++;
}
#else /* SMP */
{
struct td_sched *ts;
ts = td->td_sched;
THREAD_LOCK_ASSERT(td, MA_OWNED);
KASSERT((td->td_inhibitors == 0),
("sched_add: trying to run inhibited thread"));
KASSERT((TD_CAN_RUN(td) || TD_IS_RUNNING(td)),
("sched_add: bad thread state"));
KASSERT(td->td_flags & TDF_INMEM,
("sched_add: thread swapped out"));
KTR_STATE2(KTR_SCHED, "thread", sched_tdname(td), "runq add",
"prio:%d", td->td_priority, KTR_ATTR_LINKED,
sched_tdname(curthread));
KTR_POINT1(KTR_SCHED, "thread", sched_tdname(curthread), "wokeup",
KTR_ATTR_LINKED, sched_tdname(td));
/*
* Now that the thread is moving to the run-queue, set the lock
* to the scheduler's lock.
*/
if (td->td_lock != &sched_lock) {
mtx_lock_spin(&sched_lock);
thread_lock_set(td, &sched_lock);
}
TD_SET_RUNQ(td);
CTR2(KTR_RUNQ, "sched_add: adding td_sched:%p (td:%p) to runq", ts, td);
ts->ts_runq = &runq;
/*
* If we are yielding (on the way out anyhow) or the thread
* being saved is US, then don't try be smart about preemption
* or kicking off another CPU as it won't help and may hinder.
* In the YIEDLING case, we are about to run whoever is being
* put in the queue anyhow, and in the OURSELF case, we are
* puting ourself on the run queue which also only happens
* when we are about to yield.
*/
if ((flags & SRQ_YIELDING) == 0) {
if (maybe_preempt(td))
return;
}
if ((td->td_flags & TDF_NOLOAD) == 0)
sched_load_add();
runq_add(ts->ts_runq, td, flags);
maybe_resched(td);
}
#endif /* SMP */
void
sched_rem(struct thread *td)
{
struct td_sched *ts;
ts = td->td_sched;
KASSERT(td->td_flags & TDF_INMEM,
("sched_rem: thread swapped out"));
KASSERT(TD_ON_RUNQ(td),
("sched_rem: thread not on run queue"));
mtx_assert(&sched_lock, MA_OWNED);
KTR_STATE2(KTR_SCHED, "thread", sched_tdname(td), "runq rem",
"prio:%d", td->td_priority, KTR_ATTR_LINKED,
sched_tdname(curthread));
if ((td->td_flags & TDF_NOLOAD) == 0)
sched_load_rem();
#ifdef SMP
if (ts->ts_runq != &runq)
runq_length[ts->ts_runq - runq_pcpu]--;
#endif
runq_remove(ts->ts_runq, td);
TD_SET_CAN_RUN(td);
}
/*
* Select threads to run. Note that running threads still consume a
* slot.
*/
struct thread *
sched_choose(void)
{
struct thread *td;
struct runq *rq;
mtx_assert(&sched_lock, MA_OWNED);
#ifdef SMP
struct thread *tdcpu;
rq = &runq;
td = runq_choose_fuzz(&runq, runq_fuzz);
tdcpu = runq_choose(&runq_pcpu[PCPU_GET(cpuid)]);
if (td == NULL ||
(tdcpu != NULL &&
tdcpu->td_priority < td->td_priority)) {
CTR2(KTR_RUNQ, "choosing td %p from pcpu runq %d", tdcpu,
PCPU_GET(cpuid));
td = tdcpu;
rq = &runq_pcpu[PCPU_GET(cpuid)];
} else {
CTR1(KTR_RUNQ, "choosing td_sched %p from main runq", td);
}
#else
rq = &runq;
td = runq_choose(&runq);
#endif
if (td) {
#ifdef SMP
if (td == tdcpu)
runq_length[PCPU_GET(cpuid)]--;
#endif
runq_remove(rq, td);
td->td_flags |= TDF_DIDRUN;
KASSERT(td->td_flags & TDF_INMEM,
("sched_choose: thread swapped out"));
return (td);
}
return (PCPU_GET(idlethread));
}
void
sched_preempt(struct thread *td)
{
thread_lock(td);
if (td->td_critnest > 1)
td->td_owepreempt = 1;
else
mi_switch(SW_INVOL | SW_PREEMPT | SWT_PREEMPT, NULL);
thread_unlock(td);
}
void
sched_userret(struct thread *td)
{
/*
* XXX we cheat slightly on the locking here to avoid locking in
* the usual case. Setting td_priority here is essentially an
* incomplete workaround for not setting it properly elsewhere.
* Now that some interrupt handlers are threads, not setting it
* properly elsewhere can clobber it in the window between setting
* it here and returning to user mode, so don't waste time setting
* it perfectly here.
*/
Rework the interface between priority propagation (lending) and the schedulers a bit to ensure more correct handling of priorities and fewer priority inversions: - Add two functions to the sched(9) API to handle priority lending: sched_lend_prio() and sched_unlend_prio(). The turnstile code uses these functions to ask the scheduler to lend a thread a set priority and to tell the scheduler when it thinks it is ok for a thread to stop borrowing priority. The unlend case is slightly complex in that the turnstile code tells the scheduler what the minimum priority of the thread needs to be to satisfy the requirements of any other threads blocked on locks owned by the thread in question. The scheduler then decides where the thread can go back to normal mode (if it's normal priority is high enough to satisfy the pending lock requests) or it it should continue to use the priority specified to the sched_unlend_prio() call. This involves adding a new per-thread flag TDF_BORROWING that replaces the ULE-only kse flag for priority elevation. - Schedulers now refuse to lower the priority of a thread that is currently borrowing another therad's priority. - If a scheduler changes the priority of a thread that is currently sitting on a turnstile, it will call a new function turnstile_adjust() to inform the turnstile code of the change. This function resorts the thread on the priority list of the turnstile if needed, and if the thread ends up at the head of the list (due to having the highest priority) and its priority was raised, then it will propagate that new priority to the owner of the lock it is blocked on. Some additional fixes specific to the 4BSD scheduler include: - Common code for updating the priority of a thread when the user priority of its associated kse group has been consolidated in a new static function resetpriority_thread(). One change to this function is that it will now only adjust the priority of a thread if it already has a time sharing priority, thus preserving any boosts from a tsleep() until the thread returns to userland. Also, resetpriority() no longer calls maybe_resched() on each thread in the group. Instead, the code calling resetpriority() is responsible for calling resetpriority_thread() on any threads that need to be updated. - schedcpu() now uses resetpriority_thread() instead of just calling sched_prio() directly after it updates a kse group's user priority. - sched_clock() now uses resetpriority_thread() rather than writing directly to td_priority. - sched_nice() now updates all the priorities of the threads after the group priority has been adjusted. Discussed with: bde Reviewed by: ups, jeffr Tested on: 4bsd, ule Tested on: i386, alpha, sparc64
2004-12-30 20:52:44 +00:00
KASSERT((td->td_flags & TDF_BORROWING) == 0,
("thread with borrowed priority returning to userland"));
if (td->td_priority != td->td_user_pri) {
thread_lock(td);
td->td_priority = td->td_user_pri;
td->td_base_pri = td->td_user_pri;
thread_unlock(td);
}
}
void
sched_bind(struct thread *td, int cpu)
{
struct td_sched *ts;
THREAD_LOCK_ASSERT(td, MA_OWNED|MA_NOTRECURSED);
KASSERT(td == curthread, ("sched_bind: can only bind curthread"));
ts = td->td_sched;
td->td_flags |= TDF_BOUND;
#ifdef SMP
ts->ts_runq = &runq_pcpu[cpu];
if (PCPU_GET(cpuid) == cpu)
return;
mi_switch(SW_VOL, NULL);
#endif
}
void
sched_unbind(struct thread* td)
{
THREAD_LOCK_ASSERT(td, MA_OWNED);
KASSERT(td == curthread, ("sched_unbind: can only bind curthread"));
td->td_flags &= ~TDF_BOUND;
}
int
sched_is_bound(struct thread *td)
{
THREAD_LOCK_ASSERT(td, MA_OWNED);
return (td->td_flags & TDF_BOUND);
}
void
sched_relinquish(struct thread *td)
{
thread_lock(td);
mi_switch(SW_VOL | SWT_RELINQUISH, NULL);
thread_unlock(td);
}
int
sched_load(void)
{
return (sched_tdcnt);
}
int
sched_sizeof_proc(void)
{
return (sizeof(struct proc));
}
int
sched_sizeof_thread(void)
{
return (sizeof(struct thread) + sizeof(struct td_sched));
}
fixpt_t
sched_pctcpu(struct thread *td)
{
struct td_sched *ts;
THREAD_LOCK_ASSERT(td, MA_OWNED);
ts = td->td_sched;
return (ts->ts_pctcpu);
}
Add scheduler CORE, the work I have done half a year ago, recent, I picked it up again. The scheduler is forked from ULE, but the algorithm to detect an interactive process is almost completely different with ULE, it comes from Linux paper "Understanding the Linux 2.6.8.1 CPU Scheduler", although I still use same word "score" as a priority boost in ULE scheduler. Briefly, the scheduler has following characteristic: 1. Timesharing process's nice value is seriously respected, timeslice and interaction detecting algorithm are based on nice value. 2. per-cpu scheduling queue and load balancing. 3. O(1) scheduling. 4. Some cpu affinity code in wakeup path. 5. Support POSIX SCHED_FIFO and SCHED_RR. Unlike scheduler 4BSD and ULE which using fuzzy RQ_PPQ, the scheduler uses 256 priority queues. Unlike ULE which using pull and push, the scheduelr uses pull method, the main reason is to let relative idle cpu do the work, but current the whole scheduler is protected by the big sched_lock, so the benefit is not visible, it really can be worse than nothing because all other cpu are locked out when we are doing balancing work, which the 4BSD scheduelr does not have this problem. The scheduler does not support hyperthreading very well, in fact, the scheduler does not make the difference between physical CPU and logical CPU, this should be improved in feature. The scheduler has priority inversion problem on MP machine, it is not good for realtime scheduling, it can cause realtime process starving. As a result, it seems the MySQL super-smack runs better on my Pentium-D machine when using libthr, despite on UP or SMP kernel.
2006-06-13 13:12:56 +00:00
void
Refactor timer management code with priority to one-shot operation mode. The main goal of this is to generate timer interrupts only when there is some work to do. When CPU is busy interrupts are generating at full rate of hz + stathz to fullfill scheduler and timekeeping requirements. But when CPU is idle, only minimum set of interrupts (down to 8 interrupts per second per CPU now), needed to handle scheduled callouts is executed. This allows significantly increase idle CPU sleep time, increasing effect of static power-saving technologies. Also it should reduce host CPU load on virtualized systems, when guest system is idle. There is set of tunables, also available as writable sysctls, allowing to control wanted event timer subsystem behavior: kern.eventtimer.timer - allows to choose event timer hardware to use. On x86 there is up to 4 different kinds of timers. Depending on whether chosen timer is per-CPU, behavior of other options slightly differs. kern.eventtimer.periodic - allows to choose periodic and one-shot operation mode. In periodic mode, current timer hardware taken as the only source of time for time events. This mode is quite alike to previous kernel behavior. One-shot mode instead uses currently selected time counter hardware to schedule all needed events one by one and program timer to generate interrupt exactly in specified time. Default value depends of chosen timer capabilities, but one-shot mode is preferred, until other is forced by user or hardware. kern.eventtimer.singlemul - in periodic mode specifies how much times higher timer frequency should be, to not strictly alias hardclock() and statclock() events. Default values are 2 and 4, but could be reduced to 1 if extra interrupts are unwanted. kern.eventtimer.idletick - makes each CPU to receive every timer interrupt independently of whether they busy or not. By default this options is disabled. If chosen timer is per-CPU and runs in periodic mode, this option has no effect - all interrupts are generating. As soon as this patch modifies cpu_idle() on some platforms, I have also refactored one on x86. Now it makes use of MONITOR/MWAIT instrunctions (if supported) under high sleep/wakeup rate, as fast alternative to other methods. It allows SMP scheduler to wake up sleeping CPUs much faster without using IPI, significantly increasing performance on some highly task-switching loads. Tested by: many (on i386, amd64, sparc64 and powerc) H/W donated by: Gheorghe Ardelean Sponsored by: iXsystems, Inc.
2010-09-13 07:25:35 +00:00
sched_tick(int cnt)
Add scheduler CORE, the work I have done half a year ago, recent, I picked it up again. The scheduler is forked from ULE, but the algorithm to detect an interactive process is almost completely different with ULE, it comes from Linux paper "Understanding the Linux 2.6.8.1 CPU Scheduler", although I still use same word "score" as a priority boost in ULE scheduler. Briefly, the scheduler has following characteristic: 1. Timesharing process's nice value is seriously respected, timeslice and interaction detecting algorithm are based on nice value. 2. per-cpu scheduling queue and load balancing. 3. O(1) scheduling. 4. Some cpu affinity code in wakeup path. 5. Support POSIX SCHED_FIFO and SCHED_RR. Unlike scheduler 4BSD and ULE which using fuzzy RQ_PPQ, the scheduler uses 256 priority queues. Unlike ULE which using pull and push, the scheduelr uses pull method, the main reason is to let relative idle cpu do the work, but current the whole scheduler is protected by the big sched_lock, so the benefit is not visible, it really can be worse than nothing because all other cpu are locked out when we are doing balancing work, which the 4BSD scheduelr does not have this problem. The scheduler does not support hyperthreading very well, in fact, the scheduler does not make the difference between physical CPU and logical CPU, this should be improved in feature. The scheduler has priority inversion problem on MP machine, it is not good for realtime scheduling, it can cause realtime process starving. As a result, it seems the MySQL super-smack runs better on my Pentium-D machine when using libthr, despite on UP or SMP kernel.
2006-06-13 13:12:56 +00:00
{
}
/*
* The actual idle process.
*/
void
sched_idletd(void *dummy)
{
struct pcpuidlestat *stat;
stat = DPCPU_PTR(idlestat);
for (;;) {
mtx_assert(&Giant, MA_NOTOWNED);
while (sched_runnable() == 0) {
cpu_idle(stat->idlecalls + stat->oldidlecalls > 64);
stat->idlecalls++;
}
mtx_lock_spin(&sched_lock);
mi_switch(SW_VOL | SWT_IDLE, NULL);
mtx_unlock_spin(&sched_lock);
}
}
/*
* A CPU is entering for the first time or a thread is exiting.
*/
void
sched_throw(struct thread *td)
{
/*
* Correct spinlock nesting. The idle thread context that we are
* borrowing was created so that it would start out with a single
* spin lock (sched_lock) held in fork_trampoline(). Since we've
* explicitly acquired locks in this function, the nesting count
* is now 2 rather than 1. Since we are nested, calling
* spinlock_exit() will simply adjust the counts without allowing
* spin lock using code to interrupt us.
*/
if (td == NULL) {
mtx_lock_spin(&sched_lock);
spinlock_exit();
} else {
lock_profile_release_lock(&sched_lock.lock_object);
MPASS(td->td_lock == &sched_lock);
}
mtx_assert(&sched_lock, MA_OWNED);
KASSERT(curthread->td_md.md_spinlock_count == 1, ("invalid count"));
PCPU_SET(switchtime, cpu_ticks());
PCPU_SET(switchticks, ticks);
cpu_throw(td, choosethread()); /* doesn't return */
}
void
sched_fork_exit(struct thread *td)
{
/*
* Finish setting up thread glue so that it begins execution in a
* non-nested critical section with sched_lock held but not recursed.
*/
td->td_oncpu = PCPU_GET(cpuid);
sched_lock.mtx_lock = (uintptr_t)td;
lock_profile_obtain_lock_success(&sched_lock.lock_object,
0, 0, __FILE__, __LINE__);
THREAD_LOCK_ASSERT(td, MA_OWNED | MA_NOTRECURSED);
}
char *
sched_tdname(struct thread *td)
{
#ifdef KTR
struct td_sched *ts;
ts = td->td_sched;
if (ts->ts_name[0] == '\0')
snprintf(ts->ts_name, sizeof(ts->ts_name),
"%s tid %d", td->td_name, td->td_tid);
return (ts->ts_name);
#else
return (td->td_name);
#endif
}
void
sched_affinity(struct thread *td)
{
#ifdef SMP
struct td_sched *ts;
int cpu;
THREAD_LOCK_ASSERT(td, MA_OWNED);
/*
* Set the TSF_AFFINITY flag if there is at least one CPU this
* thread can't run on.
*/
ts = td->td_sched;
ts->ts_flags &= ~TSF_AFFINITY;
CPU_FOREACH(cpu) {
if (!THREAD_CAN_SCHED(td, cpu)) {
ts->ts_flags |= TSF_AFFINITY;
break;
}
}
/*
* If this thread can run on all CPUs, nothing else to do.
*/
if (!(ts->ts_flags & TSF_AFFINITY))
return;
/* Pinned threads and bound threads should be left alone. */
if (td->td_pinned != 0 || td->td_flags & TDF_BOUND)
return;
switch (td->td_state) {
case TDS_RUNQ:
/*
* If we are on a per-CPU runqueue that is in the set,
* then nothing needs to be done.
*/
if (ts->ts_runq != &runq &&
THREAD_CAN_SCHED(td, ts->ts_runq - runq_pcpu))
return;
/* Put this thread on a valid per-CPU runqueue. */
sched_rem(td);
sched_add(td, SRQ_BORING);
break;
case TDS_RUNNING:
/*
* See if our current CPU is in the set. If not, force a
* context switch.
*/
if (THREAD_CAN_SCHED(td, td->td_oncpu))
return;
td->td_flags |= TDF_NEEDRESCHED;
if (td != curthread)
ipi_cpu(cpu, IPI_AST);
break;
default:
break;
}
#endif
}