/*
* Copyright (c) 2001-2007 Jakub Jermar
* All rights reserved.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
*
* - Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer.
* - 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.
* - The name of the author may not be used to endorse or promote products
* derived from this software without specific prior written permission.
*
* THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
* IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
* OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
* IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT,
* INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
* NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
* DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
* THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
* (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
* THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
*/
/** @addtogroup genericproc
* @{
*/
/**
* @file
* @brief Scheduler and load balancing.
*
* This file contains the scheduler and kcpulb kernel thread which
* performs load-balancing of per-CPU run queues.
*/
#include <proc/scheduler.h>
#include <proc/thread.h>
#include <proc/task.h>
#include <mm/frame.h>
#include <mm/page.h>
#include <mm/as.h>
#include <time/timeout.h>
#include <time/delay.h>
#include <arch/asm.h>
#include <arch/faddr.h>
#include <arch/cycle.h>
#include <atomic.h>
#include <synch/spinlock.h>
#include <config.h>
#include <context.h>
#include <fpu_context.h>
#include <func.h>
#include <arch.h>
#include <adt/list.h>
#include <panic.h>
#include <cpu.h>
#include <print.h>
#include <debug.h>
#include <arch/smp/sun4v/smp.h>
static void before_task_runs(void);
static void before_thread_runs(void);
static void after_thread_ran(void);
static void scheduler_separated_stack(void);
atomic_t nrdy; /**< Number of ready threads in the system. */
/** Carry out actions before new task runs. */
void before_task_runs(void)
{
before_task_runs_arch();
}
/** Take actions before new thread runs.
*
* Perform actions that need to be
* taken before the newly selected
* tread is passed control.
*
* THREAD->lock is locked on entry
*
*/
void before_thread_runs(void)
{
before_thread_runs_arch();
#ifdef CONFIG_FPU_LAZY
if(THREAD == CPU->fpu_owner)
fpu_enable();
else
fpu_disable();
#else
fpu_enable();
if (THREAD->fpu_context_exists)
fpu_context_restore(THREAD->saved_fpu_context);
else {
fpu_init();
THREAD->fpu_context_exists = 1;
}
#endif
}
/** Take actions after THREAD had run.
*
* Perform actions that need to be
* taken after the running thread
* had been preempted by the scheduler.
*
* THREAD->lock is locked on entry
*
*/
void after_thread_ran(void)
{
after_thread_ran_arch();
}
#ifdef CONFIG_FPU_LAZY
void scheduler_fpu_lazy_request(void)
{
restart:
fpu_enable();
spinlock_lock(&CPU->lock);
/* Save old context */
if (CPU->fpu_owner != NULL) {
spinlock_lock(&CPU->fpu_owner->lock);
fpu_context_save(CPU->fpu_owner->saved_fpu_context);
/* don't prevent migration */
CPU->fpu_owner->fpu_context_engaged = 0;
spinlock_unlock(&CPU->fpu_owner->lock);
CPU->fpu_owner = NULL;
}
spinlock_lock(&THREAD->lock);
if (THREAD->fpu_context_exists) {
fpu_context_restore(THREAD->saved_fpu_context);
} else {
/* Allocate FPU context */
if (!THREAD->saved_fpu_context) {
/* Might sleep */
spinlock_unlock(&THREAD->lock);
spinlock_unlock(&CPU->lock);
THREAD->saved_fpu_context =
(fpu_context_t *) slab_alloc(fpu_context_slab, 0);
/* We may have switched CPUs during slab_alloc */
goto restart;
}
fpu_init();
THREAD->fpu_context_exists = 1;
}
CPU->fpu_owner = THREAD;
THREAD->fpu_context_engaged = 1;
spinlock_unlock(&THREAD->lock);
spinlock_unlock(&CPU->lock);
}
#endif
/** Initialize scheduler
*
* Initialize kernel scheduler.
*
*/
void scheduler_init(void)
{
}
/** Get thread to be scheduled
*
* Get the optimal thread to be scheduled
* according to thread accounting and scheduler
* policy.
*
* @return Thread to be scheduled.
*
*/
static thread_t *find_best_thread(void)
{
thread_t *t;
runq_t *r;
int i;
ASSERT(CPU != NULL);
loop:
interrupts_enable();
if (atomic_get(&CPU->nrdy) == 0) {
/*
* For there was nothing to run, the CPU goes to sleep
* until a hardware interrupt or an IPI comes.
* This improves energy saving and hyperthreading.
*/
/*
* An interrupt might occur right now and wake up a thread.
* In such case, the CPU will continue to go to sleep
* even though there is a runnable thread.
*/
cpu_sleep();
goto loop;
}
interrupts_disable();
for (i = 0; i < RQ_COUNT; i++) {
r = &CPU->rq[i];
spinlock_lock(&r->lock);
if (r->n == 0) {
/*
* If this queue is empty, try a lower-priority queue.
*/
spinlock_unlock(&r->lock);
continue;
}
atomic_dec(&CPU->nrdy);
atomic_dec(&nrdy);
if (CPU->arch.exec_unit)
atomic_dec(&(CPU->arch.exec_unit->nrdy));
r->n--;
/*
* Take the first thread from the queue.
*/
t = list_get_instance(r->rq_head.next, thread_t, rq_link);
list_remove(&t->rq_link);
spinlock_unlock(&r->lock);
spinlock_lock(&t->lock);
t->cpu = CPU;
t->ticks = us2ticks((i + 1) * 10000);
t->priority = i; /* correct rq index */
/*
* Clear the THREAD_FLAG_STOLEN flag so that t can be migrated
* when load balancing needs emerge.
*/
t->flags &= ~THREAD_FLAG_STOLEN;
spinlock_unlock(&t->lock);
return t;
}
goto loop;
}
/** Prevent rq starvation
*
* Prevent low priority threads from starving in rq's.
*
* When the function decides to relink rq's, it reconnects
* respective pointers so that in result threads with 'pri'
* greater or equal start are moved to a higher-priority queue.
*
* @param start Threshold priority.
*
*/
static void relink_rq(int start)
{
link_t head;
runq_t *r;
int i, n;
list_initialize(&head);
spinlock_lock(&CPU->lock);
if (CPU->needs_relink > NEEDS_RELINK_MAX) {
for (i = start; i < RQ_COUNT - 1; i++) {
/* remember and empty rq[i + 1] */
r = &CPU->rq[i + 1];
spinlock_lock(&r->lock);
list_concat(&head, &r->rq_head);
n = r->n;
r->n = 0;
spinlock_unlock(&r->lock);
/* append rq[i + 1] to rq[i] */
r = &CPU->rq[i];
spinlock_lock(&r->lock);
list_concat(&r->rq_head, &head);
r->n += n;
spinlock_unlock(&r->lock);
}
CPU->needs_relink = 0;
}
spinlock_unlock(&CPU->lock);
}
/** The scheduler
*
* The thread scheduling procedure.
* Passes control directly to
* scheduler_separated_stack().
*
*/
void scheduler(void)
{
volatile ipl_t ipl;
ASSERT(CPU != NULL);
ipl = interrupts_disable();
if (atomic_get(&haltstate))
halt();
if (THREAD) {
spinlock_lock(&THREAD->lock);
/* Update thread accounting */
THREAD->cycles += get_cycle() - THREAD->last_cycle;
#ifndef CONFIG_FPU_LAZY
fpu_context_save(THREAD->saved_fpu_context);
#endif
if (!context_save(&THREAD->saved_context)) {
/*
* This is the place where threads leave scheduler();
*/
/* Save current CPU cycle */
THREAD->last_cycle = get_cycle();
spinlock_unlock(&THREAD->lock);
interrupts_restore(THREAD->saved_context.ipl);
return;
}
/*
* Interrupt priority level of preempted thread is recorded
* here to facilitate scheduler() invocations from
* interrupts_disable()'d code (e.g. waitq_sleep_timeout()).
*/
THREAD->saved_context.ipl = ipl;
}
/*
* Through the 'THE' structure, we keep track of THREAD, TASK, CPU, VM
* and preemption counter. At this point THE could be coming either
* from THREAD's or CPU's stack.
*/
the_copy(THE, (the_t *) CPU->stack);
/*
* We may not keep the old stack.
* Reason: If we kept the old stack and got blocked, for instance, in
* find_best_thread(), the old thread could get rescheduled by another
* CPU and overwrite the part of its own stack that was also used by
* the scheduler on this CPU.
*
* Moreover, we have to bypass the compiler-generated POP sequence
* which is fooled by SP being set to the very top of the stack.
* Therefore the scheduler() function continues in
* scheduler_separated_stack().
*/
context_save(&CPU->saved_context);
context_set(&CPU->saved_context, FADDR(scheduler_separated_stack),
(uintptr_t) CPU->stack, CPU_STACK_SIZE);
context_restore(&CPU->saved_context);
/* not reached */
}
/** Scheduler stack switch wrapper
*
* Second part of the scheduler() function
* using new stack. Handling the actual context
* switch to a new thread.
*
* Assume THREAD->lock is held.
*/
void scheduler_separated_stack(void)
{
int priority;
DEADLOCK_PROBE_INIT(p_joinwq);
ASSERT(CPU != NULL);
if (THREAD) {
/* must be run after the switch to scheduler stack */
after_thread_ran();
switch (THREAD->state) {
case Running:
spinlock_unlock(&THREAD->lock);
thread_ready(THREAD);
break;
case Exiting:
repeat:
if (THREAD->detached) {
thread_destroy(THREAD);
} else {
/*
* The thread structure is kept allocated until
* somebody calls thread_detach() on it.
*/
if (!spinlock_trylock(&THREAD->join_wq.lock)) {
/*
* Avoid deadlock.
*/
spinlock_unlock(&THREAD->lock);
delay(HZ);
spinlock_lock(&THREAD->lock);
DEADLOCK_PROBE(p_joinwq,
DEADLOCK_THRESHOLD);
goto repeat;
}
_waitq_wakeup_unsafe(&THREAD->join_wq,
WAKEUP_FIRST);
spinlock_unlock(&THREAD->join_wq.lock);
THREAD->state = Lingering;
spinlock_unlock(&THREAD->lock);
}
break;
case Sleeping:
/*
* Prefer the thread after it's woken up.
*/
THREAD->priority = -1;
/*
* We need to release wq->lock which we locked in
* waitq_sleep(). Address of wq->lock is kept in
* THREAD->sleep_queue.
*/
spinlock_unlock(&THREAD->sleep_queue->lock);
/*
* Check for possible requests for out-of-context
* invocation.
*/
if (THREAD->call_me) {
THREAD->call_me(THREAD->call_me_with);
THREAD->call_me = NULL;
THREAD->call_me_with = NULL;
}
spinlock_unlock(&THREAD->lock);
break;
default:
/*
* Entering state is unexpected.
*/
panic("tid%" PRIu64 ": unexpected state %s\n",
THREAD->tid, thread_states[THREAD->state]);
break;
}
THREAD = NULL;
}
THREAD = find_best_thread();
spinlock_lock(&THREAD->lock);
priority = THREAD->priority;
spinlock_unlock(&THREAD->lock);
relink_rq(priority);
/*
* If both the old and the new task are the same, lots of work is
* avoided.
*/
if (TASK != THREAD->task) {
as_t *as1 = NULL;
as_t *as2;
if (TASK) {
spinlock_lock(&TASK->lock);
as1 = TASK->as;
spinlock_unlock(&TASK->lock);
}
spinlock_lock(&THREAD->task->lock);
as2 = THREAD->task->as;
spinlock_unlock(&THREAD->task->lock);
/*
* Note that it is possible for two tasks to share one address
* space.
*/
if (as1 != as2) {
/*
* Both tasks and address spaces are different.
* Replace the old one with the new one.
*/
as_switch(as1, as2);
}
TASK = THREAD->task;
before_task_runs();
}
spinlock_lock(&THREAD->lock);
THREAD->state = Running;
#ifdef SCHEDULER_VERBOSE
printf("cpu%u: tid %" PRIu64
" (priority=%d, ticks=%" PRIu64
", nrdy=%ld)\n", CPU->id, THREAD->tid, THREAD->priority,
THREAD->ticks, atomic_get(&CPU->nrdy));
#endif
/*
* Some architectures provide late kernel PA2KA(identity)
* mapping in a page fault handler. However, the page fault
* handler uses the kernel stack of the running thread and
* therefore cannot be used to map it. The kernel stack, if
* necessary, is to be mapped in before_thread_runs(). This
* function must be executed before the switch to the new stack.
*/
before_thread_runs();
/*
* Copy the knowledge of CPU, TASK, THREAD and preemption counter to
* thread's stack.
*/
the_copy(THE, (the_t *) THREAD->kstack);
context_restore(&THREAD->saved_context);
/* not reached */
}
#ifdef CONFIG_SMP
/** Load balancing thread
*
* SMP load balancing thread, supervising thread supplies
* for the CPU it's wired to.
*
* @param arg Generic thread argument (unused).
*
*/
void kcpulb(void *arg)
{
thread_t *t;
int count, average, j, k = 0;
unsigned int i;
ipl_t ipl;
/*
* Detach kcpulb as nobody will call thread_join_timeout() on it.
*/
thread_detach(THREAD);
loop:
/*
* Work in 1s intervals.
*/
thread_sleep(1);
not_satisfied:
/*
* Calculate the number of threads that will be migrated/stolen from
* other CPU's. Note that situation can have changed between two
* passes. Each time get the most up to date counts.
*/
average = atomic_get(&nrdy) / config.cpu_active + 1;
count = average - atomic_get(&CPU->nrdy);
/* calculate number of threads to be stolen from other exec. units */
spinlock_lock(&(CPU->arch.exec_unit->proposed_nrdy_lock));
bool eu_busy = calculate_optimal_nrdy(CPU->arch.exec_unit);
unsigned int count_other_eus = CPU->arch.proposed_nrdy
- atomic_get(&(CPU->nrdy));
spinlock_unlock(&(CPU->arch.exec_unit->proposed_nrdy_lock));
/*
* If the CPU's parent core is overloaded, do not do the load
* balancing, otherwise we would migrate threads which should be
* migrated to other cores and since a thread cannot be migrated
* multiple times, it would not be migrated to the other core
* in the future.
*/
if (eu_busy)
return;
/*
* get the maximum - stole enough threads to satisfy both the need to
* have all virtual CPUs equally busy and the need to have all the
* cores equally busy
*/
if (((int) count_other_eus) > count)
count = count_other_eus;
if (count <= 0)
goto satisfied;
/*
* Searching least priority queues on all CPU's first and most priority
* queues on all CPU's last.
*/
for (j = RQ_COUNT - 1; j >= 0; j--) {
for (i = 0; i < config.cpu_active; i++) {
link_t *l;
runq_t *r;
cpu_t *cpu;
cpu = &cpus[(i + k) % config.cpu_active];
/*
* Not interested in ourselves.
* Doesn't require interrupt disabling for kcpulb has
* THREAD_FLAG_WIRED.
*/
if (CPU == cpu)
continue;
if (atomic_get(&cpu->nrdy) <= average)
continue;
ipl = interrupts_disable();
r = &cpu->rq[j];
spinlock_lock(&r->lock);
if (r->n == 0) {
spinlock_unlock(&r->lock);
interrupts_restore(ipl);
continue;
}
t = NULL;
l = r->rq_head.prev; /* search rq from the back */
while (l != &r->rq_head) {
t = list_get_instance(l, thread_t, rq_link);
/*
* We don't want to steal CPU-wired threads
* neither threads already stolen. The latter
* prevents threads from migrating between CPU's
* without ever being run. We don't want to
* steal threads whose FPU context is still in
* CPU.
*/
spinlock_lock(&t->lock);
if ((!(t->flags & (THREAD_FLAG_WIRED |
THREAD_FLAG_STOLEN))) &&
(!(t->fpu_context_engaged))) {
/*
* Remove t from r.
*/
spinlock_unlock(&t->lock);
atomic_dec(&cpu->nrdy);
atomic_dec(&nrdy);
if (cpu->arch.exec_unit)
atomic_dec(&(cpu->arch.exec_unit->nrdy));
r->n--;
list_remove(&t->rq_link);
break;
}
spinlock_unlock(&t->lock);
l = l->prev;
t = NULL;
}
spinlock_unlock(&r->lock);
if (t) {
/*
* Ready t on local CPU
*/
spinlock_lock(&t->lock);
#ifdef KCPULB_VERBOSE
printf("kcpulb%u: TID %" PRIu64
" -> cpu%u, " "nrdy=%ld, avg=%ld\n", CPU->id, t->tid,
CPU->id, atomic_get(&CPU->nrdy),
atomic_get(&nrdy) / config.cpu_active);
#endif
t->flags |= THREAD_FLAG_STOLEN;
t->state = Entering;
spinlock_unlock(&t->lock);
thread_ready(t);
interrupts_restore(ipl);
if (--count == 0)
goto satisfied;
/*
* We are not satisfied yet, focus on another
* CPU next time.
*/
k++;
continue;
}
interrupts_restore(ipl);
}
}
if (atomic_get(&CPU->nrdy)) {
/*
* Be a little bit light-weight and let migrated threads run.
*/
scheduler();
} else {
/*
* We failed to migrate a single thread.
* Give up this turn.
*/
goto loop;
}
goto not_satisfied;
satisfied:
goto loop;
}
#endif /* CONFIG_SMP */
/** Print information about threads & scheduler queues */
void sched_print_list(void)
{
ipl_t ipl;
unsigned int cpu, i;
runq_t *r;
thread_t *t;
link_t *cur;
/* We are going to mess with scheduler structures,
* let's not be interrupted */
ipl = interrupts_disable();
for (cpu = 0; cpu < config.cpu_count; cpu++) {
if (!cpus[cpu].active)
continue;
spinlock_lock(&cpus[cpu].lock);
printf("cpu%u: address=%p, nrdy=%ld, needs_relink=%" PRIc
"\n", cpus[cpu].id, &cpus[cpu], atomic_get(&cpus[cpu].nrdy),
cpus[cpu].needs_relink);
for (i = 0; i < RQ_COUNT; i++) {
r = &cpus[cpu].rq[i];
spinlock_lock(&r->lock);
if (!r->n) {
spinlock_unlock(&r->lock);
continue;
}
for (cur = r->rq_head.next; cur != &r->rq_head;
cur = cur->next) {
t = list_get_instance(cur, thread_t, rq_link);
printf("%" PRIu64
"(%s) ", t
->tid
, thread_states[t->state]);
}
spinlock_unlock(&r->lock);
}
spinlock_unlock(&cpus[cpu].lock);
}
interrupts_restore(ipl);
}
/** @}
*/