static void
poll_idle(void)
{
struct thread *td = curthread;
struct rtprio rtp;
rtp.prio = RTP_PRIO_MAX; /* lowest priority */
rtp.type = RTP_PRIO_IDLE;
PROC_SLOCK(td->td_proc);
rtp_to_pri(&rtp, td);
PROC_SUNLOCK(td->td_proc);
for (;;) {
if (poll_in_idle_loop && poll_handlers > 0) {
idlepoll_sleeping = 0;
ether_poll(poll_each_burst);
thread_lock(td);
mi_switch(SW_VOL, NULL);
thread_unlock(td);
} else {
idlepoll_sleeping = 1;
tsleep(&idlepoll_sleeping, 0, "pollid", hz * 3);
}
}
}
/*
* XXX The priority and scheduler modifications should
* be moved into published interfaces in kern/kern_sync.
*
* The permissions to modify process p were checked in "p31b_proc()".
*
*/
int
ksched_setscheduler(struct ksched *ksched,
struct thread *td, int policy, const struct sched_param *param)
{
int e = 0;
struct rtprio rtp;
switch(policy)
{
case SCHED_RR:
case SCHED_FIFO:
if (param->sched_priority >= P1B_PRIO_MIN &&
param->sched_priority <= P1B_PRIO_MAX)
{
rtp.prio = p4prio_to_rtpprio(param->sched_priority);
rtp.type = (policy == SCHED_FIFO)
? RTP_PRIO_FIFO : RTP_PRIO_REALTIME;
rtp_to_pri(&rtp, td);
}
else
e = EPERM;
break;
case SCHED_OTHER:
if (param->sched_priority >= 0 &&
param->sched_priority <= (PRI_MAX_TIMESHARE - PRI_MIN_TIMESHARE)) {
rtp.type = RTP_PRIO_NORMAL;
rtp.prio = p4prio_to_tsprio(param->sched_priority);
rtp_to_pri(&rtp, td);
} else
e = EINVAL;
break;
default:
e = EINVAL;
break;
}
return e;
}
static void
vm_pagezero(void __unused *arg)
{
struct rtprio rtp;
struct thread *td;
int pages, pri;
td = curthread;
rtp.prio = RTP_PRIO_MAX;
rtp.type = RTP_PRIO_IDLE;
pages = 0;
mtx_lock_spin(&sched_lock);
rtp_to_pri(&rtp, td->td_ksegrp);
pri = td->td_priority;
mtx_unlock_spin(&sched_lock);
idlezero_enable = idlezero_enable_default;
for (;;) {
if (vm_page_zero_check()) {
pages += vm_page_zero_idle();
#ifndef PREEMPTION
if (pages > idlezero_maxrun || sched_runnable()) {
mtx_lock_spin(&sched_lock);
mi_switch(SW_VOL, NULL);
mtx_unlock_spin(&sched_lock);
pages = 0;
}
#endif
} else {
vm_page_lock_queues();
wakeup_needed = TRUE;
msleep(&zero_state, &vm_page_queue_mtx, PDROP | pri,
"pgzero", hz * 300);
pages = 0;
}
}
}
static void
vm_pagezero(void)
{
struct thread *td;
struct proc *p;
struct rtprio rtp;
int pages = 0;
int pri;
td = curthread;
p = td->td_proc;
rtp.prio = RTP_PRIO_MAX;
rtp.type = RTP_PRIO_IDLE;
mtx_lock_spin(&sched_lock);
rtp_to_pri(&rtp, td->td_ksegrp);
pri = td->td_priority;
mtx_unlock_spin(&sched_lock);
PROC_LOCK(p);
p->p_flag |= P_NOLOAD;
PROC_UNLOCK(p);
for (;;) {
if (vm_page_zero_check()) {
pages += vm_page_zero_idle();
if (pages > idlezero_maxrun || sched_runnable()) {
mtx_lock_spin(&sched_lock);
td->td_proc->p_stats->p_ru.ru_nvcsw++;
mi_switch();
mtx_unlock_spin(&sched_lock);
pages = 0;
}
} else {
tsleep(&zero_state, pri, "pgzero", hz * 300);
pages = 0;
}
}
}
static int
create_thread(struct thread *td, mcontext_t *ctx,
void (*start_func)(void *), void *arg,
char *stack_base, size_t stack_size,
char *tls_base,
long *child_tid, long *parent_tid,
int flags, struct rtprio *rtp)
{
stack_t stack;
struct thread *newtd;
struct proc *p;
int error;
p = td->td_proc;
/* Have race condition but it is cheap. */
if (p->p_numthreads >= max_threads_per_proc) {
++max_threads_hits;
return (EPROCLIM);
}
if (rtp != NULL) {
switch(rtp->type) {
case RTP_PRIO_REALTIME:
case RTP_PRIO_FIFO:
/* Only root can set scheduler policy */
if (priv_check(td, PRIV_SCHED_SETPOLICY) != 0)
return (EPERM);
if (rtp->prio > RTP_PRIO_MAX)
return (EINVAL);
break;
case RTP_PRIO_NORMAL:
rtp->prio = 0;
break;
default:
return (EINVAL);
}
}
#ifdef RACCT
PROC_LOCK(td->td_proc);
error = racct_add(p, RACCT_NTHR, 1);
PROC_UNLOCK(td->td_proc);
if (error != 0)
return (EPROCLIM);
#endif
/* Initialize our td */
newtd = thread_alloc(0);
if (newtd == NULL) {
error = ENOMEM;
goto fail;
}
cpu_set_upcall(newtd, td);
/*
* Try the copyout as soon as we allocate the td so we don't
* have to tear things down in a failure case below.
* Here we copy out tid to two places, one for child and one
* for parent, because pthread can create a detached thread,
* if parent wants to safely access child tid, it has to provide
* its storage, because child thread may exit quickly and
* memory is freed before parent thread can access it.
*/
if ((child_tid != NULL &&
suword_lwpid(child_tid, newtd->td_tid)) ||
(parent_tid != NULL &&
suword_lwpid(parent_tid, newtd->td_tid))) {
thread_free(newtd);
error = EFAULT;
goto fail;
}
bzero(&newtd->td_startzero,
__rangeof(struct thread, td_startzero, td_endzero));
bcopy(&td->td_startcopy, &newtd->td_startcopy,
__rangeof(struct thread, td_startcopy, td_endcopy));
newtd->td_proc = td->td_proc;
newtd->td_ucred = crhold(td->td_ucred);
if (ctx != NULL) { /* old way to set user context */
error = set_mcontext(newtd, ctx);
if (error != 0) {
thread_free(newtd);
crfree(td->td_ucred);
goto fail;
}
} else {
/* Set up our machine context. */
stack.ss_sp = stack_base;
stack.ss_size = stack_size;
/* Set upcall address to user thread entry function. */
cpu_set_upcall_kse(newtd, start_func, arg, &stack);
/* Setup user TLS address and TLS pointer register. */
error = cpu_set_user_tls(newtd, tls_base);
if (error != 0) {
thread_free(newtd);
crfree(td->td_ucred);
goto fail;
}
}
PROC_LOCK(td->td_proc);
td->td_proc->p_flag |= P_HADTHREADS;
thread_link(newtd, p);
bcopy(p->p_comm, newtd->td_name, sizeof(newtd->td_name));
thread_lock(td);
/* let the scheduler know about these things. */
sched_fork_thread(td, newtd);
thread_unlock(td);
if (P_SHOULDSTOP(p))
newtd->td_flags |= TDF_ASTPENDING | TDF_NEEDSUSPCHK;
PROC_UNLOCK(p);
tidhash_add(newtd);
thread_lock(newtd);
if (rtp != NULL) {
if (!(td->td_pri_class == PRI_TIMESHARE &&
rtp->type == RTP_PRIO_NORMAL)) {
rtp_to_pri(rtp, newtd);
sched_prio(newtd, newtd->td_user_pri);
} /* ignore timesharing class */
}
TD_SET_CAN_RUN(newtd);
sched_add(newtd, SRQ_BORING);
thread_unlock(newtd);
return (0);
fail:
#ifdef RACCT
PROC_LOCK(p);
racct_sub(p, RACCT_NTHR, 1);
PROC_UNLOCK(p);
#endif
return (error);
}
int
thread_create(struct thread *td, struct rtprio *rtp,
int (*initialize_thread)(struct thread *, void *), void *thunk)
{
struct thread *newtd;
struct proc *p;
int error;
p = td->td_proc;
if (rtp != NULL) {
switch(rtp->type) {
case RTP_PRIO_REALTIME:
case RTP_PRIO_FIFO:
/* Only root can set scheduler policy */
if (priv_check(td, PRIV_SCHED_SETPOLICY) != 0)
return (EPERM);
if (rtp->prio > RTP_PRIO_MAX)
return (EINVAL);
break;
case RTP_PRIO_NORMAL:
rtp->prio = 0;
break;
default:
return (EINVAL);
}
}
#ifdef RACCT
if (racct_enable) {
PROC_LOCK(p);
error = racct_add(p, RACCT_NTHR, 1);
PROC_UNLOCK(p);
if (error != 0)
return (EPROCLIM);
}
#endif
/* Initialize our td */
error = kern_thr_alloc(p, 0, &newtd);
if (error)
goto fail;
cpu_set_upcall(newtd, td);
bzero(&newtd->td_startzero,
__rangeof(struct thread, td_startzero, td_endzero));
bcopy(&td->td_startcopy, &newtd->td_startcopy,
__rangeof(struct thread, td_startcopy, td_endcopy));
newtd->td_proc = td->td_proc;
thread_cow_get(newtd, td);
error = initialize_thread(newtd, thunk);
if (error != 0) {
thread_cow_free(newtd);
thread_free(newtd);
goto fail;
}
PROC_LOCK(p);
p->p_flag |= P_HADTHREADS;
thread_link(newtd, p);
bcopy(p->p_comm, newtd->td_name, sizeof(newtd->td_name));
newtd->td_pax = p->p_pax;
thread_lock(td);
/* let the scheduler know about these things. */
sched_fork_thread(td, newtd);
thread_unlock(td);
if (P_SHOULDSTOP(p))
newtd->td_flags |= TDF_ASTPENDING | TDF_NEEDSUSPCHK;
if (p->p_flag2 & P2_LWP_EVENTS)
newtd->td_dbgflags |= TDB_BORN;
/*
* Copy the existing thread VM policy into the new thread.
*/
vm_domain_policy_localcopy(&newtd->td_vm_dom_policy,
&td->td_vm_dom_policy);
PROC_UNLOCK(p);
tidhash_add(newtd);
thread_lock(newtd);
if (rtp != NULL) {
if (!(td->td_pri_class == PRI_TIMESHARE &&
rtp->type == RTP_PRIO_NORMAL)) {
rtp_to_pri(rtp, newtd);
sched_prio(newtd, newtd->td_user_pri);
} /* ignore timesharing class */
}
TD_SET_CAN_RUN(newtd);
sched_add(newtd, SRQ_BORING);
thread_unlock(newtd);
return (0);
fail:
#ifdef RACCT
if (racct_enable) {
PROC_LOCK(p);
racct_sub(p, RACCT_NTHR, 1);
PROC_UNLOCK(p);
}
#endif
return (error);
}
int
thread_create(struct thread *td, struct rtprio *rtp,
int (*initialize_thread)(struct thread *, void *), void *thunk)
{
struct thread *newtd;
struct proc *p;
int error;
p = td->td_proc;
if (rtp != NULL) {
switch(rtp->type) {
case RTP_PRIO_REALTIME:
case RTP_PRIO_FIFO:
/* Only root can set scheduler policy */
if (priv_check(td, PRIV_SCHED_SETPOLICY) != 0)
return (EPERM);
if (rtp->prio > RTP_PRIO_MAX)
return (EINVAL);
break;
case RTP_PRIO_NORMAL:
rtp->prio = 0;
break;
default:
return (EINVAL);
}
}
#ifdef RACCT
if (racct_enable) {
PROC_LOCK(p);
error = racct_add(p, RACCT_NTHR, 1);
PROC_UNLOCK(p);
if (error != 0)
return (EPROCLIM);
}
#endif
/* Initialize our td */
error = kern_thr_alloc(p, 0, &newtd);
if (error)
goto fail;
cpu_copy_thread(newtd, td);
bzero(&newtd->td_startzero,
__rangeof(struct thread, td_startzero, td_endzero));
bcopy(&td->td_startcopy, &newtd->td_startcopy,
__rangeof(struct thread, td_startcopy, td_endcopy));
newtd->td_proc = td->td_proc;
newtd->td_rb_list = newtd->td_rbp_list = newtd->td_rb_inact = 0;
thread_cow_get(newtd, td);
error = initialize_thread(newtd, thunk);
if (error != 0) {
thread_cow_free(newtd);
thread_free(newtd);
goto fail;
}
PROC_LOCK(p);
p->p_flag |= P_HADTHREADS;
thread_link(newtd, p);
bcopy(p->p_comm, newtd->td_name, sizeof(newtd->td_name));
thread_lock(td);
/* let the scheduler know about these things. */
sched_fork_thread(td, newtd);
thread_unlock(td);
if (P_SHOULDSTOP(p))
newtd->td_flags |= TDF_ASTPENDING | TDF_NEEDSUSPCHK;
if (p->p_ptevents & PTRACE_LWP)
newtd->td_dbgflags |= TDB_BORN;
PROC_UNLOCK(p);
#ifdef HWPMC_HOOKS
if (PMC_PROC_IS_USING_PMCS(p))
PMC_CALL_HOOK(newtd, PMC_FN_THR_CREATE, NULL);
else if (PMC_SYSTEM_SAMPLING_ACTIVE())
PMC_CALL_HOOK_UNLOCKED(newtd, PMC_FN_THR_CREATE_LOG, NULL);
#endif
tidhash_add(newtd);
thread_lock(newtd);
if (rtp != NULL) {
if (!(td->td_pri_class == PRI_TIMESHARE &&
rtp->type == RTP_PRIO_NORMAL)) {
rtp_to_pri(rtp, newtd);
sched_prio(newtd, newtd->td_user_pri);
} /* ignore timesharing class */
}
TD_SET_CAN_RUN(newtd);
sched_add(newtd, SRQ_BORING);
thread_unlock(newtd);
return (0);
fail:
#ifdef RACCT
if (racct_enable) {
PROC_LOCK(p);
racct_sub(p, RACCT_NTHR, 1);
PROC_UNLOCK(p);
}
#endif
return (error);
}