static int
proc_sum(struct proc *p, fixpt_t *estcpup)
{
struct thread *td;
int estcpu;
int val;
val = 0;
estcpu = 0;
FOREACH_THREAD_IN_PROC(p, td) {
thread_lock(td);
if (TD_ON_RUNQ(td) ||
TD_IS_RUNNING(td))
val = 1;
estcpu += sched_pctcpu(td);
thread_unlock(td);
}
/*
* Remove a thread from its KSEGRP's run queue.
* This in turn may remove it from a KSE if it was already assigned
* to one, possibly causing a new thread to be assigned to the KSE
* and the KSE getting a new priority.
*/
static void
remrunqueue(struct thread *td)
{
struct thread *td2, *td3;
struct ksegrp *kg;
struct kse *ke;
mtx_assert(&sched_lock, MA_OWNED);
KASSERT((TD_ON_RUNQ(td)), ("remrunqueue: Bad state on run queue"));
kg = td->td_ksegrp;
ke = td->td_kse;
CTR1(KTR_RUNQ, "remrunqueue: td%p", td);
TD_SET_CAN_RUN(td);
/*
* If it is not a threaded process, take the shortcut.
*/
if ((td->td_proc->p_flag & P_HADTHREADS) == 0) {
/* remve from sys run queue and free up a slot */
sched_rem(td);
ke->ke_state = KES_THREAD;
return;
}
td3 = TAILQ_PREV(td, threadqueue, td_runq);
TAILQ_REMOVE(&kg->kg_runq, td, td_runq);
kg->kg_runnable--;
if (ke->ke_state == KES_ONRUNQ) {
/*
* This thread has been assigned to the system run queue.
* We need to dissociate it and try assign the
* KSE to the next available thread. Then, we should
* see if we need to move the KSE in the run queues.
*/
sched_rem(td);
ke->ke_state = KES_THREAD;
td2 = kg->kg_last_assigned;
KASSERT((td2 != NULL), ("last assigned has wrong value"));
if (td2 == td)
kg->kg_last_assigned = td3;
/* slot_fill(kg); */ /* will replace it with another */
}
}
/*
* Change the priority of a thread that is on the run queue.
*/
void
adjustrunqueue( struct thread *td, int newpri)
{
struct ksegrp *kg;
struct kse *ke;
mtx_assert(&sched_lock, MA_OWNED);
KASSERT((TD_ON_RUNQ(td)), ("adjustrunqueue: Bad state on run queue"));
ke = td->td_kse;
CTR1(KTR_RUNQ, "adjustrunqueue: td%p", td);
/*
* If it is not a threaded process, take the shortcut.
*/
if ((td->td_proc->p_flag & P_HADTHREADS) == 0) {
/* We only care about the kse in the run queue. */
td->td_priority = newpri;
if (ke->ke_rqindex != (newpri / RQ_PPQ)) {
sched_rem(td);
sched_add(td, SRQ_BORING);
}
return;
}
/* It is a threaded process */
kg = td->td_ksegrp;
if (ke->ke_state == KES_ONRUNQ) {
if (kg->kg_last_assigned == td) {
kg->kg_last_assigned =
TAILQ_PREV(td, threadqueue, td_runq);
}
sched_rem(td);
}
TAILQ_REMOVE(&kg->kg_runq, td, td_runq);
kg->kg_runnable--;
TD_SET_CAN_RUN(td);
td->td_priority = newpri;
setrunqueue(td, SRQ_BORING);
}
/* 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;
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;
}
/*
* 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);
}
/*
* 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
}
/*
* Original vm_pageout_oom, will be called if LRU pageout_oom will fail
*/
static void
original_vm_pageout_oom(int shortage)
{
struct proc *p, *bigproc;
vm_offset_t size, bigsize;
struct thread *td;
struct vmspace *vm;
/*
* We keep the process bigproc locked once we find it to keep anyone
* from messing with it; however, there is a possibility of
* deadlock if process B is bigproc and one of it's child processes
* attempts to propagate a signal to B while we are waiting for A's
* lock while walking this list. To avoid this, we don't block on
* the process lock but just skip a process if it is already locked.
*/
bigproc = NULL;
bigsize = 0;
sx_slock(&allproc_lock);
FOREACH_PROC_IN_SYSTEM(p) {
int breakout;
if (PROC_TRYLOCK(p) == 0)
continue;
/*
* If this is a system, protected or killed process, skip it.
*/
if (p->p_state != PRS_NORMAL ||
(p->p_flag & (P_INEXEC | P_PROTECTED | P_SYSTEM)) ||
(p->p_pid == 1) || P_KILLED(p) ||
((p->p_pid < 48) && (swap_pager_avail != 0))) {
PROC_UNLOCK(p);
continue;
}
/*
* If the process is in a non-running type state,
* don't touch it. Check all the threads individually.
*/
breakout = 0;
FOREACH_THREAD_IN_PROC(p, td) {
thread_lock(td);
if (!TD_ON_RUNQ(td) &&
!TD_IS_RUNNING(td) &&
!TD_IS_SLEEPING(td)) {
thread_unlock(td);
breakout = 1;
break;
}
thread_unlock(td);
}
if (breakout) {
PROC_UNLOCK(p);
continue;
}
/*
* get the process size
*/
vm = vmspace_acquire_ref(p);
if (vm == NULL) {
PROC_UNLOCK(p);
continue;
}
if (!vm_map_trylock_read(&vm->vm_map)) {
vmspace_free(vm);
PROC_UNLOCK(p);
continue;
}
size = vmspace_swap_count(vm);
vm_map_unlock_read(&vm->vm_map);
if (shortage == VM_OOM_MEM)
size += vmspace_resident_count(vm);
vmspace_free(vm);
/*
* if the this process is bigger than the biggest one
* remember it.
*/
if (size > bigsize) {
if (bigproc != NULL)
PROC_UNLOCK(bigproc);
bigproc = p;
bigsize = size;
} else
PROC_UNLOCK(p);
}
/*
* Read proc's from memory file into buffer bp, which has space to hold
* at most maxcnt procs.
*/
static int
kvm_proclist(kvm_t *kd, int what, int arg, struct proc *p,
struct kinfo_proc *bp, int maxcnt)
{
int cnt = 0;
struct kinfo_proc kinfo_proc, *kp;
struct pgrp pgrp;
struct session sess;
struct cdev t_cdev;
struct tty tty;
struct vmspace vmspace;
struct sigacts sigacts;
#if 0
struct pstats pstats;
#endif
struct ucred ucred;
struct prison pr;
struct thread mtd;
struct proc proc;
struct proc pproc;
struct sysentvec sysent;
char svname[KI_EMULNAMELEN];
kp = &kinfo_proc;
kp->ki_structsize = sizeof(kinfo_proc);
/*
* Loop on the processes. this is completely broken because we need to be
* able to loop on the threads and merge the ones that are the same process some how.
*/
for (; cnt < maxcnt && p != NULL; p = LIST_NEXT(&proc, p_list)) {
memset(kp, 0, sizeof *kp);
if (KREAD(kd, (u_long)p, &proc)) {
_kvm_err(kd, kd->program, "can't read proc at %p", p);
return (-1);
}
if (proc.p_state == PRS_NEW)
continue;
if (proc.p_state != PRS_ZOMBIE) {
if (KREAD(kd, (u_long)TAILQ_FIRST(&proc.p_threads),
&mtd)) {
_kvm_err(kd, kd->program,
"can't read thread at %p",
TAILQ_FIRST(&proc.p_threads));
return (-1);
}
}
if (KREAD(kd, (u_long)proc.p_ucred, &ucred) == 0) {
kp->ki_ruid = ucred.cr_ruid;
kp->ki_svuid = ucred.cr_svuid;
kp->ki_rgid = ucred.cr_rgid;
kp->ki_svgid = ucred.cr_svgid;
kp->ki_cr_flags = ucred.cr_flags;
if (ucred.cr_ngroups > KI_NGROUPS) {
kp->ki_ngroups = KI_NGROUPS;
kp->ki_cr_flags |= KI_CRF_GRP_OVERFLOW;
} else
kp->ki_ngroups = ucred.cr_ngroups;
kvm_read(kd, (u_long)ucred.cr_groups, kp->ki_groups,
kp->ki_ngroups * sizeof(gid_t));
kp->ki_uid = ucred.cr_uid;
if (ucred.cr_prison != NULL) {
if (KREAD(kd, (u_long)ucred.cr_prison, &pr)) {
_kvm_err(kd, kd->program,
"can't read prison at %p",
ucred.cr_prison);
return (-1);
}
kp->ki_jid = pr.pr_id;
}
}
switch(what & ~KERN_PROC_INC_THREAD) {
case KERN_PROC_GID:
if (kp->ki_groups[0] != (gid_t)arg)
continue;
break;
case KERN_PROC_PID:
if (proc.p_pid != (pid_t)arg)
continue;
break;
case KERN_PROC_RGID:
if (kp->ki_rgid != (gid_t)arg)
continue;
break;
case KERN_PROC_UID:
if (kp->ki_uid != (uid_t)arg)
continue;
break;
case KERN_PROC_RUID:
if (kp->ki_ruid != (uid_t)arg)
continue;
break;
}
/*
* We're going to add another proc to the set. If this
* will overflow the buffer, assume the reason is because
* nprocs (or the proc list) is corrupt and declare an error.
*/
if (cnt >= maxcnt) {
_kvm_err(kd, kd->program, "nprocs corrupt");
return (-1);
}
/*
* gather kinfo_proc
*/
kp->ki_paddr = p;
kp->ki_addr = 0; /* XXX uarea */
/* kp->ki_kstack = proc.p_thread.td_kstack; XXXKSE */
kp->ki_args = proc.p_args;
kp->ki_tracep = proc.p_tracevp;
kp->ki_textvp = proc.p_textvp;
kp->ki_fd = proc.p_fd;
kp->ki_vmspace = proc.p_vmspace;
if (proc.p_sigacts != NULL) {
if (KREAD(kd, (u_long)proc.p_sigacts, &sigacts)) {
_kvm_err(kd, kd->program,
"can't read sigacts at %p", proc.p_sigacts);
return (-1);
}
kp->ki_sigignore = sigacts.ps_sigignore;
kp->ki_sigcatch = sigacts.ps_sigcatch;
}
#if 0
if ((proc.p_flag & P_INMEM) && proc.p_stats != NULL) {
if (KREAD(kd, (u_long)proc.p_stats, &pstats)) {
_kvm_err(kd, kd->program,
"can't read stats at %x", proc.p_stats);
return (-1);
}
kp->ki_start = pstats.p_start;
/*
* XXX: The times here are probably zero and need
* to be calculated from the raw data in p_rux and
* p_crux.
*/
kp->ki_rusage = pstats.p_ru;
kp->ki_childstime = pstats.p_cru.ru_stime;
kp->ki_childutime = pstats.p_cru.ru_utime;
/* Some callers want child-times in a single value */
timeradd(&kp->ki_childstime, &kp->ki_childutime,
&kp->ki_childtime);
}
#endif
if (proc.p_oppid)
kp->ki_ppid = proc.p_oppid;
else if (proc.p_pptr) {
if (KREAD(kd, (u_long)proc.p_pptr, &pproc)) {
_kvm_err(kd, kd->program,
"can't read pproc at %p", proc.p_pptr);
return (-1);
}
kp->ki_ppid = pproc.p_pid;
} else
kp->ki_ppid = 0;
if (proc.p_pgrp == NULL)
goto nopgrp;
if (KREAD(kd, (u_long)proc.p_pgrp, &pgrp)) {
_kvm_err(kd, kd->program, "can't read pgrp at %p",
proc.p_pgrp);
return (-1);
}
kp->ki_pgid = pgrp.pg_id;
kp->ki_jobc = pgrp.pg_jobc;
if (KREAD(kd, (u_long)pgrp.pg_session, &sess)) {
_kvm_err(kd, kd->program, "can't read session at %p",
pgrp.pg_session);
return (-1);
}
kp->ki_sid = sess.s_sid;
(void)memcpy(kp->ki_login, sess.s_login,
sizeof(kp->ki_login));
kp->ki_kiflag = sess.s_ttyvp ? KI_CTTY : 0;
if (sess.s_leader == p)
kp->ki_kiflag |= KI_SLEADER;
if ((proc.p_flag & P_CONTROLT) && sess.s_ttyp != NULL) {
if (KREAD(kd, (u_long)sess.s_ttyp, &tty)) {
_kvm_err(kd, kd->program,
"can't read tty at %p", sess.s_ttyp);
return (-1);
}
if (tty.t_dev != NULL) {
if (KREAD(kd, (u_long)tty.t_dev, &t_cdev)) {
_kvm_err(kd, kd->program,
"can't read cdev at %p",
tty.t_dev);
return (-1);
}
#if 0
kp->ki_tdev = t_cdev.si_udev;
#else
kp->ki_tdev = NODEV;
#endif
}
if (tty.t_pgrp != NULL) {
if (KREAD(kd, (u_long)tty.t_pgrp, &pgrp)) {
_kvm_err(kd, kd->program,
"can't read tpgrp at %p",
tty.t_pgrp);
return (-1);
}
kp->ki_tpgid = pgrp.pg_id;
} else
kp->ki_tpgid = -1;
if (tty.t_session != NULL) {
if (KREAD(kd, (u_long)tty.t_session, &sess)) {
_kvm_err(kd, kd->program,
"can't read session at %p",
tty.t_session);
return (-1);
}
kp->ki_tsid = sess.s_sid;
}
} else {
nopgrp:
kp->ki_tdev = NODEV;
}
if ((proc.p_state != PRS_ZOMBIE) && mtd.td_wmesg)
(void)kvm_read(kd, (u_long)mtd.td_wmesg,
kp->ki_wmesg, WMESGLEN);
(void)kvm_read(kd, (u_long)proc.p_vmspace,
(char *)&vmspace, sizeof(vmspace));
kp->ki_size = vmspace.vm_map.size;
/*
* Approximate the kernel's method of calculating
* this field.
*/
#define pmap_resident_count(pm) ((pm)->pm_stats.resident_count)
kp->ki_rssize = pmap_resident_count(&vmspace.vm_pmap);
kp->ki_swrss = vmspace.vm_swrss;
kp->ki_tsize = vmspace.vm_tsize;
kp->ki_dsize = vmspace.vm_dsize;
kp->ki_ssize = vmspace.vm_ssize;
switch (what & ~KERN_PROC_INC_THREAD) {
case KERN_PROC_PGRP:
if (kp->ki_pgid != (pid_t)arg)
continue;
break;
case KERN_PROC_SESSION:
if (kp->ki_sid != (pid_t)arg)
continue;
break;
case KERN_PROC_TTY:
if ((proc.p_flag & P_CONTROLT) == 0 ||
kp->ki_tdev != (dev_t)arg)
continue;
break;
}
if (proc.p_comm[0] != 0)
strlcpy(kp->ki_comm, proc.p_comm, MAXCOMLEN);
(void)kvm_read(kd, (u_long)proc.p_sysent, (char *)&sysent,
sizeof(sysent));
(void)kvm_read(kd, (u_long)sysent.sv_name, (char *)&svname,
sizeof(svname));
if (svname[0] != 0)
strlcpy(kp->ki_emul, svname, KI_EMULNAMELEN);
if ((proc.p_state != PRS_ZOMBIE) &&
(mtd.td_blocked != 0)) {
kp->ki_kiflag |= KI_LOCKBLOCK;
if (mtd.td_lockname)
(void)kvm_read(kd,
(u_long)mtd.td_lockname,
kp->ki_lockname, LOCKNAMELEN);
kp->ki_lockname[LOCKNAMELEN] = 0;
}
kp->ki_runtime = cputick2usec(proc.p_rux.rux_runtime);
kp->ki_pid = proc.p_pid;
kp->ki_siglist = proc.p_siglist;
SIGSETOR(kp->ki_siglist, mtd.td_siglist);
kp->ki_sigmask = mtd.td_sigmask;
kp->ki_xstat = KW_EXITCODE(proc.p_xexit, proc.p_xsig);
kp->ki_acflag = proc.p_acflag;
kp->ki_lock = proc.p_lock;
if (proc.p_state != PRS_ZOMBIE) {
kp->ki_swtime = (ticks - proc.p_swtick) / hz;
kp->ki_flag = proc.p_flag;
kp->ki_sflag = 0;
kp->ki_nice = proc.p_nice;
kp->ki_traceflag = proc.p_traceflag;
if (proc.p_state == PRS_NORMAL) {
if (TD_ON_RUNQ(&mtd) ||
TD_CAN_RUN(&mtd) ||
TD_IS_RUNNING(&mtd)) {
kp->ki_stat = SRUN;
} else if (mtd.td_state ==
TDS_INHIBITED) {
if (P_SHOULDSTOP(&proc)) {
kp->ki_stat = SSTOP;
} else if (
TD_IS_SLEEPING(&mtd)) {
kp->ki_stat = SSLEEP;
} else if (TD_ON_LOCK(&mtd)) {
kp->ki_stat = SLOCK;
} else {
kp->ki_stat = SWAIT;
}
}
} else {
kp->ki_stat = SIDL;
}
/* Stuff from the thread */
kp->ki_pri.pri_level = mtd.td_priority;
kp->ki_pri.pri_native = mtd.td_base_pri;
kp->ki_lastcpu = mtd.td_lastcpu;
kp->ki_wchan = mtd.td_wchan;
kp->ki_oncpu = mtd.td_oncpu;
if (mtd.td_name[0] != '\0')
strlcpy(kp->ki_tdname, mtd.td_name, sizeof(kp->ki_tdname));
kp->ki_pctcpu = 0;
kp->ki_rqindex = 0;
/*
* Note: legacy fields; wraps at NO_CPU_OLD or the
* old max CPU value as appropriate
*/
if (mtd.td_lastcpu == NOCPU)
kp->ki_lastcpu_old = NOCPU_OLD;
else if (mtd.td_lastcpu > MAXCPU_OLD)
kp->ki_lastcpu_old = MAXCPU_OLD;
else
kp->ki_lastcpu_old = mtd.td_lastcpu;
if (mtd.td_oncpu == NOCPU)
kp->ki_oncpu_old = NOCPU_OLD;
else if (mtd.td_oncpu > MAXCPU_OLD)
kp->ki_oncpu_old = MAXCPU_OLD;
else
kp->ki_oncpu_old = mtd.td_oncpu;
} else {
kp->ki_stat = SZOMB;
}
kp->ki_tdev_freebsd11 = kp->ki_tdev; /* truncate */
bcopy(&kinfo_proc, bp, sizeof(kinfo_proc));
++bp;
++cnt;
}
return (cnt);
}