kqueue: Add simple per-cpu cache for the knote

[dragonfly.git] / sys / kern / kern_event.c

/*-
 * Copyright (c) 1999,2000,2001 Jonathan Lemon <jlemon@FreeBSD.org>
 * All rights reserved.
 *
 * Redistribution and use in source and binary forms, with or without
 * modification, are permitted provided that the following conditions
 * are met:
 * 1. Redistributions of source code must retain the above copyright
 *    notice, this list of conditions and the following disclaimer.
 * 2. Redistributions in binary form must reproduce the above copyright
 *    notice, this list of conditions and the following disclaimer in the
 *    documentation and/or other materials provided with the distribution.
 *
 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR 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 AUTHOR 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.
 *
 * $FreeBSD: src/sys/kern/kern_event.c,v 1.2.2.10 2004/04/04 07:03:14 cperciva Exp $
 */

#include <sys/param.h>
#include <sys/systm.h>
#include <sys/kernel.h>
#include <sys/proc.h>
#include <sys/malloc.h> 
#include <sys/unistd.h>
#include <sys/file.h>
#include <sys/lock.h>
#include <sys/fcntl.h>
#include <sys/queue.h>
#include <sys/event.h>
#include <sys/eventvar.h>
#include <sys/protosw.h>
#include <sys/socket.h>
#include <sys/socketvar.h>
#include <sys/stat.h>
#include <sys/sysctl.h>
#include <sys/sysproto.h>
#include <sys/thread.h>
#include <sys/uio.h>
#include <sys/signalvar.h>
#include <sys/filio.h>
#include <sys/ktr.h>

#include <sys/thread2.h>
#include <sys/file2.h>
#include <sys/mplock2.h>

#define EVENT_REGISTER  1
#define EVENT_PROCESS   2

MALLOC_DEFINE(M_KQUEUE, "kqueue", "memory for kqueue system");

struct kevent_copyin_args {
        struct kevent_args      *ka;
        int                     pchanges;
};

#define KNOTE_CACHE_MAX         8

struct knote_cache_list {
        struct klist            knote_cache;
        int                     knote_cache_cnt;
} __cachealign;

static int      kqueue_scan(struct kqueue *kq, struct kevent *kevp, int count,
                    struct knote *marker);
static int      kqueue_read(struct file *fp, struct uio *uio,
                    struct ucred *cred, int flags);
static int      kqueue_write(struct file *fp, struct uio *uio,
                    struct ucred *cred, int flags);
static int      kqueue_ioctl(struct file *fp, u_long com, caddr_t data,
                    struct ucred *cred, struct sysmsg *msg);
static int      kqueue_kqfilter(struct file *fp, struct knote *kn);
static int      kqueue_stat(struct file *fp, struct stat *st,
                    struct ucred *cred);
static int      kqueue_close(struct file *fp);
static void     kqueue_wakeup(struct kqueue *kq);
static int      filter_attach(struct knote *kn);
static int      filter_event(struct knote *kn, long hint);

/*
 * MPSAFE
 */
static struct fileops kqueueops = {
        .fo_read = kqueue_read,
        .fo_write = kqueue_write,
        .fo_ioctl = kqueue_ioctl,
        .fo_kqfilter = kqueue_kqfilter,
        .fo_stat = kqueue_stat,
        .fo_close = kqueue_close,
        .fo_shutdown = nofo_shutdown
};

static void     knote_attach(struct knote *kn);
static void     knote_drop(struct knote *kn);
static void     knote_detach_and_drop(struct knote *kn);
static void     knote_enqueue(struct knote *kn);
static void     knote_dequeue(struct knote *kn);
static struct   knote *knote_alloc(void);
static void     knote_free(struct knote *kn);

static void     filt_kqdetach(struct knote *kn);
static int      filt_kqueue(struct knote *kn, long hint);
static int      filt_procattach(struct knote *kn);
static void     filt_procdetach(struct knote *kn);
static int      filt_proc(struct knote *kn, long hint);
static int      filt_fileattach(struct knote *kn);
static void     filt_timerexpire(void *knx);
static int      filt_timerattach(struct knote *kn);
static void     filt_timerdetach(struct knote *kn);
static int      filt_timer(struct knote *kn, long hint);
static int      filt_userattach(struct knote *kn);
static void     filt_userdetach(struct knote *kn);
static int      filt_user(struct knote *kn, long hint);
static void     filt_usertouch(struct knote *kn, struct kevent *kev,
                                u_long type);

static struct filterops file_filtops =
        { FILTEROP_ISFD | FILTEROP_MPSAFE, filt_fileattach, NULL, NULL };
static struct filterops kqread_filtops =
        { FILTEROP_ISFD | FILTEROP_MPSAFE, NULL, filt_kqdetach, filt_kqueue };
static struct filterops proc_filtops =
        { 0, filt_procattach, filt_procdetach, filt_proc };
static struct filterops timer_filtops =
        { FILTEROP_MPSAFE, filt_timerattach, filt_timerdetach, filt_timer };
static struct filterops user_filtops =
        { FILTEROP_MPSAFE, filt_userattach, filt_userdetach, filt_user };

static int              kq_ncallouts = 0;
static int              kq_calloutmax = (4 * 1024);
SYSCTL_INT(_kern, OID_AUTO, kq_calloutmax, CTLFLAG_RW,
    &kq_calloutmax, 0, "Maximum number of callouts allocated for kqueue");
static int              kq_checkloop = 1000000;
SYSCTL_INT(_kern, OID_AUTO, kq_checkloop, CTLFLAG_RW,
    &kq_checkloop, 0, "Maximum number of loops for kqueue scan");

#define KNOTE_ACTIVATE(kn) do {                                         \
        kn->kn_status |= KN_ACTIVE;                                     \
        if ((kn->kn_status & (KN_QUEUED | KN_DISABLED)) == 0)           \
                knote_enqueue(kn);                                      \
} while(0)

#define KN_HASHSIZE             64              /* XXX should be tunable */
#define KN_HASH(val, mask)      (((val) ^ (val >> 8)) & (mask))

extern struct filterops aio_filtops;
extern struct filterops sig_filtops;

/*
 * Table for for all system-defined filters.
 */
static struct filterops *sysfilt_ops[] = {
        &file_filtops,                  /* EVFILT_READ */
        &file_filtops,                  /* EVFILT_WRITE */
        &aio_filtops,                   /* EVFILT_AIO */
        &file_filtops,                  /* EVFILT_VNODE */
        &proc_filtops,                  /* EVFILT_PROC */
        &sig_filtops,                   /* EVFILT_SIGNAL */
        &timer_filtops,                 /* EVFILT_TIMER */
        &file_filtops,                  /* EVFILT_EXCEPT */
        &user_filtops,                  /* EVFILT_USER */
};

static struct knote_cache_list  knote_cache_lists[MAXCPU];

/*
 * Acquire a knote, return non-zero on success, 0 on failure.
 *
 * If we cannot acquire the knote we sleep and return 0.  The knote
 * may be stale on return in this case and the caller must restart
 * whatever loop they are in.
 *
 * Related kq token must be held.
 */
static __inline int
knote_acquire(struct knote *kn)
{
        if (kn->kn_status & KN_PROCESSING) {
                kn->kn_status |= KN_WAITING | KN_REPROCESS;
                tsleep(kn, 0, "kqepts", hz);
                /* knote may be stale now */
                return(0);
        }
        kn->kn_status |= KN_PROCESSING;
        return(1);
}

/*
 * Release an acquired knote, clearing KN_PROCESSING and handling any
 * KN_REPROCESS events.
 *
 * Caller must be holding the related kq token
 *
 * Non-zero is returned if the knote is destroyed or detached.
 */
static __inline void
knote_release(struct knote *kn)
{
        while (kn->kn_status & KN_REPROCESS) {
                kn->kn_status &= ~KN_REPROCESS;
                if (kn->kn_status & KN_WAITING) {
                        kn->kn_status &= ~KN_WAITING;
                        wakeup(kn);
                }
                if (kn->kn_status & KN_DELETING) {
                        knote_detach_and_drop(kn);
                        return;
                        /* NOT REACHED */
                }
                if (filter_event(kn, 0))
                        KNOTE_ACTIVATE(kn);
        }
        kn->kn_status &= ~KN_PROCESSING;
        /* kn should not be accessed anymore */
}

static int
filt_fileattach(struct knote *kn)
{
        return (fo_kqfilter(kn->kn_fp, kn));
}

/*
 * MPSAFE
 */
static int
kqueue_kqfilter(struct file *fp, struct knote *kn)
{
        struct kqueue *kq = (struct kqueue *)kn->kn_fp->f_data;

        if (kn->kn_filter != EVFILT_READ)
                return (EOPNOTSUPP);

        kn->kn_fop = &kqread_filtops;
        knote_insert(&kq->kq_kqinfo.ki_note, kn);
        return (0);
}

static void
filt_kqdetach(struct knote *kn)
{
        struct kqueue *kq = (struct kqueue *)kn->kn_fp->f_data;

        knote_remove(&kq->kq_kqinfo.ki_note, kn);
}

/*ARGSUSED*/
static int
filt_kqueue(struct knote *kn, long hint)
{
        struct kqueue *kq = (struct kqueue *)kn->kn_fp->f_data;

        kn->kn_data = kq->kq_count;
        return (kn->kn_data > 0);
}

static int
filt_procattach(struct knote *kn)
{
        struct proc *p;
        int immediate;

        immediate = 0;
        p = pfind(kn->kn_id);
        if (p == NULL && (kn->kn_sfflags & NOTE_EXIT)) {
                p = zpfind(kn->kn_id);
                immediate = 1;
        }
        if (p == NULL) {
                return (ESRCH);
        }
        if (!PRISON_CHECK(curthread->td_ucred, p->p_ucred)) {
                if (p)
                        PRELE(p);
                return (EACCES);
        }

        lwkt_gettoken(&p->p_token);
        kn->kn_ptr.p_proc = p;
        kn->kn_flags |= EV_CLEAR;               /* automatically set */

        /*
         * internal flag indicating registration done by kernel
         */
        if (kn->kn_flags & EV_FLAG1) {
                kn->kn_data = kn->kn_sdata;             /* ppid */
                kn->kn_fflags = NOTE_CHILD;
                kn->kn_flags &= ~EV_FLAG1;
        }

        knote_insert(&p->p_klist, kn);

        /*
         * Immediately activate any exit notes if the target process is a
         * zombie.  This is necessary to handle the case where the target
         * process, e.g. a child, dies before the kevent is negistered.
         */
        if (immediate && filt_proc(kn, NOTE_EXIT))
                KNOTE_ACTIVATE(kn);
        lwkt_reltoken(&p->p_token);
        PRELE(p);

        return (0);
}

/*
 * The knote may be attached to a different process, which may exit,
 * leaving nothing for the knote to be attached to.  So when the process
 * exits, the knote is marked as DETACHED and also flagged as ONESHOT so
 * it will be deleted when read out.  However, as part of the knote deletion,
 * this routine is called, so a check is needed to avoid actually performing
 * a detach, because the original process does not exist any more.
 */
static void
filt_procdetach(struct knote *kn)
{
        struct proc *p;

        if (kn->kn_status & KN_DETACHED)
                return;
        p = kn->kn_ptr.p_proc;
        knote_remove(&p->p_klist, kn);
}

static int
filt_proc(struct knote *kn, long hint)
{
        u_int event;

        /*
         * mask off extra data
         */
        event = (u_int)hint & NOTE_PCTRLMASK;

        /*
         * if the user is interested in this event, record it.
         */
        if (kn->kn_sfflags & event)
                kn->kn_fflags |= event;

        /*
         * Process is gone, so flag the event as finished.  Detach the
         * knote from the process now because the process will be poof,
         * gone later on.
         */
        if (event == NOTE_EXIT) {
                struct proc *p = kn->kn_ptr.p_proc;
                if ((kn->kn_status & KN_DETACHED) == 0) {
                        PHOLD(p);
                        knote_remove(&p->p_klist, kn);
                        kn->kn_status |= KN_DETACHED;
                        kn->kn_data = p->p_xstat;
                        kn->kn_ptr.p_proc = NULL;
                        PRELE(p);
                }
                kn->kn_flags |= (EV_EOF | EV_NODATA | EV_ONESHOT); 
                return (1);
        }

        /*
         * process forked, and user wants to track the new process,
         * so attach a new knote to it, and immediately report an
         * event with the parent's pid.
         */
        if ((event == NOTE_FORK) && (kn->kn_sfflags & NOTE_TRACK)) {
                struct kevent kev;
                int error;

                /*
                 * register knote with new process.
                 */
                kev.ident = hint & NOTE_PDATAMASK;      /* pid */
                kev.filter = kn->kn_filter;
                kev.flags = kn->kn_flags | EV_ADD | EV_ENABLE | EV_FLAG1;
                kev.fflags = kn->kn_sfflags;
                kev.data = kn->kn_id;                   /* parent */
                kev.udata = kn->kn_kevent.udata;        /* preserve udata */
                error = kqueue_register(kn->kn_kq, &kev);
                if (error)
                        kn->kn_fflags |= NOTE_TRACKERR;
        }

        return (kn->kn_fflags != 0);
}

static void
filt_timerreset(struct knote *kn)
{
        struct callout *calloutp;
        struct timeval tv;
        int tticks;

        tv.tv_sec = kn->kn_sdata / 1000;
        tv.tv_usec = (kn->kn_sdata % 1000) * 1000;
        tticks = tvtohz_high(&tv);
        calloutp = (struct callout *)kn->kn_hook;
        callout_reset(calloutp, tticks, filt_timerexpire, kn);
}

/*
 * The callout interlocks with callout_terminate() but can still
 * race a deletion so if KN_DELETING is set we just don't touch
 * the knote.
 */
static void
filt_timerexpire(void *knx)
{
        struct knote *kn = knx;
        struct kqueue *kq = kn->kn_kq;

        lwkt_getpooltoken(kq);

        /*
         * Open knote_acquire(), since we can't sleep in callout,
         * however, we do need to record this expiration.
         */
        kn->kn_data++;
        if (kn->kn_status & KN_PROCESSING) {
                kn->kn_status |= KN_REPROCESS;
                if ((kn->kn_status & KN_DELETING) == 0 &&
                    (kn->kn_flags & EV_ONESHOT) == 0)
                        filt_timerreset(kn);
                lwkt_relpooltoken(kq);
                return;
        }
        KASSERT((kn->kn_status & KN_DELETING) == 0,
            ("acquire a deleting knote %#x", kn->kn_status));
        kn->kn_status |= KN_PROCESSING;

        KNOTE_ACTIVATE(kn);
        if ((kn->kn_flags & EV_ONESHOT) == 0)
                filt_timerreset(kn);

        knote_release(kn);

        lwkt_relpooltoken(kq);
}

/*
 * data contains amount of time to sleep, in milliseconds
 */ 
static int
filt_timerattach(struct knote *kn)
{
        struct callout *calloutp;
        int prev_ncallouts;

        prev_ncallouts = atomic_fetchadd_int(&kq_ncallouts, 1);
        if (prev_ncallouts >= kq_calloutmax) {
                atomic_subtract_int(&kq_ncallouts, 1);
                kn->kn_hook = NULL;
                return (ENOMEM);
        }

        kn->kn_flags |= EV_CLEAR;               /* automatically set */
        calloutp = kmalloc(sizeof(*calloutp), M_KQUEUE, M_WAITOK);
        callout_init_mp(calloutp);
        kn->kn_hook = (caddr_t)calloutp;

        filt_timerreset(kn);
        return (0);
}

/*
 * This function is called with the knote flagged locked but it is
 * still possible to race a callout event due to the callback blocking.
 * We must call callout_terminate() instead of callout_stop() to deal
 * with the race.
 */
static void
filt_timerdetach(struct knote *kn)
{
        struct callout *calloutp;

        calloutp = (struct callout *)kn->kn_hook;
        callout_terminate(calloutp);
        kfree(calloutp, M_KQUEUE);
        atomic_subtract_int(&kq_ncallouts, 1);
}

static int
filt_timer(struct knote *kn, long hint)
{

        return (kn->kn_data != 0);
}

/*
 * EVFILT_USER
 */
static int
filt_userattach(struct knote *kn)
{
        kn->kn_hook = NULL;
        if (kn->kn_fflags & NOTE_TRIGGER)
                kn->kn_ptr.hookid = 1;
        else
                kn->kn_ptr.hookid = 0;
        return 0;
}

static void
filt_userdetach(struct knote *kn)
{
        /* nothing to do */
}

static int
filt_user(struct knote *kn, long hint)
{
        return (kn->kn_ptr.hookid);
}

static void
filt_usertouch(struct knote *kn, struct kevent *kev, u_long type)
{
        u_int ffctrl;

        switch (type) {
        case EVENT_REGISTER:
                if (kev->fflags & NOTE_TRIGGER)
                        kn->kn_ptr.hookid = 1;

                ffctrl = kev->fflags & NOTE_FFCTRLMASK;
                kev->fflags &= NOTE_FFLAGSMASK;
                switch (ffctrl) {
                case NOTE_FFNOP:
                        break;

                case NOTE_FFAND:
                        kn->kn_sfflags &= kev->fflags;
                        break;

                case NOTE_FFOR:
                        kn->kn_sfflags |= kev->fflags;
                        break;

                case NOTE_FFCOPY:
                        kn->kn_sfflags = kev->fflags;
                        break;

                default:
                        /* XXX Return error? */
                        break;
                }
                kn->kn_sdata = kev->data;

                /*
                 * This is not the correct use of EV_CLEAR in an event
                 * modification, it should have been passed as a NOTE instead.
                 * But we need to maintain compatibility with Apple & FreeBSD.
                 *
                 * Note however that EV_CLEAR can still be used when doing
                 * the initial registration of the event and works as expected
                 * (clears the event on reception).
                 */
                if (kev->flags & EV_CLEAR) {
                        kn->kn_ptr.hookid = 0;
                        kn->kn_data = 0;
                        kn->kn_fflags = 0;
                }
                break;

        case EVENT_PROCESS:
                *kev = kn->kn_kevent;
                kev->fflags = kn->kn_sfflags;
                kev->data = kn->kn_sdata;
                if (kn->kn_flags & EV_CLEAR) {
                        kn->kn_ptr.hookid = 0;
                        /* kn_data, kn_fflags handled by parent */
                }
                break;

        default:
                panic("filt_usertouch() - invalid type (%ld)", type);
                break;
        }
}

/*
 * Initialize a kqueue.
 *
 * NOTE: The lwp/proc code initializes a kqueue for select/poll ops.
 *
 * MPSAFE
 */
void
kqueue_init(struct kqueue *kq, struct filedesc *fdp)
{
        TAILQ_INIT(&kq->kq_knpend);
        TAILQ_INIT(&kq->kq_knlist);
        kq->kq_count = 0;
        kq->kq_fdp = fdp;
        SLIST_INIT(&kq->kq_kqinfo.ki_note);
}

/*
 * Terminate a kqueue.  Freeing the actual kq itself is left up to the
 * caller (it might be embedded in a lwp so we don't do it here).
 *
 * The kq's knlist must be completely eradicated so block on any
 * processing races.
 */
void
kqueue_terminate(struct kqueue *kq)
{
        struct lwkt_token *tok;
        struct knote *kn;

        tok = lwkt_token_pool_lookup(kq);
        lwkt_gettoken(tok);
        while ((kn = TAILQ_FIRST(&kq->kq_knlist)) != NULL) {
                if (knote_acquire(kn))
                        knote_detach_and_drop(kn);
        }
        lwkt_reltoken(tok);

        if (kq->kq_knhash) {
                hashdestroy(kq->kq_knhash, M_KQUEUE, kq->kq_knhashmask);
                kq->kq_knhash = NULL;
                kq->kq_knhashmask = 0;
        }
}

/*
 * MPSAFE
 */
int
sys_kqueue(struct kqueue_args *uap)
{
        struct thread *td = curthread;
        struct kqueue *kq;
        struct file *fp;
        int fd, error;

        error = falloc(td->td_lwp, &fp, &fd);
        if (error)
                return (error);
        fp->f_flag = FREAD | FWRITE;
        fp->f_type = DTYPE_KQUEUE;
        fp->f_ops = &kqueueops;

        kq = kmalloc(sizeof(struct kqueue), M_KQUEUE, M_WAITOK | M_ZERO);
        kqueue_init(kq, td->td_proc->p_fd);
        fp->f_data = kq;

        fsetfd(kq->kq_fdp, fp, fd);
        uap->sysmsg_result = fd;
        fdrop(fp);
        return (error);
}

/*
 * Copy 'count' items into the destination list pointed to by uap->eventlist.
 */
static int
kevent_copyout(void *arg, struct kevent *kevp, int count, int *res)
{
        struct kevent_copyin_args *kap;
        int error;

        kap = (struct kevent_copyin_args *)arg;

        error = copyout(kevp, kap->ka->eventlist, count * sizeof(*kevp));
        if (error == 0) {
                kap->ka->eventlist += count;
                *res += count;
        } else {
                *res = -1;
        }

        return (error);
}

/*
 * Copy at most 'max' items from the list pointed to by kap->changelist,
 * return number of items in 'events'.
 */
static int
kevent_copyin(void *arg, struct kevent *kevp, int max, int *events)
{
        struct kevent_copyin_args *kap;
        int error, count;

        kap = (struct kevent_copyin_args *)arg;

        count = min(kap->ka->nchanges - kap->pchanges, max);
        error = copyin(kap->ka->changelist, kevp, count * sizeof *kevp);
        if (error == 0) {
                kap->ka->changelist += count;
                kap->pchanges += count;
                *events = count;
        }

        return (error);
}

/*
 * MPSAFE
 */
int
kern_kevent(struct kqueue *kq, int nevents, int *res, void *uap,
            k_copyin_fn kevent_copyinfn, k_copyout_fn kevent_copyoutfn,
            struct timespec *tsp_in)
{
        struct kevent *kevp;
        struct timespec *tsp, ats;
        int i, n, total, error, nerrors = 0;
        int lres;
        int limit = kq_checkloop;
        struct kevent kev[KQ_NEVENTS];
        struct knote marker;
        struct lwkt_token *tok;

        if (tsp_in == NULL || tsp_in->tv_sec || tsp_in->tv_nsec)
                atomic_set_int(&curthread->td_mpflags, TDF_MP_BATCH_DEMARC);

        tsp = tsp_in;
        *res = 0;

        for (;;) {
                n = 0;
                error = kevent_copyinfn(uap, kev, KQ_NEVENTS, &n);
                if (error)
                        return error;
                if (n == 0)
                        break;
                for (i = 0; i < n; i++) {
                        kevp = &kev[i];
                        kevp->flags &= ~EV_SYSFLAGS;
                        error = kqueue_register(kq, kevp);

                        /*
                         * If a registration returns an error we
                         * immediately post the error.  The kevent()
                         * call itself will fail with the error if
                         * no space is available for posting.
                         *
                         * Such errors normally bypass the timeout/blocking
                         * code.  However, if the copyoutfn function refuses
                         * to post the error (see sys_poll()), then we
                         * ignore it too.
                         */
                        if (error || (kevp->flags & EV_RECEIPT)) {
                                kevp->flags = EV_ERROR;
                                kevp->data = error;
                                lres = *res;
                                kevent_copyoutfn(uap, kevp, 1, res);
                                if (*res < 0) {
                                        return error;
                                } else if (lres != *res) {
                                        nevents--;
                                        nerrors++;
                                }
                        }
                }
        }
        if (nerrors)
                return 0;

        /*
         * Acquire/wait for events - setup timeout
         */
        if (tsp != NULL) {
                if (tsp->tv_sec || tsp->tv_nsec) {
                        getnanouptime(&ats);
                        timespecadd(tsp, &ats);         /* tsp = target time */
                }
        }

        /*
         * Loop as required.
         *
         * Collect as many events as we can. Sleeping on successive
         * loops is disabled if copyoutfn has incremented (*res).
         *
         * The loop stops if an error occurs, all events have been
         * scanned (the marker has been reached), or fewer than the
         * maximum number of events is found.
         *
         * The copyoutfn function does not have to increment (*res) in
         * order for the loop to continue.
         *
         * NOTE: doselect() usually passes 0x7FFFFFFF for nevents.
         */
        total = 0;
        error = 0;
        marker.kn_filter = EVFILT_MARKER;
        marker.kn_status = KN_PROCESSING;
        tok = lwkt_token_pool_lookup(kq);
        lwkt_gettoken(tok);
        TAILQ_INSERT_TAIL(&kq->kq_knpend, &marker, kn_tqe);
        lwkt_reltoken(tok);
        while ((n = nevents - total) > 0) {
                if (n > KQ_NEVENTS)
                        n = KQ_NEVENTS;

                /*
                 * If no events are pending sleep until timeout (if any)
                 * or an event occurs.
                 *
                 * After the sleep completes the marker is moved to the
                 * end of the list, making any received events available
                 * to our scan.
                 */
                if (kq->kq_count == 0 && *res == 0) {
                        int timeout;

                        if (tsp == NULL) {
                                timeout = 0;
                        } else if (tsp->tv_sec == 0 && tsp->tv_nsec == 0) {
                                error = EWOULDBLOCK;
                                break;
                        } else {
                                struct timespec atx = *tsp;

                                getnanouptime(&ats);
                                timespecsub(&atx, &ats);
                                if (atx.tv_sec < 0) {
                                        error = EWOULDBLOCK;
                                        break;
                                } else {
                                        timeout = atx.tv_sec > 24 * 60 * 60 ?
                                            24 * 60 * 60 * hz :
                                            tstohz_high(&atx);
                                }
                        }

                        lwkt_gettoken(tok);
                        if (kq->kq_count == 0) {
                                kq->kq_sleep_cnt++;
                                error = tsleep(kq, PCATCH, "kqread", timeout);

                                /* don't restart after signals... */
                                if (error == ERESTART)
                                        error = EINTR;
                                if (error) {
                                        lwkt_reltoken(tok);
                                        break;
                                }

                                TAILQ_REMOVE(&kq->kq_knpend, &marker, kn_tqe);
                                TAILQ_INSERT_TAIL(&kq->kq_knpend, &marker,
                                    kn_tqe);
                        }
                        lwkt_reltoken(tok);
                }

                /*
                 * Process all received events
                 * Account for all non-spurious events in our total
                 */
                i = kqueue_scan(kq, kev, n, &marker);
                if (i) {
                        lres = *res;
                        error = kevent_copyoutfn(uap, kev, i, res);
                        total += *res - lres;
                        if (error)
                                break;
                }
                if (limit && --limit == 0)
                        panic("kqueue: checkloop failed i=%d", i);

                /*
                 * Normally when fewer events are returned than requested
                 * we can stop.  However, if only spurious events were
                 * collected the copyout will not bump (*res) and we have
                 * to continue.
                 */
                if (i < n && *res)
                        break;

                /*
                 * Deal with an edge case where spurious events can cause
                 * a loop to occur without moving the marker.  This can
                 * prevent kqueue_scan() from picking up new events which
                 * race us.  We must be sure to move the marker for this
                 * case.
                 *
                 * NOTE: We do not want to move the marker if events
                 *       were scanned because normal kqueue operations
                 *       may reactivate events.  Moving the marker in
                 *       that case could result in duplicates for the
                 *       same event.
                 */
                if (i == 0) {
                        lwkt_gettoken(tok);
                        TAILQ_REMOVE(&kq->kq_knpend, &marker, kn_tqe);
                        TAILQ_INSERT_TAIL(&kq->kq_knpend, &marker, kn_tqe);
                        lwkt_reltoken(tok);
                }
        }
        lwkt_gettoken(tok);
        TAILQ_REMOVE(&kq->kq_knpend, &marker, kn_tqe);
        lwkt_reltoken(tok);

        /* Timeouts do not return EWOULDBLOCK. */
        if (error == EWOULDBLOCK)
                error = 0;
        return error;
}

/*
 * MPALMOSTSAFE
 */
int
sys_kevent(struct kevent_args *uap)
{
        struct thread *td = curthread;
        struct proc *p = td->td_proc;
        struct timespec ts, *tsp;
        struct kqueue *kq;
        struct file *fp = NULL;
        struct kevent_copyin_args *kap, ka;
        int error;

        if (uap->timeout) {
                error = copyin(uap->timeout, &ts, sizeof(ts));
                if (error)
                        return (error);
                tsp = &ts;
        } else {
                tsp = NULL;
        }
        fp = holdfp(p->p_fd, uap->fd, -1);
        if (fp == NULL)
                return (EBADF);
        if (fp->f_type != DTYPE_KQUEUE) {
                fdrop(fp);
                return (EBADF);
        }

        kq = (struct kqueue *)fp->f_data;

        kap = &ka;
        kap->ka = uap;
        kap->pchanges = 0;

        error = kern_kevent(kq, uap->nevents, &uap->sysmsg_result, kap,
                            kevent_copyin, kevent_copyout, tsp);

        fdrop(fp);

        return (error);
}

int
kqueue_register(struct kqueue *kq, struct kevent *kev)
{
        struct filedesc *fdp = kq->kq_fdp;
        struct klist *list = NULL;
        struct filterops *fops;
        struct file *fp = NULL;
        struct knote *kn = NULL;
        struct thread *td;
        int error = 0;
        struct knote_cache_list *cache_list;

        if (kev->filter < 0) {
                if (kev->filter + EVFILT_SYSCOUNT < 0)
                        return (EINVAL);
                fops = sysfilt_ops[~kev->filter];       /* to 0-base index */
        } else {
                /*
                 * XXX
                 * filter attach routine is responsible for insuring that
                 * the identifier can be attached to it.
                 */
                return (EINVAL);
        }

        if (fops->f_flags & FILTEROP_ISFD) {
                /* validate descriptor */
                fp = holdfp(fdp, kev->ident, -1);
                if (fp == NULL)
                        return (EBADF);
        }

        cache_list = &knote_cache_lists[mycpuid];
        if (SLIST_EMPTY(&cache_list->knote_cache)) {
                struct knote *new_kn;

                new_kn = knote_alloc();
                SLIST_INSERT_HEAD(&cache_list->knote_cache, new_kn, kn_link);
                cache_list->knote_cache_cnt++;
        }

        td = curthread;
        lwkt_getpooltoken(kq);

        /*
         * Make sure that only one thread can register event on this kqueue,
         * so that we would not suffer any race, even if the registration
         * blocked, i.e. kq token was released, and the kqueue was shared
         * between threads (this should be rare though).
         */
        while (__predict_false(kq->kq_regtd != NULL && kq->kq_regtd != td)) {
                kq->kq_state |= KQ_REGWAIT;
                tsleep(&kq->kq_regtd, 0, "kqreg", 0);
        }
        if (__predict_false(kq->kq_regtd != NULL)) {
                /* Recursive calling of kqueue_register() */
                td = NULL;
        } else {
                /* Owner of the kq_regtd, i.e. td != NULL */
                kq->kq_regtd = td;
        }

        if (fp != NULL) {
                list = &fp->f_klist;
        } else if (kq->kq_knhashmask) {
                list = &kq->kq_knhash[
                    KN_HASH((u_long)kev->ident, kq->kq_knhashmask)];
        }
        if (list != NULL) {
                lwkt_getpooltoken(list);
again:
                SLIST_FOREACH(kn, list, kn_link) {
                        if (kn->kn_kq == kq &&
                            kn->kn_filter == kev->filter &&
                            kn->kn_id == kev->ident) {
                                if (knote_acquire(kn) == 0)
                                        goto again;
                                break;
                        }
                }
                lwkt_relpooltoken(list);
        }

        /*
         * NOTE: At this point if kn is non-NULL we will have acquired
         *       it and set KN_PROCESSING.
         */
        if (kn == NULL && ((kev->flags & EV_ADD) == 0)) {
                error = ENOENT;
                goto done;
        }

        /*
         * kn now contains the matching knote, or NULL if no match
         */
        if (kev->flags & EV_ADD) {
                if (kn == NULL) {
                        kn = SLIST_FIRST(&cache_list->knote_cache);
                        if (kn == NULL) {
                                kn = knote_alloc();
                        } else {
                                SLIST_REMOVE_HEAD(&cache_list->knote_cache,
                                    kn_link);
                                cache_list->knote_cache_cnt--;
                        }
                        kn->kn_fp = fp;
                        kn->kn_kq = kq;
                        kn->kn_fop = fops;

                        /*
                         * apply reference count to knote structure, and
                         * do not release it at the end of this routine.
                         */
                        fp = NULL;

                        kn->kn_sfflags = kev->fflags;
                        kn->kn_sdata = kev->data;
                        kev->fflags = 0;
                        kev->data = 0;
                        kn->kn_kevent = *kev;

                        /*
                         * KN_PROCESSING prevents the knote from getting
                         * ripped out from under us while we are trying
                         * to attach it, in case the attach blocks.
                         */
                        kn->kn_status = KN_PROCESSING;
                        knote_attach(kn);
                        if ((error = filter_attach(kn)) != 0) {
                                kn->kn_status |= KN_DELETING | KN_REPROCESS;
                                knote_drop(kn);
                                goto done;
                        }

                        /*
                         * Interlock against close races which either tried
                         * to remove our knote while we were blocked or missed
                         * it entirely prior to our attachment.  We do not
                         * want to end up with a knote on a closed descriptor.
                         */
                        if ((fops->f_flags & FILTEROP_ISFD) &&
                            checkfdclosed(fdp, kev->ident, kn->kn_fp)) {
                                kn->kn_status |= KN_DELETING | KN_REPROCESS;
                        }
                } else {
                        /*
                         * The user may change some filter values after the
                         * initial EV_ADD, but doing so will not reset any 
                         * filter which have already been triggered.
                         */
                        KKASSERT(kn->kn_status & KN_PROCESSING);
                        if (fops == &user_filtops) {
                                filt_usertouch(kn, kev, EVENT_REGISTER);
                        } else {
                                kn->kn_sfflags = kev->fflags;
                                kn->kn_sdata = kev->data;
                                kn->kn_kevent.udata = kev->udata;
                        }
                }

                /*
                 * Execute the filter event to immediately activate the
                 * knote if necessary.  If reprocessing events are pending
                 * due to blocking above we do not run the filter here
                 * but instead let knote_release() do it.  Otherwise we
                 * might run the filter on a deleted event.
                 */
                if ((kn->kn_status & KN_REPROCESS) == 0) {
                        if (filter_event(kn, 0))
                                KNOTE_ACTIVATE(kn);
                }
        } else if (kev->flags & EV_DELETE) {
                /*
                 * Delete the existing knote
                 */
                knote_detach_and_drop(kn);
                goto done;
        } else {
                /*
                 * Modify an existing event.
                 *
                 * The user may change some filter values after the
                 * initial EV_ADD, but doing so will not reset any
                 * filter which have already been triggered.
                 */
                KKASSERT(kn->kn_status & KN_PROCESSING);
                if (fops == &user_filtops) {
                        filt_usertouch(kn, kev, EVENT_REGISTER);
                } else {
                        kn->kn_sfflags = kev->fflags;
                        kn->kn_sdata = kev->data;
                        kn->kn_kevent.udata = kev->udata;
                }

                /*
                 * Execute the filter event to immediately activate the
                 * knote if necessary.  If reprocessing events are pending
                 * due to blocking above we do not run the filter here
                 * but instead let knote_release() do it.  Otherwise we
                 * might run the filter on a deleted event.
                 */
                if ((kn->kn_status & KN_REPROCESS) == 0) {
                        if (filter_event(kn, 0))
                                KNOTE_ACTIVATE(kn);
                }
        }

        /*
         * Disablement does not deactivate a knote here.
         */
        if ((kev->flags & EV_DISABLE) &&
            ((kn->kn_status & KN_DISABLED) == 0)) {
                kn->kn_status |= KN_DISABLED;
        }

        /*
         * Re-enablement may have to immediately enqueue an active knote.
         */
        if ((kev->flags & EV_ENABLE) && (kn->kn_status & KN_DISABLED)) {
                kn->kn_status &= ~KN_DISABLED;
                if ((kn->kn_status & KN_ACTIVE) &&
                    ((kn->kn_status & KN_QUEUED) == 0)) {
                        knote_enqueue(kn);
                }
        }

        /*
         * Handle any required reprocessing
         */
        knote_release(kn);
        /* kn may be invalid now */

done:
        if (td != NULL) { /* Owner of the kq_regtd */
                kq->kq_regtd = NULL;
                if (__predict_false(kq->kq_state & KQ_REGWAIT)) {
                        kq->kq_state &= ~KQ_REGWAIT;
                        wakeup(&kq->kq_regtd);
                }
        }
        lwkt_relpooltoken(kq);
        if (fp != NULL)
                fdrop(fp);
        return (error);
}

/*
 * Scan the kqueue, return the number of active events placed in kevp up
 * to count.
 *
 * Continuous mode events may get recycled, do not continue scanning past
 * marker unless no events have been collected.
 */
static int
kqueue_scan(struct kqueue *kq, struct kevent *kevp, int count,
            struct knote *marker)
{
        struct knote *kn, local_marker;
        int total;

        total = 0;
        local_marker.kn_filter = EVFILT_MARKER;
        local_marker.kn_status = KN_PROCESSING;

        lwkt_getpooltoken(kq);

        /*
         * Collect events.
         */
        TAILQ_INSERT_HEAD(&kq->kq_knpend, &local_marker, kn_tqe);
        while (count) {
                kn = TAILQ_NEXT(&local_marker, kn_tqe);
                if (kn->kn_filter == EVFILT_MARKER) {
                        /* Marker reached, we are done */
                        if (kn == marker)
                                break;

                        /* Move local marker past some other threads marker */
                        kn = TAILQ_NEXT(kn, kn_tqe);
                        TAILQ_REMOVE(&kq->kq_knpend, &local_marker, kn_tqe);
                        TAILQ_INSERT_BEFORE(kn, &local_marker, kn_tqe);
                        continue;
                }

                /*
                 * We can't skip a knote undergoing processing, otherwise
                 * we risk not returning it when the user process expects
                 * it should be returned.  Sleep and retry.
                 */
                if (knote_acquire(kn) == 0)
                        continue;

                /*
                 * Remove the event for processing.
                 *
                 * WARNING!  We must leave KN_QUEUED set to prevent the
                 *           event from being KNOTE_ACTIVATE()d while
                 *           the queue state is in limbo, in case we
                 *           block.
                 */
                TAILQ_REMOVE(&kq->kq_knpend, kn, kn_tqe);
                kq->kq_count--;

                /*
                 * We have to deal with an extremely important race against
                 * file descriptor close()s here.  The file descriptor can
                 * disappear MPSAFE, and there is a small window of
                 * opportunity between that and the call to knote_fdclose().
                 *
                 * If we hit that window here while doselect or dopoll is
                 * trying to delete a spurious event they will not be able
                 * to match up the event against a knote and will go haywire.
                 */
                if ((kn->kn_fop->f_flags & FILTEROP_ISFD) &&
                    checkfdclosed(kq->kq_fdp, kn->kn_kevent.ident, kn->kn_fp)) {
                        kn->kn_status |= KN_DELETING | KN_REPROCESS;
                }

                if (kn->kn_status & KN_DISABLED) {
                        /*
                         * If disabled we ensure the event is not queued
                         * but leave its active bit set.  On re-enablement
                         * the event may be immediately triggered.
                         */
                        kn->kn_status &= ~KN_QUEUED;
                } else if ((kn->kn_flags & EV_ONESHOT) == 0 &&
                           (kn->kn_status & KN_DELETING) == 0 &&
                           filter_event(kn, 0) == 0) {
                        /*
                         * If not running in one-shot mode and the event
                         * is no longer present we ensure it is removed
                         * from the queue and ignore it.
                         */
                        kn->kn_status &= ~(KN_QUEUED | KN_ACTIVE);
                } else {
                        /*
                         * Post the event
                         */
                        if (kn->kn_fop == &user_filtops)
                                filt_usertouch(kn, kevp, EVENT_PROCESS);
                        else
                                *kevp = kn->kn_kevent;
                        ++kevp;
                        ++total;
                        --count;

                        if (kn->kn_flags & EV_ONESHOT) {
                                kn->kn_status &= ~KN_QUEUED;
                                kn->kn_status |= KN_DELETING | KN_REPROCESS;
                        } else {
                                if (kn->kn_flags & (EV_CLEAR | EV_DISPATCH)) {
                                        if (kn->kn_flags & EV_CLEAR) {
                                                kn->kn_data = 0;
                                                kn->kn_fflags = 0;
                                        }
                                        if (kn->kn_flags & EV_DISPATCH) {
                                                kn->kn_status |= KN_DISABLED;
                                        }
                                        kn->kn_status &= ~(KN_QUEUED |
                                                           KN_ACTIVE);
                                } else {
                                        TAILQ_INSERT_TAIL(&kq->kq_knpend, kn, kn_tqe);
                                        kq->kq_count++;
                                }
                        }
                }

                /*
                 * Handle any post-processing states
                 */
                knote_release(kn);
        }
        TAILQ_REMOVE(&kq->kq_knpend, &local_marker, kn_tqe);

        lwkt_relpooltoken(kq);
        return (total);
}

/*
 * XXX
 * This could be expanded to call kqueue_scan, if desired.
 *
 * MPSAFE
 */
static int
kqueue_read(struct file *fp, struct uio *uio, struct ucred *cred, int flags)
{
        return (ENXIO);
}

/*
 * MPSAFE
 */
static int
kqueue_write(struct file *fp, struct uio *uio, struct ucred *cred, int flags)
{
        return (ENXIO);
}

/*
 * MPALMOSTSAFE
 */
static int
kqueue_ioctl(struct file *fp, u_long com, caddr_t data,
             struct ucred *cred, struct sysmsg *msg)
{
        struct lwkt_token *tok;
        struct kqueue *kq;
        int error;

        kq = (struct kqueue *)fp->f_data;
        tok = lwkt_token_pool_lookup(kq);
        lwkt_gettoken(tok);

        switch(com) {
        case FIOASYNC:
                if (*(int *)data)
                        kq->kq_state |= KQ_ASYNC;
                else
                        kq->kq_state &= ~KQ_ASYNC;
                error = 0;
                break;
        case FIOSETOWN:
                error = fsetown(*(int *)data, &kq->kq_sigio);
                break;
        default:
                error = ENOTTY;
                break;
        }
        lwkt_reltoken(tok);
        return (error);
}

/*
 * MPSAFE
 */
static int
kqueue_stat(struct file *fp, struct stat *st, struct ucred *cred)
{
        struct kqueue *kq = (struct kqueue *)fp->f_data;

        bzero((void *)st, sizeof(*st));
        st->st_size = kq->kq_count;
        st->st_blksize = sizeof(struct kevent);
        st->st_mode = S_IFIFO;
        return (0);
}

/*
 * MPSAFE
 */
static int
kqueue_close(struct file *fp)
{
        struct kqueue *kq = (struct kqueue *)fp->f_data;

        kqueue_terminate(kq);

        fp->f_data = NULL;
        funsetown(&kq->kq_sigio);

        kfree(kq, M_KQUEUE);
        return (0);
}

static void
kqueue_wakeup(struct kqueue *kq)
{
        if (kq->kq_sleep_cnt) {
                if (kq->kq_sleep_cnt == 1)
                        wakeup_one(kq);
                else
                        wakeup(kq);
                kq->kq_sleep_cnt = 0;
        }
        KNOTE(&kq->kq_kqinfo.ki_note, 0);
}

/*
 * Calls filterops f_attach function, acquiring mplock if filter is not
 * marked as FILTEROP_MPSAFE.
 *
 * Caller must be holding the related kq token
 */
static int
filter_attach(struct knote *kn)
{
        int ret;

        if (kn->kn_fop->f_flags & FILTEROP_MPSAFE) {
                ret = kn->kn_fop->f_attach(kn);
        } else {
                get_mplock();
                ret = kn->kn_fop->f_attach(kn);
                rel_mplock();
        }
        return (ret);
}

/*
 * Detach the knote and drop it, destroying the knote.
 *
 * Calls filterops f_detach function, acquiring mplock if filter is not
 * marked as FILTEROP_MPSAFE.
 *
 * Caller must be holding the related kq token
 */
static void
knote_detach_and_drop(struct knote *kn)
{
        kn->kn_status |= KN_DELETING | KN_REPROCESS;
        if (kn->kn_fop->f_flags & FILTEROP_MPSAFE) {
                kn->kn_fop->f_detach(kn);
        } else {
                get_mplock();
                kn->kn_fop->f_detach(kn);
                rel_mplock();
        }
        knote_drop(kn);
}

/*
 * Calls filterops f_event function, acquiring mplock if filter is not
 * marked as FILTEROP_MPSAFE.
 *
 * If the knote is in the middle of being created or deleted we cannot
 * safely call the filter op.
 *
 * Caller must be holding the related kq token
 */
static int
filter_event(struct knote *kn, long hint)
{
        int ret;

        if (kn->kn_fop->f_flags & FILTEROP_MPSAFE) {
                ret = kn->kn_fop->f_event(kn, hint);
        } else {
                get_mplock();
                ret = kn->kn_fop->f_event(kn, hint);
                rel_mplock();
        }
        return (ret);
}

/*
 * Walk down a list of knotes, activating them if their event has triggered.
 *
 * If we encounter any knotes which are undergoing processing we just mark
 * them for reprocessing and do not try to [re]activate the knote.  However,
 * if a hint is being passed we have to wait and that makes things a bit
 * sticky.
 */
void
knote(struct klist *list, long hint)
{
        struct knote *kn, marker;

        marker.kn_filter = EVFILT_MARKER;
        marker.kn_status = KN_PROCESSING;

        lwkt_getpooltoken(list);
        if (SLIST_EMPTY(list)) {
                lwkt_relpooltoken(list);
                return;
        }

        SLIST_INSERT_HEAD(list, &marker, kn_next);
        while ((kn = SLIST_NEXT(&marker, kn_next)) != NULL) {
                struct kqueue *kq;
                int last_knote = 0;

                if (kn->kn_filter == EVFILT_MARKER) {
                        /* Skip marker */
                        SLIST_REMOVE(list, &marker, knote, kn_next);
                        if (SLIST_NEXT(kn, kn_next) == NULL)
                                goto done;
                        SLIST_INSERT_AFTER(kn, &marker, kn_next);
                        continue;
                }

                kq = kn->kn_kq;
                lwkt_getpooltoken(kq);

                if (kn != SLIST_NEXT(&marker, kn_next) || kn->kn_kq != kq) {
                        /*
                         * Don't move the marker; check the knote after
                         * the marker again.
                         */
                        lwkt_relpooltoken(kq);
                        continue;
                }

                if (kn->kn_status & KN_PROCESSING) {
                        /*
                         * Someone else is processing the knote, ask the
                         * other thread to reprocess it and don't mess
                         * with it otherwise.
                         */
                        if (hint == 0) {
                                /*
                                 * Move the marker w/ the kq token, so that
                                 * this knote will not be ripped behind our
                                 * back.
                                 */
                                SLIST_REMOVE(list, &marker, knote, kn_next);
                                if (SLIST_NEXT(kn, kn_next) != NULL)
                                        SLIST_INSERT_AFTER(kn, &marker, kn_next);
                                else
                                        last_knote = 1;
                                kn->kn_status |= KN_REPROCESS;
                                lwkt_relpooltoken(kq);

                                if (last_knote)
                                        goto done;
                                continue;
                        }

                        /*
                         * If the hint is non-zero we have to wait or risk
                         * losing the state the caller is trying to update.
                         */
                        kn->kn_status |= KN_WAITING | KN_REPROCESS;
                        tsleep(kn, 0, "knotec", hz);

                        /*
                         * Don't move the marker; check this knote again,
                         * hopefully it is still after the marker.  Or it
                         * was deleted and we would check the next knote.
                         */
                        lwkt_relpooltoken(kq);
                        continue;
                }

                /*
                 * Become the reprocessing master ourselves.
                 */
                KASSERT((kn->kn_status & KN_DELETING) == 0,
                    ("acquire a deleting knote %#x", kn->kn_status));
                kn->kn_status |= KN_PROCESSING;

                /* Move the marker */
                SLIST_REMOVE(list, &marker, knote, kn_next);
                if (SLIST_NEXT(kn, kn_next) != NULL)
                        SLIST_INSERT_AFTER(kn, &marker, kn_next);
                else
                        last_knote = 1;

                /*
                 * If hint is non-zero running the event is mandatory
                 * so do it whether reprocessing is set or not.
                 */
                if (filter_event(kn, hint))
                        KNOTE_ACTIVATE(kn);

                knote_release(kn);
                lwkt_relpooltoken(kq);

                if (last_knote)
                        goto done;
        }
        SLIST_REMOVE(list, &marker, knote, kn_next);
done:
        lwkt_relpooltoken(list);
}

/*
 * Insert knote at head of klist.
 *
 * This function may only be called via a filter function and thus
 * kq_token should already be held and marked for processing.
 */
void
knote_insert(struct klist *klist, struct knote *kn)
{
        lwkt_getpooltoken(klist);
        KKASSERT(kn->kn_status & KN_PROCESSING);
        SLIST_INSERT_HEAD(klist, kn, kn_next);
        lwkt_relpooltoken(klist);
}

/*
 * Remove knote from a klist
 *
 * This function may only be called via a filter function and thus
 * kq_token should already be held and marked for processing.
 */
void
knote_remove(struct klist *klist, struct knote *kn)
{
        lwkt_getpooltoken(klist);
        KKASSERT(kn->kn_status & KN_PROCESSING);
        SLIST_REMOVE(klist, kn, knote, kn_next);
        lwkt_relpooltoken(klist);
}

void
knote_assume_knotes(struct kqinfo *src, struct kqinfo *dst,
                    struct filterops *ops, void *hook)
{
        struct knote *kn, marker;
        int has_note;

        marker.kn_filter = EVFILT_MARKER;
        marker.kn_status = KN_PROCESSING;

        lwkt_getpooltoken(&src->ki_note);
        if (SLIST_EMPTY(&src->ki_note)) {
                lwkt_relpooltoken(&src->ki_note);
                return;
        }
        lwkt_getpooltoken(&dst->ki_note);

restart:
        has_note = 0;
        SLIST_INSERT_HEAD(&src->ki_note, &marker, kn_next);
        while ((kn = SLIST_NEXT(&marker, kn_next)) != NULL) {
                struct kqueue *kq;

                if (kn->kn_filter == EVFILT_MARKER) {
                        /* Skip marker */
                        SLIST_REMOVE(&src->ki_note, &marker, knote, kn_next);
                        SLIST_INSERT_AFTER(kn, &marker, kn_next);
                        continue;
                }

                kq = kn->kn_kq;
                lwkt_getpooltoken(kq);

                if (kn != SLIST_NEXT(&marker, kn_next) || kn->kn_kq != kq) {
                        /*
                         * Don't move the marker; check the knote after
                         * the marker again.
                         */
                        lwkt_relpooltoken(kq);
                        continue;
                }

                /* Move marker */
                SLIST_REMOVE(&src->ki_note, &marker, knote, kn_next);
                SLIST_INSERT_AFTER(kn, &marker, kn_next);

                has_note = 1;
                if (knote_acquire(kn)) {
                        knote_remove(&src->ki_note, kn);
                        kn->kn_fop = ops;
                        kn->kn_hook = hook;
                        knote_insert(&dst->ki_note, kn);
                        knote_release(kn);
                        /* kn may be invalid now */
                }
                lwkt_relpooltoken(kq);
        }
        SLIST_REMOVE(&src->ki_note, &marker, knote, kn_next);
        if (has_note) {
                /* Keep draining, until nothing left */
                goto restart;
        }

        lwkt_relpooltoken(&dst->ki_note);
        lwkt_relpooltoken(&src->ki_note);
}

/*
 * Remove all knotes referencing a specified fd
 */
void
knote_fdclose(struct file *fp, struct filedesc *fdp, int fd)
{
        struct kqueue *kq;
        struct knote *kn;
        struct knote *kntmp;

        lwkt_getpooltoken(&fp->f_klist);
restart:
        SLIST_FOREACH(kn, &fp->f_klist, kn_link) {
                if (kn->kn_kq->kq_fdp == fdp && kn->kn_id == fd) {
                        kq = kn->kn_kq;
                        lwkt_getpooltoken(kq);

                        /* temporary verification hack */
                        SLIST_FOREACH(kntmp, &fp->f_klist, kn_link) {
                                if (kn == kntmp)
                                        break;
                        }
                        if (kn != kntmp || kn->kn_kq->kq_fdp != fdp ||
                            kn->kn_id != fd || kn->kn_kq != kq) {
                                lwkt_relpooltoken(kq);
                                goto restart;
                        }
                        if (knote_acquire(kn))
                                knote_detach_and_drop(kn);
                        lwkt_relpooltoken(kq);
                        goto restart;
                }
        }
        lwkt_relpooltoken(&fp->f_klist);
}

/*
 * Low level attach function.
 *
 * The knote should already be marked for processing.
 * Caller must hold the related kq token.
 */
static void
knote_attach(struct knote *kn)
{
        struct klist *list;
        struct kqueue *kq = kn->kn_kq;

        if (kn->kn_fop->f_flags & FILTEROP_ISFD) {
                KKASSERT(kn->kn_fp);
                list = &kn->kn_fp->f_klist;
        } else {
                if (kq->kq_knhashmask == 0)
                        kq->kq_knhash = hashinit(KN_HASHSIZE, M_KQUEUE,
                                                 &kq->kq_knhashmask);
                list = &kq->kq_knhash[KN_HASH(kn->kn_id, kq->kq_knhashmask)];
        }
        lwkt_getpooltoken(list);
        SLIST_INSERT_HEAD(list, kn, kn_link);
        lwkt_relpooltoken(list);
        TAILQ_INSERT_HEAD(&kq->kq_knlist, kn, kn_kqlink);
}

/*
 * Low level drop function.
 *
 * The knote should already be marked for processing.
 * Caller must hold the related kq token.
 */
static void
knote_drop(struct knote *kn)
{
        struct kqueue *kq;
        struct klist *list;

        kq = kn->kn_kq;

        if (kn->kn_fop->f_flags & FILTEROP_ISFD)
                list = &kn->kn_fp->f_klist;
        else
                list = &kq->kq_knhash[KN_HASH(kn->kn_id, kq->kq_knhashmask)];

        lwkt_getpooltoken(list);
        SLIST_REMOVE(list, kn, knote, kn_link);
        lwkt_relpooltoken(list);
        TAILQ_REMOVE(&kq->kq_knlist, kn, kn_kqlink);
        if (kn->kn_status & KN_QUEUED)
                knote_dequeue(kn);
        if (kn->kn_fop->f_flags & FILTEROP_ISFD) {
                fdrop(kn->kn_fp);
                kn->kn_fp = NULL;
        }
        knote_free(kn);
}

/*
 * Low level enqueue function.
 *
 * The knote should already be marked for processing.
 * Caller must be holding the kq token
 */
static void
knote_enqueue(struct knote *kn)
{
        struct kqueue *kq = kn->kn_kq;

        KASSERT((kn->kn_status & KN_QUEUED) == 0, ("knote already queued"));
        TAILQ_INSERT_TAIL(&kq->kq_knpend, kn, kn_tqe);
        kn->kn_status |= KN_QUEUED;
        ++kq->kq_count;

        /*
         * Send SIGIO on request (typically set up as a mailbox signal)
         */
        if (kq->kq_sigio && (kq->kq_state & KQ_ASYNC) && kq->kq_count == 1)
                pgsigio(kq->kq_sigio, SIGIO, 0);

        kqueue_wakeup(kq);
}

/*
 * Low level dequeue function.
 *
 * The knote should already be marked for processing.
 * Caller must be holding the kq token
 */
static void
knote_dequeue(struct knote *kn)
{
        struct kqueue *kq = kn->kn_kq;

        KASSERT(kn->kn_status & KN_QUEUED, ("knote not queued"));
        TAILQ_REMOVE(&kq->kq_knpend, kn, kn_tqe);
        kn->kn_status &= ~KN_QUEUED;
        kq->kq_count--;
}

static struct knote *
knote_alloc(void)
{
        return kmalloc(sizeof(struct knote), M_KQUEUE, M_WAITOK);
}

static void
knote_free(struct knote *kn)
{
        struct knote_cache_list *cache_list;

        cache_list = &knote_cache_lists[mycpuid];
        if (cache_list->knote_cache_cnt < KNOTE_CACHE_MAX) {
                SLIST_INSERT_HEAD(&cache_list->knote_cache, kn, kn_link);
                cache_list->knote_cache_cnt++;
                return;
        }
        kfree(kn, M_KQUEUE);
}