diff --git a/sys/kern/kern_event.c b/sys/kern/kern_event.c index f1ee8de4e7ee..4b4522704ad0 100644 --- a/sys/kern/kern_event.c +++ b/sys/kern/kern_event.c @@ -1,2853 +1,2857 @@ /*- * SPDX-License-Identifier: BSD-2-Clause-FreeBSD * * Copyright (c) 1999,2000,2001 Jonathan Lemon * Copyright 2004 John-Mark Gurney * Copyright (c) 2009 Apple, Inc. * 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. */ #include __FBSDID("$FreeBSD$"); #include "opt_ktrace.h" #include "opt_kqueue.h" #ifdef COMPAT_FREEBSD11 #define _WANT_FREEBSD11_KEVENT #endif #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #ifdef KTRACE #include #endif #include #include static MALLOC_DEFINE(M_KQUEUE, "kqueue", "memory for kqueue system"); /* * This lock is used if multiple kq locks are required. This possibly * should be made into a per proc lock. */ static struct mtx kq_global; MTX_SYSINIT(kq_global, &kq_global, "kqueue order", MTX_DEF); #define KQ_GLOBAL_LOCK(lck, haslck) do { \ if (!haslck) \ mtx_lock(lck); \ haslck = 1; \ } while (0) #define KQ_GLOBAL_UNLOCK(lck, haslck) do { \ if (haslck) \ mtx_unlock(lck); \ haslck = 0; \ } while (0) TASKQUEUE_DEFINE_THREAD(kqueue_ctx); static int kevent_copyout(void *arg, struct kevent *kevp, int count); static int kevent_copyin(void *arg, struct kevent *kevp, int count); static int kqueue_register(struct kqueue *kq, struct kevent *kev, struct thread *td, int mflag); static int kqueue_acquire(struct file *fp, struct kqueue **kqp); static void kqueue_release(struct kqueue *kq, int locked); static void kqueue_destroy(struct kqueue *kq); static void kqueue_drain(struct kqueue *kq, struct thread *td); static int kqueue_expand(struct kqueue *kq, struct filterops *fops, uintptr_t ident, int mflag); static void kqueue_task(void *arg, int pending); static int kqueue_scan(struct kqueue *kq, int maxevents, struct kevent_copyops *k_ops, const struct timespec *timeout, struct kevent *keva, struct thread *td); static void kqueue_wakeup(struct kqueue *kq); static struct filterops *kqueue_fo_find(int filt); static void kqueue_fo_release(int filt); struct g_kevent_args; static int kern_kevent_generic(struct thread *td, struct g_kevent_args *uap, struct kevent_copyops *k_ops, const char *struct_name); static fo_ioctl_t kqueue_ioctl; static fo_poll_t kqueue_poll; static fo_kqfilter_t kqueue_kqfilter; static fo_stat_t kqueue_stat; static fo_close_t kqueue_close; static fo_fill_kinfo_t kqueue_fill_kinfo; static struct fileops kqueueops = { .fo_read = invfo_rdwr, .fo_write = invfo_rdwr, .fo_truncate = invfo_truncate, .fo_ioctl = kqueue_ioctl, .fo_poll = kqueue_poll, .fo_kqfilter = kqueue_kqfilter, .fo_stat = kqueue_stat, .fo_close = kqueue_close, .fo_chmod = invfo_chmod, .fo_chown = invfo_chown, .fo_sendfile = invfo_sendfile, .fo_fill_kinfo = kqueue_fill_kinfo, }; static int knote_attach(struct knote *kn, struct kqueue *kq); static void knote_drop(struct knote *kn, struct thread *td); static void knote_drop_detached(struct knote *kn, struct thread *td); static void knote_enqueue(struct knote *kn); static void knote_dequeue(struct knote *kn); static void knote_init(void); static struct knote *knote_alloc(int mflag); 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 void filt_timerexpire_l(struct knote *kn, bool proc_locked); static int filt_timerattach(struct knote *kn); static void filt_timerdetach(struct knote *kn); static void filt_timerstart(struct knote *kn, sbintime_t to); static void filt_timertouch(struct knote *kn, struct kevent *kev, u_long type); static int filt_timervalidate(struct knote *kn, sbintime_t *to); 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 = { .f_isfd = 1, .f_attach = filt_fileattach, }; static struct filterops kqread_filtops = { .f_isfd = 1, .f_detach = filt_kqdetach, .f_event = filt_kqueue, }; /* XXX - move to kern_proc.c? */ static struct filterops proc_filtops = { .f_isfd = 0, .f_attach = filt_procattach, .f_detach = filt_procdetach, .f_event = filt_proc, }; static struct filterops timer_filtops = { .f_isfd = 0, .f_attach = filt_timerattach, .f_detach = filt_timerdetach, .f_event = filt_timer, .f_touch = filt_timertouch, }; static struct filterops user_filtops = { .f_attach = filt_userattach, .f_detach = filt_userdetach, .f_event = filt_user, .f_touch = filt_usertouch, }; static uma_zone_t knote_zone; static unsigned int __exclusive_cache_line kq_ncallouts; static unsigned int kq_calloutmax = 4 * 1024; SYSCTL_UINT(_kern, OID_AUTO, kq_calloutmax, CTLFLAG_RW, &kq_calloutmax, 0, "Maximum number of callouts allocated for kqueue"); /* XXX - ensure not influx ? */ #define KNOTE_ACTIVATE(kn, islock) do { \ if ((islock)) \ mtx_assert(&(kn)->kn_kq->kq_lock, MA_OWNED); \ else \ KQ_LOCK((kn)->kn_kq); \ (kn)->kn_status |= KN_ACTIVE; \ if (((kn)->kn_status & (KN_QUEUED | KN_DISABLED)) == 0) \ knote_enqueue((kn)); \ if (!(islock)) \ KQ_UNLOCK((kn)->kn_kq); \ } while (0) #define KQ_LOCK(kq) do { \ mtx_lock(&(kq)->kq_lock); \ } while (0) #define KQ_FLUX_WAKEUP(kq) do { \ if (((kq)->kq_state & KQ_FLUXWAIT) == KQ_FLUXWAIT) { \ (kq)->kq_state &= ~KQ_FLUXWAIT; \ wakeup((kq)); \ } \ } while (0) #define KQ_UNLOCK_FLUX(kq) do { \ KQ_FLUX_WAKEUP(kq); \ mtx_unlock(&(kq)->kq_lock); \ } while (0) #define KQ_UNLOCK(kq) do { \ mtx_unlock(&(kq)->kq_lock); \ } while (0) #define KQ_OWNED(kq) do { \ mtx_assert(&(kq)->kq_lock, MA_OWNED); \ } while (0) #define KQ_NOTOWNED(kq) do { \ mtx_assert(&(kq)->kq_lock, MA_NOTOWNED); \ } while (0) static struct knlist * kn_list_lock(struct knote *kn) { struct knlist *knl; knl = kn->kn_knlist; if (knl != NULL) knl->kl_lock(knl->kl_lockarg); return (knl); } static void kn_list_unlock(struct knlist *knl) { bool do_free; if (knl == NULL) return; do_free = knl->kl_autodestroy && knlist_empty(knl); knl->kl_unlock(knl->kl_lockarg); if (do_free) { knlist_destroy(knl); free(knl, M_KQUEUE); } } static bool kn_in_flux(struct knote *kn) { return (kn->kn_influx > 0); } static void kn_enter_flux(struct knote *kn) { KQ_OWNED(kn->kn_kq); MPASS(kn->kn_influx < INT_MAX); kn->kn_influx++; } static bool kn_leave_flux(struct knote *kn) { KQ_OWNED(kn->kn_kq); MPASS(kn->kn_influx > 0); kn->kn_influx--; return (kn->kn_influx == 0); } #define KNL_ASSERT_LOCK(knl, islocked) do { \ if (islocked) \ KNL_ASSERT_LOCKED(knl); \ else \ KNL_ASSERT_UNLOCKED(knl); \ } while (0) #ifdef INVARIANTS #define KNL_ASSERT_LOCKED(knl) do { \ knl->kl_assert_lock((knl)->kl_lockarg, LA_LOCKED); \ } while (0) #define KNL_ASSERT_UNLOCKED(knl) do { \ knl->kl_assert_lock((knl)->kl_lockarg, LA_UNLOCKED); \ } while (0) #else /* !INVARIANTS */ #define KNL_ASSERT_LOCKED(knl) do {} while (0) #define KNL_ASSERT_UNLOCKED(knl) do {} while (0) #endif /* INVARIANTS */ #ifndef KN_HASHSIZE #define KN_HASHSIZE 64 /* XXX should be tunable */ #endif #define KN_HASH(val, mask) (((val) ^ (val >> 8)) & (mask)) static int filt_nullattach(struct knote *kn) { return (ENXIO); }; struct filterops null_filtops = { .f_isfd = 0, .f_attach = filt_nullattach, }; /* XXX - make SYSINIT to add these, and move into respective modules. */ extern struct filterops sig_filtops; extern struct filterops fs_filtops; /* * Table for all system-defined filters. */ static struct mtx filterops_lock; MTX_SYSINIT(kqueue_filterops, &filterops_lock, "protect sysfilt_ops", MTX_DEF); static struct { struct filterops *for_fop; int for_nolock; int for_refcnt; } sysfilt_ops[EVFILT_SYSCOUNT] = { { &file_filtops, 1 }, /* EVFILT_READ */ { &file_filtops, 1 }, /* EVFILT_WRITE */ { &null_filtops }, /* EVFILT_AIO */ { &file_filtops, 1 }, /* EVFILT_VNODE */ { &proc_filtops, 1 }, /* EVFILT_PROC */ { &sig_filtops, 1 }, /* EVFILT_SIGNAL */ { &timer_filtops, 1 }, /* EVFILT_TIMER */ { &file_filtops, 1 }, /* EVFILT_PROCDESC */ { &fs_filtops, 1 }, /* EVFILT_FS */ { &null_filtops }, /* EVFILT_LIO */ { &user_filtops, 1 }, /* EVFILT_USER */ { &null_filtops }, /* EVFILT_SENDFILE */ { &file_filtops, 1 }, /* EVFILT_EMPTY */ }; /* * Simple redirection for all cdevsw style objects to call their fo_kqfilter * method. */ static int filt_fileattach(struct knote *kn) { return (fo_kqfilter(kn->kn_fp, kn)); } /*ARGSUSED*/ static int kqueue_kqfilter(struct file *fp, struct knote *kn) { struct kqueue *kq = kn->kn_fp->f_data; if (kn->kn_filter != EVFILT_READ) return (EINVAL); kn->kn_status |= KN_KQUEUE; kn->kn_fop = &kqread_filtops; knlist_add(&kq->kq_sel.si_note, kn, 0); return (0); } static void filt_kqdetach(struct knote *kn) { struct kqueue *kq = kn->kn_fp->f_data; knlist_remove(&kq->kq_sel.si_note, kn, 0); } /*ARGSUSED*/ static int filt_kqueue(struct knote *kn, long hint) { struct kqueue *kq = kn->kn_fp->f_data; kn->kn_data = kq->kq_count; return (kn->kn_data > 0); } /* XXX - move to kern_proc.c? */ static int filt_procattach(struct knote *kn) { struct proc *p; int error; bool exiting, immediate; exiting = immediate = false; if (kn->kn_sfflags & NOTE_EXIT) p = pfind_any(kn->kn_id); else p = pfind(kn->kn_id); if (p == NULL) return (ESRCH); if (p->p_flag & P_WEXIT) exiting = true; if ((error = p_cansee(curthread, p))) { PROC_UNLOCK(p); return (error); } kn->kn_ptr.p_proc = p; kn->kn_flags |= EV_CLEAR; /* automatically set */ /* * Internal flag indicating registration done by kernel for the * purposes of getting a NOTE_CHILD notification. */ if (kn->kn_flags & EV_FLAG2) { kn->kn_flags &= ~EV_FLAG2; kn->kn_data = kn->kn_sdata; /* ppid */ kn->kn_fflags = NOTE_CHILD; kn->kn_sfflags &= ~(NOTE_EXIT | NOTE_EXEC | NOTE_FORK); immediate = true; /* Force immediate activation of child note. */ } /* * Internal flag indicating registration done by kernel (for other than * NOTE_CHILD). */ if (kn->kn_flags & EV_FLAG1) { kn->kn_flags &= ~EV_FLAG1; } knlist_add(p->p_klist, kn, 1); /* * Immediately activate any child notes or, in the case of a zombie * target process, exit notes. The latter is necessary to handle the * case where the target process, e.g. a child, dies before the kevent * is registered. */ if (immediate || (exiting && filt_proc(kn, NOTE_EXIT))) KNOTE_ACTIVATE(kn, 0); PROC_UNLOCK(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. */ /* XXX - move to kern_proc.c? */ static void filt_procdetach(struct knote *kn) { knlist_remove(kn->kn_knlist, kn, 0); kn->kn_ptr.p_proc = NULL; } /* XXX - move to kern_proc.c? */ static int filt_proc(struct knote *kn, long hint) { struct proc *p; u_int event; p = kn->kn_ptr.p_proc; if (p == NULL) /* already activated, from attach filter */ return (0); /* 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. */ if (event == NOTE_EXIT) { kn->kn_flags |= EV_EOF | EV_ONESHOT; kn->kn_ptr.p_proc = NULL; if (kn->kn_fflags & NOTE_EXIT) kn->kn_data = KW_EXITCODE(p->p_xexit, p->p_xsig); if (kn->kn_fflags == 0) kn->kn_flags |= EV_DROP; return (1); } return (kn->kn_fflags != 0); } /* * Called when the process forked. It mostly does the same as the * knote(), activating all knotes registered to be activated when the * process forked. Additionally, for each knote attached to the * parent, check whether user wants to track the new process. If so * attach a new knote to it, and immediately report an event with the * child's pid. */ void knote_fork(struct knlist *list, int pid) { struct kqueue *kq; struct knote *kn; struct kevent kev; int error; MPASS(list != NULL); KNL_ASSERT_LOCKED(list); if (SLIST_EMPTY(&list->kl_list)) return; memset(&kev, 0, sizeof(kev)); SLIST_FOREACH(kn, &list->kl_list, kn_selnext) { kq = kn->kn_kq; KQ_LOCK(kq); if (kn_in_flux(kn) && (kn->kn_status & KN_SCAN) == 0) { KQ_UNLOCK(kq); continue; } /* * The same as knote(), activate the event. */ if ((kn->kn_sfflags & NOTE_TRACK) == 0) { if (kn->kn_fop->f_event(kn, NOTE_FORK)) KNOTE_ACTIVATE(kn, 1); KQ_UNLOCK(kq); continue; } /* * The NOTE_TRACK case. In addition to the activation * of the event, we need to register new events to * track the child. Drop the locks in preparation for * the call to kqueue_register(). */ kn_enter_flux(kn); KQ_UNLOCK(kq); list->kl_unlock(list->kl_lockarg); /* * Activate existing knote and register tracking knotes with * new process. * * First register a knote to get just the child notice. This * must be a separate note from a potential NOTE_EXIT * notification since both NOTE_CHILD and NOTE_EXIT are defined * to use the data field (in conflicting ways). */ kev.ident = pid; kev.filter = kn->kn_filter; kev.flags = kn->kn_flags | EV_ADD | EV_ENABLE | EV_ONESHOT | EV_FLAG2; kev.fflags = kn->kn_sfflags; kev.data = kn->kn_id; /* parent */ kev.udata = kn->kn_kevent.udata;/* preserve udata */ error = kqueue_register(kq, &kev, NULL, M_NOWAIT); if (error) kn->kn_fflags |= NOTE_TRACKERR; /* * Then register another knote to track other potential events * from the new process. */ kev.ident = 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(kq, &kev, NULL, M_NOWAIT); if (error) kn->kn_fflags |= NOTE_TRACKERR; if (kn->kn_fop->f_event(kn, NOTE_FORK)) KNOTE_ACTIVATE(kn, 0); list->kl_lock(list->kl_lockarg); KQ_LOCK(kq); kn_leave_flux(kn); KQ_UNLOCK_FLUX(kq); } } /* * XXX: EVFILT_TIMER should perhaps live in kern_time.c beside the * interval timer support code. */ #define NOTE_TIMER_PRECMASK \ (NOTE_SECONDS | NOTE_MSECONDS | NOTE_USECONDS | NOTE_NSECONDS) static sbintime_t timer2sbintime(int64_t data, int flags) { int64_t secs; /* * Macros for converting to the fractional second portion of an * sbintime_t using 64bit multiplication to improve precision. */ #define NS_TO_SBT(ns) (((ns) * (((uint64_t)1 << 63) / 500000000)) >> 32) #define US_TO_SBT(us) (((us) * (((uint64_t)1 << 63) / 500000)) >> 32) #define MS_TO_SBT(ms) (((ms) * (((uint64_t)1 << 63) / 500)) >> 32) switch (flags & NOTE_TIMER_PRECMASK) { case NOTE_SECONDS: #ifdef __LP64__ if (data > (SBT_MAX / SBT_1S)) return (SBT_MAX); #endif return ((sbintime_t)data << 32); case NOTE_MSECONDS: /* FALLTHROUGH */ case 0: if (data >= 1000) { secs = data / 1000; #ifdef __LP64__ if (secs > (SBT_MAX / SBT_1S)) return (SBT_MAX); #endif return (secs << 32 | MS_TO_SBT(data % 1000)); } return (MS_TO_SBT(data)); case NOTE_USECONDS: if (data >= 1000000) { secs = data / 1000000; #ifdef __LP64__ if (secs > (SBT_MAX / SBT_1S)) return (SBT_MAX); #endif return (secs << 32 | US_TO_SBT(data % 1000000)); } return (US_TO_SBT(data)); case NOTE_NSECONDS: if (data >= 1000000000) { secs = data / 1000000000; #ifdef __LP64__ if (secs > (SBT_MAX / SBT_1S)) return (SBT_MAX); #endif return (secs << 32 | NS_TO_SBT(data % 1000000000)); } return (NS_TO_SBT(data)); default: break; } return (-1); } struct kq_timer_cb_data { struct callout c; struct proc *p; struct knote *kn; int cpuid; int flags; TAILQ_ENTRY(kq_timer_cb_data) link; sbintime_t next; /* next timer event fires at */ sbintime_t to; /* precalculated timer period, 0 for abs */ }; #define KQ_TIMER_CB_ENQUEUED 0x01 static void kqtimer_sched_callout(struct kq_timer_cb_data *kc) { callout_reset_sbt_on(&kc->c, kc->next, 0, filt_timerexpire, kc->kn, kc->cpuid, C_ABSOLUTE); } void kqtimer_proc_continue(struct proc *p) { struct kq_timer_cb_data *kc, *kc1; struct bintime bt; sbintime_t now; PROC_LOCK_ASSERT(p, MA_OWNED); getboottimebin(&bt); now = bttosbt(bt); TAILQ_FOREACH_SAFE(kc, &p->p_kqtim_stop, link, kc1) { TAILQ_REMOVE(&p->p_kqtim_stop, kc, link); kc->flags &= ~KQ_TIMER_CB_ENQUEUED; if (kc->next <= now) filt_timerexpire_l(kc->kn, true); else kqtimer_sched_callout(kc); } } static void filt_timerexpire_l(struct knote *kn, bool proc_locked) { struct kq_timer_cb_data *kc; struct proc *p; uint64_t delta; sbintime_t now; kc = kn->kn_ptr.p_v; if ((kn->kn_flags & EV_ONESHOT) != 0 || kc->to == 0) { kn->kn_data++; KNOTE_ACTIVATE(kn, 0); return; } now = sbinuptime(); if (now >= kc->next) { delta = (now - kc->next) / kc->to; if (delta == 0) delta = 1; kn->kn_data += delta; kc->next += delta * kc->to; if (now >= kc->next) /* overflow */ kc->next = now + kc->to; KNOTE_ACTIVATE(kn, 0); /* XXX - handle locking */ } /* * Initial check for stopped kc->p is racy. It is fine to * miss the set of the stop flags, at worst we would schedule * one more callout. On the other hand, it is not fine to not * schedule when we we missed clearing of the flags, we * recheck them under the lock and observe consistent state. */ p = kc->p; if (P_SHOULDSTOP(p) || P_KILLED(p)) { if (!proc_locked) PROC_LOCK(p); if (P_SHOULDSTOP(p) || P_KILLED(p)) { if ((kc->flags & KQ_TIMER_CB_ENQUEUED) == 0) { kc->flags |= KQ_TIMER_CB_ENQUEUED; TAILQ_INSERT_TAIL(&p->p_kqtim_stop, kc, link); } if (!proc_locked) PROC_UNLOCK(p); return; } if (!proc_locked) PROC_UNLOCK(p); } kqtimer_sched_callout(kc); } static void filt_timerexpire(void *knx) { filt_timerexpire_l(knx, false); } /* * data contains amount of time to sleep */ static int filt_timervalidate(struct knote *kn, sbintime_t *to) { struct bintime bt; sbintime_t sbt; if (kn->kn_sdata < 0) return (EINVAL); if (kn->kn_sdata == 0 && (kn->kn_flags & EV_ONESHOT) == 0) kn->kn_sdata = 1; /* * The only fflags values supported are the timer unit * (precision) and the absolute time indicator. */ if ((kn->kn_sfflags & ~(NOTE_TIMER_PRECMASK | NOTE_ABSTIME)) != 0) return (EINVAL); *to = timer2sbintime(kn->kn_sdata, kn->kn_sfflags); if (*to < 0) return (EINVAL); if ((kn->kn_sfflags & NOTE_ABSTIME) != 0) { getboottimebin(&bt); sbt = bttosbt(bt); *to = MAX(0, *to - sbt); } return (0); } static int filt_timerattach(struct knote *kn) { struct kq_timer_cb_data *kc; sbintime_t to; int error; to = -1; error = filt_timervalidate(kn, &to); if (error != 0) return (error); KASSERT(to > 0 || (kn->kn_flags & EV_ONESHOT) != 0 || (kn->kn_sfflags & NOTE_ABSTIME) != 0, ("%s: periodic timer has a calculated zero timeout", __func__)); KASSERT(to >= 0, ("%s: timer has a calculated negative timeout", __func__)); if (atomic_fetchadd_int(&kq_ncallouts, 1) + 1 > kq_calloutmax) { atomic_subtract_int(&kq_ncallouts, 1); return (ENOMEM); } if ((kn->kn_sfflags & NOTE_ABSTIME) == 0) kn->kn_flags |= EV_CLEAR; /* automatically set */ kn->kn_status &= ~KN_DETACHED; /* knlist_add clears it */ kn->kn_ptr.p_v = kc = malloc(sizeof(*kc), M_KQUEUE, M_WAITOK); kc->kn = kn; kc->p = curproc; kc->cpuid = PCPU_GET(cpuid); kc->flags = 0; callout_init(&kc->c, 1); filt_timerstart(kn, to); return (0); } static void filt_timerstart(struct knote *kn, sbintime_t to) { struct kq_timer_cb_data *kc; kc = kn->kn_ptr.p_v; if ((kn->kn_sfflags & NOTE_ABSTIME) != 0) { kc->next = to; kc->to = 0; } else { kc->next = to + sbinuptime(); kc->to = to; } kqtimer_sched_callout(kc); } static void filt_timerdetach(struct knote *kn) { struct kq_timer_cb_data *kc; unsigned int old __unused; bool pending; kc = kn->kn_ptr.p_v; do { callout_drain(&kc->c); /* * kqtimer_proc_continue() might have rescheduled this callout. * Double-check, using the process mutex as an interlock. */ PROC_LOCK(kc->p); if ((kc->flags & KQ_TIMER_CB_ENQUEUED) != 0) { kc->flags &= ~KQ_TIMER_CB_ENQUEUED; TAILQ_REMOVE(&kc->p->p_kqtim_stop, kc, link); } pending = callout_pending(&kc->c); PROC_UNLOCK(kc->p); } while (pending); free(kc, M_KQUEUE); old = atomic_fetchadd_int(&kq_ncallouts, -1); KASSERT(old > 0, ("Number of callouts cannot become negative")); kn->kn_status |= KN_DETACHED; /* knlist_remove sets it */ } static void filt_timertouch(struct knote *kn, struct kevent *kev, u_long type) { struct kq_timer_cb_data *kc; struct kqueue *kq; sbintime_t to; int error; switch (type) { case EVENT_REGISTER: /* Handle re-added timers that update data/fflags */ if (kev->flags & EV_ADD) { kc = kn->kn_ptr.p_v; /* Drain any existing callout. */ callout_drain(&kc->c); /* Throw away any existing undelivered record * of the timer expiration. This is done under * the presumption that if a process is * re-adding this timer with new parameters, * it is no longer interested in what may have * happened under the old parameters. If it is * interested, it can wait for the expiration, * delete the old timer definition, and then * add the new one. * * This has to be done while the kq is locked: * - if enqueued, dequeue * - make it no longer active * - clear the count of expiration events */ kq = kn->kn_kq; KQ_LOCK(kq); if (kn->kn_status & KN_QUEUED) knote_dequeue(kn); kn->kn_status &= ~KN_ACTIVE; kn->kn_data = 0; KQ_UNLOCK(kq); /* Reschedule timer based on new data/fflags */ kn->kn_sfflags = kev->fflags; kn->kn_sdata = kev->data; error = filt_timervalidate(kn, &to); if (error != 0) { kn->kn_flags |= EV_ERROR; kn->kn_data = error; } else filt_timerstart(kn, to); } break; case EVENT_PROCESS: *kev = kn->kn_kevent; if (kn->kn_flags & EV_CLEAR) { kn->kn_data = 0; kn->kn_fflags = 0; } break; default: panic("filt_timertouch() - invalid type (%ld)", type); break; } } static int filt_timer(struct knote *kn, long hint) { return (kn->kn_data != 0); } static int filt_userattach(struct knote *kn) { /* * EVFILT_USER knotes are not attached to anything in the kernel. */ kn->kn_hook = NULL; if (kn->kn_fflags & NOTE_TRIGGER) kn->kn_hookid = 1; else kn->kn_hookid = 0; return (0); } static void filt_userdetach(__unused struct knote *kn) { /* * EVFILT_USER knotes are not attached to anything in the kernel. */ } static int filt_user(struct knote *kn, __unused long hint) { return (kn->kn_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_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; if (kev->flags & EV_CLEAR) { kn->kn_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_hookid = 0; kn->kn_data = 0; kn->kn_fflags = 0; } break; default: panic("filt_usertouch() - invalid type (%ld)", type); break; } } int sys_kqueue(struct thread *td, struct kqueue_args *uap) { return (kern_kqueue(td, 0, NULL)); } static void kqueue_init(struct kqueue *kq) { mtx_init(&kq->kq_lock, "kqueue", NULL, MTX_DEF | MTX_DUPOK); TAILQ_INIT(&kq->kq_head); knlist_init_mtx(&kq->kq_sel.si_note, &kq->kq_lock); TASK_INIT(&kq->kq_task, 0, kqueue_task, kq); } int kern_kqueue(struct thread *td, int flags, struct filecaps *fcaps) { struct filedesc *fdp; struct kqueue *kq; struct file *fp; struct ucred *cred; int fd, error; fdp = td->td_proc->p_fd; cred = td->td_ucred; if (!chgkqcnt(cred->cr_ruidinfo, 1, lim_cur(td, RLIMIT_KQUEUES))) return (ENOMEM); error = falloc_caps(td, &fp, &fd, flags, fcaps); if (error != 0) { chgkqcnt(cred->cr_ruidinfo, -1, 0); return (error); } /* An extra reference on `fp' has been held for us by falloc(). */ kq = malloc(sizeof *kq, M_KQUEUE, M_WAITOK | M_ZERO); kqueue_init(kq); kq->kq_fdp = fdp; kq->kq_cred = crhold(cred); FILEDESC_XLOCK(fdp); TAILQ_INSERT_HEAD(&fdp->fd_kqlist, kq, kq_list); FILEDESC_XUNLOCK(fdp); finit(fp, FREAD | FWRITE, DTYPE_KQUEUE, kq, &kqueueops); fdrop(fp, td); td->td_retval[0] = fd; return (0); } struct g_kevent_args { int fd; const void *changelist; int nchanges; void *eventlist; int nevents; const struct timespec *timeout; }; int sys_kevent(struct thread *td, struct kevent_args *uap) { struct kevent_copyops k_ops = { .arg = uap, .k_copyout = kevent_copyout, .k_copyin = kevent_copyin, .kevent_size = sizeof(struct kevent), }; struct g_kevent_args gk_args = { .fd = uap->fd, .changelist = uap->changelist, .nchanges = uap->nchanges, .eventlist = uap->eventlist, .nevents = uap->nevents, .timeout = uap->timeout, }; return (kern_kevent_generic(td, &gk_args, &k_ops, "kevent")); } static int kern_kevent_generic(struct thread *td, struct g_kevent_args *uap, struct kevent_copyops *k_ops, const char *struct_name) { struct timespec ts, *tsp; #ifdef KTRACE struct kevent *eventlist = uap->eventlist; #endif int error; if (uap->timeout != NULL) { error = copyin(uap->timeout, &ts, sizeof(ts)); if (error) return (error); tsp = &ts; } else tsp = NULL; #ifdef KTRACE if (KTRPOINT(td, KTR_STRUCT_ARRAY)) ktrstructarray(struct_name, UIO_USERSPACE, uap->changelist, uap->nchanges, k_ops->kevent_size); #endif error = kern_kevent(td, uap->fd, uap->nchanges, uap->nevents, k_ops, tsp); #ifdef KTRACE if (error == 0 && KTRPOINT(td, KTR_STRUCT_ARRAY)) ktrstructarray(struct_name, UIO_USERSPACE, eventlist, td->td_retval[0], k_ops->kevent_size); #endif 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) { struct kevent_args *uap; int error; KASSERT(count <= KQ_NEVENTS, ("count (%d) > KQ_NEVENTS", count)); uap = (struct kevent_args *)arg; error = copyout(kevp, uap->eventlist, count * sizeof *kevp); if (error == 0) uap->eventlist += count; return (error); } /* * Copy 'count' items from the list pointed to by uap->changelist. */ static int kevent_copyin(void *arg, struct kevent *kevp, int count) { struct kevent_args *uap; int error; KASSERT(count <= KQ_NEVENTS, ("count (%d) > KQ_NEVENTS", count)); uap = (struct kevent_args *)arg; error = copyin(uap->changelist, kevp, count * sizeof *kevp); if (error == 0) uap->changelist += count; return (error); } #ifdef COMPAT_FREEBSD11 static int kevent11_copyout(void *arg, struct kevent *kevp, int count) { struct freebsd11_kevent_args *uap; struct freebsd11_kevent kev11; int error, i; KASSERT(count <= KQ_NEVENTS, ("count (%d) > KQ_NEVENTS", count)); uap = (struct freebsd11_kevent_args *)arg; for (i = 0; i < count; i++) { kev11.ident = kevp->ident; kev11.filter = kevp->filter; kev11.flags = kevp->flags; kev11.fflags = kevp->fflags; kev11.data = kevp->data; kev11.udata = kevp->udata; error = copyout(&kev11, uap->eventlist, sizeof(kev11)); if (error != 0) break; uap->eventlist++; kevp++; } return (error); } /* * Copy 'count' items from the list pointed to by uap->changelist. */ static int kevent11_copyin(void *arg, struct kevent *kevp, int count) { struct freebsd11_kevent_args *uap; struct freebsd11_kevent kev11; int error, i; KASSERT(count <= KQ_NEVENTS, ("count (%d) > KQ_NEVENTS", count)); uap = (struct freebsd11_kevent_args *)arg; for (i = 0; i < count; i++) { error = copyin(uap->changelist, &kev11, sizeof(kev11)); if (error != 0) break; kevp->ident = kev11.ident; kevp->filter = kev11.filter; kevp->flags = kev11.flags; kevp->fflags = kev11.fflags; kevp->data = (uintptr_t)kev11.data; kevp->udata = kev11.udata; bzero(&kevp->ext, sizeof(kevp->ext)); uap->changelist++; kevp++; } return (error); } int freebsd11_kevent(struct thread *td, struct freebsd11_kevent_args *uap) { struct kevent_copyops k_ops = { .arg = uap, .k_copyout = kevent11_copyout, .k_copyin = kevent11_copyin, .kevent_size = sizeof(struct freebsd11_kevent), }; struct g_kevent_args gk_args = { .fd = uap->fd, .changelist = uap->changelist, .nchanges = uap->nchanges, .eventlist = uap->eventlist, .nevents = uap->nevents, .timeout = uap->timeout, }; return (kern_kevent_generic(td, &gk_args, &k_ops, "freebsd11_kevent")); } #endif int kern_kevent(struct thread *td, int fd, int nchanges, int nevents, struct kevent_copyops *k_ops, const struct timespec *timeout) { cap_rights_t rights; struct file *fp; int error; cap_rights_init_zero(&rights); if (nchanges > 0) cap_rights_set_one(&rights, CAP_KQUEUE_CHANGE); if (nevents > 0) cap_rights_set_one(&rights, CAP_KQUEUE_EVENT); error = fget(td, fd, &rights, &fp); if (error != 0) return (error); error = kern_kevent_fp(td, fp, nchanges, nevents, k_ops, timeout); fdrop(fp, td); return (error); } static int kqueue_kevent(struct kqueue *kq, struct thread *td, int nchanges, int nevents, struct kevent_copyops *k_ops, const struct timespec *timeout) { struct kevent keva[KQ_NEVENTS]; struct kevent *kevp, *changes; int i, n, nerrors, error; if (nchanges < 0) return (EINVAL); nerrors = 0; while (nchanges > 0) { n = nchanges > KQ_NEVENTS ? KQ_NEVENTS : nchanges; error = k_ops->k_copyin(k_ops->arg, keva, n); if (error) return (error); changes = keva; for (i = 0; i < n; i++) { kevp = &changes[i]; if (!kevp->filter) continue; kevp->flags &= ~EV_SYSFLAGS; error = kqueue_register(kq, kevp, td, M_WAITOK); if (error || (kevp->flags & EV_RECEIPT)) { if (nevents == 0) return (error); kevp->flags = EV_ERROR; kevp->data = error; (void)k_ops->k_copyout(k_ops->arg, kevp, 1); nevents--; nerrors++; } } nchanges -= n; } if (nerrors) { td->td_retval[0] = nerrors; return (0); } return (kqueue_scan(kq, nevents, k_ops, timeout, keva, td)); } int kern_kevent_fp(struct thread *td, struct file *fp, int nchanges, int nevents, struct kevent_copyops *k_ops, const struct timespec *timeout) { struct kqueue *kq; int error; error = kqueue_acquire(fp, &kq); if (error != 0) return (error); error = kqueue_kevent(kq, td, nchanges, nevents, k_ops, timeout); kqueue_release(kq, 0); return (error); } /* * Performs a kevent() call on a temporarily created kqueue. This can be * used to perform one-shot polling, similar to poll() and select(). */ int kern_kevent_anonymous(struct thread *td, int nevents, struct kevent_copyops *k_ops) { struct kqueue kq = {}; int error; kqueue_init(&kq); kq.kq_refcnt = 1; error = kqueue_kevent(&kq, td, nevents, nevents, k_ops, NULL); kqueue_drain(&kq, td); kqueue_destroy(&kq); return (error); } int kqueue_add_filteropts(int filt, struct filterops *filtops) { int error; error = 0; if (filt > 0 || filt + EVFILT_SYSCOUNT < 0) { printf( "trying to add a filterop that is out of range: %d is beyond %d\n", ~filt, EVFILT_SYSCOUNT); return EINVAL; } mtx_lock(&filterops_lock); if (sysfilt_ops[~filt].for_fop != &null_filtops && sysfilt_ops[~filt].for_fop != NULL) error = EEXIST; else { sysfilt_ops[~filt].for_fop = filtops; sysfilt_ops[~filt].for_refcnt = 0; } mtx_unlock(&filterops_lock); return (error); } int kqueue_del_filteropts(int filt) { int error; error = 0; if (filt > 0 || filt + EVFILT_SYSCOUNT < 0) return EINVAL; mtx_lock(&filterops_lock); if (sysfilt_ops[~filt].for_fop == &null_filtops || sysfilt_ops[~filt].for_fop == NULL) error = EINVAL; else if (sysfilt_ops[~filt].for_refcnt != 0) error = EBUSY; else { sysfilt_ops[~filt].for_fop = &null_filtops; sysfilt_ops[~filt].for_refcnt = 0; } mtx_unlock(&filterops_lock); return error; } static struct filterops * kqueue_fo_find(int filt) { if (filt > 0 || filt + EVFILT_SYSCOUNT < 0) return NULL; if (sysfilt_ops[~filt].for_nolock) return sysfilt_ops[~filt].for_fop; mtx_lock(&filterops_lock); sysfilt_ops[~filt].for_refcnt++; if (sysfilt_ops[~filt].for_fop == NULL) sysfilt_ops[~filt].for_fop = &null_filtops; mtx_unlock(&filterops_lock); return sysfilt_ops[~filt].for_fop; } static void kqueue_fo_release(int filt) { if (filt > 0 || filt + EVFILT_SYSCOUNT < 0) return; if (sysfilt_ops[~filt].for_nolock) return; mtx_lock(&filterops_lock); KASSERT(sysfilt_ops[~filt].for_refcnt > 0, ("filter object refcount not valid on release")); sysfilt_ops[~filt].for_refcnt--; mtx_unlock(&filterops_lock); } /* * A ref to kq (obtained via kqueue_acquire) must be held. */ static int kqueue_register(struct kqueue *kq, struct kevent *kev, struct thread *td, int mflag) { struct filterops *fops; struct file *fp; struct knote *kn, *tkn; struct knlist *knl; int error, filt, event; int haskqglobal, filedesc_unlock; if ((kev->flags & (EV_ENABLE | EV_DISABLE)) == (EV_ENABLE | EV_DISABLE)) return (EINVAL); fp = NULL; kn = NULL; knl = NULL; error = 0; haskqglobal = 0; filedesc_unlock = 0; filt = kev->filter; fops = kqueue_fo_find(filt); if (fops == NULL) return EINVAL; if (kev->flags & EV_ADD) { /* Reject an invalid flag pair early */ if (kev->flags & EV_KEEPUDATA) { tkn = NULL; error = EINVAL; goto done; } /* * Prevent waiting with locks. Non-sleepable * allocation failures are handled in the loop, only * if the spare knote appears to be actually required. */ tkn = knote_alloc(mflag); } else { tkn = NULL; } findkn: if (fops->f_isfd) { KASSERT(td != NULL, ("td is NULL")); if (kev->ident > INT_MAX) error = EBADF; else error = fget(td, kev->ident, &cap_event_rights, &fp); if (error) goto done; if ((kev->flags & EV_ADD) == EV_ADD && kqueue_expand(kq, fops, kev->ident, M_NOWAIT) != 0) { /* try again */ fdrop(fp, td); fp = NULL; error = kqueue_expand(kq, fops, kev->ident, mflag); if (error) goto done; goto findkn; } if (fp->f_type == DTYPE_KQUEUE) { /* * If we add some intelligence about what we are doing, * we should be able to support events on ourselves. * We need to know when we are doing this to prevent * getting both the knlist lock and the kq lock since * they are the same thing. */ if (fp->f_data == kq) { error = EINVAL; goto done; } /* * Pre-lock the filedesc before the global * lock mutex, see the comment in * kqueue_close(). */ FILEDESC_XLOCK(td->td_proc->p_fd); filedesc_unlock = 1; KQ_GLOBAL_LOCK(&kq_global, haskqglobal); } KQ_LOCK(kq); if (kev->ident < kq->kq_knlistsize) { SLIST_FOREACH(kn, &kq->kq_knlist[kev->ident], kn_link) if (kev->filter == kn->kn_filter) break; } } else { if ((kev->flags & EV_ADD) == EV_ADD) { error = kqueue_expand(kq, fops, kev->ident, mflag); if (error != 0) goto done; } KQ_LOCK(kq); /* * If possible, find an existing knote to use for this kevent. */ if (kev->filter == EVFILT_PROC && (kev->flags & (EV_FLAG1 | EV_FLAG2)) != 0) { /* This is an internal creation of a process tracking * note. Don't attempt to coalesce this with an * existing note. */ ; } else if (kq->kq_knhashmask != 0) { struct klist *list; list = &kq->kq_knhash[ KN_HASH((u_long)kev->ident, kq->kq_knhashmask)]; SLIST_FOREACH(kn, list, kn_link) if (kev->ident == kn->kn_id && kev->filter == kn->kn_filter) break; } } /* knote is in the process of changing, wait for it to stabilize. */ if (kn != NULL && kn_in_flux(kn)) { KQ_GLOBAL_UNLOCK(&kq_global, haskqglobal); if (filedesc_unlock) { FILEDESC_XUNLOCK(td->td_proc->p_fd); filedesc_unlock = 0; } kq->kq_state |= KQ_FLUXWAIT; msleep(kq, &kq->kq_lock, PSOCK | PDROP, "kqflxwt", 0); if (fp != NULL) { fdrop(fp, td); fp = NULL; } goto findkn; } /* * kn now contains the matching knote, or NULL if no match */ if (kn == NULL) { if (kev->flags & EV_ADD) { kn = tkn; tkn = NULL; if (kn == NULL) { KQ_UNLOCK(kq); error = ENOMEM; goto done; } kn->kn_fp = fp; kn->kn_kq = kq; kn->kn_fop = fops; /* * apply reference counts to knote structure, and * do not release it at the end of this routine. */ fops = NULL; fp = NULL; kn->kn_sfflags = kev->fflags; kn->kn_sdata = kev->data; kev->fflags = 0; kev->data = 0; kn->kn_kevent = *kev; kn->kn_kevent.flags &= ~(EV_ADD | EV_DELETE | EV_ENABLE | EV_DISABLE | EV_FORCEONESHOT); kn->kn_status = KN_DETACHED; if ((kev->flags & EV_DISABLE) != 0) kn->kn_status |= KN_DISABLED; kn_enter_flux(kn); error = knote_attach(kn, kq); KQ_UNLOCK(kq); if (error != 0) { tkn = kn; goto done; } if ((error = kn->kn_fop->f_attach(kn)) != 0) { knote_drop_detached(kn, td); goto done; } knl = kn_list_lock(kn); goto done_ev_add; } else { /* No matching knote and the EV_ADD flag is not set. */ KQ_UNLOCK(kq); error = ENOENT; goto done; } } if (kev->flags & EV_DELETE) { kn_enter_flux(kn); KQ_UNLOCK(kq); knote_drop(kn, td); goto done; } if (kev->flags & EV_FORCEONESHOT) { kn->kn_flags |= EV_ONESHOT; KNOTE_ACTIVATE(kn, 1); } if ((kev->flags & EV_ENABLE) != 0) kn->kn_status &= ~KN_DISABLED; else if ((kev->flags & EV_DISABLE) != 0) kn->kn_status |= KN_DISABLED; /* * The user may change some filter values after the initial EV_ADD, * but doing so will not reset any filter which has already been * triggered. */ kn->kn_status |= KN_SCAN; kn_enter_flux(kn); KQ_UNLOCK(kq); knl = kn_list_lock(kn); if ((kev->flags & EV_KEEPUDATA) == 0) kn->kn_kevent.udata = kev->udata; if (!fops->f_isfd && fops->f_touch != NULL) { fops->f_touch(kn, kev, EVENT_REGISTER); } else { kn->kn_sfflags = kev->fflags; kn->kn_sdata = kev->data; } done_ev_add: /* * We can get here with kn->kn_knlist == NULL. This can happen when * the initial attach event decides that the event is "completed" * already, e.g., filt_procattach() is called on a zombie process. It * will call filt_proc() which will remove it from the list, and NULL * kn_knlist. * * KN_DISABLED will be stable while the knote is in flux, so the * unlocked read will not race with an update. */ if ((kn->kn_status & KN_DISABLED) == 0) event = kn->kn_fop->f_event(kn, 0); else event = 0; KQ_LOCK(kq); if (event) kn->kn_status |= KN_ACTIVE; if ((kn->kn_status & (KN_ACTIVE | KN_DISABLED | KN_QUEUED)) == KN_ACTIVE) knote_enqueue(kn); kn->kn_status &= ~KN_SCAN; kn_leave_flux(kn); kn_list_unlock(knl); KQ_UNLOCK_FLUX(kq); done: KQ_GLOBAL_UNLOCK(&kq_global, haskqglobal); if (filedesc_unlock) FILEDESC_XUNLOCK(td->td_proc->p_fd); if (fp != NULL) fdrop(fp, td); knote_free(tkn); if (fops != NULL) kqueue_fo_release(filt); return (error); } static int kqueue_acquire(struct file *fp, struct kqueue **kqp) { int error; struct kqueue *kq; error = 0; kq = fp->f_data; if (fp->f_type != DTYPE_KQUEUE || kq == NULL) return (EBADF); *kqp = kq; KQ_LOCK(kq); if ((kq->kq_state & KQ_CLOSING) == KQ_CLOSING) { KQ_UNLOCK(kq); return (EBADF); } kq->kq_refcnt++; KQ_UNLOCK(kq); return error; } static void kqueue_release(struct kqueue *kq, int locked) { if (locked) KQ_OWNED(kq); else KQ_LOCK(kq); kq->kq_refcnt--; if (kq->kq_refcnt == 1) wakeup(&kq->kq_refcnt); if (!locked) KQ_UNLOCK(kq); } void kqueue_drain_schedtask(void) { taskqueue_quiesce(taskqueue_kqueue_ctx); } static void kqueue_schedtask(struct kqueue *kq) { struct thread *td; KQ_OWNED(kq); KASSERT(((kq->kq_state & KQ_TASKDRAIN) != KQ_TASKDRAIN), ("scheduling kqueue task while draining")); if ((kq->kq_state & KQ_TASKSCHED) != KQ_TASKSCHED) { taskqueue_enqueue(taskqueue_kqueue_ctx, &kq->kq_task); kq->kq_state |= KQ_TASKSCHED; td = curthread; thread_lock(td); td->td_flags |= TDF_ASTPENDING | TDF_KQTICKLED; thread_unlock(td); } } /* * Expand the kq to make sure we have storage for fops/ident pair. * * Return 0 on success (or no work necessary), return errno on failure. */ static int kqueue_expand(struct kqueue *kq, struct filterops *fops, uintptr_t ident, int mflag) { struct klist *list, *tmp_knhash, *to_free; u_long tmp_knhashmask; int error, fd, size; KQ_NOTOWNED(kq); error = 0; to_free = NULL; if (fops->f_isfd) { fd = ident; if (kq->kq_knlistsize <= fd) { size = kq->kq_knlistsize; while (size <= fd) size += KQEXTENT; list = malloc(size * sizeof(*list), M_KQUEUE, mflag); if (list == NULL) return ENOMEM; KQ_LOCK(kq); if ((kq->kq_state & KQ_CLOSING) != 0) { to_free = list; error = EBADF; } else if (kq->kq_knlistsize > fd) { to_free = list; } else { if (kq->kq_knlist != NULL) { bcopy(kq->kq_knlist, list, kq->kq_knlistsize * sizeof(*list)); to_free = kq->kq_knlist; kq->kq_knlist = NULL; } bzero((caddr_t)list + kq->kq_knlistsize * sizeof(*list), (size - kq->kq_knlistsize) * sizeof(*list)); kq->kq_knlistsize = size; kq->kq_knlist = list; } KQ_UNLOCK(kq); } } else { if (kq->kq_knhashmask == 0) { tmp_knhash = hashinit_flags(KN_HASHSIZE, M_KQUEUE, &tmp_knhashmask, (mflag & M_WAITOK) != 0 ? HASH_WAITOK : HASH_NOWAIT); if (tmp_knhash == NULL) return (ENOMEM); KQ_LOCK(kq); if ((kq->kq_state & KQ_CLOSING) != 0) { to_free = tmp_knhash; error = EBADF; } else if (kq->kq_knhashmask == 0) { kq->kq_knhash = tmp_knhash; kq->kq_knhashmask = tmp_knhashmask; } else { to_free = tmp_knhash; } KQ_UNLOCK(kq); } } free(to_free, M_KQUEUE); KQ_NOTOWNED(kq); return (error); } static void kqueue_task(void *arg, int pending) { struct kqueue *kq; int haskqglobal; haskqglobal = 0; kq = arg; KQ_GLOBAL_LOCK(&kq_global, haskqglobal); KQ_LOCK(kq); KNOTE_LOCKED(&kq->kq_sel.si_note, 0); kq->kq_state &= ~KQ_TASKSCHED; if ((kq->kq_state & KQ_TASKDRAIN) == KQ_TASKDRAIN) { wakeup(&kq->kq_state); } KQ_UNLOCK(kq); KQ_GLOBAL_UNLOCK(&kq_global, haskqglobal); } /* * Scan, update kn_data (if not ONESHOT), and copyout triggered events. * We treat KN_MARKER knotes as if they are in flux. */ static int kqueue_scan(struct kqueue *kq, int maxevents, struct kevent_copyops *k_ops, const struct timespec *tsp, struct kevent *keva, struct thread *td) { struct kevent *kevp; struct knote *kn, *marker; struct knlist *knl; sbintime_t asbt, rsbt; int count, error, haskqglobal, influx, nkev, touch; count = maxevents; nkev = 0; error = 0; haskqglobal = 0; if (maxevents == 0) goto done_nl; if (maxevents < 0) { error = EINVAL; goto done_nl; } rsbt = 0; if (tsp != NULL) { if (!timespecvalid_interval(tsp)) { error = EINVAL; goto done_nl; } if (timespecisset(tsp)) { if (tsp->tv_sec <= INT32_MAX) { rsbt = tstosbt(*tsp); if (TIMESEL(&asbt, rsbt)) asbt += tc_tick_sbt; if (asbt <= SBT_MAX - rsbt) asbt += rsbt; else asbt = 0; rsbt >>= tc_precexp; } else asbt = 0; } else asbt = -1; } else asbt = 0; marker = knote_alloc(M_WAITOK); marker->kn_status = KN_MARKER; KQ_LOCK(kq); retry: kevp = keva; if (kq->kq_count == 0) { if (asbt == -1) { error = EWOULDBLOCK; } else { kq->kq_state |= KQ_SLEEP; error = msleep_sbt(kq, &kq->kq_lock, PSOCK | PCATCH, "kqread", asbt, rsbt, C_ABSOLUTE); } if (error == 0) goto retry; /* don't restart after signals... */ if (error == ERESTART) error = EINTR; else if (error == EWOULDBLOCK) error = 0; goto done; } TAILQ_INSERT_TAIL(&kq->kq_head, marker, kn_tqe); influx = 0; while (count) { KQ_OWNED(kq); kn = TAILQ_FIRST(&kq->kq_head); if ((kn->kn_status == KN_MARKER && kn != marker) || kn_in_flux(kn)) { if (influx) { influx = 0; KQ_FLUX_WAKEUP(kq); } kq->kq_state |= KQ_FLUXWAIT; error = msleep(kq, &kq->kq_lock, PSOCK, "kqflxwt", 0); continue; } TAILQ_REMOVE(&kq->kq_head, kn, kn_tqe); if ((kn->kn_status & KN_DISABLED) == KN_DISABLED) { kn->kn_status &= ~KN_QUEUED; kq->kq_count--; continue; } if (kn == marker) { KQ_FLUX_WAKEUP(kq); if (count == maxevents) goto retry; goto done; } KASSERT(!kn_in_flux(kn), ("knote %p is unexpectedly in flux", kn)); if ((kn->kn_flags & EV_DROP) == EV_DROP) { kn->kn_status &= ~KN_QUEUED; kn_enter_flux(kn); kq->kq_count--; KQ_UNLOCK(kq); /* * We don't need to lock the list since we've * marked it as in flux. */ knote_drop(kn, td); KQ_LOCK(kq); continue; } else if ((kn->kn_flags & EV_ONESHOT) == EV_ONESHOT) { kn->kn_status &= ~KN_QUEUED; kn_enter_flux(kn); kq->kq_count--; KQ_UNLOCK(kq); /* * We don't need to lock the list since we've * marked the knote as being in flux. */ *kevp = kn->kn_kevent; knote_drop(kn, td); KQ_LOCK(kq); kn = NULL; } else { kn->kn_status |= KN_SCAN; kn_enter_flux(kn); KQ_UNLOCK(kq); if ((kn->kn_status & KN_KQUEUE) == KN_KQUEUE) KQ_GLOBAL_LOCK(&kq_global, haskqglobal); knl = kn_list_lock(kn); if (kn->kn_fop->f_event(kn, 0) == 0) { KQ_LOCK(kq); KQ_GLOBAL_UNLOCK(&kq_global, haskqglobal); kn->kn_status &= ~(KN_QUEUED | KN_ACTIVE | KN_SCAN); kn_leave_flux(kn); kq->kq_count--; kn_list_unlock(knl); influx = 1; continue; } touch = (!kn->kn_fop->f_isfd && kn->kn_fop->f_touch != NULL); if (touch) kn->kn_fop->f_touch(kn, kevp, EVENT_PROCESS); else *kevp = kn->kn_kevent; KQ_LOCK(kq); KQ_GLOBAL_UNLOCK(&kq_global, haskqglobal); if (kn->kn_flags & (EV_CLEAR | EV_DISPATCH)) { /* * Manually clear knotes who weren't * 'touch'ed. */ if (touch == 0 && 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); kq->kq_count--; } else TAILQ_INSERT_TAIL(&kq->kq_head, kn, kn_tqe); kn->kn_status &= ~KN_SCAN; kn_leave_flux(kn); kn_list_unlock(knl); influx = 1; } /* we are returning a copy to the user */ kevp++; nkev++; count--; if (nkev == KQ_NEVENTS) { influx = 0; KQ_UNLOCK_FLUX(kq); error = k_ops->k_copyout(k_ops->arg, keva, nkev); nkev = 0; kevp = keva; KQ_LOCK(kq); if (error) break; } } TAILQ_REMOVE(&kq->kq_head, marker, kn_tqe); done: KQ_OWNED(kq); KQ_UNLOCK_FLUX(kq); knote_free(marker); done_nl: KQ_NOTOWNED(kq); if (nkev != 0) error = k_ops->k_copyout(k_ops->arg, keva, nkev); td->td_retval[0] = maxevents - count; return (error); } /*ARGSUSED*/ static int kqueue_ioctl(struct file *fp, u_long cmd, void *data, struct ucred *active_cred, struct thread *td) { /* * Enabling sigio causes two major problems: * 1) infinite recursion: * Synopsys: kevent is being used to track signals and have FIOASYNC * set. On receipt of a signal this will cause a kqueue to recurse * into itself over and over. Sending the sigio causes the kqueue * to become ready, which in turn posts sigio again, forever. * Solution: this can be solved by setting a flag in the kqueue that * we have a SIGIO in progress. * 2) locking problems: * Synopsys: Kqueue is a leaf subsystem, but adding signalling puts * us above the proc and pgrp locks. * Solution: Post a signal using an async mechanism, being sure to * record a generation count in the delivery so that we do not deliver * a signal to the wrong process. * * Note, these two mechanisms are somewhat mutually exclusive! */ #if 0 struct kqueue *kq; kq = fp->f_data; switch (cmd) { case FIOASYNC: if (*(int *)data) { kq->kq_state |= KQ_ASYNC; } else { kq->kq_state &= ~KQ_ASYNC; } return (0); case FIOSETOWN: return (fsetown(*(int *)data, &kq->kq_sigio)); case FIOGETOWN: *(int *)data = fgetown(&kq->kq_sigio); return (0); } #endif return (ENOTTY); } /*ARGSUSED*/ static int kqueue_poll(struct file *fp, int events, struct ucred *active_cred, struct thread *td) { struct kqueue *kq; int revents = 0; int error; if ((error = kqueue_acquire(fp, &kq))) return POLLERR; KQ_LOCK(kq); if (events & (POLLIN | POLLRDNORM)) { if (kq->kq_count) { revents |= events & (POLLIN | POLLRDNORM); } else { selrecord(td, &kq->kq_sel); if (SEL_WAITING(&kq->kq_sel)) kq->kq_state |= KQ_SEL; } } kqueue_release(kq, 1); KQ_UNLOCK(kq); return (revents); } /*ARGSUSED*/ static int kqueue_stat(struct file *fp, struct stat *st, struct ucred *active_cred) { bzero((void *)st, sizeof *st); /* * We no longer return kq_count because the unlocked value is useless. * If you spent all this time getting the count, why not spend your * syscall better by calling kevent? * * XXX - This is needed for libc_r. */ st->st_mode = S_IFIFO; return (0); } static void kqueue_drain(struct kqueue *kq, struct thread *td) { struct knote *kn; int i; KQ_LOCK(kq); KASSERT((kq->kq_state & KQ_CLOSING) != KQ_CLOSING, ("kqueue already closing")); kq->kq_state |= KQ_CLOSING; if (kq->kq_refcnt > 1) msleep(&kq->kq_refcnt, &kq->kq_lock, PSOCK, "kqclose", 0); KASSERT(kq->kq_refcnt == 1, ("other refs are out there!")); KASSERT(knlist_empty(&kq->kq_sel.si_note), ("kqueue's knlist not empty")); for (i = 0; i < kq->kq_knlistsize; i++) { while ((kn = SLIST_FIRST(&kq->kq_knlist[i])) != NULL) { if (kn_in_flux(kn)) { kq->kq_state |= KQ_FLUXWAIT; msleep(kq, &kq->kq_lock, PSOCK, "kqclo1", 0); continue; } kn_enter_flux(kn); KQ_UNLOCK(kq); knote_drop(kn, td); KQ_LOCK(kq); } } if (kq->kq_knhashmask != 0) { for (i = 0; i <= kq->kq_knhashmask; i++) { while ((kn = SLIST_FIRST(&kq->kq_knhash[i])) != NULL) { if (kn_in_flux(kn)) { kq->kq_state |= KQ_FLUXWAIT; msleep(kq, &kq->kq_lock, PSOCK, "kqclo2", 0); continue; } kn_enter_flux(kn); KQ_UNLOCK(kq); knote_drop(kn, td); KQ_LOCK(kq); } } } if ((kq->kq_state & KQ_TASKSCHED) == KQ_TASKSCHED) { kq->kq_state |= KQ_TASKDRAIN; msleep(&kq->kq_state, &kq->kq_lock, PSOCK, "kqtqdr", 0); } if ((kq->kq_state & KQ_SEL) == KQ_SEL) { selwakeuppri(&kq->kq_sel, PSOCK); if (!SEL_WAITING(&kq->kq_sel)) kq->kq_state &= ~KQ_SEL; } KQ_UNLOCK(kq); } static void kqueue_destroy(struct kqueue *kq) { KASSERT(kq->kq_fdp == NULL, ("kqueue still attached to a file descriptor")); seldrain(&kq->kq_sel); knlist_destroy(&kq->kq_sel.si_note); mtx_destroy(&kq->kq_lock); if (kq->kq_knhash != NULL) free(kq->kq_knhash, M_KQUEUE); if (kq->kq_knlist != NULL) free(kq->kq_knlist, M_KQUEUE); funsetown(&kq->kq_sigio); } /*ARGSUSED*/ static int kqueue_close(struct file *fp, struct thread *td) { struct kqueue *kq = fp->f_data; struct filedesc *fdp; int error; int filedesc_unlock; if ((error = kqueue_acquire(fp, &kq))) return error; kqueue_drain(kq, td); /* * We could be called due to the knote_drop() doing fdrop(), * called from kqueue_register(). In this case the global * lock is owned, and filedesc sx is locked before, to not * take the sleepable lock after non-sleepable. */ fdp = kq->kq_fdp; kq->kq_fdp = NULL; if (!sx_xlocked(FILEDESC_LOCK(fdp))) { FILEDESC_XLOCK(fdp); filedesc_unlock = 1; } else filedesc_unlock = 0; TAILQ_REMOVE(&fdp->fd_kqlist, kq, kq_list); if (filedesc_unlock) FILEDESC_XUNLOCK(fdp); kqueue_destroy(kq); chgkqcnt(kq->kq_cred->cr_ruidinfo, -1, 0); crfree(kq->kq_cred); free(kq, M_KQUEUE); fp->f_data = NULL; return (0); } static int kqueue_fill_kinfo(struct file *fp, struct kinfo_file *kif, struct filedesc *fdp) { + struct kqueue *kq = fp->f_data; kif->kf_type = KF_TYPE_KQUEUE; + kif->kf_un.kf_kqueue.kf_kqueue_addr = (uintptr_t)kq; + kif->kf_un.kf_kqueue.kf_kqueue_count = kq->kq_count; + kif->kf_un.kf_kqueue.kf_kqueue_state = kq->kq_state; return (0); } static void kqueue_wakeup(struct kqueue *kq) { KQ_OWNED(kq); if ((kq->kq_state & KQ_SLEEP) == KQ_SLEEP) { kq->kq_state &= ~KQ_SLEEP; wakeup(kq); } if ((kq->kq_state & KQ_SEL) == KQ_SEL) { selwakeuppri(&kq->kq_sel, PSOCK); if (!SEL_WAITING(&kq->kq_sel)) kq->kq_state &= ~KQ_SEL; } if (!knlist_empty(&kq->kq_sel.si_note)) kqueue_schedtask(kq); if ((kq->kq_state & KQ_ASYNC) == KQ_ASYNC) { pgsigio(&kq->kq_sigio, SIGIO, 0); } } /* * Walk down a list of knotes, activating them if their event has triggered. * * There is a possibility to optimize in the case of one kq watching another. * Instead of scheduling a task to wake it up, you could pass enough state * down the chain to make up the parent kqueue. Make this code functional * first. */ void knote(struct knlist *list, long hint, int lockflags) { struct kqueue *kq; struct knote *kn, *tkn; int error; if (list == NULL) return; KNL_ASSERT_LOCK(list, lockflags & KNF_LISTLOCKED); if ((lockflags & KNF_LISTLOCKED) == 0) list->kl_lock(list->kl_lockarg); /* * If we unlock the list lock (and enter influx), we can * eliminate the kqueue scheduling, but this will introduce * four lock/unlock's for each knote to test. Also, marker * would be needed to keep iteration position, since filters * or other threads could remove events. */ SLIST_FOREACH_SAFE(kn, &list->kl_list, kn_selnext, tkn) { kq = kn->kn_kq; KQ_LOCK(kq); if (kn_in_flux(kn) && (kn->kn_status & KN_SCAN) == 0) { /* * Do not process the influx notes, except for * the influx coming from the kq unlock in the * kqueue_scan(). In the later case, we do * not interfere with the scan, since the code * fragment in kqueue_scan() locks the knlist, * and cannot proceed until we finished. */ KQ_UNLOCK(kq); } else if ((lockflags & KNF_NOKQLOCK) != 0) { kn_enter_flux(kn); KQ_UNLOCK(kq); error = kn->kn_fop->f_event(kn, hint); KQ_LOCK(kq); kn_leave_flux(kn); if (error) KNOTE_ACTIVATE(kn, 1); KQ_UNLOCK_FLUX(kq); } else { if (kn->kn_fop->f_event(kn, hint)) KNOTE_ACTIVATE(kn, 1); KQ_UNLOCK(kq); } } if ((lockflags & KNF_LISTLOCKED) == 0) list->kl_unlock(list->kl_lockarg); } /* * add a knote to a knlist */ void knlist_add(struct knlist *knl, struct knote *kn, int islocked) { KNL_ASSERT_LOCK(knl, islocked); KQ_NOTOWNED(kn->kn_kq); KASSERT(kn_in_flux(kn), ("knote %p not in flux", kn)); KASSERT((kn->kn_status & KN_DETACHED) != 0, ("knote %p was not detached", kn)); if (!islocked) knl->kl_lock(knl->kl_lockarg); SLIST_INSERT_HEAD(&knl->kl_list, kn, kn_selnext); if (!islocked) knl->kl_unlock(knl->kl_lockarg); KQ_LOCK(kn->kn_kq); kn->kn_knlist = knl; kn->kn_status &= ~KN_DETACHED; KQ_UNLOCK(kn->kn_kq); } static void knlist_remove_kq(struct knlist *knl, struct knote *kn, int knlislocked, int kqislocked) { KASSERT(!kqislocked || knlislocked, ("kq locked w/o knl locked")); KNL_ASSERT_LOCK(knl, knlislocked); mtx_assert(&kn->kn_kq->kq_lock, kqislocked ? MA_OWNED : MA_NOTOWNED); KASSERT(kqislocked || kn_in_flux(kn), ("knote %p not in flux", kn)); KASSERT((kn->kn_status & KN_DETACHED) == 0, ("knote %p was already detached", kn)); if (!knlislocked) knl->kl_lock(knl->kl_lockarg); SLIST_REMOVE(&knl->kl_list, kn, knote, kn_selnext); kn->kn_knlist = NULL; if (!knlislocked) kn_list_unlock(knl); if (!kqislocked) KQ_LOCK(kn->kn_kq); kn->kn_status |= KN_DETACHED; if (!kqislocked) KQ_UNLOCK(kn->kn_kq); } /* * remove knote from the specified knlist */ void knlist_remove(struct knlist *knl, struct knote *kn, int islocked) { knlist_remove_kq(knl, kn, islocked, 0); } int knlist_empty(struct knlist *knl) { KNL_ASSERT_LOCKED(knl); return (SLIST_EMPTY(&knl->kl_list)); } static struct mtx knlist_lock; MTX_SYSINIT(knlist_lock, &knlist_lock, "knlist lock for lockless objects", MTX_DEF); static void knlist_mtx_lock(void *arg); static void knlist_mtx_unlock(void *arg); static void knlist_mtx_lock(void *arg) { mtx_lock((struct mtx *)arg); } static void knlist_mtx_unlock(void *arg) { mtx_unlock((struct mtx *)arg); } static void knlist_mtx_assert_lock(void *arg, int what) { if (what == LA_LOCKED) mtx_assert((struct mtx *)arg, MA_OWNED); else mtx_assert((struct mtx *)arg, MA_NOTOWNED); } static void knlist_rw_rlock(void *arg) { rw_rlock((struct rwlock *)arg); } static void knlist_rw_runlock(void *arg) { rw_runlock((struct rwlock *)arg); } static void knlist_rw_assert_lock(void *arg, int what) { if (what == LA_LOCKED) rw_assert((struct rwlock *)arg, RA_LOCKED); else rw_assert((struct rwlock *)arg, RA_UNLOCKED); } void knlist_init(struct knlist *knl, void *lock, void (*kl_lock)(void *), void (*kl_unlock)(void *), void (*kl_assert_lock)(void *, int)) { if (lock == NULL) knl->kl_lockarg = &knlist_lock; else knl->kl_lockarg = lock; if (kl_lock == NULL) knl->kl_lock = knlist_mtx_lock; else knl->kl_lock = kl_lock; if (kl_unlock == NULL) knl->kl_unlock = knlist_mtx_unlock; else knl->kl_unlock = kl_unlock; if (kl_assert_lock == NULL) knl->kl_assert_lock = knlist_mtx_assert_lock; else knl->kl_assert_lock = kl_assert_lock; knl->kl_autodestroy = 0; SLIST_INIT(&knl->kl_list); } void knlist_init_mtx(struct knlist *knl, struct mtx *lock) { knlist_init(knl, lock, NULL, NULL, NULL); } struct knlist * knlist_alloc(struct mtx *lock) { struct knlist *knl; knl = malloc(sizeof(struct knlist), M_KQUEUE, M_WAITOK); knlist_init_mtx(knl, lock); return (knl); } void knlist_init_rw_reader(struct knlist *knl, struct rwlock *lock) { knlist_init(knl, lock, knlist_rw_rlock, knlist_rw_runlock, knlist_rw_assert_lock); } void knlist_destroy(struct knlist *knl) { KASSERT(KNLIST_EMPTY(knl), ("destroying knlist %p with knotes on it", knl)); } void knlist_detach(struct knlist *knl) { KNL_ASSERT_LOCKED(knl); knl->kl_autodestroy = 1; if (knlist_empty(knl)) { knlist_destroy(knl); free(knl, M_KQUEUE); } } /* * Even if we are locked, we may need to drop the lock to allow any influx * knotes time to "settle". */ void knlist_cleardel(struct knlist *knl, struct thread *td, int islocked, int killkn) { struct knote *kn, *kn2; struct kqueue *kq; KASSERT(!knl->kl_autodestroy, ("cleardel for autodestroy %p", knl)); if (islocked) KNL_ASSERT_LOCKED(knl); else { KNL_ASSERT_UNLOCKED(knl); again: /* need to reacquire lock since we have dropped it */ knl->kl_lock(knl->kl_lockarg); } SLIST_FOREACH_SAFE(kn, &knl->kl_list, kn_selnext, kn2) { kq = kn->kn_kq; KQ_LOCK(kq); if (kn_in_flux(kn)) { KQ_UNLOCK(kq); continue; } knlist_remove_kq(knl, kn, 1, 1); if (killkn) { kn_enter_flux(kn); KQ_UNLOCK(kq); knote_drop_detached(kn, td); } else { /* Make sure cleared knotes disappear soon */ kn->kn_flags |= EV_EOF | EV_ONESHOT; KQ_UNLOCK(kq); } kq = NULL; } if (!SLIST_EMPTY(&knl->kl_list)) { /* there are still in flux knotes remaining */ kn = SLIST_FIRST(&knl->kl_list); kq = kn->kn_kq; KQ_LOCK(kq); KASSERT(kn_in_flux(kn), ("knote removed w/o list lock")); knl->kl_unlock(knl->kl_lockarg); kq->kq_state |= KQ_FLUXWAIT; msleep(kq, &kq->kq_lock, PSOCK | PDROP, "kqkclr", 0); kq = NULL; goto again; } if (islocked) KNL_ASSERT_LOCKED(knl); else { knl->kl_unlock(knl->kl_lockarg); KNL_ASSERT_UNLOCKED(knl); } } /* * Remove all knotes referencing a specified fd must be called with FILEDESC * lock. This prevents a race where a new fd comes along and occupies the * entry and we attach a knote to the fd. */ void knote_fdclose(struct thread *td, int fd) { struct filedesc *fdp = td->td_proc->p_fd; struct kqueue *kq; struct knote *kn; int influx; FILEDESC_XLOCK_ASSERT(fdp); /* * We shouldn't have to worry about new kevents appearing on fd * since filedesc is locked. */ TAILQ_FOREACH(kq, &fdp->fd_kqlist, kq_list) { KQ_LOCK(kq); again: influx = 0; while (kq->kq_knlistsize > fd && (kn = SLIST_FIRST(&kq->kq_knlist[fd])) != NULL) { if (kn_in_flux(kn)) { /* someone else might be waiting on our knote */ if (influx) wakeup(kq); kq->kq_state |= KQ_FLUXWAIT; msleep(kq, &kq->kq_lock, PSOCK, "kqflxwt", 0); goto again; } kn_enter_flux(kn); KQ_UNLOCK(kq); influx = 1; knote_drop(kn, td); KQ_LOCK(kq); } KQ_UNLOCK_FLUX(kq); } } static int knote_attach(struct knote *kn, struct kqueue *kq) { struct klist *list; KASSERT(kn_in_flux(kn), ("knote %p not marked influx", kn)); KQ_OWNED(kq); if ((kq->kq_state & KQ_CLOSING) != 0) return (EBADF); if (kn->kn_fop->f_isfd) { if (kn->kn_id >= kq->kq_knlistsize) return (ENOMEM); list = &kq->kq_knlist[kn->kn_id]; } else { if (kq->kq_knhash == NULL) return (ENOMEM); list = &kq->kq_knhash[KN_HASH(kn->kn_id, kq->kq_knhashmask)]; } SLIST_INSERT_HEAD(list, kn, kn_link); return (0); } static void knote_drop(struct knote *kn, struct thread *td) { if ((kn->kn_status & KN_DETACHED) == 0) kn->kn_fop->f_detach(kn); knote_drop_detached(kn, td); } static void knote_drop_detached(struct knote *kn, struct thread *td) { struct kqueue *kq; struct klist *list; kq = kn->kn_kq; KASSERT((kn->kn_status & KN_DETACHED) != 0, ("knote %p still attached", kn)); KQ_NOTOWNED(kq); KQ_LOCK(kq); KASSERT(kn->kn_influx == 1, ("knote_drop called on %p with influx %d", kn, kn->kn_influx)); if (kn->kn_fop->f_isfd) list = &kq->kq_knlist[kn->kn_id]; else list = &kq->kq_knhash[KN_HASH(kn->kn_id, kq->kq_knhashmask)]; if (!SLIST_EMPTY(list)) SLIST_REMOVE(list, kn, knote, kn_link); if (kn->kn_status & KN_QUEUED) knote_dequeue(kn); KQ_UNLOCK_FLUX(kq); if (kn->kn_fop->f_isfd) { fdrop(kn->kn_fp, td); kn->kn_fp = NULL; } kqueue_fo_release(kn->kn_kevent.filter); kn->kn_fop = NULL; knote_free(kn); } static void knote_enqueue(struct knote *kn) { struct kqueue *kq = kn->kn_kq; KQ_OWNED(kn->kn_kq); KASSERT((kn->kn_status & KN_QUEUED) == 0, ("knote already queued")); TAILQ_INSERT_TAIL(&kq->kq_head, kn, kn_tqe); kn->kn_status |= KN_QUEUED; kq->kq_count++; kqueue_wakeup(kq); } static void knote_dequeue(struct knote *kn) { struct kqueue *kq = kn->kn_kq; KQ_OWNED(kn->kn_kq); KASSERT(kn->kn_status & KN_QUEUED, ("knote not queued")); TAILQ_REMOVE(&kq->kq_head, kn, kn_tqe); kn->kn_status &= ~KN_QUEUED; kq->kq_count--; } static void knote_init(void) { knote_zone = uma_zcreate("KNOTE", sizeof(struct knote), NULL, NULL, NULL, NULL, UMA_ALIGN_PTR, 0); } SYSINIT(knote, SI_SUB_PSEUDO, SI_ORDER_ANY, knote_init, NULL); static struct knote * knote_alloc(int mflag) { return (uma_zalloc(knote_zone, mflag | M_ZERO)); } static void knote_free(struct knote *kn) { uma_zfree(knote_zone, kn); } /* * Register the kev w/ the kq specified by fd. */ int kqfd_register(int fd, struct kevent *kev, struct thread *td, int mflag) { struct kqueue *kq; struct file *fp; cap_rights_t rights; int error; error = fget(td, fd, cap_rights_init_one(&rights, CAP_KQUEUE_CHANGE), &fp); if (error != 0) return (error); if ((error = kqueue_acquire(fp, &kq)) != 0) goto noacquire; error = kqueue_register(kq, kev, td, mflag); kqueue_release(kq, 0); noacquire: fdrop(fp, td); return (error); } diff --git a/sys/kern/sys_eventfd.c b/sys/kern/sys_eventfd.c index 8ee8421296e5..10566b486f30 100644 --- a/sys/kern/sys_eventfd.c +++ b/sys/kern/sys_eventfd.c @@ -1,346 +1,347 @@ /*- * SPDX-License-Identifier: BSD-2-Clause-FreeBSD * * Copyright (c) 2014 Dmitry Chagin * * 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. */ #include __FBSDID("$FreeBSD$"); #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include _Static_assert(EFD_CLOEXEC == O_CLOEXEC, "Mismatched EFD_CLOEXEC"); _Static_assert(EFD_NONBLOCK == O_NONBLOCK, "Mismatched EFD_NONBLOCK"); MALLOC_DEFINE(M_EVENTFD, "eventfd", "eventfd structures"); static fo_rdwr_t eventfd_read; static fo_rdwr_t eventfd_write; static fo_ioctl_t eventfd_ioctl; static fo_poll_t eventfd_poll; static fo_kqfilter_t eventfd_kqfilter; static fo_stat_t eventfd_stat; static fo_close_t eventfd_close; static fo_fill_kinfo_t eventfd_fill_kinfo; static struct fileops eventfdops = { .fo_read = eventfd_read, .fo_write = eventfd_write, .fo_truncate = invfo_truncate, .fo_ioctl = eventfd_ioctl, .fo_poll = eventfd_poll, .fo_kqfilter = eventfd_kqfilter, .fo_stat = eventfd_stat, .fo_close = eventfd_close, .fo_chmod = invfo_chmod, .fo_chown = invfo_chown, .fo_sendfile = invfo_sendfile, .fo_fill_kinfo = eventfd_fill_kinfo, .fo_flags = DFLAG_PASSABLE }; static void filt_eventfddetach(struct knote *kn); static int filt_eventfdread(struct knote *kn, long hint); static int filt_eventfdwrite(struct knote *kn, long hint); static struct filterops eventfd_rfiltops = { .f_isfd = 1, .f_detach = filt_eventfddetach, .f_event = filt_eventfdread }; static struct filterops eventfd_wfiltops = { .f_isfd = 1, .f_detach = filt_eventfddetach, .f_event = filt_eventfdwrite }; struct eventfd { eventfd_t efd_count; uint32_t efd_flags; struct selinfo efd_sel; struct mtx efd_lock; }; int eventfd_create_file(struct thread *td, struct file *fp, uint32_t initval, int flags) { struct eventfd *efd; int fflags; AUDIT_ARG_FFLAGS(flags); AUDIT_ARG_VALUE(initval); efd = malloc(sizeof(*efd), M_EVENTFD, M_WAITOK | M_ZERO); efd->efd_flags = flags; efd->efd_count = initval; mtx_init(&efd->efd_lock, "eventfd", NULL, MTX_DEF); knlist_init_mtx(&efd->efd_sel.si_note, &efd->efd_lock); fflags = FREAD | FWRITE; if ((flags & EFD_NONBLOCK) != 0) fflags |= FNONBLOCK; finit(fp, fflags, DTYPE_EVENTFD, efd, &eventfdops); return (0); } static int eventfd_close(struct file *fp, struct thread *td) { struct eventfd *efd; efd = fp->f_data; seldrain(&efd->efd_sel); knlist_destroy(&efd->efd_sel.si_note); mtx_destroy(&efd->efd_lock); free(efd, M_EVENTFD); return (0); } static int eventfd_read(struct file *fp, struct uio *uio, struct ucred *active_cred, int flags, struct thread *td) { struct eventfd *efd; eventfd_t count; int error; if (uio->uio_resid < sizeof(eventfd_t)) return (EINVAL); error = 0; efd = fp->f_data; mtx_lock(&efd->efd_lock); while (error == 0 && efd->efd_count == 0) { if ((fp->f_flag & FNONBLOCK) != 0) { mtx_unlock(&efd->efd_lock); return (EAGAIN); } error = mtx_sleep(&efd->efd_count, &efd->efd_lock, PCATCH, "efdrd", 0); } if (error == 0) { MPASS(efd->efd_count > 0); if ((efd->efd_flags & EFD_SEMAPHORE) != 0) { count = 1; --efd->efd_count; } else { count = efd->efd_count; efd->efd_count = 0; } KNOTE_LOCKED(&efd->efd_sel.si_note, 0); selwakeup(&efd->efd_sel); wakeup(&efd->efd_count); mtx_unlock(&efd->efd_lock); error = uiomove(&count, sizeof(eventfd_t), uio); } else mtx_unlock(&efd->efd_lock); return (error); } static int eventfd_write(struct file *fp, struct uio *uio, struct ucred *active_cred, int flags, struct thread *td) { struct eventfd *efd; eventfd_t count; int error; if (uio->uio_resid < sizeof(eventfd_t)) return (EINVAL); error = uiomove(&count, sizeof(eventfd_t), uio); if (error != 0) return (error); if (count == UINT64_MAX) return (EINVAL); efd = fp->f_data; mtx_lock(&efd->efd_lock); retry: if (UINT64_MAX - efd->efd_count <= count) { if ((fp->f_flag & FNONBLOCK) != 0) { mtx_unlock(&efd->efd_lock); /* Do not not return the number of bytes written */ uio->uio_resid += sizeof(eventfd_t); return (EAGAIN); } error = mtx_sleep(&efd->efd_count, &efd->efd_lock, PCATCH, "efdwr", 0); if (error == 0) goto retry; } if (error == 0) { MPASS(UINT64_MAX - efd->efd_count > count); efd->efd_count += count; KNOTE_LOCKED(&efd->efd_sel.si_note, 0); selwakeup(&efd->efd_sel); wakeup(&efd->efd_count); } mtx_unlock(&efd->efd_lock); return (error); } static int eventfd_poll(struct file *fp, int events, struct ucred *active_cred, struct thread *td) { struct eventfd *efd; int revents; efd = fp->f_data; revents = 0; mtx_lock(&efd->efd_lock); if ((events & (POLLIN | POLLRDNORM)) != 0 && efd->efd_count > 0) revents |= events & (POLLIN | POLLRDNORM); if ((events & (POLLOUT | POLLWRNORM)) != 0 && UINT64_MAX - 1 > efd->efd_count) revents |= events & (POLLOUT | POLLWRNORM); if (revents == 0) selrecord(td, &efd->efd_sel); mtx_unlock(&efd->efd_lock); return (revents); } static int eventfd_kqfilter(struct file *fp, struct knote *kn) { struct eventfd *efd = fp->f_data; mtx_lock(&efd->efd_lock); switch (kn->kn_filter) { case EVFILT_READ: kn->kn_fop = &eventfd_rfiltops; break; case EVFILT_WRITE: kn->kn_fop = &eventfd_wfiltops; break; default: mtx_unlock(&efd->efd_lock); return (EINVAL); } kn->kn_hook = efd; knlist_add(&efd->efd_sel.si_note, kn, 1); mtx_unlock(&efd->efd_lock); return (0); } static void filt_eventfddetach(struct knote *kn) { struct eventfd *efd = kn->kn_hook; mtx_lock(&efd->efd_lock); knlist_remove(&efd->efd_sel.si_note, kn, 1); mtx_unlock(&efd->efd_lock); } static int filt_eventfdread(struct knote *kn, long hint) { struct eventfd *efd = kn->kn_hook; int ret; mtx_assert(&efd->efd_lock, MA_OWNED); kn->kn_data = (int64_t)efd->efd_count; ret = efd->efd_count > 0; return (ret); } static int filt_eventfdwrite(struct knote *kn, long hint) { struct eventfd *efd = kn->kn_hook; int ret; mtx_assert(&efd->efd_lock, MA_OWNED); kn->kn_data = (int64_t)(UINT64_MAX - 1 - efd->efd_count); ret = UINT64_MAX - 1 > efd->efd_count; return (ret); } static int eventfd_ioctl(struct file *fp, u_long cmd, void *data, struct ucred *active_cred, struct thread *td) { switch (cmd) { case FIONBIO: case FIOASYNC: return (0); } return (ENOTTY); } static int eventfd_stat(struct file *fp, struct stat *st, struct ucred *active_cred) { bzero((void *)st, sizeof *st); st->st_mode = S_IFIFO; return (0); } static int eventfd_fill_kinfo(struct file *fp, struct kinfo_file *kif, struct filedesc *fdp) { struct eventfd *efd = fp->f_data; kif->kf_type = KF_TYPE_EVENTFD; mtx_lock(&efd->efd_lock); kif->kf_un.kf_eventfd.kf_eventfd_value = efd->efd_count; kif->kf_un.kf_eventfd.kf_eventfd_flags = efd->efd_flags; + kif->kf_un.kf_eventfd.kf_eventfd_addr = (uintptr_t)efd; mtx_unlock(&efd->efd_lock); return (0); } diff --git a/sys/kern/sys_pipe.c b/sys/kern/sys_pipe.c index ed93a2b7f094..d9e7d97d01e1 100644 --- a/sys/kern/sys_pipe.c +++ b/sys/kern/sys_pipe.c @@ -1,1847 +1,1850 @@ /*- * SPDX-License-Identifier: BSD-4-Clause * * Copyright (c) 1996 John S. Dyson * Copyright (c) 2012 Giovanni Trematerra * 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 immediately at the beginning of the file, without modification, * 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. * 3. Absolutely no warranty of function or purpose is made by the author * John S. Dyson. * 4. Modifications may be freely made to this file if the above conditions * are met. */ /* * This file contains a high-performance replacement for the socket-based * pipes scheme originally used in FreeBSD/4.4Lite. It does not support * all features of sockets, but does do everything that pipes normally * do. */ /* * This code has two modes of operation, a small write mode and a large * write mode. The small write mode acts like conventional pipes with * a kernel buffer. If the buffer is less than PIPE_MINDIRECT, then the * "normal" pipe buffering is done. If the buffer is between PIPE_MINDIRECT * and PIPE_SIZE in size, the sending process pins the underlying pages in * memory, and the receiving process copies directly from these pinned pages * in the sending process. * * If the sending process receives a signal, it is possible that it will * go away, and certainly its address space can change, because control * is returned back to the user-mode side. In that case, the pipe code * arranges to copy the buffer supplied by the user process, to a pageable * kernel buffer, and the receiving process will grab the data from the * pageable kernel buffer. Since signals don't happen all that often, * the copy operation is normally eliminated. * * The constant PIPE_MINDIRECT is chosen to make sure that buffering will * happen for small transfers so that the system will not spend all of * its time context switching. * * In order to limit the resource use of pipes, two sysctls exist: * * kern.ipc.maxpipekva - This is a hard limit on the amount of pageable * address space available to us in pipe_map. This value is normally * autotuned, but may also be loader tuned. * * kern.ipc.pipekva - This read-only sysctl tracks the current amount of * memory in use by pipes. * * Based on how large pipekva is relative to maxpipekva, the following * will happen: * * 0% - 50%: * New pipes are given 16K of memory backing, pipes may dynamically * grow to as large as 64K where needed. * 50% - 75%: * New pipes are given 4K (or PAGE_SIZE) of memory backing, * existing pipes may NOT grow. * 75% - 100%: * New pipes are given 4K (or PAGE_SIZE) of memory backing, * existing pipes will be shrunk down to 4K whenever possible. * * Resizing may be disabled by setting kern.ipc.piperesizeallowed=0. If * that is set, the only resize that will occur is the 0 -> SMALL_PIPE_SIZE * resize which MUST occur for reverse-direction pipes when they are * first used. * * Additional information about the current state of pipes may be obtained * from kern.ipc.pipes, kern.ipc.pipefragretry, kern.ipc.pipeallocfail, * and kern.ipc.piperesizefail. * * Locking rules: There are two locks present here: A mutex, used via * PIPE_LOCK, and a flag, used via pipelock(). All locking is done via * the flag, as mutexes can not persist over uiomove. The mutex * exists only to guard access to the flag, and is not in itself a * locking mechanism. Also note that there is only a single mutex for * both directions of a pipe. * * As pipelock() may have to sleep before it can acquire the flag, it * is important to reread all data after a call to pipelock(); everything * in the structure may have changed. */ #include __FBSDID("$FreeBSD$"); #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include /* * Use this define if you want to disable *fancy* VM things. Expect an * approx 30% decrease in transfer rate. This could be useful for * NetBSD or OpenBSD. */ /* #define PIPE_NODIRECT */ #define PIPE_PEER(pipe) \ (((pipe)->pipe_type & PIPE_TYPE_NAMED) ? (pipe) : ((pipe)->pipe_peer)) /* * interfaces to the outside world */ static fo_rdwr_t pipe_read; static fo_rdwr_t pipe_write; static fo_truncate_t pipe_truncate; static fo_ioctl_t pipe_ioctl; static fo_poll_t pipe_poll; static fo_kqfilter_t pipe_kqfilter; static fo_stat_t pipe_stat; static fo_close_t pipe_close; static fo_chmod_t pipe_chmod; static fo_chown_t pipe_chown; static fo_fill_kinfo_t pipe_fill_kinfo; struct fileops pipeops = { .fo_read = pipe_read, .fo_write = pipe_write, .fo_truncate = pipe_truncate, .fo_ioctl = pipe_ioctl, .fo_poll = pipe_poll, .fo_kqfilter = pipe_kqfilter, .fo_stat = pipe_stat, .fo_close = pipe_close, .fo_chmod = pipe_chmod, .fo_chown = pipe_chown, .fo_sendfile = invfo_sendfile, .fo_fill_kinfo = pipe_fill_kinfo, .fo_flags = DFLAG_PASSABLE }; static void filt_pipedetach(struct knote *kn); static void filt_pipedetach_notsup(struct knote *kn); static int filt_pipenotsup(struct knote *kn, long hint); static int filt_piperead(struct knote *kn, long hint); static int filt_pipewrite(struct knote *kn, long hint); static struct filterops pipe_nfiltops = { .f_isfd = 1, .f_detach = filt_pipedetach_notsup, .f_event = filt_pipenotsup }; static struct filterops pipe_rfiltops = { .f_isfd = 1, .f_detach = filt_pipedetach, .f_event = filt_piperead }; static struct filterops pipe_wfiltops = { .f_isfd = 1, .f_detach = filt_pipedetach, .f_event = filt_pipewrite }; /* * Default pipe buffer size(s), this can be kind-of large now because pipe * space is pageable. The pipe code will try to maintain locality of * reference for performance reasons, so small amounts of outstanding I/O * will not wipe the cache. */ #define MINPIPESIZE (PIPE_SIZE/3) #define MAXPIPESIZE (2*PIPE_SIZE/3) static long amountpipekva; static int pipefragretry; static int pipeallocfail; static int piperesizefail; static int piperesizeallowed = 1; static long pipe_mindirect = PIPE_MINDIRECT; SYSCTL_LONG(_kern_ipc, OID_AUTO, maxpipekva, CTLFLAG_RDTUN | CTLFLAG_NOFETCH, &maxpipekva, 0, "Pipe KVA limit"); SYSCTL_LONG(_kern_ipc, OID_AUTO, pipekva, CTLFLAG_RD, &amountpipekva, 0, "Pipe KVA usage"); SYSCTL_INT(_kern_ipc, OID_AUTO, pipefragretry, CTLFLAG_RD, &pipefragretry, 0, "Pipe allocation retries due to fragmentation"); SYSCTL_INT(_kern_ipc, OID_AUTO, pipeallocfail, CTLFLAG_RD, &pipeallocfail, 0, "Pipe allocation failures"); SYSCTL_INT(_kern_ipc, OID_AUTO, piperesizefail, CTLFLAG_RD, &piperesizefail, 0, "Pipe resize failures"); SYSCTL_INT(_kern_ipc, OID_AUTO, piperesizeallowed, CTLFLAG_RW, &piperesizeallowed, 0, "Pipe resizing allowed"); static void pipeinit(void *dummy __unused); static void pipeclose(struct pipe *cpipe); static void pipe_free_kmem(struct pipe *cpipe); static int pipe_create(struct pipe *pipe, bool backing); static int pipe_paircreate(struct thread *td, struct pipepair **p_pp); static __inline int pipelock(struct pipe *cpipe, int catch); static __inline void pipeunlock(struct pipe *cpipe); static void pipe_timestamp(struct timespec *tsp); #ifndef PIPE_NODIRECT static int pipe_build_write_buffer(struct pipe *wpipe, struct uio *uio); static void pipe_destroy_write_buffer(struct pipe *wpipe); static int pipe_direct_write(struct pipe *wpipe, struct uio *uio); static void pipe_clone_write_buffer(struct pipe *wpipe); #endif static int pipespace(struct pipe *cpipe, int size); static int pipespace_new(struct pipe *cpipe, int size); static int pipe_zone_ctor(void *mem, int size, void *arg, int flags); static int pipe_zone_init(void *mem, int size, int flags); static void pipe_zone_fini(void *mem, int size); static uma_zone_t pipe_zone; static struct unrhdr64 pipeino_unr; static dev_t pipedev_ino; SYSINIT(vfs, SI_SUB_VFS, SI_ORDER_ANY, pipeinit, NULL); static void pipeinit(void *dummy __unused) { pipe_zone = uma_zcreate("pipe", sizeof(struct pipepair), pipe_zone_ctor, NULL, pipe_zone_init, pipe_zone_fini, UMA_ALIGN_PTR, 0); KASSERT(pipe_zone != NULL, ("pipe_zone not initialized")); new_unrhdr64(&pipeino_unr, 1); pipedev_ino = devfs_alloc_cdp_inode(); KASSERT(pipedev_ino > 0, ("pipe dev inode not initialized")); } static int sysctl_handle_pipe_mindirect(SYSCTL_HANDLER_ARGS) { int error = 0; long tmp_pipe_mindirect = pipe_mindirect; error = sysctl_handle_long(oidp, &tmp_pipe_mindirect, arg2, req); if (error != 0 || req->newptr == NULL) return (error); /* * Don't allow pipe_mindirect to be set so low that we violate * atomicity requirements. */ if (tmp_pipe_mindirect <= PIPE_BUF) return (EINVAL); pipe_mindirect = tmp_pipe_mindirect; return (0); } SYSCTL_OID(_kern_ipc, OID_AUTO, pipe_mindirect, CTLTYPE_LONG | CTLFLAG_RW, &pipe_mindirect, 0, sysctl_handle_pipe_mindirect, "L", "Minimum write size triggering VM optimization"); static int pipe_zone_ctor(void *mem, int size, void *arg, int flags) { struct pipepair *pp; struct pipe *rpipe, *wpipe; KASSERT(size == sizeof(*pp), ("pipe_zone_ctor: wrong size")); pp = (struct pipepair *)mem; /* * We zero both pipe endpoints to make sure all the kmem pointers * are NULL, flag fields are zero'd, etc. We timestamp both * endpoints with the same time. */ rpipe = &pp->pp_rpipe; bzero(rpipe, sizeof(*rpipe)); pipe_timestamp(&rpipe->pipe_ctime); rpipe->pipe_atime = rpipe->pipe_mtime = rpipe->pipe_ctime; wpipe = &pp->pp_wpipe; bzero(wpipe, sizeof(*wpipe)); wpipe->pipe_ctime = rpipe->pipe_ctime; wpipe->pipe_atime = wpipe->pipe_mtime = rpipe->pipe_ctime; rpipe->pipe_peer = wpipe; rpipe->pipe_pair = pp; wpipe->pipe_peer = rpipe; wpipe->pipe_pair = pp; /* * Mark both endpoints as present; they will later get free'd * one at a time. When both are free'd, then the whole pair * is released. */ rpipe->pipe_present = PIPE_ACTIVE; wpipe->pipe_present = PIPE_ACTIVE; /* * Eventually, the MAC Framework may initialize the label * in ctor or init, but for now we do it elswhere to avoid * blocking in ctor or init. */ pp->pp_label = NULL; return (0); } static int pipe_zone_init(void *mem, int size, int flags) { struct pipepair *pp; KASSERT(size == sizeof(*pp), ("pipe_zone_init: wrong size")); pp = (struct pipepair *)mem; mtx_init(&pp->pp_mtx, "pipe mutex", NULL, MTX_DEF | MTX_NEW); return (0); } static void pipe_zone_fini(void *mem, int size) { struct pipepair *pp; KASSERT(size == sizeof(*pp), ("pipe_zone_fini: wrong size")); pp = (struct pipepair *)mem; mtx_destroy(&pp->pp_mtx); } static int pipe_paircreate(struct thread *td, struct pipepair **p_pp) { struct pipepair *pp; struct pipe *rpipe, *wpipe; int error; *p_pp = pp = uma_zalloc(pipe_zone, M_WAITOK); #ifdef MAC /* * The MAC label is shared between the connected endpoints. As a * result mac_pipe_init() and mac_pipe_create() are called once * for the pair, and not on the endpoints. */ mac_pipe_init(pp); mac_pipe_create(td->td_ucred, pp); #endif rpipe = &pp->pp_rpipe; wpipe = &pp->pp_wpipe; knlist_init_mtx(&rpipe->pipe_sel.si_note, PIPE_MTX(rpipe)); knlist_init_mtx(&wpipe->pipe_sel.si_note, PIPE_MTX(wpipe)); /* * Only the forward direction pipe is backed by big buffer by * default. */ error = pipe_create(rpipe, true); if (error != 0) goto fail; error = pipe_create(wpipe, false); if (error != 0) { /* * This cleanup leaves the pipe inode number for rpipe * still allocated, but never used. We do not free * inode numbers for opened pipes, which is required * for correctness because numbers must be unique. * But also it avoids any memory use by the unr * allocator, so stashing away the transient inode * number is reasonable. */ pipe_free_kmem(rpipe); goto fail; } rpipe->pipe_state |= PIPE_DIRECTOK; wpipe->pipe_state |= PIPE_DIRECTOK; return (0); fail: knlist_destroy(&rpipe->pipe_sel.si_note); knlist_destroy(&wpipe->pipe_sel.si_note); #ifdef MAC mac_pipe_destroy(pp); #endif uma_zfree(pipe_zone, pp); return (error); } int pipe_named_ctor(struct pipe **ppipe, struct thread *td) { struct pipepair *pp; int error; error = pipe_paircreate(td, &pp); if (error != 0) return (error); pp->pp_rpipe.pipe_type |= PIPE_TYPE_NAMED; *ppipe = &pp->pp_rpipe; return (0); } void pipe_dtor(struct pipe *dpipe) { struct pipe *peer; peer = (dpipe->pipe_type & PIPE_TYPE_NAMED) != 0 ? dpipe->pipe_peer : NULL; funsetown(&dpipe->pipe_sigio); pipeclose(dpipe); if (peer != NULL) { funsetown(&peer->pipe_sigio); pipeclose(peer); } } /* * Get a timestamp. * * This used to be vfs_timestamp but the higher precision is unnecessary and * can very negatively affect performance in virtualized environments (e.g., on * vms running on amd64 when using the rdtscp instruction). */ static void pipe_timestamp(struct timespec *tsp) { getnanotime(tsp); } /* * The pipe system call for the DTYPE_PIPE type of pipes. If we fail, let * the zone pick up the pieces via pipeclose(). */ int kern_pipe(struct thread *td, int fildes[2], int flags, struct filecaps *fcaps1, struct filecaps *fcaps2) { struct file *rf, *wf; struct pipe *rpipe, *wpipe; struct pipepair *pp; int fd, fflags, error; error = pipe_paircreate(td, &pp); if (error != 0) return (error); rpipe = &pp->pp_rpipe; wpipe = &pp->pp_wpipe; error = falloc_caps(td, &rf, &fd, flags, fcaps1); if (error) { pipeclose(rpipe); pipeclose(wpipe); return (error); } /* An extra reference on `rf' has been held for us by falloc_caps(). */ fildes[0] = fd; fflags = FREAD | FWRITE; if ((flags & O_NONBLOCK) != 0) fflags |= FNONBLOCK; /* * Warning: once we've gotten past allocation of the fd for the * read-side, we can only drop the read side via fdrop() in order * to avoid races against processes which manage to dup() the read * side while we are blocked trying to allocate the write side. */ finit(rf, fflags, DTYPE_PIPE, rpipe, &pipeops); error = falloc_caps(td, &wf, &fd, flags, fcaps2); if (error) { fdclose(td, rf, fildes[0]); fdrop(rf, td); /* rpipe has been closed by fdrop(). */ pipeclose(wpipe); return (error); } /* An extra reference on `wf' has been held for us by falloc_caps(). */ finit(wf, fflags, DTYPE_PIPE, wpipe, &pipeops); fdrop(wf, td); fildes[1] = fd; fdrop(rf, td); return (0); } #ifdef COMPAT_FREEBSD10 /* ARGSUSED */ int freebsd10_pipe(struct thread *td, struct freebsd10_pipe_args *uap __unused) { int error; int fildes[2]; error = kern_pipe(td, fildes, 0, NULL, NULL); if (error) return (error); td->td_retval[0] = fildes[0]; td->td_retval[1] = fildes[1]; return (0); } #endif int sys_pipe2(struct thread *td, struct pipe2_args *uap) { int error, fildes[2]; if (uap->flags & ~(O_CLOEXEC | O_NONBLOCK)) return (EINVAL); error = kern_pipe(td, fildes, uap->flags, NULL, NULL); if (error) return (error); error = copyout(fildes, uap->fildes, 2 * sizeof(int)); if (error) { (void)kern_close(td, fildes[0]); (void)kern_close(td, fildes[1]); } return (error); } /* * Allocate kva for pipe circular buffer, the space is pageable * This routine will 'realloc' the size of a pipe safely, if it fails * it will retain the old buffer. * If it fails it will return ENOMEM. */ static int pipespace_new(struct pipe *cpipe, int size) { caddr_t buffer; int error, cnt, firstseg; static int curfail = 0; static struct timeval lastfail; KASSERT(!mtx_owned(PIPE_MTX(cpipe)), ("pipespace: pipe mutex locked")); KASSERT(!(cpipe->pipe_state & PIPE_DIRECTW), ("pipespace: resize of direct writes not allowed")); retry: cnt = cpipe->pipe_buffer.cnt; if (cnt > size) size = cnt; size = round_page(size); buffer = (caddr_t) vm_map_min(pipe_map); error = vm_map_find(pipe_map, NULL, 0, (vm_offset_t *)&buffer, size, 0, VMFS_ANY_SPACE, VM_PROT_RW, VM_PROT_RW, 0); if (error != KERN_SUCCESS) { if (cpipe->pipe_buffer.buffer == NULL && size > SMALL_PIPE_SIZE) { size = SMALL_PIPE_SIZE; pipefragretry++; goto retry; } if (cpipe->pipe_buffer.buffer == NULL) { pipeallocfail++; if (ppsratecheck(&lastfail, &curfail, 1)) printf("kern.ipc.maxpipekva exceeded; see tuning(7)\n"); } else { piperesizefail++; } return (ENOMEM); } /* copy data, then free old resources if we're resizing */ if (cnt > 0) { if (cpipe->pipe_buffer.in <= cpipe->pipe_buffer.out) { firstseg = cpipe->pipe_buffer.size - cpipe->pipe_buffer.out; bcopy(&cpipe->pipe_buffer.buffer[cpipe->pipe_buffer.out], buffer, firstseg); if ((cnt - firstseg) > 0) bcopy(cpipe->pipe_buffer.buffer, &buffer[firstseg], cpipe->pipe_buffer.in); } else { bcopy(&cpipe->pipe_buffer.buffer[cpipe->pipe_buffer.out], buffer, cnt); } } pipe_free_kmem(cpipe); cpipe->pipe_buffer.buffer = buffer; cpipe->pipe_buffer.size = size; cpipe->pipe_buffer.in = cnt; cpipe->pipe_buffer.out = 0; cpipe->pipe_buffer.cnt = cnt; atomic_add_long(&amountpipekva, cpipe->pipe_buffer.size); return (0); } /* * Wrapper for pipespace_new() that performs locking assertions. */ static int pipespace(struct pipe *cpipe, int size) { KASSERT(cpipe->pipe_state & PIPE_LOCKFL, ("Unlocked pipe passed to pipespace")); return (pipespace_new(cpipe, size)); } /* * lock a pipe for I/O, blocking other access */ static __inline int pipelock(struct pipe *cpipe, int catch) { int error, prio; PIPE_LOCK_ASSERT(cpipe, MA_OWNED); prio = PRIBIO; if (catch) prio |= PCATCH; while (cpipe->pipe_state & PIPE_LOCKFL) { KASSERT(cpipe->pipe_waiters >= 0, ("%s: bad waiter count %d", __func__, cpipe->pipe_waiters)); cpipe->pipe_waiters++; error = msleep(&cpipe->pipe_waiters, PIPE_MTX(cpipe), prio, "pipelk", 0); cpipe->pipe_waiters--; if (error != 0) return (error); } cpipe->pipe_state |= PIPE_LOCKFL; return (0); } /* * unlock a pipe I/O lock */ static __inline void pipeunlock(struct pipe *cpipe) { PIPE_LOCK_ASSERT(cpipe, MA_OWNED); KASSERT(cpipe->pipe_state & PIPE_LOCKFL, ("Unlocked pipe passed to pipeunlock")); KASSERT(cpipe->pipe_waiters >= 0, ("%s: bad waiter count %d", __func__, cpipe->pipe_waiters)); cpipe->pipe_state &= ~PIPE_LOCKFL; if (cpipe->pipe_waiters > 0) wakeup_one(&cpipe->pipe_waiters); } void pipeselwakeup(struct pipe *cpipe) { PIPE_LOCK_ASSERT(cpipe, MA_OWNED); if (cpipe->pipe_state & PIPE_SEL) { selwakeuppri(&cpipe->pipe_sel, PSOCK); if (!SEL_WAITING(&cpipe->pipe_sel)) cpipe->pipe_state &= ~PIPE_SEL; } if ((cpipe->pipe_state & PIPE_ASYNC) && cpipe->pipe_sigio) pgsigio(&cpipe->pipe_sigio, SIGIO, 0); KNOTE_LOCKED(&cpipe->pipe_sel.si_note, 0); } /* * Initialize and allocate VM and memory for pipe. The structure * will start out zero'd from the ctor, so we just manage the kmem. */ static int pipe_create(struct pipe *pipe, bool large_backing) { int error; error = pipespace_new(pipe, !large_backing || amountpipekva > maxpipekva / 2 ? SMALL_PIPE_SIZE : PIPE_SIZE); if (error == 0) pipe->pipe_ino = alloc_unr64(&pipeino_unr); return (error); } /* ARGSUSED */ static int pipe_read(struct file *fp, struct uio *uio, struct ucred *active_cred, int flags, struct thread *td) { struct pipe *rpipe; int error; int nread = 0; int size; rpipe = fp->f_data; PIPE_LOCK(rpipe); ++rpipe->pipe_busy; error = pipelock(rpipe, 1); if (error) goto unlocked_error; #ifdef MAC error = mac_pipe_check_read(active_cred, rpipe->pipe_pair); if (error) goto locked_error; #endif if (amountpipekva > (3 * maxpipekva) / 4) { if ((rpipe->pipe_state & PIPE_DIRECTW) == 0 && rpipe->pipe_buffer.size > SMALL_PIPE_SIZE && rpipe->pipe_buffer.cnt <= SMALL_PIPE_SIZE && piperesizeallowed == 1) { PIPE_UNLOCK(rpipe); pipespace(rpipe, SMALL_PIPE_SIZE); PIPE_LOCK(rpipe); } } while (uio->uio_resid) { /* * normal pipe buffer receive */ if (rpipe->pipe_buffer.cnt > 0) { size = rpipe->pipe_buffer.size - rpipe->pipe_buffer.out; if (size > rpipe->pipe_buffer.cnt) size = rpipe->pipe_buffer.cnt; if (size > uio->uio_resid) size = uio->uio_resid; PIPE_UNLOCK(rpipe); error = uiomove( &rpipe->pipe_buffer.buffer[rpipe->pipe_buffer.out], size, uio); PIPE_LOCK(rpipe); if (error) break; rpipe->pipe_buffer.out += size; if (rpipe->pipe_buffer.out >= rpipe->pipe_buffer.size) rpipe->pipe_buffer.out = 0; rpipe->pipe_buffer.cnt -= size; /* * If there is no more to read in the pipe, reset * its pointers to the beginning. This improves * cache hit stats. */ if (rpipe->pipe_buffer.cnt == 0) { rpipe->pipe_buffer.in = 0; rpipe->pipe_buffer.out = 0; } nread += size; #ifndef PIPE_NODIRECT /* * Direct copy, bypassing a kernel buffer. */ } else if ((size = rpipe->pipe_pages.cnt) != 0) { if (size > uio->uio_resid) size = (u_int) uio->uio_resid; PIPE_UNLOCK(rpipe); error = uiomove_fromphys(rpipe->pipe_pages.ms, rpipe->pipe_pages.pos, size, uio); PIPE_LOCK(rpipe); if (error) break; nread += size; rpipe->pipe_pages.pos += size; rpipe->pipe_pages.cnt -= size; if (rpipe->pipe_pages.cnt == 0) { rpipe->pipe_state &= ~PIPE_WANTW; wakeup(rpipe); } #endif } else { /* * detect EOF condition * read returns 0 on EOF, no need to set error */ if (rpipe->pipe_state & PIPE_EOF) break; /* * If the "write-side" has been blocked, wake it up now. */ if (rpipe->pipe_state & PIPE_WANTW) { rpipe->pipe_state &= ~PIPE_WANTW; wakeup(rpipe); } /* * Break if some data was read. */ if (nread > 0) break; /* * Unlock the pipe buffer for our remaining processing. * We will either break out with an error or we will * sleep and relock to loop. */ pipeunlock(rpipe); /* * Handle non-blocking mode operation or * wait for more data. */ if (fp->f_flag & FNONBLOCK) { error = EAGAIN; } else { rpipe->pipe_state |= PIPE_WANTR; if ((error = msleep(rpipe, PIPE_MTX(rpipe), PRIBIO | PCATCH, "piperd", 0)) == 0) error = pipelock(rpipe, 1); } if (error) goto unlocked_error; } } #ifdef MAC locked_error: #endif pipeunlock(rpipe); /* XXX: should probably do this before getting any locks. */ if (error == 0) pipe_timestamp(&rpipe->pipe_atime); unlocked_error: --rpipe->pipe_busy; /* * PIPE_WANT processing only makes sense if pipe_busy is 0. */ if ((rpipe->pipe_busy == 0) && (rpipe->pipe_state & PIPE_WANT)) { rpipe->pipe_state &= ~(PIPE_WANT|PIPE_WANTW); wakeup(rpipe); } else if (rpipe->pipe_buffer.cnt < MINPIPESIZE) { /* * Handle write blocking hysteresis. */ if (rpipe->pipe_state & PIPE_WANTW) { rpipe->pipe_state &= ~PIPE_WANTW; wakeup(rpipe); } } /* * Only wake up writers if there was actually something read. * Otherwise, when calling read(2) at EOF, a spurious wakeup occurs. */ if (nread > 0 && rpipe->pipe_buffer.size - rpipe->pipe_buffer.cnt >= PIPE_BUF) pipeselwakeup(rpipe); PIPE_UNLOCK(rpipe); if (nread > 0) td->td_ru.ru_msgrcv++; return (error); } #ifndef PIPE_NODIRECT /* * Map the sending processes' buffer into kernel space and wire it. * This is similar to a physical write operation. */ static int pipe_build_write_buffer(struct pipe *wpipe, struct uio *uio) { u_int size; int i; PIPE_LOCK_ASSERT(wpipe, MA_OWNED); KASSERT((wpipe->pipe_state & PIPE_DIRECTW) == 0, ("%s: PIPE_DIRECTW set on %p", __func__, wpipe)); KASSERT(wpipe->pipe_pages.cnt == 0, ("%s: pipe map for %p contains residual data", __func__, wpipe)); if (uio->uio_iov->iov_len > wpipe->pipe_buffer.size) size = wpipe->pipe_buffer.size; else size = uio->uio_iov->iov_len; wpipe->pipe_state |= PIPE_DIRECTW; PIPE_UNLOCK(wpipe); i = vm_fault_quick_hold_pages(&curproc->p_vmspace->vm_map, (vm_offset_t)uio->uio_iov->iov_base, size, VM_PROT_READ, wpipe->pipe_pages.ms, PIPENPAGES); PIPE_LOCK(wpipe); if (i < 0) { wpipe->pipe_state &= ~PIPE_DIRECTW; return (EFAULT); } wpipe->pipe_pages.npages = i; wpipe->pipe_pages.pos = ((vm_offset_t) uio->uio_iov->iov_base) & PAGE_MASK; wpipe->pipe_pages.cnt = size; uio->uio_iov->iov_len -= size; uio->uio_iov->iov_base = (char *)uio->uio_iov->iov_base + size; if (uio->uio_iov->iov_len == 0) uio->uio_iov++; uio->uio_resid -= size; uio->uio_offset += size; return (0); } /* * Unwire the process buffer. */ static void pipe_destroy_write_buffer(struct pipe *wpipe) { PIPE_LOCK_ASSERT(wpipe, MA_OWNED); KASSERT((wpipe->pipe_state & PIPE_DIRECTW) != 0, ("%s: PIPE_DIRECTW not set on %p", __func__, wpipe)); KASSERT(wpipe->pipe_pages.cnt == 0, ("%s: pipe map for %p contains residual data", __func__, wpipe)); wpipe->pipe_state &= ~PIPE_DIRECTW; vm_page_unhold_pages(wpipe->pipe_pages.ms, wpipe->pipe_pages.npages); wpipe->pipe_pages.npages = 0; } /* * In the case of a signal, the writing process might go away. This * code copies the data into the circular buffer so that the source * pages can be freed without loss of data. */ static void pipe_clone_write_buffer(struct pipe *wpipe) { struct uio uio; struct iovec iov; int size; int pos; PIPE_LOCK_ASSERT(wpipe, MA_OWNED); KASSERT((wpipe->pipe_state & PIPE_DIRECTW) != 0, ("%s: PIPE_DIRECTW not set on %p", __func__, wpipe)); size = wpipe->pipe_pages.cnt; pos = wpipe->pipe_pages.pos; wpipe->pipe_pages.cnt = 0; wpipe->pipe_buffer.in = size; wpipe->pipe_buffer.out = 0; wpipe->pipe_buffer.cnt = size; PIPE_UNLOCK(wpipe); iov.iov_base = wpipe->pipe_buffer.buffer; iov.iov_len = size; uio.uio_iov = &iov; uio.uio_iovcnt = 1; uio.uio_offset = 0; uio.uio_resid = size; uio.uio_segflg = UIO_SYSSPACE; uio.uio_rw = UIO_READ; uio.uio_td = curthread; uiomove_fromphys(wpipe->pipe_pages.ms, pos, size, &uio); PIPE_LOCK(wpipe); pipe_destroy_write_buffer(wpipe); } /* * This implements the pipe buffer write mechanism. Note that only * a direct write OR a normal pipe write can be pending at any given time. * If there are any characters in the pipe buffer, the direct write will * be deferred until the receiving process grabs all of the bytes from * the pipe buffer. Then the direct mapping write is set-up. */ static int pipe_direct_write(struct pipe *wpipe, struct uio *uio) { int error; retry: PIPE_LOCK_ASSERT(wpipe, MA_OWNED); if ((wpipe->pipe_state & PIPE_EOF) != 0) { error = EPIPE; goto error1; } if (wpipe->pipe_state & PIPE_DIRECTW) { if (wpipe->pipe_state & PIPE_WANTR) { wpipe->pipe_state &= ~PIPE_WANTR; wakeup(wpipe); } pipeselwakeup(wpipe); wpipe->pipe_state |= PIPE_WANTW; pipeunlock(wpipe); error = msleep(wpipe, PIPE_MTX(wpipe), PRIBIO | PCATCH, "pipdww", 0); pipelock(wpipe, 0); if (error != 0) goto error1; goto retry; } if (wpipe->pipe_buffer.cnt > 0) { if (wpipe->pipe_state & PIPE_WANTR) { wpipe->pipe_state &= ~PIPE_WANTR; wakeup(wpipe); } pipeselwakeup(wpipe); wpipe->pipe_state |= PIPE_WANTW; pipeunlock(wpipe); error = msleep(wpipe, PIPE_MTX(wpipe), PRIBIO | PCATCH, "pipdwc", 0); pipelock(wpipe, 0); if (error != 0) goto error1; goto retry; } error = pipe_build_write_buffer(wpipe, uio); if (error) { goto error1; } while (wpipe->pipe_pages.cnt != 0 && (wpipe->pipe_state & PIPE_EOF) == 0) { if (wpipe->pipe_state & PIPE_WANTR) { wpipe->pipe_state &= ~PIPE_WANTR; wakeup(wpipe); } pipeselwakeup(wpipe); wpipe->pipe_state |= PIPE_WANTW; pipeunlock(wpipe); error = msleep(wpipe, PIPE_MTX(wpipe), PRIBIO | PCATCH, "pipdwt", 0); pipelock(wpipe, 0); if (error != 0) break; } if ((wpipe->pipe_state & PIPE_EOF) != 0) { wpipe->pipe_pages.cnt = 0; pipe_destroy_write_buffer(wpipe); pipeselwakeup(wpipe); error = EPIPE; } else if (error == EINTR || error == ERESTART) { pipe_clone_write_buffer(wpipe); } else { pipe_destroy_write_buffer(wpipe); } KASSERT((wpipe->pipe_state & PIPE_DIRECTW) == 0, ("pipe %p leaked PIPE_DIRECTW", wpipe)); return (error); error1: wakeup(wpipe); return (error); } #endif static int pipe_write(struct file *fp, struct uio *uio, struct ucred *active_cred, int flags, struct thread *td) { struct pipe *wpipe, *rpipe; ssize_t orig_resid; int desiredsize, error; rpipe = fp->f_data; wpipe = PIPE_PEER(rpipe); PIPE_LOCK(rpipe); error = pipelock(wpipe, 1); if (error) { PIPE_UNLOCK(rpipe); return (error); } /* * detect loss of pipe read side, issue SIGPIPE if lost. */ if (wpipe->pipe_present != PIPE_ACTIVE || (wpipe->pipe_state & PIPE_EOF)) { pipeunlock(wpipe); PIPE_UNLOCK(rpipe); return (EPIPE); } #ifdef MAC error = mac_pipe_check_write(active_cred, wpipe->pipe_pair); if (error) { pipeunlock(wpipe); PIPE_UNLOCK(rpipe); return (error); } #endif ++wpipe->pipe_busy; /* Choose a larger size if it's advantageous */ desiredsize = max(SMALL_PIPE_SIZE, wpipe->pipe_buffer.size); while (desiredsize < wpipe->pipe_buffer.cnt + uio->uio_resid) { if (piperesizeallowed != 1) break; if (amountpipekva > maxpipekva / 2) break; if (desiredsize == BIG_PIPE_SIZE) break; desiredsize = desiredsize * 2; } /* Choose a smaller size if we're in a OOM situation */ if (amountpipekva > (3 * maxpipekva) / 4 && wpipe->pipe_buffer.size > SMALL_PIPE_SIZE && wpipe->pipe_buffer.cnt <= SMALL_PIPE_SIZE && piperesizeallowed == 1) desiredsize = SMALL_PIPE_SIZE; /* Resize if the above determined that a new size was necessary */ if (desiredsize != wpipe->pipe_buffer.size && (wpipe->pipe_state & PIPE_DIRECTW) == 0) { PIPE_UNLOCK(wpipe); pipespace(wpipe, desiredsize); PIPE_LOCK(wpipe); } MPASS(wpipe->pipe_buffer.size != 0); orig_resid = uio->uio_resid; while (uio->uio_resid) { int space; if (wpipe->pipe_state & PIPE_EOF) { error = EPIPE; break; } #ifndef PIPE_NODIRECT /* * If the transfer is large, we can gain performance if * we do process-to-process copies directly. * If the write is non-blocking, we don't use the * direct write mechanism. * * The direct write mechanism will detect the reader going * away on us. */ if (uio->uio_segflg == UIO_USERSPACE && uio->uio_iov->iov_len >= pipe_mindirect && wpipe->pipe_buffer.size >= pipe_mindirect && (fp->f_flag & FNONBLOCK) == 0) { error = pipe_direct_write(wpipe, uio); if (error != 0) break; continue; } #endif /* * Pipe buffered writes cannot be coincidental with * direct writes. We wait until the currently executing * direct write is completed before we start filling the * pipe buffer. We break out if a signal occurs or the * reader goes away. */ if (wpipe->pipe_pages.cnt != 0) { if (wpipe->pipe_state & PIPE_WANTR) { wpipe->pipe_state &= ~PIPE_WANTR; wakeup(wpipe); } pipeselwakeup(wpipe); wpipe->pipe_state |= PIPE_WANTW; pipeunlock(wpipe); error = msleep(wpipe, PIPE_MTX(rpipe), PRIBIO | PCATCH, "pipbww", 0); pipelock(wpipe, 0); if (error != 0) break; continue; } space = wpipe->pipe_buffer.size - wpipe->pipe_buffer.cnt; /* Writes of size <= PIPE_BUF must be atomic. */ if ((space < uio->uio_resid) && (orig_resid <= PIPE_BUF)) space = 0; if (space > 0) { int size; /* Transfer size */ int segsize; /* first segment to transfer */ /* * Transfer size is minimum of uio transfer * and free space in pipe buffer. */ if (space > uio->uio_resid) size = uio->uio_resid; else size = space; /* * First segment to transfer is minimum of * transfer size and contiguous space in * pipe buffer. If first segment to transfer * is less than the transfer size, we've got * a wraparound in the buffer. */ segsize = wpipe->pipe_buffer.size - wpipe->pipe_buffer.in; if (segsize > size) segsize = size; /* Transfer first segment */ PIPE_UNLOCK(rpipe); error = uiomove(&wpipe->pipe_buffer.buffer[wpipe->pipe_buffer.in], segsize, uio); PIPE_LOCK(rpipe); if (error == 0 && segsize < size) { KASSERT(wpipe->pipe_buffer.in + segsize == wpipe->pipe_buffer.size, ("Pipe buffer wraparound disappeared")); /* * Transfer remaining part now, to * support atomic writes. Wraparound * happened. */ PIPE_UNLOCK(rpipe); error = uiomove( &wpipe->pipe_buffer.buffer[0], size - segsize, uio); PIPE_LOCK(rpipe); } if (error == 0) { wpipe->pipe_buffer.in += size; if (wpipe->pipe_buffer.in >= wpipe->pipe_buffer.size) { KASSERT(wpipe->pipe_buffer.in == size - segsize + wpipe->pipe_buffer.size, ("Expected wraparound bad")); wpipe->pipe_buffer.in = size - segsize; } wpipe->pipe_buffer.cnt += size; KASSERT(wpipe->pipe_buffer.cnt <= wpipe->pipe_buffer.size, ("Pipe buffer overflow")); } if (error != 0) break; continue; } else { /* * If the "read-side" has been blocked, wake it up now. */ if (wpipe->pipe_state & PIPE_WANTR) { wpipe->pipe_state &= ~PIPE_WANTR; wakeup(wpipe); } /* * don't block on non-blocking I/O */ if (fp->f_flag & FNONBLOCK) { error = EAGAIN; break; } /* * We have no more space and have something to offer, * wake up select/poll. */ pipeselwakeup(wpipe); wpipe->pipe_state |= PIPE_WANTW; pipeunlock(wpipe); error = msleep(wpipe, PIPE_MTX(rpipe), PRIBIO | PCATCH, "pipewr", 0); pipelock(wpipe, 0); if (error != 0) break; continue; } } --wpipe->pipe_busy; if ((wpipe->pipe_busy == 0) && (wpipe->pipe_state & PIPE_WANT)) { wpipe->pipe_state &= ~(PIPE_WANT | PIPE_WANTR); wakeup(wpipe); } else if (wpipe->pipe_buffer.cnt > 0) { /* * If we have put any characters in the buffer, we wake up * the reader. */ if (wpipe->pipe_state & PIPE_WANTR) { wpipe->pipe_state &= ~PIPE_WANTR; wakeup(wpipe); } } /* * Don't return EPIPE if any byte was written. * EINTR and other interrupts are handled by generic I/O layer. * Do not pretend that I/O succeeded for obvious user error * like EFAULT. */ if (uio->uio_resid != orig_resid && error == EPIPE) error = 0; if (error == 0) pipe_timestamp(&wpipe->pipe_mtime); /* * We have something to offer, * wake up select/poll. */ if (wpipe->pipe_buffer.cnt) pipeselwakeup(wpipe); pipeunlock(wpipe); PIPE_UNLOCK(rpipe); if (uio->uio_resid != orig_resid) td->td_ru.ru_msgsnd++; return (error); } /* ARGSUSED */ static int pipe_truncate(struct file *fp, off_t length, struct ucred *active_cred, struct thread *td) { struct pipe *cpipe; int error; cpipe = fp->f_data; if (cpipe->pipe_type & PIPE_TYPE_NAMED) error = vnops.fo_truncate(fp, length, active_cred, td); else error = invfo_truncate(fp, length, active_cred, td); return (error); } /* * we implement a very minimal set of ioctls for compatibility with sockets. */ static int pipe_ioctl(struct file *fp, u_long cmd, void *data, struct ucred *active_cred, struct thread *td) { struct pipe *mpipe = fp->f_data; int error; PIPE_LOCK(mpipe); #ifdef MAC error = mac_pipe_check_ioctl(active_cred, mpipe->pipe_pair, cmd, data); if (error) { PIPE_UNLOCK(mpipe); return (error); } #endif error = 0; switch (cmd) { case FIONBIO: break; case FIOASYNC: if (*(int *)data) { mpipe->pipe_state |= PIPE_ASYNC; } else { mpipe->pipe_state &= ~PIPE_ASYNC; } break; case FIONREAD: if (!(fp->f_flag & FREAD)) { *(int *)data = 0; PIPE_UNLOCK(mpipe); return (0); } if (mpipe->pipe_pages.cnt != 0) *(int *)data = mpipe->pipe_pages.cnt; else *(int *)data = mpipe->pipe_buffer.cnt; break; case FIOSETOWN: PIPE_UNLOCK(mpipe); error = fsetown(*(int *)data, &mpipe->pipe_sigio); goto out_unlocked; case FIOGETOWN: *(int *)data = fgetown(&mpipe->pipe_sigio); break; /* This is deprecated, FIOSETOWN should be used instead. */ case TIOCSPGRP: PIPE_UNLOCK(mpipe); error = fsetown(-(*(int *)data), &mpipe->pipe_sigio); goto out_unlocked; /* This is deprecated, FIOGETOWN should be used instead. */ case TIOCGPGRP: *(int *)data = -fgetown(&mpipe->pipe_sigio); break; default: error = ENOTTY; break; } PIPE_UNLOCK(mpipe); out_unlocked: return (error); } static int pipe_poll(struct file *fp, int events, struct ucred *active_cred, struct thread *td) { struct pipe *rpipe; struct pipe *wpipe; int levents, revents; #ifdef MAC int error; #endif revents = 0; rpipe = fp->f_data; wpipe = PIPE_PEER(rpipe); PIPE_LOCK(rpipe); #ifdef MAC error = mac_pipe_check_poll(active_cred, rpipe->pipe_pair); if (error) goto locked_error; #endif if (fp->f_flag & FREAD && events & (POLLIN | POLLRDNORM)) if (rpipe->pipe_pages.cnt > 0 || rpipe->pipe_buffer.cnt > 0) revents |= events & (POLLIN | POLLRDNORM); if (fp->f_flag & FWRITE && events & (POLLOUT | POLLWRNORM)) if (wpipe->pipe_present != PIPE_ACTIVE || (wpipe->pipe_state & PIPE_EOF) || ((wpipe->pipe_state & PIPE_DIRECTW) == 0 && ((wpipe->pipe_buffer.size - wpipe->pipe_buffer.cnt) >= PIPE_BUF || wpipe->pipe_buffer.size == 0))) revents |= events & (POLLOUT | POLLWRNORM); levents = events & (POLLIN | POLLINIGNEOF | POLLPRI | POLLRDNORM | POLLRDBAND); if (rpipe->pipe_type & PIPE_TYPE_NAMED && fp->f_flag & FREAD && levents && fp->f_pipegen == rpipe->pipe_wgen) events |= POLLINIGNEOF; if ((events & POLLINIGNEOF) == 0) { if (rpipe->pipe_state & PIPE_EOF) { if (fp->f_flag & FREAD) revents |= (events & (POLLIN | POLLRDNORM)); if (wpipe->pipe_present != PIPE_ACTIVE || (wpipe->pipe_state & PIPE_EOF)) revents |= POLLHUP; } } if (revents == 0) { /* * Add ourselves regardless of eventmask as we have to return * POLLHUP even if it was not asked for. */ if ((fp->f_flag & FREAD) != 0) { selrecord(td, &rpipe->pipe_sel); if (SEL_WAITING(&rpipe->pipe_sel)) rpipe->pipe_state |= PIPE_SEL; } if ((fp->f_flag & FWRITE) != 0 && wpipe->pipe_present == PIPE_ACTIVE) { selrecord(td, &wpipe->pipe_sel); if (SEL_WAITING(&wpipe->pipe_sel)) wpipe->pipe_state |= PIPE_SEL; } } #ifdef MAC locked_error: #endif PIPE_UNLOCK(rpipe); return (revents); } /* * We shouldn't need locks here as we're doing a read and this should * be a natural race. */ static int pipe_stat(struct file *fp, struct stat *ub, struct ucred *active_cred) { struct pipe *pipe; #ifdef MAC int error; #endif pipe = fp->f_data; #ifdef MAC if (mac_pipe_check_stat_enabled()) { PIPE_LOCK(pipe); error = mac_pipe_check_stat(active_cred, pipe->pipe_pair); PIPE_UNLOCK(pipe); if (error) { return (error); } } #endif /* For named pipes ask the underlying filesystem. */ if (pipe->pipe_type & PIPE_TYPE_NAMED) { return (vnops.fo_stat(fp, ub, active_cred)); } bzero(ub, sizeof(*ub)); ub->st_mode = S_IFIFO; ub->st_blksize = PAGE_SIZE; if (pipe->pipe_pages.cnt != 0) ub->st_size = pipe->pipe_pages.cnt; else ub->st_size = pipe->pipe_buffer.cnt; ub->st_blocks = howmany(ub->st_size, ub->st_blksize); ub->st_atim = pipe->pipe_atime; ub->st_mtim = pipe->pipe_mtime; ub->st_ctim = pipe->pipe_ctime; ub->st_uid = fp->f_cred->cr_uid; ub->st_gid = fp->f_cred->cr_gid; ub->st_dev = pipedev_ino; ub->st_ino = pipe->pipe_ino; /* * Left as 0: st_nlink, st_rdev, st_flags, st_gen. */ return (0); } /* ARGSUSED */ static int pipe_close(struct file *fp, struct thread *td) { if (fp->f_vnode != NULL) return vnops.fo_close(fp, td); fp->f_ops = &badfileops; pipe_dtor(fp->f_data); fp->f_data = NULL; return (0); } static int pipe_chmod(struct file *fp, mode_t mode, struct ucred *active_cred, struct thread *td) { struct pipe *cpipe; int error; cpipe = fp->f_data; if (cpipe->pipe_type & PIPE_TYPE_NAMED) error = vn_chmod(fp, mode, active_cred, td); else error = invfo_chmod(fp, mode, active_cred, td); return (error); } static int pipe_chown(struct file *fp, uid_t uid, gid_t gid, struct ucred *active_cred, struct thread *td) { struct pipe *cpipe; int error; cpipe = fp->f_data; if (cpipe->pipe_type & PIPE_TYPE_NAMED) error = vn_chown(fp, uid, gid, active_cred, td); else error = invfo_chown(fp, uid, gid, active_cred, td); return (error); } static int pipe_fill_kinfo(struct file *fp, struct kinfo_file *kif, struct filedesc *fdp) { struct pipe *pi; if (fp->f_type == DTYPE_FIFO) return (vn_fill_kinfo(fp, kif, fdp)); kif->kf_type = KF_TYPE_PIPE; pi = fp->f_data; kif->kf_un.kf_pipe.kf_pipe_addr = (uintptr_t)pi; kif->kf_un.kf_pipe.kf_pipe_peer = (uintptr_t)pi->pipe_peer; kif->kf_un.kf_pipe.kf_pipe_buffer_cnt = pi->pipe_buffer.cnt; + kif->kf_un.kf_pipe.kf_pipe_buffer_in = pi->pipe_buffer.in; + kif->kf_un.kf_pipe.kf_pipe_buffer_out = pi->pipe_buffer.out; + kif->kf_un.kf_pipe.kf_pipe_buffer_size = pi->pipe_buffer.size; return (0); } static void pipe_free_kmem(struct pipe *cpipe) { KASSERT(!mtx_owned(PIPE_MTX(cpipe)), ("pipe_free_kmem: pipe mutex locked")); if (cpipe->pipe_buffer.buffer != NULL) { atomic_subtract_long(&amountpipekva, cpipe->pipe_buffer.size); vm_map_remove(pipe_map, (vm_offset_t)cpipe->pipe_buffer.buffer, (vm_offset_t)cpipe->pipe_buffer.buffer + cpipe->pipe_buffer.size); cpipe->pipe_buffer.buffer = NULL; } #ifndef PIPE_NODIRECT { cpipe->pipe_pages.cnt = 0; cpipe->pipe_pages.pos = 0; cpipe->pipe_pages.npages = 0; } #endif } /* * shutdown the pipe */ static void pipeclose(struct pipe *cpipe) { #ifdef MAC struct pipepair *pp; #endif struct pipe *ppipe; KASSERT(cpipe != NULL, ("pipeclose: cpipe == NULL")); PIPE_LOCK(cpipe); pipelock(cpipe, 0); #ifdef MAC pp = cpipe->pipe_pair; #endif /* * If the other side is blocked, wake it up saying that * we want to close it down. */ cpipe->pipe_state |= PIPE_EOF; while (cpipe->pipe_busy) { wakeup(cpipe); cpipe->pipe_state |= PIPE_WANT; pipeunlock(cpipe); msleep(cpipe, PIPE_MTX(cpipe), PRIBIO, "pipecl", 0); pipelock(cpipe, 0); } pipeselwakeup(cpipe); /* * Disconnect from peer, if any. */ ppipe = cpipe->pipe_peer; if (ppipe->pipe_present == PIPE_ACTIVE) { ppipe->pipe_state |= PIPE_EOF; wakeup(ppipe); pipeselwakeup(ppipe); } /* * Mark this endpoint as free. Release kmem resources. We * don't mark this endpoint as unused until we've finished * doing that, or the pipe might disappear out from under * us. */ PIPE_UNLOCK(cpipe); pipe_free_kmem(cpipe); PIPE_LOCK(cpipe); cpipe->pipe_present = PIPE_CLOSING; pipeunlock(cpipe); /* * knlist_clear() may sleep dropping the PIPE_MTX. Set the * PIPE_FINALIZED, that allows other end to free the * pipe_pair, only after the knotes are completely dismantled. */ knlist_clear(&cpipe->pipe_sel.si_note, 1); cpipe->pipe_present = PIPE_FINALIZED; seldrain(&cpipe->pipe_sel); knlist_destroy(&cpipe->pipe_sel.si_note); /* * If both endpoints are now closed, release the memory for the * pipe pair. If not, unlock. */ if (ppipe->pipe_present == PIPE_FINALIZED) { PIPE_UNLOCK(cpipe); #ifdef MAC mac_pipe_destroy(pp); #endif uma_zfree(pipe_zone, cpipe->pipe_pair); } else PIPE_UNLOCK(cpipe); } /*ARGSUSED*/ static int pipe_kqfilter(struct file *fp, struct knote *kn) { struct pipe *cpipe; /* * If a filter is requested that is not supported by this file * descriptor, don't return an error, but also don't ever generate an * event. */ if ((kn->kn_filter == EVFILT_READ) && !(fp->f_flag & FREAD)) { kn->kn_fop = &pipe_nfiltops; return (0); } if ((kn->kn_filter == EVFILT_WRITE) && !(fp->f_flag & FWRITE)) { kn->kn_fop = &pipe_nfiltops; return (0); } cpipe = fp->f_data; PIPE_LOCK(cpipe); switch (kn->kn_filter) { case EVFILT_READ: kn->kn_fop = &pipe_rfiltops; break; case EVFILT_WRITE: kn->kn_fop = &pipe_wfiltops; if (cpipe->pipe_peer->pipe_present != PIPE_ACTIVE) { /* other end of pipe has been closed */ PIPE_UNLOCK(cpipe); return (EPIPE); } cpipe = PIPE_PEER(cpipe); break; default: if ((cpipe->pipe_type & PIPE_TYPE_NAMED) != 0) { PIPE_UNLOCK(cpipe); return (vnops.fo_kqfilter(fp, kn)); } PIPE_UNLOCK(cpipe); return (EINVAL); } kn->kn_hook = cpipe; knlist_add(&cpipe->pipe_sel.si_note, kn, 1); PIPE_UNLOCK(cpipe); return (0); } static void filt_pipedetach(struct knote *kn) { struct pipe *cpipe = kn->kn_hook; PIPE_LOCK(cpipe); knlist_remove(&cpipe->pipe_sel.si_note, kn, 1); PIPE_UNLOCK(cpipe); } /*ARGSUSED*/ static int filt_piperead(struct knote *kn, long hint) { struct file *fp = kn->kn_fp; struct pipe *rpipe = kn->kn_hook; PIPE_LOCK_ASSERT(rpipe, MA_OWNED); kn->kn_data = rpipe->pipe_buffer.cnt; if (kn->kn_data == 0) kn->kn_data = rpipe->pipe_pages.cnt; if ((rpipe->pipe_state & PIPE_EOF) != 0 && ((rpipe->pipe_type & PIPE_TYPE_NAMED) == 0 || fp->f_pipegen != rpipe->pipe_wgen)) { kn->kn_flags |= EV_EOF; return (1); } kn->kn_flags &= ~EV_EOF; return (kn->kn_data > 0); } /*ARGSUSED*/ static int filt_pipewrite(struct knote *kn, long hint) { struct pipe *wpipe = kn->kn_hook; /* * If this end of the pipe is closed, the knote was removed from the * knlist and the list lock (i.e., the pipe lock) is therefore not held. */ if (wpipe->pipe_present == PIPE_ACTIVE || (wpipe->pipe_type & PIPE_TYPE_NAMED) != 0) { PIPE_LOCK_ASSERT(wpipe, MA_OWNED); if (wpipe->pipe_state & PIPE_DIRECTW) { kn->kn_data = 0; } else if (wpipe->pipe_buffer.size > 0) { kn->kn_data = wpipe->pipe_buffer.size - wpipe->pipe_buffer.cnt; } else { kn->kn_data = PIPE_BUF; } } if (wpipe->pipe_present != PIPE_ACTIVE || (wpipe->pipe_state & PIPE_EOF)) { kn->kn_flags |= EV_EOF; return (1); } kn->kn_flags &= ~EV_EOF; return (kn->kn_data >= PIPE_BUF); } static void filt_pipedetach_notsup(struct knote *kn) { } static int filt_pipenotsup(struct knote *kn, long hint) { return (0); } diff --git a/sys/kern/sys_socket.c b/sys/kern/sys_socket.c index 859c6d593da0..7f39e58f4ce4 100644 --- a/sys/kern/sys_socket.c +++ b/sys/kern/sys_socket.c @@ -1,849 +1,851 @@ /*- * SPDX-License-Identifier: BSD-3-Clause * * Copyright (c) 1982, 1986, 1990, 1993 * The Regents of the University of California. 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. * 3. Neither the name of the University nor the names of its contributors * may be used to endorse or promote products derived from this software * without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * * @(#)sys_socket.c 8.1 (Berkeley) 6/10/93 */ #include __FBSDID("$FreeBSD$"); #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include /* XXX */ #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include static SYSCTL_NODE(_kern_ipc, OID_AUTO, aio, CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, "socket AIO stats"); static int empty_results; SYSCTL_INT(_kern_ipc_aio, OID_AUTO, empty_results, CTLFLAG_RD, &empty_results, 0, "socket operation returned EAGAIN"); static int empty_retries; SYSCTL_INT(_kern_ipc_aio, OID_AUTO, empty_retries, CTLFLAG_RD, &empty_retries, 0, "socket operation retries"); static fo_rdwr_t soo_read; static fo_rdwr_t soo_write; static fo_ioctl_t soo_ioctl; static fo_poll_t soo_poll; extern fo_kqfilter_t soo_kqfilter; static fo_stat_t soo_stat; static fo_close_t soo_close; static fo_fill_kinfo_t soo_fill_kinfo; static fo_aio_queue_t soo_aio_queue; static void soo_aio_cancel(struct kaiocb *job); struct fileops socketops = { .fo_read = soo_read, .fo_write = soo_write, .fo_truncate = invfo_truncate, .fo_ioctl = soo_ioctl, .fo_poll = soo_poll, .fo_kqfilter = soo_kqfilter, .fo_stat = soo_stat, .fo_close = soo_close, .fo_chmod = invfo_chmod, .fo_chown = invfo_chown, .fo_sendfile = invfo_sendfile, .fo_fill_kinfo = soo_fill_kinfo, .fo_aio_queue = soo_aio_queue, .fo_flags = DFLAG_PASSABLE }; static int soo_read(struct file *fp, struct uio *uio, struct ucred *active_cred, int flags, struct thread *td) { struct socket *so = fp->f_data; int error; #ifdef MAC error = mac_socket_check_receive(active_cred, so); if (error) return (error); #endif error = soreceive(so, 0, uio, 0, 0, 0); return (error); } static int soo_write(struct file *fp, struct uio *uio, struct ucred *active_cred, int flags, struct thread *td) { struct socket *so = fp->f_data; int error; #ifdef MAC error = mac_socket_check_send(active_cred, so); if (error) return (error); #endif error = sosend(so, 0, uio, 0, 0, 0, uio->uio_td); if (error == EPIPE && (so->so_options & SO_NOSIGPIPE) == 0) { PROC_LOCK(uio->uio_td->td_proc); tdsignal(uio->uio_td, SIGPIPE); PROC_UNLOCK(uio->uio_td->td_proc); } return (error); } static int soo_ioctl(struct file *fp, u_long cmd, void *data, struct ucred *active_cred, struct thread *td) { struct socket *so = fp->f_data; int error = 0; switch (cmd) { case FIONBIO: SOCK_LOCK(so); if (*(int *)data) so->so_state |= SS_NBIO; else so->so_state &= ~SS_NBIO; SOCK_UNLOCK(so); break; case FIOASYNC: if (*(int *)data) { SOCK_LOCK(so); so->so_state |= SS_ASYNC; if (SOLISTENING(so)) { so->sol_sbrcv_flags |= SB_ASYNC; so->sol_sbsnd_flags |= SB_ASYNC; } else { SOCK_RECVBUF_LOCK(so); so->so_rcv.sb_flags |= SB_ASYNC; SOCK_RECVBUF_UNLOCK(so); SOCK_SENDBUF_LOCK(so); so->so_snd.sb_flags |= SB_ASYNC; SOCK_SENDBUF_UNLOCK(so); } SOCK_UNLOCK(so); } else { SOCK_LOCK(so); so->so_state &= ~SS_ASYNC; if (SOLISTENING(so)) { so->sol_sbrcv_flags &= ~SB_ASYNC; so->sol_sbsnd_flags &= ~SB_ASYNC; } else { SOCK_RECVBUF_LOCK(so); so->so_rcv.sb_flags &= ~SB_ASYNC; SOCK_RECVBUF_UNLOCK(so); SOCK_SENDBUF_LOCK(so); so->so_snd.sb_flags &= ~SB_ASYNC; SOCK_SENDBUF_UNLOCK(so); } SOCK_UNLOCK(so); } break; case FIONREAD: SOCK_RECVBUF_LOCK(so); if (SOLISTENING(so)) { error = EINVAL; } else { *(int *)data = sbavail(&so->so_rcv) - so->so_rcv.sb_ctl; } SOCK_RECVBUF_UNLOCK(so); break; case FIONWRITE: /* Unlocked read. */ if (SOLISTENING(so)) { error = EINVAL; } else { *(int *)data = sbavail(&so->so_snd); } break; case FIONSPACE: /* Unlocked read. */ if (SOLISTENING(so)) { error = EINVAL; } else { if ((so->so_snd.sb_hiwat < sbused(&so->so_snd)) || (so->so_snd.sb_mbmax < so->so_snd.sb_mbcnt)) { *(int *)data = 0; } else { *(int *)data = sbspace(&so->so_snd); } } break; case FIOSETOWN: error = fsetown(*(int *)data, &so->so_sigio); break; case FIOGETOWN: *(int *)data = fgetown(&so->so_sigio); break; case SIOCSPGRP: error = fsetown(-(*(int *)data), &so->so_sigio); break; case SIOCGPGRP: *(int *)data = -fgetown(&so->so_sigio); break; case SIOCATMARK: /* Unlocked read. */ if (SOLISTENING(so)) { error = EINVAL; } else { *(int *)data = (so->so_rcv.sb_state & SBS_RCVATMARK) != 0; } break; default: /* * Interface/routing/protocol specific ioctls: interface and * routing ioctls should have a different entry since a * socket is unnecessary. */ if (IOCGROUP(cmd) == 'i') error = ifioctl(so, cmd, data, td); else if (IOCGROUP(cmd) == 'r') { CURVNET_SET(so->so_vnet); error = rtioctl_fib(cmd, data, so->so_fibnum); CURVNET_RESTORE(); } else { CURVNET_SET(so->so_vnet); error = ((*so->so_proto->pr_usrreqs->pru_control) (so, cmd, data, 0, td)); CURVNET_RESTORE(); } break; } return (error); } static int soo_poll(struct file *fp, int events, struct ucred *active_cred, struct thread *td) { struct socket *so = fp->f_data; #ifdef MAC int error; error = mac_socket_check_poll(active_cred, so); if (error) return (error); #endif return (sopoll(so, events, fp->f_cred, td)); } static int soo_stat(struct file *fp, struct stat *ub, struct ucred *active_cred) { struct socket *so = fp->f_data; int error; bzero((caddr_t)ub, sizeof (*ub)); ub->st_mode = S_IFSOCK; #ifdef MAC error = mac_socket_check_stat(active_cred, so); if (error) return (error); #endif SOCK_LOCK(so); if (!SOLISTENING(so)) { struct sockbuf *sb; /* * If SBS_CANTRCVMORE is set, but there's still data left * in the receive buffer, the socket is still readable. */ sb = &so->so_rcv; SOCK_RECVBUF_LOCK(so); if ((sb->sb_state & SBS_CANTRCVMORE) == 0 || sbavail(sb)) ub->st_mode |= S_IRUSR | S_IRGRP | S_IROTH; ub->st_size = sbavail(sb) - sb->sb_ctl; SOCK_RECVBUF_UNLOCK(so); sb = &so->so_snd; SOCK_SENDBUF_LOCK(so); if ((sb->sb_state & SBS_CANTSENDMORE) == 0) ub->st_mode |= S_IWUSR | S_IWGRP | S_IWOTH; SOCK_SENDBUF_UNLOCK(so); } ub->st_uid = so->so_cred->cr_uid; ub->st_gid = so->so_cred->cr_gid; error = so->so_proto->pr_usrreqs->pru_sense(so, ub); SOCK_UNLOCK(so); return (error); } /* * API socket close on file pointer. We call soclose() to close the socket * (including initiating closing protocols). soclose() will sorele() the * file reference but the actual socket will not go away until the socket's * ref count hits 0. */ static int soo_close(struct file *fp, struct thread *td) { int error = 0; struct socket *so; so = fp->f_data; fp->f_ops = &badfileops; fp->f_data = NULL; if (so) error = soclose(so); return (error); } static int soo_fill_kinfo(struct file *fp, struct kinfo_file *kif, struct filedesc *fdp) { struct sockaddr *sa; struct inpcb *inpcb; struct unpcb *unpcb; struct socket *so; int error; kif->kf_type = KF_TYPE_SOCKET; so = fp->f_data; CURVNET_SET(so->so_vnet); kif->kf_un.kf_sock.kf_sock_domain0 = so->so_proto->pr_domain->dom_family; kif->kf_un.kf_sock.kf_sock_type0 = so->so_type; kif->kf_un.kf_sock.kf_sock_protocol0 = so->so_proto->pr_protocol; kif->kf_un.kf_sock.kf_sock_pcb = (uintptr_t)so->so_pcb; switch (kif->kf_un.kf_sock.kf_sock_domain0) { case AF_INET: case AF_INET6: - if (kif->kf_un.kf_sock.kf_sock_protocol0 == IPPROTO_TCP) { - if (so->so_pcb != NULL) { - inpcb = (struct inpcb *)(so->so_pcb); - kif->kf_un.kf_sock.kf_sock_inpcb = - (uintptr_t)inpcb->inp_ppcb; - kif->kf_un.kf_sock.kf_sock_sendq = - sbused(&so->so_snd); - kif->kf_un.kf_sock.kf_sock_recvq = - sbused(&so->so_rcv); - } + if (so->so_pcb != NULL) { + inpcb = (struct inpcb *)(so->so_pcb); + kif->kf_un.kf_sock.kf_sock_inpcb = + (uintptr_t)inpcb->inp_ppcb; } + kif->kf_un.kf_sock.kf_sock_rcv_sb_state = + so->so_rcv.sb_state; + kif->kf_un.kf_sock.kf_sock_snd_sb_state = + so->so_snd.sb_state; + kif->kf_un.kf_sock.kf_sock_sendq = + sbused(&so->so_snd); + kif->kf_un.kf_sock.kf_sock_recvq = + sbused(&so->so_rcv); break; case AF_UNIX: if (so->so_pcb != NULL) { unpcb = (struct unpcb *)(so->so_pcb); if (unpcb->unp_conn) { kif->kf_un.kf_sock.kf_sock_unpconn = (uintptr_t)unpcb->unp_conn; kif->kf_un.kf_sock.kf_sock_rcv_sb_state = so->so_rcv.sb_state; kif->kf_un.kf_sock.kf_sock_snd_sb_state = so->so_snd.sb_state; kif->kf_un.kf_sock.kf_sock_sendq = sbused(&so->so_snd); kif->kf_un.kf_sock.kf_sock_recvq = sbused(&so->so_rcv); } } break; } error = so->so_proto->pr_usrreqs->pru_sockaddr(so, &sa); if (error == 0 && sa->sa_len <= sizeof(kif->kf_un.kf_sock.kf_sa_local)) { bcopy(sa, &kif->kf_un.kf_sock.kf_sa_local, sa->sa_len); free(sa, M_SONAME); } error = so->so_proto->pr_usrreqs->pru_peeraddr(so, &sa); if (error == 0 && sa->sa_len <= sizeof(kif->kf_un.kf_sock.kf_sa_peer)) { bcopy(sa, &kif->kf_un.kf_sock.kf_sa_peer, sa->sa_len); free(sa, M_SONAME); } strncpy(kif->kf_path, so->so_proto->pr_domain->dom_name, sizeof(kif->kf_path)); CURVNET_RESTORE(); return (0); } /* * Use the 'backend3' field in AIO jobs to store the amount of data * completed by the AIO job so far. */ #define aio_done backend3 static STAILQ_HEAD(, task) soaio_jobs; static struct mtx soaio_jobs_lock; static struct task soaio_kproc_task; static int soaio_starting, soaio_idle, soaio_queued; static struct unrhdr *soaio_kproc_unr; static int soaio_max_procs = MAX_AIO_PROCS; SYSCTL_INT(_kern_ipc_aio, OID_AUTO, max_procs, CTLFLAG_RW, &soaio_max_procs, 0, "Maximum number of kernel processes to use for async socket IO"); static int soaio_num_procs; SYSCTL_INT(_kern_ipc_aio, OID_AUTO, num_procs, CTLFLAG_RD, &soaio_num_procs, 0, "Number of active kernel processes for async socket IO"); static int soaio_target_procs = TARGET_AIO_PROCS; SYSCTL_INT(_kern_ipc_aio, OID_AUTO, target_procs, CTLFLAG_RD, &soaio_target_procs, 0, "Preferred number of ready kernel processes for async socket IO"); static int soaio_lifetime; SYSCTL_INT(_kern_ipc_aio, OID_AUTO, lifetime, CTLFLAG_RW, &soaio_lifetime, 0, "Maximum lifetime for idle aiod"); static void soaio_kproc_loop(void *arg) { struct proc *p; struct vmspace *myvm; struct task *task; int error, id, pending; id = (intptr_t)arg; /* * Grab an extra reference on the daemon's vmspace so that it * doesn't get freed by jobs that switch to a different * vmspace. */ p = curproc; myvm = vmspace_acquire_ref(p); mtx_lock(&soaio_jobs_lock); MPASS(soaio_starting > 0); soaio_starting--; for (;;) { while (!STAILQ_EMPTY(&soaio_jobs)) { task = STAILQ_FIRST(&soaio_jobs); STAILQ_REMOVE_HEAD(&soaio_jobs, ta_link); soaio_queued--; pending = task->ta_pending; task->ta_pending = 0; mtx_unlock(&soaio_jobs_lock); task->ta_func(task->ta_context, pending); mtx_lock(&soaio_jobs_lock); } MPASS(soaio_queued == 0); if (p->p_vmspace != myvm) { mtx_unlock(&soaio_jobs_lock); vmspace_switch_aio(myvm); mtx_lock(&soaio_jobs_lock); continue; } soaio_idle++; error = mtx_sleep(&soaio_idle, &soaio_jobs_lock, 0, "-", soaio_lifetime); soaio_idle--; if (error == EWOULDBLOCK && STAILQ_EMPTY(&soaio_jobs) && soaio_num_procs > soaio_target_procs) break; } soaio_num_procs--; mtx_unlock(&soaio_jobs_lock); free_unr(soaio_kproc_unr, id); kproc_exit(0); } static void soaio_kproc_create(void *context, int pending) { struct proc *p; int error, id; mtx_lock(&soaio_jobs_lock); for (;;) { if (soaio_num_procs < soaio_target_procs) { /* Must create */ } else if (soaio_num_procs >= soaio_max_procs) { /* * Hit the limit on kernel processes, don't * create another one. */ break; } else if (soaio_queued <= soaio_idle + soaio_starting) { /* * No more AIO jobs waiting for a process to be * created, so stop. */ break; } soaio_starting++; mtx_unlock(&soaio_jobs_lock); id = alloc_unr(soaio_kproc_unr); error = kproc_create(soaio_kproc_loop, (void *)(intptr_t)id, &p, 0, 0, "soaiod%d", id); if (error != 0) { free_unr(soaio_kproc_unr, id); mtx_lock(&soaio_jobs_lock); soaio_starting--; break; } mtx_lock(&soaio_jobs_lock); soaio_num_procs++; } mtx_unlock(&soaio_jobs_lock); } void soaio_enqueue(struct task *task) { mtx_lock(&soaio_jobs_lock); MPASS(task->ta_pending == 0); task->ta_pending++; STAILQ_INSERT_TAIL(&soaio_jobs, task, ta_link); soaio_queued++; if (soaio_queued <= soaio_idle) wakeup_one(&soaio_idle); else if (soaio_num_procs < soaio_max_procs) taskqueue_enqueue(taskqueue_thread, &soaio_kproc_task); mtx_unlock(&soaio_jobs_lock); } static void soaio_init(void) { soaio_lifetime = AIOD_LIFETIME_DEFAULT; STAILQ_INIT(&soaio_jobs); mtx_init(&soaio_jobs_lock, "soaio jobs", NULL, MTX_DEF); soaio_kproc_unr = new_unrhdr(1, INT_MAX, NULL); TASK_INIT(&soaio_kproc_task, 0, soaio_kproc_create, NULL); } SYSINIT(soaio, SI_SUB_VFS, SI_ORDER_ANY, soaio_init, NULL); static __inline int soaio_ready(struct socket *so, struct sockbuf *sb) { return (sb == &so->so_rcv ? soreadable(so) : sowriteable(so)); } static void soaio_process_job(struct socket *so, sb_which which, struct kaiocb *job) { struct ucred *td_savedcred; struct thread *td; struct sockbuf *sb = sobuf(so, which); #ifdef MAC struct file *fp = job->fd_file; #endif size_t cnt, done, job_total_nbytes __diagused; long ru_before; int error, flags; SOCK_BUF_UNLOCK(so, which); aio_switch_vmspace(job); td = curthread; retry: td_savedcred = td->td_ucred; td->td_ucred = job->cred; job_total_nbytes = job->uiop->uio_resid + job->aio_done; done = job->aio_done; cnt = job->uiop->uio_resid; job->uiop->uio_offset = 0; job->uiop->uio_td = td; flags = MSG_NBIO; /* * For resource usage accounting, only count a completed request * as a single message to avoid counting multiple calls to * sosend/soreceive on a blocking socket. */ if (sb == &so->so_rcv) { ru_before = td->td_ru.ru_msgrcv; #ifdef MAC error = mac_socket_check_receive(fp->f_cred, so); if (error == 0) #endif error = soreceive(so, NULL, job->uiop, NULL, NULL, &flags); if (td->td_ru.ru_msgrcv != ru_before) job->msgrcv = 1; } else { if (!TAILQ_EMPTY(&sb->sb_aiojobq)) flags |= MSG_MORETOCOME; ru_before = td->td_ru.ru_msgsnd; #ifdef MAC error = mac_socket_check_send(fp->f_cred, so); if (error == 0) #endif error = sosend(so, NULL, job->uiop, NULL, NULL, flags, td); if (td->td_ru.ru_msgsnd != ru_before) job->msgsnd = 1; if (error == EPIPE && (so->so_options & SO_NOSIGPIPE) == 0) { PROC_LOCK(job->userproc); kern_psignal(job->userproc, SIGPIPE); PROC_UNLOCK(job->userproc); } } done += cnt - job->uiop->uio_resid; job->aio_done = done; td->td_ucred = td_savedcred; if (error == EWOULDBLOCK) { /* * The request was either partially completed or not * completed at all due to racing with a read() or * write() on the socket. If the socket is * non-blocking, return with any partial completion. * If the socket is blocking or if no progress has * been made, requeue this request at the head of the * queue to try again when the socket is ready. */ MPASS(done != job_total_nbytes); SOCK_BUF_LOCK(so, which); if (done == 0 || !(so->so_state & SS_NBIO)) { empty_results++; if (soaio_ready(so, sb)) { empty_retries++; SOCK_BUF_UNLOCK(so, which); goto retry; } if (!aio_set_cancel_function(job, soo_aio_cancel)) { SOCK_BUF_UNLOCK(so, which); if (done != 0) aio_complete(job, done, 0); else aio_cancel(job); SOCK_BUF_LOCK(so, which); } else { TAILQ_INSERT_HEAD(&sb->sb_aiojobq, job, list); } return; } SOCK_BUF_UNLOCK(so, which); } if (done != 0 && (error == ERESTART || error == EINTR || error == EWOULDBLOCK)) error = 0; if (error) aio_complete(job, -1, error); else aio_complete(job, done, 0); SOCK_BUF_LOCK(so, which); } static void soaio_process_sb(struct socket *so, sb_which which) { struct kaiocb *job; struct sockbuf *sb = sobuf(so, which); CURVNET_SET(so->so_vnet); SOCK_BUF_LOCK(so, which); while (!TAILQ_EMPTY(&sb->sb_aiojobq) && soaio_ready(so, sb)) { job = TAILQ_FIRST(&sb->sb_aiojobq); TAILQ_REMOVE(&sb->sb_aiojobq, job, list); if (!aio_clear_cancel_function(job)) continue; soaio_process_job(so, which, job); } /* * If there are still pending requests, the socket must not be * ready so set SB_AIO to request a wakeup when the socket * becomes ready. */ if (!TAILQ_EMPTY(&sb->sb_aiojobq)) sb->sb_flags |= SB_AIO; sb->sb_flags &= ~SB_AIO_RUNNING; SOCK_BUF_UNLOCK(so, which); sorele(so); CURVNET_RESTORE(); } void soaio_rcv(void *context, int pending) { struct socket *so; so = context; soaio_process_sb(so, SO_RCV); } void soaio_snd(void *context, int pending) { struct socket *so; so = context; soaio_process_sb(so, SO_SND); } void sowakeup_aio(struct socket *so, sb_which which) { struct sockbuf *sb = sobuf(so, which); SOCK_BUF_LOCK_ASSERT(so, which); sb->sb_flags &= ~SB_AIO; if (sb->sb_flags & SB_AIO_RUNNING) return; sb->sb_flags |= SB_AIO_RUNNING; soref(so); soaio_enqueue(&sb->sb_aiotask); } static void soo_aio_cancel(struct kaiocb *job) { struct socket *so; struct sockbuf *sb; long done; int opcode; sb_which which; so = job->fd_file->f_data; opcode = job->uaiocb.aio_lio_opcode; if (opcode & LIO_READ) { sb = &so->so_rcv; which = SO_RCV; } else { MPASS(opcode & LIO_WRITE); sb = &so->so_snd; which = SO_SND; } SOCK_BUF_LOCK(so, which); if (!aio_cancel_cleared(job)) TAILQ_REMOVE(&sb->sb_aiojobq, job, list); if (TAILQ_EMPTY(&sb->sb_aiojobq)) sb->sb_flags &= ~SB_AIO; SOCK_BUF_UNLOCK(so, which); done = job->aio_done; if (done != 0) aio_complete(job, done, 0); else aio_cancel(job); } static int soo_aio_queue(struct file *fp, struct kaiocb *job) { struct socket *so; struct sockbuf *sb; sb_which which; int error; so = fp->f_data; error = (*so->so_proto->pr_usrreqs->pru_aio_queue)(so, job); if (error == 0) return (0); /* Lock through the socket, since this may be a listening socket. */ switch (job->uaiocb.aio_lio_opcode & (LIO_WRITE | LIO_READ)) { case LIO_READ: SOCK_RECVBUF_LOCK(so); sb = &so->so_rcv; which = SO_RCV; break; case LIO_WRITE: SOCK_SENDBUF_LOCK(so); sb = &so->so_snd; which = SO_SND; break; default: return (EINVAL); } if (SOLISTENING(so)) { SOCK_BUF_UNLOCK(so, which); return (EINVAL); } if (!aio_set_cancel_function(job, soo_aio_cancel)) panic("new job was cancelled"); TAILQ_INSERT_TAIL(&sb->sb_aiojobq, job, list); if (!(sb->sb_flags & SB_AIO_RUNNING)) { if (soaio_ready(so, sb)) sowakeup_aio(so, which); else sb->sb_flags |= SB_AIO; } SOCK_BUF_UNLOCK(so, which); return (0); } diff --git a/sys/sys/param.h b/sys/sys/param.h index 963f7f89c930..d8110f08a385 100644 --- a/sys/sys/param.h +++ b/sys/sys/param.h @@ -1,392 +1,392 @@ /*- * SPDX-License-Identifier: BSD-3-Clause * * Copyright (c) 1982, 1986, 1989, 1993 * The Regents of the University of California. All rights reserved. * (c) UNIX System Laboratories, Inc. * All or some portions of this file are derived from material licensed * to the University of California by American Telephone and Telegraph * Co. or Unix System Laboratories, Inc. and are reproduced herein with * the permission of UNIX System Laboratories, Inc. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * 3. Neither the name of the University nor the names of its contributors * may be used to endorse or promote products derived from this software * without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * * @(#)param.h 8.3 (Berkeley) 4/4/95 * $FreeBSD$ */ #ifndef _SYS_PARAM_H_ #define _SYS_PARAM_H_ #include #define BSD 199506 /* System version (year & month). */ #define BSD4_3 1 #define BSD4_4 1 /* * __FreeBSD_version numbers are documented in the Porter's Handbook. * If you bump the version for any reason, you should update the documentation * there. * Currently this lives here in the doc/ repository: * * documentation/content/en/books/porters-handbook/versions/_index.adoc * * Encoding: Rxx * 'R' is in the range 0 to 4 if this is a release branch or * X.0-CURRENT before releng/X.0 is created, otherwise 'R' is * in the range 5 to 9. * Short hand: MMmmXXX * * __FreeBSD_version is bumped every time there's a change in the base system * that's noteworthy. A noteworthy change is any change which changes the * kernel's KBI in -CURRENT, one that changes some detail about the system that * external software (or the ports system) would want to know about, one that * adds a system call, one that adds or deletes a shipped library, a security * fix, or similar change not specifically noted here. Bumps should be limited * to one per day / a couple per week except for security fixes. * * The approved way to obtain this from a shell script is: * awk '/^\#define[[:space:]]*__FreeBSD_version/ {print $3}' * Other methods to parse this file may work, but are not guaranteed against * future changes. The above script works back to FreeBSD 3.x when this macro * was introduced. This number is propagated to other places needing it that * cannot include sys/param.h and should only be updated here. */ #undef __FreeBSD_version -#define __FreeBSD_version 1400061 +#define __FreeBSD_version 1400062 /* * __FreeBSD_kernel__ indicates that this system uses the kernel of FreeBSD, * which by definition is always true on FreeBSD. This macro is also defined * on other systems that use the kernel of FreeBSD, such as GNU/kFreeBSD. * * It is tempting to use this macro in userland code when we want to enable * kernel-specific routines, and in fact it's fine to do this in code that * is part of FreeBSD itself. However, be aware that as presence of this * macro is still not widespread (e.g. older FreeBSD versions, 3rd party * compilers, etc), it is STRONGLY DISCOURAGED to check for this macro in * external applications without also checking for __FreeBSD__ as an * alternative. */ #undef __FreeBSD_kernel__ #define __FreeBSD_kernel__ #if defined(_KERNEL) || defined(IN_RTLD) #define P_OSREL_SIGWAIT 700000 #define P_OSREL_SIGSEGV 700004 #define P_OSREL_MAP_ANON 800104 #define P_OSREL_MAP_FSTRICT 1100036 #define P_OSREL_SHUTDOWN_ENOTCONN 1100077 #define P_OSREL_MAP_GUARD 1200035 #define P_OSREL_WRFSBASE 1200041 #define P_OSREL_CK_CYLGRP 1200046 #define P_OSREL_VMTOTAL64 1200054 #define P_OSREL_CK_SUPERBLOCK 1300000 #define P_OSREL_CK_INODE 1300005 #define P_OSREL_POWERPC_NEW_AUX_ARGS 1300070 #define P_OSREL_MAJOR(x) ((x) / 100000) #endif #ifndef LOCORE #include #endif /* * Machine-independent constants (some used in following include files). * Redefined constants are from POSIX 1003.1 limits file. * * MAXCOMLEN should be >= sizeof(ac_comm) (see ) */ #include #define MAXCOMLEN 19 /* max command name remembered */ #define MAXINTERP PATH_MAX /* max interpreter file name length */ #define MAXLOGNAME 33 /* max login name length (incl. NUL) */ #define MAXUPRC CHILD_MAX /* max simultaneous processes */ #define NCARGS ARG_MAX /* max bytes for an exec function */ #define NGROUPS (NGROUPS_MAX+1) /* max number groups */ #define NOFILE OPEN_MAX /* max open files per process */ #define NOGROUP 65535 /* marker for empty group set member */ #define MAXHOSTNAMELEN 256 /* max hostname size */ #define SPECNAMELEN 255 /* max length of devicename */ /* More types and definitions used throughout the kernel. */ #ifdef _KERNEL #include #include #ifndef LOCORE #include #include #endif #ifndef FALSE #define FALSE 0 #endif #ifndef TRUE #define TRUE 1 #endif #endif #ifndef _KERNEL #ifndef LOCORE /* Signals. */ #include #endif #endif /* Machine type dependent parameters. */ #include #ifndef _KERNEL #include #endif #ifndef DEV_BSHIFT #define DEV_BSHIFT 9 /* log2(DEV_BSIZE) */ #endif #define DEV_BSIZE (1<>PAGE_SHIFT) #endif /* * btodb() is messy and perhaps slow because `bytes' may be an off_t. We * want to shift an unsigned type to avoid sign extension and we don't * want to widen `bytes' unnecessarily. Assume that the result fits in * a daddr_t. */ #ifndef btodb #define btodb(bytes) /* calculates (bytes / DEV_BSIZE) */ \ (sizeof (bytes) > sizeof(long) \ ? (daddr_t)((unsigned long long)(bytes) >> DEV_BSHIFT) \ : (daddr_t)((unsigned long)(bytes) >> DEV_BSHIFT)) #endif #ifndef dbtob #define dbtob(db) /* calculates (db * DEV_BSIZE) */ \ ((off_t)(db) << DEV_BSHIFT) #endif #define PRIMASK 0x0ff #define PCATCH 0x100 /* OR'd with pri for tsleep to check signals */ #define PDROP 0x200 /* OR'd with pri to stop re-entry of interlock mutex */ #define PNOLOCK 0x400 /* OR'd with pri to allow sleeping w/o a lock */ #define PRILASTFLAG 0x400 /* Last flag defined above */ #define NZERO 0 /* default "nice" */ #define NBBY 8 /* number of bits in a byte */ #define NBPW sizeof(int) /* number of bytes per word (integer) */ #define CMASK 022 /* default file mask: S_IWGRP|S_IWOTH */ #define NODEV (dev_t)(-1) /* non-existent device */ /* * File system parameters and macros. * * MAXBSIZE - Filesystems are made out of blocks of at most MAXBSIZE bytes * per block. MAXBSIZE may be made larger without effecting * any existing filesystems as long as it does not exceed MAXPHYS, * and may be made smaller at the risk of not being able to use * filesystems which require a block size exceeding MAXBSIZE. * * MAXBCACHEBUF - Maximum size of a buffer in the buffer cache. This must * be >= MAXBSIZE and can be set differently for different * architectures by defining it in . * Making this larger allows NFS to do larger reads/writes. * * BKVASIZE - Nominal buffer space per buffer, in bytes. BKVASIZE is the * minimum KVM memory reservation the kernel is willing to make. * Filesystems can of course request smaller chunks. Actual * backing memory uses a chunk size of a page (PAGE_SIZE). * The default value here can be overridden on a per-architecture * basis by defining it in . * * If you make BKVASIZE too small you risk seriously fragmenting * the buffer KVM map which may slow things down a bit. If you * make it too big the kernel will not be able to optimally use * the KVM memory reserved for the buffer cache and will wind * up with too-few buffers. * * The default is 16384, roughly 2x the block size used by a * normal UFS filesystem. */ #define MAXBSIZE 65536 /* must be power of 2 */ #ifndef MAXBCACHEBUF #define MAXBCACHEBUF MAXBSIZE /* must be a power of 2 >= MAXBSIZE */ #endif #ifndef BKVASIZE #define BKVASIZE 16384 /* must be power of 2 */ #endif #define BKVAMASK (BKVASIZE-1) /* * MAXPATHLEN defines the longest permissible path length after expanding * symbolic links. It is used to allocate a temporary buffer from the buffer * pool in which to do the name expansion, hence should be a power of two, * and must be less than or equal to MAXBSIZE. MAXSYMLINKS defines the * maximum number of symbolic links that may be expanded in a path name. * It should be set high enough to allow all legitimate uses, but halt * infinite loops reasonably quickly. */ #define MAXPATHLEN PATH_MAX #define MAXSYMLINKS 32 /* Bit map related macros. */ #define setbit(a,i) (((unsigned char *)(a))[(i)/NBBY] |= 1<<((i)%NBBY)) #define clrbit(a,i) (((unsigned char *)(a))[(i)/NBBY] &= ~(1<<((i)%NBBY))) #define isset(a,i) \ (((const unsigned char *)(a))[(i)/NBBY] & (1<<((i)%NBBY))) #define isclr(a,i) \ ((((const unsigned char *)(a))[(i)/NBBY] & (1<<((i)%NBBY))) == 0) /* Macros for counting and rounding. */ #ifndef howmany #define howmany(x, y) (((x)+((y)-1))/(y)) #endif #define nitems(x) (sizeof((x)) / sizeof((x)[0])) #define rounddown(x, y) (((x)/(y))*(y)) #define rounddown2(x, y) __align_down(x, y) /* if y is power of two */ #define roundup(x, y) ((((x)+((y)-1))/(y))*(y)) /* to any y */ #define roundup2(x, y) __align_up(x, y) /* if y is powers of two */ #define powerof2(x) ((((x)-1)&(x))==0) /* Macros for min/max. */ #define MIN(a,b) (((a)<(b))?(a):(b)) #define MAX(a,b) (((a)>(b))?(a):(b)) #ifdef _KERNEL /* * Basic byte order function prototypes for non-inline functions. */ #ifndef LOCORE #ifndef _BYTEORDER_PROTOTYPED #define _BYTEORDER_PROTOTYPED __BEGIN_DECLS __uint32_t htonl(__uint32_t); __uint16_t htons(__uint16_t); __uint32_t ntohl(__uint32_t); __uint16_t ntohs(__uint16_t); __END_DECLS #endif #endif #ifndef _BYTEORDER_FUNC_DEFINED #define _BYTEORDER_FUNC_DEFINED #define htonl(x) __htonl(x) #define htons(x) __htons(x) #define ntohl(x) __ntohl(x) #define ntohs(x) __ntohs(x) #endif /* !_BYTEORDER_FUNC_DEFINED */ #endif /* _KERNEL */ /* * Scale factor for scaled integers used to count %cpu time and load avgs. * * The number of CPU `tick's that map to a unique `%age' can be expressed * by the formula (1 / (2 ^ (FSHIFT - 11))). Since the intermediate * calculation is done with 64-bit precision, the maximum load average that can * be calculated is approximately 2^32 / FSCALE. * * For the scheduler to maintain a 1:1 mapping of CPU `tick' to `%age', * FSHIFT must be at least 11. This gives a maximum load avg of 2 million. */ #define FSHIFT 11 /* bits to right of fixed binary point */ #define FSCALE (1<> (PAGE_SHIFT - DEV_BSHIFT)) #define ctodb(db) /* calculates pages to devblks */ \ ((db) << (PAGE_SHIFT - DEV_BSHIFT)) /* * Old spelling of __containerof(). */ #define member2struct(s, m, x) \ ((struct s *)(void *)((char *)(x) - offsetof(struct s, m))) /* * Access a variable length array that has been declared as a fixed * length array. */ #define __PAST_END(array, offset) (((__typeof__(*(array)) *)(array))[offset]) #endif /* _SYS_PARAM_H_ */ diff --git a/sys/sys/user.h b/sys/sys/user.h index cf64f8e8eda4..d3257ffdf0f0 100644 --- a/sys/sys/user.h +++ b/sys/sys/user.h @@ -1,674 +1,682 @@ /*- * SPDX-License-Identifier: BSD-3-Clause * * Copyright (c) 1982, 1986, 1989, 1991, 1993 * The Regents of the University of California. * Copyright (c) 2007 Robert N. M. Watson * 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. * 3. Neither the name of the University nor the names of its contributors * may be used to endorse or promote products derived from this software * without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * * @(#)user.h 8.2 (Berkeley) 9/23/93 * $FreeBSD$ */ #ifndef _SYS_USER_H_ #define _SYS_USER_H_ #include #ifndef _KERNEL /* stuff that *used* to be included by user.h, or is now needed */ #include #include #include #include #include #include #include #include #include #include /* XXX */ #include /* XXX */ #include /* XXX */ #include /* XXX */ #endif /* !_KERNEL */ #ifndef _SYS_RESOURCEVAR_H_ #include #endif #ifndef _SYS_SIGNALVAR_H_ #include #endif #ifndef _SYS_SOCKET_VAR_H_ #include #endif #include /* * KERN_PROC subtype ops return arrays of selected proc structure entries: * * This struct includes several arrays of spare space, with different arrays * for different standard C-types. When adding new variables to this struct, * the space for byte-aligned data should be taken from the ki_sparestring, * pointers from ki_spareptrs, word-aligned data from ki_spareints, and * doubleword-aligned data from ki_sparelongs. Make sure the space for new * variables come from the array which matches the size and alignment of * those variables on ALL hardware platforms, and then adjust the appropriate * KI_NSPARE_* value(s) to match. * * Always verify that sizeof(struct kinfo_proc) == KINFO_PROC_SIZE on all * platforms after you have added new variables. Note that if you change * the value of KINFO_PROC_SIZE, then many userland programs will stop * working until they are recompiled! * * Once you have added the new field, you will need to add code to initialize * it in two places: function fill_kinfo_proc in sys/kern/kern_proc.c and * function kvm_proclist in lib/libkvm/kvm_proc.c . */ #define KI_NSPARE_INT 2 #define KI_NSPARE_LONG 12 #define KI_NSPARE_PTR 5 #ifndef _KERNEL #ifndef KINFO_PROC_SIZE #error "Unknown architecture" #endif #endif /* !_KERNEL */ #define WMESGLEN 8 /* size of returned wchan message */ #define LOCKNAMELEN 8 /* size of returned lock name */ #define TDNAMLEN 16 /* size of returned thread name */ #define COMMLEN 19 /* size of returned ki_comm name */ #define KI_EMULNAMELEN 16 /* size of returned ki_emul */ #define KI_NGROUPS 16 /* number of groups in ki_groups */ #define LOGNAMELEN 17 /* size of returned ki_login */ #define LOGINCLASSLEN 17 /* size of returned ki_loginclass */ #ifndef BURN_BRIDGES #define OCOMMLEN TDNAMLEN #define ki_ocomm ki_tdname #endif /* Flags for the process credential. */ #define KI_CRF_CAPABILITY_MODE 0x00000001 /* * Steal a bit from ki_cr_flags to indicate that the cred had more than * KI_NGROUPS groups. */ #define KI_CRF_GRP_OVERFLOW 0x80000000 struct kinfo_proc { int ki_structsize; /* size of this structure */ int ki_layout; /* reserved: layout identifier */ struct pargs *ki_args; /* address of command arguments */ struct proc *ki_paddr; /* address of proc */ struct user *ki_addr; /* kernel virtual addr of u-area */ struct vnode *ki_tracep; /* pointer to trace file */ struct vnode *ki_textvp; /* pointer to executable file */ struct filedesc *ki_fd; /* pointer to open file info */ struct vmspace *ki_vmspace; /* pointer to kernel vmspace struct */ const void *ki_wchan; /* sleep address */ pid_t ki_pid; /* Process identifier */ pid_t ki_ppid; /* parent process id */ pid_t ki_pgid; /* process group id */ pid_t ki_tpgid; /* tty process group id */ pid_t ki_sid; /* Process session ID */ pid_t ki_tsid; /* Terminal session ID */ short ki_jobc; /* job control counter */ short ki_spare_short1; /* unused (just here for alignment) */ uint32_t ki_tdev_freebsd11; /* controlling tty dev */ sigset_t ki_siglist; /* Signals arrived but not delivered */ sigset_t ki_sigmask; /* Current signal mask */ sigset_t ki_sigignore; /* Signals being ignored */ sigset_t ki_sigcatch; /* Signals being caught by user */ uid_t ki_uid; /* effective user id */ uid_t ki_ruid; /* Real user id */ uid_t ki_svuid; /* Saved effective user id */ gid_t ki_rgid; /* Real group id */ gid_t ki_svgid; /* Saved effective group id */ short ki_ngroups; /* number of groups */ short ki_spare_short2; /* unused (just here for alignment) */ gid_t ki_groups[KI_NGROUPS]; /* groups */ vm_size_t ki_size; /* virtual size */ segsz_t ki_rssize; /* current resident set size in pages */ segsz_t ki_swrss; /* resident set size before last swap */ segsz_t ki_tsize; /* text size (pages) XXX */ segsz_t ki_dsize; /* data size (pages) XXX */ segsz_t ki_ssize; /* stack size (pages) */ u_short ki_xstat; /* Exit status for wait & stop signal */ u_short ki_acflag; /* Accounting flags */ fixpt_t ki_pctcpu; /* %cpu for process during ki_swtime */ u_int ki_estcpu; /* Time averaged value of ki_cpticks */ u_int ki_slptime; /* Time since last blocked */ u_int ki_swtime; /* Time swapped in or out */ u_int ki_cow; /* number of copy-on-write faults */ u_int64_t ki_runtime; /* Real time in microsec */ struct timeval ki_start; /* starting time */ struct timeval ki_childtime; /* time used by process children */ long ki_flag; /* P_* flags */ long ki_kiflag; /* KI_* flags (below) */ int ki_traceflag; /* Kernel trace points */ char ki_stat; /* S* process status */ signed char ki_nice; /* Process "nice" value */ char ki_lock; /* Process lock (prevent swap) count */ char ki_rqindex; /* Run queue index */ u_char ki_oncpu_old; /* Which cpu we are on (legacy) */ u_char ki_lastcpu_old; /* Last cpu we were on (legacy) */ char ki_tdname[TDNAMLEN+1]; /* thread name */ char ki_wmesg[WMESGLEN+1]; /* wchan message */ char ki_login[LOGNAMELEN+1]; /* setlogin name */ char ki_lockname[LOCKNAMELEN+1]; /* lock name */ char ki_comm[COMMLEN+1]; /* command name */ char ki_emul[KI_EMULNAMELEN+1]; /* emulation name */ char ki_loginclass[LOGINCLASSLEN+1]; /* login class */ char ki_moretdname[MAXCOMLEN-TDNAMLEN+1]; /* more thread name */ /* * When adding new variables, take space for char-strings from the * front of ki_sparestrings, and ints from the end of ki_spareints. * That way the spare room from both arrays will remain contiguous. */ char ki_sparestrings[46]; /* spare string space */ int ki_spareints[KI_NSPARE_INT]; /* spare room for growth */ uint64_t ki_tdev; /* controlling tty dev */ int ki_oncpu; /* Which cpu we are on */ int ki_lastcpu; /* Last cpu we were on */ int ki_tracer; /* Pid of tracing process */ int ki_flag2; /* P2_* flags */ int ki_fibnum; /* Default FIB number */ u_int ki_cr_flags; /* Credential flags */ int ki_jid; /* Process jail ID */ int ki_numthreads; /* XXXKSE number of threads in total */ lwpid_t ki_tid; /* XXXKSE thread id */ struct priority ki_pri; /* process priority */ struct rusage ki_rusage; /* process rusage statistics */ /* XXX - most fields in ki_rusage_ch are not (yet) filled in */ struct rusage ki_rusage_ch; /* rusage of children processes */ struct pcb *ki_pcb; /* kernel virtual addr of pcb */ void *ki_kstack; /* kernel virtual addr of stack */ void *ki_udata; /* User convenience pointer */ struct thread *ki_tdaddr; /* address of thread */ /* * When adding new variables, take space for pointers from the * front of ki_spareptrs, and longs from the end of ki_sparelongs. * That way the spare room from both arrays will remain contiguous. */ struct pwddesc *ki_pd; /* pointer to process paths info */ void *ki_spareptrs[KI_NSPARE_PTR]; /* spare room for growth */ long ki_sparelongs[KI_NSPARE_LONG]; /* spare room for growth */ long ki_sflag; /* PS_* flags */ long ki_tdflags; /* XXXKSE kthread flag */ }; void fill_kinfo_proc(struct proc *, struct kinfo_proc *); /* XXX - the following two defines are temporary */ #define ki_childstime ki_rusage_ch.ru_stime #define ki_childutime ki_rusage_ch.ru_utime /* * Legacy PS_ flag. This moved to p_flag but is maintained for * compatibility. */ #define PS_INMEM 0x00001 /* Loaded into memory. */ /* ki_sessflag values */ #define KI_CTTY 0x00000001 /* controlling tty vnode active */ #define KI_SLEADER 0x00000002 /* session leader */ #define KI_LOCKBLOCK 0x00000004 /* proc blocked on lock ki_lockname */ /* * This used to be the per-process structure containing data that * isn't needed in core when the process is swapped out, but now it * remains only for the benefit of a.out core dumps. */ struct user { struct pstats u_stats; /* *p_stats */ struct kinfo_proc u_kproc; /* eproc */ }; /* * The KERN_PROC_FILE sysctl allows a process to dump the file descriptor * array of another process. */ #define KF_ATTR_VALID 0x0001 #define KF_TYPE_NONE 0 #define KF_TYPE_VNODE 1 #define KF_TYPE_SOCKET 2 #define KF_TYPE_PIPE 3 #define KF_TYPE_FIFO 4 #define KF_TYPE_KQUEUE 5 /* was KF_TYPE_CRYPTO 6 */ #define KF_TYPE_MQUEUE 7 #define KF_TYPE_SHM 8 #define KF_TYPE_SEM 9 #define KF_TYPE_PTS 10 #define KF_TYPE_PROCDESC 11 #define KF_TYPE_DEV 12 #define KF_TYPE_EVENTFD 13 #define KF_TYPE_UNKNOWN 255 #define KF_VTYPE_VNON 0 #define KF_VTYPE_VREG 1 #define KF_VTYPE_VDIR 2 #define KF_VTYPE_VBLK 3 #define KF_VTYPE_VCHR 4 #define KF_VTYPE_VLNK 5 #define KF_VTYPE_VSOCK 6 #define KF_VTYPE_VFIFO 7 #define KF_VTYPE_VBAD 8 #define KF_VTYPE_UNKNOWN 255 #define KF_FD_TYPE_CWD -1 /* Current working directory */ #define KF_FD_TYPE_ROOT -2 /* Root directory */ #define KF_FD_TYPE_JAIL -3 /* Jail directory */ #define KF_FD_TYPE_TRACE -4 /* Ktrace vnode */ #define KF_FD_TYPE_TEXT -5 /* Text vnode */ #define KF_FD_TYPE_CTTY -6 /* Controlling terminal */ #define KF_FLAG_READ 0x00000001 #define KF_FLAG_WRITE 0x00000002 #define KF_FLAG_APPEND 0x00000004 #define KF_FLAG_ASYNC 0x00000008 #define KF_FLAG_FSYNC 0x00000010 #define KF_FLAG_NONBLOCK 0x00000020 #define KF_FLAG_DIRECT 0x00000040 #define KF_FLAG_HASLOCK 0x00000080 #define KF_FLAG_SHLOCK 0x00000100 #define KF_FLAG_EXLOCK 0x00000200 #define KF_FLAG_NOFOLLOW 0x00000400 #define KF_FLAG_CREAT 0x00000800 #define KF_FLAG_TRUNC 0x00001000 #define KF_FLAG_EXCL 0x00002000 #define KF_FLAG_EXEC 0x00004000 /* * Old format. Has variable hidden padding due to alignment. * This is a compatibility hack for pre-build 7.1 packages. */ #if defined(__amd64__) #define KINFO_OFILE_SIZE 1328 #endif #if defined(__i386__) #define KINFO_OFILE_SIZE 1324 #endif struct kinfo_ofile { int kf_structsize; /* Size of kinfo_file. */ int kf_type; /* Descriptor type. */ int kf_fd; /* Array index. */ int kf_ref_count; /* Reference count. */ int kf_flags; /* Flags. */ /* XXX Hidden alignment padding here on amd64 */ off_t kf_offset; /* Seek location. */ int kf_vnode_type; /* Vnode type. */ int kf_sock_domain; /* Socket domain. */ int kf_sock_type; /* Socket type. */ int kf_sock_protocol; /* Socket protocol. */ char kf_path[PATH_MAX]; /* Path to file, if any. */ struct sockaddr_storage kf_sa_local; /* Socket address. */ struct sockaddr_storage kf_sa_peer; /* Peer address. */ }; #if defined(__amd64__) || defined(__i386__) /* * This size should never be changed. If you really need to, you must provide * backward ABI compatibility by allocating a new sysctl MIB that will return * the new structure. The current structure has to be returned by the current * sysctl MIB. See how it is done for the kinfo_ofile structure. */ #define KINFO_FILE_SIZE 1392 #endif struct kinfo_file { int kf_structsize; /* Variable size of record. */ int kf_type; /* Descriptor type. */ int kf_fd; /* Array index. */ int kf_ref_count; /* Reference count. */ int kf_flags; /* Flags. */ int kf_pad0; /* Round to 64 bit alignment. */ int64_t kf_offset; /* Seek location. */ union { struct { /* API compatibility with FreeBSD < 12. */ int kf_vnode_type; int kf_sock_domain; int kf_sock_type; int kf_sock_protocol; struct sockaddr_storage kf_sa_local; struct sockaddr_storage kf_sa_peer; }; union { struct { /* Sendq size */ uint32_t kf_sock_sendq; /* Socket domain. */ int kf_sock_domain0; /* Socket type. */ int kf_sock_type0; /* Socket protocol. */ int kf_sock_protocol0; /* Socket address. */ struct sockaddr_storage kf_sa_local; /* Peer address. */ struct sockaddr_storage kf_sa_peer; /* Address of so_pcb. */ uint64_t kf_sock_pcb; /* Address of inp_ppcb. */ uint64_t kf_sock_inpcb; /* Address of unp_conn. */ uint64_t kf_sock_unpconn; /* Send buffer state. */ uint16_t kf_sock_snd_sb_state; /* Receive buffer state. */ uint16_t kf_sock_rcv_sb_state; /* Recvq size. */ uint32_t kf_sock_recvq; } kf_sock; struct { /* Vnode type. */ int kf_file_type; /* Space for future use */ int kf_spareint[3]; uint64_t kf_spareint64[30]; /* Vnode filesystem id. */ uint64_t kf_file_fsid; /* File device. */ uint64_t kf_file_rdev; /* Global file id. */ uint64_t kf_file_fileid; /* File size. */ uint64_t kf_file_size; /* Vnode filesystem id, FreeBSD 11 compat. */ uint32_t kf_file_fsid_freebsd11; /* File device, FreeBSD 11 compat. */ uint32_t kf_file_rdev_freebsd11; /* File mode. */ uint16_t kf_file_mode; /* Round to 64 bit alignment. */ uint16_t kf_file_pad0; uint32_t kf_file_pad1; } kf_file; struct { uint32_t kf_spareint[4]; uint64_t kf_spareint64[32]; uint32_t kf_sem_value; uint16_t kf_sem_mode; } kf_sem; struct { uint32_t kf_spareint[4]; uint64_t kf_spareint64[32]; uint64_t kf_pipe_addr; uint64_t kf_pipe_peer; uint32_t kf_pipe_buffer_cnt; - /* Round to 64 bit alignment. */ - uint32_t kf_pipe_pad0[3]; + uint32_t kf_pipe_buffer_in; + uint32_t kf_pipe_buffer_out; + uint32_t kf_pipe_buffer_size; } kf_pipe; struct { uint32_t kf_spareint[4]; uint64_t kf_spareint64[32]; uint32_t kf_pts_dev_freebsd11; uint32_t kf_pts_pad0; uint64_t kf_pts_dev; /* Round to 64 bit alignment. */ uint32_t kf_pts_pad1[4]; } kf_pts; struct { uint32_t kf_spareint[4]; uint64_t kf_spareint64[32]; pid_t kf_pid; } kf_proc; struct { uint64_t kf_eventfd_value; uint32_t kf_eventfd_flags; + uint32_t kf_eventfd_spareint[3]; + uint64_t kf_eventfd_addr; } kf_eventfd; + struct { + uint64_t kf_kqueue_addr; + int32_t kf_kqueue_count; + int32_t kf_kqueue_state; + } kf_kqueue; } kf_un; }; uint16_t kf_status; /* Status flags. */ uint16_t kf_pad1; /* Round to 32 bit alignment. */ int _kf_ispare0; /* Space for more stuff. */ cap_rights_t kf_cap_rights; /* Capability rights. */ uint64_t _kf_cap_spare; /* Space for future cap_rights_t. */ /* Truncated before copyout in sysctl */ char kf_path[PATH_MAX]; /* Path to file, if any. */ }; struct kinfo_lockf { int kl_structsize; /* Variable size of record. */ int kl_rw; int kl_type; int kl_pid; int kl_sysid; int kl_pad0; uint64_t kl_file_fsid; uint64_t kl_file_rdev; uint64_t kl_file_fileid; off_t kl_start; off_t kl_len; /* len == 0 till the EOF */ char kl_path[PATH_MAX]; }; #define KLOCKF_RW_READ 0x01 #define KLOCKF_RW_WRITE 0x02 #define KLOCKF_TYPE_FLOCK 0x01 #define KLOCKF_TYPE_PID 0x02 #define KLOCKF_TYPE_REMOTE 0x03 /* * The KERN_PROC_VMMAP sysctl allows a process to dump the VM layout of * another process as a series of entries. */ #define KVME_TYPE_NONE 0 #define KVME_TYPE_DEFAULT 1 #define KVME_TYPE_VNODE 2 #define KVME_TYPE_SWAP 3 #define KVME_TYPE_DEVICE 4 #define KVME_TYPE_PHYS 5 #define KVME_TYPE_DEAD 6 #define KVME_TYPE_SG 7 #define KVME_TYPE_MGTDEVICE 8 #define KVME_TYPE_GUARD 9 #define KVME_TYPE_UNKNOWN 255 #define KVME_PROT_READ 0x00000001 #define KVME_PROT_WRITE 0x00000002 #define KVME_PROT_EXEC 0x00000004 #define KVME_FLAG_COW 0x00000001 #define KVME_FLAG_NEEDS_COPY 0x00000002 #define KVME_FLAG_NOCOREDUMP 0x00000004 #define KVME_FLAG_SUPER 0x00000008 #define KVME_FLAG_GROWS_UP 0x00000010 #define KVME_FLAG_GROWS_DOWN 0x00000020 #define KVME_FLAG_USER_WIRED 0x00000040 #if defined(__amd64__) #define KINFO_OVMENTRY_SIZE 1168 #endif #if defined(__i386__) #define KINFO_OVMENTRY_SIZE 1128 #endif struct kinfo_ovmentry { int kve_structsize; /* Size of kinfo_vmmapentry. */ int kve_type; /* Type of map entry. */ void *kve_start; /* Starting address. */ void *kve_end; /* Finishing address. */ int kve_flags; /* Flags on map entry. */ int kve_resident; /* Number of resident pages. */ int kve_private_resident; /* Number of private pages. */ int kve_protection; /* Protection bitmask. */ int kve_ref_count; /* VM obj ref count. */ int kve_shadow_count; /* VM obj shadow count. */ char kve_path[PATH_MAX]; /* Path to VM obj, if any. */ void *_kve_pspare[8]; /* Space for more stuff. */ off_t kve_offset; /* Mapping offset in object */ uint64_t kve_fileid; /* inode number if vnode */ uint32_t kve_fsid; /* dev_t of vnode location */ int _kve_ispare[3]; /* Space for more stuff. */ }; #if defined(__amd64__) || defined(__i386__) #define KINFO_VMENTRY_SIZE 1160 #endif struct kinfo_vmentry { int kve_structsize; /* Variable size of record. */ int kve_type; /* Type of map entry. */ uint64_t kve_start; /* Starting address. */ uint64_t kve_end; /* Finishing address. */ uint64_t kve_offset; /* Mapping offset in object */ uint64_t kve_vn_fileid; /* inode number if vnode */ uint32_t kve_vn_fsid_freebsd11; /* dev_t of vnode location */ int kve_flags; /* Flags on map entry. */ int kve_resident; /* Number of resident pages. */ int kve_private_resident; /* Number of private pages. */ int kve_protection; /* Protection bitmask. */ int kve_ref_count; /* VM obj ref count. */ int kve_shadow_count; /* VM obj shadow count. */ int kve_vn_type; /* Vnode type. */ uint64_t kve_vn_size; /* File size. */ uint32_t kve_vn_rdev_freebsd11; /* Device id if device. */ uint16_t kve_vn_mode; /* File mode. */ uint16_t kve_status; /* Status flags. */ union { uint64_t _kve_vn_fsid; /* dev_t of vnode location */ uint64_t _kve_obj; /* handle of anon obj */ } kve_type_spec; uint64_t kve_vn_rdev; /* Device id if device. */ int _kve_ispare[8]; /* Space for more stuff. */ /* Truncated before copyout in sysctl */ char kve_path[PATH_MAX]; /* Path to VM obj, if any. */ }; #define kve_vn_fsid kve_type_spec._kve_vn_fsid #define kve_obj kve_type_spec._kve_obj /* * The "vm.objects" sysctl provides a list of all VM objects in the system * via an array of these entries. */ struct kinfo_vmobject { int kvo_structsize; /* Variable size of record. */ int kvo_type; /* Object type: KVME_TYPE_*. */ uint64_t kvo_size; /* Object size in pages. */ uint64_t kvo_vn_fileid; /* inode number if vnode. */ uint32_t kvo_vn_fsid_freebsd11; /* dev_t of vnode location. */ int kvo_ref_count; /* Reference count. */ int kvo_shadow_count; /* Shadow count. */ int kvo_memattr; /* Memory attribute. */ uint64_t kvo_resident; /* Number of resident pages. */ uint64_t kvo_active; /* Number of active pages. */ uint64_t kvo_inactive; /* Number of inactive pages. */ union { uint64_t _kvo_vn_fsid; uint64_t _kvo_backing_obj; /* Handle for the backing obj */ } kvo_type_spec; /* Type-specific union */ uint64_t kvo_me; /* Uniq handle for anon obj */ uint64_t _kvo_qspare[6]; uint32_t kvo_swapped; /* Number of swapped pages */ uint32_t _kvo_ispare[7]; char kvo_path[PATH_MAX]; /* Pathname, if any. */ }; #define kvo_vn_fsid kvo_type_spec._kvo_vn_fsid #define kvo_backing_obj kvo_type_spec._kvo_backing_obj /* * The KERN_PROC_KSTACK sysctl allows a process to dump the kernel stacks of * another process as a series of entries. Each stack is represented by a * series of symbol names and offsets as generated by stack_sbuf_print(9). */ #define KKST_MAXLEN 1024 #define KKST_STATE_STACKOK 0 /* Stack is valid. */ #define KKST_STATE_SWAPPED 1 /* Stack swapped out. */ #define KKST_STATE_RUNNING 2 /* Stack ephemeral. */ #if defined(__amd64__) || defined(__i386__) #define KINFO_KSTACK_SIZE 1096 #endif struct kinfo_kstack { lwpid_t kkst_tid; /* ID of thread. */ int kkst_state; /* Validity of stack. */ char kkst_trace[KKST_MAXLEN]; /* String representing stack. */ int _kkst_ispare[16]; /* Space for more stuff. */ }; struct kinfo_sigtramp { void *ksigtramp_start; void *ksigtramp_end; void *ksigtramp_spare[4]; }; #define KMAP_FLAG_WIREFUTURE 0x01 /* all future mappings wil be wired */ #define KMAP_FLAG_ASLR 0x02 /* ASLR is applied to mappings */ #define KMAP_FLAG_ASLR_IGNSTART 0x04 /* ASLR may map into sbrk grow region */ #define KMAP_FLAG_WXORX 0x08 /* W^X mapping policy is enforced */ #define KMAP_FLAG_ASLR_STACK 0x10 /* the stack location is randomized */ struct kinfo_vm_layout { uintptr_t kvm_min_user_addr; uintptr_t kvm_max_user_addr; uintptr_t kvm_text_addr; size_t kvm_text_size; uintptr_t kvm_data_addr; size_t kvm_data_size; uintptr_t kvm_stack_addr; size_t kvm_stack_size; int kvm_map_flags; uintptr_t kvm_spare[14]; }; #ifdef _KERNEL /* Flags for kern_proc_out function. */ #define KERN_PROC_NOTHREADS 0x1 #define KERN_PROC_MASK32 0x2 /* Flags for kern_proc_filedesc_out. */ #define KERN_FILEDESC_PACK_KINFO 0x00000001U /* Flags for kern_proc_vmmap_out. */ #define KERN_VMMAP_PACK_KINFO 0x00000001U struct sbuf; /* * The kern_proc out functions are helper functions to dump process * miscellaneous kinfo structures to sbuf. The main consumers are KERN_PROC * sysctls but they may also be used by other kernel subsystems. * * The functions manipulate the process locking state and expect the process * to be locked on enter. On return the process is unlocked. */ int kern_proc_filedesc_out(struct proc *p, struct sbuf *sb, ssize_t maxlen, int flags); int kern_proc_cwd_out(struct proc *p, struct sbuf *sb, ssize_t maxlen); int kern_proc_out(struct proc *p, struct sbuf *sb, int flags); int kern_proc_vmmap_out(struct proc *p, struct sbuf *sb, ssize_t maxlen, int flags); int vntype_to_kinfo(int vtype); void pack_kinfo(struct kinfo_file *kif); #endif /* !_KERNEL */ #endif