]> git.saurik.com Git - apple/xnu.git/blobdiff - bsd/kern/kern_event.c
xnu-1699.22.73.tar.gz
[apple/xnu.git] / bsd / kern / kern_event.c
index 5d195dcf0d628e111cd4dbaec3529a842b84fc8e..632501473870d4e4edeb15f3f177c8fab69fcc51 100644 (file)
@@ -92,6 +92,9 @@
 #include <libkern/libkern.h>
 #include "net/net_str_id.h"
 
+#include <mach/task.h>
+#include <kern/vm_pressure.h>
+
 MALLOC_DEFINE(M_KQUEUE, "kqueue", "memory for kqueue system");
 
 #define KQ_EVENT NULL
@@ -140,6 +143,8 @@ static void kevent_continue(struct kqueue *kq, void *data, int error);
 static void    kqueue_scan_continue(void *contp, wait_result_t wait_result);
 static int     kqueue_process(struct kqueue *kq, kevent_callback_t callback,
                               void *data, int *countp, struct proc *p);
+static int     kqueue_begin_processing(struct kqueue *kq);
+static void    kqueue_end_processing(struct kqueue *kq);
 static int     knote_process(struct knote *kn, kevent_callback_t callback,
                              void *data, struct kqtailq *inprocessp, struct proc *p);
 static void    knote_put(struct knote *kn);
@@ -183,6 +188,15 @@ static struct filterops proc_filtops = {
         .f_event = filt_proc,
 };
 
+static int filt_vmattach(struct knote *kn);
+static void filt_vmdetach(struct knote *kn);
+static int filt_vm(struct knote *kn, long hint);
+static struct filterops vm_filtops = {
+       .f_attach = filt_vmattach,
+       .f_detach = filt_vmdetach,
+       .f_event = filt_vm,
+};
+
 extern struct filterops fs_filtops;
 
 extern struct filterops sig_filtops;
@@ -238,11 +252,6 @@ static struct filterops user_filtops = {
         .f_touch = filt_usertouch,
 };
 
-#if CONFIG_AUDIT
-/* Audit session filter */
-extern struct filterops audit_session_filtops;
-#endif
-
 /*
  * Table for for all system-defined filters.
  */
@@ -261,11 +270,8 @@ static struct filterops *sysfilt_ops[] = {
        &machport_filtops,              /* EVFILT_MACHPORT */
        &fs_filtops,                    /* EVFILT_FS */
        &user_filtops,                  /* EVFILT_USER */
-#if CONFIG_AUDIT
-       &audit_session_filtops,         /* EVFILT_SESSION */
-#else
-       &bad_filtops,
-#endif
+       &bad_filtops,                   /* unused */
+       &vm_filtops,                    /* EVFILT_VM */
 };
 
 /*
@@ -455,6 +461,7 @@ static int
 filt_procattach(struct knote *kn)
 {
        struct proc *p;
+       pid_t selfpid = (pid_t)0;
 
        assert(PID_MAX < NOTE_PDATAMASK);
        
@@ -466,6 +473,16 @@ filt_procattach(struct knote *kn)
                return (ESRCH);
        }
 
+       if ((kn->kn_sfflags & NOTE_EXIT) != 0) {
+               selfpid = proc_selfpid();
+               /* check for validity of NOTE_EXISTATUS */
+               if (((kn->kn_sfflags & NOTE_EXITSTATUS) != 0) && 
+                       ((p->p_ppid != selfpid) && (((p->p_lflag & P_LTRACED) == 0) || (p->p_oppid != selfpid)))) {
+                       proc_rele(p);
+                       return(EACCES);
+               }
+       }
+
        proc_klist_lock();
 
        kn->kn_flags |= EV_CLEAR;       /* automatically set */
@@ -524,12 +541,57 @@ filt_proc(struct knote *kn, long hint)
                if (event == NOTE_REAP || (event == NOTE_EXIT && !(kn->kn_sfflags & NOTE_REAP))) {
                        kn->kn_flags |= (EV_EOF | EV_ONESHOT);
                }
+               if ((event == NOTE_EXIT) && ((kn->kn_sfflags & NOTE_EXITSTATUS) != 0)) {
+                       kn->kn_fflags |= NOTE_EXITSTATUS;
+                       kn->kn_data = (hint & NOTE_PDATAMASK);
+               }
+               if ((event == NOTE_RESOURCEEND) && ((kn->kn_sfflags & NOTE_RESOURCEEND) != 0)) {
+                       kn->kn_fflags |= NOTE_RESOURCEEND;
+                       kn->kn_data = (hint & NOTE_PDATAMASK);
+               }
        }
 
        /* atomic check, no locking need when called from above */
        return (kn->kn_fflags != 0); 
 }
 
+/*
+ * Virtual memory kevents
+ *
+ * author: Matt Jacobson [matthew_jacobson@apple.com]
+ */
+
+static int
+filt_vmattach(struct knote *kn)
+{      
+       /* 
+        * The note will be cleared once the information has been flushed to the client. 
+        * If there is still pressure, we will be re-alerted.
+        */
+       kn->kn_flags |= EV_CLEAR; 
+       
+       return vm_knote_register(kn);
+}
+
+static void
+filt_vmdetach(struct knote *kn)
+{
+       vm_knote_unregister(kn);
+}
+
+static int
+filt_vm(struct knote *kn, long hint)
+{
+       /* hint == 0 means this is just an alive? check (always true) */
+       if (hint != 0) { 
+               /* If this knote is interested in the event specified in hint... */
+               if ((kn->kn_sfflags & hint) != 0) { 
+                       kn->kn_fflags |= hint;
+               }
+       }
+       
+       return (kn->kn_fflags != 0);
+}
 
 /*
  * filt_timervalidate - process data from user
@@ -872,7 +934,7 @@ 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_flags & EV_TRIGGER) {
+       if (kn->kn_fflags & NOTE_TRIGGER) {
                kn->kn_hookid = 1;
        } else {
                kn->kn_hookid = 0;
@@ -895,10 +957,10 @@ filt_user(struct knote *kn, __unused long hint)
 static void
 filt_usertouch(struct knote *kn, struct kevent64_s *kev, long type)
 {
-        int ffctrl;
+        uint32_t ffctrl;
         switch (type) {
         case EVENT_REGISTER:
-                if (kev->fflags & NOTE_TRIGGER || kev->flags & EV_TRIGGER) {
+                if (kev->fflags & NOTE_TRIGGER) {
                         kn->kn_hookid = 1;
                 }
 
@@ -1511,6 +1573,7 @@ kevent_register(struct kqueue *kq, struct kevent64_s *kev, __unused struct proc
                        error = fops->f_attach(kn);
 
                        kqlock(kq);
+
                        if (error != 0) {
                                /*
                                 * Failed to attach correctly, so drop.
@@ -1594,11 +1657,6 @@ kevent_register(struct kqueue *kq, struct kevent64_s *kev, __unused struct proc
                 */
                if (!fops->f_isfd && fops->f_touch != NULL)
                        fops->f_touch(kn, kev, EVENT_REGISTER);
-
-               /* We may need to push some info down to a networked filesystem */
-               if (kn->kn_filter == EVFILT_VNODE) {
-                       vnode_knoteupdate(kn);
-               }
        }
        /* still have use ref on knote */
 
@@ -1770,6 +1828,47 @@ knote_process(struct knote       *kn,
        return error;
 }
 
+/*
+ * Return 0 to indicate that processing should proceed,
+ * -1 if there is nothing to process.
+ *
+ * Called with kqueue locked and returns the same way,
+ * but may drop lock temporarily.
+ */
+static int
+kqueue_begin_processing(struct kqueue *kq)
+{
+       for (;;) {
+               if (kq->kq_count == 0) {
+                       return -1;
+               }
+
+               /* if someone else is processing the queue, wait */
+               if (kq->kq_nprocess != 0) {
+                       wait_queue_assert_wait((wait_queue_t)kq->kq_wqs, &kq->kq_nprocess, THREAD_UNINT, 0);
+                       kq->kq_state |= KQ_PROCWAIT;
+                       kqunlock(kq);
+                       thread_block(THREAD_CONTINUE_NULL);
+                       kqlock(kq);
+               } else {
+                       kq->kq_nprocess = 1;
+                       return 0;
+               }
+       }
+}
+
+/*
+ * Called with kqueue lock held.
+ */
+static void
+kqueue_end_processing(struct kqueue *kq)
+{
+       kq->kq_nprocess = 0;
+       if (kq->kq_state & KQ_PROCWAIT) {
+               kq->kq_state &= ~KQ_PROCWAIT;
+               wait_queue_wakeup_all((wait_queue_t)kq->kq_wqs, &kq->kq_nprocess, THREAD_AWAKENED);
+       }
+}
 
 /*
  * kqueue_process - process the triggered events in a kqueue
@@ -1799,23 +1898,13 @@ kqueue_process(struct kqueue *kq,
        int error;
 
         TAILQ_INIT(&inprocess);
- restart:
-       if (kq->kq_count == 0) {
+
+       if (kqueue_begin_processing(kq) == -1) {
                *countp = 0;
+               /* Nothing to process */
                return 0;
        }
 
-       /* if someone else is processing the queue, wait */
-       if (hw_atomic_add(&kq->kq_nprocess, 1) != 1) {
-               hw_atomic_sub(&kq->kq_nprocess, 1);
-               wait_queue_assert_wait((wait_queue_t)kq->kq_wqs, &kq->kq_nprocess, THREAD_UNINT, 0);
-               kq->kq_state |= KQ_PROCWAIT;
-               kqunlock(kq);
-               thread_block(THREAD_CONTINUE_NULL);
-               kqlock(kq);
-               goto restart;
-       }
-
        /*
         * Clear any pre-posted status from previous runs, so we only
         * detect events that occur during this run.
@@ -1850,11 +1939,8 @@ kqueue_process(struct kqueue *kq,
                kn->kn_tq = &kq->kq_head;
                TAILQ_INSERT_TAIL(&kq->kq_head, kn, kn_tqe);
        }
-       hw_atomic_sub(&kq->kq_nprocess, 1);
-       if (kq->kq_state & KQ_PROCWAIT) {
-               kq->kq_state &= ~KQ_PROCWAIT;
-               wait_queue_wakeup_all((wait_queue_t)kq->kq_wqs, &kq->kq_nprocess, THREAD_AWAKENED);
-       }
+
+       kqueue_end_processing(kq);
 
        *countp = nevents;
        return error;
@@ -2044,11 +2130,15 @@ static int
 kqueue_select(struct fileproc *fp, int which, void *wql, __unused vfs_context_t ctx)
 {
        struct kqueue *kq = (struct kqueue *)fp->f_data;
-       int again;
-
+       struct knote *kn;
+       struct kqtailq inprocessq;
+       int retnum = 0;
+       
        if (which != FREAD)
                return 0;
 
+       TAILQ_INIT(&inprocessq);
+
        kqlock(kq);
        /* 
         * If this is the first pass, link the wait queue associated with the
@@ -2067,11 +2157,12 @@ kqueue_select(struct fileproc *fp, int which, void *wql, __unused vfs_context_t
                                        (wait_queue_link_t)wql);
        }
 
- retry:
-       again = 0;
-       if (kq->kq_count != 0) {
-               struct knote *kn;
+       if (kqueue_begin_processing(kq) == -1) {
+               kqunlock(kq);
+               return 0;
+       }
 
+       if (kq->kq_count != 0) {
                /*
                 * there is something queued - but it might be a
                 * KN_STAYQUEUED knote, which may or may not have
@@ -2079,31 +2170,42 @@ kqueue_select(struct fileproc *fp, int which, void *wql, __unused vfs_context_t
                 * list of knotes to see, and peek at the stay-
                 * queued ones to be really sure.
                 */
-               TAILQ_FOREACH(kn, &kq->kq_head, kn_tqe) {
-                       int retnum = 0;
-                       if ((kn->kn_status & KN_STAYQUEUED) == 0 ||
-                           (retnum = kn->kn_fop->f_peek(kn)) > 0) {
-                               kqunlock(kq);
-                               return 1;
+               while ((kn = (struct knote*)TAILQ_FIRST(&kq->kq_head)) != NULL) {
+                       if ((kn->kn_status & KN_STAYQUEUED) == 0) {
+                               retnum = 1;
+                               goto out;
                        }
-                       if (retnum < 0)
-                               again++;
+
+                       TAILQ_REMOVE(&kq->kq_head, kn, kn_tqe);
+                       TAILQ_INSERT_TAIL(&inprocessq, kn, kn_tqe);
+
+                       if (kqlock2knoteuse(kq, kn)) {
+                               unsigned peek;
+
+                               peek = kn->kn_fop->f_peek(kn);
+                               if (knoteuse2kqlock(kq, kn)) {
+                                       if (peek > 0) {
+                                               retnum = 1;
+                                               goto out;
+                                       }
+                               } else {
+                                       retnum = 0;
+                               }
+                       } 
                }
        }
 
-       /*
-        * If we stumbled across a knote that couldn't be peeked at,
-        * we have to drop the kq lock and try again.
-        */
-       if (again > 0) {
-               kqunlock(kq);
-               mutex_pause(0);
-               kqlock(kq);
-               goto retry;
+out:
+       /* Return knotes to active queue */
+       while ((kn = TAILQ_FIRST(&inprocessq)) != NULL) {
+               TAILQ_REMOVE(&inprocessq, kn, kn_tqe);
+               kn->kn_tq = &kq->kq_head;
+               TAILQ_INSERT_TAIL(&kq->kq_head, kn, kn_tqe);
        }
 
+       kqueue_end_processing(kq);
        kqunlock(kq);
-       return 0;
+       return retnum;
 }
 
 /*
@@ -2312,10 +2414,7 @@ knote_link_wait_queue(struct knote *kn, struct wait_queue *wq)
 
        kr = wait_queue_link(wq, kq->kq_wqs);
        if (kr == KERN_SUCCESS) {
-               kqlock(kq);
-               kn->kn_status |= KN_STAYQUEUED;
-               knote_enqueue(kn);
-               kqunlock(kq);
+               knote_markstayqueued(kn);
                return 0;
        } else {
                return ENOMEM;
@@ -2531,6 +2630,7 @@ knote_init(void)
 
        /* Initialize the timer filter lock */
        lck_mtx_init(&_filt_timerlock, kq_lck_grp, kq_lck_attr);
+       lck_mtx_init(&vm_pressure_klist_mutex, kq_lck_grp, kq_lck_attr);
 }
 SYSINIT(knote, SI_SUB_PSEUDO, SI_ORDER_ANY, knote_init, NULL)
 
@@ -2843,3 +2943,12 @@ fill_kqueueinfo(struct kqueue *kq, struct kqueue_info * kinfo)
        return(0);
 }
 
+
+void
+knote_markstayqueued(struct knote *kn)
+{
+       kqlock(kn->kn_kq);
+       kn->kn_status |= KN_STAYQUEUED;
+       knote_enqueue(kn);
+       kqunlock(kn->kn_kq);
+}