xnu-3789.70.16.tar.gz

[apple/xnu.git] / bsd / kern / sys_generic.c

diff --git a/bsd/kern/sys_generic.c b/bsd/kern/sys_generic.c
index 1fb49eb91dcd3961b85a932a2103b87c470a3468..2d1431763c7c23d2bce4e28d18f4f9386c3641a3 100644 (file)
--- a/bsd/kern/sys_generic.c
+++ b/bsd/kern/sys_generic.c
@@ -146,7 +146,6 @@ void evpipefree(struct pipe *);
 void postpipeevent(struct pipe *, int);
 void postevent(struct socket *, struct sockbuf *, int);
 extern kern_return_t IOBSDGetPlatformUUID(__darwin_uuid_t uuid, mach_timespec_t timeoutp);
 void postpipeevent(struct pipe *, int);
 void postevent(struct socket *, struct sockbuf *, int);
 extern kern_return_t IOBSDGetPlatformUUID(__darwin_uuid_t uuid, mach_timespec_t timeoutp);

-extern void delay(int);

 
 int rd_uio(struct proc *p, int fdes, uio_t uio, user_ssize_t *retval);
 int wr_uio(struct proc *p, struct fileproc *fp, uio_t uio, user_ssize_t *retval);
 
 int rd_uio(struct proc *p, int fdes, uio_t uio, user_ssize_t *retval);
 int wr_uio(struct proc *p, struct fileproc *fp, uio_t uio, user_ssize_t *retval);


@@ -171,7 +170,7 @@ void select_waitq_init(void);
 void
 select_waitq_init(void)
 {
 void
 select_waitq_init(void)
 {

-       waitq_init(&select_conflict_queue, SYNC_POLICY_FIFO | SYNC_POLICY_DISABLE_IRQ);
+       waitq_init(&select_conflict_queue, SYNC_POLICY_FIFO);

 }
 
 #define f_flag f_fglob->fg_flag
 }
 
 #define f_flag f_fglob->fg_flag


@@ -1204,12 +1203,12 @@ select_internal(struct proc *p, struct select_nocancel_args *uap, uint64_t timeo
                        panic("can't allocate %ld bytes for wqstate buffer",
                              uth->uu_wqstate_sz);
                waitq_set_init(uth->uu_wqset,
                        panic("can't allocate %ld bytes for wqstate buffer",
                              uth->uu_wqstate_sz);
                waitq_set_init(uth->uu_wqset,

-                              SYNC_POLICY_FIFO|SYNC_POLICY_PREPOST|SYNC_POLICY_DISABLE_IRQ, NULL);
+                              SYNC_POLICY_FIFO|SYNC_POLICY_PREPOST, NULL, NULL);

        }
 
        if (!waitq_set_is_valid(uth->uu_wqset))
                waitq_set_init(uth->uu_wqset,
        }
 
        if (!waitq_set_is_valid(uth->uu_wqset))
                waitq_set_init(uth->uu_wqset,

-                              SYNC_POLICY_FIFO|SYNC_POLICY_PREPOST|SYNC_POLICY_DISABLE_IRQ, NULL);
+                              SYNC_POLICY_FIFO|SYNC_POLICY_PREPOST, NULL, NULL);

 
        /* the last chunk of our buffer is an array of waitq pointers */
        seldata->wqp = (uint64_t *)((char *)(uth->uu_wqset) + ALIGN(sizeof(struct waitq_set)));
 
        /* the last chunk of our buffer is an array of waitq pointers */
        seldata->wqp = (uint64_t *)((char *)(uth->uu_wqset) + ALIGN(sizeof(struct waitq_set)));


@@ -1691,7 +1690,7 @@ poll_nocancel(struct proc *p, struct poll_nocancel_args *uap, int32_t *retval)
            (nfds > p->p_rlimit[RLIMIT_NOFILE].rlim_cur && (proc_suser(p) || nfds > FD_SETSIZE)))
                return (EINVAL);
 
            (nfds > p->p_rlimit[RLIMIT_NOFILE].rlim_cur && (proc_suser(p) || nfds > FD_SETSIZE)))
                return (EINVAL);
 

-       kq = kqueue_alloc(p);
+       kq = kqueue_alloc(p, 0);

        if (kq == NULL)
                return (EAGAIN);
 
        if (kq == NULL)
                return (EAGAIN);
 


@@ -1728,7 +1727,6 @@ poll_nocancel(struct proc *p, struct poll_nocancel_args *uap, int32_t *retval)
        OSBitOrAtomic(P_SELECT, &p->p_flag);
        for (i = 0; i < nfds; i++) {
                short events = fds[i].events;
        OSBitOrAtomic(P_SELECT, &p->p_flag);
        for (i = 0; i < nfds; i++) {
                short events = fds[i].events;

-               int kerror = 0;

 
                /* per spec, ignore fd values below zero */
                if (fds[i].fd < 0) {
 
                /* per spec, ignore fd values below zero */
                if (fds[i].fd < 0) {


@@ -1747,19 +1745,19 @@ poll_nocancel(struct proc *p, struct poll_nocancel_args *uap, int32_t *retval)
                        kev.filter = EVFILT_READ;
                        if (events & ( POLLPRI | POLLRDBAND ))
                                kev.flags |= EV_OOBAND;
                        kev.filter = EVFILT_READ;
                        if (events & ( POLLPRI | POLLRDBAND ))
                                kev.flags |= EV_OOBAND;

-                       kerror = kevent_register(kq, &kev, p);
+                       kevent_register(kq, &kev, p);

                }
 
                /* Handle output events */
                }
 
                /* Handle output events */

-               if (kerror == 0 &&
-                   events & ( POLLOUT | POLLWRNORM | POLLWRBAND )) {
+               if ((kev.flags & EV_ERROR) == 0 &&
+                   (events & ( POLLOUT | POLLWRNORM | POLLWRBAND ))) {

                        kev.filter = EVFILT_WRITE;
                        kev.filter = EVFILT_WRITE;

-                       kerror = kevent_register(kq, &kev, p);
+                       kevent_register(kq, &kev, p);

                }
 
                /* Handle BSD extension vnode events */
                }
 
                /* Handle BSD extension vnode events */

-               if (kerror == 0 &&
-                   events & ( POLLEXTEND | POLLATTRIB | POLLNLINK | POLLWRITE )) {
+               if ((kev.flags & EV_ERROR) == 0 &&
+                   (events & ( POLLEXTEND | POLLATTRIB | POLLNLINK | POLLWRITE ))) {

                        kev.filter = EVFILT_VNODE;
                        kev.fflags = 0;
                        if (events & POLLEXTEND)
                        kev.filter = EVFILT_VNODE;
                        kev.fflags = 0;
                        if (events & POLLEXTEND)


@@ -1770,25 +1768,40 @@ poll_nocancel(struct proc *p, struct poll_nocancel_args *uap, int32_t *retval)
                                kev.fflags |= NOTE_LINK;
                        if (events & POLLWRITE)
                                kev.fflags |= NOTE_WRITE;
                                kev.fflags |= NOTE_LINK;
                        if (events & POLLWRITE)
                                kev.fflags |= NOTE_WRITE;

-                       kerror = kevent_register(kq, &kev, p);
+                       kevent_register(kq, &kev, p);

                }
 
                }
 

-               if (kerror != 0) {
+               if (kev.flags & EV_ERROR) {

                        fds[i].revents = POLLNVAL;
                        rfds++;
                } else
                        fds[i].revents = 0;
        }
 
                        fds[i].revents = POLLNVAL;
                        rfds++;
                } else
                        fds[i].revents = 0;
        }
 

-       /* Did we have any trouble registering? */
-       if (rfds > 0)
+       /*
+        * Did we have any trouble registering?
+        * If user space passed 0 FDs, then respect any timeout value passed.
+        * This is an extremely inefficient sleep. If user space passed one or
+        * more FDs, and we had trouble registering _all_ of them, then bail
+        * out. If a subset of the provided FDs failed to register, then we
+        * will still call the kqueue_scan function.
+        */
+       if (nfds && (rfds == nfds))

                goto done;
 
                goto done;
 

+       /*
+        * If any events have trouble registering, an event has fired and we
+        * shouldn't wait for events in kqueue_scan -- use the current time as
+        * the deadline.
+        */
+       if (rfds)
+               getmicrouptime(&atv);
+

        /* scan for, and possibly wait for, the kevents to trigger */
        cont->pca_fds = uap->fds;
        cont->pca_nfds = nfds;
        cont->pca_rfds = rfds;
        /* scan for, and possibly wait for, the kevents to trigger */
        cont->pca_fds = uap->fds;
        cont->pca_nfds = nfds;
        cont->pca_rfds = rfds;

-       error = kqueue_scan(kq, poll_callback, NULL, cont, &atv, p);
+       error = kqueue_scan(kq, poll_callback, NULL, cont, NULL, &atv, p);

        rfds = cont->pca_rfds;
 
  done:
        rfds = cont->pca_rfds;
 
  done:


@@ -2104,7 +2117,7 @@ selrecord(__unused struct proc *selector, struct selinfo *sip, void *s_data)
                return;
 
        if ((sip->si_flags & SI_INITED) == 0) {
                return;
 
        if ((sip->si_flags & SI_INITED) == 0) {

-               waitq_init(&sip->si_waitq, SYNC_POLICY_FIFO | SYNC_POLICY_DISABLE_IRQ);
+               waitq_init(&sip->si_waitq, SYNC_POLICY_FIFO);

                sip->si_flags |= SI_INITED;
                sip->si_flags &= ~SI_CLEAR;
        }
                sip->si_flags |= SI_INITED;
                sip->si_flags &= ~SI_CLEAR;
        }


@@ -3159,10 +3172,15 @@ ledger(struct proc *p, struct ledger_args *args, __unused int32_t *retval)
                error = copyin(args->arg3, (char *)&len, sizeof (len));
        else if (args->cmd == LEDGER_TEMPLATE_INFO)
                error = copyin(args->arg2, (char *)&len, sizeof (len));
                error = copyin(args->arg3, (char *)&len, sizeof (len));
        else if (args->cmd == LEDGER_TEMPLATE_INFO)
                error = copyin(args->arg2, (char *)&len, sizeof (len));

-#ifdef LEDGER_DEBUG

        else if (args->cmd == LEDGER_LIMIT)
        else if (args->cmd == LEDGER_LIMIT)

+#ifdef LEDGER_DEBUG

                error = copyin(args->arg2, (char *)&lla, sizeof (lla));
                error = copyin(args->arg2, (char *)&lla, sizeof (lla));

+#else
+               return (EINVAL);

 #endif
 #endif

+       else if ((args->cmd < 0) || (args->cmd > LEDGER_MAX_CMD))
+               return (EINVAL);
+

        if (error)
                return (error);
        if (len < 0)
        if (error)
                return (error);
        if (len < 0)


@@ -3213,7 +3231,7 @@ ledger(struct proc *p, struct ledger_args *args, __unused int32_t *retval)
 
                        rval = ledger_get_task_entry_info_multiple(task, &buf, &len);
                        proc_rele(proc);
 
                        rval = ledger_get_task_entry_info_multiple(task, &buf, &len);
                        proc_rele(proc);

-                       if ((rval == 0) && (len > 0)) {
+                       if ((rval == 0) && (len >= 0)) {

                                sz = len * sizeof (struct ledger_entry_info);
                                rval = copyout(buf, args->arg2, sz);
                                kfree(buf, sz);
                                sz = len * sizeof (struct ledger_entry_info);
                                rval = copyout(buf, args->arg2, sz);
                                kfree(buf, sz);


@@ -3228,7 +3246,7 @@ ledger(struct proc *p, struct ledger_args *args, __unused int32_t *retval)
                        int sz;
 
                        rval = ledger_template_info(&buf, &len);
                        int sz;
 
                        rval = ledger_template_info(&buf, &len);

-                       if ((rval == 0) && (len > 0)) {
+                       if ((rval == 0) && (len >= 0)) {

                                sz = len * sizeof (struct ledger_template_info);
                                rval = copyout(buf, args->arg1, sz);
                                kfree(buf, sz);
                                sz = len * sizeof (struct ledger_template_info);
                                rval = copyout(buf, args->arg1, sz);
                                kfree(buf, sz);


@@ -3239,6 +3257,8 @@ ledger(struct proc *p, struct ledger_args *args, __unused int32_t *retval)
                }
 
                default:
                }
 
                default:

+                       panic("ledger syscall logic error -- command type %d", args->cmd);
+                       proc_rele(proc);

                        rval = EINVAL;
        }
 
                        rval = EINVAL;
        }
 


@@ -3298,7 +3318,7 @@ static inline struct waitq_set *sysctl_get_wqset(int idx)
        gwqs = (struct g_wqset *)kalloc(sizeof(*gwqs));
        assert(gwqs != NULL);
 
        gwqs = (struct g_wqset *)kalloc(sizeof(*gwqs));
        assert(gwqs != NULL);
 

-       gwqs->wqset = waitq_set_alloc(SYNC_POLICY_FIFO|SYNC_POLICY_PREPOST|SYNC_POLICY_DISABLE_IRQ);
+       gwqs->wqset = waitq_set_alloc(SYNC_POLICY_FIFO|SYNC_POLICY_PREPOST, NULL);

        enqueue_tail(&g_wqset_list, &gwqs->link);
        printf("[WQ]: created new waitq set 0x%llx\n", wqset_id(gwqs->wqset));
 
        enqueue_tail(&g_wqset_list, &gwqs->link);
        printf("[WQ]: created new waitq set 0x%llx\n", wqset_id(gwqs->wqset));
 


@@ -3317,7 +3337,7 @@ static inline struct waitq *global_test_waitq(int idx)
        if (!g_wq_init) {
                g_wq_init = 1;
                for (int i = 0; i < MAX_GLOBAL_TEST_QUEUES; i++)
        if (!g_wq_init) {
                g_wq_init = 1;
                for (int i = 0; i < MAX_GLOBAL_TEST_QUEUES; i++)

-                       waitq_init(&g_wq[i], SYNC_POLICY_FIFO|SYNC_POLICY_DISABLE_IRQ);
+                       waitq_init(&g_wq[i], SYNC_POLICY_FIFO);

        }
 
        return &g_wq[idx % MAX_GLOBAL_TEST_QUEUES];
        }
 
        return &g_wq[idx % MAX_GLOBAL_TEST_QUEUES];