]> git.saurik.com Git - apple/xnu.git/blobdiff - bsd/kern/kern_descrip.c
xnu-2782.30.5.tar.gz
[apple/xnu.git] / bsd / kern / kern_descrip.c
index 1ec789877ebb6bda029ed592e256c7b7b20662c1..722fcfdff6b0bdd75b32afe6944b9db43c56f875 100644 (file)
@@ -1,23 +1,29 @@
 /*
 /*
- * Copyright (c) 2000-2004 Apple Computer, Inc. All rights reserved.
+ * Copyright (c) 2000-2014 Apple Inc. All rights reserved.
  *
  *
- * @APPLE_LICENSE_HEADER_START@
+ * @APPLE_OSREFERENCE_LICENSE_HEADER_START@
  * 
  * 
- * The contents of this file constitute Original Code as defined in and
- * are subject to the Apple Public Source License Version 1.1 (the
- * "License").  You may not use this file except in compliance with the
- * License.  Please obtain a copy of the License at
- * http://www.apple.com/publicsource and read it before using this file.
+ * This file contains Original Code and/or Modifications of Original Code
+ * as defined in and that are subject to the Apple Public Source License
+ * Version 2.0 (the 'License'). You may not use this file except in
+ * compliance with the License. The rights granted to you under the License
+ * may not be used to create, or enable the creation or redistribution of,
+ * unlawful or unlicensed copies of an Apple operating system, or to
+ * circumvent, violate, or enable the circumvention or violation of, any
+ * terms of an Apple operating system software license agreement.
  * 
  * 
- * This Original Code and all software distributed under the License are
- * distributed on an "AS IS" basis, WITHOUT WARRANTY OF ANY KIND, EITHER
+ * Please obtain a copy of the License at
+ * http://www.opensource.apple.com/apsl/ and read it before using this file.
+ * 
+ * The Original Code and all software distributed under the License are
+ * distributed on an 'AS IS' basis, WITHOUT WARRANTY OF ANY KIND, EITHER
  * EXPRESS OR IMPLIED, AND APPLE HEREBY DISCLAIMS ALL SUCH WARRANTIES,
  * INCLUDING WITHOUT LIMITATION, ANY WARRANTIES OF MERCHANTABILITY,
  * EXPRESS OR IMPLIED, AND APPLE HEREBY DISCLAIMS ALL SUCH WARRANTIES,
  * INCLUDING WITHOUT LIMITATION, ANY WARRANTIES OF MERCHANTABILITY,
- * FITNESS FOR A PARTICULAR PURPOSE OR NON-INFRINGEMENT.  Please see the
- * License for the specific language governing rights and limitations
- * under the License.
+ * FITNESS FOR A PARTICULAR PURPOSE, QUIET ENJOYMENT OR NON-INFRINGEMENT.
+ * Please see the License for the specific language governing rights and
+ * limitations under the License.
  * 
  * 
- * @APPLE_LICENSE_HEADER_END@
+ * @APPLE_OSREFERENCE_LICENSE_HEADER_END@
  */
 /* Copyright (c) 1995, 1997 Apple Computer, Inc. All Rights Reserved */
 /*
  */
 /* Copyright (c) 1995, 1997 Apple Computer, Inc. All Rights Reserved */
 /*
  *
  *     @(#)kern_descrip.c      8.8 (Berkeley) 2/14/95
  */
  *
  *     @(#)kern_descrip.c      8.8 (Berkeley) 2/14/95
  */
+/*
+ * NOTICE: This file was modified by SPARTA, Inc. in 2006 to introduce
+ * support for mandatory and extensible security protections.  This notice
+ * is included in support of clause 2.2 (b) of the Apple Public License,
+ * Version 2.0.
+ */
 
 #include <sys/param.h>
 #include <sys/systm.h>
 
 #include <sys/param.h>
 #include <sys/systm.h>
@@ -68,6 +80,7 @@
 #include <sys/proc_internal.h>
 #include <sys/kauth.h>
 #include <sys/file_internal.h>
 #include <sys/proc_internal.h>
 #include <sys/kauth.h>
 #include <sys/file_internal.h>
+#include <sys/guarded.h>
 #include <sys/socket.h>
 #include <sys/socketvar.h>
 #include <sys/stat.h>
 #include <sys/socket.h>
 #include <sys/socketvar.h>
 #include <sys/stat.h>
 #include <sys/resourcevar.h>
 #include <sys/aio_kern.h>
 #include <sys/ev.h>
 #include <sys/resourcevar.h>
 #include <sys/aio_kern.h>
 #include <sys/ev.h>
-#include <kern/lock.h>
+#include <kern/locks.h>
+#include <sys/uio_internal.h>
+#include <sys/codesign.h>
+#include <sys/codedir_internal.h>
 
 
-#include <bsm/audit_kernel.h>
+#include <security/audit/audit.h>
 
 #include <sys/mount_internal.h>
 #include <sys/kdebug.h>
 #include <sys/sysproto.h>
 #include <sys/pipe.h>
 
 #include <sys/mount_internal.h>
 #include <sys/kdebug.h>
 #include <sys/sysproto.h>
 #include <sys/pipe.h>
+#include <sys/spawn.h>
 #include <kern/kern_types.h>
 #include <kern/kalloc.h>
 #include <libkern/OSAtomic.h>
 
 #include <kern/kern_types.h>
 #include <kern/kalloc.h>
 #include <libkern/OSAtomic.h>
 
+#include <sys/ubc_internal.h>
+
+#include <kern/ipc_misc.h>
+#include <vm/vm_protos.h>
+
+#include <mach/mach_port.h>
+#include <stdbool.h>
+
+#if CONFIG_PROTECT
+#include <sys/cprotect.h>
+#endif
+#include <hfs/hfs.h>
+
+kern_return_t ipc_object_copyin(ipc_space_t, mach_port_name_t,
+    mach_msg_type_name_t, ipc_port_t *);
+void ipc_port_release_send(ipc_port_t);
+
 struct psemnode;
 struct pshmnode;
 
 struct psemnode;
 struct pshmnode;
 
-int fdopen(dev_t dev, int mode, int type, struct proc *p);
-int ogetdtablesize(struct proc *p, void *uap, register_t *retval);
-int finishdup(struct proc * p, struct filedesc *fdp, int old, int new, register_t *retval);
+static int finishdup(proc_t p,
+    struct filedesc *fdp, int old, int new, int flags, int32_t *retval);
 
 
-int closef(struct fileglob *fg, struct proc *p);
-int falloc_locked(struct proc *p, struct fileproc **resultfp, int *resultfd, int locked);
-void fddrop(struct proc *p, int fd);
-int fdgetf_noref(struct proc *p, int fd, struct fileproc **resultfp);
+int falloc_locked(proc_t p, struct fileproc **resultfp, int *resultfd, vfs_context_t ctx, int locked);
 void fg_drop(struct fileproc * fp);
 void fg_free(struct fileglob *fg);
 void fg_ref(struct fileproc * fp);
 void fg_drop(struct fileproc * fp);
 void fg_free(struct fileglob *fg);
 void fg_ref(struct fileproc * fp);
-int fp_getfpshm(struct proc *p, int fd, struct fileproc **resultfp, struct pshmnode  **resultpshm);
+void fileport_releasefg(struct fileglob *fg);
+
+/* flags for close_internal_locked */
+#define FD_DUP2RESV 1
+
+/* We don't want these exported */
+
+__private_extern__
+int unlink1(vfs_context_t, struct nameidata *, int);
+
+static void _fdrelse(struct proc * p, int fd);
 
 
-static int closef_finish(struct fileproc *fp, struct fileglob *fg, struct proc *p);
 
 extern void file_lock_init(void);
 
 extern void file_lock_init(void);
-extern int is_suser(void);
-extern int kqueue_stat(struct fileproc *fp, struct stat *st, struct proc *p);
-extern int soo_stat(struct socket *so, struct stat *ub);
-extern int vn_path_package_check(vnode_t vp, char *path, int pathlen, int *component);
 
 extern kauth_scope_t   kauth_scope_fileop;
 
 
 extern kauth_scope_t   kauth_scope_fileop;
 
+/* Conflict wait queue for when selects collide (opaque type) */
+extern struct wait_queue select_conflict_queue;
+
 #define f_flag f_fglob->fg_flag
 #define f_flag f_fglob->fg_flag
-#define f_type f_fglob->fg_type
+#define f_type f_fglob->fg_ops->fo_type
 #define f_msgcount f_fglob->fg_msgcount
 #define f_cred f_fglob->fg_cred
 #define f_ops f_fglob->fg_ops
 #define f_offset f_fglob->fg_offset
 #define f_data f_fglob->fg_data
 #define f_msgcount f_fglob->fg_msgcount
 #define f_cred f_fglob->fg_cred
 #define f_ops f_fglob->fg_ops
 #define f_offset f_fglob->fg_offset
 #define f_data f_fglob->fg_data
+#define CHECK_ADD_OVERFLOW_INT64L(x, y) \
+               (((((x) > 0) && ((y) > 0) && ((x) > LLONG_MAX - (y))) || \
+               (((x) < 0) && ((y) < 0) && ((x) < LLONG_MIN - (y)))) \
+               ? 1 : 0)
 /*
  * Descriptor management.
  */
 /*
  * Descriptor management.
  */
-struct filelist filehead;      /* head of list of open files */
 struct fmsglist fmsghead;      /* head of list of open files */
 struct fmsglist fmsg_ithead;   /* head of list of open files */
 int nfiles;                    /* actual number of open files */
 struct fmsglist fmsghead;      /* head of list of open files */
 struct fmsglist fmsg_ithead;   /* head of list of open files */
 int nfiles;                    /* actual number of open files */
@@ -139,99 +180,349 @@ lck_grp_t * file_lck_grp;
 lck_attr_t * file_lck_attr;
 
 lck_mtx_t * uipc_lock;
 lck_attr_t * file_lck_attr;
 
 lck_mtx_t * uipc_lock;
-lck_mtx_t * file_iterate_lcok;
-lck_mtx_t * file_flist_lock;
 
 
 
 
+/*
+ * check_file_seek_range
+ *
+ * Description: Checks if seek offsets are in the range of 0 to LLONG_MAX.
+ *
+ * Parameters:  fl             Flock structure.
+ *             cur_file_offset Current offset in the file.
+ *
+ * Returns:    0               on Success.
+ *             EOVERFLOW       on overflow.
+ *             EINVAL          on offset less than zero.
+ */
+
+static int
+check_file_seek_range(struct flock *fl, off_t cur_file_offset)
+{
+       if (fl->l_whence == SEEK_CUR) {
+               /* Check if the start marker is beyond LLONG_MAX. */
+               if (CHECK_ADD_OVERFLOW_INT64L(fl->l_start, cur_file_offset)) {
+                       /* Check if start marker is negative */
+                       if (fl->l_start < 0) {
+                               return EINVAL;
+                       }
+                       return EOVERFLOW;
+               }
+               /* Check if the start marker is negative. */
+               if (fl->l_start + cur_file_offset < 0) {
+                       return EINVAL;
+               }
+               /* Check if end marker is beyond LLONG_MAX. */
+               if ((fl->l_len > 0) && (CHECK_ADD_OVERFLOW_INT64L(fl->l_start + 
+                       cur_file_offset, fl->l_len - 1))) {
+                       return EOVERFLOW;
+               }
+               /* Check if the end marker is negative. */
+               if ((fl->l_len <= 0) && (fl->l_start + cur_file_offset +
+                       fl->l_len < 0)) {
+                       return EINVAL;
+               }
+       } else if (fl->l_whence == SEEK_SET) {
+               /* Check if the start marker is negative. */
+               if (fl->l_start < 0) {
+                       return EINVAL;
+               }
+               /* Check if the end marker is beyond LLONG_MAX. */
+               if ((fl->l_len > 0) && 
+                   CHECK_ADD_OVERFLOW_INT64L(fl->l_start, fl->l_len - 1)) {
+                       return EOVERFLOW;
+               }
+               /* Check if the end marker is negative. */
+               if ((fl->l_len < 0) &&  fl->l_start + fl->l_len < 0) {
+                       return EINVAL;
+               }
+       }
+       return 0;
+}
+
+
+/*
+ * file_lock_init
+ *
+ * Description:        Initialize the file lock group and the uipc and flist locks
+ *
+ * Parameters: (void)
+ *
+ * Returns:    void
+ *
+ * Notes:      Called at system startup from bsd_init().
+ */
 void
 file_lock_init(void)
 {
 void
 file_lock_init(void)
 {
-
        /* allocate file lock group attribute and group */
        file_lck_grp_attr= lck_grp_attr_alloc_init();
        /* allocate file lock group attribute and group */
        file_lck_grp_attr= lck_grp_attr_alloc_init();
-       lck_grp_attr_setstat(file_lck_grp_attr);
 
        file_lck_grp = lck_grp_alloc_init("file",  file_lck_grp_attr);
 
        /* Allocate file lock attribute */
        file_lck_attr = lck_attr_alloc_init();
 
        file_lck_grp = lck_grp_alloc_init("file",  file_lck_grp_attr);
 
        /* Allocate file lock attribute */
        file_lck_attr = lck_attr_alloc_init();
-       //lck_attr_setdebug(file_lck_attr);
 
        uipc_lock = lck_mtx_alloc_init(file_lck_grp, file_lck_attr);
 
        uipc_lock = lck_mtx_alloc_init(file_lck_grp, file_lck_attr);
-       file_iterate_lcok = lck_mtx_alloc_init(file_lck_grp, file_lck_attr);
-       file_flist_lock = lck_mtx_alloc_init(file_lck_grp, file_lck_attr);
+}
 
 
-       
 
 
+/*
+ * proc_fdlock, proc_fdlock_spin
+ *
+ * Description:        Lock to control access to the per process struct fileproc
+ *             and struct filedesc
+ *
+ * Parameters: p                               Process to take the lock on
+ *
+ * Returns:    void
+ *
+ * Notes:      The lock is initialized in forkproc() and destroyed in
+ *             reap_child_process().
+ */
+void
+proc_fdlock(proc_t p)
+{
+       lck_mtx_lock(&p->p_fdmlock);
 }
 
 }
 
+void
+proc_fdlock_spin(proc_t p)
+{
+       lck_mtx_lock_spin(&p->p_fdmlock);
+}
 
 void
 
 void
-proc_fdlock(struct proc *p)
+proc_fdlock_assert(proc_t p, int assertflags)
 {
 {
-       lck_mtx_lock(&p->p_fdmlock);
+       lck_mtx_assert(&p->p_fdmlock, assertflags);
 }
 
 }
 
+
+/*
+ * proc_fdunlock
+ *
+ * Description:        Unlock the lock previously locked by a call to proc_fdlock()
+ *
+ * Parameters: p                               Process to drop the lock on
+ *
+ * Returns:    void
+ */
 void
 void
-proc_fdunlock(struct proc *p)
+proc_fdunlock(proc_t p)
 {
        lck_mtx_unlock(&p->p_fdmlock);
 }
 
 {
        lck_mtx_unlock(&p->p_fdmlock);
 }
 
+
 /*
  * System calls on descriptors.
  */
 
 /*
  * System calls on descriptors.
  */
 
+
+/*
+ * getdtablesize
+ *
+ * Description:        Returns the per process maximum size of the descriptor table
+ *
+ * Parameters: p                               Process being queried
+ *             retval                          Pointer to the call return area
+ *
+ * Returns:    0                               Success
+ *
+ * Implicit returns:
+ *             *retval (modified)              Size of dtable
+ */
 int
 int
-getdtablesize(struct proc *p, __unused struct getdtablesize_args *uap, register_t *retval)
+getdtablesize(proc_t p, __unused struct getdtablesize_args *uap, int32_t *retval)
 {
 {
-       proc_fdlock(p);
+       proc_fdlock_spin(p);
        *retval = min((int)p->p_rlimit[RLIMIT_NOFILE].rlim_cur, maxfiles);
        proc_fdunlock(p);
 
        return (0);
 }
 
        *retval = min((int)p->p_rlimit[RLIMIT_NOFILE].rlim_cur, maxfiles);
        proc_fdunlock(p);
 
        return (0);
 }
 
-int
-ogetdtablesize(struct proc *p, __unused void *uap, register_t *retval)
+
+void
+procfdtbl_reservefd(struct proc * p, int fd)
 {
 {
-       proc_fdlock(p);
-       *retval = min((int)p->p_rlimit[RLIMIT_NOFILE].rlim_cur, NOFILE);
-       proc_fdunlock(p);
+       p->p_fd->fd_ofiles[fd] = NULL;
+        p->p_fd->fd_ofileflags[fd] |= UF_RESERVED;
+}
 
 
-       return (0);
+void
+procfdtbl_markclosefd(struct proc * p, int fd)
+{
+        p->p_fd->fd_ofileflags[fd] |= (UF_RESERVED | UF_CLOSING);
 }
 
 }
 
-static __inline__ void
-_fdrelse(struct filedesc *fdp, int fd)
+void
+procfdtbl_releasefd(struct proc * p, int fd, struct fileproc * fp)
+{
+       if (fp != NULL)
+               p->p_fd->fd_ofiles[fd] = fp;
+        p->p_fd->fd_ofileflags[fd] &= ~UF_RESERVED;
+       if ((p->p_fd->fd_ofileflags[fd] & UF_RESVWAIT) == UF_RESVWAIT) {
+               p->p_fd->fd_ofileflags[fd] &= ~UF_RESVWAIT;
+               wakeup(&p->p_fd);
+       }
+}
+
+void 
+procfdtbl_waitfd(struct proc * p, int fd)
+{
+        p->p_fd->fd_ofileflags[fd] |= UF_RESVWAIT;
+       msleep(&p->p_fd, &p->p_fdmlock, PRIBIO, "ftbl_waitfd", NULL);
+}
+
+
+void
+procfdtbl_clearfd(struct proc * p, int fd)
+{
+       int waiting;
+
+       waiting = (p->p_fd->fd_ofileflags[fd] & UF_RESVWAIT);
+       p->p_fd->fd_ofiles[fd] = NULL;       
+       p->p_fd->fd_ofileflags[fd] = 0;
+       if ( waiting == UF_RESVWAIT) {
+               wakeup(&p->p_fd);
+       }
+}
+
+/*
+ * _fdrelse
+ *
+ * Description:        Inline utility function to free an fd in a filedesc
+ *
+ * Parameters: fdp                             Pointer to filedesc fd lies in
+ *             fd                              fd to free
+ *             reserv                          fd should be reserved
+ *
+ * Returns:    void
+ *
+ * Locks:      Assumes proc_fdlock for process pointing to fdp is held by
+ *             the caller
+ */
+static void
+_fdrelse(struct proc * p, int fd)
 {
 {
+       struct filedesc *fdp = p->p_fd;
+       int nfd = 0;
+
        if (fd < fdp->fd_freefile)
                fdp->fd_freefile = fd;
 #if DIAGNOSTIC
        if (fd > fdp->fd_lastfile)
        if (fd < fdp->fd_freefile)
                fdp->fd_freefile = fd;
 #if DIAGNOSTIC
        if (fd > fdp->fd_lastfile)
-               panic("fdrelse: fd_lastfile inconsistent");
+               panic("fdrelse: fd_lastfile inconsistent");
 #endif
 #endif
-       fdp->fd_ofiles[fd] = NULL;
-       fdp->fd_ofileflags[fd] = 0;
+       procfdtbl_clearfd(p, fd);
 
 
-       while ((fd = fdp->fd_lastfile) > 0 &&
-                       fdp->fd_ofiles[fd] == NULL &&
-                       !(fdp->fd_ofileflags[fd] & UF_RESERVED))
+       while ((nfd = fdp->fd_lastfile) > 0 &&
+                       fdp->fd_ofiles[nfd] == NULL &&
+                       !(fdp->fd_ofileflags[nfd] & UF_RESERVED))
                fdp->fd_lastfile--;
 }
 
                fdp->fd_lastfile--;
 }
 
+
+int
+fd_rdwr(
+       int fd,
+       enum uio_rw rw,
+       uint64_t base,
+       int64_t len,
+       enum uio_seg segflg,
+       off_t   offset,
+       int     io_flg,
+       int64_t *aresid)
+{
+        struct fileproc *fp;
+       proc_t  p;
+        int error = 0;
+       int flags = 0;
+       int spacetype;
+       uio_t auio = NULL;
+       char uio_buf[ UIO_SIZEOF(1) ];
+       struct vfs_context context = *(vfs_context_current());
+       bool wrote_some = false;
+
+       p = current_proc();
+
+        error = fp_lookup(p, fd, &fp, 0);
+        if (error)
+                return(error);
+
+       if (fp->f_type != DTYPE_VNODE && fp->f_type != DTYPE_PIPE && fp->f_type != DTYPE_SOCKET) {
+               error = EINVAL;
+               goto out;
+       }
+       if (rw == UIO_WRITE && !(fp->f_flag & FWRITE)) {
+                error = EBADF;
+               goto out;
+       }
+       
+       if (rw == UIO_READ && !(fp->f_flag & FREAD)) {
+               error = EBADF;
+               goto out;
+       }
+       
+       context.vc_ucred = fp->f_fglob->fg_cred;
+
+       if (UIO_SEG_IS_USER_SPACE(segflg))
+               spacetype = proc_is64bit(p) ? UIO_USERSPACE64 : UIO_USERSPACE32;
+       else
+               spacetype = UIO_SYSSPACE;
+
+       auio = uio_createwithbuffer(1, offset, spacetype, rw, &uio_buf[0], sizeof(uio_buf));
+
+       uio_addiov(auio, base, len);
+
+       if ( !(io_flg & IO_APPEND))
+               flags = FOF_OFFSET;
+
+       if (rw == UIO_WRITE) {
+               user_ssize_t orig_resid = uio_resid(auio);
+               error = fo_write(fp, auio, flags, &context);
+               wrote_some = uio_resid(auio) < orig_resid;
+       } else
+               error = fo_read(fp, auio, flags, &context);
+
+       if (aresid)
+               *aresid = uio_resid(auio);
+       else {
+               if (uio_resid(auio) && error == 0)
+                       error = EIO;
+       }
+out:
+        if (wrote_some)
+                fp_drop_written(p, fd, fp);
+        else
+                fp_drop(p, fd, fp, 0);
+
+       return error;
+}
+
+
+
 /*
 /*
- * Duplicate a file descriptor.
+ * dup
+ *
+ * Description:        Duplicate a file descriptor.
+ *
+ * Parameters: p                               Process performing the dup
+ *             uap->fd                         The fd to dup
+ *             retval                          Pointer to the call return area
+ *
+ * Returns:    0                               Success
+ *             !0                              Errno
+ *
+ * Implicit returns:
+ *             *retval (modified)              The new descriptor
  */
  */
-/* ARGSUSED */
 int
 int
-dup(p, uap, retval)
-       struct proc *p;
-       struct dup_args *uap;
-       register_t *retval;
+dup(proc_t p, struct dup_args *uap, int32_t *retval)
 {
 {
-       register struct filedesc *fdp = p->p_fd;
-       register int old = uap->fd;
+       struct filedesc *fdp = p->p_fd;
+       int old = uap->fd;
        int new, error;
        struct fileproc *fp;
 
        int new, error;
        struct fileproc *fp;
 
@@ -240,12 +531,18 @@ dup(p, uap, retval)
                proc_fdunlock(p);
                return(error);
        }
                proc_fdunlock(p);
                return(error);
        }
+       if (FP_ISGUARDED(fp, GUARD_DUP)) {
+               error = fp_guard_exception(p, old, fp, kGUARD_EXC_DUP);
+               (void) fp_drop(p, old, fp, 1);
+               proc_fdunlock(p);
+               return (error);
+       }
        if ( (error = fdalloc(p, 0, &new)) ) {
                fp_drop(p, old, fp, 1);
                proc_fdunlock(p);
                return (error);
        }
        if ( (error = fdalloc(p, 0, &new)) ) {
                fp_drop(p, old, fp, 1);
                proc_fdunlock(p);
                return (error);
        }
-       error = finishdup(p, fdp, old, new, retval);
+       error = finishdup(p, fdp, old, new, 0, retval);
        fp_drop(p, old, fp, 1);
        proc_fdunlock(p);
 
        fp_drop(p, old, fp, 1);
        proc_fdunlock(p);
 
@@ -253,28 +550,44 @@ dup(p, uap, retval)
 }
 
 /*
 }
 
 /*
- * Duplicate a file descriptor to a particular value.
+ * dup2
+ *
+ * Description:        Duplicate a file descriptor to a particular value.
+ *
+ * Parameters: p                               Process performing the dup
+ *             uap->from                       The fd to dup
+ *             uap->to                         The fd to dup it to
+ *             retval                          Pointer to the call return area
+ *
+ * Returns:    0                               Success
+ *             !0                              Errno
+ *
+ * Implicit returns:
+ *             *retval (modified)              The new descriptor
  */
  */
-/* ARGSUSED */
 int
 int
-dup2(p, uap, retval)
-       struct proc *p;
-       struct dup2_args *uap;
-       register_t *retval;
+dup2(proc_t p, struct dup2_args *uap, int32_t *retval)
 {
 {
-       register struct filedesc *fdp = p->p_fd;
-       register int old = uap->from, new = uap->to;
+       struct filedesc *fdp = p->p_fd;
+       int old = uap->from, new = uap->to;
        int i, error;
        int i, error;
-       struct fileproc *fp;
+       struct fileproc *fp, *nfp;
 
        proc_fdlock(p);
 
 
        proc_fdlock(p);
 
+startover:
        if ( (error = fp_lookup(p, old, &fp, 1)) ) {
                proc_fdunlock(p);
                return(error);
        }
        if ( (error = fp_lookup(p, old, &fp, 1)) ) {
                proc_fdunlock(p);
                return(error);
        }
+       if (FP_ISGUARDED(fp, GUARD_DUP)) {
+               error = fp_guard_exception(p, old, fp, kGUARD_EXC_DUP);
+               (void) fp_drop(p, old, fp, 1);
+               proc_fdunlock(p);
+               return (error);
+       }
        if (new < 0 ||
        if (new < 0 ||
-               new >= p->p_rlimit[RLIMIT_NOFILE].rlim_cur ||
+               (rlim_t)new >= p->p_rlimit[RLIMIT_NOFILE].rlim_cur ||
            new >= maxfiles) {
                fp_drop(p, old, fp, 1);
                proc_fdunlock(p);
            new >= maxfiles) {
                fp_drop(p, old, fp, 1);
                proc_fdunlock(p);
@@ -293,60 +606,165 @@ dup2(p, uap, retval)
                        return (error);
                }
                if (new != i) {
                        return (error);
                }
                if (new != i) {
-                       _fdrelse(fdp, i);
+                       fdrelse(p, i);
                        goto closeit;
                }
        } else {
                        goto closeit;
                }
        } else {
-               struct fileproc **fpp;
-               char flags;
 closeit:
 closeit:
-               flags = fdp->fd_ofileflags[new];
-               if ((flags & (UF_RESERVED | UF_CLOSING)) == UF_RESERVED) {
+               while ((fdp->fd_ofileflags[new] & UF_RESERVED) == UF_RESERVED)  {
+                               fp_drop(p, old, fp, 1);
+                               procfdtbl_waitfd(p, new);
+#if DIAGNOSTIC
+                               proc_fdlock_assert(p, LCK_MTX_ASSERT_OWNED);
+#endif
+                               goto startover;
+               }
+
+               if ((fdp->fd_ofiles[new] != NULL) &&
+                   ((error = fp_lookup(p, new, &nfp, 1)) == 0)) {
                        fp_drop(p, old, fp, 1);
                        fp_drop(p, old, fp, 1);
-                       proc_fdunlock(p);
-                       return (EBADF);
+                       if (FP_ISGUARDED(nfp, GUARD_CLOSE)) {
+                               error = fp_guard_exception(p,
+                                   new, nfp, kGUARD_EXC_CLOSE);
+                               (void) fp_drop(p, new, nfp, 1);
+                               proc_fdunlock(p);
+                               return (error);
+                       }
+                       (void)close_internal_locked(p, new, nfp, FD_DUP2RESV);
+#if DIAGNOSTIC
+                       proc_fdlock_assert(p, LCK_MTX_ASSERT_OWNED);
+#endif
+                       procfdtbl_clearfd(p, new);
+                       goto startover;
+               } else  {
+#if DIAGNOSTIC
+                       if (fdp->fd_ofiles[new] != NULL)
+                               panic("dup2: no ref on fileproc %d", new);
+#endif
+                       procfdtbl_reservefd(p, new);
                }
 
                }
 
-               /*
-                * dup2() must succeed even if the close has an error.
-                */
-               if (*(fpp = &fdp->fd_ofiles[new])) {
-                       struct fileproc *nfp = *fpp;
+#if DIAGNOSTIC
+               proc_fdlock_assert(p, LCK_MTX_ASSERT_OWNED);
+#endif
 
 
-                       close_internal(p, new, nfp, (CLOSEINT_LOCKED | CLOSEINT_WAITONCLOSE | CLOSEINT_NOFDRELSE | CLOSEINT_NOFDNOREF));
-                       *fpp = NULL;
-               }
        }
        }
-       error = finishdup(p, fdp, old, new, retval);
+#if DIAGNOSTIC
+       if (fdp->fd_ofiles[new] != 0)
+               panic("dup2: overwriting fd_ofiles with new %d", new);
+       if ((fdp->fd_ofileflags[new] & UF_RESERVED) == 0)
+               panic("dup2: unreserved fileflags with new %d", new);
+#endif
+       error = finishdup(p, fdp, old, new, 0, retval);
        fp_drop(p, old, fp, 1);
        proc_fdunlock(p);
 
        return(error);
 }
 
        fp_drop(p, old, fp, 1);
        proc_fdunlock(p);
 
        return(error);
 }
 
+
+/*
+ * fcntl
+ *
+ * Description:        The file control system call.
+ *
+ * Parameters: p                               Process performing the fcntl
+ *             uap->fd                         The fd to operate against
+ *             uap->cmd                        The command to perform
+ *             uap->arg                        Pointer to the command argument
+ *             retval                          Pointer to the call return area
+ *
+ * Returns:    0                               Success
+ *             !0                              Errno (see fcntl_nocancel)
+ *
+ * Implicit returns:
+ *             *retval (modified)              fcntl return value (if any)
+ *
+ * Notes:      This system call differs from fcntl_nocancel() in that it
+ *             tests for cancellation prior to performing a potentially
+ *             blocking operation.
+ */
+int
+fcntl(proc_t p, struct fcntl_args *uap, int32_t *retval)
+{
+       __pthread_testcancel(1);
+       return(fcntl_nocancel(p, (struct fcntl_nocancel_args *)uap, retval));
+}
+
+
 /*
 /*
- * The file control system call.
+ * fcntl_nocancel
+ *
+ * Description:        A non-cancel-testing file control system call.
+ *
+ * Parameters: p                               Process performing the fcntl
+ *             uap->fd                         The fd to operate against
+ *             uap->cmd                        The command to perform
+ *             uap->arg                        Pointer to the command argument
+ *             retval                          Pointer to the call return area
+ *
+ * Returns:    0                               Success
+ *             EINVAL
+ *     fp_lookup:EBADF                         Bad file descriptor
+ * [F_DUPFD]
+ *     fdalloc:EMFILE
+ *     fdalloc:ENOMEM
+ *     finishdup:EBADF
+ *     finishdup:ENOMEM
+ * [F_SETOWN]
+ *             ESRCH
+ * [F_SETLK]
+ *             EBADF
+ *             EOVERFLOW
+ *     copyin:EFAULT
+ *     vnode_getwithref:???
+ *     VNOP_ADVLOCK:???
+ *     msleep:ETIMEDOUT
+ * [F_GETLK]
+ *             EBADF
+ *             EOVERFLOW
+ *     copyin:EFAULT
+ *     copyout:EFAULT
+ *     vnode_getwithref:???
+ *     VNOP_ADVLOCK:???
+ * [F_PREALLOCATE]
+ *             EBADF
+ *             EINVAL
+ *     copyin:EFAULT
+ *     copyout:EFAULT
+ *     vnode_getwithref:???
+ *     VNOP_ALLOCATE:???
+ * [F_SETSIZE,F_RDADVISE]
+ *             EBADF
+ *     copyin:EFAULT
+ *     vnode_getwithref:???
+ * [F_RDAHEAD,F_NOCACHE]
+ *             EBADF
+ *     vnode_getwithref:???
+ * [???]
+ *
+ * Implicit returns:
+ *             *retval (modified)              fcntl return value (if any)
  */
 int
  */
 int
-fcntl(p, uap, retval)
-       struct proc *p;
-       struct fcntl_args *uap;
-       register_t *retval;
+fcntl_nocancel(proc_t p, struct fcntl_nocancel_args *uap, int32_t *retval)
 {
        int fd = uap->fd;
        struct filedesc *fdp = p->p_fd;
        struct fileproc *fp;
        char *pop;
 {
        int fd = uap->fd;
        struct filedesc *fdp = p->p_fd;
        struct fileproc *fp;
        char *pop;
-       struct vnode *vp;
+       struct vnode *vp = NULLVP;      /* for AUDIT_ARG() at end */
        int i, tmp, error, error2, flg = F_POSIX;
        struct flock fl;
        int i, tmp, error, error2, flg = F_POSIX;
        struct flock fl;
+       struct flocktimeout fltimeout;
+       struct timespec *timeout = NULL;
        struct vfs_context context;
        off_t offset;
        int newmin;
        daddr64_t lbn, bn;
        struct vfs_context context;
        off_t offset;
        int newmin;
        daddr64_t lbn, bn;
-       int devBlockSize = 0;
        unsigned int fflag;
        user_addr_t argp;
        unsigned int fflag;
        user_addr_t argp;
+       boolean_t is64bit;
 
        AUDIT_ARG(fd, uap->fd);
        AUDIT_ARG(cmd, uap->cmd);
 
        AUDIT_ARG(fd, uap->fd);
        AUDIT_ARG(cmd, uap->cmd);
@@ -356,28 +774,45 @@ fcntl(p, uap, retval)
                proc_fdunlock(p);
                return(error);
        }
                proc_fdunlock(p);
                return(error);
        }
-       context.vc_proc = p;
+       context.vc_thread = current_thread();
        context.vc_ucred = fp->f_cred;
        context.vc_ucred = fp->f_cred;
-       if (proc_is64bit(p)) {
+
+       is64bit = proc_is64bit(p);
+       if (is64bit) {
                argp = uap->arg;
        }
        else {
                argp = uap->arg;
        }
        else {
-               /* since the arg parameter is defined as a long but may be either
-                * a long or a pointer we must take care to handle sign extension 
-                * issues.  Our sys call munger will sign extend a long when we are
-                * called from a 32-bit process.  Since we can never have an address
-                * greater than 32-bits from a 32-bit process we lop off the top 
-                * 32-bits to avoid getting the wrong address
+               /*
+                * Since the arg parameter is defined as a long but may be
+                * either a long or a pointer we must take care to handle
+                * sign extension issues.  Our sys call munger will sign
+                * extend a long when we are called from a 32-bit process.
+                * Since we can never have an address greater than 32-bits
+                * from a 32-bit process we lop off the top 32-bits to avoid
+                * getting the wrong address
                 */
                 */
-               argp = CAST_USER_ADDR_T(uap->arg);
+               argp = CAST_USER_ADDR_T((uint32_t)uap->arg);
        }
 
        pop = &fdp->fd_ofileflags[fd];
 
        }
 
        pop = &fdp->fd_ofileflags[fd];
 
+#if CONFIG_MACF
+       error = mac_file_check_fcntl(proc_ucred(p), fp->f_fglob, uap->cmd,
+           uap->arg);
+       if (error)
+               goto out;
+#endif
+
        switch (uap->cmd) {
 
        case F_DUPFD:
        switch (uap->cmd) {
 
        case F_DUPFD:
-               newmin = CAST_DOWN(int, uap->arg);
+       case F_DUPFD_CLOEXEC:
+               if (FP_ISGUARDED(fp, GUARD_DUP)) {
+                       error = fp_guard_exception(p, fd, fp, kGUARD_EXC_DUP);
+                       goto out;
+               }
+               newmin = CAST_DOWN_EXPLICIT(int, uap->arg); /* arg is an int, so we won't lose bits */
+               AUDIT_ARG(value32, newmin);
                if ((u_int)newmin >= p->p_rlimit[RLIMIT_NOFILE].rlim_cur ||
                    newmin >= maxfiles) {
                        error = EINVAL;
                if ((u_int)newmin >= p->p_rlimit[RLIMIT_NOFILE].rlim_cur ||
                    newmin >= maxfiles) {
                        error = EINVAL;
@@ -385,17 +820,27 @@ fcntl(p, uap, retval)
                }
                if ( (error = fdalloc(p, newmin, &i)) )
                        goto out;
                }
                if ( (error = fdalloc(p, newmin, &i)) )
                        goto out;
-               error = finishdup(p, fdp, fd, i, retval);
+               error = finishdup(p, fdp, fd, i,
+                   uap->cmd == F_DUPFD_CLOEXEC ? UF_EXCLOSE : 0, retval);
                goto out;
 
        case F_GETFD:
                goto out;
 
        case F_GETFD:
-               *retval = (*pop & UF_EXCLOSE)? 1 : 0;
+               *retval = (*pop & UF_EXCLOSE)? FD_CLOEXEC : 0;
                error = 0;
                goto out;
 
        case F_SETFD:
                error = 0;
                goto out;
 
        case F_SETFD:
-               *pop = (*pop &~ UF_EXCLOSE) |
-                       (uap->arg & 1)? UF_EXCLOSE : 0;
+               AUDIT_ARG(value32, uap->arg);
+               if (uap->arg & FD_CLOEXEC)
+                       *pop |= UF_EXCLOSE;
+               else {
+                       if (FILEPROC_TYPE(fp) == FTYPE_GUARDED) {
+                               error = fp_guard_exception(p,
+                                   fd, fp, kGUARD_EXC_NOCLOEXEC);
+                               goto out;
+                       }
+                       *pop &= ~UF_EXCLOSE;
+               }
                error = 0;
                goto out;
 
                error = 0;
                goto out;
 
@@ -406,19 +851,20 @@ fcntl(p, uap, retval)
 
        case F_SETFL:
                fp->f_flag &= ~FCNTLFLAGS;
 
        case F_SETFL:
                fp->f_flag &= ~FCNTLFLAGS;
-               tmp = CAST_DOWN(int, uap->arg);
+               tmp = CAST_DOWN_EXPLICIT(int, uap->arg); /* arg is an int, so we won't lose bits */
+               AUDIT_ARG(value32, tmp);
                fp->f_flag |= FFLAGS(tmp) & FCNTLFLAGS;
                tmp = fp->f_flag & FNONBLOCK;
                fp->f_flag |= FFLAGS(tmp) & FCNTLFLAGS;
                tmp = fp->f_flag & FNONBLOCK;
-               error = fo_ioctl(fp, FIONBIO, (caddr_t)&tmp, p);
+               error = fo_ioctl(fp, FIONBIO, (caddr_t)&tmp, &context);
                if (error)
                        goto out;
                tmp = fp->f_flag & FASYNC;
                if (error)
                        goto out;
                tmp = fp->f_flag & FASYNC;
-               error = fo_ioctl(fp, FIOASYNC, (caddr_t)&tmp, p);
+               error = fo_ioctl(fp, FIOASYNC, (caddr_t)&tmp, &context);
                if (!error)
                        goto out;
                fp->f_flag &= ~FNONBLOCK;
                tmp = 0;
                if (!error)
                        goto out;
                fp->f_flag &= ~FNONBLOCK;
                tmp = 0;
-               (void)fo_ioctl(fp, FIONBIO, (caddr_t)&tmp, p);
+               (void)fo_ioctl(fp, FIONBIO, (caddr_t)&tmp, &context);
                goto out;
 
        case F_GETOWN:
                goto out;
 
        case F_GETOWN:
@@ -427,35 +873,76 @@ fcntl(p, uap, retval)
                        error = 0;
                        goto out;
                }
                        error = 0;
                        goto out;
                }
-               error = fo_ioctl(fp, (int)TIOCGPGRP, (caddr_t)retval, p);
+               error = fo_ioctl(fp, (int)TIOCGPGRP, (caddr_t)retval, &context);
                *retval = -*retval;
                goto out;
 
        case F_SETOWN:
                *retval = -*retval;
                goto out;
 
        case F_SETOWN:
-               tmp = CAST_DOWN(pid_t, uap->arg);
+               tmp = CAST_DOWN_EXPLICIT(pid_t, uap->arg); /* arg is an int, so we won't lose bits */
+               AUDIT_ARG(value32, tmp);
                if (fp->f_type == DTYPE_SOCKET) {
                        ((struct socket *)fp->f_data)->so_pgid = tmp;
                        error =0;
                        goto out;
                }
                if (fp->f_type == DTYPE_PIPE) {
                if (fp->f_type == DTYPE_SOCKET) {
                        ((struct socket *)fp->f_data)->so_pgid = tmp;
                        error =0;
                        goto out;
                }
                if (fp->f_type == DTYPE_PIPE) {
-                       error =  fo_ioctl(fp, (int)TIOCSPGRP, (caddr_t)&tmp, p);
+                       error =  fo_ioctl(fp, TIOCSPGRP, (caddr_t)&tmp, &context);
                        goto out;
                }
 
                if (tmp <= 0) {
                        tmp = -tmp;
                } else {
                        goto out;
                }
 
                if (tmp <= 0) {
                        tmp = -tmp;
                } else {
-                       struct proc *p1 = pfind(tmp);
+                       proc_t p1 = proc_find(tmp);
                        if (p1 == 0) {
                                error = ESRCH;
                                goto out;
                        }
                        if (p1 == 0) {
                                error = ESRCH;
                                goto out;
                        }
-                       tmp = (int)p1->p_pgrp->pg_id;
+                       tmp = (int)p1->p_pgrpid;
+                       proc_rele(p1);
+               }
+               error =  fo_ioctl(fp, (int)TIOCSPGRP, (caddr_t)&tmp, &context);
+               goto out;
+
+       case F_SETNOSIGPIPE:
+               tmp = CAST_DOWN_EXPLICIT(int, uap->arg);
+               if (fp->f_type == DTYPE_SOCKET) {
+#if SOCKETS
+                       error = sock_setsockopt((struct socket *)fp->f_data,
+                           SOL_SOCKET, SO_NOSIGPIPE, &tmp, sizeof (tmp));
+#else
+                       error = EINVAL;
+#endif
+               } else {
+                       struct fileglob *fg = fp->f_fglob;
+
+                       lck_mtx_lock_spin(&fg->fg_lock);
+                       if (tmp)
+                               fg->fg_lflags |= FG_NOSIGPIPE;
+                       else
+                               fg->fg_lflags &= FG_NOSIGPIPE;
+                       lck_mtx_unlock(&fg->fg_lock);
+                       error = 0;
+               }
+               goto out;
+
+       case F_GETNOSIGPIPE:
+               if (fp->f_type == DTYPE_SOCKET) {
+#if SOCKETS
+                       int retsize = sizeof (*retval);
+                       error = sock_getsockopt((struct socket *)fp->f_data,
+                           SOL_SOCKET, SO_NOSIGPIPE, retval, &retsize);
+#else
+                       error = EINVAL;
+#endif
+               } else {
+                       *retval = (fp->f_fglob->fg_lflags & FG_NOSIGPIPE) ?
+                               1 : 0;
+                       error = 0;
                }
                }
-               error =  fo_ioctl(fp, (int)TIOCSPGRP, (caddr_t)&tmp, p);
                goto out;
 
                goto out;
 
+       case F_SETLKWTIMEOUT:
        case F_SETLKW:
                flg |= F_WAIT;
                /* Fall into F_SETLK */
        case F_SETLKW:
                flg |= F_WAIT;
                /* Fall into F_SETLK */
@@ -472,16 +959,41 @@ fcntl(p, uap, retval)
                proc_fdunlock(p);
 
                /* Copy in the lock structure */
                proc_fdunlock(p);
 
                /* Copy in the lock structure */
-               error = copyin(argp, (caddr_t)&fl, sizeof (fl));
+               if (uap->cmd == F_SETLKWTIMEOUT) {
+                       error = copyin(argp, (caddr_t) &fltimeout, sizeof(fltimeout));
+                       if (error) {
+                               goto outdrop;
+                       }
+                       fl = fltimeout.fl;
+                       timeout = &fltimeout.timeout;
+               } else {
+                       error = copyin(argp, (caddr_t)&fl, sizeof(fl));
+                       if (error) {
+                               goto outdrop;
+                       }
+               }
+
+               /* Check starting byte and ending byte for EOVERFLOW in SEEK_CUR */
+               /* and ending byte for EOVERFLOW in SEEK_SET */
+               error = check_file_seek_range(&fl, offset);
                if (error) {
                        goto outdrop;
                }
                if (error) {
                        goto outdrop;
                }
+
                if ( (error = vnode_getwithref(vp)) ) {
                        goto outdrop;
                }
                if (fl.l_whence == SEEK_CUR)
                        fl.l_start += offset;
 
                if ( (error = vnode_getwithref(vp)) ) {
                        goto outdrop;
                }
                if (fl.l_whence == SEEK_CUR)
                        fl.l_start += offset;
 
+#if CONFIG_MACF
+               error = mac_file_check_lock(proc_ucred(p), fp->f_fglob,
+                   F_SETLK, &fl);
+               if (error) {
+                       (void)vnode_put(vp);
+                       goto outdrop;
+               }
+#endif
                switch (fl.l_type) {
 
                case F_RDLCK:
                switch (fl.l_type) {
 
                case F_RDLCK:
@@ -490,8 +1002,9 @@ fcntl(p, uap, retval)
                                error = EBADF;
                                goto outdrop;
                        }
                                error = EBADF;
                                goto outdrop;
                        }
+                       // XXX UInt32 unsafe for LP64 kernel
                        OSBitOrAtomic(P_LADVLOCK, &p->p_ladvflag);
                        OSBitOrAtomic(P_LADVLOCK, &p->p_ladvflag);
-                       error = VNOP_ADVLOCK(vp, (caddr_t)p, F_SETLK, &fl, flg, &context);
+                       error = VNOP_ADVLOCK(vp, (caddr_t)p, F_SETLK, &fl, flg, &context, timeout);
                        (void)vnode_put(vp);
                        goto outdrop;
 
                        (void)vnode_put(vp);
                        goto outdrop;
 
@@ -501,14 +1014,15 @@ fcntl(p, uap, retval)
                                error = EBADF;
                                goto outdrop;
                        }
                                error = EBADF;
                                goto outdrop;
                        }
+                       // XXX UInt32 unsafe for LP64 kernel
                        OSBitOrAtomic(P_LADVLOCK, &p->p_ladvflag);
                        OSBitOrAtomic(P_LADVLOCK, &p->p_ladvflag);
-                       error = VNOP_ADVLOCK(vp, (caddr_t)p, F_SETLK, &fl, flg, &context);
+                       error = VNOP_ADVLOCK(vp, (caddr_t)p, F_SETLK, &fl, flg, &context, timeout);
                        (void)vnode_put(vp);
                        goto outdrop;
 
                case F_UNLCK:
                        error = VNOP_ADVLOCK(vp, (caddr_t)p, F_UNLCK, &fl,
                        (void)vnode_put(vp);
                        goto outdrop;
 
                case F_UNLCK:
                        error = VNOP_ADVLOCK(vp, (caddr_t)p, F_UNLCK, &fl,
-                               F_POSIX, &context);
+                               F_POSIX, &context, timeout);
                        (void)vnode_put(vp);
                        goto outdrop;
 
                        (void)vnode_put(vp);
                        goto outdrop;
 
@@ -529,20 +1043,57 @@ fcntl(p, uap, retval)
                proc_fdunlock(p);
 
                /* Copy in the lock structure */
                proc_fdunlock(p);
 
                /* Copy in the lock structure */
-               error = copyin(argp, (caddr_t)&fl, sizeof (fl));
+               error = copyin(argp, (caddr_t)&fl, sizeof(fl));
                if (error)
                        goto outdrop;
 
                if (error)
                        goto outdrop;
 
+               /* Check starting byte and ending byte for EOVERFLOW in SEEK_CUR */
+               /* and ending byte for EOVERFLOW in SEEK_SET */
+               error = check_file_seek_range(&fl, offset);
+               if (error) {
+                       goto outdrop;
+               }
+
+               if ((fl.l_whence == SEEK_SET) && (fl.l_start < 0)) {
+                       error = EINVAL;
+                       goto outdrop;
+               }
+
+               switch (fl.l_type) {
+               case F_RDLCK:
+               case F_UNLCK:
+               case F_WRLCK:
+                       break;
+               default:
+                       error = EINVAL;
+                       goto outdrop;
+               }
+
+               switch (fl.l_whence) {
+               case SEEK_CUR:
+               case SEEK_SET:
+               case SEEK_END:
+                       break;
+               default:
+                       error = EINVAL;
+                       goto outdrop;
+               }
+
                if ( (error = vnode_getwithref(vp)) == 0 ) {
                        if (fl.l_whence == SEEK_CUR)
                                fl.l_start += offset;
 
                if ( (error = vnode_getwithref(vp)) == 0 ) {
                        if (fl.l_whence == SEEK_CUR)
                                fl.l_start += offset;
 
-                       error = VNOP_ADVLOCK(vp, (caddr_t)p, F_GETLK, &fl, F_POSIX, &context);
+#if CONFIG_MACF
+                       error = mac_file_check_lock(proc_ucred(p), fp->f_fglob,
+                           uap->cmd, &fl);
+                       if (error == 0)
+#endif
+                       error = VNOP_ADVLOCK(vp, (caddr_t)p, uap->cmd, &fl, F_POSIX, &context, NULL);
 
                        (void)vnode_put(vp);
 
                        if (error == 0)
 
                        (void)vnode_put(vp);
 
                        if (error == 0)
-                               error = copyout((caddr_t)&fl, argp, sizeof (fl));
+                               error = copyout((caddr_t)&fl, argp, sizeof(fl));
                }
                goto outdrop;
 
                }
                goto outdrop;
 
@@ -564,7 +1115,7 @@ fcntl(p, uap, retval)
                        goto outdrop;
                }
 
                        goto outdrop;
                }
 
-               error = copyin(argp, (caddr_t)&alloc_struct, sizeof (alloc_struct));
+               error = copyin(argp, (caddr_t)&alloc_struct, sizeof(alloc_struct));
                if (error)
                        goto outdrop;
 
                if (error)
                        goto outdrop;
 
@@ -624,7 +1175,7 @@ fcntl(p, uap, retval)
                                              &context);
                        (void)vnode_put(vp);
 
                                              &context);
                        (void)vnode_put(vp);
 
-                       error2 = copyout((caddr_t)&alloc_struct, argp, sizeof (alloc_struct));
+                       error2 = copyout((caddr_t)&alloc_struct, argp, sizeof(alloc_struct));
 
                        if (error == 0)
                                error = error2;
 
                        if (error == 0)
                                error = error2;
@@ -637,32 +1188,42 @@ fcntl(p, uap, retval)
                        error = EBADF;
                        goto out;
                }
                        error = EBADF;
                        goto out;
                }
+               vp = (struct vnode *)fp->f_data;
                proc_fdunlock(p);
 
                error = copyin(argp, (caddr_t)&offset, sizeof (off_t));
                if (error)
                        goto outdrop;
                proc_fdunlock(p);
 
                error = copyin(argp, (caddr_t)&offset, sizeof (off_t));
                if (error)
                        goto outdrop;
+               AUDIT_ARG(value64, offset);
 
 
-               /*
-                * Make sure that we are root.  Growing a file
-                * without zero filling the data is a security hole 
-                * root would have access anyway so we'll allow it
-                */
+               error = vnode_getwithref(vp);
+               if (error)
+                       goto outdrop;
 
 
-               if (!is_suser()) {
-                       error = EACCES;
+#if CONFIG_MACF
+               error = mac_vnode_check_truncate(&context,
+                   fp->f_fglob->fg_cred, vp);
+               if (error) {
+                       (void)vnode_put(vp);
                        goto outdrop;
                }
                        goto outdrop;
                }
-               vp = (struct vnode *)fp->f_data;
-
-               if ( (error = vnode_getwithref(vp)) == 0 ) {
-                       /*
+#endif
+               /*
+                * Make sure that we are root.  Growing a file
+                * without zero filling the data is a security hole 
+                * root would have access anyway so we'll allow it
+                */
+               if (!kauth_cred_issuser(kauth_cred_get())) {
+                       error = EACCES;
+               } else {
+                       /*
                         * set the file size
                         */
                         * set the file size
                         */
-                       error = vnode_setsize(vp, offset, IO_NOZEROFILL, &context);
-
-                       (void)vnode_put(vp);
+                       error = vnode_setsize(vp, offset, IO_NOZEROFILL,
+                           &context);
                }
                }
+
+               (void)vnode_put(vp);
                goto outdrop;
 
        case F_RDAHEAD:
                goto outdrop;
 
        case F_RDAHEAD:
@@ -670,29 +1231,62 @@ fcntl(p, uap, retval)
                        error = EBADF;
                        goto out;
                }
                        error = EBADF;
                        goto out;
                }
-               vp = (struct vnode *)fp->f_data;
-               proc_fdunlock(p);
+               if (uap->arg)
+                       fp->f_fglob->fg_flag &= ~FNORDAHEAD;
+               else
+                       fp->f_fglob->fg_flag |= FNORDAHEAD;
 
 
-               if ( (error = vnode_getwithref(vp)) == 0) {
-                       if (uap->arg)
-                               vnode_clearnoreadahead(vp);
-                       else
-                               vnode_setnoreadahead(vp);
+               goto out;
 
 
-                       (void)vnode_put(vp);
+       case F_NOCACHE:
+               if (fp->f_type != DTYPE_VNODE) {
+                       error = EBADF;
+                       goto out;
                }
                }
-               goto outdrop;
+               if (uap->arg)
+                       fp->f_fglob->fg_flag |= FNOCACHE;
+               else
+                       fp->f_fglob->fg_flag &= ~FNOCACHE;
 
 
-       case F_NOCACHE:
+               goto out;
+
+       case F_NODIRECT:
+               if (fp->f_type != DTYPE_VNODE) {
+                       error = EBADF;
+                       goto out;
+               }
+               if (uap->arg)
+                       fp->f_fglob->fg_flag |= FNODIRECT;
+               else
+                       fp->f_fglob->fg_flag &= ~FNODIRECT;
+
+               goto out;
+
+       case F_SINGLE_WRITER:
                if (fp->f_type != DTYPE_VNODE) {
                        error = EBADF;
                        goto out;
                }
                if (fp->f_type != DTYPE_VNODE) {
                        error = EBADF;
                        goto out;
                }
+               if (uap->arg)
+                       fp->f_fglob->fg_flag |= FSINGLE_WRITER;
+               else
+                       fp->f_fglob->fg_flag &= ~FSINGLE_WRITER;
+
+               goto out;
+
+       case F_GLOBAL_NOCACHE:
+               if (fp->f_type != DTYPE_VNODE) {
+                       error = EBADF;
+                       goto out;
+               }
                vp = (struct vnode *)fp->f_data;
                proc_fdunlock(p);
 
                if ( (error = vnode_getwithref(vp)) == 0 ) {
                vp = (struct vnode *)fp->f_data;
                proc_fdunlock(p);
 
                if ( (error = vnode_getwithref(vp)) == 0 ) {
-                       if (uap->arg)
+
+                       *retval = vnode_isnocache(vp);
+
+                       if (uap->arg)
                                vnode_setnocache(vp);
                        else
                                vnode_clearnocache(vp);
                                vnode_setnocache(vp);
                        else
                                vnode_clearnocache(vp);
@@ -701,9 +1295,9 @@ fcntl(p, uap, retval)
                }
                goto outdrop;
 
                }
                goto outdrop;
 
-       case F_GLOBAL_NOCACHE:
-               if (fp->f_type != DTYPE_VNODE) {
-                       error = EBADF;
+       case F_CHECK_OPENEVT:
+               if (fp->f_type != DTYPE_VNODE) {
+                       error = EBADF;
                        goto out;
                }
                vp = (struct vnode *)fp->f_data;
                        goto out;
                }
                vp = (struct vnode *)fp->f_data;
@@ -711,12 +1305,12 @@ fcntl(p, uap, retval)
 
                if ( (error = vnode_getwithref(vp)) == 0 ) {
 
 
                if ( (error = vnode_getwithref(vp)) == 0 ) {
 
-                       *retval = vnode_isnocache(vp);
+                       *retval = vnode_is_openevt(vp);
 
 
-                       if (uap->arg)
-                               vnode_setnocache(vp);
+                       if (uap->arg)
+                               vnode_set_openevt(vp);
                        else
                        else
-                               vnode_clearnocache(vp);
+                               vnode_clear_openevt(vp);
 
                        (void)vnode_put(vp);
                }
 
                        (void)vnode_put(vp);
                }
@@ -732,7 +1326,7 @@ fcntl(p, uap, retval)
                vp = (struct vnode *)fp->f_data;
                proc_fdunlock(p);
 
                vp = (struct vnode *)fp->f_data;
                proc_fdunlock(p);
 
-               if ( (error = copyin(argp, (caddr_t)&ra_struct, sizeof (ra_struct))) )
+               if ( (error = copyin(argp, (caddr_t)&ra_struct, sizeof(ra_struct))) )
                        goto outdrop;
                if ( (error = vnode_getwithref(vp)) == 0 ) {
                        error = VNOP_IOCTL(vp, F_RDADVISE, (caddr_t)&ra_struct, 0, &context);
                        goto outdrop;
                if ( (error = vnode_getwithref(vp)) == 0 ) {
                        error = VNOP_IOCTL(vp, F_RDADVISE, (caddr_t)&ra_struct, 0, &context);
@@ -742,94 +1336,1029 @@ fcntl(p, uap, retval)
                goto outdrop;
                }
 
                goto outdrop;
                }
 
-       case F_READBOOTSTRAP:
-       case F_WRITEBOOTSTRAP: {
-               fbootstraptransfer_t fbt_struct;
-               user_fbootstraptransfer_t user_fbt_struct;
-               int     sizeof_struct;
-               caddr_t boot_structp;
+        case F_FLUSH_DATA:
+
+                if (fp->f_type != DTYPE_VNODE) {
+                        error = EBADF;
+                        goto out;
+                }
+                vp = (struct vnode *)fp->f_data;
+                proc_fdunlock(p);
+
+                if ( (error = vnode_getwithref(vp)) == 0 ) {
+                        error = cluster_push(vp, 0);
+
+                        (void)vnode_put(vp);
+                }
+                goto outdrop;
+
+       case F_LOG2PHYS:
+       case F_LOG2PHYS_EXT: {
+               struct log2phys l2p_struct;    /* structure for allocate command */
+               int devBlockSize;
+
+               off_t file_offset = 0;
+               size_t a_size = 0;
+               size_t run = 0;
+
+               if (uap->cmd == F_LOG2PHYS_EXT) {
+                       error = copyin(argp, (caddr_t)&l2p_struct, sizeof(l2p_struct));
+                       if (error)
+                               goto out;
+                       file_offset = l2p_struct.l2p_devoffset;
+               } else {
+                       file_offset = fp->f_offset;
+               }
+               if (fp->f_type != DTYPE_VNODE) {
+                       error = EBADF;
+                       goto out;
+               }
+               vp = (struct vnode *)fp->f_data;
+               proc_fdunlock(p);
+               if ( (error = vnode_getwithref(vp)) ) {
+                       goto outdrop;
+               }
+               error = VNOP_OFFTOBLK(vp, file_offset, &lbn);
+               if (error) {
+                       (void)vnode_put(vp);
+                       goto outdrop;
+               }
+               error = VNOP_BLKTOOFF(vp, lbn, &offset);
+               if (error) {
+                       (void)vnode_put(vp);
+                       goto outdrop;
+               }
+               devBlockSize = vfs_devblocksize(vnode_mount(vp));
+               if (uap->cmd == F_LOG2PHYS_EXT) {
+#if defined(__LP64__) 
+                       a_size = l2p_struct.l2p_contigbytes;
+#else
+                       if ((l2p_struct.l2p_contigbytes > SIZE_MAX) || (l2p_struct.l2p_contigbytes < 0)) {
+                               /* size_t is 32-bit on a 32-bit kernel, therefore 
+                                * assigning l2p_contigbytes to a_size may have 
+                                * caused integer overflow.  We, therefore, return 
+                                * an error here instead of calculating incorrect 
+                                * value. 
+                                */
+                               printf ("fcntl: F_LOG2PHYS_EXT: l2p_contigbytes=%lld will overflow, returning error\n", l2p_struct.l2p_contigbytes);
+                               error = EFBIG;
+                               goto outdrop;
+                       } else {
+                               a_size = l2p_struct.l2p_contigbytes;
+                       }
+#endif
+               } else {
+                       a_size = devBlockSize;
+               }
+               
+               error = VNOP_BLOCKMAP(vp, offset, a_size, &bn, &run, NULL, 0, &context);
+
+               (void)vnode_put(vp);
+
+               if (!error) {
+                       l2p_struct.l2p_flags = 0;       /* for now */
+                       if (uap->cmd == F_LOG2PHYS_EXT) {
+                               l2p_struct.l2p_contigbytes = run - (file_offset - offset);
+                       } else {
+                               l2p_struct.l2p_contigbytes = 0; /* for now */
+                       }
+
+                       /*
+                        * The block number being -1 suggests that the file offset is not backed
+                        * by any real blocks on-disk.  As a result, just let it be passed back up wholesale.
+                        */
+                       if (bn == -1) {
+                               /* Don't multiply it by the block size */
+                               l2p_struct.l2p_devoffset = bn;
+                       }
+                       else {
+                               l2p_struct.l2p_devoffset = bn * devBlockSize;
+                               l2p_struct.l2p_devoffset += file_offset - offset;
+                       }
+                       error = copyout((caddr_t)&l2p_struct, argp, sizeof(l2p_struct));
+               }
+               goto outdrop;
+               }
+       case F_GETPATH: {
+               char *pathbufp;
+               int pathlen;
+
+               if (fp->f_type != DTYPE_VNODE) {
+                       error = EBADF;
+                       goto out;
+               }
+               vp = (struct vnode *)fp->f_data;
+               proc_fdunlock(p);
+
+               pathlen = MAXPATHLEN;
+               MALLOC(pathbufp, char *, pathlen, M_TEMP, M_WAITOK);
+               if (pathbufp == NULL) {
+                       error = ENOMEM;
+                       goto outdrop;
+               }
+               if ( (error = vnode_getwithref(vp)) == 0 ) {
+                       error = vn_getpath(vp, pathbufp, &pathlen);
+                       (void)vnode_put(vp);
+
+                       if (error == 0)
+                               error = copyout((caddr_t)pathbufp, argp, pathlen);
+               }
+               FREE(pathbufp, M_TEMP);
+               goto outdrop;
+       }
+
+       case F_PATHPKG_CHECK: {
+               char *pathbufp;
+               size_t pathlen;
+
+               if (fp->f_type != DTYPE_VNODE) {
+                       error = EBADF;
+                       goto out;
+               }
+               vp = (struct vnode *)fp->f_data;
+               proc_fdunlock(p);
+
+               pathlen = MAXPATHLEN;
+               pathbufp = kalloc(MAXPATHLEN);
+
+               if ( (error = copyinstr(argp, pathbufp, MAXPATHLEN, &pathlen)) == 0 ) {
+                       if ( (error = vnode_getwithref(vp)) == 0 ) {
+                               AUDIT_ARG(text, pathbufp);
+                               error = vn_path_package_check(vp, pathbufp, pathlen, retval);
+
+                               (void)vnode_put(vp);
+                       }
+               }
+               kfree(pathbufp, MAXPATHLEN);
+               goto outdrop;
+       }
+
+       case F_CHKCLEAN:   // used by regression tests to see if all dirty pages got cleaned by fsync()
+       case F_FULLFSYNC:  // fsync + flush the journal + DKIOCSYNCHRONIZECACHE
+       case F_FREEZE_FS:  // freeze all other fs operations for the fs of this fd
+       case F_THAW_FS: {  // thaw all frozen fs operations for the fs of this fd
+               if (fp->f_type != DTYPE_VNODE) {
+                       error = EBADF;
+                       goto out;
+               }
+               vp = (struct vnode *)fp->f_data;
+               proc_fdunlock(p);
+
+               if ( (error = vnode_getwithref(vp)) == 0 ) {
+                       error = VNOP_IOCTL(vp, uap->cmd, (caddr_t)NULL, 0, &context);
+
+                       (void)vnode_put(vp);
+               }
+               break;
+       }
+
+       /*
+        * SPI (private) for opening a file starting from a dir fd
+        */
+       case F_OPENFROM: {
+               struct user_fopenfrom fopen;
+               struct vnode_attr va;
+               struct nameidata nd;
+               int cmode;
+
+               /* Check if this isn't a valid file descriptor */
+               if ((fp->f_type != DTYPE_VNODE) ||
+                   (fp->f_flag & FREAD) == 0) {
+                       error = EBADF;
+                       goto out;
+               }
+               vp = (struct vnode *)fp->f_data;
+               proc_fdunlock(p);
+
+               if (vnode_getwithref(vp)) {
+                       error = ENOENT;
+                       goto outdrop;
+               }
+               
+               /* Only valid for directories */
+               if (vp->v_type != VDIR) {
+                       vnode_put(vp);
+                       error = ENOTDIR;
+                       goto outdrop;
+               }
+
+               /* Get flags, mode and pathname arguments. */
+               if (IS_64BIT_PROCESS(p)) {
+                       error = copyin(argp, &fopen, sizeof(fopen));
+               } else {
+                       struct user32_fopenfrom fopen32;
+
+                       error = copyin(argp, &fopen32, sizeof(fopen32));
+                       fopen.o_flags = fopen32.o_flags;
+                       fopen.o_mode = fopen32.o_mode;
+                       fopen.o_pathname = CAST_USER_ADDR_T(fopen32.o_pathname);
+               }
+               if (error) {
+                       vnode_put(vp);
+                       goto outdrop;
+               }
+               AUDIT_ARG(fflags, fopen.o_flags);
+               AUDIT_ARG(mode, fopen.o_mode);
+               VATTR_INIT(&va);
+               /* Mask off all but regular access permissions */
+               cmode = ((fopen.o_mode &~ fdp->fd_cmask) & ALLPERMS) & ~S_ISTXT;
+               VATTR_SET(&va, va_mode, cmode & ACCESSPERMS);
+
+               /* Start the lookup relative to the file descriptor's vnode. */
+               NDINIT(&nd, LOOKUP, OP_OPEN, USEDVP | FOLLOW | AUDITVNPATH1, UIO_USERSPACE,
+                      fopen.o_pathname, &context);
+               nd.ni_dvp = vp;
+
+               error = open1(&context, &nd, fopen.o_flags, &va,
+                             fileproc_alloc_init, NULL, retval);
+
+               vnode_put(vp);
+               break;
+       }
+       /*
+        * SPI (private) for unlinking a file starting from a dir fd
+        */
+       case F_UNLINKFROM: {
+               struct nameidata nd;
+               user_addr_t pathname;
+
+               /* Check if this isn't a valid file descriptor */
+               if ((fp->f_type != DTYPE_VNODE) ||
+                   (fp->f_flag & FREAD) == 0) {
+                       error = EBADF;
+                       goto out;
+               }
+               vp = (struct vnode *)fp->f_data;
+               proc_fdunlock(p);
+
+               if (vnode_getwithref(vp)) {
+                       error = ENOENT;
+                       goto outdrop;
+               }
+               
+               /* Only valid for directories */
+               if (vp->v_type != VDIR) {
+                       vnode_put(vp);
+                       error = ENOTDIR;
+                       goto outdrop;
+               }
+
+               /* Get flags, mode and pathname arguments. */
+               if (IS_64BIT_PROCESS(p)) {
+                       pathname = (user_addr_t)argp;
+               } else {
+                       pathname = CAST_USER_ADDR_T(argp);
+               }
+
+               /* Start the lookup relative to the file descriptor's vnode. */
+               NDINIT(&nd, DELETE, OP_UNLINK, USEDVP | AUDITVNPATH1, UIO_USERSPACE,
+                      pathname, &context);
+               nd.ni_dvp = vp;
+
+               error = unlink1(&context, &nd, 0);
+               
+               vnode_put(vp);
+               break;
+
+       }
+
+       case F_ADDSIGS:
+       case F_ADDFILESIGS:
+       {
+               struct user_fsignatures fs;
+               kern_return_t kr;
+               vm_offset_t kernel_blob_addr;
+               vm_size_t kernel_blob_size;
+
+               if (fp->f_type != DTYPE_VNODE) {
+                       error = EBADF;
+                       goto out;
+               }
+               vp = (struct vnode *)fp->f_data;
+               proc_fdunlock(p);
+               error = vnode_getwithref(vp);
+               if (error)
+                       goto outdrop;
+
+               if (IS_64BIT_PROCESS(p)) {
+                       error = copyin(argp, &fs, sizeof (fs));
+               } else {
+                       struct user32_fsignatures fs32;
+
+                       error = copyin(argp, &fs32, sizeof (fs32));
+                       fs.fs_file_start = fs32.fs_file_start;
+                       fs.fs_blob_start = CAST_USER_ADDR_T(fs32.fs_blob_start);
+                       fs.fs_blob_size = fs32.fs_blob_size;
+               }
+
+               if (error) {
+                       vnode_put(vp);
+                       goto outdrop;
+               }
+
+               if(ubc_cs_blob_get(vp, CPU_TYPE_ANY, fs.fs_file_start))
+               {
+                       vnode_put(vp);
+                       goto outdrop;
+               }
+/*
+ * An arbitrary limit, to prevent someone from mapping in a 20GB blob.  This should cover
+ * our use cases for the immediate future, but note that at the time of this commit, some
+ * platforms are nearing 2MB blob sizes (with a prior soft limit of 2.5MB).
+ *
+ * We should consider how we can manage this more effectively; the above means that some
+ * platforms are using megabytes of memory for signing data; it merely hasn't crossed the
+ * threshold considered ridiculous at the time of this change.
+ */
+#define CS_MAX_BLOB_SIZE (10ULL * 1024ULL * 1024ULL)
+               if (fs.fs_blob_size > CS_MAX_BLOB_SIZE) {
+                       error = E2BIG;
+                       vnode_put(vp);
+                       goto outdrop;
+               }
+
+               kernel_blob_size = CAST_DOWN(vm_size_t, fs.fs_blob_size);
+               kr = ubc_cs_blob_allocate(&kernel_blob_addr, &kernel_blob_size);
+               if (kr != KERN_SUCCESS) {
+                       error = ENOMEM;
+                       vnode_put(vp);
+                       goto outdrop;
+               }
+
+               if(uap->cmd == F_ADDSIGS) {
+                       error = copyin(fs.fs_blob_start,
+                                      (void *) kernel_blob_addr,
+                                      kernel_blob_size);
+               } else /* F_ADDFILESIGS */ {
+                       error = vn_rdwr(UIO_READ,
+                                       vp,
+                                       (caddr_t) kernel_blob_addr,
+                                       kernel_blob_size,
+                                        fs.fs_file_start + fs.fs_blob_start,
+                                       UIO_SYSSPACE,
+                                       0,
+                                       kauth_cred_get(),
+                                       0,
+                                       p);
+               }
+               
+               if (error) {
+                       ubc_cs_blob_deallocate(kernel_blob_addr,
+                                              kernel_blob_size);
+                       vnode_put(vp);
+                       goto outdrop;
+               }
+
+               error = ubc_cs_blob_add(
+                       vp,
+                       CPU_TYPE_ANY,   /* not for a specific architecture */
+                       fs.fs_file_start,
+                       kernel_blob_addr,
+                       kernel_blob_size);
+               if (error) {
+                       ubc_cs_blob_deallocate(kernel_blob_addr,
+                                              kernel_blob_size);
+               } else {
+                       /* ubc_blob_add() has consumed "kernel_blob_addr" */
+#if CHECK_CS_VALIDATION_BITMAP
+                       ubc_cs_validation_bitmap_allocate( vp );
+#endif
+               }
+
+               (void) vnode_put(vp);
+               break;
+       }
+       case F_FINDSIGS: {
+#ifdef SECURE_KERNEL
+               error = ENOTSUP;
+#else /* !SECURE_KERNEL */
+               off_t offsetMacho;
+
+               if (fp->f_type != DTYPE_VNODE) {
+                       error = EBADF;
+                       goto out;
+               }
+               vp = (struct vnode *)fp->f_data;
+               proc_fdunlock(p);
+               error = vnode_getwithref(vp);
+               if (error)
+                       goto outdrop;
+
+               error = copyin(argp, &offsetMacho, sizeof(offsetMacho));
+               if (error) {
+                       (void)vnode_put(vp);
+                       goto outdrop;
+               }
+
+#if CONFIG_MACF
+               error = mac_vnode_find_sigs(p, vp, offsetMacho);
+#else
+               error = EPERM;
+#endif
+               if (error) {
+                       (void)vnode_put(vp);
+                       goto outdrop;
+               }
+#endif /* SECURE_KERNEL */
+               break;
+       }
+#if CONFIG_PROTECT
+       case F_GETPROTECTIONCLASS: {
+               int class = 0;
+               
+               if (fp->f_type != DTYPE_VNODE) {
+                       error = EBADF;
+                       goto out;
+               }
+               vp = (struct vnode *)fp->f_data;
+
+               proc_fdunlock(p);
+
+               if (vnode_getwithref(vp)) {
+                       error = ENOENT;
+                       goto outdrop;
+               }
+       
+               error = cp_vnode_getclass (vp, &class);
+               if (error == 0) {
+                       *retval = class;
+               }
+
+               vnode_put(vp);
+               break;
+       }
+       
+       case F_SETPROTECTIONCLASS: {
+               /* tmp must be a valid PROTECTION_CLASS_* */
+               tmp = CAST_DOWN_EXPLICIT(uint32_t, uap->arg);
+               
+               if (fp->f_type != DTYPE_VNODE) {
+                       error = EBADF;
+                       goto out;
+               }
+               vp = (struct vnode *)fp->f_data;
+
+               proc_fdunlock(p);
+       
+               if (vnode_getwithref(vp)) {
+                       error = ENOENT;
+                       goto outdrop;
+               }       
+               
+               /* Only go forward if you have write access */
+               vfs_context_t ctx = vfs_context_current();
+               if(vnode_authorize(vp, NULLVP, (KAUTH_VNODE_ACCESS | KAUTH_VNODE_WRITE_DATA), ctx) != 0) {
+                       vnode_put(vp);
+                       error = EBADF;
+                       goto outdrop;
+               }
+               error = cp_vnode_setclass (vp, tmp);
+               vnode_put(vp);
+               break;
+       }       
+
+       case F_TRANSCODEKEY: {
+               
+               char *backup_keyp = NULL;
+               unsigned backup_key_len = CP_MAX_WRAPPEDKEYSIZE;
+
+               if (fp->f_type != DTYPE_VNODE) {
+                       error = EBADF;
+                       goto out;
+               }
+               
+               vp = (struct vnode *)fp->f_data;
+               proc_fdunlock(p);
+
+               if (vnode_getwithref(vp)) {
+                       error = ENOENT;
+                       goto outdrop;
+               }       
+
+               MALLOC(backup_keyp, char *, backup_key_len, M_TEMP, M_WAITOK);
+               if (backup_keyp == NULL) {
+                       error = ENOMEM;
+                       goto outdrop;
+               }
+
+               error = cp_vnode_transcode (vp, backup_keyp, &backup_key_len);
+               vnode_put(vp);
+
+               if (error == 0) {
+                       error = copyout((caddr_t)backup_keyp, argp, backup_key_len);
+                       *retval = backup_key_len;
+               }
+
+               FREE(backup_keyp, M_TEMP);
+
+               break;
+       }       
+
+       case F_GETPROTECTIONLEVEL:  {
+               uint32_t cp_version = 0;
+
+               if (fp->f_type != DTYPE_VNODE) {
+                       error = EBADF; 
+                       goto out;
+               }
+
+               vp = (struct vnode*) fp->f_data;
+               proc_fdunlock (p);
+
+               if (vnode_getwithref(vp)) {
+                       error = ENOENT;
+                       goto outdrop;
+               }
+
+               /*
+                * if cp_get_major_vers fails, error will be set to proper errno 
+                * and cp_version will still be 0.
+                */
+
+               error = cp_get_root_major_vers (vp, &cp_version);
+               *retval = cp_version;
+
+               vnode_put (vp);
+               break;
+       }
+
+       case F_GETDEFAULTPROTLEVEL:  {
+               uint32_t cp_default = 0;
+
+               if (fp->f_type != DTYPE_VNODE) {
+                       error = EBADF; 
+                       goto out;
+               }
+
+               vp = (struct vnode*) fp->f_data;
+               proc_fdunlock (p);
+
+               if (vnode_getwithref(vp)) {
+                       error = ENOENT;
+                       goto outdrop;
+               }
+
+               /*
+                * if cp_get_major_vers fails, error will be set to proper errno 
+                * and cp_version will still be 0.
+                */
+
+               error = cp_get_default_level(vp, &cp_default);
+               *retval = cp_default;
+
+               vnode_put (vp);
+               break;
+       }
+
+       
+#endif /* CONFIG_PROTECT */
+
+       case F_MOVEDATAEXTENTS: {
+               struct fileproc *fp2 = NULL;
+               struct vnode *src_vp = NULLVP;
+               struct vnode *dst_vp = NULLVP;
+               /* We need to grab the 2nd FD out of the argments before moving on. */
+               int fd2 = CAST_DOWN_EXPLICIT(int32_t, uap->arg);
+               
+               if (fp->f_type != DTYPE_VNODE) {
+                       error = EBADF;
+                       goto out;
+               }
+
+               /* For now, special case HFS+ only, since this is SPI. */
+               src_vp = (struct vnode *)fp->f_data;
+               if (src_vp->v_tag != VT_HFS) {
+                       error = EINVAL;
+                       goto out;
+               }
+
+               /*
+                * Get the references before we start acquiring iocounts on the vnodes, 
+                * while we still hold the proc fd lock
+                */
+               if ( (error = fp_lookup(p, fd2, &fp2, 1)) ) {
+                       error = EBADF;
+                       goto out;
+               }
+               if (fp2->f_type != DTYPE_VNODE) {
+                       fp_drop(p, fd2, fp2, 1);
+                       error = EBADF;
+                       goto out;
+               }
+               dst_vp = (struct vnode *)fp2->f_data;
+               if (dst_vp->v_tag != VT_HFS) {
+                       fp_drop(p, fd2, fp2, 1);
+                       error = EINVAL;
+                       goto out;
+               }
+
+#if CONFIG_MACF
+               /* Re-do MAC checks against the new FD, pass in a fake argument */
+               error = mac_file_check_fcntl(proc_ucred(p), fp2->f_fglob, uap->cmd, 0);
+               if (error) {
+                       fp_drop(p, fd2, fp2, 1);
+                       goto out;
+               }
+#endif
+               /* Audit the 2nd FD */
+               AUDIT_ARG(fd, fd2);
+
+               proc_fdunlock(p);
+
+               if (vnode_getwithref(src_vp)) {
+                       fp_drop(p, fd2, fp2, 0);
+                       error = ENOENT;
+                       goto outdrop;
+               }       
+               if (vnode_getwithref(dst_vp)) {
+                       vnode_put (src_vp);
+                       fp_drop(p, fd2, fp2, 0);
+                       error = ENOENT;
+                       goto outdrop;
+               }       
+               
+               /* 
+                * Basic asserts; validate they are not the same and that
+                * both live on the same filesystem.
+                */
+               if (dst_vp == src_vp) {
+                       vnode_put (src_vp);
+                       vnode_put (dst_vp);
+                       fp_drop (p, fd2, fp2, 0);
+                       error = EINVAL;
+                       goto outdrop;
+               }       
+
+               if (dst_vp->v_mount != src_vp->v_mount) {
+                       vnode_put (src_vp);
+                       vnode_put (dst_vp);
+                       fp_drop (p, fd2, fp2, 0);
+                       error = EXDEV;
+                       goto outdrop;
+               }
+
+               /* Now we have a legit pair of FDs.  Go to work */
+
+               /* Now check for write access to the target files */
+               if(vnode_authorize(src_vp, NULLVP, 
+                                                  (KAUTH_VNODE_ACCESS | KAUTH_VNODE_WRITE_DATA), &context) != 0) {
+                       vnode_put(src_vp);
+                       vnode_put(dst_vp);
+                       fp_drop(p, fd2, fp2, 0);
+                       error = EBADF;
+                       goto outdrop;
+               }
+               
+               if(vnode_authorize(dst_vp, NULLVP, 
+                                                  (KAUTH_VNODE_ACCESS | KAUTH_VNODE_WRITE_DATA), &context) != 0) {
+                       vnode_put(src_vp);
+                       vnode_put(dst_vp);
+                       fp_drop(p, fd2, fp2, 0);
+                       error = EBADF;
+                       goto outdrop;
+               }
+                       
+               /* Verify that both vps point to files and not directories */
+               if ( !vnode_isreg(src_vp) || !vnode_isreg(dst_vp)) {
+                       error = EINVAL;
+                       vnode_put (src_vp);
+                       vnode_put (dst_vp);
+                       fp_drop (p, fd2, fp2, 0);
+                       goto outdrop;
+               }
+
+               /* 
+                * The exchangedata syscall handler passes in 0 for the flags to VNOP_EXCHANGE.
+                * We'll pass in our special bit indicating that the new behavior is expected
+                */
+               
+               error = VNOP_EXCHANGE(src_vp, dst_vp, FSOPT_EXCHANGE_DATA_ONLY, &context);
+               
+               vnode_put (src_vp);
+               vnode_put (dst_vp);
+               fp_drop(p, fd2, fp2, 0);
+               break;
+       }
+       
+       /* 
+        * SPI for making a file compressed.
+        */
+       case F_MAKECOMPRESSED: {
+               uint32_t gcounter = CAST_DOWN_EXPLICIT(uint32_t, uap->arg);
+
+               if (fp->f_type != DTYPE_VNODE) {
+                       error = EBADF; 
+                       goto out;
+               }
+
+               vp = (struct vnode*) fp->f_data;
+               proc_fdunlock (p);
+
+               /* get the vnode */
+               if (vnode_getwithref(vp)) {
+                       error = ENOENT;
+                       goto outdrop;
+               }
+
+               /* Is it a file? */
+               if ((vnode_isreg(vp) == 0) && (vnode_islnk(vp) == 0)) {
+                       vnode_put(vp);
+                       error = EBADF;
+                       goto outdrop;
+               }
+
+               /* invoke ioctl to pass off to FS */
+               /* Only go forward if you have write access */ 
+               vfs_context_t ctx = vfs_context_current();
+               if(vnode_authorize(vp, NULLVP, (KAUTH_VNODE_ACCESS | KAUTH_VNODE_WRITE_DATA), ctx) != 0) {
+                       vnode_put(vp);
+                       error = EBADF;
+                       goto outdrop;
+               }
+
+               error = VNOP_IOCTL(vp, uap->cmd, (caddr_t)&gcounter, 0, &context);
+
+               vnode_put (vp);
+               break;                             
+       }
+                       
+       /*
+        * SPI (private) for indicating to a filesystem that subsequent writes to
+        * the open FD will written to the Fastflow.
+        */
+       case F_SET_GREEDY_MODE:
+               /* intentionally drop through to the same handler as F_SETSTATIC.
+                * both fcntls should pass the argument and their selector into VNOP_IOCTL.
+                */
+
+       /*
+        * SPI (private) for indicating to a filesystem that subsequent writes to
+        * the open FD will represent static content.
+        */
+       case F_SETSTATICCONTENT: {
+               caddr_t ioctl_arg = NULL;
+
+               if (uap->arg) {
+                       ioctl_arg = (caddr_t) 1;
+               }
+
+               if (fp->f_type != DTYPE_VNODE) {
+                       error = EBADF;
+                       goto out;
+               }
+               vp = (struct vnode *)fp->f_data;
+               proc_fdunlock(p);
+
+               error = vnode_getwithref(vp);
+               if (error) {
+                       error = ENOENT;
+                       goto outdrop;
+               }
+
+               /* Only go forward if you have write access */
+               vfs_context_t ctx = vfs_context_current();
+               if(vnode_authorize(vp, NULLVP, (KAUTH_VNODE_ACCESS | KAUTH_VNODE_WRITE_DATA), ctx) != 0) {
+                       vnode_put(vp);
+                       error = EBADF;
+                       goto outdrop;
+               }
+
+               error = VNOP_IOCTL(vp, uap->cmd, ioctl_arg, 0, &context);
+               (void)vnode_put(vp);
+       
+               break;
+       }
+
+       /*
+        * SPI (private) for indicating to the lower level storage driver that the
+        * subsequent writes should be of a particular IO type (burst, greedy, static),
+        * or other flavors that may be necessary.
+        */
+       case F_SETIOTYPE: {
+               caddr_t param_ptr; 
+               uint32_t param;
+
+               if (uap->arg) {
+                       /* extract 32 bits of flags from userland */
+                       param_ptr = (caddr_t) uap->arg;
+                       param = (uint32_t) param_ptr;
+               }
+               else {
+                       /* If no argument is specified, error out */
+                       error = EINVAL;
+                       goto out;
+               }
+               
+               /* 
+                * Validate the different types of flags that can be specified: 
+                * all of them are mutually exclusive for now.
+                */
+               switch (param) {
+                       case F_IOTYPE_ISOCHRONOUS:
+                               break;
+
+                       default:
+                               error = EINVAL;
+                               goto out;
+               }
+
+
+               if (fp->f_type != DTYPE_VNODE) {
+                       error = EBADF;
+                       goto out;
+               }
+               vp = (struct vnode *)fp->f_data;
+               proc_fdunlock(p);
+
+               error = vnode_getwithref(vp);
+               if (error) {
+                       error = ENOENT;
+                       goto outdrop;
+               }
+
+               /* Only go forward if you have write access */
+               vfs_context_t ctx = vfs_context_current();
+               if(vnode_authorize(vp, NULLVP, (KAUTH_VNODE_ACCESS | KAUTH_VNODE_WRITE_DATA), ctx) != 0) {
+                       vnode_put(vp);
+                       error = EBADF;
+                       goto outdrop;
+               }
+
+               error = VNOP_IOCTL(vp, uap->cmd, param_ptr, 0, &context);
+               (void)vnode_put(vp);
+
+               break;
+       }
+
+       
+       /*
+        * Extract the CodeDirectory of the vnode associated with
+        * the file descriptor and copy it back to user space
+        */
+       case F_GETCODEDIR: {
+               struct user_fcodeblobs args;
 
                if (fp->f_type != DTYPE_VNODE) {
                        error = EBADF;
                        goto out;
                }
 
                if (fp->f_type != DTYPE_VNODE) {
                        error = EBADF;
                        goto out;
                }
+
                vp = (struct vnode *)fp->f_data;
                proc_fdunlock(p);
 
                vp = (struct vnode *)fp->f_data;
                proc_fdunlock(p);
 
+               if ((fp->f_flag & FREAD) == 0) {
+                       error = EBADF;
+                       goto outdrop;
+               }
+
                if (IS_64BIT_PROCESS(p)) {
                if (IS_64BIT_PROCESS(p)) {
-                       sizeof_struct = sizeof(user_fbt_struct);
-                       boot_structp = (caddr_t) &user_fbt_struct;
+                       struct user64_fcodeblobs args64;
+
+                       error = copyin(argp, &args64, sizeof(args64));
+                       if (error) 
+                               goto outdrop;
+
+                       args.f_cd_hash = args64.f_cd_hash;
+                       args.f_hash_size = args64.f_hash_size;
+                       args.f_cd_buffer = args64.f_cd_buffer;
+                       args.f_cd_size = args64.f_cd_size;
+                       args.f_out_size = args64.f_out_size;
+                       args.f_arch = args64.f_arch;
+               } else {
+                       struct user32_fcodeblobs args32;
+
+                       error = copyin(argp, &args32, sizeof(args32));
+                       if (error)
+                               goto outdrop;
+
+                       args.f_cd_hash = CAST_USER_ADDR_T(args32.f_cd_hash);
+                       args.f_hash_size = args32.f_hash_size;
+                       args.f_cd_buffer = CAST_USER_ADDR_T(args32.f_cd_buffer);
+                       args.f_cd_size = args32.f_cd_size;
+                       args.f_out_size = CAST_USER_ADDR_T(args32.f_out_size);
+                       args.f_arch = args32.f_arch;
                }
                }
-               else {
-                       sizeof_struct = sizeof(fbt_struct);
-                       boot_structp = (caddr_t) &fbt_struct;
+
+               if (vp->v_ubcinfo == NULL) {
+                       error = EINVAL;
+                       goto outdrop;
                }
                }
-               error = copyin(argp, boot_structp, sizeof_struct);
-               if (error)
+
+               struct cs_blob *t_blob = vp->v_ubcinfo->cs_blobs;
+
+               /*
+                * This call fails if there is no cs_blob corresponding to the
+                * vnode, or if there are multiple cs_blobs present, and the caller
+                * did not specify which cpu_type they want the cs_blob for
+                */
+               if (t_blob == NULL) {
+                       error = ENOENT; /* there is no codesigning blob for this process */
                        goto outdrop;
                        goto outdrop;
-               if ( (error = vnode_getwithref(vp)) ) {
+               } else if (args.f_arch == 0 && t_blob->csb_next != NULL) {
+                       error = ENOENT; /* too many architectures and none specified */
                        goto outdrop;
                }
                        goto outdrop;
                }
-               if (uap->cmd == F_WRITEBOOTSTRAP) {
-                       /*
-                        * Make sure that we are root.  Updating the
-                        * bootstrap on a disk could be a security hole
-                        */
-                       if (!is_suser()) {
-                               (void)vnode_put(vp);
-                               error = EACCES;
+
+               /* If the user specified an architecture, find the right blob */
+               if (args.f_arch != 0) {
+                       while (t_blob) {
+                               if (t_blob->csb_cpu_type == args.f_arch)
+                                       break;
+                               t_blob = t_blob->csb_next;
+                       }
+                       /* The cpu_type the user requested could not be found */
+                       if (t_blob == NULL) {
+                               error = ENOENT;
                                goto outdrop;
                        }
                }
                                goto outdrop;
                        }
                }
-               if (strcmp(vnode_mount(vp)->mnt_vfsstat.f_fstypename, "hfs") != 0) {
-                       error = EINVAL;
-               } else {
-                       /*
-                        * call vnop_ioctl to handle the I/O
-                        */
-                       error = VNOP_IOCTL(vp, uap->cmd, boot_structp, 0, &context);
+
+               const CS_SuperBlob *super_blob = (void *)t_blob->csb_mem_kaddr;
+               const CS_CodeDirectory *cd = findCodeDirectory(super_blob,
+                                                        (char *) super_blob,
+                                                        (char *) super_blob + t_blob->csb_mem_size);
+               if (cd == NULL) {
+                       error = ENOENT;
+                       goto outdrop;
                }
                }
-               (void)vnode_put(vp);
-               goto outdrop;
-       }
-       case F_LOG2PHYS: {
-               struct log2phys l2p_struct;    /* structure for allocate command */
 
 
+               uint64_t buffer_size = ntohl(cd->length);
+
+               if (buffer_size > UINT_MAX) {
+                       error = ERANGE;
+                       goto outdrop;
+               }
+
+               error = copyout(&buffer_size, args.f_out_size, sizeof(unsigned int));
+               if (error) 
+                       goto outdrop;
+
+               if (sizeof(t_blob->csb_sha1) > args.f_hash_size ||
+                                       buffer_size > args.f_cd_size) {
+                       error = ERANGE;
+                       goto outdrop;
+               }
+
+               error = copyout(t_blob->csb_sha1, args.f_cd_hash, sizeof(t_blob->csb_sha1));
+               if (error) 
+                       goto outdrop;
+               error = copyout(cd, args.f_cd_buffer, buffer_size);
+               if (error) 
+                       goto outdrop;
+
+               break;
+       }
+       
+       /* 
+        * Set the vnode pointed to by 'fd'
+        * and tag it as the (potentially future) backing store
+        * for another filesystem
+        */
+       case F_SETBACKINGSTORE: {
                if (fp->f_type != DTYPE_VNODE) {
                        error = EBADF;
                        goto out;
                }
                if (fp->f_type != DTYPE_VNODE) {
                        error = EBADF;
                        goto out;
                }
+               
                vp = (struct vnode *)fp->f_data;
                vp = (struct vnode *)fp->f_data;
-               proc_fdunlock(p);
-               if ( (error = vnode_getwithref(vp)) ) {
-                       goto outdrop;
+
+               if (vp->v_tag != VT_HFS) {
+                       error = EINVAL;
+                       goto out;
                }
                }
-               error = VNOP_OFFTOBLK(vp, fp->f_offset, &lbn);
-               if (error) {
-                       (void)vnode_put(vp);
+               proc_fdunlock(p);
+
+               if (vnode_getwithref(vp)) {
+                       error = ENOENT;
                        goto outdrop;
                }
                        goto outdrop;
                }
-               error = VNOP_BLKTOOFF(vp, lbn, &offset);
-               if (error) {
-                       (void)vnode_put(vp);
+               
+               /* only proceed if you have write access */
+               vfs_context_t ctx = vfs_context_current();
+               if(vnode_authorize(vp, NULLVP, (KAUTH_VNODE_ACCESS | KAUTH_VNODE_WRITE_DATA), ctx) != 0) {
+                       vnode_put(vp);
+                       error = EBADF;
                        goto outdrop;
                }
                        goto outdrop;
                }
-               devBlockSize = vfs_devblocksize(vnode_mount(vp));
-
-               error = VNOP_BLOCKMAP(vp, offset, devBlockSize, &bn, NULL, NULL, 0, &context);
 
 
-               (void)vnode_put(vp);
-
-               if (!error) {
-                       l2p_struct.l2p_flags = 0;       /* for now */
-                       l2p_struct.l2p_contigbytes = 0; /* for now */
-                       l2p_struct.l2p_devoffset = bn * devBlockSize;
-                       l2p_struct.l2p_devoffset += fp->f_offset - offset;
-                       error = copyout((caddr_t)&l2p_struct, argp, sizeof (l2p_struct));
+               
+               /* If arg != 0, set, otherwise unset */
+               if (uap->arg) {
+                       error = VNOP_IOCTL (vp, uap->cmd, (caddr_t)1, 0, &context);
                }
                }
-               goto outdrop;
+               else {
+                       error = VNOP_IOCTL (vp, uap->cmd, (caddr_t)NULL, 0, &context);
                }
                }
-       case F_GETPATH: {
+               
+               vnode_put(vp);
+               break;
+       }
+
+       /* 
+        * like F_GETPATH, but special semantics for
+        * the mobile time machine handler.
+        */
+       case F_GETPATH_MTMINFO: {
                char *pathbufp;
                int pathlen;
 
                char *pathbufp;
                int pathlen;
 
@@ -847,45 +2376,56 @@ fcntl(p, uap, retval)
                        goto outdrop;
                }
                if ( (error = vnode_getwithref(vp)) == 0 ) {
                        goto outdrop;
                }
                if ( (error = vnode_getwithref(vp)) == 0 ) {
-                       error = vn_getpath(vp, pathbufp, &pathlen);
-                       (void)vnode_put(vp);
+                       int backingstore = 0;
+                       
+                       /* Check for error from vn_getpath before moving on */
+                       if ((error = vn_getpath(vp, pathbufp, &pathlen)) == 0) {
+                               if (vp->v_tag == VT_HFS) {
+                                       error = VNOP_IOCTL (vp, uap->cmd, (caddr_t) &backingstore, 0, &context);
+                               }
+                               (void)vnode_put(vp);
 
 
-                       if (error == 0)
-                               error = copyout((caddr_t)pathbufp, argp, pathlen);
+                               if (error == 0) {
+                                       error = copyout((caddr_t)pathbufp, argp, pathlen);
+                               }
+                               if (error == 0) {
+                                       /* 
+                                        * If the copyout was successful, now check to ensure
+                                        * that this vnode is not a BACKINGSTORE vnode.  mtmd
+                                        * wants the path regardless.
+                                        */
+                                       if (backingstore) {
+                                               error = EBUSY;
+                                       }
+                               }
+                       } else
+                               (void)vnode_put(vp);
                }
                FREE(pathbufp, M_TEMP);
                goto outdrop;
        }
 
                }
                FREE(pathbufp, M_TEMP);
                goto outdrop;
        }
 
-       case F_PATHPKG_CHECK: {
-               char *pathbufp;
-               size_t pathlen;
-
-               if (fp->f_type != DTYPE_VNODE) {
-                       error = EBADF;
+       default:
+               /*
+                * This is an fcntl() that we d not recognize at this level;
+                * if this is a vnode, we send it down into the VNOP_IOCTL
+                * for this vnode; this can include special devices, and will
+                * effectively overload fcntl() to send ioctl()'s.
+                */
+               if((uap->cmd & IOC_VOID) && (uap->cmd & IOC_INOUT)){
+            error = EINVAL;
                        goto out;
                }
                        goto out;
                }
-               vp = (struct vnode *)fp->f_data;
-               proc_fdunlock(p);
-
-               pathlen = MAXPATHLEN;
-               pathbufp = kalloc(MAXPATHLEN);
-
-               if ( (error = copyinstr(argp, pathbufp, MAXPATHLEN, &pathlen)) == 0 ) {
-                       if ( (error = vnode_getwithref(vp)) == 0 ) {
-                               error = vn_path_package_check(vp, pathbufp, pathlen, retval);
-
-                               (void)vnode_put(vp);
-                       }
+               
+               /* Catch any now-invalid fcntl() selectors */
+               switch (uap->cmd) {
+                       case F_MARKDEPENDENCY:
+                               error = EINVAL;
+                               goto out;
+                       default:
+                               break;
                }
                }
-               kfree(pathbufp, MAXPATHLEN);
-               goto outdrop;
-       }
 
 
-       case F_CHKCLEAN:   // used by regression tests to see if all dirty pages got cleaned by fsync()
-       case F_FULLFSYNC:  // fsync + flush the journal + DKIOCSYNCHRONIZECACHE
-       case F_FREEZE_FS:  // freeze all other fs operations for the fs of this fd
-       case F_THAW_FS: {  // thaw all frozen fs operations for the fs of this fd
                if (fp->f_type != DTYPE_VNODE) {
                        error = EBADF;
                        goto out;
                if (fp->f_type != DTYPE_VNODE) {
                        error = EBADF;
                        goto out;
@@ -894,35 +2434,87 @@ fcntl(p, uap, retval)
                proc_fdunlock(p);
 
                if ( (error = vnode_getwithref(vp)) == 0 ) {
                proc_fdunlock(p);
 
                if ( (error = vnode_getwithref(vp)) == 0 ) {
-                       error = VNOP_IOCTL(vp, uap->cmd, (caddr_t)NULL, 0, &context);
-
-                       (void)vnode_put(vp);
-               }
-               break;
-       }
-           
-       default:
-               if (uap->cmd < FCNTL_FS_SPECIFIC_BASE) {
-                       error = EINVAL;
-                       goto out;
-               }
-
-               // if it's a fs-specific fcntl() then just pass it through
+#define STK_PARAMS 128
+                       char stkbuf[STK_PARAMS];
+                       unsigned int size;
+                       caddr_t data, memp;
+                       /*
+                        * For this to work properly, we have to copy in the
+                        * ioctl() cmd argument if there is one; we must also
+                        * check that a command parameter, if present, does
+                        * not exceed the maximum command length dictated by
+                        * the number of bits we have available in the command
+                        * to represent a structure length.  Finally, we have
+                        * to copy the results back out, if it is that type of
+                        * ioctl().
+                        */
+                       size = IOCPARM_LEN(uap->cmd);
+                       if (size > IOCPARM_MAX) {
+                               (void)vnode_put(vp);
+                               error = EINVAL;
+                               break;
+                       }
 
 
-               if (fp->f_type != DTYPE_VNODE) {
-                       error = EBADF;
-                       goto out;
-               }
-               vp = (struct vnode *)fp->f_data;
-               proc_fdunlock(p);
+                       memp = NULL;
+                       if (size > sizeof (stkbuf)) {
+                               if ((memp = (caddr_t)kalloc(size)) == 0) {
+                                       (void)vnode_put(vp);
+                                       error = ENOMEM;
+                                       goto outdrop;
+                               }
+                               data = memp;
+                       } else {
+                               data = &stkbuf[0];
+                       }
+                       
+                       if (uap->cmd & IOC_IN) {
+                               if (size) {
+                                       /* structure */
+                                       error = copyin(argp, data, size);
+                                       if (error) {
+                                               (void)vnode_put(vp);
+                                               if (memp)
+                                                       kfree(memp, size);
+                                               goto outdrop;
+                                       }
+
+                                       /* Bzero the section beyond that which was needed */
+                                       if (size <= sizeof(stkbuf)) {
+                                               bzero ( (((uint8_t*)data) + size), (sizeof(stkbuf) - size));
+                                       }
+                               } else {
+                                       /* int */
+                                       if (is64bit) {
+                                               *(user_addr_t *)data = argp;
+                                       } else {
+                                               *(uint32_t *)data = (uint32_t)argp;
+                                       }
+                               };
+                       } else if ((uap->cmd & IOC_OUT) && size) {
+                               /*
+                                * Zero the buffer so the user always
+                                * gets back something deterministic.
+                                */
+                               bzero(data, size);
+                       } else if (uap->cmd & IOC_VOID) {
+                               if (is64bit) {
+                                   *(user_addr_t *)data = argp;
+                               } else {
+                                   *(uint32_t *)data = (uint32_t)argp;
+                               }
+                       }
 
 
-               if ( (error = vnode_getwithref(vp)) == 0 ) {
-                       error = VNOP_IOCTL(vp, uap->cmd, CAST_DOWN(caddr_t, argp), 0, &context);
+                       error = VNOP_IOCTL(vp, uap->cmd, CAST_DOWN(caddr_t, data), 0, &context);
 
                        (void)vnode_put(vp);
 
                        (void)vnode_put(vp);
+
+                       /* Copy any output data to user */
+                       if (error == 0 && (uap->cmd & IOC_OUT) && size) 
+                               error = copyout(data, argp, size);
+                       if (memp)
+                               kfree(memp, size);
                }
                break;
                }
                break;
-       
        }
 
 outdrop:
        }
 
 outdrop:
@@ -935,46 +2527,122 @@ out:
        return(error);
 }
 
        return(error);
 }
 
+
 /*
 /*
- * Common code for dup, dup2, and fcntl(F_DUPFD).
+ * finishdup
+ *
+ * Description:        Common code for dup, dup2, and fcntl(F_DUPFD).
+ *
+ * Parameters: p                               Process performing the dup
+ *             old                             The fd to dup
+ *             new                             The fd to dup it to
+ *             fd_flags                        Flags to augment the new fd
+ *             retval                          Pointer to the call return area
+ *
+ * Returns:    0                               Success
+ *             EBADF
+ *             ENOMEM
+ *
+ * Implicit returns:
+ *             *retval (modified)              The new descriptor
+ *
+ * Locks:      Assumes proc_fdlock for process pointing to fdp is held by
+ *             the caller
+ *
+ * Notes:      This function may drop and reacquire this lock; it is unsafe
+ *             for a caller to assume that other state protected by the lock
+ *             has not been subsequently changed out from under it.
  */
 int
  */
 int
-finishdup(struct proc * p, struct filedesc *fdp, int old, int new, register_t *retval)
+finishdup(proc_t p,
+    struct filedesc *fdp, int old, int new, int fd_flags, int32_t *retval)
 {
        struct fileproc *nfp;
        struct fileproc *ofp;
 {
        struct fileproc *nfp;
        struct fileproc *ofp;
+#if CONFIG_MACF
+       int error;
+#endif
 
 
+#if DIAGNOSTIC
+       proc_fdlock_assert(p, LCK_MTX_ASSERT_OWNED);
+#endif
        if ((ofp = fdp->fd_ofiles[old]) == NULL ||
        if ((ofp = fdp->fd_ofiles[old]) == NULL ||
-                       (fdp->fd_ofileflags[old] & UF_RESERVED)) {
-               _fdrelse(fdp, new);
+           (fdp->fd_ofileflags[old] & UF_RESERVED)) {
+               fdrelse(p, new);
                return (EBADF);
        }
        fg_ref(ofp);
                return (EBADF);
        }
        fg_ref(ofp);
+
+#if CONFIG_MACF
+       error = mac_file_check_dup(proc_ucred(p), ofp->f_fglob, new);
+       if (error) {
+               fg_drop(ofp);
+               fdrelse(p, new);
+               return (error);
+       }
+#endif
+
        proc_fdunlock(p);
 
        proc_fdunlock(p);
 
-       MALLOC_ZONE(nfp, struct fileproc *, sizeof(struct fileproc), M_FILEPROC, M_WAITOK);
-       bzero(nfp, sizeof(struct fileproc));
+       nfp = fileproc_alloc_init(NULL);
 
        proc_fdlock(p);
 
        proc_fdlock(p);
-       nfp->f_flags = ofp->f_flags;
+
+       if (nfp == NULL) {
+               fg_drop(ofp);
+               fdrelse(p, new);
+               return (ENOMEM);
+       }
+
        nfp->f_fglob = ofp->f_fglob;
        nfp->f_fglob = ofp->f_fglob;
-       nfp->f_iocount = 0;
 
 
-       fdp->fd_ofiles[new] = nfp;
-       fdp->fd_ofileflags[new] = fdp->fd_ofileflags[old] &~ UF_EXCLOSE;
+#if DIAGNOSTIC
+       if (fdp->fd_ofiles[new] != 0)
+               panic("finishdup: overwriting fd_ofiles with new %d", new);
+       if ((fdp->fd_ofileflags[new] & UF_RESERVED) == 0)
+               panic("finishdup: unreserved fileflags with new %d", new);
+#endif
+
        if (new > fdp->fd_lastfile)
                fdp->fd_lastfile = new;
        if (new > fdp->fd_lastfile)
                fdp->fd_lastfile = new;
+       *fdflags(p, new) |= fd_flags;
+       procfdtbl_releasefd(p, new, nfp);
        *retval = new;
        return (0);
 }
 
 
        *retval = new;
        return (0);
 }
 
 
+/*
+ * close
+ *
+ * Description:        The implementation of the close(2) system call
+ *
+ * Parameters: p                       Process in whose per process file table
+ *                                     the close is to occur
+ *             uap->fd                 fd to be closed
+ *             retval                  <unused>
+ *
+ * Returns:    0                       Success
+ *     fp_lookup:EBADF                 Bad file descriptor
+ *      fp_guard_exception:???          Guarded file descriptor
+ *     close_internal:EBADF
+ *     close_internal:???              Anything returnable by a per-fileops
+ *                                     close function
+ */
+int
+close(proc_t p, struct close_args *uap, int32_t *retval)
+{
+       __pthread_testcancel(1);
+       return(close_nocancel(p, (struct close_nocancel_args *)uap, retval));
+}
+
+
 int
 int
-close(struct proc *p, struct close_args *uap, __unused register_t *retval)
+close_nocancel(proc_t p, struct close_nocancel_args *uap, __unused int32_t *retval)
 {
        struct fileproc *fp;
        int fd = uap->fd;
 {
        struct fileproc *fp;
        int fd = uap->fd;
-       int error =0;
+       int error;
 
        AUDIT_SYSCLOSE(p, fd);
 
 
        AUDIT_SYSCLOSE(p, fd);
 
@@ -985,50 +2653,70 @@ close(struct proc *p, struct close_args *uap, __unused register_t *retval)
                return(error);
        }
 
                return(error);
        }
 
-       error = close_internal(p, fd, fp, CLOSEINT_LOCKED | CLOSEINT_WAITONCLOSE);
+       if (FP_ISGUARDED(fp, GUARD_CLOSE)) {
+               error = fp_guard_exception(p, fd, fp, kGUARD_EXC_CLOSE);
+               (void) fp_drop(p, fd, fp, 1);
+               proc_fdunlock(p);
+               return (error);
+       }
+
+       error = close_internal_locked(p, fd, fp, 0);
 
        proc_fdunlock(p);
 
 
        proc_fdunlock(p);
 
-       return(error);
+       return (error);
 }
 
 
 /*
 }
 
 
 /*
+ * close_internal_locked
+ *
  * Close a file descriptor.
  * Close a file descriptor.
+ *
+ * Parameters: p                       Process in whose per process file table
+ *                                     the close is to occur
+ *             fd                      fd to be closed
+ *             fp                      fileproc associated with the fd
+ *
+ * Returns:    0                       Success
+ *             EBADF                   fd already in close wait state
+ *     closef_locked:???               Anything returnable by a per-fileops
+ *                                     close function
+ *
+ * Locks:      Assumes proc_fdlock for process is held by the caller and returns
+ *             with lock held
+ *
+ * Notes:      This function may drop and reacquire this lock; it is unsafe
+ *             for a caller to assume that other state protected by the lock
+ *             has not been subsequently changed out from under it.
  */
 int
  */
 int
-close_internal(struct proc *p, int fd, struct fileproc *fp, int flags)
+close_internal_locked(proc_t p, int fd, struct fileproc *fp, int flags)
 {
        struct filedesc *fdp = p->p_fd;
        int error =0;
 {
        struct filedesc *fdp = p->p_fd;
        int error =0;
-       int locked = flags & CLOSEINT_LOCKED;
-       int waitonclose = flags & CLOSEINT_WAITONCLOSE;
-       int norelse = flags & CLOSEINT_NOFDRELSE;
-       int nofdref = flags & CLOSEINT_NOFDNOREF;
-       int slpstate = PRIBIO;
+       int resvfd = flags & FD_DUP2RESV;
 
 
-       if (!locked)
-               proc_fdlock(p);
 
 
-       /* Keep people from using the filedesc while we are closing it */
-       fdp->fd_ofileflags[fd] |= UF_RESERVED;
+#if DIAGNOSTIC
+       proc_fdlock_assert(p, LCK_MTX_ASSERT_OWNED);
+#endif
 
 
-       fdp->fd_ofileflags[fd] |= UF_CLOSING;
+       /* Keep people from using the filedesc while we are closing it */
+       procfdtbl_markclosefd(p, fd);
 
 
 
 
-       if ((waitonclose && ((fp->f_flags & FP_CLOSING) == FP_CLOSING))) {
-                       if (nofdref == 0)
-                               fp_drop(p, fd, fp, 1);
-                       fp->f_flags |= FP_WAITCLOSE;
-                       if (!locked)
-                               slpstate |= PDROP;
-                       msleep(&fp->f_flags, &p->p_fdmlock, slpstate, "close wait",0) ; 
-                       return(EBADF);
+       if ((fp->f_flags & FP_CLOSING) == FP_CLOSING) {
+               panic("close_internal_locked: being called on already closing fd");
        }
 
        }
 
+
+#if DIAGNOSTIC
+       if ((fdp->fd_ofileflags[fd] & UF_RESERVED) == 0)
+               panic("close_internal: unreserved fileflags with fd %d", fd);
+#endif
+
        fp->f_flags |= FP_CLOSING;
        fp->f_flags |= FP_CLOSING;
-       if (nofdref)
-               fp->f_iocount++;
 
        if ( (fp->f_flags & FP_AIOISSUED) || kauth_authorize_fileop_has_listeners() ) {
 
 
        if ( (fp->f_flags & FP_AIOISSUED) || kauth_authorize_fileop_has_listeners() ) {
 
@@ -1063,65 +2751,117 @@ close_internal(struct proc *p, int fd, struct fileproc *fp, int flags)
        if (fp->f_flags & FP_WAITEVENT) 
                (void)waitevent_close(p, fp);
 
        if (fp->f_flags & FP_WAITEVENT) 
                (void)waitevent_close(p, fp);
 
-       if ((fp->f_flags & FP_INCHRREAD) == 0)
-               fileproc_drain(p, fp);
-       if (norelse == 0)
-               _fdrelse(fdp, fd);
+       fileproc_drain(p, fp);
+
+       if (resvfd == 0) {
+               _fdrelse(p, fd);
+       } else {
+               procfdtbl_reservefd(p, fd);
+       }
+
        error = closef_locked(fp, fp->f_fglob, p);
        if ((fp->f_flags & FP_WAITCLOSE) == FP_WAITCLOSE)
                wakeup(&fp->f_flags);
        fp->f_flags &= ~(FP_WAITCLOSE | FP_CLOSING);
 
        error = closef_locked(fp, fp->f_fglob, p);
        if ((fp->f_flags & FP_WAITCLOSE) == FP_WAITCLOSE)
                wakeup(&fp->f_flags);
        fp->f_flags &= ~(FP_WAITCLOSE | FP_CLOSING);
 
-       if (!locked)
-               proc_fdunlock(p);
+       proc_fdunlock(p);
+
+       fileproc_free(fp);      
+
+       proc_fdlock(p);
+
+#if DIAGNOSTIC
+       if (resvfd != 0) {
+               if ((fdp->fd_ofileflags[fd] & UF_RESERVED) == 0)
+                       panic("close with reserved fd returns with freed fd:%d: proc: %p", fd, p);
+       }
+#endif
 
 
-       FREE_ZONE(fp, sizeof *fp, M_FILEPROC);  
        return(error);
 }
 
        return(error);
 }
 
+
 /*
 /*
- * Return status information about a file descriptor.
+ * fstat1
+ *
+ * Description:        Return status information about a file descriptor.
+ *
+ * Parameters: p                               The process doing the fstat
+ *             fd                              The fd to stat
+ *             ub                              The user stat buffer
+ *             xsecurity                       The user extended security
+ *                                             buffer, or 0 if none
+ *             xsecurity_size                  The size of xsecurity, or 0
+ *                                             if no xsecurity
+ *             isstat64                        Flag to indicate 64 bit version
+ *                                             for inode size, etc.
+ *
+ * Returns:    0                               Success
+ *             EBADF
+ *             EFAULT
+ *     fp_lookup:EBADF                         Bad file descriptor
+ *     vnode_getwithref:???
+ *     copyout:EFAULT
+ *     vnode_getwithref:???
+ *     vn_stat:???
+ *     soo_stat:???
+ *     pipe_stat:???
+ *     pshm_stat:???
+ *     kqueue_stat:???
  *
  *
- * XXX switch on node type is bogus; need a stat in struct fileops instead.
+ * Notes:      Internal implementation for all other fstat() related
+ *             functions
+ *
+ *             XXX switch on node type is bogus; need a stat in struct
+ *             XXX fileops instead.
  */
 static int
  */
 static int
-fstat1(struct proc *p, int fd, user_addr_t ub, user_addr_t xsecurity, user_addr_t xsecurity_size)
+fstat1(proc_t p, int fd, user_addr_t ub, user_addr_t xsecurity, user_addr_t xsecurity_size, int isstat64)
 {
        struct fileproc *fp;
 {
        struct fileproc *fp;
-       struct stat sb;
-       struct user_stat user_sb;
+       union {
+               struct stat sb;
+               struct stat64 sb64;
+       } source;
+       union {
+               struct user64_stat user64_sb;
+               struct user32_stat user32_sb;
+               struct user64_stat64 user64_sb64;
+               struct user32_stat64 user32_sb64;
+       } dest;
        int error, my_size;
        int error, my_size;
-       int funnel_state;
-       short type;
+       file_type_t type;
        caddr_t data;
        kauth_filesec_t fsec;
        caddr_t data;
        kauth_filesec_t fsec;
-       ssize_t xsecurity_bufsize;
-       int entrycount;
-       struct vfs_context context;
+       user_size_t xsecurity_bufsize;
+       vfs_context_t ctx = vfs_context_current();
+       void * sbptr;
 
 
        AUDIT_ARG(fd, fd);
 
 
 
        AUDIT_ARG(fd, fd);
 
-       if ((error = fp_lookup(p, fd, &fp, 0)) != 0)
+       if ((error = fp_lookup(p, fd, &fp, 0)) != 0) {
                return(error);
                return(error);
+       }
        type = fp->f_type;
        data = fp->f_data;
        fsec = KAUTH_FILESEC_NONE;
 
        type = fp->f_type;
        data = fp->f_data;
        fsec = KAUTH_FILESEC_NONE;
 
+       sbptr = (void *)&source;
+
        switch (type) {
 
        case DTYPE_VNODE:
        switch (type) {
 
        case DTYPE_VNODE:
-               context.vc_proc = current_proc();
-               context.vc_ucred = kauth_cred_get();
                if ((error = vnode_getwithref((vnode_t)data)) == 0) {
                        /*
                if ((error = vnode_getwithref((vnode_t)data)) == 0) {
                        /*
-                        * If the caller has the file open, and is not requesting extended security,
-                        * we are going to let them get the basic stat information.
+                        * If the caller has the file open, and is not
+                        * requesting extended security information, we are
+                        * going to let them get the basic stat information.
                         */
                        if (xsecurity == USER_ADDR_NULL) {
                         */
                        if (xsecurity == USER_ADDR_NULL) {
-                               error = vn_stat_noauth((vnode_t)data, &sb, NULL, &context);
+                               error = vn_stat_noauth((vnode_t)data, sbptr, NULL, isstat64, ctx);
                        } else {
                        } else {
-                               error = vn_stat((vnode_t)data, &sb, &fsec, &context);
+                               error = vn_stat((vnode_t)data, sbptr, &fsec, isstat64, ctx);
                        }
 
                        AUDIT_ARG(vnpath, (struct vnode *)data, ARG_VNODE1);
                        }
 
                        AUDIT_ARG(vnpath, (struct vnode *)data, ARG_VNODE1);
@@ -1129,43 +2869,60 @@ fstat1(struct proc *p, int fd, user_addr_t ub, user_addr_t xsecurity, user_addr_
                }
                break;
 
                }
                break;
 
+#if SOCKETS
        case DTYPE_SOCKET:
        case DTYPE_SOCKET:
-               error = soo_stat((struct socket *)data, &sb);
+               error = soo_stat((struct socket *)data, sbptr, isstat64);
                break;
                break;
+#endif /* SOCKETS */
 
        case DTYPE_PIPE:
 
        case DTYPE_PIPE:
-               error = pipe_stat((void *)data, &sb);
+               error = pipe_stat((void *)data, sbptr, isstat64);
                break;
 
        case DTYPE_PSXSHM:
                break;
 
        case DTYPE_PSXSHM:
-               error = pshm_stat((void *)data, &sb);
+               error = pshm_stat((void *)data, sbptr, isstat64);
                break;
 
        case DTYPE_KQUEUE:
                break;
 
        case DTYPE_KQUEUE:
-               funnel_state = thread_funnel_set(kernel_flock, TRUE);
-               error = kqueue_stat(fp, &sb, p);
-               thread_funnel_set(kernel_flock, funnel_state);
+               error = kqueue_stat((void *)data, sbptr, isstat64, p);
                break;
 
        default:
                error = EBADF;
                goto out;
        }
                break;
 
        default:
                error = EBADF;
                goto out;
        }
-       /* Zap spare fields */
-       sb.st_lspare = 0;
-       sb.st_qspare[0] = 0LL;
-       sb.st_qspare[1] = 0LL;
        if (error == 0) {
                caddr_t sbp;
        if (error == 0) {
                caddr_t sbp;
-               if (IS_64BIT_PROCESS(current_proc())) {
-                       munge_stat(&sb, &user_sb); 
-                       my_size = sizeof(user_sb);
-                       sbp = (caddr_t)&user_sb;
-               }
-               else {
-                       my_size = sizeof(sb);
-                       sbp = (caddr_t)&sb;
+
+               if (isstat64 != 0) {
+                       source.sb64.st_lspare = 0;
+                       source.sb64.st_qspare[0] = 0LL;
+                       source.sb64.st_qspare[1] = 0LL;
+
+                       if (IS_64BIT_PROCESS(current_proc())) {
+                               munge_user64_stat64(&source.sb64, &dest.user64_sb64); 
+                               my_size = sizeof(dest.user64_sb64);
+                               sbp = (caddr_t)&dest.user64_sb64;
+                       } else {
+                               munge_user32_stat64(&source.sb64, &dest.user32_sb64); 
+                               my_size = sizeof(dest.user32_sb64);
+                               sbp = (caddr_t)&dest.user32_sb64;
+                       }
+               } else {
+                       source.sb.st_lspare = 0;
+                       source.sb.st_qspare[0] = 0LL;
+                       source.sb.st_qspare[1] = 0LL;
+                       if (IS_64BIT_PROCESS(current_proc())) {
+                               munge_user64_stat(&source.sb, &dest.user64_sb); 
+                               my_size = sizeof(dest.user64_sb);
+                               sbp = (caddr_t)&dest.user64_sb;
+                       } else {
+                               munge_user32_stat(&source.sb, &dest.user32_sb); 
+                               my_size = sizeof(dest.user32_sb);
+                               sbp = (caddr_t)&dest.user32_sb;
+                       }
                }
                }
+
                error = copyout(sbp, ub, my_size);
        }
 
                error = copyout(sbp, ub, my_size);
        }
 
@@ -1200,33 +2957,119 @@ out:
        return (error);
 }
 
        return (error);
 }
 
+
+/*
+ * fstat_extended
+ *
+ * Description:        Extended version of fstat supporting returning extended
+ *             security information
+ *
+ * Parameters: p                               The process doing the fstat
+ *             uap->fd                         The fd to stat
+ *             uap->ub                         The user stat buffer
+ *             uap->xsecurity                  The user extended security
+ *                                             buffer, or 0 if none
+ *             uap->xsecurity_size             The size of xsecurity, or 0
+ *
+ * Returns:    0                               Success
+ *             !0                              Errno (see fstat1)
+ */
+int
+fstat_extended(proc_t p, struct fstat_extended_args *uap, __unused int32_t *retval)
+{
+       return(fstat1(p, uap->fd, uap->ub, uap->xsecurity, uap->xsecurity_size, 0));
+}
+
+/*
+ * fstat
+ *
+ * Description:        Get file status for the file associated with fd
+ *
+ * Parameters: p                               The process doing the fstat
+ *             uap->fd                         The fd to stat
+ *             uap->ub                         The user stat buffer
+ *
+ * Returns:    0                               Success
+ *             !0                              Errno (see fstat1)
+ */
+int
+fstat(proc_t p, register struct fstat_args *uap, __unused int32_t *retval)
+{
+       return(fstat1(p, uap->fd, uap->ub, 0, 0, 0));
+}
+
+
+/*
+ * fstat64_extended
+ *
+ * Description:        Extended version of fstat64 supporting returning extended
+ *             security information
+ *
+ * Parameters: p                               The process doing the fstat
+ *             uap->fd                         The fd to stat
+ *             uap->ub                         The user stat buffer
+ *             uap->xsecurity                  The user extended security
+ *                                             buffer, or 0 if none
+ *             uap->xsecurity_size             The size of xsecurity, or 0
+ *
+ * Returns:    0                               Success
+ *             !0                              Errno (see fstat1)
+ */
 int
 int
-fstat_extended(struct proc *p, struct fstat_extended_args *uap, __unused register_t *retval)
+fstat64_extended(proc_t p, struct fstat64_extended_args *uap, __unused int32_t *retval)
 {
 {
-       return(fstat1(p, uap->fd, uap->ub, uap->xsecurity, uap->xsecurity_size));
+       return(fstat1(p, uap->fd, uap->ub, uap->xsecurity, uap->xsecurity_size, 1));
 }
  
 }
  
+
+/*
+ * fstat64
+ *
+ * Description:        Get 64 bit version of the file status for the file associated
+ *             with fd
+ *
+ * Parameters: p                               The process doing the fstat
+ *             uap->fd                         The fd to stat
+ *             uap->ub                         The user stat buffer
+ *
+ * Returns:    0                               Success
+ *             !0                              Errno (see fstat1)
+ */
 int
 int
-fstat(struct proc *p, register struct fstat_args *uap, __unused register_t *retval)
+fstat64(proc_t p, register struct fstat64_args *uap, __unused int32_t *retval)
 {
 {
-       return(fstat1(p, uap->fd, uap->ub, 0, 0));
+       return(fstat1(p, uap->fd, uap->ub, 0, 0, 1));
 }
 
 }
 
+
 /*
 /*
- * Return pathconf information about a file descriptor.
+ * fpathconf
+ *
+ * Description:        Return pathconf information about a file descriptor.
+ *
+ * Parameters: p                               Process making the request
+ *             uap->fd                         fd to get information about
+ *             uap->name                       Name of information desired
+ *             retval                          Pointer to the call return area
+ *
+ * Returns:    0                               Success
+ *             EINVAL
+ *     fp_lookup:EBADF                         Bad file descriptor
+ *     vnode_getwithref:???
+ *     vn_pathconf:???
+ *
+ * Implicit returns:
+ *             *retval (modified)              Returned information (numeric)
  */
 int
  */
 int
-fpathconf(p, uap, retval)
-       struct proc *p;
-       register struct fpathconf_args *uap;
-       register_t *retval;
+fpathconf(proc_t p, struct fpathconf_args *uap, int32_t *retval)
 {
        int fd = uap->fd;
        struct fileproc *fp;
        struct vnode *vp;
 {
        int fd = uap->fd;
        struct fileproc *fp;
        struct vnode *vp;
-       struct vfs_context context;
        int error = 0;
        int error = 0;
-       short type;
+       file_type_t type;
        caddr_t data;
 
 
        caddr_t data;
 
 
@@ -1239,8 +3082,8 @@ fpathconf(p, uap, retval)
        switch (type) {
 
        case DTYPE_SOCKET:
        switch (type) {
 
        case DTYPE_SOCKET:
-               if (uap->name != _PC_PIPE_BUF) {
-                       error = EINVAL;
+               if (uap->name != _PC_PIPE_BUF) {
+                       error = EINVAL;
                        goto out;
                }
                *retval = PIPE_BUF;
                        goto out;
                }
                *retval = PIPE_BUF;
@@ -1248,7 +3091,11 @@ fpathconf(p, uap, retval)
                goto out;
 
        case DTYPE_PIPE:
                goto out;
 
        case DTYPE_PIPE:
-               *retval = PIPE_BUF;
+               if (uap->name != _PC_PIPE_BUF) {
+                       error = EINVAL;
+                       goto out;
+               }
+               *retval = PIPE_BUF;
                error = 0;
                goto out;
 
                error = 0;
                goto out;
 
@@ -1258,22 +3105,16 @@ fpathconf(p, uap, retval)
                if ( (error = vnode_getwithref(vp)) == 0) {
                        AUDIT_ARG(vnpath, vp, ARG_VNODE1);
 
                if ( (error = vnode_getwithref(vp)) == 0) {
                        AUDIT_ARG(vnpath, vp, ARG_VNODE1);
 
-                       context.vc_proc = p;
-                       context.vc_ucred = kauth_cred_get();
-
-                       error = vn_pathconf(vp, uap->name, retval, &context);
+                       error = vn_pathconf(vp, uap->name, retval, vfs_context_current());
 
                        (void)vnode_put(vp);
                }
                goto out;
 
 
                        (void)vnode_put(vp);
                }
                goto out;
 
-       case DTYPE_PSXSHM:
-       case DTYPE_KQUEUE:
+       default:
                error = EINVAL;
                goto out;
 
                error = EINVAL;
                goto out;
 
-       default:
-               panic("fpathconf (unrecognized - %d)", type);
        }
        /*NOTREACHED*/
 out:
        }
        /*NOTREACHED*/
 out:
@@ -1282,37 +3123,56 @@ out:
 }
 
 /*
 }
 
 /*
- * Allocate a file descriptor for the process.
+ * Statistics counter for the number of times a process calling fdalloc()
+ * has resulted in an expansion of the per process open file table.
+ *
+ * XXX This would likely be of more use if it were per process
  */
 int fdexpand;
 
  */
 int fdexpand;
 
+
+/*
+ * fdalloc
+ *
+ * Description:        Allocate a file descriptor for the process.
+ *
+ * Parameters: p                               Process to allocate the fd in
+ *             want                            The fd we would prefer to get
+ *             result                          Pointer to fd we got
+ *
+ * Returns:    0                               Success
+ *             EMFILE
+ *             ENOMEM
+ *
+ * Implicit returns:
+ *             *result (modified)              The fd which was allocated
+ */
 int
 int
-fdalloc(p, want, result)
-       struct proc *p;
-       int want;
-       int *result;
+fdalloc(proc_t p, int want, int *result)
 {
 {
-       register struct filedesc *fdp = p->p_fd;
-       register int i;
+       struct filedesc *fdp = p->p_fd;
+       int i;
        int lim, last, numfiles, oldnfiles;
        struct fileproc **newofiles, **ofiles;
        int lim, last, numfiles, oldnfiles;
        struct fileproc **newofiles, **ofiles;
-       char *newofileflags, *ofileflags;
+       char *newofileflags;
 
        /*
         * Search for a free descriptor starting at the higher
         * of want or fd_freefile.  If that fails, consider
         * expanding the ofile array.
         */
 
        /*
         * Search for a free descriptor starting at the higher
         * of want or fd_freefile.  If that fails, consider
         * expanding the ofile array.
         */
+#if DIAGNOSTIC
+       proc_fdlock_assert(p, LCK_MTX_ASSERT_OWNED);
+#endif
+
        lim = min((int)p->p_rlimit[RLIMIT_NOFILE].rlim_cur, maxfiles);
        for (;;) {
                last = min(fdp->fd_nfiles, lim);
                if ((i = want) < fdp->fd_freefile)
                        i = fdp->fd_freefile;
        lim = min((int)p->p_rlimit[RLIMIT_NOFILE].rlim_cur, maxfiles);
        for (;;) {
                last = min(fdp->fd_nfiles, lim);
                if ((i = want) < fdp->fd_freefile)
                        i = fdp->fd_freefile;
-               ofiles = &fdp->fd_ofiles[i];
-               ofileflags = &fdp->fd_ofileflags[i];
                for (; i < last; i++) {
                for (; i < last; i++) {
-                       if (*ofiles == NULL && !(*ofileflags & UF_RESERVED)) {
-                               *ofileflags = UF_RESERVED;
+                       if (fdp->fd_ofiles[i] == NULL && !(fdp->fd_ofileflags[i] & UF_RESERVED)) {
+                               procfdtbl_reservefd(p, i);
                                if (i > fdp->fd_lastfile)
                                        fdp->fd_lastfile = i;
                                if (want <= fdp->fd_freefile)
                                if (i > fdp->fd_lastfile)
                                        fdp->fd_lastfile = i;
                                if (want <= fdp->fd_freefile)
@@ -1320,7 +3180,6 @@ fdalloc(p, want, result)
                                *result = i;
                                return (0);
                        }
                                *result = i;
                                return (0);
                        }
-                       ofiles++; ofileflags++;
                }
 
                /*
                }
 
                /*
@@ -1353,15 +3212,15 @@ fdalloc(p, want, result)
                 */
                oldnfiles = fdp->fd_nfiles;
                (void) memcpy(newofiles, fdp->fd_ofiles,
                 */
                oldnfiles = fdp->fd_nfiles;
                (void) memcpy(newofiles, fdp->fd_ofiles,
-                               oldnfiles * sizeof *fdp->fd_ofiles);
+                               oldnfiles * sizeof(*fdp->fd_ofiles));
                (void) memset(&newofiles[oldnfiles], 0,
                (void) memset(&newofiles[oldnfiles], 0,
-                               (numfiles - oldnfiles) * sizeof *fdp->fd_ofiles);
+                               (numfiles - oldnfiles) * sizeof(*fdp->fd_ofiles));
 
                (void) memcpy(newofileflags, fdp->fd_ofileflags,
 
                (void) memcpy(newofileflags, fdp->fd_ofileflags,
-                               oldnfiles * sizeof *fdp->fd_ofileflags);
+                               oldnfiles * sizeof(*fdp->fd_ofileflags));
                (void) memset(&newofileflags[oldnfiles], 0,
                                (numfiles - oldnfiles) *
                (void) memset(&newofileflags[oldnfiles], 0,
                                (numfiles - oldnfiles) *
-                                               sizeof *fdp->fd_ofileflags);
+                                               sizeof(*fdp->fd_ofileflags));
                ofiles = fdp->fd_ofiles;
                fdp->fd_ofiles = newofiles;
                fdp->fd_ofileflags = newofileflags;
                ofiles = fdp->fd_ofiles;
                fdp->fd_ofiles = newofiles;
                fdp->fd_ofileflags = newofileflags;
@@ -1371,14 +3230,26 @@ fdalloc(p, want, result)
        }
 }
 
        }
 }
 
+
 /*
 /*
- * Check to see whether n user file descriptors
- * are available to the process p.
+ * fdavail
+ *
+ * Description:        Check to see whether n user file descriptors are available
+ *             to the process p.
+ *
+ * Parameters: p                               Process to check in
+ *             n                               The number of fd's desired
+ *
+ * Returns:    0                               No
+ *             1                               Yes
+ *
+ * Locks:      Assumes proc_fdlock for process is held by the caller
+ *
+ * Notes:      The answer only remains valid so long as the proc_fdlock is
+ *             held by the caller.
  */
 int
  */
 int
-fdavail(p, n)
-       struct proc *p;
-       int n;
+fdavail(proc_t p, int n)
 {
        struct filedesc *fdp = p->p_fd;
        struct fileproc **fpp;
 {
        struct filedesc *fdp = p->p_fd;
        struct fileproc **fpp;
@@ -1396,45 +3267,50 @@ fdavail(p, n)
        return (0);
 }
 
        return (0);
 }
 
-void
-fdrelse(p, fd)
-       struct proc *p;
-       int fd;
-{
-       _fdrelse(p->p_fd, fd);
-}
 
 
+/*
+ * fdrelse
+ *
+ * Description:        Legacy KPI wrapper function for _fdrelse
+ *
+ * Parameters: p                               Process in which fd lives
+ *             fd                              fd to free
+ *
+ * Returns:    void
+ *
+ * Locks:      Assumes proc_fdlock for process is held by the caller
+ */
 void
 void
-fddrop(p, fd)
-       struct proc *p;
-       int fd;
+fdrelse(proc_t p, int fd)
 {
 {
-       struct filedesc *fdp = p->p_fd;
-       struct fileproc *fp;
-
-       if (fd < fdp->fd_freefile)
-               fdp->fd_freefile = fd;
-#if DIAGNOSTIC
-       if (fd > fdp->fd_lastfile)
-               panic("fdrelse: fd_lastfile inconsistent");
-#endif
-       fp = fdp->fd_ofiles[fd];
-       fdp->fd_ofiles[fd] = NULL;
-       fdp->fd_ofileflags[fd] = 0;
-
-       while ((fd = fdp->fd_lastfile) > 0 &&
-                       fdp->fd_ofiles[fd] == NULL &&
-                       !(fdp->fd_ofileflags[fd] & UF_RESERVED))
-               fdp->fd_lastfile--;
-       FREE_ZONE(fp, sizeof *fp, M_FILEPROC);
+       _fdrelse(p, fd);
 }
 
 
 }
 
 
+/*
+ * fdgetf_noref
+ *
+ * Description:        Get the fileproc pointer for the given fd from the per process
+ *             open file table without taking an explicit reference on it.
+ *
+ * Parameters: p                               Process containing fd
+ *             fd                              fd to obtain fileproc for
+ *             resultfp                        Pointer to pointer return area
+ *
+ * Returns:    0                               Success
+ *             EBADF
+ *
+ * Implicit returns:
+ *             *resultfp (modified)            Pointer to fileproc pointer
+ *
+ * Locks:      Assumes proc_fdlock for process is held by the caller
+ *
+ * Notes:      Because there is no reference explicitly taken, the returned
+ *             fileproc pointer is only valid so long as the proc_fdlock
+ *             remains held by the caller.
+ */
 int
 int
-fdgetf_noref(p, fd, resultfp)
-       struct proc *p;
-       int fd;
-       struct fileproc **resultfp;
+fdgetf_noref(proc_t p, int fd, struct fileproc **resultfp)
 {
        struct filedesc *fdp = p->p_fd;
        struct fileproc *fp;
 {
        struct filedesc *fdp = p->p_fd;
        struct fileproc *fp;
@@ -1450,36 +3326,99 @@ fdgetf_noref(p, fd, resultfp)
 }
 
 
 }
 
 
-/* should be called only when proc_fdlock is held */
-void
-fp_setflags(proc_t p, struct fileproc * fp, int flags)
+/*
+ * fp_getfvp
+ *
+ * Description:        Get fileproc and vnode pointer for a given fd from the per
+ *             process open file table of the specified process, and if
+ *             successful, increment the f_iocount
+ *
+ * Parameters: p                               Process in which fd lives
+ *             fd                              fd to get information for
+ *             resultfp                        Pointer to result fileproc
+ *                                             pointer area, or 0 if none
+ *             resultvp                        Pointer to result vnode pointer
+ *                                             area, or 0 if none
+ *
+ * Returns:    0                               Success
+ *             EBADF                           Bad file descriptor
+ *             ENOTSUP                         fd does not refer to a vnode
+ *
+ * Implicit returns:
+ *             *resultfp (modified)            Fileproc pointer
+ *             *resultvp (modified)            vnode pointer
+ *
+ * Notes:      The resultfp and resultvp fields are optional, and may be
+ *             independently specified as NULL to skip returning information
+ *
+ * Locks:      Internally takes and releases proc_fdlock
+ */
+int
+fp_getfvp(proc_t p, int fd, struct fileproc **resultfp, struct vnode **resultvp)
 {
 {
-       proc_fdlock(p);
-       fp->f_flags |= flags;
-       proc_fdunlock(p);
-}
+       struct filedesc *fdp = p->p_fd;
+       struct fileproc *fp;
 
 
-void
-fp_clearflags(proc_t p, struct fileproc * fp, int flags)
-{
+       proc_fdlock_spin(p);
+       if (fd < 0 || fd >= fdp->fd_nfiles ||
+                       (fp = fdp->fd_ofiles[fd]) == NULL ||
+                       (fdp->fd_ofileflags[fd] & UF_RESERVED)) {
+               proc_fdunlock(p);
+               return (EBADF);
+       }
+       if (fp->f_type != DTYPE_VNODE) {
+               proc_fdunlock(p);
+               return(ENOTSUP);
+       }
+       fp->f_iocount++;
 
 
-       proc_fdlock(p);
-       if (fp)
-               fp->f_flags &= ~flags;
+       if (resultfp)
+               *resultfp = fp;
+       if (resultvp)
+               *resultvp = (struct vnode *)fp->f_data;
        proc_fdunlock(p);
        proc_fdunlock(p);
+
+       return (0);
 }
 
 }
 
+
+/*
+ * fp_getfvpandvid
+ *
+ * Description:        Get fileproc, vnode pointer, and vid for a given fd from the
+ *             per process open file table of the specified process, and if
+ *             successful, increment the f_iocount
+ *
+ * Parameters: p                               Process in which fd lives
+ *             fd                              fd to get information for
+ *             resultfp                        Pointer to result fileproc
+ *                                             pointer area, or 0 if none
+ *             resultvp                        Pointer to result vnode pointer
+ *                                             area, or 0 if none
+ *             vidp                            Pointer to resuld vid area
+ *
+ * Returns:    0                               Success
+ *             EBADF                           Bad file descriptor
+ *             ENOTSUP                         fd does not refer to a vnode
+ *
+ * Implicit returns:
+ *             *resultfp (modified)            Fileproc pointer
+ *             *resultvp (modified)            vnode pointer
+ *             *vidp                           vid value
+ *
+ * Notes:      The resultfp and resultvp fields are optional, and may be
+ *             independently specified as NULL to skip returning information
+ *
+ * Locks:      Internally takes and releases proc_fdlock
+ */
 int
 int
-fp_getfvp(p, fd, resultfp, resultvp)
-       struct proc *p;
-       int fd;
-       struct fileproc **resultfp;
-       struct vnode  **resultvp;
+fp_getfvpandvid(proc_t p, int fd, struct fileproc **resultfp,
+               struct vnode **resultvp, uint32_t *vidp)
 {
        struct filedesc *fdp = p->p_fd;
        struct fileproc *fp;
 
 {
        struct filedesc *fdp = p->p_fd;
        struct fileproc *fp;
 
-       proc_fdlock(p);
+       proc_fdlock_spin(p);
        if (fd < 0 || fd >= fdp->fd_nfiles ||
                        (fp = fdp->fd_ofiles[fd]) == NULL ||
                        (fdp->fd_ofileflags[fd] & UF_RESERVED)) {
        if (fd < 0 || fd >= fdp->fd_nfiles ||
                        (fp = fdp->fd_ofiles[fd]) == NULL ||
                        (fdp->fd_ofileflags[fd] & UF_RESERVED)) {
@@ -1496,6 +3435,8 @@ fp_getfvp(p, fd, resultfp, resultvp)
                *resultfp = fp;
        if (resultvp)
                *resultvp = (struct vnode *)fp->f_data;
                *resultfp = fp;
        if (resultvp)
                *resultvp = (struct vnode *)fp->f_data;
+       if (vidp)
+               *vidp = (uint32_t)vnode_vid((struct vnode *)fp->f_data);
        proc_fdunlock(p);
 
        return (0);
        proc_fdunlock(p);
 
        return (0);
@@ -1503,24 +3444,38 @@ fp_getfvp(p, fd, resultfp, resultvp)
 
 
 /*
 
 
 /*
+ * fp_getfsock
+ *
+ * Description:        Get fileproc and socket pointer for a given fd from the
+ *             per process open file table of the specified process, and if
+ *             successful, increment the f_iocount
+ *
+ * Parameters: p                               Process in which fd lives
+ *             fd                              fd to get information for
+ *             resultfp                        Pointer to result fileproc
+ *                                             pointer area, or 0 if none
+ *             results                         Pointer to result socket
+ *                                             pointer area, or 0 if none
+ *
  * Returns:    EBADF                   The file descriptor is invalid
  *             EOPNOTSUPP              The file descriptor is not a socket
  *             0                       Success
  *
  * Returns:    EBADF                   The file descriptor is invalid
  *             EOPNOTSUPP              The file descriptor is not a socket
  *             0                       Success
  *
+ * Implicit returns:
+ *             *resultfp (modified)            Fileproc pointer
+ *             *results (modified)             socket pointer
+ *
  * Notes:      EOPNOTSUPP should probably be ENOTSOCK; this function is only
  *             ever called from accept1().
  */
 int
  * Notes:      EOPNOTSUPP should probably be ENOTSOCK; this function is only
  *             ever called from accept1().
  */
 int
-fp_getfsock(p, fd, resultfp, results)
-       struct proc *p;
-       int fd;
-       struct fileproc **resultfp;
-       struct socket  **results;
+fp_getfsock(proc_t p, int fd, struct fileproc **resultfp,
+           struct socket **results)
 {
        struct filedesc *fdp = p->p_fd;
        struct fileproc *fp;
 
 {
        struct filedesc *fdp = p->p_fd;
        struct fileproc *fp;
 
-       proc_fdlock(p);
+       proc_fdlock_spin(p);
        if (fd < 0 || fd >= fdp->fd_nfiles ||
                        (fp = fdp->fd_ofiles[fd]) == NULL ||
                        (fdp->fd_ofileflags[fd] & UF_RESERVED)) {
        if (fd < 0 || fd >= fdp->fd_nfiles ||
                        (fp = fdp->fd_ofiles[fd]) == NULL ||
                        (fdp->fd_ofileflags[fd] & UF_RESERVED)) {
@@ -1543,17 +3498,39 @@ fp_getfsock(p, fd, resultfp, results)
 }
 
 
 }
 
 
+/*
+ * fp_getfkq
+ *
+ * Description:        Get fileproc and kqueue pointer for a given fd from the
+ *             per process open file table of the specified process, and if
+ *             successful, increment the f_iocount
+ *
+ * Parameters: p                               Process in which fd lives
+ *             fd                              fd to get information for
+ *             resultfp                        Pointer to result fileproc
+ *                                             pointer area, or 0 if none
+ *             resultkq                        Pointer to result kqueue
+ *                                             pointer area, or 0 if none
+ *
+ * Returns:    EBADF                   The file descriptor is invalid
+ *             EBADF                   The file descriptor is not a socket
+ *             0                       Success
+ *
+ * Implicit returns:
+ *             *resultfp (modified)            Fileproc pointer
+ *             *resultkq (modified)            kqueue pointer
+ *
+ * Notes:      The second EBADF should probably be something else to make
+ *             the error condition distinct.
+ */
 int
 int
-fp_getfkq(p, fd, resultfp, resultkq)
-       struct proc *p;
-       int fd;
-       struct fileproc **resultfp;
-       struct kqueue  **resultkq;
+fp_getfkq(proc_t p, int fd, struct fileproc **resultfp,
+         struct kqueue **resultkq)
 {
        struct filedesc *fdp = p->p_fd;
        struct fileproc *fp;
 
 {
        struct filedesc *fdp = p->p_fd;
        struct fileproc *fp;
 
-       proc_fdlock(p);
+       proc_fdlock_spin(p);
        if ( fd < 0 || fd >= fdp->fd_nfiles ||
                        (fp = fdp->fd_ofiles[fd]) == NULL ||
                        (fdp->fd_ofileflags[fd] & UF_RESERVED)) {
        if ( fd < 0 || fd >= fdp->fd_nfiles ||
                        (fp = fdp->fd_ofiles[fd]) == NULL ||
                        (fdp->fd_ofileflags[fd] & UF_RESERVED)) {
@@ -1575,17 +3552,42 @@ fp_getfkq(p, fd, resultfp, resultkq)
        return (0);
 }
 
        return (0);
 }
 
+
+/*
+ * fp_getfpshm
+ *
+ * Description:        Get fileproc and POSIX shared memory pointer for a given fd
+ *             from the per process open file table of the specified process
+ *             and if successful, increment the f_iocount
+ *
+ * Parameters: p                               Process in which fd lives
+ *             fd                              fd to get information for
+ *             resultfp                        Pointer to result fileproc
+ *                                             pointer area, or 0 if none
+ *             resultpshm                      Pointer to result POSIX
+ *                                             shared memory pointer
+ *                                             pointer area, or 0 if none
+ *
+ * Returns:    EBADF                   The file descriptor is invalid
+ *             EBADF                   The file descriptor is not a POSIX
+ *                                     shared memory area
+ *             0                       Success
+ *
+ * Implicit returns:
+ *             *resultfp (modified)            Fileproc pointer
+ *             *resultpshm (modified)          POSIX shared memory pointer
+ *
+ * Notes:      The second EBADF should probably be something else to make
+ *             the error condition distinct.
+ */
 int
 int
-fp_getfpshm(p, fd, resultfp, resultpshm)
-       struct proc *p;
-       int fd;
-       struct fileproc **resultfp;
-       struct pshmnode  **resultpshm;
+fp_getfpshm(proc_t p, int fd, struct fileproc **resultfp,
+           struct pshmnode **resultpshm)
 {
        struct filedesc *fdp = p->p_fd;
        struct fileproc *fp;
 
 {
        struct filedesc *fdp = p->p_fd;
        struct fileproc *fp;
 
-       proc_fdlock(p);
+       proc_fdlock_spin(p);
        if (fd < 0 || fd >= fdp->fd_nfiles ||
                        (fp = fdp->fd_ofiles[fd]) == NULL ||
                        (fdp->fd_ofileflags[fd] & UF_RESERVED)) {
        if (fd < 0 || fd >= fdp->fd_nfiles ||
                        (fp = fdp->fd_ofiles[fd]) == NULL ||
                        (fdp->fd_ofileflags[fd] & UF_RESERVED)) {
@@ -1609,17 +3611,50 @@ fp_getfpshm(p, fd, resultfp, resultpshm)
 }
 
 
 }
 
 
+/*
+ * fp_getfsem
+ *
+ * Description:        Get fileproc and POSIX semaphore pointer for a given fd from
+ *             the per process open file table of the specified process
+ *             and if successful, increment the f_iocount
+ *
+ * Parameters: p                               Process in which fd lives
+ *             fd                              fd to get information for
+ *             resultfp                        Pointer to result fileproc
+ *                                             pointer area, or 0 if none
+ *             resultpsem                      Pointer to result POSIX
+ *                                             semaphore pointer area, or
+ *                                             0 if none
+ *
+ * Returns:    EBADF                   The file descriptor is invalid
+ *             EBADF                   The file descriptor is not a POSIX
+ *                                     semaphore
+ *             0                       Success
+ *
+ * Implicit returns:
+ *             *resultfp (modified)            Fileproc pointer
+ *             *resultpsem (modified)          POSIX semaphore pointer
+ *
+ * Notes:      The second EBADF should probably be something else to make
+ *             the error condition distinct.
+ *
+ *             In order to support unnamed POSIX semaphores, the named
+ *             POSIX semaphores will have to move out of the per-process
+ *             open filetable, and into a global table that is shared with
+ *             unnamed POSIX semaphores, since unnamed POSIX semaphores
+ *             are typically used by declaring instances in shared memory,
+ *             and there's no other way to do this without changing the
+ *             underlying type, which would introduce binary compatibility
+ *             issues.
+ */
 int
 int
-fp_getfpsem(p, fd, resultfp, resultpsem)
-       struct proc *p;
-       int fd;
-       struct fileproc **resultfp;
-       struct psemnode  **resultpsem;
+fp_getfpsem(proc_t p, int fd, struct fileproc **resultfp,
+           struct psemnode **resultpsem)
 {
        struct filedesc *fdp = p->p_fd;
        struct fileproc *fp;
 
 {
        struct filedesc *fdp = p->p_fd;
        struct fileproc *fp;
 
-       proc_fdlock(p);
+       proc_fdlock_spin(p);
        if (fd < 0 || fd >= fdp->fd_nfiles ||
                        (fp = fdp->fd_ofiles[fd]) == NULL ||
                        (fdp->fd_ofileflags[fd] & UF_RESERVED)) {
        if (fd < 0 || fd >= fdp->fd_nfiles ||
                        (fp = fdp->fd_ofiles[fd]) == NULL ||
                        (fdp->fd_ofileflags[fd] & UF_RESERVED)) {
@@ -1640,21 +3675,97 @@ fp_getfpsem(p, fd, resultfp, resultpsem)
 
        return (0);
 }
 
        return (0);
 }
+
+
+/*
+ * fp_getfpipe
+ *
+ * Description:        Get fileproc and pipe pointer for a given fd from the
+ *             per process open file table of the specified process
+ *             and if successful, increment the f_iocount
+ *
+ * Parameters: p                               Process in which fd lives
+ *             fd                              fd to get information for
+ *             resultfp                        Pointer to result fileproc
+ *                                             pointer area, or 0 if none
+ *             resultpipe                      Pointer to result pipe
+ *                                             pointer area, or 0 if none
+ *
+ * Returns:    EBADF                   The file descriptor is invalid
+ *             EBADF                   The file descriptor is not a socket
+ *             0                       Success
+ *
+ * Implicit returns:
+ *             *resultfp (modified)            Fileproc pointer
+ *             *resultpipe (modified)          pipe pointer
+ *
+ * Notes:      The second EBADF should probably be something else to make
+ *             the error condition distinct.
+ */
 int
 int
-fp_lookup(p, fd, resultfp, locked)
-       struct proc *p;
-       int fd;
-       struct fileproc **resultfp;
-       int locked;
+fp_getfpipe(proc_t p, int fd, struct fileproc **resultfp,
+           struct pipe **resultpipe)
 {
        struct filedesc *fdp = p->p_fd;
        struct fileproc *fp;
 
 {
        struct filedesc *fdp = p->p_fd;
        struct fileproc *fp;
 
-       if (!locked)
-               proc_fdlock(p);
+       proc_fdlock_spin(p);
        if (fd < 0 || fd >= fdp->fd_nfiles ||
                        (fp = fdp->fd_ofiles[fd]) == NULL ||
                        (fdp->fd_ofileflags[fd] & UF_RESERVED)) {
        if (fd < 0 || fd >= fdp->fd_nfiles ||
                        (fp = fdp->fd_ofiles[fd]) == NULL ||
                        (fdp->fd_ofileflags[fd] & UF_RESERVED)) {
+               proc_fdunlock(p);
+               return (EBADF);
+       }
+       if (fp->f_type != DTYPE_PIPE) {
+               proc_fdunlock(p);
+               return(EBADF);
+       }
+       fp->f_iocount++;
+
+       if (resultfp)
+               *resultfp = fp;
+       if (resultpipe)
+               *resultpipe = (struct pipe *)fp->f_data;
+       proc_fdunlock(p);
+
+       return (0);
+}
+
+/*
+ * fp_lookup
+ *
+ * Description:        Get fileproc pointer for a given fd from the per process
+ *             open file table of the specified process and if successful,
+ *             increment the f_iocount
+ *
+ * Parameters: p                               Process in which fd lives
+ *             fd                              fd to get information for
+ *             resultfp                        Pointer to result fileproc
+ *                                             pointer area, or 0 if none
+ *             locked                          !0 if the caller holds the
+ *                                             proc_fdlock, 0 otherwise
+ *
+ * Returns:    0                       Success
+ *             EBADF                   Bad file descriptor
+ *
+ * Implicit returns:
+ *             *resultfp (modified)            Fileproc pointer
+ *
+ * Locks:      If the argument 'locked' is non-zero, then the caller is
+ *             expected to have taken and held the proc_fdlock; if it is
+ *             zero, than this routine internally takes and drops this lock.
+ */
+int
+fp_lookup(proc_t p, int fd, struct fileproc **resultfp, int locked)
+{
+       struct filedesc *fdp = p->p_fd;
+       struct fileproc *fp;
+
+       if (!locked)
+               proc_fdlock_spin(p);
+       if (fd < 0 || fdp == NULL || fd >= fdp->fd_nfiles ||
+                       (fp = fdp->fd_ofiles[fd]) == NULL ||
+                       (fdp->fd_ofileflags[fd] & UF_RESERVED)) {
                if (!locked)
                        proc_fdunlock(p);
                return (EBADF);
                if (!locked)
                        proc_fdunlock(p);
                return (EBADF);
@@ -1669,12 +3780,98 @@ fp_lookup(p, fd, resultfp, locked)
        return (0);
 }
 
        return (0);
 }
 
+
+/*
+ * fp_tryswap
+ * 
+ * Description: Swap the fileproc pointer for a given fd with a new
+ *             fileproc pointer in the per-process open file table of
+ *             the specified process.  The fdlock must be held at entry.
+ *
+ * Parameters:  p              Process containing the fd
+ *             fd              The fd of interest
+ *             nfp             Pointer to the newfp
+ *
+ * Returns:    0               Success
+ *             EBADF           Bad file descriptor
+ *             EINTR           Interrupted
+ *             EKEEPLOOKING    f_iocount changed while lock was dropped.
+ */
+int
+fp_tryswap(proc_t p, int fd, struct fileproc *nfp)
+{
+       struct fileproc *fp;
+       int error;
+
+       proc_fdlock_assert(p, LCK_MTX_ASSERT_OWNED);
+
+       if (0 != (error = fp_lookup(p, fd, &fp, 1)))
+               return (error);
+       /*
+        * At this point, our caller (change_guardedfd_np) has
+        * one f_iocount reference, and we just took another
+        * one to begin the replacement.
+        */
+       if (fp->f_iocount < 2) {
+               panic("f_iocount too small %d", fp->f_iocount);
+       } else if (2 == fp->f_iocount) {
+
+               /* Copy the contents of *fp, preserving the "type" of *nfp */
+
+               nfp->f_flags = (nfp->f_flags & FP_TYPEMASK) |
+                       (fp->f_flags & ~FP_TYPEMASK);
+               nfp->f_iocount = fp->f_iocount;
+               nfp->f_fglob = fp->f_fglob;
+               nfp->f_waddr = fp->f_waddr;
+
+               p->p_fd->fd_ofiles[fd] = nfp;
+               (void) fp_drop(p, fd, nfp, 1);
+       } else {
+               /*
+                * Wait for all other active references to evaporate.
+                */
+               p->p_fpdrainwait = 1;
+               error = msleep(&p->p_fpdrainwait, &p->p_fdmlock,
+                   PRIBIO | PCATCH, "tryswap fpdrain", NULL);
+               if (0 == error) {
+                       /*
+                        * Return an "internal" errno to trigger a full
+                        * reevaluation of the change-guard attempt.
+                        */
+                       error = EKEEPLOOKING;
+                       printf("%s: lookup collision fd %d\n", __func__, fd);
+               }
+               (void) fp_drop(p, fd, fp, 1);
+       }
+       return (error);
+}
+
+
+/*
+ * fp_drop_written
+ *
+ * Description:        Set the FP_WRITTEN flag on the fileproc and drop the I/O
+ *             reference previously taken by calling fp_lookup et. al.
+ *
+ * Parameters: p                               Process in which the fd lives
+ *             fd                              fd associated with the fileproc
+ *             fp                              fileproc on which to set the
+ *                                             flag and drop the reference
+ *
+ * Returns:    0                               Success
+ *     fp_drop:EBADF                           Bad file descriptor
+ *
+ * Locks:      This function internally takes and drops the proc_fdlock for
+ *             the supplied process
+ *
+ * Notes:      The fileproc must correspond to the fd in the supplied proc
+ */
 int
 fp_drop_written(proc_t p, int fd, struct fileproc *fp)
 {
         int error;
 
 int
 fp_drop_written(proc_t p, int fd, struct fileproc *fp)
 {
         int error;
 
-       proc_fdlock(p);
+       proc_fdlock_spin(p);
 
        fp->f_flags |= FP_WRITTEN;
        
 
        fp->f_flags |= FP_WRITTEN;
        
@@ -1686,12 +3883,31 @@ fp_drop_written(proc_t p, int fd, struct fileproc *fp)
 }
 
 
 }
 
 
+/*
+ * fp_drop_event
+ *
+ * Description:        Set the FP_WAITEVENT flag on the fileproc and drop the I/O
+ *             reference previously taken by calling fp_lookup et. al.
+ *
+ * Parameters: p                               Process in which the fd lives
+ *             fd                              fd associated with the fileproc
+ *             fp                              fileproc on which to set the
+ *                                             flag and drop the reference
+ *
+ * Returns:    0                               Success
+ *     fp_drop:EBADF                           Bad file descriptor
+ *
+ * Locks:      This function internally takes and drops the proc_fdlock for
+ *             the supplied process
+ *
+ * Notes:      The fileproc must correspond to the fd in the supplied proc
+ */
 int
 fp_drop_event(proc_t p, int fd, struct fileproc *fp)
 {
         int error;
 
 int
 fp_drop_event(proc_t p, int fd, struct fileproc *fp)
 {
         int error;
 
-       proc_fdlock(p);
+       proc_fdlock_spin(p);
 
        fp->f_flags |= FP_WAITEVENT;
        
 
        fp->f_flags |= FP_WAITEVENT;
        
@@ -1702,17 +3918,38 @@ fp_drop_event(proc_t p, int fd, struct fileproc *fp)
        return (error);
 }
 
        return (error);
 }
 
+
+/*
+ * fp_drop
+ *
+ * Description:        Drop the I/O reference previously taken by calling fp_lookup
+ *             et. al.
+ *
+ * Parameters: p                               Process in which the fd lives
+ *             fd                              fd associated with the fileproc
+ *             fp                              fileproc on which to set the
+ *                                             flag and drop the reference
+ *             locked                          flag to internally take and
+ *                                             drop proc_fdlock if it is not
+ *                                             already held by the caller
+ *
+ * Returns:    0                               Success
+ *             EBADF                           Bad file descriptor
+ *
+ * Locks:      This function internally takes and drops the proc_fdlock for
+ *             the supplied process if 'locked' is non-zero, and assumes that
+ *             the caller already holds this lock if 'locked' is non-zero.
+ *
+ * Notes:      The fileproc must correspond to the fd in the supplied proc
+ */
 int
 int
-fp_drop(p, fd, fp, locked)
-       struct proc *p;
-       int fd;
-       struct fileproc *fp;
-       int locked;
+fp_drop(proc_t p, int fd, struct fileproc *fp, int locked)
 {
        struct filedesc *fdp = p->p_fd;
 {
        struct filedesc *fdp = p->p_fd;
+       int     needwakeup = 0;
 
        if (!locked)
 
        if (!locked)
-               proc_fdlock(p);
+               proc_fdlock_spin(p);
         if ((fp == FILEPROC_NULL) && (fd < 0 || fd >= fdp->fd_nfiles ||
                        (fp = fdp->fd_ofiles[fd]) == NULL ||
                        ((fdp->fd_ofileflags[fd] & UF_RESERVED) &&
         if ((fp == FILEPROC_NULL) && (fd < 0 || fd >= fdp->fd_nfiles ||
                        (fp = fdp->fd_ofiles[fd]) == NULL ||
                        ((fdp->fd_ofileflags[fd] & UF_RESERVED) &&
@@ -1723,24 +3960,67 @@ fp_drop(p, fd, fp, locked)
        }
        fp->f_iocount--;
 
        }
        fp->f_iocount--;
 
-       if (p->p_fpdrainwait && fp->f_iocount == 0) {
-               p->p_fpdrainwait = 0;
-               wakeup(&p->p_fpdrainwait);
+       if (fp->f_iocount == 0) {
+               if (fp->f_flags & FP_SELCONFLICT)
+                       fp->f_flags &= ~FP_SELCONFLICT;
+
+               if (p->p_fpdrainwait) {
+                       p->p_fpdrainwait = 0;
+                       needwakeup = 1;
+               }
        }
        if (!locked)
                proc_fdunlock(p);
        }
        if (!locked)
                proc_fdunlock(p);
+       if (needwakeup)
+               wakeup(&p->p_fpdrainwait);
                
        return (0);
 }
 
                
        return (0);
 }
 
+
+/*
+ * file_vnode
+ *
+ * Description:        Given an fd, look it up in the current process's per process
+ *             open file table, and return its internal vnode pointer.
+ *
+ * Parameters: fd                              fd to obtain vnode from
+ *             vpp                             pointer to vnode return area
+ *
+ * Returns:    0                               Success
+ *             EINVAL                          The fd does not refer to a
+ *                                             vnode fileproc entry
+ *     fp_lookup:EBADF                         Bad file descriptor
+ *
+ * Implicit returns:
+ *             *vpp (modified)                 Returned vnode pointer
+ *
+ * Locks:      This function internally takes and drops the proc_fdlock for
+ *             the current process
+ *
+ * Notes:      If successful, this function increments the f_iocount on the
+ *             fd's corresponding fileproc.
+ *
+ *             The fileproc referenced is not returned; because of this, care
+ *             must be taken to not drop the last reference (e.g. by closing
+ *             the file).  This is inherently unsafe, since the reference may
+ *             not be recoverable from the vnode, if there is a subsequent
+ *             close that destroys the associate fileproc.  The caller should
+ *             therefore retain their own reference on the fileproc so that
+ *             the f_iocount can be dropped subsequently.  Failure to do this
+ *             can result in the returned pointer immediately becoming invalid
+ *             following the call.
+ *
+ *             Use of this function is discouraged.
+ */
 int
 file_vnode(int fd, struct vnode **vpp)
 {
 int
 file_vnode(int fd, struct vnode **vpp)
 {
-       struct proc * p = current_proc();
+       proc_t p = current_proc();
        struct fileproc *fp;
        int error;
        
        struct fileproc *fp;
        int error;
        
-       proc_fdlock(p);
+       proc_fdlock_spin(p);
        if ( (error = fp_lookup(p, fd, &fp, 1)) ) {
                proc_fdunlock(p);
                return(error);
        if ( (error = fp_lookup(p, fd, &fp, 1)) ) {
                proc_fdunlock(p);
                return(error);
@@ -1750,21 +4030,123 @@ file_vnode(int fd, struct vnode **vpp)
                proc_fdunlock(p);
                return(EINVAL);
        }
                proc_fdunlock(p);
                return(EINVAL);
        }
-       *vpp = (struct vnode *)fp->f_data;
+       if (vpp != NULL)
+               *vpp = (struct vnode *)fp->f_data;
+       proc_fdunlock(p);
+
+       return(0);
+}
+
+
+/*
+ * file_vnode_withvid
+ *
+ * Description:        Given an fd, look it up in the current process's per process
+ *             open file table, and return its internal vnode pointer.
+ *
+ * Parameters: fd                              fd to obtain vnode from
+ *             vpp                             pointer to vnode return area
+ *             vidp                            pointer to vid of the returned vnode
+ *
+ * Returns:    0                               Success
+ *             EINVAL                          The fd does not refer to a
+ *                                             vnode fileproc entry
+ *     fp_lookup:EBADF                         Bad file descriptor
+ *
+ * Implicit returns:
+ *             *vpp (modified)                 Returned vnode pointer
+ *
+ * Locks:      This function internally takes and drops the proc_fdlock for
+ *             the current process
+ *
+ * Notes:      If successful, this function increments the f_iocount on the
+ *             fd's corresponding fileproc.
+ *
+ *             The fileproc referenced is not returned; because of this, care
+ *             must be taken to not drop the last reference (e.g. by closing
+ *             the file).  This is inherently unsafe, since the reference may
+ *             not be recoverable from the vnode, if there is a subsequent
+ *             close that destroys the associate fileproc.  The caller should
+ *             therefore retain their own reference on the fileproc so that
+ *             the f_iocount can be dropped subsequently.  Failure to do this
+ *             can result in the returned pointer immediately becoming invalid
+ *             following the call.
+ *
+ *             Use of this function is discouraged.
+ */
+int
+file_vnode_withvid(int fd, struct vnode **vpp, uint32_t * vidp)
+{
+       proc_t p = current_proc();
+       struct fileproc *fp;
+       vnode_t vp;
+       int error;
+       
+       proc_fdlock_spin(p);
+       if ( (error = fp_lookup(p, fd, &fp, 1)) ) {
+               proc_fdunlock(p);
+               return(error);
+       }
+       if (fp->f_type != DTYPE_VNODE) {
+               fp_drop(p, fd, fp,1);
+               proc_fdunlock(p);
+               return(EINVAL);
+       }
+       vp = (struct vnode *)fp->f_data;
+       if (vpp != NULL) 
+               *vpp = vp;
+
+       if ((vidp != NULL) && (vp != NULLVP)) 
+               *vidp = (uint32_t)vp->v_id;
+
        proc_fdunlock(p);
 
        return(0);
 }
 
 
        proc_fdunlock(p);
 
        return(0);
 }
 
 
+/*
+ * file_socket
+ *
+ * Description:        Given an fd, look it up in the current process's per process
+ *             open file table, and return its internal socket pointer.
+ *
+ * Parameters: fd                              fd to obtain vnode from
+ *             sp                              pointer to socket return area
+ *
+ * Returns:    0                               Success
+ *             ENOTSOCK                        Not a socket
+ *             fp_lookup:EBADF                 Bad file descriptor
+ *
+ * Implicit returns:
+ *             *sp (modified)                  Returned socket pointer
+ *
+ * Locks:      This function internally takes and drops the proc_fdlock for
+ *             the current process
+ *
+ * Notes:      If successful, this function increments the f_iocount on the
+ *             fd's corresponding fileproc.
+ *
+ *             The fileproc referenced is not returned; because of this, care
+ *             must be taken to not drop the last reference (e.g. by closing
+ *             the file).  This is inherently unsafe, since the reference may
+ *             not be recoverable from the socket, if there is a subsequent
+ *             close that destroys the associate fileproc.  The caller should
+ *             therefore retain their own reference on the fileproc so that
+ *             the f_iocount can be dropped subsequently.  Failure to do this
+ *             can result in the returned pointer immediately becoming invalid
+ *             following the call.
+ *
+ *             Use of this function is discouraged.
+ */
 int
 file_socket(int fd, struct socket **sp)
 {
 int
 file_socket(int fd, struct socket **sp)
 {
-       struct proc * p = current_proc();
+       proc_t p = current_proc();
        struct fileproc *fp;
        int error;
        
        struct fileproc *fp;
        int error;
        
-       proc_fdlock(p);
+       proc_fdlock_spin(p);
        if ( (error = fp_lookup(p, fd, &fp, 1)) ) {
                proc_fdunlock(p);
                return(error);
        if ( (error = fp_lookup(p, fd, &fp, 1)) ) {
                proc_fdunlock(p);
                return(error);
@@ -1780,15 +4162,39 @@ file_socket(int fd, struct socket **sp)
        return(0);
 }
 
        return(0);
 }
 
+
+/*
+ * file_flags
+ *
+ * Description:        Given an fd, look it up in the current process's per process
+ *             open file table, and return its fileproc's flags field.
+ *
+ * Parameters: fd                              fd whose flags are to be
+ *                                             retrieved
+ *             flags                           pointer to flags data area
+ *
+ * Returns:    0                               Success
+ *             ENOTSOCK                        Not a socket
+ *             fp_lookup:EBADF                 Bad file descriptor
+ *
+ * Implicit returns:
+ *             *flags (modified)               Returned flags field
+ *
+ * Locks:      This function internally takes and drops the proc_fdlock for
+ *             the current process
+ *
+ * Notes:      This function will internally increment and decrement the
+ *             f_iocount of the fileproc as part of its operation.
+ */
 int
 int
-file_flags(int fd, int * flags)
+file_flags(int fd, int *flags)
 {
 
 {
 
-       struct proc * p = current_proc();
+       proc_t p = current_proc();
        struct fileproc *fp;
        int error;
        
        struct fileproc *fp;
        int error;
        
-       proc_fdlock(p);
+       proc_fdlock_spin(p);
        if ( (error = fp_lookup(p, fd, &fp, 1)) ) {
                proc_fdunlock(p);
                return(error);
        if ( (error = fp_lookup(p, fd, &fp, 1)) ) {
                proc_fdunlock(p);
                return(error);
@@ -1801,13 +4207,50 @@ file_flags(int fd, int * flags)
 }
 
 
 }
 
 
+/*
+ * file_drop
+ *
+ * Description:        Drop an iocount reference on an fd, and wake up any waiters
+ *             for draining (i.e. blocked in fileproc_drain() called during
+ *             the last attempt to close a file).
+ *
+ * Parameters: fd                              fd on which an ioreference is
+ *                                             to be dropped
+ *
+ * Returns:    0                               Success
+ *             EBADF                           Bad file descriptor
+ *
+ * Description:        Given an fd, look it up in the current process's per process
+ *             open file table, and drop it's fileproc's f_iocount by one
+ *
+ * Notes:      This is intended as a corresponding operation to the functions
+ *             file_vnode() and file_socket() operations.
+ *
+ *             Technically, the close reference is supposed to be protected
+ *             by a fileproc_drain(), however, a drain will only block if
+ *             the fd refers to a character device, and that device has had
+ *             preparefileread() called on it.  If it refers to something
+ *             other than a character device, then the drain will occur and
+ *             block each close attempt, rather than merely the last close.
+ *
+ *             Since it's possible for an fd that refers to a character
+ *             device to have an intermediate close followed by an open to
+ *             cause a different file to correspond to that descriptor,
+ *             unless there was a cautionary reference taken on the fileproc,
+ *             this is an inherently unsafe function.  This happens in the
+ *             case where multiple fd's in a process refer to the same
+ *             character device (e.g. stdin/out/err pointing to a tty, etc.).
+ *
+ *             Use of this function is discouraged.
+ */
 int 
 file_drop(int fd)
 {
        struct fileproc *fp;
 int 
 file_drop(int fd)
 {
        struct fileproc *fp;
-       struct proc *p = current_proc();
+       proc_t p = current_proc();
+       int     needwakeup = 0;
 
 
-       proc_fdlock(p);
+       proc_fdlock_spin(p);
        if (fd < 0 || fd >= p->p_fd->fd_nfiles ||
                        (fp = p->p_fd->fd_ofiles[fd]) == NULL ||
                        ((p->p_fd->fd_ofileflags[fd] & UF_RESERVED) &&
        if (fd < 0 || fd >= p->p_fd->fd_nfiles ||
                        (fp = p->p_fd->fd_ofiles[fd]) == NULL ||
                        ((p->p_fd->fd_ofileflags[fd] & UF_RESERVED) &&
@@ -1817,42 +4260,142 @@ file_drop(int fd)
        }
        fp->f_iocount --;
 
        }
        fp->f_iocount --;
 
-       if (p->p_fpdrainwait && fp->f_iocount == 0) {
-               p->p_fpdrainwait = 0;
-               wakeup(&p->p_fpdrainwait);
+       if (fp->f_iocount == 0) {
+               if (fp->f_flags & FP_SELCONFLICT)
+                       fp->f_flags &= ~FP_SELCONFLICT;
+
+               if (p->p_fpdrainwait) {
+                       p->p_fpdrainwait = 0;
+                       needwakeup = 1;
+               }
        }
        proc_fdunlock(p);
        }
        proc_fdunlock(p);
+
+       if (needwakeup)
+               wakeup(&p->p_fpdrainwait);
        return(0);
        return(0);
+}
 
 
 
 
+static int falloc_withalloc_locked(proc_t, struct fileproc **, int *,
+    vfs_context_t, struct fileproc * (*)(void *), void *, int);
+
+/*
+ * falloc
+ *
+ * Description:        Allocate an entry in the per process open file table and
+ *             return the corresponding fileproc and fd.
+ *
+ * Parameters: p                               The process in whose open file
+ *                                             table the fd is to be allocated
+ *             resultfp                        Pointer to fileproc pointer
+ *                                             return area
+ *             resultfd                        Pointer to fd return area
+ *             ctx                             VFS context
+ *
+ * Returns:    0                               Success
+ *     falloc:ENFILE                           Too many open files in system
+ *     falloc:EMFILE                           Too many open files in process
+ *     falloc:ENOMEM                           M_FILEPROC or M_FILEGLOB zone
+ *                                             exhausted
+ *
+ * Implicit returns:
+ *             *resultfd (modified)            Returned fileproc pointer
+ *             *resultfd (modified)            Returned fd
+ *
+ * Locks:      This function takes and drops the proc_fdlock; if this lock
+ *             is already held, use falloc_locked() instead.
+ *
+ * Notes:      This function takes separate process and context arguments
+ *             solely to support kern_exec.c; otherwise, it would take
+ *             neither, and expect falloc_locked() to use the
+ *             vfs_context_current() routine internally.
+ */
+int
+falloc(proc_t p, struct fileproc **resultfp, int *resultfd, vfs_context_t ctx)
+{
+       return (falloc_withalloc(p, resultfp, resultfd, ctx,
+           fileproc_alloc_init, NULL));
 }
 
 }
 
+/*
+ * Like falloc, but including the fileproc allocator and create-args
+ */
 int
 int
-falloc(p, resultfp, resultfd )
-       struct proc *p;
-       struct fileproc **resultfp;
-       int *resultfd;
+falloc_withalloc(proc_t p, struct fileproc **resultfp, int *resultfd,
+    vfs_context_t ctx, fp_allocfn_t fp_zalloc, void *arg)
 {
        int error;
 
        proc_fdlock(p);
 {
        int error;
 
        proc_fdlock(p);
-       error = falloc_locked(p, resultfp, resultfd, 1);
+       error = falloc_withalloc_locked(p,
+           resultfp, resultfd, ctx, fp_zalloc, arg, 1);
        proc_fdunlock(p);
 
        proc_fdunlock(p);
 
-       return(error);
+       return (error);
 }
 }
+
+/*
+ * "uninitialized" ops -- ensure fg->fg_ops->fo_type always exists
+ */
+static const struct fileops uninitops;
+
 /*
 /*
+ * falloc_locked
+ *
  * Create a new open file structure and allocate
  * Create a new open file structure and allocate
- * a file decriptor for the process that refers to it.
+ * a file descriptor for the process that refers to it.
+ *
+ * Returns:    0                       Success
+ *
+ * Description:        Allocate an entry in the per process open file table and
+ *             return the corresponding fileproc and fd.
+ *
+ * Parameters: p                               The process in whose open file
+ *                                             table the fd is to be allocated
+ *             resultfp                        Pointer to fileproc pointer
+ *                                             return area
+ *             resultfd                        Pointer to fd return area
+ *             ctx                             VFS context
+ *             locked                          Flag to indicate whether the
+ *                                             caller holds proc_fdlock
+ *
+ * Returns:    0                               Success
+ *             ENFILE                          Too many open files in system
+ *             fdalloc:EMFILE                  Too many open files in process
+ *             ENOMEM                          M_FILEPROC or M_FILEGLOB zone
+ *                                             exhausted
+ *     fdalloc:ENOMEM
+ *
+ * Implicit returns:
+ *             *resultfd (modified)            Returned fileproc pointer
+ *             *resultfd (modified)            Returned fd
+ *
+ * Locks:      If the parameter 'locked' is zero, this function takes and
+ *             drops the proc_fdlock; if non-zero, the caller must hold the
+ *             lock.
+ *
+ * Notes:      If you intend to use a non-zero 'locked' parameter, use the
+ *             utility function falloc() instead.
+ *
+ *             This function takes separate process and context arguments
+ *             solely to support kern_exec.c; otherwise, it would take
+ *             neither, and use the vfs_context_current() routine internally.
  */
 int
  */
 int
-falloc_locked(p, resultfp, resultfd, locked)
-       struct proc *p;
-       struct fileproc **resultfp;
-       int *resultfd;
-       int locked;
+falloc_locked(proc_t p, struct fileproc **resultfp, int *resultfd,
+             vfs_context_t ctx, int locked)
+{
+       return (falloc_withalloc_locked(p, resultfp, resultfd, ctx,
+           fileproc_alloc_init, NULL, locked));
+}
+
+static int
+falloc_withalloc_locked(proc_t p, struct fileproc **resultfp, int *resultfd,
+    vfs_context_t ctx, fp_allocfn_t fp_zalloc, void *crarg,
+    int locked)
 {
 {
-       struct fileproc *fp, *fq;
+       struct fileproc *fp;
        struct fileglob *fg;
        int error, nfd;
 
        struct fileglob *fg;
        int error, nfd;
 
@@ -1869,6 +4412,15 @@ falloc_locked(p, resultfp, resultfd, locked)
                tablefull("file");
                return (ENFILE);
        }
                tablefull("file");
                return (ENFILE);
        }
+#if CONFIG_MACF
+       error = mac_file_check_create(proc_ucred(p));
+       if (error) {
+               if (!locked)
+                       proc_fdunlock(p);
+               return (error);
+       }
+#endif
+
        /*
         * Allocate a new file descriptor.
         * If the process has file descriptor zero open, add to the list
        /*
         * Allocate a new file descriptor.
         * If the process has file descriptor zero open, add to the list
@@ -1877,30 +4429,41 @@ falloc_locked(p, resultfp, resultfd, locked)
         */
        proc_fdunlock(p);
 
         */
        proc_fdunlock(p);
 
-       MALLOC_ZONE(fp, struct fileproc *, sizeof(struct fileproc), M_FILEPROC, M_WAITOK);
+       fp = (*fp_zalloc)(crarg);
+       if (fp == NULL) {
+               if (locked)
+                       proc_fdlock(p);
+               return (ENOMEM);
+       }
        MALLOC_ZONE(fg, struct fileglob *, sizeof(struct fileglob), M_FILEGLOB, M_WAITOK);
        MALLOC_ZONE(fg, struct fileglob *, sizeof(struct fileglob), M_FILEGLOB, M_WAITOK);
-       bzero(fp, sizeof(struct fileproc));
+       if (fg == NULL) {
+               fileproc_free(fp);
+               if (locked)
+                       proc_fdlock(p);
+               return (ENOMEM);
+       }
        bzero(fg, sizeof(struct fileglob));
        lck_mtx_init(&fg->fg_lock, file_lck_grp, file_lck_attr);
 
        fp->f_iocount = 1;
        fg->fg_count = 1;
        bzero(fg, sizeof(struct fileglob));
        lck_mtx_init(&fg->fg_lock, file_lck_grp, file_lck_attr);
 
        fp->f_iocount = 1;
        fg->fg_count = 1;
+       fg->fg_ops = &uninitops;
        fp->f_fglob = fg;
        fp->f_fglob = fg;
+#if CONFIG_MACF
+       mac_file_label_init(fg);
+#endif
 
 
-       proc_fdlock(p);
+       kauth_cred_ref(ctx->vc_ucred);
 
 
-       fp->f_cred = kauth_cred_proc_ref(p);
+       proc_fdlock(p);
 
 
-       lck_mtx_lock(file_flist_lock);
+       fp->f_cred = ctx->vc_ucred;
 
 
-       nfiles++;
+#if CONFIG_MACF
+       mac_file_label_associate(fp->f_cred, fg);
+#endif
 
 
-       if ( (fq = p->p_fd->fd_ofiles[0]) ) {
-               LIST_INSERT_AFTER(fq->f_fglob, fg, f_list);
-       } else {
-               LIST_INSERT_HEAD(&filehead, fg, f_list);
-       }
-       lck_mtx_unlock(file_flist_lock);
+       OSAddAtomic(1, &nfiles);
 
        p->p_fd->fd_ofiles[nfd] = fp;
 
 
        p->p_fd->fd_ofiles[nfd] = fp;
 
@@ -1915,69 +4478,156 @@ falloc_locked(p, resultfp, resultfd, locked)
        return (0);
 }
 
        return (0);
 }
 
+
 /*
 /*
- * Free a file structure.
+ * fg_free
+ *
+ * Description:        Free a file structure; drop the global open file count, and
+ *             drop the credential reference, if the fileglob has one, and
+ *             destroy the instance mutex before freeing
+ *
+ * Parameters: fg                              Pointer to fileglob to be
+ *                                             freed
+ *
+ * Returns:    void
  */
 void
  */
 void
-fg_free(fg)
-       struct fileglob *fg;
+fg_free(struct fileglob *fg)
 {
 {
-       kauth_cred_t cred;
+       OSAddAtomic(-1, &nfiles);
 
 
-       lck_mtx_lock(file_flist_lock);
-       LIST_REMOVE(fg, f_list);
-       nfiles--;
-       lck_mtx_unlock(file_flist_lock);
+       if (fg->fg_vn_data) {
+               fg_vn_data_free(fg->fg_vn_data);
+               fg->fg_vn_data = NULL;
+       }
 
 
-       cred = fg->fg_cred;
-       if (cred != NOCRED) {
-               fg->fg_cred = NOCRED;
-               kauth_cred_rele(cred);
+       if (IS_VALID_CRED(fg->fg_cred)) {
+               kauth_cred_unref(&fg->fg_cred);
        }
        lck_mtx_destroy(&fg->fg_lock, file_lck_grp);
 
        }
        lck_mtx_destroy(&fg->fg_lock, file_lck_grp);
 
+#if CONFIG_MACF
+       mac_file_label_destroy(fg);
+#endif
        FREE_ZONE(fg, sizeof *fg, M_FILEGLOB);
 }
 
        FREE_ZONE(fg, sizeof *fg, M_FILEGLOB);
 }
 
+
+/*
+ * fdexec
+ *
+ * Description:        Perform close-on-exec processing for all files in a process
+ *             that are either marked as close-on-exec, or which were in the
+ *             process of being opened at the time of the execve
+ *
+ *             Also handles the case (via posix_spawn()) where -all-
+ *             files except those marked with "inherit" as treated as
+ *             close-on-exec.
+ *
+ * Parameters: p                               Pointer to process calling
+ *                                             execve
+ *
+ * Returns:    void
+ *
+ * Locks:      This function internally takes and drops proc_fdlock()
+ *
+ */
 void
 void
-fdexec(p)
-       struct proc *p;
+fdexec(proc_t p, short flags)
 {
        struct filedesc *fdp = p->p_fd;
 {
        struct filedesc *fdp = p->p_fd;
-       int i = fdp->fd_lastfile;
-       struct fileproc **fpp = &fdp->fd_ofiles[i];
-       char *flags = &fdp->fd_ofileflags[i];
-       int funnel_state;
+       int i;
+       boolean_t cloexec_default = (flags & POSIX_SPAWN_CLOEXEC_DEFAULT) != 0;
 
 
-       funnel_state = thread_funnel_set(kernel_flock, FALSE);
        proc_fdlock(p);
        proc_fdlock(p);
+       for (i = fdp->fd_lastfile; i >= 0; i--) {
+
+               struct fileproc *fp = fdp->fd_ofiles[i];
+               char *flagp = &fdp->fd_ofileflags[i];
 
 
-       while (i >= 0) {
-               if ((*flags & (UF_RESERVED|UF_EXCLOSE)) == UF_EXCLOSE) {
-                       struct fileproc *fp = *fpp;
+               if (fp && cloexec_default) {
+                       /*
+                        * Reverse the usual semantics of file descriptor
+                        * inheritance - all of them should be closed
+                        * except files marked explicitly as "inherit" and
+                        * not marked close-on-exec.
+                        */
+                       if ((*flagp & (UF_EXCLOSE|UF_INHERIT)) != UF_INHERIT)
+                               *flagp |= UF_EXCLOSE;
+                       *flagp &= ~UF_INHERIT;
+               }
 
 
+               if (
+                   ((*flagp & (UF_RESERVED|UF_EXCLOSE)) == UF_EXCLOSE)
+#if CONFIG_MACF
+                   || (fp && mac_file_check_inherit(proc_ucred(p), fp->f_fglob))
+#endif
+               ) {
                         if (i < fdp->fd_knlistsize)
                                 knote_fdclose(p, i);
                         if (i < fdp->fd_knlistsize)
                                 knote_fdclose(p, i);
-
-                       *fpp = NULL; *flags = 0;
+                       procfdtbl_clearfd(p, i);
                        if (i == fdp->fd_lastfile && i > 0)
                                fdp->fd_lastfile--;
                        if (i == fdp->fd_lastfile && i > 0)
                                fdp->fd_lastfile--;
+                       if (i < fdp->fd_freefile)
+                               fdp->fd_freefile = i;
+
+                       /*
+                        * Wait for any third party viewers (e.g., lsof)
+                        * to release their references to this fileproc.
+                        */
+                       while (fp->f_iocount > 0) {
+                               p->p_fpdrainwait = 1;
+                               msleep(&p->p_fpdrainwait, &p->p_fdmlock, PRIBIO,
+                                   "fpdrain", NULL);
+                       }
+
                        closef_locked(fp, fp->f_fglob, p);
                        closef_locked(fp, fp->f_fglob, p);
-                       FREE_ZONE(fp, sizeof *fp, M_FILEPROC);
-               }
 
 
-               i--; fpp--; flags--;
+                       fileproc_free(fp);
+               }
        }
        proc_fdunlock(p);
        }
        proc_fdunlock(p);
-       thread_funnel_set(kernel_flock, funnel_state);
 }
 
 }
 
+
 /*
 /*
- * Copy a filedesc structure.
+ * fdcopy
+ *
+ * Description:        Copy a filedesc structure.  This is normally used as part of
+ *             forkproc() when forking a new process, to copy the per process
+ *             open file table over to the new process.
+ *
+ * Parameters: p                               Process whose open file table
+ *                                             is to be copied (parent)
+ *             uth_cdir                        Per thread current working
+ *                                             cirectory, or NULL
+ *
+ * Returns:    NULL                            Copy failed
+ *             !NULL                           Pointer to new struct filedesc
+ *
+ * Locks:      This function internally takes and drops proc_fdlock()
+ *
+ * Notes:      Files are copied directly, ignoring the new resource limits
+ *             for the process that's being copied into.  Since the descriptor
+ *             references are just additional references, this does not count
+ *             against the number of open files on the system.
+ *
+ *             The struct filedesc includes the current working directory,
+ *             and the current root directory, if the process is chroot'ed.
+ *
+ *             If the exec was called by a thread using a per thread current
+ *             working directory, we inherit the working directory from the
+ *             thread making the call, rather than from the process.
+ *
+ *             In the case of a failure to obtain a reference, for most cases,
+ *             the file entry will be silently dropped.  There's an exception
+ *             for the case of a chroot dir, since a failure to to obtain a
+ *             reference there would constitute an "escape" from the chroot
+ *             environment, which must not be allowed.  In that case, we will
+ *             deny the execve() operation, rather than allowing the escape.
  */
 struct filedesc *
  */
 struct filedesc *
-fdcopy(p)
-       struct proc *p;
+fdcopy(proc_t p, vnode_t uth_cdir)
 {
        struct filedesc *newfdp, *fdp = p->p_fd;
        int i;
 {
        struct filedesc *newfdp, *fdp = p->p_fd;
        int i;
@@ -1985,7 +4635,7 @@ fdcopy(p)
        vnode_t v_dir;
 
        MALLOC_ZONE(newfdp, struct filedesc *,
        vnode_t v_dir;
 
        MALLOC_ZONE(newfdp, struct filedesc *,
-                       sizeof *newfdp, M_FILEDESC, M_WAITOK);
+                       sizeof(*newfdp), M_FILEDESC, M_WAITOK);
        if (newfdp == NULL)
                return(NULL);
 
        if (newfdp == NULL)
                return(NULL);
 
@@ -1994,10 +4644,18 @@ fdcopy(p)
        /*
         * the FD_CHROOT flag will be inherited via this copy
         */
        /*
         * the FD_CHROOT flag will be inherited via this copy
         */
-       (void) memcpy(newfdp, fdp, sizeof *newfdp);
+       (void) memcpy(newfdp, fdp, sizeof(*newfdp));
+
+       /*
+        * If we are running with per-thread current working directories,
+        * inherit the new current working directory from the current thread
+        * instead, before we take our references.
+        */
+       if (uth_cdir != NULLVP)
+               newfdp->fd_cdir = uth_cdir;
 
        /*
 
        /*
-        * for both fd_cdir and fd_rdir make sure we get
+        * For both fd_cdir and fd_rdir make sure we get
         * a valid reference... if we can't, than set
         * set the pointer(s) to NULL in the child... this
         * will keep us from using a non-referenced vp
         * a valid reference... if we can't, than set
         * set the pointer(s) to NULL in the child... this
         * will keep us from using a non-referenced vp
@@ -2020,7 +4678,7 @@ fdcopy(p)
                 * our reference from the parent also
                 * since the vnode has gone DEAD making
                 * it useless... by dropping it we'll
                 * our reference from the parent also
                 * since the vnode has gone DEAD making
                 * it useless... by dropping it we'll
-                * be that much closer to recyling it
+                * be that much closer to recycling it
                 */
                vnode_rele(fdp->fd_cdir);
                fdp->fd_cdir = NULL;
                 */
                vnode_rele(fdp->fd_cdir);
                fdp->fd_cdir = NULL;
@@ -2031,23 +4689,25 @@ fdcopy(p)
                        if ( (vnode_ref(v_dir)) )
                                newfdp->fd_rdir = NULL;
                        vnode_put(v_dir);
                        if ( (vnode_ref(v_dir)) )
                                newfdp->fd_rdir = NULL;
                        vnode_put(v_dir);
-               } else
+               } else {
                        newfdp->fd_rdir = NULL;
                        newfdp->fd_rdir = NULL;
+               }
        }
        }
+       /* Coming from a chroot environment and unable to get a reference... */
        if (newfdp->fd_rdir == NULL && fdp->fd_rdir) {
                /*
        if (newfdp->fd_rdir == NULL && fdp->fd_rdir) {
                /*
-                * we couldn't get a new reference on
-                * the root directory being
-                * inherited... we might as well drop
-                * our reference from the parent also
-                * since the vnode has gone DEAD making
-                * it useless... by dropping it we'll
-                * be that much closer to recyling it
+                * We couldn't get a new reference on
+                * the chroot directory being
+                * inherited... this is fatal, since
+                * otherwise it would constitute an
+                * escape from a chroot environment by
+                * the new process.
                 */
                 */
-               vnode_rele(fdp->fd_rdir);
-               fdp->fd_rdir = NULL;
+               if (newfdp->fd_cdir)
+                       vnode_rele(newfdp->fd_cdir);
+               FREE_ZONE(newfdp, sizeof *newfdp, M_FILEDESC);
+               return(NULL);
        }
        }
-       newfdp->fd_refcnt = 1;
 
        /*
         * If the number of open files fits in the internal arrays
 
        /*
         * If the number of open files fits in the internal arrays
@@ -2077,9 +4737,10 @@ fdcopy(p)
                if (newfdp->fd_rdir)
                        vnode_rele(newfdp->fd_rdir);
 
                if (newfdp->fd_rdir)
                        vnode_rele(newfdp->fd_rdir);
 
-               FREE_ZONE(newfdp, sizeof *newfdp, M_FILEDESC);
+               FREE_ZONE(newfdp, sizeof(*newfdp), M_FILEDESC);
                return(NULL);
        }
                return(NULL);
        }
+       (void) memset(newfdp->fd_ofiles, 0, i * OFILESIZE);
        proc_fdlock(p);
 
        newfdp->fd_ofileflags = (char *) &newfdp->fd_ofiles[i];
        proc_fdlock(p);
 
        newfdp->fd_ofileflags = (char *) &newfdp->fd_ofiles[i];
@@ -2090,18 +4751,23 @@ fdcopy(p)
                char *flags;
 
                (void) memcpy(newfdp->fd_ofiles, fdp->fd_ofiles,
                char *flags;
 
                (void) memcpy(newfdp->fd_ofiles, fdp->fd_ofiles,
-                                       i * sizeof *fdp->fd_ofiles);
+                                       (newfdp->fd_lastfile + 1) * sizeof(*fdp->fd_ofiles));
                (void) memcpy(newfdp->fd_ofileflags, fdp->fd_ofileflags,
                (void) memcpy(newfdp->fd_ofileflags, fdp->fd_ofileflags,
-                                       i * sizeof *fdp->fd_ofileflags);
+                                       (newfdp->fd_lastfile + 1) * sizeof(*fdp->fd_ofileflags));
 
                /*
                 * kq descriptors cannot be copied.
                 */
                if (newfdp->fd_knlistsize != -1) {
                        fpp = &newfdp->fd_ofiles[newfdp->fd_lastfile];
 
                /*
                 * kq descriptors cannot be copied.
                 */
                if (newfdp->fd_knlistsize != -1) {
                        fpp = &newfdp->fd_ofiles[newfdp->fd_lastfile];
-                       for (i = newfdp->fd_lastfile; i >= 0; i--, fpp--) {
+                       flags = &newfdp->fd_ofileflags[newfdp->fd_lastfile];
+                       for (i = newfdp->fd_lastfile;
+                           i >= 0; i--, fpp--, flags--) {
+                               if (*flags & UF_RESERVED)
+                                       continue;       /* (removed below) */
                                if (*fpp != NULL && (*fpp)->f_type == DTYPE_KQUEUE) {
                                        *fpp = NULL;
                                if (*fpp != NULL && (*fpp)->f_type == DTYPE_KQUEUE) {
                                        *fpp = NULL;
+                                       *flags = 0;
                                        if (i < newfdp->fd_freefile)
                                                newfdp->fd_freefile = i;
                                }
                                        if (i < newfdp->fd_freefile)
                                                newfdp->fd_freefile = i;
                                }
@@ -2116,33 +4782,56 @@ fdcopy(p)
                fpp = newfdp->fd_ofiles;
                flags = newfdp->fd_ofileflags;
 
                fpp = newfdp->fd_ofiles;
                flags = newfdp->fd_ofileflags;
 
-               for (i = newfdp->fd_lastfile; i-- >= 0; fpp++, flags++)
-                       if ((ofp = *fpp) != NULL && !(*flags & UF_RESERVED)) {
-                               MALLOC_ZONE(fp, struct fileproc *, sizeof(struct fileproc), M_FILEPROC, M_WAITOK);
-                               bzero(fp, sizeof(struct fileproc));
-                               fp->f_flags = ofp->f_flags;
-                               //fp->f_iocount = ofp->f_iocount;
-                               fp->f_iocount = 0;
-                               fp->f_fglob = ofp->f_fglob;
-                               (void)fg_ref(fp);
-                               *fpp = fp;
+               for (i = newfdp->fd_lastfile + 1; --i >= 0; fpp++, flags++)
+                       if ((ofp = *fpp) != NULL &&
+                           0 == (*flags & (UF_FORKCLOSE|UF_RESERVED))) {
+#if DEBUG
+                               if (FILEPROC_TYPE(ofp) != FTYPE_SIMPLE)
+                                       panic("complex fileproc");
+#endif
+                               fp = fileproc_alloc_init(NULL);
+                               if (fp == NULL) {
+                                       /*
+                                        * XXX no room to copy, unable to
+                                        * XXX safely unwind state at present
+                                        */
+                                       *fpp = NULL;
+                               } else {
+                                       fp->f_flags |=
+                                           (ofp->f_flags & ~FP_TYPEMASK);
+                                       fp->f_fglob = ofp->f_fglob;
+                                       (void)fg_ref(fp);
+                                       *fpp = fp;
+                               }
                        } else {
                        } else {
+                               if (i < newfdp->fd_freefile)
+                                       newfdp->fd_freefile = i;
                                *fpp = NULL;
                                *flags = 0;
                        }
                                *fpp = NULL;
                                *flags = 0;
                        }
-       } else
-               (void) memset(newfdp->fd_ofiles, 0, i * OFILESIZE);
+       }
 
        proc_fdunlock(p);
        return (newfdp);
 }
 
 
        proc_fdunlock(p);
        return (newfdp);
 }
 
+
 /*
 /*
- * Release a filedesc structure.
+ * fdfree
+ *
+ * Description:        Release a filedesc (per process open file table) structure;
+ *             this is done on process exit(), or from forkproc_free() if
+ *             the fork fails for some reason subsequent to a successful
+ *             call to fdcopy()
+ *
+ * Parameters: p                               Pointer to process going away
+ *
+ * Returns:    void
+ *
+ * Locks:      This function internally takes and drops proc_fdlock()
  */
 void
  */
 void
-fdfree(p)
-       struct proc *p;
+fdfree(proc_t p)
 {
        struct filedesc *fdp;
        struct fileproc *fp;
 {
        struct filedesc *fdp;
        struct fileproc *fp;
@@ -2150,37 +4839,31 @@ fdfree(p)
 
        proc_fdlock(p);
 
 
        proc_fdlock(p);
 
-       /* Certain daemons might not have file descriptors */
-       fdp = p->p_fd;
-
-       if ((fdp == NULL) || (--fdp->fd_refcnt > 0)) {
+       if (p == kernproc || NULL == (fdp = p->p_fd)) {
                proc_fdunlock(p);
                return;
        }
                proc_fdunlock(p);
                return;
        }
-       if (fdp->fd_refcnt == 0xffff)
-               panic("fdfree: bad fd_refcnt");
 
 
-       /* Last reference: the structure can't change out from under us */
+       extern struct filedesc filedesc0;
+
+       if (&filedesc0 == fdp)
+               panic("filedesc0");
 
        if (fdp->fd_nfiles > 0 && fdp->fd_ofiles) {
                for (i = fdp->fd_lastfile; i >= 0; i--) {
                        if ((fp = fdp->fd_ofiles[i]) != NULL) {
                          
                          if (fdp->fd_ofileflags[i] & UF_RESERVED)
 
        if (fdp->fd_nfiles > 0 && fdp->fd_ofiles) {
                for (i = fdp->fd_lastfile; i >= 0; i--) {
                        if ((fp = fdp->fd_ofiles[i]) != NULL) {
                          
                          if (fdp->fd_ofileflags[i] & UF_RESERVED)
-                           panic("fdfree: found fp with UF_RESERVED\n");
+                               panic("fdfree: found fp with UF_RESERVED");
 
 
-                               /* closef drops the iocount ... */
-                               if ((fp->f_flags & FP_INCHRREAD) != 0) 
-                                       fp->f_iocount++;
-                           fdp->fd_ofiles[i] = NULL;
-                               fdp->fd_ofileflags[i] |= UF_RESERVED;
+                               procfdtbl_reservefd(p, i);
 
                                if (i < fdp->fd_knlistsize)
                                        knote_fdclose(p, i);
                                if (fp->f_flags & FP_WAITEVENT) 
                                        (void)waitevent_close(p, fp);
                                (void) closef_locked(fp, fp->f_fglob, p);
 
                                if (i < fdp->fd_knlistsize)
                                        knote_fdclose(p, i);
                                if (fp->f_flags & FP_WAITEVENT) 
                                        (void)waitevent_close(p, fp);
                                (void) closef_locked(fp, fp->f_fglob, p);
-                               FREE_ZONE(fp, sizeof *fp, M_FILEPROC);
+                               fileproc_free(fp);
                        }
                }
                FREE_ZONE(fdp->fd_ofiles, fdp->fd_nfiles * OFILESIZE, M_OFILETABL);
                        }
                }
                FREE_ZONE(fdp->fd_ofiles, fdp->fd_nfiles * OFILESIZE, M_OFILETABL);
@@ -2195,7 +4878,7 @@ fdfree(p)
        if (fdp->fd_rdir)
                vnode_rele(fdp->fd_rdir);
 
        if (fdp->fd_rdir)
                vnode_rele(fdp->fd_rdir);
 
-       proc_fdlock(p);
+       proc_fdlock_spin(p);
        p->p_fd = NULL;
        proc_fdunlock(p);
 
        p->p_fd = NULL;
        proc_fdunlock(p);
 
@@ -2204,72 +4887,31 @@ fdfree(p)
        if (fdp->fd_knhash)
                FREE(fdp->fd_knhash, M_KQUEUE);
 
        if (fdp->fd_knhash)
                FREE(fdp->fd_knhash, M_KQUEUE);
 
-       FREE_ZONE(fdp, sizeof *fdp, M_FILEDESC);
-}
-
-static int
-closef_finish(fp, fg, p)
-       struct fileproc *fp;
-       struct fileglob *fg;
-       struct proc *p;
-{
-       struct vnode *vp;
-       struct flock lf;
-       int error;
-       struct vfs_context context;
-
-       if ((fg->fg_flag & FHASLOCK) && fg->fg_type == DTYPE_VNODE) {
-               lf.l_whence = SEEK_SET;
-               lf.l_start = 0;
-               lf.l_len = 0;
-               lf.l_type = F_UNLCK;
-               vp = (struct vnode *)fg->fg_data;
-               context.vc_proc = p;
-               context.vc_ucred = fg->fg_cred;
-
-               (void) VNOP_ADVLOCK(vp, (caddr_t)fg, F_UNLCK, &lf, F_FLOCK, &context);
-       }
-       if (fg->fg_ops)
-               error = fo_close(fg, p);
-       else
-               error = 0;
-
-       if (((fp != (struct fileproc *)0) && ((fp->f_flags & FP_INCHRREAD) != 0))) {
-               proc_fdlock(p);
-               if ( ((fp->f_flags & FP_INCHRREAD) != 0) ) {
-                       fileproc_drain(p, fp);
-               }
-               proc_fdunlock(p);
-       }
-       fg_free(fg);
-
-       return (error);
+       FREE_ZONE(fdp, sizeof(*fdp), M_FILEDESC);
 }
 
 }
 
-int
-closef(fg, p)
-       struct fileglob *fg;
-       struct proc *p;
-{
-       int error;
-
-       proc_fdlock(p);
-       error = closef_locked((struct fileproc *)0, fg, p);
-       proc_fdunlock(p);
-
-       return(error);
-}
 /*
 /*
- * Internal form of close.
- * Decrement reference count on file structure.
- * Note: p may be NULL when closing a file
- * that was being passed in a message.
+ * closef_locked
+ *
+ * Description:        Internal form of closef; called with proc_fdlock held
+ *
+ * Parameters: fp                      Pointer to fileproc for fd
+ *             fg                      Pointer to fileglob for fd
+ *             p                       Pointer to proc structure
+ *
+ * Returns:    0                       Success
+ *     closef_finish:???               Anything returnable by a per-fileops
+ *                                     close function
+ *
+ * Note:       Decrements reference count on file structure; if this was the
+ *             last reference, then closef_finish() is called
+ *
+ *             p and fp are allowed to  be NULL when closing a file that was
+ *             being passed in a message (but only if we are called when this
+ *             is NOT the last reference).
  */
 int
  */
 int
-closef_locked(fp, fg, p)
-       struct fileproc *fp;
-       struct fileglob *fg;
-       struct proc *p;
+closef_locked(struct fileproc *fp, struct fileglob *fg, proc_t p)
 {
        struct vnode *vp;
        struct flock lf;
 {
        struct vnode *vp;
        struct flock lf;
@@ -2279,6 +4921,14 @@ closef_locked(fp, fg, p)
        if (fg == NULL) {
                return (0);
        }
        if (fg == NULL) {
                return (0);
        }
+
+       /* Set up context with cred stashed in fg */
+       if (p == current_proc())
+               context.vc_thread = current_thread();
+       else
+               context.vc_thread = NULL;
+       context.vc_ucred = fg->fg_cred;
+
        /*
         * POSIX record locking dictates that any close releases ALL
         * locks owned by this process.  This is handled by setting
        /*
         * POSIX record locking dictates that any close releases ALL
         * locks owned by this process.  This is handled by setting
@@ -2287,7 +4937,8 @@ closef_locked(fp, fg, p)
         * If the descriptor was in a message, POSIX-style locks
         * aren't passed with the descriptor.
         */
         * If the descriptor was in a message, POSIX-style locks
         * aren't passed with the descriptor.
         */
-       if (p && (p->p_ladvflag & P_LADVLOCK) && fg->fg_type == DTYPE_VNODE) {
+       if (p && (p->p_ladvflag & P_LADVLOCK) &&
+           DTYPE_VNODE == FILEGLOB_DTYPE(fg)) {
                proc_fdunlock(p);
 
                lf.l_whence = SEEK_SET;
                proc_fdunlock(p);
 
                lf.l_whence = SEEK_SET;
@@ -2297,23 +4948,22 @@ closef_locked(fp, fg, p)
                vp = (struct vnode *)fg->fg_data;
 
                if ( (error = vnode_getwithref(vp)) == 0 ) {
                vp = (struct vnode *)fg->fg_data;
 
                if ( (error = vnode_getwithref(vp)) == 0 ) {
-                       context.vc_proc = p;
-                       context.vc_ucred = fg->fg_cred;
-                       (void) VNOP_ADVLOCK(vp, (caddr_t)p, F_UNLCK, &lf, F_POSIX, &context);
-
+                       (void) VNOP_ADVLOCK(vp, (caddr_t)p, F_UNLCK, &lf, F_POSIX, &context, NULL);
                        (void)vnode_put(vp);
                }
                proc_fdlock(p);
        }
                        (void)vnode_put(vp);
                }
                proc_fdlock(p);
        }
-       lck_mtx_lock(&fg->fg_lock);
+       lck_mtx_lock_spin(&fg->fg_lock);
        fg->fg_count--;
 
        if (fg->fg_count > 0) {
                lck_mtx_unlock(&fg->fg_lock);
                return (0);
        }
        fg->fg_count--;
 
        if (fg->fg_count > 0) {
                lck_mtx_unlock(&fg->fg_lock);
                return (0);
        }
+#if DIAGNOSTIC
        if (fg->fg_count != 0)
        if (fg->fg_count != 0)
-               panic("fg: being freed with bad fg_count (%d)", fg, fg->fg_count);
+               panic("fg %p: being freed with bad fg_count (%d)", fg, fg->fg_count);
+#endif
 
        if (fp && (fp->f_flags & FP_WRITTEN))
                fg->fg_flag |= FWASWRITTEN;
 
        if (fp && (fp->f_flags & FP_WRITTEN))
                fg->fg_flag |= FWASWRITTEN;
@@ -2321,63 +4971,143 @@ closef_locked(fp, fg, p)
        fg->fg_lflags |= FG_TERM;
        lck_mtx_unlock(&fg->fg_lock);
 
        fg->fg_lflags |= FG_TERM;
        lck_mtx_unlock(&fg->fg_lock);
 
-       proc_fdunlock(p);
-       error = closef_finish(fp, fg, p);
-       proc_fdlock(p);
+       if (p)
+               proc_fdunlock(p);
+
+       /* Since we ensure that fg->fg_ops is always initialized, 
+        * it is safe to invoke fo_close on the fg */
+       error = fo_close(fg, &context);
+
+       fg_free(fg);
+       
+       if (p)
+               proc_fdlock(p);
 
        return(error);
 }
 
 
 
        return(error);
 }
 
 
-extern int selwait;
+/*
+ * fileproc_drain
+ *
+ * Description:        Drain out pending I/O operations
+ *
+ * Parameters: p                               Process closing this file
+ *             fp                              fileproc struct for the open
+ *                                             instance on the file
+ *
+ * Returns:    void
+ *
+ * Locks:      Assumes the caller holds the proc_fdlock
+ *
+ * Notes:      For character devices, this occurs on the last close of the
+ *             device; for all other file descriptors, this occurs on each
+ *             close to prevent fd's from being closed out from under
+ *             operations currently in progress and blocked
+ *
+ * See Also:   file_vnode(), file_socket(), file_drop(), and the cautions
+ *             regarding their use and interaction with this function.
+ */
 void
 void
-fileproc_drain(struct proc *p, struct fileproc * fp)
+fileproc_drain(proc_t p, struct fileproc * fp)
 {
 {
+       struct vfs_context context;
+
+       context.vc_thread = proc_thread(p);     /* XXX */
+       context.vc_ucred = fp->f_fglob->fg_cred;
+
        fp->f_iocount-- ; /* (the one the close holds) */
 
        while (fp->f_iocount) {
        fp->f_iocount-- ; /* (the one the close holds) */
 
        while (fp->f_iocount) {
-               if (((fp->f_flags & FP_INSELECT)== FP_INSELECT)) {
-                       wait_queue_wakeup_all((wait_queue_t)fp->f_waddr, &selwait, THREAD_INTERRUPTED);
-               } else  {
-                       if (fp->f_fglob->fg_ops->fo_drain) {
-                               (*fp->f_fglob->fg_ops->fo_drain)(fp, p);
-                       }
+
+               lck_mtx_convert_spin(&p->p_fdmlock);
+
+               if (fp->f_fglob->fg_ops->fo_drain) {
+                       (*fp->f_fglob->fg_ops->fo_drain)(fp, &context);
+               }
+               if ((fp->f_flags & FP_INSELECT) == FP_INSELECT) {
+                       if (wait_queue_wakeup_all((wait_queue_t)fp->f_waddr, NULL, THREAD_INTERRUPTED) == KERN_INVALID_ARGUMENT)
+                               panic("bad wait queue for wait_queue_wakeup_all %p", fp->f_waddr);
+               } 
+               if ((fp->f_flags & FP_SELCONFLICT) == FP_SELCONFLICT) {
+                       if (wait_queue_wakeup_all(&select_conflict_queue, NULL, THREAD_INTERRUPTED) == KERN_INVALID_ARGUMENT)
+                               panic("bad select_conflict_queue");
                }
                p->p_fpdrainwait = 1;
 
                }
                p->p_fpdrainwait = 1;
 
-               msleep(&p->p_fpdrainwait, &p->p_fdmlock, PRIBIO, "fpdrain",0);
+               msleep(&p->p_fpdrainwait, &p->p_fdmlock, PRIBIO, "fpdrain", NULL);
 
 
-               //panic("successful wait after drain\n");
        }
        }
+#if DIAGNOSTIC
+       if ((fp->f_flags & FP_INSELECT) != 0)
+               panic("FP_INSELECT set on drained fp");
+#endif
+       if ((fp->f_flags & FP_SELCONFLICT) == FP_SELCONFLICT)
+               fp->f_flags &= ~FP_SELCONFLICT;
 }
 
 }
 
+
+/*
+ * fp_free
+ *
+ * Description:        Release the fd and free the fileproc associated with the fd
+ *             in the per process open file table of the specified process;
+ *             these values must correspond.
+ *
+ * Parameters: p                               Process containing fd
+ *             fd                              fd to be released
+ *             fp                              fileproc to be freed
+ *
+ * Returns:    0                               Success
+ *
+ * Notes:      XXX function should be void - no one interprets the returns
+ *             XXX code
+ */
 int
 int
-fp_free(struct proc * p, int fd, struct fileproc * fp)
+fp_free(proc_t p, int fd, struct fileproc * fp)
 {
 {
-        proc_fdlock(p);
+        proc_fdlock_spin(p);
        fdrelse(p, fd);
         proc_fdunlock(p);
 
        fg_free(fp->f_fglob);
        fdrelse(p, fd);
         proc_fdunlock(p);
 
        fg_free(fp->f_fglob);
-       FREE_ZONE(fp, sizeof *fp, M_FILEPROC);
+       fileproc_free(fp);
+       return(0);
 }
 
 
 /*
 }
 
 
 /*
- * Apply an advisory lock on a file descriptor.
+ * flock
+ *
+ * Description:        Apply an advisory lock on a file descriptor.
+ *
+ * Parameters: p                               Process making request
+ *             uap->fd                         fd on which the lock is to be
+ *                                             attempted
+ *             uap->how                        (Un)Lock bits, including type
+ *             retval                          Pointer to the call return area
+ *             
+ * Returns:    0                               Success
+ *     fp_getfvp:EBADF                         Bad file descriptor
+ *     fp_getfvp:ENOTSUP                       fd does not refer to a vnode
+ *     vnode_getwithref:???
+ *     VNOP_ADVLOCK:???
  *
  *
- * Just attempt to get a record lock of the requested type on
- * the entire file (l_whence = SEEK_SET, l_start = 0, l_len = 0).
+ * Implicit returns:
+ *             *retval (modified)              Size of dtable
+ *
+ * Notes:      Just attempt to get a record lock of the requested type on
+ *             the entire file (l_whence = SEEK_SET, l_start = 0, l_len = 0).
  */
 int
  */
 int
-flock(struct proc *p, register struct flock_args *uap, __unused register_t *retval)
+flock(proc_t p, struct flock_args *uap, __unused int32_t *retval)
 {
        int fd = uap->fd;
        int how = uap->how;
        struct fileproc *fp;
        struct vnode *vp;
        struct flock lf;
 {
        int fd = uap->fd;
        int how = uap->how;
        struct fileproc *fp;
        struct vnode *vp;
        struct flock lf;
-       struct vfs_context context;
+       vfs_context_t ctx = vfs_context_current();
        int error=0;
 
        AUDIT_ARG(fd, uap->fd);
        int error=0;
 
        AUDIT_ARG(fd, uap->fd);
@@ -2389,16 +5119,13 @@ flock(struct proc *p, register struct flock_args *uap, __unused register_t *retv
        }
        AUDIT_ARG(vnpath, vp, ARG_VNODE1);
 
        }
        AUDIT_ARG(vnpath, vp, ARG_VNODE1);
 
-       context.vc_proc = p;
-       context.vc_ucred = fp->f_cred;
-
        lf.l_whence = SEEK_SET;
        lf.l_start = 0;
        lf.l_len = 0;
        if (how & LOCK_UN) {
                lf.l_type = F_UNLCK;
                fp->f_flag &= ~FHASLOCK;
        lf.l_whence = SEEK_SET;
        lf.l_start = 0;
        lf.l_len = 0;
        if (how & LOCK_UN) {
                lf.l_type = F_UNLCK;
                fp->f_flag &= ~FHASLOCK;
-               error = VNOP_ADVLOCK(vp, (caddr_t)fp->f_fglob, F_UNLCK, &lf, F_FLOCK, &context);
+               error = VNOP_ADVLOCK(vp, (caddr_t)fp->f_fglob, F_UNLCK, &lf, F_FLOCK, ctx, NULL);
                goto out;
        }
        if (how & LOCK_EX)
                goto out;
        }
        if (how & LOCK_EX)
@@ -2409,12 +5136,17 @@ flock(struct proc *p, register struct flock_args *uap, __unused register_t *retv
                error = EBADF;
                goto out;
        }
                error = EBADF;
                goto out;
        }
+#if CONFIG_MACF
+       error = mac_file_check_lock(proc_ucred(p), fp->f_fglob, F_SETLK, &lf);
+       if (error)
+               goto out;
+#endif
        fp->f_flag |= FHASLOCK;
        if (how & LOCK_NB) {
        fp->f_flag |= FHASLOCK;
        if (how & LOCK_NB) {
-               error = VNOP_ADVLOCK(vp, (caddr_t)fp->f_fglob, F_SETLK, &lf, F_FLOCK, &context);
+               error = VNOP_ADVLOCK(vp, (caddr_t)fp->f_fglob, F_SETLK, &lf, F_FLOCK, ctx, NULL);
                goto out;       
        }
                goto out;       
        }
-       error = VNOP_ADVLOCK(vp, (caddr_t)fp->f_fglob, F_SETLK, &lf, F_FLOCK|F_WAIT, &context);
+       error = VNOP_ADVLOCK(vp, (caddr_t)fp->f_fglob, F_SETLK, &lf, F_FLOCK|F_WAIT, ctx, NULL);
 out:
        (void)vnode_put(vp);
 out1:
 out:
        (void)vnode_put(vp);
 out1:
@@ -2424,42 +5156,212 @@ out1:
 }
 
 /*
 }
 
 /*
- * File Descriptor pseudo-device driver (/dev/fd/).
+ * fileport_makeport
+ *
+ * Description: Obtain a Mach send right for a given file descriptor.
+ *
+ * Parameters: p               Process calling fileport
+ *             uap->fd         The fd to reference
+ *             uap->portnamep  User address at which to place port name.
  *
  *
- * Opening minor device N dup()s the file (if any) connected to file
- * descriptor N belonging to the calling process.  Note that this driver
- * consists of only the ``open()'' routine, because all subsequent
- * references to this file will be direct to the other driver.
+ * Returns:    0               Success.
+ *                     EBADF           Bad file descriptor.
+ *                     EINVAL          File descriptor had type that cannot be sent, misc. other errors.
+ *                     EFAULT          Address at which to store port name is not valid.
+ *                     EAGAIN          Resource shortage.
+ *
+ * Implicit returns:
+ *             On success, name of send right is stored at user-specified address.             
  */
 int
  */
 int
-fdopen(dev_t dev, __unused int mode, __unused int type, struct proc *p)
+fileport_makeport(proc_t p, struct fileport_makeport_args *uap,
+    __unused int *retval)
 {
 {
+       int err;
+       int fd = uap->fd;
+       user_addr_t user_portaddr = uap->portnamep;
+       struct fileproc *fp = FILEPROC_NULL;
+       struct fileglob *fg = NULL;
+       ipc_port_t fileport;
+       mach_port_name_t name = MACH_PORT_NULL;
+
+       err = fp_lookup(p, fd, &fp, 0);
+       if (err != 0) {
+               goto out;
+       }
 
 
-       /*
-        * XXX Kludge: set curproc->p_dupfd to contain the value of the
-        * the file descriptor being sought for duplication. The error 
-        * return ensures that the vnode for this device will be released
-        * by vn_open. Open will detect this special error and take the
-        * actions in dupfdopen below. Other callers of vn_open or vnop_open
-        * will simply report the error.
-        */
-       p->p_dupfd = minor(dev);
-       return (ENODEV);
+       if (!filetype_issendable(fp->f_type)) {
+               err = EINVAL;
+               goto out;
+       }
+
+       if (FP_ISGUARDED(fp, GUARD_FILEPORT)) {
+               proc_fdlock(p);
+               err = fp_guard_exception(p, fd, fp, kGUARD_EXC_FILEPORT);
+               proc_fdunlock(p);
+               goto out;
+       }
+
+       /* Dropped when port is deallocated */
+       fg = fp->f_fglob;
+       fg_ref(fp);
+
+       /* Allocate and initialize a port */
+       fileport = fileport_alloc(fg);
+       if (fileport == IPC_PORT_NULL) {
+               err = EAGAIN;
+               fg_drop(fp);
+               goto out;
+       }
+       
+       /* Add an entry.  Deallocates port on failure. */
+       name = ipc_port_copyout_send(fileport, get_task_ipcspace(p->task));
+       if (!MACH_PORT_VALID(name)) {
+               err = EINVAL;
+               goto out;
+       } 
+       
+       err = copyout(&name, user_portaddr, sizeof(mach_port_name_t));
+       if (err != 0) {
+               goto out;
+       }
+
+       /* Tag the fileglob for debugging purposes */
+       lck_mtx_lock_spin(&fg->fg_lock);
+       fg->fg_lflags |= FG_PORTMADE;
+       lck_mtx_unlock(&fg->fg_lock);
+
+       fp_drop(p, fd, fp, 0);
+
+       return 0;
+
+out:
+       if (MACH_PORT_VALID(name)) {
+               /* Don't care if another thread races us to deallocate the entry */
+               (void) mach_port_deallocate(get_task_ipcspace(p->task), name);
+       }
+
+       if (fp != FILEPROC_NULL) {
+               fp_drop(p, fd, fp, 0);
+       }
+
+       return err;
+}
+
+void
+fileport_releasefg(struct fileglob *fg)
+{
+       (void)closef_locked(NULL, fg, PROC_NULL);
+
+       return;
 }
 
 }
 
+
 /*
 /*
- * Duplicate the specified descriptor to a free descriptor.
+ * fileport_makefd
+ *
+ * Description: Obtain the file descriptor for a given Mach send right.
+ *
+ * Parameters: p               Process calling fileport
+ *             uap->port       Name of send right to file port.
+ *
+ * Returns:    0               Success
+ *             EINVAL          Invalid Mach port name, or port is not for a file.
+ *     fdalloc:EMFILE
+ *     fdalloc:ENOMEM          Unable to allocate fileproc or extend file table.
+ *
+ * Implicit returns:
+ *             *retval (modified)              The new descriptor
  */
 int
  */
 int
-dupfdopen(fdp, indx, dfd, mode, error)
-       register struct filedesc *fdp;
-       register int indx, dfd;
-       int mode;
-       int error;
+fileport_makefd(proc_t p, struct fileport_makefd_args *uap, int32_t *retval)
+{
+       struct fileglob *fg;
+       struct fileproc *fp = FILEPROC_NULL;
+       ipc_port_t port = IPC_PORT_NULL;
+       mach_port_name_t send = uap->port;
+       kern_return_t res;
+       int fd;
+       int err;
+
+       res = ipc_object_copyin(get_task_ipcspace(p->task),
+                       send, MACH_MSG_TYPE_COPY_SEND, &port);
+
+       if (res != KERN_SUCCESS) {
+               err = EINVAL;
+               goto out;
+       }
+
+       fg = fileport_port_to_fileglob(port);
+       if (fg == NULL) {
+               err = EINVAL;
+               goto out;
+       }
+
+       fp = fileproc_alloc_init(NULL);
+       if (fp == FILEPROC_NULL) {
+               err = ENOMEM;
+               goto out;
+       }
+
+       fp->f_fglob = fg;
+       fg_ref(fp);
+
+       proc_fdlock(p);
+       err = fdalloc(p, 0, &fd);
+       if (err != 0) {
+               proc_fdunlock(p);
+               goto out;
+       }
+       *fdflags(p, fd) |= UF_EXCLOSE;
+
+       procfdtbl_releasefd(p, fd, fp);
+       proc_fdunlock(p);
+
+       *retval = fd;
+       err = 0;
+out:
+       if ((fp != NULL) && (0 != err)) {
+               fileproc_free(fp);
+       } 
+
+       if (IPC_PORT_NULL != port) {
+               ipc_port_release_send(port);
+       }
+
+       return err;
+}
+
+
+/*
+ * dupfdopen
+ *
+ * Description:        Duplicate the specified descriptor to a free descriptor;
+ *             this is the second half of fdopen(), above.
+ *
+ * Parameters: fdp                             filedesc pointer to fill in
+ *             indx                            fd to dup to
+ *             dfd                             fd to dup from
+ *             mode                            mode to set on new fd
+ *             error                           command code
+ *
+ * Returns:    0                               Success
+ *             EBADF                           Source fd is bad
+ *             EACCES                          Requested mode not allowed
+ *             !0                              'error', if not ENODEV or
+ *                                             ENXIO
+ *
+ * Notes:      XXX This is not thread safe; see fdopen() above
+ */
+int
+dupfdopen(struct filedesc *fdp, int indx, int dfd, int flags, int error)
 {
        struct fileproc *wfp;
        struct fileproc *fp;
 {
        struct fileproc *wfp;
        struct fileproc *fp;
-       struct proc * p = current_proc();
+#if CONFIG_MACF
+       int myerror;
+#endif
+       proc_t p = current_proc();
 
        /*
         * If the to-be-dup'd fd number is greater than the allowed number
 
        /*
         * If the to-be-dup'd fd number is greater than the allowed number
@@ -2478,6 +5380,13 @@ dupfdopen(fdp, indx, dfd, mode, error)
                proc_fdunlock(p);
                return (EBADF);
        }
                proc_fdunlock(p);
                return (EBADF);
        }
+#if CONFIG_MACF
+       myerror = mac_file_check_dup(proc_ucred(p), wfp->f_fglob, dfd);
+       if (myerror) {
+               proc_fdunlock(p);
+               return (myerror);
+       }
+#endif
        /*
         * There are two cases of interest here.
         *
        /*
         * There are two cases of interest here.
         *
@@ -2492,11 +5401,18 @@ dupfdopen(fdp, indx, dfd, mode, error)
         */
        switch (error) {
        case ENODEV:
         */
        switch (error) {
        case ENODEV:
+               if (FP_ISGUARDED(wfp, GUARD_DUP)) {
+                       int err = fp_guard_exception(p,
+                           dfd, wfp, kGUARD_EXC_DUP);
+                       proc_fdunlock(p);
+                       return (err);
+               }
+
                /*
                 * Check that the mode the file is being opened for is a
                 * subset of the mode of the existing descriptor.
                 */
                /*
                 * Check that the mode the file is being opened for is a
                 * subset of the mode of the existing descriptor.
                 */
-               if (((mode & (FREAD|FWRITE)) | wfp->f_flag) != wfp->f_flag) {
+               if (((flags & (FREAD|FWRITE)) | wfp->f_flag) != wfp->f_flag) {
                        proc_fdunlock(p);
                        return (EACCES);
                }
                        proc_fdunlock(p);
                        return (EACCES);
                }
@@ -2508,29 +5424,10 @@ dupfdopen(fdp, indx, dfd, mode, error)
                        fg_free(fp->f_fglob);
                fp->f_fglob = wfp->f_fglob;
 
                        fg_free(fp->f_fglob);
                fp->f_fglob = wfp->f_fglob;
 
-               fdp->fd_ofileflags[indx] = fdp->fd_ofileflags[dfd];
-
-               proc_fdunlock(p);
-               return (0);
-
-       case ENXIO:
-               /*
-                * Steal away the file pointer from dfd, and stuff it into indx.
-                */
-               if (indx > fdp->fd_lastfile)
-                       fdp->fd_lastfile = indx;
-
-               if (fp->f_fglob)
-                       fg_free(fp->f_fglob);
-               fp->f_fglob = wfp->f_fglob;
-
-               fdp->fd_ofileflags[indx] = fdp->fd_ofileflags[dfd];
-               _fdrelse(fdp, dfd);
+               fdp->fd_ofileflags[indx] = fdp->fd_ofileflags[dfd] |
+                       (flags & O_CLOEXEC) ? UF_EXCLOSE : 0;
 
                proc_fdunlock(p);
 
                proc_fdunlock(p);
-
-               FREE_ZONE(wfp, sizeof *fp, M_FILEPROC); 
-
                return (0);
 
        default:
                return (0);
 
        default:
@@ -2540,6 +5437,19 @@ dupfdopen(fdp, indx, dfd, mode, error)
        /* NOTREACHED */
 }
 
        /* NOTREACHED */
 }
 
+
+/*
+ * fg_ref
+ *
+ * Description:        Add a reference to a fileglob by fileproc
+ *
+ * Parameters: fp                              fileproc containing fileglob
+ *                                             pointer
+ *
+ * Returns:    void
+ *
+ * Notes:      XXX Should use OSAddAtomic?
+ */
 void
 fg_ref(struct fileproc * fp)
 {
 void
 fg_ref(struct fileproc * fp)
 {
@@ -2547,33 +5457,68 @@ fg_ref(struct fileproc * fp)
 
        fg = fp->f_fglob;
 
 
        fg = fp->f_fglob;
 
-       lck_mtx_lock(&fg->fg_lock);
+       lck_mtx_lock_spin(&fg->fg_lock);
+
+#if DIAGNOSTIC
+       if ((fp->f_flags & ~((unsigned int)FP_VALID_FLAGS)) != 0)
+               panic("fg_ref: invalid bits on fp %p", fp);
+
+       if (fg->fg_count == 0)
+               panic("fg_ref: adding fgcount to zeroed fg: fp %p fg %p",
+                   fp, fg);
+#endif
        fg->fg_count++;
        lck_mtx_unlock(&fg->fg_lock);
 }
 
        fg->fg_count++;
        lck_mtx_unlock(&fg->fg_lock);
 }
 
+
+/*
+ * fg_drop
+ *
+ * Description:        Remove a reference to a fileglob by fileproc
+ *
+ * Parameters: fp                              fileproc containing fileglob
+ *                                             pointer
+ *
+ * Returns:    void
+ *
+ * Notes:      XXX Should use OSAddAtomic?
+ */
 void
 fg_drop(struct fileproc * fp)
 {
        struct fileglob *fg;
 
        fg = fp->f_fglob;
 void
 fg_drop(struct fileproc * fp)
 {
        struct fileglob *fg;
 
        fg = fp->f_fglob;
-       lck_mtx_lock(&fg->fg_lock);
+       lck_mtx_lock_spin(&fg->fg_lock);
        fg->fg_count--;
        lck_mtx_unlock(&fg->fg_lock);
 }
 
        fg->fg_count--;
        lck_mtx_unlock(&fg->fg_lock);
 }
 
-
+#if SOCKETS
+/*
+ * fg_insertuipc
+ *
+ * Description:        Insert fileglob onto message queue
+ *
+ * Parameters: fg                              Fileglob pointer to insert
+ *
+ * Returns:    void
+ *
+ * Locks:      Takes and drops fg_lock, potentially many times
+ */
 void
 fg_insertuipc(struct fileglob * fg)
 {
 void
 fg_insertuipc(struct fileglob * fg)
 {
-int insertque = 0;
+       int insertque = 0;
 
 
-       lck_mtx_lock(&fg->fg_lock);
+       lck_mtx_lock_spin(&fg->fg_lock);
 
        while (fg->fg_lflags & FG_RMMSGQ) {
 
        while (fg->fg_lflags & FG_RMMSGQ) {
+               lck_mtx_convert_spin(&fg->fg_lock);
+
                fg->fg_lflags |= FG_WRMMSGQ;
                fg->fg_lflags |= FG_WRMMSGQ;
-               msleep(&fg->fg_lflags, &fg->fg_lock, 0, "fg_insertuipc", 0);
+               msleep(&fg->fg_lflags, &fg->fg_lock, 0, "fg_insertuipc", NULL);
        }
 
        fg->fg_count++;
        }
 
        fg->fg_count++;
@@ -2585,7 +5530,8 @@ int insertque = 0;
        lck_mtx_unlock(&fg->fg_lock);
 
        if (insertque) {
        lck_mtx_unlock(&fg->fg_lock);
 
        if (insertque) {
-               lck_mtx_lock(uipc_lock);
+               lck_mtx_lock_spin(uipc_lock);
+               unp_gc_wait();
                LIST_INSERT_HEAD(&fmsghead, fg, f_msglist);
                lck_mtx_unlock(uipc_lock);
                lck_mtx_lock(&fg->fg_lock);
                LIST_INSERT_HEAD(&fmsghead, fg, f_msglist);
                lck_mtx_unlock(uipc_lock);
                lck_mtx_lock(&fg->fg_lock);
@@ -2599,15 +5545,29 @@ int insertque = 0;
 
 }
 
 
 }
 
+
+/*
+ * fg_removeuipc
+ *
+ * Description:        Remove fileglob from message queue
+ *
+ * Parameters: fg                              Fileglob pointer to remove
+ *
+ * Returns:    void
+ *
+ * Locks:      Takes and drops fg_lock, potentially many times
+ */
 void
 fg_removeuipc(struct fileglob * fg)
 {
 void
 fg_removeuipc(struct fileglob * fg)
 {
-int removeque = 0;
+       int removeque = 0;
 
 
-       lck_mtx_lock(&fg->fg_lock);
+       lck_mtx_lock_spin(&fg->fg_lock);
        while (fg->fg_lflags & FG_INSMSGQ) {
        while (fg->fg_lflags & FG_INSMSGQ) {
+               lck_mtx_convert_spin(&fg->fg_lock);
+
                fg->fg_lflags |= FG_WINSMSGQ;
                fg->fg_lflags |= FG_WINSMSGQ;
-               msleep(&fg->fg_lflags, &fg->fg_lock, 0, "fg_removeuipc", 0);
+               msleep(&fg->fg_lflags, &fg->fg_lock, 0, "fg_removeuipc", NULL);
        }
        fg->fg_msgcount--;
        if (fg->fg_msgcount == 0) {
        }
        fg->fg_msgcount--;
        if (fg->fg_msgcount == 0) {
@@ -2617,7 +5577,8 @@ int removeque = 0;
        lck_mtx_unlock(&fg->fg_lock);
 
        if (removeque) {
        lck_mtx_unlock(&fg->fg_lock);
 
        if (removeque) {
-               lck_mtx_lock(uipc_lock);
+               lck_mtx_lock_spin(uipc_lock);
+               unp_gc_wait();
                LIST_REMOVE(fg, f_msglist);
                lck_mtx_unlock(uipc_lock);
                lck_mtx_lock(&fg->fg_lock);
                LIST_REMOVE(fg, f_msglist);
                lck_mtx_unlock(uipc_lock);
                lck_mtx_lock(&fg->fg_lock);
@@ -2629,46 +5590,187 @@ int removeque = 0;
                lck_mtx_unlock(&fg->fg_lock);
        }
 }
                lck_mtx_unlock(&fg->fg_lock);
        }
 }
+#endif /* SOCKETS */
 
 
-
+/*
+ * fo_read
+ *
+ * Description:        Generic fileops read indirected through the fileops pointer
+ *             in the fileproc structure
+ *
+ * Parameters: fp                              fileproc structure pointer
+ *             uio                             user I/O structure pointer
+ *             flags                           FOF_ flags
+ *             ctx                             VFS context for operation
+ *
+ * Returns:    0                               Success
+ *             !0                              Errno from read
+ */
 int
 int
-fo_read(struct fileproc *fp, struct uio *uio, kauth_cred_t cred, int flags, struct proc *p)
+fo_read(struct fileproc *fp, struct uio *uio, int flags, vfs_context_t ctx)
 {
 {
-       return ((*fp->f_ops->fo_read)(fp, uio, cred, flags, p));
+       return ((*fp->f_ops->fo_read)(fp, uio, flags, ctx));
 }
 
 }
 
+
+/*
+ * fo_write
+ *
+ * Description:        Generic fileops write indirected through the fileops pointer
+ *             in the fileproc structure
+ *
+ * Parameters: fp                              fileproc structure pointer
+ *             uio                             user I/O structure pointer
+ *             flags                           FOF_ flags
+ *             ctx                             VFS context for operation
+ *
+ * Returns:    0                               Success
+ *             !0                              Errno from write
+ */
 int
 int
-fo_write(struct fileproc *fp, struct uio *uio, kauth_cred_t cred, int flags, struct proc *p)
+fo_write(struct fileproc *fp, struct uio *uio, int flags, vfs_context_t ctx)
 {
 {
-       return((*fp->f_ops->fo_write)(fp, uio, cred, flags, p));
+       return((*fp->f_ops->fo_write)(fp, uio, flags, ctx));
 }
 
 }
 
+
+/*
+ * fo_ioctl
+ *
+ * Description:        Generic fileops ioctl indirected through the fileops pointer
+ *             in the fileproc structure
+ *
+ * Parameters: fp                              fileproc structure pointer
+ *             com                             ioctl command
+ *             data                            pointer to internalized copy
+ *                                             of user space ioctl command
+ *                                             parameter data in kernel space
+ *             ctx                             VFS context for operation
+ *
+ * Returns:    0                               Success
+ *             !0                              Errno from ioctl
+ *
+ * Locks:      The caller is assumed to have held the proc_fdlock; this
+ *             function releases and reacquires this lock.  If the caller
+ *             accesses data protected by this lock prior to calling this
+ *             function, it will need to revalidate/reacquire any cached
+ *             protected data obtained prior to the call.
+ */
 int 
 int 
-fo_ioctl(struct fileproc *fp, u_long com, caddr_t data, struct proc *p)
+fo_ioctl(struct fileproc *fp, u_long com, caddr_t data, vfs_context_t ctx)
 {
 {
-int error;
+       int error;
 
 
-       proc_fdunlock(p);
-       error = (*fp->f_ops->fo_ioctl)(fp, com, data, p);
-       proc_fdlock(p);
+       proc_fdunlock(vfs_context_proc(ctx));
+       error = (*fp->f_ops->fo_ioctl)(fp, com, data, ctx);
+       proc_fdlock(vfs_context_proc(ctx));
        return(error);
 }       
 
        return(error);
 }       
 
+
+/*
+ * fo_select
+ *
+ * Description:        Generic fileops select indirected through the fileops pointer
+ *             in the fileproc structure
+ *
+ * Parameters: fp                              fileproc structure pointer
+ *             which                           select which
+ *             wql                             pointer to wait queue list
+ *             ctx                             VFS context for operation
+ *
+ * Returns:    0                               Success
+ *             !0                              Errno from select
+ */
 int
 int
-fo_select(struct fileproc *fp, int which, void *wql, struct proc *p)
+fo_select(struct fileproc *fp, int which, void *wql, vfs_context_t ctx)
 {       
 {       
-       return((*fp->f_ops->fo_select)(fp, which, wql, p));
+       return((*fp->f_ops->fo_select)(fp, which, wql, ctx));
 }
 
 }
 
+
+/*
+ * fo_close
+ *
+ * Description:        Generic fileops close indirected through the fileops pointer
+ *             in the fileproc structure
+ *
+ * Parameters: fp                              fileproc structure pointer for
+ *                                             file to close
+ *             ctx                             VFS context for operation
+ *
+ * Returns:    0                               Success
+ *             !0                              Errno from close
+ */
 int
 int
-fo_close(struct fileglob *fg, struct proc *p)
+fo_close(struct fileglob *fg, vfs_context_t ctx)
 {       
 {       
-       return((*fg->fg_ops->fo_close)(fg, p));
+       return((*fg->fg_ops->fo_close)(fg, ctx));
 }
 
 }
 
+
+/*
+ * fo_kqfilter
+ *
+ * Description:        Generic fileops kqueue filter indirected through the fileops
+ *             pointer in the fileproc structure
+ *
+ * Parameters: fp                              fileproc structure pointer
+ *             kn                              pointer to knote to filter on
+ *             ctx                             VFS context for operation
+ *
+ * Returns:    0                               Success
+ *             !0                              Errno from kqueue filter
+ */
 int
 int
-fo_kqfilter(struct fileproc *fp, struct knote *kn, struct proc *p)
+fo_kqfilter(struct fileproc *fp, struct knote *kn, vfs_context_t ctx)
+{
+        return ((*fp->f_ops->fo_kqfilter)(fp, kn, ctx));
+}
+
+/*
+ * The ability to send a file descriptor to another
+ * process is opt-in by file type.
+ */
+boolean_t
+filetype_issendable(file_type_t fdtype) 
+{
+       switch (fdtype) {
+               case DTYPE_VNODE:
+               case DTYPE_SOCKET:
+               case DTYPE_PIPE:
+               case DTYPE_PSXSHM:
+                       return TRUE;
+               default:
+                       /* DTYPE_KQUEUE, DTYPE_FSEVENTS, DTYPE_PSXSEM */
+                       return FALSE;
+       }
+}
+
+
+struct fileproc *
+fileproc_alloc_init(__unused void *arg)
 {
 {
-        return ((*fp->f_ops->fo_kqfilter)(fp, kn, p));
+       struct fileproc *fp;
+
+       MALLOC_ZONE(fp, struct fileproc *, sizeof (*fp), M_FILEPROC, M_WAITOK);
+       if (fp)
+               bzero(fp, sizeof (*fp));
+
+       return (fp);
 }
 
 }
 
+void
+fileproc_free(struct fileproc *fp)
+{
+       switch (FILEPROC_TYPE(fp)) {
+       case FTYPE_SIMPLE:
+               FREE_ZONE(fp, sizeof (*fp), M_FILEPROC);
+               break;
+       case FTYPE_GUARDED:
+               guarded_fileproc_free(fp);
+               break;
+       default:
+               panic("%s: corrupt fp %p flags %x", __func__, fp, fp->f_flags);
+       }
+}