]> git.saurik.com Git - apple/xnu.git/blobdiff - osfmk/ipc/mach_port.c
xnu-4570.71.2.tar.gz
[apple/xnu.git] / osfmk / ipc / mach_port.c
index 28d0fbc87d2fdc10c0be4d93f9bfd9a538f9e048..ced4e638475fffac43bf8b7b656d6c2062a41a46 100644 (file)
@@ -71,7 +71,6 @@
  */
 
 #include <mach_debug.h>
-#include <mach_rt.h>
 
 #include <mach/port.h>
 #include <mach/kern_return.h>
@@ -84,6 +83,7 @@
 #include <kern/counters.h>
 #include <kern/thread.h>
 #include <kern/kalloc.h>
+#include <kern/exc_guard.h>
 #include <mach/mach_port_server.h>
 #include <vm/vm_map.h>
 #include <vm/vm_kern.h>
 #include <ipc/ipc_pset.h>
 #include <ipc/ipc_right.h>
 #include <ipc/ipc_kmsg.h>
-#include <ipc/ipc_labelh.h>
 #include <kern/misc_protos.h>
 #include <security/mac_mach_internal.h>
 
-#include <mach/security_server.h>
+#if IMPORTANCE_INHERITANCE
+#include <ipc/ipc_importance.h>
+#endif
+
 
 /*
  * Forward declarations
@@ -114,12 +116,23 @@ void mach_port_names_helper(
 
 void mach_port_gst_helper(
        ipc_pset_t              pset,
-       ipc_port_t              port,
        ipc_entry_num_t         maxnames,
        mach_port_name_t        *names,
        ipc_entry_num_t         *actualp);
 
 
+kern_return_t
+mach_port_guard_exception(
+       mach_port_name_t        name,
+       uint64_t                inguard,
+       uint64_t                portguard,
+       unsigned                reason);
+
+/* Needs port locked */
+void mach_port_get_status_helper(
+       ipc_port_t              port,
+       mach_port_status_t      *status);
+
 /* Zeroed template of qos flags */
 
 static mach_port_qos_t qos_template;
@@ -128,6 +141,9 @@ static mach_port_qos_t      qos_template;
  *     Routine:        mach_port_names_helper
  *     Purpose:
  *             A helper function for mach_port_names.
+ *
+ *     Conditions:
+ *             Space containing entry is [at least] read-locked.
  */
 
 void
@@ -141,44 +157,51 @@ mach_port_names_helper(
 {
        ipc_entry_bits_t bits;
        ipc_port_request_index_t request;
-       mach_port_type_t type;
+       mach_port_type_t type = 0;
        ipc_entry_num_t actual;
+       ipc_port_t port;
 
        bits = entry->ie_bits;
        request = entry->ie_request;
-       if (bits & MACH_PORT_TYPE_SEND_RIGHTS) {
-               ipc_port_t port;
-               boolean_t died;
+       __IGNORE_WCASTALIGN(port = (ipc_port_t) entry->ie_object);
 
-               port = (ipc_port_t) entry->ie_object;
-               assert(port != IP_NULL);
+       if (bits & MACH_PORT_TYPE_RECEIVE) {
+               assert(IP_VALID(port));
 
-               /*
-                *      The timestamp serializes mach_port_names
-                *      with ipc_port_destroy.  If the port died,
-                *      but after mach_port_names started, pretend
-                *      that it isn't dead.
-                */
+               if (request != IE_REQ_NONE) {
+                       ip_lock(port);
+                       assert(ip_active(port));
+                       type |= ipc_port_request_type(port, name, request);
+                       ip_unlock(port);
+               }
 
-               ip_lock(port);
-               died = (!ip_active(port) &&
-                       IP_TIMESTAMP_ORDER(port->ip_timestamp, timestamp));
-               ip_unlock(port);
+       } else if (bits & MACH_PORT_TYPE_SEND_RIGHTS) {
+               mach_port_type_t reqtype;
 
-               if (died) {
-                       /* pretend this is a dead-name entry */
+               assert(IP_VALID(port));
+               ip_lock(port);
 
+               reqtype = (request != IE_REQ_NONE) ?
+                         ipc_port_request_type(port, name, request) : 0;
+               
+               /*
+                * If the port is alive, or was alive when the mach_port_names
+                * started, then return that fact.  Otherwise, pretend we found
+                * a dead name entry.
+                */
+               if (ip_active(port) || IP_TIMESTAMP_ORDER(timestamp, port->ip_timestamp)) {
+                       type |= reqtype;
+               } else {
                        bits &= ~(IE_BITS_TYPE_MASK);
                        bits |= MACH_PORT_TYPE_DEAD_NAME;
-                       if (request != 0)
+                       /* account for additional reference for dead-name notification */
+                       if (reqtype != 0)
                                bits++;
-                       request = 0;
                }
+               ip_unlock(port);
        }
 
-       type = IE_BITS_TYPE(bits);
-       if (request != 0)
-               type |= MACH_PORT_TYPE_DNREQUEST;
+       type |= IE_BITS_TYPE(bits);
 
        actual = *actualp;
        names[actual] = name;
@@ -211,7 +234,6 @@ mach_port_names(
        mach_port_type_t        **typesp,
        mach_msg_type_number_t  *typesCnt)
 {
-       ipc_tree_entry_t tentry;
        ipc_entry_t table;
        ipc_entry_num_t tsize;
        mach_port_index_t index;
@@ -228,7 +250,7 @@ mach_port_names(
        vm_map_copy_t memory2;  /* copied-in memory, for types */
 
        /* safe simplifying assumption */
-       assert_static(sizeof(mach_port_name_t) == sizeof(mach_port_type_t));
+       static_assert(sizeof(mach_port_name_t) == sizeof(mach_port_type_t));
 
        if (space == IS_NULL)
                return KERN_INVALID_TASK;
@@ -240,7 +262,7 @@ mach_port_names(
                vm_size_t size_needed;
 
                is_read_lock(space);
-               if (!space->is_active) {
+               if (!is_active(space)) {
                        is_read_unlock(space);
                        if (size != 0) {
                                kmem_free(ipc_kernel_map, addr1, size);
@@ -250,9 +272,10 @@ mach_port_names(
                }
 
                /* upper bound on number of names in the space */
-
-               bound = space->is_table_size + space->is_tree_total;
-               size_needed = round_page(bound * sizeof(mach_port_name_t));
+               bound = space->is_table_size;
+               size_needed = vm_map_round_page(
+                       (bound * sizeof(mach_port_name_t)),
+                       VM_MAP_PAGE_MASK(ipc_kernel_map));
 
                if (size_needed <= size)
                        break;
@@ -265,11 +288,11 @@ mach_port_names(
                }
                size = size_needed;
 
-               kr = vm_allocate(ipc_kernel_map, &addr1, size, VM_FLAGS_ANYWHERE);
+               kr = vm_allocate_kernel(ipc_kernel_map, &addr1, size, VM_FLAGS_ANYWHERE, VM_KERN_MEMORY_IPC);
                if (kr != KERN_SUCCESS)
                        return KERN_RESOURCE_SHORTAGE;
 
-               kr = vm_allocate(ipc_kernel_map, &addr2, size, VM_FLAGS_ANYWHERE);
+               kr = vm_allocate_kernel(ipc_kernel_map, &addr2, size, VM_FLAGS_ANYWHERE, VM_KERN_MEMORY_IPC);
                if (kr != KERN_SUCCESS) {
                        kmem_free(ipc_kernel_map, addr1, size);
                        return KERN_RESOURCE_SHORTAGE;
@@ -277,18 +300,29 @@ mach_port_names(
 
                /* can't fault while we hold locks */
 
-               kr = vm_map_wire(ipc_kernel_map, vm_map_trunc_page(addr1),
-                                vm_map_round_page(addr1 + size),
-                                VM_PROT_READ|VM_PROT_WRITE, FALSE);
+               kr = vm_map_wire_kernel(
+                       ipc_kernel_map,
+                       vm_map_trunc_page(addr1,
+                                         VM_MAP_PAGE_MASK(ipc_kernel_map)),
+                       vm_map_round_page(addr1 + size,
+                                         VM_MAP_PAGE_MASK(ipc_kernel_map)),
+                       VM_PROT_READ|VM_PROT_WRITE, VM_KERN_MEMORY_IPC,
+                       FALSE);
                if (kr != KERN_SUCCESS) {
                        kmem_free(ipc_kernel_map, addr1, size);
                        kmem_free(ipc_kernel_map, addr2, size);
                        return KERN_RESOURCE_SHORTAGE;
                }
 
-               kr = vm_map_wire(ipc_kernel_map, vm_map_trunc_page(addr2),
-                                vm_map_round_page(addr2 + size),
-                                VM_PROT_READ|VM_PROT_WRITE, FALSE);
+               kr = vm_map_wire_kernel(
+                       ipc_kernel_map,
+                       vm_map_trunc_page(addr2,
+                                         VM_MAP_PAGE_MASK(ipc_kernel_map)),
+                       vm_map_round_page(addr2 + size,
+                                         VM_MAP_PAGE_MASK(ipc_kernel_map)),
+                       VM_PROT_READ|VM_PROT_WRITE,
+                       VM_KERN_MEMORY_IPC,
+                       FALSE);
                if (kr != KERN_SUCCESS) {
                        kmem_free(ipc_kernel_map, addr1, size);
                        kmem_free(ipc_kernel_map, addr2, size);
@@ -320,17 +354,6 @@ mach_port_names(
                }
        }
 
-       for (tentry = ipc_splay_traverse_start(&space->is_tree);
-           tentry != ITE_NULL;
-           tentry = ipc_splay_traverse_next(&space->is_tree, FALSE)) {
-               ipc_entry_t entry = &tentry->ite_entry;
-               mach_port_name_t name = tentry->ite_name;
-
-               assert(IE_BITS_TYPE(tentry->ite_bits) != MACH_PORT_TYPE_NONE);
-               mach_port_names_helper(timestamp, entry, name, names,
-                                      types, &actual);
-       }
-       ipc_splay_traverse_finish(&space->is_tree);
        is_read_unlock(space);
 
        if (actual == 0) {
@@ -346,19 +369,31 @@ mach_port_names(
                vm_size_t vm_size_used;
 
                size_used = actual * sizeof(mach_port_name_t);
-               vm_size_used = round_page(size_used);
+               vm_size_used =
+                       vm_map_round_page(size_used,
+                                         VM_MAP_PAGE_MASK(ipc_kernel_map));
 
                /*
                 *      Make used memory pageable and get it into
                 *      copied-in form.  Free any unused memory.
                 */
 
-               kr = vm_map_unwire(ipc_kernel_map, vm_map_trunc_page(addr1),
-                                  vm_map_round_page(addr1 + vm_size_used), FALSE);
+               kr = vm_map_unwire(
+                       ipc_kernel_map,
+                       vm_map_trunc_page(addr1,
+                                         VM_MAP_PAGE_MASK(ipc_kernel_map)),
+                       vm_map_round_page(addr1 + vm_size_used,
+                                         VM_MAP_PAGE_MASK(ipc_kernel_map)),
+                       FALSE);
                assert(kr == KERN_SUCCESS);
 
-               kr = vm_map_unwire(ipc_kernel_map, vm_map_trunc_page(addr2),
-                                  vm_map_round_page(addr2 + vm_size_used), FALSE);
+               kr = vm_map_unwire(
+                       ipc_kernel_map,
+                       vm_map_trunc_page(addr2,
+                                         VM_MAP_PAGE_MASK(ipc_kernel_map)),
+                       vm_map_round_page(addr2 + vm_size_used,
+                                         VM_MAP_PAGE_MASK(ipc_kernel_map)),
+                       FALSE);
                assert(kr == KERN_SUCCESS);
 
                kr = vm_map_copyin(ipc_kernel_map, (vm_map_address_t)addr1,
@@ -431,12 +466,16 @@ mach_port_type(
        kr = ipc_right_lookup_write(space, name, &entry);
        if (kr != KERN_SUCCESS)
                return kr;
-       /* space is write-locked and active */
 
+       /* space is write-locked and active */
        kr = ipc_right_info(space, name, entry, typep, &urefs);
-       if (kr == KERN_SUCCESS)
-               is_write_unlock(space);
        /* space is unlocked */
+
+#if 1
+        /* JMM - workaround rdar://problem/9121297 (CF being too picky on these bits). */
+        *typep &= ~(MACH_PORT_TYPE_SPREQUEST | MACH_PORT_TYPE_SPREQUEST_DELAYED);
+#endif
+
        return kr;
 }
 
@@ -455,26 +494,21 @@ mach_port_type(
  *             KERN_INVALID_VALUE      The nname isn't a legal name.
  *             KERN_NAME_EXISTS        The nname already denotes a right.
  *             KERN_RESOURCE_SHORTAGE  Couldn't allocate memory.
+ *
+ *      This interface is obsolete and always returns
+ *      KERN_NOT_SUPPORTED.
  */
 
 kern_return_t
 mach_port_rename(
-       ipc_space_t             space,
-       mach_port_name_t        oname,
-       mach_port_name_t        nname)
+       __unused ipc_space_t            space,
+       __unused mach_port_name_t       oname,
+       __unused mach_port_name_t       nname)
 {
-       if (space == IS_NULL)
-               return KERN_INVALID_TASK;
-
-       if (!MACH_PORT_VALID(oname))
-               return KERN_INVALID_NAME;
-
-       if (!MACH_PORT_VALID(nname))
-               return KERN_INVALID_VALUE;
-
-       return ipc_object_rename(space, oname, nname);
+       return KERN_NOT_SUPPORTED;
 }
 
+
 /*
  *     Routine:        mach_port_allocate_name [kernel call]
  *     Purpose:
@@ -635,8 +669,6 @@ mach_port_allocate_full(
        if (qosp->name) {
                if (!MACH_PORT_VALID (*namep))
                        return (KERN_INVALID_VALUE);
-               if (is_fast_space (space))
-                       return (KERN_FAILURE);
        }
 
        if (qosp->prealloc) {
@@ -644,9 +676,11 @@ mach_port_allocate_full(
                        return KERN_RESOURCE_SHORTAGE;
                } else {
                        mach_msg_size_t size = qosp->len + MAX_TRAILER_SIZE;
+
                        if (right != MACH_PORT_RIGHT_RECEIVE)
                                return (KERN_INVALID_VALUE);
-                       kmsg = (ipc_kmsg_t)ipc_kmsg_alloc(size);
+
+                       kmsg = (ipc_kmsg_t)ipc_kmsg_prealloc(size);
                        if (kmsg == IKM_NULL)
                                return (KERN_RESOURCE_SHORTAGE);
                }
@@ -733,8 +767,7 @@ mach_port_destroy(
                return kr;
        /* space is write-locked and active */
 
-       kr = ipc_right_destroy(space, name, entry); 
-       is_write_unlock(space);
+       kr = ipc_right_destroy(space, name, entry, TRUE, 0); /* unlocks space */
        return kr;
 }
 
@@ -742,8 +775,8 @@ mach_port_destroy(
  *     Routine:        mach_port_deallocate [kernel call]
  *     Purpose:
  *             Deallocates a user reference from a send right,
- *             send-once right, or a dead-name right.  May
- *             deallocate the right, if this is the last uref,
+ *             send-once right, dead-name right or a port_set right.
+ *             May deallocate the right, if this is the last uref,
  *             and destroy the name, if it doesn't denote
  *             other rights.
  *     Conditions:
@@ -826,12 +859,13 @@ mach_port_get_refs(
        kr = ipc_right_lookup_write(space, name, &entry);
        if (kr != KERN_SUCCESS)
                return kr;
+
        /* space is write-locked and active */
+       kr = ipc_right_info(space, name, entry, &type, &urefs);
+       /* space is unlocked */
 
-       kr = ipc_right_info(space, name, entry, &type, &urefs); /* unlocks */
        if (kr != KERN_SUCCESS)
-               return kr;      /* space is unlocked */
-       is_write_unlock(space);
+               return kr;      
 
        if (type & MACH_PORT_TYPE(right))
                switch (right) {
@@ -912,6 +946,95 @@ mach_port_mod_refs(
 }
 
 
+/*
+ *     Routine:        mach_port_peek [kernel call]
+ *     Purpose:
+ *             Peek at the message queue for the specified receive
+ *             right and return info about a message in the queue.
+ *
+ *             On input, seqnop points to a sequence number value
+ *             to match the message being peeked. If zero is specified
+ *             as the seqno, the first message in the queue will be
+ *             peeked.
+ *
+ *             Only the following trailer types are currently supported:
+ *                     MACH_RCV_TRAILER_TYPE(MACH_MSG_TRAILER_FORMAT_0)
+ *
+ *                             or'ed with one of these element types:
+ *                     MACH_RCV_TRAILER_ELEMENTS(MACH_RCV_TRAILER_NULL)
+ *                     MACH_RCV_TRAILER_ELEMENTS(MACH_RCV_TRAILER_SEQNO)
+ *                     MACH_RCV_TRAILER_ELEMENTS(MACH_RCV_TRAILER_SENDER)
+ *                     MACH_RCV_TRAILER_ELEMENTS(MACH_RCV_TRAILER_AUDIT)
+ *
+ *             On input, the value pointed to by trailer_sizep must be
+ *             large enough to hold the requested trailer size.
+ *
+ *             The message sequence number, id, size, requested trailer info
+ *             and requested trailer size are returned in their respective
+ *             output parameters upon success.
+ *
+ *     Conditions:
+ *             Nothing locked.
+ *     Returns:
+ *             KERN_SUCCESS            Matching message found, out parameters set.
+ *             KERN_INVALID_TASK       The space is null or dead.
+ *             KERN_INVALID_NAME       The name doesn't denote a right.
+ *             KERN_INVALID_RIGHT      Name doesn't denote receive rights.
+ *             KERN_INVALID_VALUE      The input parameter values are out of bounds.
+ *             KERN_FAILURE            The requested message was not found.
+ */
+
+kern_return_t
+mach_port_peek(
+       ipc_space_t                     space,
+       mach_port_name_t                name,
+       mach_msg_trailer_type_t         trailer_type,
+       mach_port_seqno_t               *seqnop,
+       mach_msg_size_t                 *msg_sizep,
+       mach_msg_id_t                   *msg_idp,
+       mach_msg_trailer_info_t         trailer_infop,
+       mach_msg_type_number_t          *trailer_sizep)
+{
+       ipc_port_t port;
+       kern_return_t kr;
+       boolean_t found;
+       mach_msg_max_trailer_t max_trailer;
+
+       if (space == IS_NULL)
+               return KERN_INVALID_TASK;
+
+       if (!MACH_PORT_VALID(name))
+               return KERN_INVALID_RIGHT;
+
+       /*
+        * We don't allow anything greater than the audit trailer - to avoid
+        * leaking the context pointer and to avoid variable-sized context issues.
+        */
+       if (GET_RCV_ELEMENTS(trailer_type) > MACH_RCV_TRAILER_AUDIT ||
+           REQUESTED_TRAILER_SIZE(TRUE, trailer_type) > *trailer_sizep) 
+               return KERN_INVALID_VALUE;
+
+       *trailer_sizep = REQUESTED_TRAILER_SIZE(TRUE, trailer_type);
+
+       kr = ipc_port_translate_receive(space, name, &port);
+       if (kr != KERN_SUCCESS)
+               return kr;
+
+       /* Port locked and active */
+
+       found = ipc_mqueue_peek(&port->ip_messages, seqnop,
+                               msg_sizep, msg_idp, &max_trailer, NULL);
+       ip_unlock(port);
+
+       if (found != TRUE)
+               return KERN_FAILURE;
+
+       max_trailer.msgh_seqno = *seqnop;
+       memcpy(trailer_infop, &max_trailer, *trailer_sizep);
+
+       return KERN_SUCCESS;
+}
+
 /*
  *     Routine:        mach_port_set_mscount [kernel call]
  *     Purpose:
@@ -1010,7 +1133,7 @@ kern_return_t
 mach_port_get_context(
        ipc_space_t             space,
        mach_port_name_t        name,
-       mach_vm_address_t       *context)
+       mach_vm_address_t       *context)
 {
        ipc_port_t port;
        kern_return_t kr;
@@ -1025,8 +1148,13 @@ mach_port_get_context(
        if (kr != KERN_SUCCESS)
                return kr;
 
-       /* port is locked and active */
-       *context = port->ip_context;
+       /* Port locked and active */
+
+       /* For strictly guarded ports, return empty context (which acts as guard) */
+       if (port->ip_strict_guard)
+               *context = 0;
+       else
+               *context = port->ip_context;
 
        ip_unlock(port);
        return KERN_SUCCESS;
@@ -1051,7 +1179,7 @@ kern_return_t
 mach_port_set_context(
        ipc_space_t             space,
        mach_port_name_t        name,
-       mach_vm_address_t       context)
+       mach_vm_address_t       context)
 {
        ipc_port_t port;
        kern_return_t kr;
@@ -1067,6 +1195,14 @@ mach_port_set_context(
                return kr;
 
        /* port is locked and active */
+       if(port->ip_strict_guard) {
+               uint64_t portguard = port->ip_context;
+               ip_unlock(port);
+               /* For strictly guarded ports, disallow overwriting context; Raise Exception */
+               mach_port_guard_exception(name, context, portguard, kGUARD_EXC_SET_CONTEXT);
+               return KERN_INVALID_ARGUMENT;
+       }
+
        port->ip_context = context;
 
        ip_unlock(port);
@@ -1074,42 +1210,6 @@ mach_port_set_context(
 }
 
 
-/*
- *     Routine:        mach_port_gst_helper
- *     Purpose:
- *             A helper function for mach_port_get_set_status.
- */
-
-void
-mach_port_gst_helper(
-       ipc_pset_t              pset,
-       ipc_port_t              port,
-       ipc_entry_num_t         maxnames,
-       mach_port_name_t        *names,
-       ipc_entry_num_t         *actualp)
-{
-       mach_port_name_t name;
-
-       assert(port != IP_NULL);
-
-       ip_lock(port);
-       assert(ip_active(port));
-
-       name = port->ip_receiver_name;
-       assert(name != MACH_PORT_NULL);
-
-       ip_unlock(port);
-
-       if (ipc_pset_member(pset, port)) {
-               ipc_entry_num_t actual = *actualp;
-
-               if (actual < maxnames)
-                       names[actual] = name;
-
-               *actualp = actual+1;
-       }
-}
-
 /*
  *     Routine:        mach_port_get_set_status [kernel call]
  *     Purpose:
@@ -1147,86 +1247,51 @@ mach_port_get_set_status(
        if (!MACH_PORT_VALID(name))
                return KERN_INVALID_RIGHT;
 
-       size = PAGE_SIZE;       /* initial guess */
+       size = VM_MAP_PAGE_SIZE(ipc_kernel_map);        /* initial guess */
 
        for (;;) {
-               ipc_tree_entry_t tentry;
-               ipc_entry_t entry, table;
-               ipc_entry_num_t tsize;
-               mach_port_index_t index;
                mach_port_name_t *names;
+               ipc_object_t psobj;
                ipc_pset_t pset;
 
-               kr = vm_allocate(ipc_kernel_map, &addr, size, VM_FLAGS_ANYWHERE);
+               kr = vm_allocate_kernel(ipc_kernel_map, &addr, size, VM_FLAGS_ANYWHERE, VM_KERN_MEMORY_IPC);
                if (kr != KERN_SUCCESS)
                        return KERN_RESOURCE_SHORTAGE;
 
                /* can't fault while we hold locks */
 
-               kr = vm_map_wire(ipc_kernel_map, addr, addr + size,
-                                    VM_PROT_READ|VM_PROT_WRITE, FALSE);
+               kr = vm_map_wire_kernel(ipc_kernel_map, addr, addr + size,
+                                    VM_PROT_READ|VM_PROT_WRITE, VM_KERN_MEMORY_IPC, FALSE);
                assert(kr == KERN_SUCCESS);
 
-               kr = ipc_right_lookup_read(space, name, &entry);
+               kr = ipc_object_translate(space, name, MACH_PORT_RIGHT_PORT_SET, &psobj);
                if (kr != KERN_SUCCESS) {
                        kmem_free(ipc_kernel_map, addr, size);
                        return kr;
                }
-               /* space is read-locked and active */
 
-               if (IE_BITS_TYPE(entry->ie_bits) != MACH_PORT_TYPE_PORT_SET) {
-                       is_read_unlock(space);
-                       kmem_free(ipc_kernel_map, addr, size);
-                       return KERN_INVALID_RIGHT;
-               }
-
-               pset = (ipc_pset_t) entry->ie_object;
-               assert(pset != IPS_NULL);
-               /* the port set must be active */
+               /* just use a portset reference from here on out */
+               __IGNORE_WCASTALIGN(pset = (ipc_pset_t) psobj);
+               ips_reference(pset);
+               ips_unlock(pset); 
 
                names = (mach_port_name_t *) addr;
                maxnames = (ipc_entry_num_t)(size / sizeof(mach_port_name_t));
-               actual = 0;
-
-               table = space->is_table;
-               tsize = space->is_table_size;
-
-               for (index = 0; index < tsize; index++) {
-                       ipc_entry_t ientry = &table[index];
-
-                       if (ientry->ie_bits & MACH_PORT_TYPE_RECEIVE) {
-                               ipc_port_t port =
-                                       (ipc_port_t) ientry->ie_object;
-
-                               mach_port_gst_helper(pset, port,
-                                                    maxnames, names, &actual);
-                       }
-               }
-
-               for (tentry = ipc_splay_traverse_start(&space->is_tree);
-                   tentry != ITE_NULL;
-                   tentry = ipc_splay_traverse_next(&space->is_tree,FALSE)) {
-                       ipc_entry_bits_t bits = tentry->ite_bits;
 
-                       assert(IE_BITS_TYPE(bits) != MACH_PORT_TYPE_NONE);
+               ipc_mqueue_set_gather_member_names(space, &pset->ips_messages, maxnames, names, &actual);
 
-                       if (bits & MACH_PORT_TYPE_RECEIVE) {
-                           ipc_port_t port = (ipc_port_t) tentry->ite_object;
-
-                           mach_port_gst_helper(pset, port, maxnames,
-                                                names, &actual);
-                       }
-               }
-               ipc_splay_traverse_finish(&space->is_tree);
-               is_read_unlock(space);
+               /* release the portset reference */
+               ips_release(pset);
 
                if (actual <= maxnames)
                        break;
 
                /* didn't have enough memory; allocate more */
-
                kmem_free(ipc_kernel_map, addr, size);
-               size = round_page(actual * sizeof(mach_port_name_t)) + PAGE_SIZE;
+               size = vm_map_round_page(
+                       (actual * sizeof(mach_port_name_t)),
+                        VM_MAP_PAGE_MASK(ipc_kernel_map)) +
+                       VM_MAP_PAGE_SIZE(ipc_kernel_map);
        }
 
        if (actual == 0) {
@@ -1238,15 +1303,22 @@ mach_port_get_set_status(
                vm_size_t vm_size_used;
 
                size_used = actual * sizeof(mach_port_name_t);
-               vm_size_used = round_page(size_used);
+               vm_size_used = vm_map_round_page(
+                       size_used,
+                       VM_MAP_PAGE_MASK(ipc_kernel_map));
 
                /*
                 *      Make used memory pageable and get it into
                 *      copied-in form.  Free any unused memory.
                 */
 
-               kr = vm_map_unwire(ipc_kernel_map, vm_map_trunc_page(addr), 
-                                  vm_map_round_page(addr + vm_size_used), FALSE);
+               kr = vm_map_unwire(
+                       ipc_kernel_map,
+                       vm_map_trunc_page(addr,
+                                         VM_MAP_PAGE_MASK(ipc_kernel_map)), 
+                       vm_map_round_page(addr + vm_size_used,
+                                         VM_MAP_PAGE_MASK(ipc_kernel_map)),
+                       FALSE);
                assert(kr == KERN_SUCCESS);
 
                kr = vm_map_copyin(ipc_kernel_map, (vm_map_address_t)addr,
@@ -1294,6 +1366,8 @@ mach_port_move_member(
        ipc_port_t port;
        ipc_pset_t nset;
        kern_return_t kr;
+       uint64_t wq_link_id = 0;
+       uint64_t wq_reserved_prepost = 0;
 
        if (space == IS_NULL)
                return KERN_INVALID_TASK;
@@ -1301,20 +1375,37 @@ mach_port_move_member(
        if (!MACH_PORT_VALID(member))
                return KERN_INVALID_RIGHT;
 
-       if (after == MACH_PORT_DEAD)
+       if (after == MACH_PORT_DEAD) {
                return KERN_INVALID_RIGHT;
+       } else if (after == MACH_PORT_NULL) {
+               wq_link_id = 0;
+       } else {
+               /*
+                * We reserve both a link, and
+                * enough prepost objects to complete
+                * the set move atomically - we can't block
+                * while we're holding the space lock, and
+                * the ipc_pset_add calls ipc_mqueue_add
+                * which may have to prepost this port onto
+                * this set.
+                */
+               wq_link_id = waitq_link_reserve(NULL);
+               wq_reserved_prepost = waitq_prepost_reserve(NULL, 10,
+                                                           WAITQ_DONT_LOCK);
+       }
 
        kr = ipc_right_lookup_read(space, member, &entry);
        if (kr != KERN_SUCCESS)
-               return kr;
+               goto done;
        /* space is read-locked and active */
 
        if ((entry->ie_bits & MACH_PORT_TYPE_RECEIVE) == 0) {
                is_read_unlock(space);
-               return KERN_INVALID_RIGHT;
+               kr = KERN_INVALID_RIGHT;
+               goto done;
        }
 
-       port = (ipc_port_t) entry->ie_object;
+       __IGNORE_WCASTALIGN(port = (ipc_port_t) entry->ie_object);
        assert(port != IP_NULL);
 
        if (after == MACH_PORT_NULL)
@@ -1323,27 +1414,40 @@ mach_port_move_member(
                entry = ipc_entry_lookup(space, after);
                if (entry == IE_NULL) {
                        is_read_unlock(space);
-                       return KERN_INVALID_NAME;
+                       kr = KERN_INVALID_NAME;
+                       goto done;
                }
 
                if ((entry->ie_bits & MACH_PORT_TYPE_PORT_SET) == 0) {
                        is_read_unlock(space);
-                       return KERN_INVALID_RIGHT;
+                       kr = KERN_INVALID_RIGHT;
+                       goto done;
                }
 
-               nset = (ipc_pset_t) entry->ie_object;
+               __IGNORE_WCASTALIGN(nset = (ipc_pset_t) entry->ie_object);
                assert(nset != IPS_NULL);
        }
        ip_lock(port);
+       assert(ip_active(port));
        ipc_pset_remove_from_all(port);
 
        if (nset != IPS_NULL) {
                ips_lock(nset);
-               kr = ipc_pset_add(nset, port);
+               kr = ipc_pset_add(nset, port, &wq_link_id, &wq_reserved_prepost);
                ips_unlock(nset);
        }
        ip_unlock(port);
        is_read_unlock(space);
+
+ done:
+
+       /*
+        * on success the ipc_pset_add() will consume the wq_link_id
+        * value (resetting it to 0), so this function is always safe to call.
+        */
+       waitq_link_release(wq_link_id);
+       waitq_prepost_release_reserve(wq_reserved_prepost);
+
        return kr;
 }
 
@@ -1447,6 +1551,12 @@ mach_port_request_notification(
                        return kr;
                /* port is locked and active */
 
+               /* you cannot register for port death notifications on a kobject */
+               if (ip_kotype(port) != IKOT_NONE) {
+                       ip_unlock(port);
+                       return KERN_INVALID_RIGHT;
+               }
+
                ipc_port_pdrequest(port, notify, &previous);
                /* port is unlocked */
 
@@ -1470,6 +1580,18 @@ mach_port_request_notification(
                break;
            }
 
+           case MACH_NOTIFY_SEND_POSSIBLE:
+
+               if (!MACH_PORT_VALID(name)) {
+                       return KERN_INVALID_ARGUMENT;
+               }
+
+               kr = ipc_right_request_alloc(space, name, sync != 0,
+                                            TRUE, notify, previousp);
+               if (kr != KERN_SUCCESS)
+                       return kr;
+               break;
+
            case MACH_NOTIFY_DEAD_NAME:
 
                if (!MACH_PORT_VALID(name)) {
@@ -1481,8 +1603,8 @@ mach_port_request_notification(
                        return KERN_INVALID_ARGUMENT;
                }
 
-               kr = ipc_right_dnrequest(space, name, sync != 0,
-                                        notify, previousp);
+               kr = ipc_right_request_alloc(space, name, sync != 0,
+                                            FALSE, notify, previousp);
                if (kr != KERN_SUCCESS)
                        return kr;
                break;
@@ -1584,6 +1706,53 @@ mach_port_extract_right(
        return kr;
 }
 
+/*
+ *     Routine:        mach_port_get_status_helper [helper]
+ *     Purpose:
+ *             Populates a mach_port_status_t structure with
+ *             port information.
+ *     Conditions:
+ *             Port needs to be locked
+ *     Returns:
+ *             None.
+ */
+void mach_port_get_status_helper(
+       ipc_port_t              port,
+       mach_port_status_t      *statusp)
+{
+       imq_lock(&port->ip_messages);
+       /* don't leak set IDs, just indicate that the port is in one or not */
+       statusp->mps_pset = !!(port->ip_in_pset);
+       statusp->mps_seqno = port->ip_messages.imq_seqno;
+       statusp->mps_qlimit = port->ip_messages.imq_qlimit;
+       statusp->mps_msgcount = port->ip_messages.imq_msgcount;
+       imq_unlock(&port->ip_messages);
+
+       statusp->mps_mscount = port->ip_mscount;
+       statusp->mps_sorights = port->ip_sorights;
+       statusp->mps_srights = port->ip_srights > 0;
+       statusp->mps_pdrequest = port->ip_pdrequest != IP_NULL;
+       statusp->mps_nsrequest = port->ip_nsrequest != IP_NULL;
+       statusp->mps_flags = 0;
+       if (port->ip_impdonation) {
+               statusp->mps_flags |= MACH_PORT_STATUS_FLAG_IMP_DONATION;
+               if (port->ip_tempowner) {
+                       statusp->mps_flags |= MACH_PORT_STATUS_FLAG_TEMPOWNER;
+                       if (IIT_NULL != port->ip_imp_task) {
+                               statusp->mps_flags |= MACH_PORT_STATUS_FLAG_TASKPTR;
+                       }
+               }
+       }
+       if (port->ip_guarded) {
+               statusp->mps_flags |= MACH_PORT_STATUS_FLAG_GUARDED;
+               if (port->ip_strict_guard) {
+                       statusp->mps_flags |= MACH_PORT_STATUS_FLAG_STRICT_GUARD;
+               }
+       }
+       return;
+}
+
+
 
 kern_return_t
 mach_port_get_attributes(
@@ -1623,41 +1792,23 @@ mach_port_get_attributes(
         }
 
         case MACH_PORT_RECEIVE_STATUS: {
-                mach_port_status_t *statusp = (mach_port_status_t *)info;
-               spl_t s;
-                
-                if (*count < MACH_PORT_RECEIVE_STATUS_COUNT)
-                        return KERN_FAILURE;
-                
+               mach_port_status_t *statusp = (mach_port_status_t *)info;
+               
+               if (*count < MACH_PORT_RECEIVE_STATUS_COUNT)
+                       return KERN_FAILURE;
+
                if (!MACH_PORT_VALID(name))
                        return KERN_INVALID_RIGHT;
 
-                kr = ipc_port_translate_receive(space, name, &port);
-                if (kr != KERN_SUCCESS)
-                        return kr;
-                /* port is locked and active */
-
-               statusp->mps_pset = port->ip_pset_count;
-
-               s = splsched();
-               imq_lock(&port->ip_messages);
-               statusp->mps_seqno = port->ip_messages.imq_seqno;
-                statusp->mps_qlimit = port->ip_messages.imq_qlimit;
-                statusp->mps_msgcount = port->ip_messages.imq_msgcount;
-               imq_unlock(&port->ip_messages);
-               splx(s);
-
-                statusp->mps_mscount = port->ip_mscount;
-                statusp->mps_sorights = port->ip_sorights;
-                statusp->mps_srights = port->ip_srights > 0;
-                statusp->mps_pdrequest = port->ip_pdrequest != IP_NULL;
-                statusp->mps_nsrequest = port->ip_nsrequest != IP_NULL;
-               statusp->mps_flags = 0;
-
-                *count = MACH_PORT_RECEIVE_STATUS_COUNT;
-                ip_unlock(port);
-                break;
-        }
+               kr = ipc_port_translate_receive(space, name, &port);
+               if (kr != KERN_SUCCESS)
+                       return kr;
+               /* port is locked and active */
+               mach_port_get_status_helper(port, statusp);
+               *count = MACH_PORT_RECEIVE_STATUS_COUNT;
+               ip_unlock(port);
+               break;
+       }
        
        case MACH_PORT_DNREQUESTS_SIZE: {
                ipc_port_request_t      table;
@@ -1675,7 +1826,7 @@ mach_port_get_attributes(
                         return kr;
                 /* port is locked and active */
                
-               table = port->ip_dnrequests;
+               table = port->ip_requests;
                if (table == IPR_NULL)
                        *(int *)info = 0;
                else
@@ -1685,6 +1836,25 @@ mach_port_get_attributes(
                break;
        }
 
+       case MACH_PORT_INFO_EXT: {
+               mach_port_info_ext_t *mp_info = (mach_port_info_ext_t *)info;
+               if (*count < MACH_PORT_INFO_EXT_COUNT)
+                       return KERN_FAILURE;
+                       
+               if (!MACH_PORT_VALID(name))
+                       return KERN_INVALID_RIGHT;
+               
+               kr = ipc_port_translate_receive(space, name, &port);
+               if (kr != KERN_SUCCESS)
+                       return kr;
+               /* port is locked and active */
+               mach_port_get_status_helper(port, &mp_info->mpie_status);
+               mp_info->mpie_boost_cnt = port->ip_impcount;
+               *count = MACH_PORT_INFO_EXT_COUNT;
+               ip_unlock(port);
+               break;
+       }
+
         default:
                return KERN_INVALID_ARGUMENT;
                 /*NOTREACHED*/
@@ -1742,11 +1912,95 @@ mach_port_set_attributes(
                         return kr;
                 /* port is locked and active */
                
-               kr = ipc_port_dngrow(port, *(int *)info);
+               kr = ipc_port_request_grow(port, *(int *)info);
                if (kr != KERN_SUCCESS)
                        return kr;
                break;
        }
+       case MACH_PORT_TEMPOWNER:
+               if (!MACH_PORT_VALID(name))
+                       return KERN_INVALID_RIGHT;
+
+               ipc_importance_task_t release_imp_task = IIT_NULL;
+               natural_t assertcnt = 0;
+
+               kr = ipc_port_translate_receive(space, name, &port);
+               if (kr != KERN_SUCCESS)
+                       return kr;
+               /* port is locked and active */
+
+               /* 
+                * don't allow temp-owner importance donation if user
+                * associated it with a kobject already (timer, host_notify target),
+                * or is a special reply port.
+                */
+               if (is_ipc_kobject(ip_kotype(port)) || port->ip_specialreply) {
+                       ip_unlock(port);
+                       return KERN_INVALID_ARGUMENT;
+               }
+
+               if (port->ip_tempowner != 0) {
+                       if (IIT_NULL != port->ip_imp_task) {
+                               release_imp_task = port->ip_imp_task;
+                               port->ip_imp_task = IIT_NULL;
+                               assertcnt = port->ip_impcount;
+                       }
+               } else {
+                       assertcnt = port->ip_impcount;
+               }
+
+               port->ip_impdonation = 1;
+               port->ip_tempowner = 1;
+               ip_unlock(port);
+
+#if IMPORTANCE_INHERITANCE
+               /* drop assertions from previous destination task */
+               if (release_imp_task != IIT_NULL) {
+                       assert(ipc_importance_task_is_any_receiver_type(release_imp_task));
+                       if (assertcnt > 0)
+                               ipc_importance_task_drop_internal_assertion(release_imp_task, assertcnt);
+                       ipc_importance_task_release(release_imp_task);
+               } else if (assertcnt > 0) {
+                       release_imp_task = current_task()->task_imp_base;
+                       if (release_imp_task != IIT_NULL &&
+                           ipc_importance_task_is_any_receiver_type(release_imp_task)) {
+                               ipc_importance_task_drop_internal_assertion(release_imp_task, assertcnt);
+                       }
+               }
+#else
+               if (release_imp_task != IIT_NULL)
+                       ipc_importance_task_release(release_imp_task);
+#endif /* IMPORTANCE_INHERITANCE */
+
+               break;
+
+#if IMPORTANCE_INHERITANCE
+       case MACH_PORT_DENAP_RECEIVER:
+       case MACH_PORT_IMPORTANCE_RECEIVER:
+               if (!MACH_PORT_VALID(name))
+                       return KERN_INVALID_RIGHT;
+
+               kr = ipc_port_translate_receive(space, name, &port);
+               if (kr != KERN_SUCCESS)
+                       return kr;
+
+               /* 
+                * don't allow importance donation if user associated
+                * it with a kobject already (timer, host_notify target),
+                * or is a special reply port.
+                */
+               if (is_ipc_kobject(ip_kotype(port)) || port->ip_specialreply) {
+                       ip_unlock(port);
+                       return KERN_INVALID_ARGUMENT;
+               }
+
+               /* port is locked and active */
+               port->ip_impdonation = 1;
+               ip_unlock(port);
+
+               break;
+#endif /* IMPORTANCE_INHERITANCE */
+
         default:
                return KERN_INVALID_ARGUMENT;
                 /*NOTREACHED*/
@@ -1782,6 +2036,8 @@ mach_port_insert_member(
        ipc_object_t obj;
        ipc_object_t psobj;
        kern_return_t kr;
+       uint64_t wq_link_id;
+       uint64_t wq_reserved_prepost;
 
        if (space == IS_NULL)
                return KERN_INVALID_TASK;
@@ -1789,19 +2045,31 @@ mach_port_insert_member(
        if (!MACH_PORT_VALID(name) || !MACH_PORT_VALID(psname))
                return KERN_INVALID_RIGHT;
 
+       wq_link_id = waitq_link_reserve(NULL);
+       wq_reserved_prepost = waitq_prepost_reserve(NULL, 10,
+                                                   WAITQ_DONT_LOCK);
+
        kr = ipc_object_translate_two(space, 
                                      name, MACH_PORT_RIGHT_RECEIVE, &obj,
                                      psname, MACH_PORT_RIGHT_PORT_SET, &psobj);
        if (kr != KERN_SUCCESS)
-               return kr;
+               goto done;
 
        /* obj and psobj are locked (and were locked in that order) */
        assert(psobj != IO_NULL);
        assert(obj != IO_NULL);
 
-       kr = ipc_pset_add((ipc_pset_t)psobj, (ipc_port_t)obj);
+       __IGNORE_WCASTALIGN(kr = ipc_pset_add((ipc_pset_t)psobj, (ipc_port_t)obj,
+                                           &wq_link_id, &wq_reserved_prepost));
+
        io_unlock(psobj);
        io_unlock(obj);
+
+ done:
+       /* on success, wq_link_id is reset to 0, so this is always safe */
+       waitq_link_release(wq_link_id);
+       waitq_prepost_release_reserve(wq_reserved_prepost);
+
        return kr;
 }
 
@@ -1849,9 +2117,11 @@ mach_port_extract_member(
        assert(psobj != IO_NULL);
        assert(obj != IO_NULL);
 
-       kr = ipc_pset_remove((ipc_pset_t)psobj, (ipc_port_t)obj);
+       __IGNORE_WCASTALIGN(kr = ipc_pset_remove((ipc_pset_t)psobj, (ipc_port_t)obj));
+
        io_unlock(psobj);
        io_unlock(obj);
+
        return kr;
 }
 
@@ -1867,7 +2137,16 @@ task_set_port_space(
 {
        kern_return_t kr;
        
+       if (space == IS_NULL)
+               return KERN_INVALID_TASK;
+
        is_write_lock(space);
+
+       if (!is_active(space)) {
+               is_write_unlock(space);
+               return KERN_INVALID_TASK;
+       }
+
        kr = ipc_entry_grow_table(space, table_entries);
        if (kr == KERN_SUCCESS)
                is_write_unlock(space);
@@ -1875,176 +2154,332 @@ task_set_port_space(
 }
 
 /*
- * Get a (new) label handle representing the given port's port label.
+ *     Routine:        mach_port_guard_locked [helper routine]
+ *     Purpose:
+ *             Sets a new guard for a locked port.
+ *     Conditions:
+ *             Port Locked.
+ *     Returns:
+ *             KERN_SUCCESS            Port Guarded.
+ *             KERN_INVALID_ARGUMENT   Port already contains a context/guard.
  */
-#if CONFIG_MACF_MACH
-kern_return_t
-mach_get_label(
-       ipc_space_t             space,
-       mach_port_name_t        name,
-       mach_port_name_t        *outlabel)
+static kern_return_t
+mach_port_guard_locked(
+       ipc_port_t              port,
+       uint64_t                guard,
+       boolean_t               strict)
 {
-       ipc_entry_t entry;
-       ipc_port_t port;
-       struct label outl;
-       kern_return_t kr;
-       int dead;
-
-       if (!MACH_PORT_VALID(name))
-               return KERN_INVALID_NAME;
+       if (port->ip_context)
+               return KERN_INVALID_ARGUMENT;
 
-       /* Lookup the port name in the task's space. */
-       kr = ipc_right_lookup_write(space, name, &entry);
-       if (kr != KERN_SUCCESS)
-               return kr;
+       port->ip_context = guard;
+       port->ip_guarded = 1;
+       port->ip_strict_guard = (strict)?1:0;
+       return KERN_SUCCESS;
+}
 
-       port = (ipc_port_t) entry->ie_object;
-       dead = ipc_right_check(space, port, name, entry);
-       if (dead) {
-               is_write_unlock(space);
-               return KERN_INVALID_RIGHT;
+/*
+ *     Routine:        mach_port_unguard_locked [helper routine]
+ *     Purpose:
+ *             Removes guard for a locked port.
+ *     Conditions:
+ *             Port Locked.
+ *     Returns:
+ *             KERN_SUCCESS            Port Unguarded.
+ *             KERN_INVALID_ARGUMENT   Port is either unguarded already or guard mismatch.
+ *                                     This also raises a EXC_GUARD exception.
+ */
+static kern_return_t
+mach_port_unguard_locked(
+       ipc_port_t              port,
+       mach_port_name_t        name,
+       uint64_t                guard)
+{
+       /* Port locked and active */
+       if (!port->ip_guarded) {
+               /* Port already unguarded; Raise exception */
+               mach_port_guard_exception(name, guard, 0, kGUARD_EXC_UNGUARDED);
+               return KERN_INVALID_ARGUMENT;
        }
-       /* port is now locked */
 
-       is_write_unlock(space);
-       /* Make sure we are not dealing with a label handle. */
-       if (ip_kotype(port) == IKOT_LABELH) {
-               /* already is a label handle! */
-               ip_unlock(port);
+       if (port->ip_context != guard) {
+               /* Incorrect guard; Raise exception */
+               mach_port_guard_exception(name, guard, port->ip_context, kGUARD_EXC_INCORRECT_GUARD);
                return KERN_INVALID_ARGUMENT;
        }
 
-       /* Copy the port label and stash it in a new label handle. */
-       mac_port_label_init(&outl);
-       mac_port_label_copy(&port->ip_label, &outl); 
-       kr = labelh_new_user(space, &outl, outlabel);
-       ip_unlock(port);
-
+       port->ip_context = 0;
+       port->ip_guarded = port->ip_strict_guard = 0;
        return KERN_SUCCESS;
 }
-#else
+
+
+/*
+ *     Routine:        mach_port_guard_exception [helper routine]
+ *     Purpose:
+ *             Marks the thread with AST_GUARD for mach port guard violation.
+ *             Also saves exception info in thread structure.
+ *     Conditions:
+ *             None.
+ *     Returns:
+ *             KERN_FAILURE            Thread marked with AST_GUARD.
+ */
 kern_return_t
-mach_get_label(
-        __unused ipc_space_t           space,
-        __unused mach_port_name_t      name,
-        __unused mach_port_name_t      *outlabel)
+mach_port_guard_exception(
+       mach_port_name_t        name,
+       __unused uint64_t       inguard,
+       uint64_t                        portguard,
+       unsigned                        reason)
 {
-       return KERN_INVALID_ARGUMENT;
+       mach_exception_code_t code = 0;
+       EXC_GUARD_ENCODE_TYPE(code, GUARD_TYPE_MACH_PORT);
+       EXC_GUARD_ENCODE_FLAVOR(code, reason);
+       EXC_GUARD_ENCODE_TARGET(code, name);
+       mach_exception_subcode_t subcode = (uint64_t)portguard;
+       thread_t t = current_thread();
+       thread_guard_violation(t, code, subcode);
+       return KERN_FAILURE;
 }
-#endif
+
 
 /*
- * also works on label handles
+ *     Routine:        mach_port_guard_ast
+ *     Purpose:
+ *             Raises an exception for mach port guard violation.
+ *     Conditions:
+ *             None.
+ *     Returns:
+ *             None.
  */
-#if CONFIG_MACF_MACH
+
+void
+mach_port_guard_ast(thread_t __unused t,
+       mach_exception_data_type_t code, mach_exception_data_type_t subcode)
+{
+       assert(t->task != kernel_task);
+
+       /* Raise an EXC_GUARD exception */
+       task_exception_notify(EXC_GUARD, code, subcode);
+
+       /* Terminate task which caused the exception */
+       task_bsdtask_kill(current_task());
+}
+
+/*
+ *     Routine:        mach_port_construct [kernel call]
+ *     Purpose:
+ *             Constructs a mach port with the provided set of options.
+ *     Conditions:
+ *             None.
+ *     Returns:
+ *             KERN_SUCCESS            The right is allocated.
+ *             KERN_INVALID_TASK       The space is null.
+ *             KERN_INVALID_TASK       The space is dead.
+ *             KERN_RESOURCE_SHORTAGE  Couldn't allocate memory.
+ *             KERN_NO_SPACE           No room in space for another right.
+ *             KERN_FAILURE            Illegal option values requested.
+ */
+
 kern_return_t
-mach_get_label_text(
+mach_port_construct(
        ipc_space_t             space,
-       mach_port_name_t        name,
-       labelstr_t              policies,
-       labelstr_t              outlabel)
+       mach_port_options_t     *options,
+       uint64_t                context,
+       mach_port_name_t        *name)
 {
-       ipc_entry_t entry;
-       kern_return_t kr;
-       struct label *l;
-       int dead;
-
-       if (space == IS_NULL || space->is_task == NULL)
-               return KERN_INVALID_TASK;
+       kern_return_t           kr;
+       ipc_port_t              port;
 
-       if (!MACH_PORT_VALID(name))
-               return KERN_INVALID_NAME;
+       if (space == IS_NULL)
+               return (KERN_INVALID_TASK);
 
-       kr = ipc_right_lookup_write(space, name, &entry);
+       /* Allocate a new port in the IPC space */
+       kr = ipc_port_alloc(space, name, &port);
        if (kr != KERN_SUCCESS)
                return kr;
+       
+       /* Port locked and active */
+       if (options->flags & MPO_CONTEXT_AS_GUARD) {
+               kr = mach_port_guard_locked(port, (uint64_t) context, (options->flags & MPO_STRICT));
+               /* A newly allocated and locked port should always be guarded successfully */
+               assert(kr == KERN_SUCCESS);
+       } else {
+               port->ip_context = context;
+       }
+       
+       /* Unlock port */
+       ip_unlock(port);
 
-       dead = ipc_right_check(space, (ipc_port_t) entry->ie_object, name,
-           entry);
-       if (dead) {
-               is_write_unlock(space);
-               return KERN_INVALID_RIGHT;
+       /* Set port attributes as requested */
+
+       if (options->flags & MPO_QLIMIT) {
+               kr = mach_port_set_attributes(space, *name, MACH_PORT_LIMITS_INFO,
+                                             (mach_port_info_t)&options->mpl, sizeof(options->mpl)/sizeof(int));
+               if (kr != KERN_SUCCESS)
+                       goto cleanup;   
        }
-       /* object (port) is now locked */
 
-       is_write_unlock (space);
-       l = io_getlabel(entry->ie_object);
+       if (options->flags & MPO_TEMPOWNER) {
+               kr = mach_port_set_attributes(space, *name, MACH_PORT_TEMPOWNER, NULL, 0);
+               if (kr != KERN_SUCCESS)
+                       goto cleanup;
+       }
+
+       if (options->flags & MPO_IMPORTANCE_RECEIVER) {
+               kr = mach_port_set_attributes(space, *name, MACH_PORT_IMPORTANCE_RECEIVER, NULL, 0);
+               if (kr != KERN_SUCCESS)
+                       goto cleanup;
+       }
+
+       if (options->flags & MPO_DENAP_RECEIVER) {
+               kr = mach_port_set_attributes(space, *name, MACH_PORT_DENAP_RECEIVER, NULL, 0);
+               if (kr != KERN_SUCCESS)
+                       goto cleanup;
+       }
 
-       mac_port_label_externalize(l, policies, outlabel, 512, 0);
+       if (options->flags & MPO_INSERT_SEND_RIGHT) {
+               kr = ipc_object_copyin(space, *name, MACH_MSG_TYPE_MAKE_SEND, (ipc_object_t *)&port);
+               if (kr != KERN_SUCCESS)
+                       goto cleanup;
+
+               kr = mach_port_insert_right(space, *name, port, MACH_MSG_TYPE_PORT_SEND);
+               if (kr != KERN_SUCCESS)
+                       goto cleanup;
+       }
 
-       io_unlocklabel(entry->ie_object);
-       io_unlock(entry->ie_object);
        return KERN_SUCCESS;
+
+cleanup:
+       /* Attempt to destroy port. If its already destroyed by some other thread, we're done */
+       (void) mach_port_destruct(space, *name, 0, context);
+       return kr;
 }
-#else
-kern_return_t
-mach_get_label_text(
-       __unused ipc_space_t            space,
-       __unused mach_port_name_t       name,
-       __unused labelstr_t             policies,
-       __unused labelstr_t             outlabel)
-{
-       return KERN_INVALID_ARGUMENT;
-}
-#endif
 
+/*
+ *     Routine:        mach_port_destruct [kernel call]
+ *     Purpose:
+ *             Destroys a mach port with appropriate guard
+ *     Conditions:
+ *             None.
+ *     Returns:
+ *             KERN_SUCCESS            The name is destroyed.
+ *             KERN_INVALID_TASK       The space is null.
+ *             KERN_INVALID_TASK       The space is dead.
+ *             KERN_INVALID_NAME       The name doesn't denote a right.
+ *             KERN_INVALID_RIGHT      The right isn't correct.
+ *             KERN_INVALID_VALUE      The delta for send right is incorrect.
+ *             KERN_INVALID_ARGUMENT   Port is either unguarded already or guard mismatch.
+ *                                     This also raises a EXC_GUARD exception.
+ */
 
-#if CONFIG_MACF_MACH
 kern_return_t
-mach_set_port_label(
+mach_port_destruct(
        ipc_space_t             space,
        mach_port_name_t        name,
-       labelstr_t              labelstr)
+       mach_port_delta_t       srdelta,
+       uint64_t                guard)
 {
-       ipc_entry_t entry;
-       kern_return_t kr;
-       struct label inl;
-       ipc_port_t port;
-       int rc;
+       kern_return_t           kr;
+       ipc_entry_t             entry;
 
-       if (space == IS_NULL || space->is_task == NULL)
+       if (space == IS_NULL)
                return KERN_INVALID_TASK;
 
        if (!MACH_PORT_VALID(name))
                return KERN_INVALID_NAME;
 
-       mac_port_label_init(&inl);
-       rc = mac_port_label_internalize(&inl, labelstr);
-       if (rc)
-               return KERN_INVALID_ARGUMENT;
-
+       /* Remove reference for receive right */
        kr = ipc_right_lookup_write(space, name, &entry);
        if (kr != KERN_SUCCESS)
                return kr;
+       /* space is write-locked and active */
+       kr = ipc_right_destruct(space, name, entry, srdelta, guard);    /* unlocks */
 
-       if (io_otype(entMACry->ie_object) != IOT_PORT) {
-               is_write_unlock(space);
-               return KERN_INVALID_RIGHT;
-       }
+       return kr;
+}
 
-       port = (ipc_port_t) entry->ie_object;
-       ip_lock(port);
+/*
+ *     Routine:        mach_port_guard [kernel call]
+ *     Purpose:
+ *             Guard a mach port with specified guard value.
+ *             The context field of the port is used as the guard.
+ *     Conditions:
+ *             None.
+ *     Returns:
+ *             KERN_SUCCESS            The name is destroyed.
+ *             KERN_INVALID_TASK       The space is null.
+ *             KERN_INVALID_TASK       The space is dead.
+ *             KERN_INVALID_NAME       The name doesn't denote a right.
+ *             KERN_INVALID_RIGHT      The right isn't correct.
+ *             KERN_INVALID_ARGUMENT   Port already contains a context/guard.
+ */
+kern_return_t
+mach_port_guard(
+       ipc_space_t             space,
+       mach_port_name_t        name,
+       uint64_t                guard,
+       boolean_t               strict)
+{
+       kern_return_t           kr;
+       ipc_port_t              port;
 
-       tasklabel_lock(space->is_task);
-       rc = mac_port_check_label_update(&space->is_task->maclabel,
-                                   &port->ip_label, &inl);
-       tasklabel_unlock(space->is_task);
-       if (rc)
-               kr = KERN_NO_ACCESS;
-       else
-               mac_port_label_copy(&inl, &port->ip_label);
+       if (space == IS_NULL)
+               return KERN_INVALID_TASK;
+
+       if (!MACH_PORT_VALID(name))
+               return KERN_INVALID_NAME;
+
+       /* Guard can be applied only to receive rights */
+       kr = ipc_port_translate_receive(space, name, &port);
+       if (kr != KERN_SUCCESS)
+               return kr;
 
+       /* Port locked and active */
+       kr = mach_port_guard_locked(port, guard, strict);
        ip_unlock(port);
-       is_write_unlock(space);
+
        return kr;
+
 }
-#else
+
+/*
+ *     Routine:        mach_port_unguard [kernel call]
+ *     Purpose:
+ *             Unguard a mach port with specified guard value.
+ *     Conditions:
+ *             None.
+ *     Returns:
+ *             KERN_SUCCESS            The name is destroyed.
+ *             KERN_INVALID_TASK       The space is null.
+ *             KERN_INVALID_TASK       The space is dead.
+ *             KERN_INVALID_NAME       The name doesn't denote a right.
+ *             KERN_INVALID_RIGHT      The right isn't correct.
+ *             KERN_INVALID_ARGUMENT   Port is either unguarded already or guard mismatch.
+ *                                     This also raises a EXC_GUARD exception.
+ */
 kern_return_t
-mach_set_port_label(
-       ipc_space_t             space __unused,
-       mach_port_name_t        name __unused,
-       labelstr_t              labelstr __unused)
+mach_port_unguard(
+       ipc_space_t             space,
+       mach_port_name_t        name,
+       uint64_t                guard)
 {
-       return KERN_INVALID_ARGUMENT;
+       
+       kern_return_t           kr;
+       ipc_port_t              port;
+
+       if (space == IS_NULL)
+               return KERN_INVALID_TASK;
+
+       if (!MACH_PORT_VALID(name))
+               return KERN_INVALID_NAME;
+
+       kr = ipc_port_translate_receive(space, name, &port);
+       if (kr != KERN_SUCCESS)
+               return kr;
+
+       /* Port locked and active */
+       kr = mach_port_unguard_locked(port, name, guard);
+       ip_unlock(port);
+       return kr;
 }
-#endif
+