]> git.saurik.com Git - apple/xnu.git/blobdiff - osfmk/vm/vm_object.c
xnu-792.21.3.tar.gz
[apple/xnu.git] / osfmk / vm / vm_object.c
index a0f8bff17e79b0a7a90a680f82b23ad172eed0fa..d32573f825230cdf00615bf10c8cd3080a15e680 100644 (file)
@@ -1,16 +1,19 @@
 /*
- * Copyright (c) 2000-2001 Apple Computer, Inc. All rights reserved.
+ * Copyright (c) 2000-2004 Apple Computer, Inc. All rights reserved.
  *
- * @APPLE_LICENSE_HEADER_START@
- * 
- * Copyright (c) 1999-2003 Apple Computer, Inc.  All Rights Reserved.
+ * @APPLE_OSREFERENCE_LICENSE_HEADER_START@
  * 
  * 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. Please obtain a copy of the License at
- * http://www.opensource.apple.com/apsl/ and read it before using this
- * file.
+ * 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.
+ * 
+ * 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
@@ -20,7 +23,7 @@
  * Please see the License for the specific language governing rights and
  * limitations under the License.
  * 
- * @APPLE_LICENSE_HEADER_END@
+ * @APPLE_OSREFERENCE_LICENSE_HEADER_END@
  */
 /*
  * @OSF_COPYRIGHT@
  *     Virtual memory object module.
  */
 
-#ifdef MACH_BSD
-/* remove as part of compoenent support merge */
-extern int     vnode_pager_workaround;
-#endif
-
 #include <mach_pagemap.h>
 #include <task_swapper.h>
 
@@ -72,7 +70,11 @@ extern int   vnode_pager_workaround;
 #include <mach/memory_object_default.h>
 #include <mach/memory_object_control_server.h>
 #include <mach/vm_param.h>
+
+#include <ipc/ipc_types.h>
 #include <ipc/ipc_port.h>
+
+#include <kern/kern_types.h>
 #include <kern/assert.h>
 #include <kern/lock.h>
 #include <kern/queue.h>
@@ -81,15 +83,15 @@ extern int  vnode_pager_workaround;
 #include <kern/host.h>
 #include <kern/host_statistics.h>
 #include <kern/processor.h>
+#include <kern/misc_protos.h>
+
 #include <vm/memory_object.h>
 #include <vm/vm_fault.h>
 #include <vm/vm_map.h>
 #include <vm/vm_object.h>
 #include <vm/vm_page.h>
 #include <vm/vm_pageout.h>
-#include <kern/misc_protos.h>
-
-
+#include <vm/vm_protos.h>
 
 /*
  *     Virtual memory objects maintain the actual data
@@ -165,10 +167,6 @@ extern int vnode_pager_workaround;
  */
 
 /* Forward declarations for internal functions. */
-static void _vm_object_allocate(
-                               vm_object_size_t        size,
-                               vm_object_t             object);
-
 static kern_return_t   vm_object_terminate(
                                vm_object_t     object);
 
@@ -181,9 +179,6 @@ static vm_object_t  vm_object_cache_trim(
 static void            vm_object_deactivate_all_pages(
                                vm_object_t     object);
 
-static void            vm_object_abort_activity(
-                               vm_object_t     object);
-
 static kern_return_t   vm_object_copy_call(
                                vm_object_t             src_object,
                                vm_object_offset_t      src_offset,
@@ -289,6 +284,16 @@ typedef struct vm_object_hash_entry        *vm_object_hash_entry_t;
 #define vm_object_hash(pager) \
        ((((unsigned)pager) >> VM_OBJECT_HASH_SHIFT) % VM_OBJECT_HASH_COUNT)
 
+void vm_object_hash_entry_free(
+       vm_object_hash_entry_t  entry);
+
+static void vm_object_reap(vm_object_t object);
+static void vm_object_reap_async(vm_object_t object);
+static void vm_object_reaper_thread(void);
+static queue_head_t vm_object_reaper_queue; /* protected by vm_object_cache_lock() */
+unsigned int vm_object_reap_count = 0;
+unsigned int vm_object_reap_count_async = 0;
+
 /*
  *     vm_object_hash_lookup looks up a pager in the hashtable
  *     and returns the corresponding entry, with optional removal.
@@ -354,7 +359,7 @@ void
 vm_object_hash_entry_free(
        vm_object_hash_entry_t  entry)
 {
-       zfree(vm_object_hash_zone, (vm_offset_t)entry);
+       zfree(vm_object_hash_zone, entry);
 }
 
 /*
@@ -363,7 +368,7 @@ vm_object_hash_entry_free(
  *     Returns a new object with the given size.
  */
 
-static void
+__private_extern__ void
 _vm_object_allocate(
        vm_object_size_t        size,
        vm_object_t             object)
@@ -375,9 +380,9 @@ _vm_object_allocate(
        *object = vm_object_template;
        queue_init(&object->memq);
        queue_init(&object->msr_q);
-#ifdef UBC_DEBUG
+#ifdef UPL_DEBUG
        queue_init(&object->uplq);
-#endif /* UBC_DEBUG */
+#endif /* UPL_DEBUG */
        vm_object_lock_init(object);
        object->size = size;
 }
@@ -406,7 +411,7 @@ vm_object_allocate(
 __private_extern__ void
 vm_object_bootstrap(void)
 {
-       register        i;
+       register int    i;
 
        vm_object_zone = zinit((vm_size_t) sizeof(struct vm_object),
                                round_page_32(512*1024),
@@ -414,7 +419,7 @@ vm_object_bootstrap(void)
                                "vm objects");
 
        queue_init(&vm_object_cached_list);
-       mutex_init(&vm_object_cached_lock_data, ETAP_VM_OBJ_CACHE);
+       mutex_init(&vm_object_cached_lock_data, 0);
 
        vm_object_hash_zone =
                        zinit((vm_size_t) sizeof (struct vm_object_hash_entry),
@@ -431,7 +436,7 @@ vm_object_bootstrap(void)
 
        /* memq; Lock; init after allocation */
        vm_object_template.size = 0;
-       vm_object_template.frozen_size = 0;
+       vm_object_template.memq_hint = VM_PAGE_NULL;
        vm_object_template.ref_count = 1;
 #if    TASK_SWAPPER
        vm_object_template.res_count = 1;
@@ -440,12 +445,12 @@ vm_object_bootstrap(void)
        vm_object_template.copy = VM_OBJECT_NULL;
        vm_object_template.shadow = VM_OBJECT_NULL;
        vm_object_template.shadow_offset = (vm_object_offset_t) 0;
-       vm_object_template.cow_hint = 0;
+       vm_object_template.cow_hint = ~(vm_offset_t)0;
        vm_object_template.true_share = FALSE;
 
        vm_object_template.pager = MEMORY_OBJECT_NULL;
        vm_object_template.paging_offset = 0;
-       vm_object_template.pager_request = PAGER_REQUEST_NULL;
+       vm_object_template.pager_control = MEMORY_OBJECT_CONTROL_NULL;
        /* msr_q; init after allocation */
 
        vm_object_template.copy_strategy = MEMORY_OBJECT_COPY_SYMMETRIC;
@@ -464,8 +469,7 @@ vm_object_bootstrap(void)
        vm_object_template.private = FALSE;
        vm_object_template.pageout = FALSE;
        vm_object_template.alive = TRUE;
-       vm_object_template.lock_in_progress = FALSE;
-       vm_object_template.lock_restart = FALSE;
+       vm_object_template.purgable = VM_OBJECT_NONPURGABLE;
        vm_object_template.silent_overwrite = FALSE;
        vm_object_template.advisory_pageout = FALSE;
        vm_object_template.shadowed = FALSE;
@@ -535,6 +539,24 @@ vm_object_bootstrap(void)
 #endif /* MACH_PAGEMAP */
 }
 
+void
+vm_object_reaper_init(void)
+{
+       kern_return_t   kr;
+       thread_t        thread;
+
+       queue_init(&vm_object_reaper_queue);
+       kr = kernel_thread_start_priority(
+               (thread_continue_t) vm_object_reaper_thread,
+               NULL,
+               BASEPRI_PREEMPT - 1,
+               &thread);
+       if (kr != KERN_SUCCESS) {
+               panic("failed to launch vm_object_reaper_thread kr=0x%x", kr);
+       }
+       thread_deallocate(thread);
+}
+
 __private_extern__ void
 vm_object_init(void)
 {
@@ -572,7 +594,7 @@ vm_object_deallocate(
        register vm_object_t    object)
 {
        boolean_t retry_cache_trim = FALSE;
-       vm_object_t shadow;
+       vm_object_t shadow = VM_OBJECT_NULL;
        
 //     if(object)dbgLog(object, object->ref_count, object->can_persist, 3);    /* (TEST/DEBUG) */
 //     else dbgLog(object, 0, 0, 3);   /* (TEST/DEBUG) */
@@ -585,10 +607,21 @@ vm_object_deallocate(
                 *      the object; we must lock it before removing
                 *      the object.
                 */
+               for (;;) {
+                       vm_object_cache_lock();
 
-               vm_object_cache_lock();
-               vm_object_lock(object);
-
+                       /*
+                        * if we try to take a regular lock here
+                        * we risk deadlocking against someone
+                        * holding a lock on this object while
+                        * trying to vm_object_deallocate a different
+                        * object
+                        */
+                       if (vm_object_lock_try(object))
+                               break;
+                       vm_object_cache_unlock();
+                       mutex_pause();  /* wait a bit */
+               }
                assert(object->ref_count > 0);
 
                /*
@@ -608,8 +641,21 @@ vm_object_deallocate(
                                        
                                memory_object_unmap(pager);
 
-                               vm_object_cache_lock();
-                               vm_object_lock(object);
+                               for (;;) {
+                                       vm_object_cache_lock();
+
+                                       /*
+                                        * if we try to take a regular lock here
+                                        * we risk deadlocking against someone
+                                        * holding a lock on this object while
+                                        * trying to vm_object_deallocate a different
+                                        * object
+                                        */
+                                       if (vm_object_lock_try(object))
+                                               break;
+                                       vm_object_cache_unlock();
+                                       mutex_pause();  /* wait a bit */
+                               }
                                assert(object->ref_count > 0);
                        }
                }
@@ -629,8 +675,23 @@ vm_object_deallocate(
                if ((object->ref_count > 1) || object->terminating) {
                        object->ref_count--;
                        vm_object_res_deallocate(object);
-                       vm_object_unlock(object);
                        vm_object_cache_unlock();
+
+                       if (object->ref_count == 1 &&
+                           object->shadow != VM_OBJECT_NULL) {
+                               /*
+                                * We don't use this VM object anymore.  We
+                                * would like to collapse it into its parent(s),
+                                * but we don't have any pointers back to these
+                                * parent object(s).
+                                * But we can try and collapse this object with
+                                * its own shadows, in case these are useless
+                                * too...
+                                */
+                               vm_object_collapse(object, 0);
+                       }
+
+                       vm_object_unlock(object); 
                        if (retry_cache_trim &&
                            ((object = vm_object_cache_trim(TRUE)) !=
                             VM_OBJECT_NULL)) {
@@ -747,10 +808,10 @@ vm_object_deallocate(
                         *      This object is not cachable; terminate it.
                         */
                        XPR(XPR_VM_OBJECT,
-        "vm_o_deallocate: !cacheable 0x%X res %d paging_ops %d thread 0x%lX ref %d\n",
-                               (integer_t)object, object->resident_page_count,
-                               object->paging_in_progress,
-                               (natural_t)current_thread(),object->ref_count);
+        "vm_o_deallocate: !cacheable 0x%X res %d paging_ops %d thread 0x%p ref %d\n",
+                           (integer_t)object, object->resident_page_count,
+                           object->paging_in_progress,
+                           (void *)current_thread(),object->ref_count);
 
                        VM_OBJ_RES_DECR(object);        /* XXX ? */
                        /*
@@ -886,7 +947,6 @@ static kern_return_t
 vm_object_terminate(
        register vm_object_t    object)
 {
-       memory_object_t         pager;
        register vm_page_t      p;
        vm_object_t             shadow_object;
 
@@ -918,6 +978,7 @@ vm_object_terminate(
                }
 
                vm_page_lock_queues();
+               p->busy = TRUE;
                VM_PAGE_QUEUES_REMOVE(p);
                vm_page_unlock_queues();
 
@@ -941,13 +1002,7 @@ vm_object_terminate(
                        p->dirty = pmap_is_modified(p->phys_page);
 
                if ((p->dirty || p->precious) && !p->error && object->alive) {
-                       p->busy = TRUE;
-                       vm_object_paging_begin(object);
-                       /* protect the object from re-use/caching while it */
-                       /* is unlocked */
-                       vm_object_unlock(object);
                        vm_pageout_cluster(p); /* flush page */
-                       vm_object_lock(object);
                        vm_object_paging_wait(object, THREAD_UNINT);
                        XPR(XPR_VM_OBJECT,
                            "vm_object_terminate restart, object 0x%X ref %d\n",
@@ -996,23 +1051,82 @@ vm_object_terminate(
 
        /*
         *      Detach the object from its shadow if we are the shadow's
-        *      copy.
+        *      copy. The reference we hold on the shadow must be dropped
+        *      by our caller.
         */
        if (((shadow_object = object->shadow) != VM_OBJECT_NULL) &&
            !(object->pageout)) {
                vm_object_lock(shadow_object);
-               assert((shadow_object->copy == object) ||
-                      (shadow_object->copy == VM_OBJECT_NULL));
-               shadow_object->copy = VM_OBJECT_NULL;
+               if (shadow_object->copy == object)
+                       shadow_object->copy = VM_OBJECT_NULL;
                vm_object_unlock(shadow_object);
        }
 
+       if (FALSE && object->paging_in_progress != 0) {
+               /*
+                * There are still some paging_in_progress references
+                * on this object, meaning that there are some paging
+                * or other I/O operations in progress for this VM object.
+                * Such operations take some paging_in_progress references
+                * up front to ensure that the object doesn't go away, but
+                * they may also need to acquire a reference on the VM object,
+                * to map it in kernel space, for example.  That means that
+                * they may end up releasing the last reference on the VM
+                * object, triggering its termination, while still holding
+                * paging_in_progress references.  Waiting for these
+                * pending paging_in_progress references to go away here would
+                * deadlock.
+                *
+                * To avoid deadlocking, we'll let the vm_object_reaper_thread
+                * complete the VM object termination if it still holds
+                * paging_in_progress references at this point.
+                *
+                * No new paging_in_progress should appear now that the
+                * VM object is "terminating" and not "alive".
+                */
+               vm_object_reap_async(object);
+               vm_object_cache_unlock();
+               vm_object_unlock(object);
+               return KERN_SUCCESS;
+       }
+       /* complete the VM object termination */
+       vm_object_reap(object);
+       object = VM_OBJECT_NULL;
+       /* cache lock and object lock were released by vm_object_reap() */
+
+       return KERN_SUCCESS;
+}
+
+/*
+ * vm_object_reap():
+ *
+ * Complete the termination of a VM object after it's been marked
+ * as "terminating" and "!alive" by vm_object_terminate().
+ *
+ * The VM object cache and the VM object must be locked by caller.
+ * The locks will be released on return and the VM object is no longer valid.
+ */
+void
+vm_object_reap(
+       vm_object_t object)
+{
+       memory_object_t         pager;
+       vm_page_t               p;
+
+#if DEBUG
+       mutex_assert(&vm_object_cached_lock_data, MA_OWNED);
+       mutex_assert(&object->Lock, MA_OWNED);
+#endif /* DEBUG */
+
+       vm_object_reap_count++;
+
        /*
-        *      The pageout daemon might be playing with our pages.
-        *      Now that the object is dead, it won't touch any more
-        *      pages, but some pages might already be on their way out.
-        *      Hence, we wait until the active paging activities have ceased
-        *      before we break the association with the pager itself.
+        * The pageout daemon might be playing with our pages.
+        * Now that the object is dead, it won't touch any more
+        * pages, but some pages might already be on their way out.
+        * Hence, we wait until the active paging activities have
+        * ceased before we break the association with the pager
+        * itself.
         */
        while (object->paging_in_progress != 0) {
                vm_object_cache_unlock();
@@ -1023,11 +1137,12 @@ vm_object_terminate(
                vm_object_lock(object);
        }
 
+       assert(object->paging_in_progress == 0);
        pager = object->pager;
        object->pager = MEMORY_OBJECT_NULL;
 
        if (pager != MEMORY_OBJECT_NULL)
-               memory_object_control_disable(object->pager_request);
+               memory_object_control_disable(object->pager_control);
        vm_object_cache_unlock();
 
        object->ref_count--;
@@ -1043,8 +1158,7 @@ vm_object_terminate(
         *      if some faults on this object were aborted.
         */
        if (object->pageout) {
-               assert(shadow_object != VM_OBJECT_NULL);
-               assert(shadow_object == object->shadow);
+               assert(object->shadow != VM_OBJECT_NULL);
 
                vm_pageout_object_terminate(object);
 
@@ -1057,7 +1171,7 @@ vm_object_terminate(
                        VM_PAGE_FREE(p);
                }
        } else if (!queue_empty(&object->memq)) {
-               panic("vm_object_terminate: queue just emptied isn't");
+               panic("vm_object_reap: queue just emptied isn't");
        }
 
        assert(object->paging_in_progress == 0);
@@ -1087,8 +1201,56 @@ vm_object_terminate(
        /*
         *      Free the space for the object.
         */
-       zfree(vm_object_zone, (vm_offset_t) object);
-       return KERN_SUCCESS;
+       zfree(vm_object_zone, object);
+       object = VM_OBJECT_NULL;
+}
+
+void
+vm_object_reap_async(
+       vm_object_t     object)
+{
+#if DEBUG
+       mutex_assert(&vm_object_cached_lock_data, MA_OWNED);
+       mutex_assert(&object->Lock, MA_OWNED);
+#endif /* DEBUG */
+
+       vm_object_reap_count_async++;
+
+       /* enqueue the VM object... */
+       queue_enter(&vm_object_reaper_queue, object,
+                   vm_object_t, cached_list);
+       /* ... and wake up the reaper thread */
+       thread_wakeup((event_t) &vm_object_reaper_queue);
+}
+
+void
+vm_object_reaper_thread(void)
+{
+       vm_object_t     object;
+
+       vm_object_cache_lock();
+
+       while (!queue_empty(&vm_object_reaper_queue)) {
+               queue_remove_first(&vm_object_reaper_queue,
+                                  object,
+                                  vm_object_t,
+                                  cached_list);
+               vm_object_lock(object);
+               assert(object->terminating);
+               assert(!object->alive);
+               
+               vm_object_reap(object);
+               /* cache is unlocked and object is no longer valid */
+               object = VM_OBJECT_NULL;
+
+               vm_object_cache_lock();
+       }
+
+       /* wait for more work... */
+       assert_wait((event_t) &vm_object_reaper_queue, THREAD_UNINT);
+       vm_object_cache_unlock();
+       thread_block((thread_continue_t) vm_object_reaper_thread);
+       /*NOTREACHED*/
 }
 
 /*
@@ -1150,63 +1312,6 @@ vm_object_release_pager(
        memory_object_deallocate(pager);
 }
 
-/*
- *     Routine:        vm_object_abort_activity [internal use only]
- *     Purpose:
- *             Abort paging requests pending on this object.
- *     In/out conditions:
- *             The object is locked on entry and exit.
- */
-static void
-vm_object_abort_activity(
-       vm_object_t     object)
-{
-       register
-       vm_page_t       p;
-       vm_page_t       next;
-
-       XPR(XPR_VM_OBJECT, "vm_object_abort_activity, object 0x%X\n",
-               (integer_t)object, 0, 0, 0, 0);
-
-       /*
-        *      Abort all activity that would be waiting
-        *      for a result on this memory object.
-        *
-        *      We could also choose to destroy all pages
-        *      that we have in memory for this object, but
-        *      we don't.
-        */
-
-       p = (vm_page_t) queue_first(&object->memq);
-       while (!queue_end(&object->memq, (queue_entry_t) p)) {
-               next = (vm_page_t) queue_next(&p->listq);
-
-               /*
-                *      If it's being paged in, destroy it.
-                *      If an unlock has been requested, start it again.
-                */
-
-               if (p->busy && p->absent) {
-                       VM_PAGE_FREE(p);
-               }
-                else {
-                       if (p->unlock_request != VM_PROT_NONE)
-                               p->unlock_request = VM_PROT_NONE;
-                       PAGE_WAKEUP(p);
-               }
-               
-               p = next;
-       }
-
-       /*
-        *      Wake up threads waiting for the memory object to
-        *      become ready.
-        */
-
-       object->pager_ready = TRUE;
-       vm_object_wakeup(object, VM_OBJECT_EVENT_PAGER_READY);
-}
-
 /*
  *     Routine:        vm_object_destroy
  *     Purpose:
@@ -1217,7 +1322,7 @@ vm_object_abort_activity(
 kern_return_t
 vm_object_destroy(
        vm_object_t             object,
-       kern_return_t           reason)
+       __unused kern_return_t          reason)
 {
        memory_object_t         old_pager;
 
@@ -1247,7 +1352,7 @@ vm_object_destroy(
        old_pager = object->pager;
        object->pager = MEMORY_OBJECT_NULL;
        if (old_pager != MEMORY_OBJECT_NULL)
-               memory_object_control_disable(object->pager_request);
+               memory_object_control_disable(object->pager_control);
        vm_object_cache_unlock();
 
        /*
@@ -1333,6 +1438,8 @@ vm_object_deactivate_pages(
 
                                if ((m->wire_count == 0) && (!m->private) && (!m->gobbled) && (!m->busy)) {
 
+                                       assert(!m->laundry);
+
                                        m->reference = FALSE;
                                        pmap_clear_reference(m->phys_page);
 
@@ -1344,6 +1451,10 @@ vm_object_deactivate_pages(
                                        }
                                        VM_PAGE_QUEUES_REMOVE(m);
 
+                                       assert(!m->laundry);
+                                       assert(m->object != kernel_object);
+                                       assert(m->pageq.next == NULL &&
+                                              m->pageq.prev == NULL);
                                        if(m->zero_fill) {
                                                queue_enter_first(
                                                        &vm_page_queue_zf, 
@@ -1413,21 +1524,22 @@ __private_extern__ void
 vm_object_pmap_protect(
        register vm_object_t            object,
        register vm_object_offset_t     offset,
-       vm_size_t                       size,
+       vm_object_size_t                size,
        pmap_t                          pmap,
-       vm_offset_t                     pmap_start,
+       vm_map_offset_t                 pmap_start,
        vm_prot_t                       prot)
 {
        if (object == VM_OBJECT_NULL)
            return;
-       size = round_page_64(size);
-       offset = trunc_page_64(offset);
+       size = vm_object_round_page(size);
+       offset = vm_object_trunc_page(offset);
 
        vm_object_lock(object);
 
+       assert(object->internal);
+
        while (TRUE) {
-           if (object->resident_page_count > atop_32(size) / 2 &&
-                   pmap != PMAP_NULL) {
+          if (ptoa_64(object->resident_page_count) > size/2 && pmap != PMAP_NULL) {
                vm_object_unlock(object);
                pmap_protect(pmap, pmap_start, pmap_start + size, prot);
                return;
@@ -1436,7 +1548,7 @@ vm_object_pmap_protect(
            /* if we are doing large ranges with respect to resident */
            /* page count then we should interate over pages otherwise */
            /* inverse page look-up will be faster */
-           if ((object->resident_page_count / 4) <  atop_32(size)) {
+           if (ptoa_64(object->resident_page_count / 4) <  size) {
                vm_page_t               p;
                vm_object_offset_t      end;
 
@@ -1446,11 +1558,10 @@ vm_object_pmap_protect(
                  queue_iterate(&object->memq, p, vm_page_t, listq) {
                    if (!p->fictitious &&
                        (offset <= p->offset) && (p->offset < end)) {
+                       vm_map_offset_t start;
 
-                       vm_offset_t start = pmap_start + 
-                                       (vm_offset_t)(p->offset - offset);
-
-                       pmap_protect(pmap, start, start + PAGE_SIZE, prot);
+                       start = pmap_start + p->offset - offset;
+                       pmap_protect(pmap, start, start + PAGE_SIZE_64, prot);
                    }
                  }
                } else {
@@ -1472,9 +1583,12 @@ vm_object_pmap_protect(
 
                if (pmap != PMAP_NULL) {
                        for(target_off = offset; 
-                               target_off < end; target_off += PAGE_SIZE) {
-                               if(p = vm_page_lookup(object, target_off)) {
-                                       vm_offset_t start = pmap_start + 
+                           target_off < end;
+                           target_off += PAGE_SIZE) {
+                               p = vm_page_lookup(object, target_off);
+                               if (p != VM_PAGE_NULL) {
+                                       vm_offset_t start;
+                                       start = pmap_start + 
                                                (vm_offset_t)(p->offset - offset);
                                        pmap_protect(pmap, start, 
                                                        start + PAGE_SIZE, prot);
@@ -1483,7 +1597,8 @@ vm_object_pmap_protect(
                } else {
                        for(target_off = offset; 
                                target_off < end; target_off += PAGE_SIZE) {
-                               if(p = vm_page_lookup(object, target_off)) {
+                               p = vm_page_lookup(object, target_off);
+                               if (p != VM_PAGE_NULL) {
                                        pmap_page_protect(p->phys_page,
                                                      prot & ~p->page_lock);
                                }
@@ -1597,6 +1712,7 @@ vm_object_copy_slowly(
 
        new_object = vm_object_allocate(size);
        new_offset = 0;
+       vm_object_lock(new_object);
 
        assert(size == trunc_page_64(size));    /* Will the loop terminate? */
 
@@ -1611,7 +1727,9 @@ vm_object_copy_slowly(
                while ((new_page = vm_page_alloc(new_object, new_offset))
                                == VM_PAGE_NULL) {
                        if (!vm_page_wait(interruptible)) {
+                               vm_object_unlock(new_object);
                                vm_object_deallocate(new_object);
+                               vm_object_deallocate(src_object);
                                *_result_object = VM_OBJECT_NULL;
                                return(MACH_SEND_INTERRUPTED);
                        }
@@ -1699,6 +1817,7 @@ vm_object_copy_slowly(
 
                                case VM_FAULT_INTERRUPTED:
                                        vm_page_free(new_page);
+                                       vm_object_unlock(new_object);
                                        vm_object_deallocate(new_object);
                                        vm_object_deallocate(src_object);
                                        *_result_object = VM_OBJECT_NULL;
@@ -1716,6 +1835,7 @@ vm_object_copy_slowly(
                                        vm_page_lock_queues();
                                        vm_page_free(new_page);
                                        vm_page_unlock_queues();
+                                       vm_object_unlock(new_object);
                                        vm_object_deallocate(new_object);
                                        vm_object_deallocate(src_object);
                                        *_result_object = VM_OBJECT_NULL;
@@ -1729,6 +1849,7 @@ vm_object_copy_slowly(
         *      Lose the extra reference, and return our object.
         */
 
+        vm_object_unlock(new_object);
        vm_object_deallocate(src_object);
        *_result_object = new_object;
        return(KERN_SUCCESS);
@@ -1755,8 +1876,8 @@ vm_object_copy_slowly(
 __private_extern__ boolean_t
 vm_object_copy_quickly(
        vm_object_t             *_object,               /* INOUT */
-       vm_object_offset_t      offset,                 /* IN */
-       vm_object_size_t        size,                   /* IN */
+       __unused vm_object_offset_t     offset, /* IN */
+       __unused vm_object_size_t       size,   /* IN */
        boolean_t               *_src_needs_copy,       /* OUT */
        boolean_t               *_dst_needs_copy)       /* OUT */
 {
@@ -1936,8 +2057,6 @@ static int copy_delayed_lock_collisions = 0;
 static int copy_delayed_max_collisions = 0;
 static int copy_delayed_lock_contention = 0;
 static int copy_delayed_protect_iterate = 0;
-static int copy_delayed_protect_lookup = 0;
-static int copy_delayed_protect_lookup_wait = 0;
 
 /*
  *     Routine:        vm_object_copy_delayed [internal]
@@ -2054,11 +2173,6 @@ vm_object_copy_delayed(
                         *      needed).
                         */
 
-                       if (new_copy != VM_OBJECT_NULL) {
-                               vm_object_unlock(new_copy);
-                               vm_object_deallocate(new_copy);
-                       }
-
                        if (old_copy->size < copy_size) {
                                /*
                                 * We can't perform a delayed copy if any of the
@@ -2075,6 +2189,12 @@ vm_object_copy_delayed(
                                                if (p->wire_count > 0) {
                                                        vm_object_unlock(old_copy);
                                                        vm_object_unlock(src_object);
+
+                                                       if (new_copy != VM_OBJECT_NULL) {
+                                                               vm_object_unlock(new_copy);
+                                                               vm_object_deallocate(new_copy);
+                                                       }
+
                                                        return VM_OBJECT_NULL;
                                                } else {
                                                        pmap_page_protect(p->phys_page, 
@@ -2089,6 +2209,12 @@ vm_object_copy_delayed(
                        vm_object_reference_locked(old_copy);
                        vm_object_unlock(old_copy);
                        vm_object_unlock(src_object);
+
+                       if (new_copy != VM_OBJECT_NULL) {
+                               vm_object_unlock(new_copy);
+                               vm_object_deallocate(new_copy);
+                       }
+
                        return(old_copy);
                }
 
@@ -2443,6 +2569,66 @@ vm_object_shadow(
  */
 
 #if 0
+static void            vm_object_abort_activity(
+                               vm_object_t     object);
+
+/*
+ *     Routine:        vm_object_abort_activity [internal use only]
+ *     Purpose:
+ *             Abort paging requests pending on this object.
+ *     In/out conditions:
+ *             The object is locked on entry and exit.
+ */
+static void
+vm_object_abort_activity(
+       vm_object_t     object)
+{
+       register
+       vm_page_t       p;
+       vm_page_t       next;
+
+       XPR(XPR_VM_OBJECT, "vm_object_abort_activity, object 0x%X\n",
+               (integer_t)object, 0, 0, 0, 0);
+
+       /*
+        *      Abort all activity that would be waiting
+        *      for a result on this memory object.
+        *
+        *      We could also choose to destroy all pages
+        *      that we have in memory for this object, but
+        *      we don't.
+        */
+
+       p = (vm_page_t) queue_first(&object->memq);
+       while (!queue_end(&object->memq, (queue_entry_t) p)) {
+               next = (vm_page_t) queue_next(&p->listq);
+
+               /*
+                *      If it's being paged in, destroy it.
+                *      If an unlock has been requested, start it again.
+                */
+
+               if (p->busy && p->absent) {
+                       VM_PAGE_FREE(p);
+               }
+                else {
+                       if (p->unlock_request != VM_PROT_NONE)
+                               p->unlock_request = VM_PROT_NONE;
+                       PAGE_WAKEUP(p);
+               }
+               
+               p = next;
+       }
+
+       /*
+        *      Wake up threads waiting for the memory object to
+        *      become ready.
+        */
+
+       object->pager_ready = TRUE;
+       vm_object_wakeup(object, VM_OBJECT_EVENT_PAGER_READY);
+}
+
 /*
  *     Routine:        vm_object_pager_dead
  *
@@ -2570,69 +2756,53 @@ vm_object_enter(
         *      Look for an object associated with this port.
         */
 
-restart:
        vm_object_cache_lock();
-       for (;;) {
+       do {
                entry = vm_object_hash_lookup(pager, FALSE);
 
-               /*
-                *      If a previous object is being terminated,
-                *      we must wait for the termination message
-                *      to be queued.
-                *
-                *      We set kobject to a non-null value to let the
-                *      terminator know that someone is waiting.
-                *      Among the possibilities is that the port
-                *      could die while we're waiting.  Must restart
-                *      instead of continuing the loop.
-                */
-
-               if (entry != VM_OBJECT_HASH_ENTRY_NULL) {
-                       if (entry->object != VM_OBJECT_NULL)
-                               break;
-
+               if (entry == VM_OBJECT_HASH_ENTRY_NULL) {
+                       if (new_object == VM_OBJECT_NULL) {
+                               /*
+                                *      We must unlock to create a new object;
+                                *      if we do so, we must try the lookup again.
+                                */
+                               vm_object_cache_unlock();
+                               assert(new_entry == VM_OBJECT_HASH_ENTRY_NULL);
+                               new_entry = vm_object_hash_entry_alloc(pager);
+                               new_object = vm_object_allocate(size);
+                               vm_object_cache_lock();
+                       } else {
+                               /*
+                                *      Lookup failed twice, and we have something
+                                *      to insert; set the object.
+                                */
+                               vm_object_hash_insert(new_entry);
+                               entry = new_entry;
+                               entry->object = new_object;
+                               new_entry = VM_OBJECT_HASH_ENTRY_NULL;
+                               new_object = VM_OBJECT_NULL;
+                               must_init = TRUE;
+                       }
+               } else if (entry->object == VM_OBJECT_NULL) {
+                       /*
+                        *      If a previous object is being terminated,
+                        *      we must wait for the termination message
+                        *      to be queued (and lookup the entry again).
+                        */
                        entry->waiting = TRUE;
+                       entry = VM_OBJECT_HASH_ENTRY_NULL;
                        assert_wait((event_t) pager, THREAD_UNINT);
                        vm_object_cache_unlock();
-                       thread_block((void (*)(void))0);
-                       goto restart;
+                       thread_block(THREAD_CONTINUE_NULL);
+                       vm_object_cache_lock();
                }
-
-               /*
-                *      We must unlock to create a new object;
-                *      if we do so, we must try the lookup again.
-                */
-
-               if (new_object == VM_OBJECT_NULL) {
-                       vm_object_cache_unlock();
-                       assert(new_entry == VM_OBJECT_HASH_ENTRY_NULL);
-                       new_entry = vm_object_hash_entry_alloc(pager);
-                       new_object = vm_object_allocate(size);
-                       vm_object_cache_lock();
-               } else {
-                       /*
-                        *      Lookup failed twice, and we have something
-                        *      to insert; set the object.
-                        */
-
-                       if (entry == VM_OBJECT_HASH_ENTRY_NULL) {
-                               vm_object_hash_insert(new_entry);
-                               entry = new_entry;
-                               new_entry = VM_OBJECT_HASH_ENTRY_NULL;
-                       }
-
-                       entry->object = new_object;
-                       new_object = VM_OBJECT_NULL;
-                       must_init = TRUE;
-               }
-       }
+       } while (entry == VM_OBJECT_HASH_ENTRY_NULL);
 
        object = entry->object;
        assert(object != VM_OBJECT_NULL);
 
        if (!must_init) {
                vm_object_lock(object);
-               assert(object->pager_created);
                assert(!internal || object->internal);
                if (named) {
                        assert(!object->named);
@@ -2676,16 +2846,17 @@ restart:
                vm_object_hash_entry_free(new_entry);
 
        if (must_init) {
-               pager_request_t pager_request;
+               memory_object_control_t control;
 
                /*
                 *      Allocate request port.
                 */
 
-               pager_request = memory_object_control_allocate(object);
-               assert (pager_request != PAGER_REQUEST_NULL);
+               control = memory_object_control_allocate(object);
+               assert (control != MEMORY_OBJECT_CONTROL_NULL);
 
                vm_object_lock(object);
+               assert(object != kernel_object);
 
                /*
                 *      Copy the reference we were given.
@@ -2700,7 +2871,7 @@ restart:
                        /* copy strategy invalid until set by memory manager */
                        object->copy_strategy = MEMORY_OBJECT_COPY_INVALID;
                }
-               object->pager_request = pager_request;
+               object->pager_control = control;
                object->pager_ready = FALSE;
 
                vm_object_unlock(object);
@@ -2710,7 +2881,7 @@ restart:
                 */
 
                (void) memory_object_init(pager,
-                       object->pager_request,
+                       object->pager_control,
                        PAGE_SIZE);
 
                vm_object_lock(object);
@@ -2777,6 +2948,8 @@ vm_object_pager_create(
        XPR(XPR_VM_OBJECT, "vm_object_pager_create, object 0x%X\n",
                (integer_t)object, 0,0,0,0);
 
+       assert(object != kernel_object);
+
        if (memory_manager_default_check() != KERN_SUCCESS)
                return;
 
@@ -2887,7 +3060,6 @@ vm_object_remove(
        vm_object_t     object)
 {
        memory_object_t pager;
-       pager_request_t pager_request;
 
        if ((pager = object->pager) != MEMORY_OBJECT_NULL) {
                vm_object_hash_entry_t  entry;
@@ -2914,6 +3086,8 @@ static boolean_t  vm_object_bypass_allowed = TRUE;
 static int     vm_external_discarded;
 static int     vm_external_collapsed;
 
+unsigned long vm_object_collapse_encrypted = 0;
+
 /*
  *     Routine:        vm_object_do_collapse
  *     Purpose:
@@ -2951,7 +3125,7 @@ vm_object_do_collapse(
                new_offset = (p->offset - backing_offset);
                
                assert(!p->busy || p->absent);
-               
+
                /*
                 *      If the parent has a page here, or if
                 *      this page falls outside the parent,
@@ -2963,6 +3137,19 @@ vm_object_do_collapse(
                if (p->offset < backing_offset || new_offset >= size) {
                        VM_PAGE_FREE(p);
                } else {
+                       /*
+                        * ENCRYPTED SWAP:
+                        * The encryption key includes the "pager" and the
+                        * "paging_offset".  These might not be the same in
+                        * the new object, so we can't just move an encrypted
+                        * page from one object to the other.  We can't just
+                        * decrypt the page here either, because that would drop
+                        * the object lock.
+                        * The caller should check for encrypted pages before
+                        * attempting to collapse.
+                        */
+                       ASSERT_PAGE_DECRYPTED(p);
+
                        pp = vm_page_lookup(object, new_offset);
                        if (pp == VM_PAGE_NULL) {
 
@@ -3003,8 +3190,13 @@ vm_object_do_collapse(
                }
        }
        
-       assert(object->pager == MEMORY_OBJECT_NULL ||
-              backing_object->pager == MEMORY_OBJECT_NULL);
+#if    !MACH_PAGEMAP
+       assert(!object->pager_created && object->pager == MEMORY_OBJECT_NULL
+               || (!backing_object->pager_created
+               &&  backing_object->pager == MEMORY_OBJECT_NULL));
+#else 
+        assert(!object->pager_created && object->pager == MEMORY_OBJECT_NULL);
+#endif /* !MACH_PAGEMAP */
 
        if (backing_object->pager != MEMORY_OBJECT_NULL) {
                vm_object_hash_entry_t  entry;
@@ -3017,27 +3209,26 @@ vm_object_do_collapse(
                 *      unused portion.
                 */
 
+               assert(!object->paging_in_progress);
                object->pager = backing_object->pager;
                entry = vm_object_hash_lookup(object->pager, FALSE);
                assert(entry != VM_OBJECT_HASH_ENTRY_NULL);
                entry->object = object;
                object->pager_created = backing_object->pager_created;
-               object->pager_request = backing_object->pager_request;
+               object->pager_control = backing_object->pager_control;
                object->pager_ready = backing_object->pager_ready;
                object->pager_initialized = backing_object->pager_initialized;
                object->cluster_size = backing_object->cluster_size;
                object->paging_offset =
                    backing_object->paging_offset + backing_offset;
-               if (object->pager_request != PAGER_REQUEST_NULL) {
-                       memory_object_control_collapse(object->pager_request,
+               if (object->pager_control != MEMORY_OBJECT_CONTROL_NULL) {
+                       memory_object_control_collapse(object->pager_control,
                                                       object);
                }
        }
 
        vm_object_cache_unlock();
 
-       object->paging_offset = backing_object->paging_offset + backing_offset;
-
 #if    MACH_PAGEMAP
        /*
         *      If the shadow offset is 0, the use the existence map from
@@ -3070,10 +3261,17 @@ vm_object_do_collapse(
         *      moves from within backing_object to within object.
         */
        
+       assert(!object->phys_contiguous);
+       assert(!backing_object->phys_contiguous);
        object->shadow = backing_object->shadow;
-       object->shadow_offset += backing_object->shadow_offset;
+       if (object->shadow) {
+               object->shadow_offset += backing_object->shadow_offset;
+       } else {
+               /* no shadow, therefore no shadow offset... */
+               object->shadow_offset = 0;
+       }
        assert((object->shadow == VM_OBJECT_NULL) ||
-              (object->shadow->copy == VM_OBJECT_NULL));
+              (object->shadow->copy != backing_object));
 
        /*
         *      Discard backing_object.
@@ -3093,7 +3291,7 @@ vm_object_do_collapse(
        XPR(XPR_VM_OBJECT, "vm_object_collapse, collapsed 0x%X\n",
                (integer_t)backing_object, 0,0,0,0);
 
-       zfree(vm_object_zone, (vm_offset_t) backing_object);
+       zfree(vm_object_zone, backing_object);
        
        object_collapses++;
 }
@@ -3127,8 +3325,15 @@ vm_object_do_bypass(
        vm_object_reference(backing_object->shadow);
 #endif /* TASK_SWAPPER */
 
+       assert(!object->phys_contiguous);
+       assert(!backing_object->phys_contiguous);
        object->shadow = backing_object->shadow;
-       object->shadow_offset += backing_object->shadow_offset;
+       if (object->shadow) {
+               object->shadow_offset += backing_object->shadow_offset;
+       } else {
+               /* no shadow, therefore no shadow offset... */
+               object->shadow_offset = 0;
+       }
        
        /*
         *      Backing object might have had a copy pointer
@@ -3211,17 +3416,24 @@ vm_object_do_bypass(
  *     Requires that the object be locked and the page queues be unlocked.
  *
  */
+static unsigned long vm_object_collapse_calls = 0;
+static unsigned long vm_object_collapse_objects = 0;
+static unsigned long vm_object_collapse_do_collapse = 0;
+static unsigned long vm_object_collapse_do_bypass = 0;
 __private_extern__ void
 vm_object_collapse(
-       register vm_object_t    object)
+       register vm_object_t                    object,
+       register vm_object_offset_t             hint_offset)
 {
        register vm_object_t                    backing_object;
-       register vm_object_offset_t             backing_offset;
-       register vm_object_size_t               size;
-       register vm_object_offset_t             new_offset;
-       register vm_page_t                      p;
+       register unsigned int                   rcount;
+       register unsigned int                   size;
+       vm_object_offset_t                      collapse_min_offset;
+       vm_object_offset_t                      collapse_max_offset;
+       vm_page_t                               page;
+       vm_object_t                             original_object;
 
-       vm_offset_t                     current_offset; 
+       vm_object_collapse_calls++;
 
        if (! vm_object_collapse_allowed && ! vm_object_bypass_allowed) {
                return;
@@ -3230,26 +3442,45 @@ vm_object_collapse(
        XPR(XPR_VM_OBJECT, "vm_object_collapse, obj 0x%X\n", 
                (integer_t)object, 0,0,0,0);
 
+       if (object == VM_OBJECT_NULL)
+               return;
+
+       original_object = object;
+
        while (TRUE) {
+               vm_object_collapse_objects++;
                /*
                 *      Verify that the conditions are right for either
                 *      collapse or bypass:
-                *
-                *      The object exists and no pages in it are currently
-                *      being paged out, and
                 */
-               if (object == VM_OBJECT_NULL ||
-                   object->paging_in_progress != 0 ||
-                   object->absent_count != 0)
-                       return;
 
                /*
                 *      There is a backing object, and
                 */
        
-               if ((backing_object = object->shadow) == VM_OBJECT_NULL)
+               backing_object = object->shadow;
+               if (backing_object == VM_OBJECT_NULL) {
+                       if (object != original_object) {
+                               vm_object_unlock(object);
+                       }
                        return;
+               }
        
+               /*
+                *      No pages in the object are currently
+                *      being paged out, and
+                */
+               if (object->paging_in_progress != 0 ||
+                   object->absent_count != 0) {
+                       /* try and collapse the rest of the shadow chain */
+                       vm_object_lock(backing_object);
+                       if (object != original_object) {
+                               vm_object_unlock(object);
+                       }
+                       object = backing_object;
+                       continue;
+               }
+
                vm_object_lock(backing_object);
 
                /*
@@ -3263,8 +3494,12 @@ vm_object_collapse(
        
                if (!backing_object->internal ||
                    backing_object->paging_in_progress != 0) {
-                       vm_object_unlock(backing_object);
-                       return;
+                       /* try and collapse the rest of the shadow chain */
+                       if (object != original_object) {
+                               vm_object_unlock(object);
+                       }
+                       object = backing_object;
+                       continue;
                }
        
                /*
@@ -3278,9 +3513,13 @@ vm_object_collapse(
                 *      parent object.
                 */
                if (backing_object->shadow != VM_OBJECT_NULL &&
-                   backing_object->shadow->copy != VM_OBJECT_NULL) {
-                       vm_object_unlock(backing_object);
-                       return;
+                   backing_object->shadow->copy == backing_object) {
+                       /* try and collapse the rest of the shadow chain */
+                       if (object != original_object) {
+                               vm_object_unlock(object);
+                       }
+                       object = backing_object;
+                       continue;
                }
 
                /*
@@ -3293,22 +3532,28 @@ vm_object_collapse(
                 *      object, we may be able to collapse it into the
                 *      parent.
                 *
-                *      The backing object must not have a pager
-                *      created for it, since collapsing an object
-                *      into a backing_object dumps new pages into
-                *      the backing_object that its pager doesn't
-                *      know about.
+                *      If MACH_PAGEMAP is defined:
+                *      The parent must not have a pager created for it,
+                *      since collapsing a backing_object dumps new pages
+                *      into the parent that its pager doesn't know about
+                *      (and the collapse code can't merge the existence
+                *      maps).
+                *      Otherwise:
+                *      As long as one of the objects is still not known
+                *      to the pager, we can collapse them.
                 */
-
                if (backing_object->ref_count == 1 &&
-                   ! object->pager_created &&
-                   vm_object_collapse_allowed) {
+                   (!object->pager_created 
+#if    !MACH_PAGEMAP
+                    || !backing_object->pager_created
+#endif /*!MACH_PAGEMAP */
+                   ) && vm_object_collapse_allowed) {
 
                        XPR(XPR_VM_OBJECT, 
-                  "vm_object_collapse: %x to %x, pager %x, pager_request %x\n",
+                  "vm_object_collapse: %x to %x, pager %x, pager_control %x\n",
                                (integer_t)backing_object, (integer_t)object,
                                (integer_t)backing_object->pager, 
-                               (integer_t)backing_object->pager_request, 0);
+                               (integer_t)backing_object->pager_control, 0);
 
                        /*
                         *      We need the cache lock for collapsing,
@@ -3316,10 +3561,46 @@ vm_object_collapse(
                         */
                        
                        if (! vm_object_cache_lock_try()) {
+                               if (object != original_object) {
+                                       vm_object_unlock(object);
+                               }
                                vm_object_unlock(backing_object);
                                return;
                        }
 
+                       /*
+                        * ENCRYPTED SWAP
+                        * We can't collapse the object if it contains
+                        * any encypted page, because the encryption key
+                        * includes the <object,offset> info.  We can't
+                        * drop the object lock in vm_object_do_collapse()
+                        * so we can't decrypt the page there either.
+                        */
+                       if (vm_pages_encrypted) {
+                               collapse_min_offset = object->shadow_offset;
+                               collapse_max_offset =
+                                       object->shadow_offset + object->size;
+                               queue_iterate(&backing_object->memq,
+                                             page, vm_page_t, listq) {
+                                       if (page->encrypted &&
+                                           (page->offset >=
+                                            collapse_min_offset) &&
+                                           (page->offset <
+                                            collapse_max_offset)) {
+                                               /*
+                                                * We found an encrypted page
+                                                * in the backing object,
+                                                * within the range covered 
+                                                * by the parent object: we can
+                                                * not collapse them.
+                                                */
+                                               vm_object_collapse_encrypted++;
+                                               vm_object_cache_unlock();
+                                               goto try_bypass;
+                                       }
+                               }
+                       }
+                      
                        /*
                         *      Collapse the object with its backing
                         *      object, and try again with the object's
@@ -3327,112 +3608,207 @@ vm_object_collapse(
                         */
 
                        vm_object_do_collapse(object, backing_object);
+                       vm_object_collapse_do_collapse++;
                        continue;
                }
 
-
+       try_bypass:
                /*
                 *      Collapsing the backing object was not possible
                 *      or permitted, so let's try bypassing it.
                 */
 
                if (! vm_object_bypass_allowed) {
-                       vm_object_unlock(backing_object);
-                       return;
+                       /* try and collapse the rest of the shadow chain */
+                       if (object != original_object) {
+                               vm_object_unlock(object);
+                       }
+                       object = backing_object;
+                       continue;
                }
 
 
                /*
-                *      If the backing object has a pager but no pagemap,
-                *      then we cannot bypass it, because we don't know
-                *      what pages it has.
+                *      If the object doesn't have all its pages present,
+                *      we have to make sure no pages in the backing object
+                *      "show through" before bypassing it.
                 */
-               if (backing_object->pager_created
+               size = atop(object->size);
+               rcount = object->resident_page_count;
+               if (rcount != size) {
+                       vm_object_offset_t      offset;
+                       vm_object_offset_t      backing_offset;
+                       unsigned int            backing_rcount;
+                       unsigned int            lookups = 0;
+
+                       /*
+                        *      If the backing object has a pager but no pagemap,
+                        *      then we cannot bypass it, because we don't know
+                        *      what pages it has.
+                        */
+                       if (backing_object->pager_created
 #if    MACH_PAGEMAP
-                   && (backing_object->existence_map == VM_EXTERNAL_NULL)
+                               && (backing_object->existence_map == VM_EXTERNAL_NULL)
 #endif /* MACH_PAGEMAP */
-                   ) {
-                       vm_object_unlock(backing_object);
-                       return;
-               }
+                               ) {
+                               /* try and collapse the rest of the shadow chain */
+                               if (object != original_object) {
+                                       vm_object_unlock(object);
+                               }
+                               object = backing_object;
+                               continue;
+                       }
 
-               /*
-                *      If the object has a pager but no pagemap,
-                *      then we cannot bypass it, because we don't know
-                *      what pages it has.
-                */
-               if (object->pager_created
+                       /*
+                        *      If the object has a pager but no pagemap,
+                        *      then we cannot bypass it, because we don't know
+                        *      what pages it has.
+                        */
+                       if (object->pager_created
 #if    MACH_PAGEMAP
-                   && (object->existence_map == VM_EXTERNAL_NULL)
+                               && (object->existence_map == VM_EXTERNAL_NULL)
 #endif /* MACH_PAGEMAP */
-                   ) {
-                       vm_object_unlock(backing_object);
-                       return;
-               }
+                               ) {
+                               /* try and collapse the rest of the shadow chain */
+                               if (object != original_object) {
+                                       vm_object_unlock(object);
+                               }
+                               object = backing_object;
+                               continue;
+                       }
 
-               backing_offset = object->shadow_offset;
-               size = object->size;
+                       /*
+                        *      If all of the pages in the backing object are
+                        *      shadowed by the parent object, the parent
+                        *      object no longer has to shadow the backing
+                        *      object; it can shadow the next one in the
+                        *      chain.
+                        *
+                        *      If the backing object has existence info,
+                        *      we must check examine its existence info
+                        *      as well.
+                        *
+                        */
 
-               /*
-                *      If all of the pages in the backing object are
-                *      shadowed by the parent object, the parent
-                *      object no longer has to shadow the backing
-                *      object; it can shadow the next one in the
-                *      chain.
-                *
-                *      If the backing object has existence info,
-                *      we must check examine its existence info
-                *      as well.
-                *
-                */
+                       backing_offset = object->shadow_offset;
+                       backing_rcount = backing_object->resident_page_count;
 
-               if(object->cow_hint >= size)
-                       object->cow_hint = 0;
-               current_offset = object->cow_hint;
-               while(TRUE) {
-                       if (vm_page_lookup(object, 
-                               (vm_object_offset_t)current_offset) 
-                                                       != VM_PAGE_NULL) {
-                               current_offset+=PAGE_SIZE;
-                       } else if ((object->pager_created) &&
-                               (object->existence_map != NULL) &&
-                               (vm_external_state_get(object->existence_map, 
-                               current_offset) 
-                                       != VM_EXTERNAL_STATE_ABSENT)) {
-                               current_offset+=PAGE_SIZE;
-                       } else if (vm_page_lookup(backing_object,
-                               (vm_object_offset_t)current_offset
-                               + backing_offset)!= VM_PAGE_NULL) {
-                               /* found a dependency */
-                               object->cow_hint = current_offset;
-                               vm_object_unlock(backing_object);
-                               return;
-                       } else if ((backing_object->pager_created) &&
-                               (backing_object->existence_map != NULL) &&
-                                       (vm_external_state_get(
-                                       backing_object->existence_map, 
-                                       current_offset + backing_offset) 
-                                               != VM_EXTERNAL_STATE_ABSENT)) {
-                               /* found a dependency */
-                               object->cow_hint = current_offset;
-                               vm_object_unlock(backing_object);
-                               return;
-                       } else {
-                               current_offset+=PAGE_SIZE;
+#define EXISTS_IN_OBJECT(obj, off, rc) \
+       (vm_external_state_get((obj)->existence_map, \
+        (vm_offset_t)(off)) == VM_EXTERNAL_STATE_EXISTS || \
+        ((rc) && ++lookups && vm_page_lookup((obj), (off)) != VM_PAGE_NULL && (rc)--))
+
+                       /*
+                        * Check the hint location first
+                        * (since it is often the quickest way out of here).
+                        */
+                       if (object->cow_hint != ~(vm_offset_t)0)
+                               hint_offset = (vm_object_offset_t)object->cow_hint;
+                       else
+                               hint_offset = (hint_offset > 8 * PAGE_SIZE_64) ?
+                                             (hint_offset - 8 * PAGE_SIZE_64) : 0;
+
+                       if (EXISTS_IN_OBJECT(backing_object, hint_offset +
+                                            backing_offset, backing_rcount) &&
+                           !EXISTS_IN_OBJECT(object, hint_offset, rcount)) {
+                               /* dependency right at the hint */
+                               object->cow_hint = (vm_offset_t)hint_offset;
+                               /* try and collapse the rest of the shadow chain */
+                               if (object != original_object) {
+                                       vm_object_unlock(object);
+                               }
+                               object = backing_object;
+                               continue;
                        }
-                       if(current_offset >= size) {
-                               /* wrap at end of object */
-                               current_offset = 0;
+
+                       /*
+                        * If the object's window onto the backing_object
+                        * is large compared to the number of resident
+                        * pages in the backing object, it makes sense to
+                        * walk the backing_object's resident pages first.
+                        *
+                        * NOTE: Pages may be in both the existence map and 
+                        * resident.  So, we can't permanently decrement
+                        * the rcount here because the second loop may
+                        * find the same pages in the backing object'
+                        * existence map that we found here and we would
+                        * double-decrement the rcount.  We also may or
+                        * may not have found the 
+                        */
+                       if (backing_rcount && size >
+                           ((backing_object->existence_map) ?
+                            backing_rcount : (backing_rcount >> 1))) {
+                               unsigned int rc = rcount;
+                               vm_page_t p;
+
+                               backing_rcount = backing_object->resident_page_count;
+                               p = (vm_page_t)queue_first(&backing_object->memq);
+                               do {
+                                       /* Until we get more than one lookup lock */
+                                       if (lookups > 256) {
+                                               lookups = 0;
+                                               delay(1);
+                                       }
+
+                                       offset = (p->offset - backing_offset);
+                                       if (offset < object->size &&
+                                           offset != hint_offset &&
+                                           !EXISTS_IN_OBJECT(object, offset, rc)) {
+                                               /* found a dependency */
+                                               object->cow_hint = (vm_offset_t)offset;
+                                               break;
+                                       }
+                                       p = (vm_page_t) queue_next(&p->listq);
+
+                               } while (--backing_rcount);
+                               if (backing_rcount != 0 ) {
+                                       /* try and collapse the rest of the shadow chain */
+                                       if (object != original_object) {
+                                               vm_object_unlock(object);
+                                       }
+                                       object = backing_object;
+                                       continue;
+                               }
                        }
-                       if(current_offset == object->cow_hint) {
-                               /* we are free of shadow influence */
-                               break;
+
+                       /*
+                        * Walk through the offsets looking for pages in the
+                        * backing object that show through to the object.
+                        */
+                       if (backing_rcount || backing_object->existence_map) {
+                               offset = hint_offset;
+                               
+                               while((offset =
+                                     (offset + PAGE_SIZE_64 < object->size) ?
+                                     (offset + PAGE_SIZE_64) : 0) != hint_offset) {
+
+                                       /* Until we get more than one lookup lock */
+                                       if (lookups > 256) {
+                                               lookups = 0;
+                                               delay(1);
+                                       }
+
+                                       if (EXISTS_IN_OBJECT(backing_object, offset +
+                                           backing_offset, backing_rcount) &&
+                                           !EXISTS_IN_OBJECT(object, offset, rcount)) {
+                                               /* found a dependency */
+                                               object->cow_hint = (vm_offset_t)offset;
+                                               break;
+                                       }
+                               }
+                               if (offset != hint_offset) {
+                                       /* try and collapse the rest of the shadow chain */
+                                       if (object != original_object) {
+                                               vm_object_unlock(object);
+                                       }
+                                       object = backing_object;
+                                       continue;
+                               }
                        }
                }
-               /* reset the cow_hint for any objects deeper in the chain */
-               object->cow_hint = 0;
-
 
+               /* reset the offset hint for any objects deeper in the chain */
+               object->cow_hint = (vm_offset_t)0;
 
                /*
                 *      All interesting pages in the backing object
@@ -3441,6 +3817,7 @@ vm_object_collapse(
                 */
 
                vm_object_do_bypass(object, backing_object);
+               vm_object_collapse_do_bypass++;
 
                /*
                 *      Try again with this object's new backing object.
@@ -3448,6 +3825,10 @@ vm_object_collapse(
 
                continue;
        }
+
+       if (object != original_object) {
+               vm_object_unlock(object);
+       }
 }
 
 /*
@@ -3486,8 +3867,7 @@ vm_object_page_remove(
                        if (p != VM_PAGE_NULL) {
                                assert(!p->cleaning && !p->pageout);
                                if (!p->fictitious)
-                                       pmap_page_protect(p->phys_page,
-                                                         VM_PROT_NONE);
+                                       pmap_disconnect(p->phys_page);
                                VM_PAGE_FREE(p);
                        }
                }
@@ -3500,8 +3880,7 @@ vm_object_page_remove(
                        if ((start <= p->offset) && (p->offset < end)) {
                                assert(!p->cleaning && !p->pageout);
                                if (!p->fictitious)
-                                   pmap_page_protect(p->phys_page,
-                                                     VM_PROT_NONE);
+                                       pmap_disconnect(p->phys_page);
                                VM_PAGE_FREE(p);
                        }
                        p = next;
@@ -3540,7 +3919,7 @@ vm_object_coalesce(
        register vm_object_t            prev_object,
        vm_object_t                     next_object,
        vm_object_offset_t              prev_offset,
-       vm_object_offset_t              next_offset,
+       __unused vm_object_offset_t next_offset,
        vm_object_size_t                prev_size,
        vm_object_size_t                next_size)
 {
@@ -3567,7 +3946,7 @@ vm_object_coalesce(
        /*
         *      Try to collapse the object first
         */
-       vm_object_collapse(prev_object);
+       vm_object_collapse(prev_object, prev_offset);
 
        /*
         *      Can't coalesce if pages not mapped to
@@ -3576,6 +3955,7 @@ vm_object_coalesce(
         *      . paged out
         *      . shadows another object
         *      . has a copy elsewhere
+        *      . is purgable
         *      . paging references (pages might be in page-list)
         */
 
@@ -3584,6 +3964,7 @@ vm_object_coalesce(
            (prev_object->shadow != VM_OBJECT_NULL) ||
            (prev_object->copy != VM_OBJECT_NULL) ||
            (prev_object->true_share != FALSE) ||
+           (prev_object->purgable != VM_OBJECT_NONPURGABLE) ||
            (prev_object->paging_in_progress != 0)) {
                vm_object_unlock(prev_object);
                return(FALSE);
@@ -3726,23 +4107,23 @@ vm_object_cached(
  */
 void
 vm_external_print(
-       vm_external_map_t map,
-       vm_size_t       size)
+       vm_external_map_t       emap,
+       vm_size_t               size)
 {
-       if (map == VM_EXTERNAL_NULL) {
+       if (emap == VM_EXTERNAL_NULL) {
                printf("0  ");
        } else {
                vm_size_t existence_size = stob(size);
                printf("{ size=%d, map=[", existence_size);
                if (existence_size > 0) {
-                       print_bitstring(map[0]);
+                       print_bitstring(emap[0]);
                }
                if (existence_size > 1) {
-                       print_bitstring(map[1]);
+                       print_bitstring(emap[1]);
                }
                if (existence_size > 2) {
                        printf("...");
-                       print_bitstring(map[existence_size-1]);
+                       print_bitstring(emap[existence_size-1]);
                }
                printf("] }\n");
        }
@@ -3754,8 +4135,6 @@ int
 vm_follow_object(
        vm_object_t object)
 {
-       extern db_indent;
-
        int count = 0;
        int orig_db_indent = db_indent;
 
@@ -3784,17 +4163,18 @@ vm_follow_object(
  */
 void
 vm_object_print(
-       vm_object_t     object,
-       boolean_t       have_addr,
-       int             arg_count,
-       char            *modif)
+       db_addr_t       db_addr,
+       __unused boolean_t      have_addr,
+       __unused int            arg_count,
+       __unused char           *modif)
 {
+       vm_object_t     object;
        register vm_page_t p;
-       extern db_indent;
-       char *s;
+       const char *s;
 
        register int count;
 
+       object = (vm_object_t) (long) db_addr;
        if (object == VM_OBJECT_NULL)
                return;
 
@@ -3804,7 +4184,7 @@ vm_object_print(
 
        iprintf("size=0x%x", object->size);
        printf(", cluster=0x%x", object->cluster_size);
-       printf(", frozen=0x%x", object->frozen_size);
+       printf(", memq_hint=%p", object->memq_hint);
        printf(", ref_count=%d\n", object->ref_count);
        iprintf("");
 #if    TASK_SWAPPER
@@ -3816,7 +4196,7 @@ vm_object_print(
        if (object->shadow) {
                register int i = 0;
                vm_object_t shadow = object;
-               while(shadow = shadow->shadow)
+               while((shadow = shadow->shadow))
                        i++;
                printf(" (depth %d)", i);
        }
@@ -3826,7 +4206,7 @@ vm_object_print(
 
        iprintf("pager=0x%x", object->pager);
        printf(", paging_offset=0x%x", object->paging_offset);
-       printf(", pager_request=0x%x\n", object->pager_request);
+       printf(", pager_control=0x%x\n", object->pager_control);
 
        iprintf("copy_strategy=%d[", object->copy_strategy);
        switch (object->copy_strategy) {
@@ -3902,10 +4282,11 @@ vm_object_print(
                (object->pageout ? "" : "!"),
                (object->internal ? "internal" : "external"),
                (object->temporary ? "temporary" : "permanent"));
-       iprintf("%salive, %slock_in_progress, %slock_restart, %sshadowed, %scached, %sprivate\n",
+       iprintf("%salive, %spurgable, %spurgable_volatile, %spurgable_empty, %sshadowed, %scached, %sprivate\n",
                (object->alive ? "" : "!"),
-               (object->lock_in_progress ? "" : "!"),
-               (object->lock_restart ? "" : "!"),
+               ((object->purgable != VM_OBJECT_NONPURGABLE) ? "" : "!"),
+               ((object->purgable == VM_OBJECT_PURGABLE_VOLATILE) ? "" : "!"),
+               ((object->purgable == VM_OBJECT_PURGABLE_EMPTY) ? "" : "!"),
                (object->shadowed ? "" : "!"),
                (vm_object_cached(object) ? "" : "!"),
                (object->private ? "" : "!"));
@@ -3936,7 +4317,7 @@ vm_object_print(
                        }
                        count++;
 
-                       printf("(off=0x%X,page=0x%X)", p->offset, (integer_t) p);
+                       printf("(off=0x%llX,page=%p)", p->offset, p);
                        p = (vm_page_t) queue_next(&p->listq);
                }
                if (count != 0) {
@@ -4051,10 +4432,21 @@ vm_object_populate_with_private(
                                        vm_page_unlock_queues();
                                } else if (m->phys_page != base_page) {
                                        /* pmap call to clear old mapping */
-                                       pmap_page_protect(m->phys_page,
-                                                               VM_PROT_NONE);
+                                       pmap_disconnect(m->phys_page);
                                        m->phys_page = base_page;
                                }
+
+                               /*
+                                * ENCRYPTED SWAP:
+                                * We're not pointing to the same
+                                * physical page any longer and the
+                                * contents of the new one are not
+                                * supposed to be encrypted.
+                                * XXX What happens to the original
+                                * physical page. Is it lost ?
+                                */
+                               m->encrypted = FALSE;
+
                        } else {
                                while ((m = vm_page_grab_fictitious()) 
                                                         == VM_PAGE_NULL)
@@ -4104,13 +4496,12 @@ vm_object_populate_with_private(
 
 __private_extern__ kern_return_t
 memory_object_free_from_cache(
-       host_t          host,
+       __unused host_t         host,
        int             *pager_id,
        int             *count)
 {
 
        int     object_released = 0;
-       int     i;
 
        register vm_object_t object = VM_OBJECT_NULL;
        vm_object_t shadow;
@@ -4204,7 +4595,7 @@ memory_object_create_named(
                                        VM_OBJECT_EVENT_PAGER_READY,
                                        THREAD_UNINT);
                }
-               *control = object->pager_request;
+               *control = object->pager_control;
                vm_object_unlock(object);
        }
        return (KERN_SUCCESS);
@@ -4395,6 +4786,9 @@ vm_object_release_name(
                        return KERN_SUCCESS;
                }
        }
+       /*NOTREACHED*/
+       assert(0);
+       return KERN_FAILURE;
 }
 
 
@@ -4407,8 +4801,9 @@ vm_object_lock_request(
        int                             flags,
        vm_prot_t                       prot)
 {
-       vm_object_offset_t      original_offset = offset;
-       boolean_t               should_flush=flags & MEMORY_OBJECT_DATA_FLUSH;
+       __unused boolean_t      should_flush;
+
+       should_flush = flags & MEMORY_OBJECT_DATA_FLUSH;
 
         XPR(XPR_MEMORY_OBJECT,
            "vm_o_lock_request, obj 0x%X off 0x%X size 0x%X flags %X prot %X\n",
@@ -4432,10 +4827,9 @@ vm_object_lock_request(
         */
        vm_object_lock(object);
        vm_object_paging_begin(object);
-       offset -= object->paging_offset;
 
        (void)vm_object_update(object,
-               offset, size, should_return, flags, prot);
+               offset, size, NULL, NULL, should_return, flags, prot);
 
        vm_object_paging_end(object);
        vm_object_unlock(object);
@@ -4443,7 +4837,364 @@ vm_object_lock_request(
        return (KERN_SUCCESS);
 }
 
+/*
+ * Empty a purgable object by grabbing the physical pages assigned to it and
+ * putting them on the free queue without writing them to backing store, etc.
+ * When the pages are next touched they will be demand zero-fill pages.  We
+ * skip pages which are busy, being paged in/out, wired, etc.  We do _not_
+ * skip referenced/dirty pages, pages on the active queue, etc.  We're more
+ * than happy to grab these since this is a purgable object.  We mark the
+ * object as "empty" after reaping its pages.
+ *
+ * On entry the object and page queues are locked, the object must be a
+ * purgable object with no delayed copies pending.
+ */
+unsigned int
+vm_object_purge(vm_object_t object)
+{
+       vm_page_t       p, next;
+       unsigned int    num_purged_pages;
+       vm_page_t       local_freeq;
+       unsigned long   local_freed;
+       int             purge_loop_quota;
+/* free pages as soon as we gather PURGE_BATCH_FREE_LIMIT pages to free */
+#define PURGE_BATCH_FREE_LIMIT 50
+/* release page queues lock every PURGE_LOOP_QUOTA iterations */
+#define PURGE_LOOP_QUOTA       100
+
+       num_purged_pages = 0;
+       if (object->purgable == VM_OBJECT_NONPURGABLE)
+               return num_purged_pages;
+
+       object->purgable = VM_OBJECT_PURGABLE_EMPTY;
+
+       assert(object->copy == VM_OBJECT_NULL);
+       assert(object->copy_strategy == MEMORY_OBJECT_COPY_NONE);
+       purge_loop_quota = PURGE_LOOP_QUOTA;
+
+       local_freeq = VM_PAGE_NULL;
+       local_freed = 0;
+
+       /*
+        * Go through the object's resident pages and try and discard them.
+        */
+       next = (vm_page_t)queue_first(&object->memq);
+       while (!queue_end(&object->memq, (queue_entry_t)next)) {
+               p = next;
+               next = (vm_page_t)queue_next(&next->listq);
+
+               if (purge_loop_quota-- == 0) {
+                       /*
+                        * Avoid holding the page queues lock for too long.
+                        * Let someone else take it for a while if needed.
+                        * Keep holding the object's lock to guarantee that
+                        * the object's page list doesn't change under us
+                        * while we yield.
+                        */
+                       if (local_freeq != VM_PAGE_NULL) {
+                               /*
+                                * Flush our queue of pages to free.
+                                */
+                               vm_page_free_list(local_freeq);
+                               local_freeq = VM_PAGE_NULL;
+                               local_freed = 0;
+                       }
+                       vm_page_unlock_queues();
+                       mutex_pause();
+                       vm_page_lock_queues();
+
+                       /* resume with the current page and a new quota */
+                       purge_loop_quota = PURGE_LOOP_QUOTA;
+               }
+                               
+                      
+               if (p->busy || p->cleaning || p->laundry ||
+                   p->list_req_pending) {
+                       /* page is being acted upon, so don't mess with it */
+                       continue;
+               }
+               if (p->wire_count) {
+                       /* don't discard a wired page */
+                       continue;
+               }
+
+               if (p->tabled) {
+                       /* clean up the object/offset table */
+                       vm_page_remove(p);
+               }
+               if (p->absent) {
+                       /* update the object's count of absent pages */
+                       vm_object_absent_release(object);
+               }
+
+               /* we can discard this page */
+
+               /* advertize that this page is in a transition state */
+               p->busy = TRUE;
+
+               if (p->no_isync == TRUE) {
+                       /* the page hasn't been mapped yet */
+                       /* (optimization to delay the i-cache sync) */
+               } else {
+                       /* unmap the page */
+                       int refmod_state;
+
+                       refmod_state = pmap_disconnect(p->phys_page);
+                       if (refmod_state & VM_MEM_MODIFIED) {
+                               p->dirty = TRUE;
+                       }
+               }
+
+               if (p->dirty || p->precious) {
+                       /* we saved the cost of cleaning this page ! */
+                       num_purged_pages++;
+                       vm_page_purged_count++;
+               }
+
+               /* remove page from active or inactive queue... */
+               VM_PAGE_QUEUES_REMOVE(p);
+
+               /* ... and put it on our queue of pages to free */
+               assert(!p->laundry);
+               assert(p->object != kernel_object);
+               assert(p->pageq.next == NULL &&
+                      p->pageq.prev == NULL);
+               p->pageq.next = (queue_entry_t) local_freeq;
+               local_freeq = p;
+               if (++local_freed >= PURGE_BATCH_FREE_LIMIT) {
+                       /* flush our queue of pages to free */
+                       vm_page_free_list(local_freeq);
+                       local_freeq = VM_PAGE_NULL;
+                       local_freed = 0;
+               }
+       }
+
+       /* flush our local queue of pages to free one last time */
+       if (local_freeq != VM_PAGE_NULL) {
+               vm_page_free_list(local_freeq);
+               local_freeq = VM_PAGE_NULL;
+               local_freed = 0;
+       }
+
+       return num_purged_pages;
+}
+
+/*
+ * vm_object_purgable_control() allows the caller to control and investigate the
+ * state of a purgable object.  A purgable object is created via a call to
+ * vm_allocate() with VM_FLAGS_PURGABLE specified.  A purgable object will
+ * never be coalesced with any other object -- even other purgable objects --
+ * and will thus always remain a distinct object.  A purgable object has
+ * special semantics when its reference count is exactly 1.  If its reference
+ * count is greater than 1, then a purgable object will behave like a normal
+ * object and attempts to use this interface will result in an error return
+ * of KERN_INVALID_ARGUMENT.
+ *
+ * A purgable object may be put into a "volatile" state which will make the
+ * object's pages elligable for being reclaimed without paging to backing
+ * store if the system runs low on memory.  If the pages in a volatile
+ * purgable object are reclaimed, the purgable object is said to have been
+ * "emptied."  When a purgable object is emptied the system will reclaim as
+ * many pages from the object as it can in a convenient manner (pages already
+ * en route to backing store or busy for other reasons are left as is).  When
+ * a purgable object is made volatile, its pages will generally be reclaimed
+ * before other pages in the application's working set.  This semantic is
+ * generally used by applications which can recreate the data in the object
+ * faster than it can be paged in.  One such example might be media assets
+ * which can be reread from a much faster RAID volume.
+ *
+ * A purgable object may be designated as "non-volatile" which means it will
+ * behave like all other objects in the system with pages being written to and
+ * read from backing store as needed to satisfy system memory needs.  If the
+ * object was emptied before the object was made non-volatile, that fact will
+ * be returned as the old state of the purgable object (see
+ * VM_PURGABLE_SET_STATE below).  In this case, any pages of the object which
+ * were reclaimed as part of emptying the object will be refaulted in as
+ * zero-fill on demand.  It is up to the application to note that an object
+ * was emptied and recreate the objects contents if necessary.  When a
+ * purgable object is made non-volatile, its pages will generally not be paged
+ * out to backing store in the immediate future.  A purgable object may also
+ * be manually emptied.
+ *
+ * Finally, the current state (non-volatile, volatile, volatile & empty) of a
+ * volatile purgable object may be queried at any time.  This information may
+ * be used as a control input to let the application know when the system is
+ * experiencing memory pressure and is reclaiming memory.
+ *
+ * The specified address may be any address within the purgable object.  If
+ * the specified address does not represent any object in the target task's
+ * virtual address space, then KERN_INVALID_ADDRESS will be returned.  If the
+ * object containing the specified address is not a purgable object, then
+ * KERN_INVALID_ARGUMENT will be returned.  Otherwise, KERN_SUCCESS will be
+ * returned.
+ *
+ * The control parameter may be any one of VM_PURGABLE_SET_STATE or
+ * VM_PURGABLE_GET_STATE.  For VM_PURGABLE_SET_STATE, the in/out parameter
+ * state is used to set the new state of the purgable object and return its
+ * old state.  For VM_PURGABLE_GET_STATE, the current state of the purgable
+ * object is returned in the parameter state.
+ *
+ * The in/out parameter state may be one of VM_PURGABLE_NONVOLATILE,
+ * VM_PURGABLE_VOLATILE or VM_PURGABLE_EMPTY.  These, respectively, represent
+ * the non-volatile, volatile and volatile/empty states described above.
+ * Setting the state of a purgable object to VM_PURGABLE_EMPTY will
+ * immediately reclaim as many pages in the object as can be conveniently
+ * collected (some may have already been written to backing store or be
+ * otherwise busy).
+ *
+ * The process of making a purgable object non-volatile and determining its
+ * previous state is atomic.  Thus, if a purgable object is made
+ * VM_PURGABLE_NONVOLATILE and the old state is returned as
+ * VM_PURGABLE_VOLATILE, then the purgable object's previous contents are
+ * completely intact and will remain so until the object is made volatile
+ * again.  If the old state is returned as VM_PURGABLE_EMPTY then the object
+ * was reclaimed while it was in a volatile state and its previous contents
+ * have been lost.
+ */
+/*
+ * The object must be locked.
+ */
+kern_return_t
+vm_object_purgable_control(
+       vm_object_t     object,
+       vm_purgable_t   control,
+       int             *state)
+{
+       int             old_state;
+       vm_page_t       p;
+
+       if (object == VM_OBJECT_NULL) {
+               /*
+                * Object must already be present or it can't be purgable.
+                */
+               return KERN_INVALID_ARGUMENT;
+       }
+
+       /*
+        * Get current state of the purgable object.
+        */
+       switch (object->purgable) {
+           case VM_OBJECT_NONPURGABLE:
+               return KERN_INVALID_ARGUMENT;
+    
+           case VM_OBJECT_PURGABLE_NONVOLATILE:
+               old_state = VM_PURGABLE_NONVOLATILE;
+               break;
+
+           case VM_OBJECT_PURGABLE_VOLATILE:
+               old_state = VM_PURGABLE_VOLATILE;
+               break;
+
+           case VM_OBJECT_PURGABLE_EMPTY:
+               old_state = VM_PURGABLE_EMPTY;
+               break;
+
+           default:
+               old_state = VM_PURGABLE_NONVOLATILE;
+               panic("Bad state (%d) for purgable object!\n",
+                     object->purgable);
+               /*NOTREACHED*/
+       }
+
+       /* purgable cant have delayed copies - now or in the future */
+       assert(object->copy == VM_OBJECT_NULL); 
+       assert(object->copy_strategy == MEMORY_OBJECT_COPY_NONE);
+
+       /*
+        * Execute the desired operation.
+        */
+       if (control == VM_PURGABLE_GET_STATE) {
+               *state = old_state;
+               return KERN_SUCCESS;
+       }
+
+       switch (*state) {
+       case VM_PURGABLE_NONVOLATILE:
+               vm_page_lock_queues();
+               if (object->purgable != VM_OBJECT_PURGABLE_NONVOLATILE) {
+                       assert(vm_page_purgeable_count >=
+                              object->resident_page_count);
+                       vm_page_purgeable_count -= object->resident_page_count;
+               }
+
+               object->purgable = VM_OBJECT_PURGABLE_NONVOLATILE;
 
+               /*
+                * If the object wasn't emptied, then mark all pages of the
+                * object as referenced in order to give them a complete turn
+                * of the virtual memory "clock" before becoming candidates
+                * for paging out (if the system is suffering from memory
+                * pressure).  We don't really need to set the pmap reference
+                * bits (which would be expensive) since the software copies
+                * are believed if they're set to true ...
+                */
+               if (old_state != VM_PURGABLE_EMPTY) {
+                       for (p = (vm_page_t)queue_first(&object->memq);
+                            !queue_end(&object->memq, (queue_entry_t)p);
+                            p = (vm_page_t)queue_next(&p->listq))
+                               p->reference = TRUE;
+               }
+
+               vm_page_unlock_queues();
+
+               break;
+
+       case VM_PURGABLE_VOLATILE:
+               vm_page_lock_queues();
+
+               if (object->purgable != VM_OBJECT_PURGABLE_VOLATILE &&
+                   object->purgable != VM_OBJECT_PURGABLE_EMPTY) {
+                       vm_page_purgeable_count += object->resident_page_count;
+               }
+
+               object->purgable = VM_OBJECT_PURGABLE_VOLATILE;
+
+               /*
+                * We want the newly volatile purgable object to be a
+                * candidate for the pageout scan before other pages in the
+                * application if the system is suffering from memory
+                * pressure.  To do this, we move a page of the object from
+                * the active queue onto the inactive queue in order to
+                * promote the object for early reclaim.  We only need to move
+                * a single page since the pageout scan will reap the entire
+                * purgable object if it finds a single page in a volatile
+                * state.  Obviously we don't do this if there are no pages
+                * associated with the object or we find a page of the object
+                * already on the inactive queue.
+                */
+               for (p = (vm_page_t)queue_first(&object->memq);
+                    !queue_end(&object->memq, (queue_entry_t)p);
+                    p = (vm_page_t)queue_next(&p->listq)) {
+                       if (p->inactive) {
+                               /* already a page on the inactive queue */
+                               break;
+                       }
+                       if (p->active && !p->busy) {
+                               /* found one we can move */
+                               vm_page_deactivate(p);
+                               break;
+                       }
+               }
+               vm_page_unlock_queues();
+
+               break;
+
+
+       case VM_PURGABLE_EMPTY:
+               vm_page_lock_queues();
+               if (object->purgable != VM_OBJECT_PURGABLE_VOLATILE &&
+                   object->purgable != VM_OBJECT_PURGABLE_EMPTY) {
+                       vm_page_purgeable_count += object->resident_page_count;
+               }
+               (void) vm_object_purge(object);
+               vm_page_unlock_queues();
+               break;
+
+       }
+       *state = old_state;
+
+       return KERN_SUCCESS;
+}
 
 #if    TASK_SWAPPER
 /*
@@ -4560,7 +5311,9 @@ vm_object_reference(
  * This is also needed as number of vnodes can be dynamically scaled.
  */
 kern_return_t
-adjust_vm_object_cache(vm_size_t oval, vm_size_t nval)
+adjust_vm_object_cache(
+       __unused vm_size_t oval,
+       vm_size_t nval)
 {
        vm_object_cached_max = nval;
        vm_object_cache_trim(FALSE);
@@ -4568,3 +5321,269 @@ adjust_vm_object_cache(vm_size_t oval, vm_size_t nval)
 }
 #endif /* MACH_BSD */
 
+
+/*
+ * vm_object_transpose
+ *
+ * This routine takes two VM objects of the same size and exchanges
+ * their backing store.
+ * The objects should be "quiesced" via a UPL operation with UPL_SET_IO_WIRE
+ * and UPL_BLOCK_ACCESS if they are referenced anywhere.
+ *
+ * The VM objects must not be locked by caller.
+ */
+kern_return_t
+vm_object_transpose(
+       vm_object_t             object1,
+       vm_object_t             object2,
+       vm_object_size_t        transpose_size)
+{
+       vm_object_t             tmp_object;
+       kern_return_t           retval;
+       boolean_t               object1_locked, object2_locked;
+       boolean_t               object1_paging, object2_paging;
+       vm_page_t               page;
+       vm_object_offset_t      page_offset;
+
+       tmp_object = VM_OBJECT_NULL;
+       object1_locked = FALSE; object2_locked = FALSE;
+       object1_paging = FALSE; object2_paging = FALSE;
+
+       if (object1 == object2 ||
+           object1 == VM_OBJECT_NULL ||
+           object2 == VM_OBJECT_NULL) {
+               /*
+                * If the 2 VM objects are the same, there's
+                * no point in exchanging their backing store.
+                */
+               retval = KERN_INVALID_VALUE;
+               goto done;
+       }
+
+       vm_object_lock(object1);
+       object1_locked = TRUE;
+       if (object1->copy || object1->shadow || object1->shadowed ||
+           object1->purgable != VM_OBJECT_NONPURGABLE) {
+               /*
+                * We don't deal with copy or shadow objects (yet).
+                */
+               retval = KERN_INVALID_VALUE;
+               goto done;
+       }
+       /*
+        * Since we're about to mess with the object's backing store,
+        * mark it as "paging_in_progress".  Note that this is not enough
+        * to prevent any paging activity on this object, so the caller should
+        * have "quiesced" the objects beforehand, via a UPL operation with
+        * UPL_SET_IO_WIRE (to make sure all the pages are there and wired)
+        * and UPL_BLOCK_ACCESS (to mark the pages "busy").
+        */
+       vm_object_paging_begin(object1);
+       object1_paging = TRUE;
+       vm_object_unlock(object1);
+       object1_locked = FALSE;
+
+       /*
+        * Same as above for the 2nd object...
+        */
+       vm_object_lock(object2);
+       object2_locked = TRUE;
+       if (object2->copy || object2->shadow || object2->shadowed ||
+           object2->purgable != VM_OBJECT_NONPURGABLE) {
+               retval = KERN_INVALID_VALUE;
+               goto done;
+       }
+       vm_object_paging_begin(object2);
+       object2_paging = TRUE;
+       vm_object_unlock(object2);
+       object2_locked = FALSE;
+
+       /*
+        * Allocate a temporary VM object to hold object1's contents
+        * while we copy object2 to object1.
+        */
+       tmp_object = vm_object_allocate(transpose_size);
+       vm_object_lock(tmp_object);
+       vm_object_paging_begin(tmp_object);
+       tmp_object->can_persist = FALSE;
+
+       /*
+        * Since we need to lock both objects at the same time,
+        * make sure we always lock them in the same order to
+        * avoid deadlocks.
+        */
+       if (object1 < object2) {
+               vm_object_lock(object1);
+               vm_object_lock(object2);
+       } else {
+               vm_object_lock(object2);
+               vm_object_lock(object1);
+       }
+       object1_locked = TRUE;
+       object2_locked = TRUE;
+
+       if (object1->size != object2->size ||
+           object1->size != transpose_size) {
+               /*
+                * If the 2 objects don't have the same size, we can't
+                * exchange their backing stores or one would overflow.
+                * If their size doesn't match the caller's
+                * "transpose_size", we can't do it either because the
+                * transpose operation will affect the entire span of 
+                * the objects.
+                */
+               retval = KERN_INVALID_VALUE;
+               goto done;
+       }
+
+
+       /*
+        * Transpose the lists of resident pages.
+        */
+       if (object1->phys_contiguous || queue_empty(&object1->memq)) {
+               /*
+                * No pages in object1, just transfer pages
+                * from object2 to object1.  No need to go through
+                * an intermediate object.
+                */
+               while (!queue_empty(&object2->memq)) {
+                       page = (vm_page_t) queue_first(&object2->memq);
+                       vm_page_rename(page, object1, page->offset);
+               }
+               assert(queue_empty(&object2->memq));
+       } else if (object2->phys_contiguous || queue_empty(&object2->memq)) {
+               /*
+                * No pages in object2, just transfer pages
+                * from object1 to object2.  No need to go through
+                * an intermediate object.
+                */
+               while (!queue_empty(&object1->memq)) {
+                       page = (vm_page_t) queue_first(&object1->memq);
+                       vm_page_rename(page, object2, page->offset);
+               }
+               assert(queue_empty(&object1->memq));
+       } else {
+               /* transfer object1's pages to tmp_object */
+               vm_page_lock_queues();
+               while (!queue_empty(&object1->memq)) {
+                       page = (vm_page_t) queue_first(&object1->memq);
+                       page_offset = page->offset;
+                       vm_page_remove(page);
+                       page->offset = page_offset;
+                       queue_enter(&tmp_object->memq, page, vm_page_t, listq);
+               }
+               vm_page_unlock_queues();
+               assert(queue_empty(&object1->memq));
+               /* transfer object2's pages to object1 */
+               while (!queue_empty(&object2->memq)) {
+                       page = (vm_page_t) queue_first(&object2->memq);
+                       vm_page_rename(page, object1, page->offset);
+               }
+               assert(queue_empty(&object2->memq));
+               /* transfer tmp_object's pages to object1 */
+               while (!queue_empty(&tmp_object->memq)) {
+                       page = (vm_page_t) queue_first(&tmp_object->memq);
+                       queue_remove(&tmp_object->memq, page,
+                                    vm_page_t, listq);
+                       vm_page_insert(page, object2, page->offset);
+               }
+               assert(queue_empty(&tmp_object->memq));
+       }
+
+       /* no need to transpose the size: they should be identical */
+       assert(object1->size == object2->size);
+
+#define __TRANSPOSE_FIELD(field)                               \
+MACRO_BEGIN                                                    \
+       tmp_object->field = object1->field;                     \
+       object1->field = object2->field;                        \
+       object2->field = tmp_object->field;                     \
+MACRO_END
+
+       assert(!object1->copy);
+       assert(!object2->copy);
+
+       assert(!object1->shadow);
+       assert(!object2->shadow);
+
+       __TRANSPOSE_FIELD(shadow_offset); /* used by phys_contiguous objects */
+       __TRANSPOSE_FIELD(pager);
+       __TRANSPOSE_FIELD(paging_offset);
+
+       __TRANSPOSE_FIELD(pager_control);
+       /* update the memory_objects' pointers back to the VM objects */
+       if (object1->pager_control != MEMORY_OBJECT_CONTROL_NULL) {
+               memory_object_control_collapse(object1->pager_control,
+                                              object1);
+       }
+       if (object2->pager_control != MEMORY_OBJECT_CONTROL_NULL) {
+               memory_object_control_collapse(object2->pager_control,
+                                              object2);
+       }
+               
+       __TRANSPOSE_FIELD(absent_count);
+
+       assert(object1->paging_in_progress);
+       assert(object2->paging_in_progress);
+
+       __TRANSPOSE_FIELD(pager_created);
+       __TRANSPOSE_FIELD(pager_initialized);
+       __TRANSPOSE_FIELD(pager_ready);
+       __TRANSPOSE_FIELD(pager_trusted);
+       __TRANSPOSE_FIELD(internal);
+       __TRANSPOSE_FIELD(temporary);
+       __TRANSPOSE_FIELD(private);
+       __TRANSPOSE_FIELD(pageout);
+       __TRANSPOSE_FIELD(true_share);
+       __TRANSPOSE_FIELD(phys_contiguous);
+       __TRANSPOSE_FIELD(nophyscache);
+       __TRANSPOSE_FIELD(last_alloc);
+       __TRANSPOSE_FIELD(sequential);
+       __TRANSPOSE_FIELD(cluster_size);
+       __TRANSPOSE_FIELD(existence_map);
+       __TRANSPOSE_FIELD(cow_hint);
+       __TRANSPOSE_FIELD(wimg_bits);
+
+#undef __TRANSPOSE_FIELD
+
+       retval = KERN_SUCCESS;
+
+done:
+       /*
+        * Cleanup.
+        */
+       if (tmp_object != VM_OBJECT_NULL) {
+               vm_object_paging_end(tmp_object);
+               vm_object_unlock(tmp_object);
+               /*
+                * Re-initialize the temporary object to avoid
+                * deallocating a real pager.
+                */
+               _vm_object_allocate(transpose_size, tmp_object);
+               vm_object_deallocate(tmp_object);
+               tmp_object = VM_OBJECT_NULL;
+       }
+
+       if (object1_locked) {
+               vm_object_unlock(object1);
+               object1_locked = FALSE;
+       }
+       if (object2_locked) {
+               vm_object_unlock(object2);
+               object2_locked = FALSE;
+       }
+       if (object1_paging) {
+               vm_object_lock(object1);
+               vm_object_paging_end(object1);
+               vm_object_unlock(object1);
+               object1_paging = FALSE;
+       }
+       if (object2_paging) {
+               vm_object_lock(object2);
+               vm_object_paging_end(object2);
+               vm_object_unlock(object2);
+               object2_paging = FALSE;
+       }
+
+       return retval;
+}