]> git.saurik.com Git - apple/xnu.git/blobdiff - osfmk/vm/memory_object.c
xnu-1228.5.18.tar.gz
[apple/xnu.git] / osfmk / vm / memory_object.c
index c3e0f22f0feeb67c633cc9dcec68adf2dc5a3357..c84e776c6bac2b8798a538c5eadb4180cdce6a9f 100644 (file)
@@ -1,16 +1,19 @@
 /*
- * Copyright (c) 2000 Apple Computer, Inc. All rights reserved.
+ * Copyright (c) 2000-2006 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@
 #include <vm/pmap.h>           /* For pmap_clear_modify */
 #include <vm/vm_kern.h>                /* For kernel_map, vm_move */
 #include <vm/vm_map.h>         /* For vm_map_pageable */
+#include <vm/vm_purgeable_internal.h>  /* Needed by some vm_page.h macros */
 
 #if    MACH_PAGEMAP
 #include <vm/vm_external.h>
 #endif /* MACH_PAGEMAP */
 
+#include <vm/vm_protos.h>
+
 
 memory_object_default_t        memory_manager_default = MEMORY_OBJECT_DEFAULT_NULL;
-vm_size_t              memory_manager_default_cluster = 0;
 decl_mutex_data(,      memory_manager_default_lock)
 
-/*
- *     Forward ref to file-local function:
- */
-boolean_t
-vm_object_update(vm_object_t, vm_object_offset_t, 
-               vm_size_t, memory_object_return_t, int, vm_prot_t);
-
 
 /*
  *     Routine:        memory_object_should_return_page
@@ -207,37 +205,20 @@ memory_object_lock_page(
 
        assert(!m->fictitious);
 
-       if (m->wire_count != 0) {
-               /*
-                *      If no change would take place
-                *      anyway, return successfully.
-                *
-                *      No change means:
-                *              Not flushing AND
-                *              No change to page lock [2 checks]  AND
-                *              Should not return page
-                *
-                * XXX  This doesn't handle sending a copy of a wired
-                * XXX  page to the pager, but that will require some
-                * XXX  significant surgery.
-                */
-               if (!should_flush &&
-                   (m->page_lock == prot || prot == VM_PROT_NO_CHANGE) &&
-                   ! memory_object_should_return_page(m, should_return)) {
-
-                       /*
-                        *      Restart page unlock requests,
-                        *      even though no change took place.
-                        *      [Memory managers may be expecting
-                        *      to see new requests.]
-                        */
-                       m->unlock_request = VM_PROT_NONE;
-                       PAGE_WAKEUP(m);
+       /*
+        *      If the page is wired, just clean or return the page if needed.
+        *      Wired pages don't get flushed or disconnected from the pmap.
+        */
 
-                       return(MEMORY_OBJECT_LOCK_RESULT_DONE);
+       if (m->wire_count != 0) {
+               if (memory_object_should_return_page(m, should_return)) {
+                       if (m->dirty)
+                               return(MEMORY_OBJECT_LOCK_RESULT_MUST_CLEAN);
+                       else
+                               return(MEMORY_OBJECT_LOCK_RESULT_MUST_RETURN);
                }
 
-               return(MEMORY_OBJECT_LOCK_RESULT_MUST_BLOCK);
+               return(MEMORY_OBJECT_LOCK_RESULT_DONE);
        }
 
        /*
@@ -257,36 +238,14 @@ memory_object_lock_page(
         */
 
        if (prot != VM_PROT_NO_CHANGE) {
-               if ((m->page_lock ^ prot) & prot) {
-                       pmap_page_protect(m->phys_page, VM_PROT_ALL & ~prot);
-               }
-#if 0
-               /* code associated with the vestigial 
-                * memory_object_data_unlock
-                */
-               m->page_lock = prot;
-               m->lock_supplied = TRUE;
-               if (prot != VM_PROT_NONE)
-                       m->unusual = TRUE;
-               else
-                       m->unusual = FALSE;
+               pmap_page_protect(m->phys_page, VM_PROT_ALL & ~prot);
 
-               /*
-                *      Restart any past unlock requests, even if no
-                *      change resulted.  If the manager explicitly
-                *      requested no protection change, then it is assumed
-                *      to be remembering past requests.
-                */
-
-               m->unlock_request = VM_PROT_NONE;
-#endif /* 0 */
                PAGE_WAKEUP(m);
        }
 
        /*
         *      Handle page returning.
         */
-
        if (memory_object_should_return_page(m, should_return)) {
 
                /*
@@ -298,12 +257,12 @@ memory_object_lock_page(
                 *      maps now.
                 */
                
-               vm_page_lock_queues();
+               vm_page_lockspin_queues();
                VM_PAGE_QUEUES_REMOVE(m);
                vm_page_unlock_queues();
 
                if (!should_flush)
-                       pmap_page_protect(m->phys_page, VM_PROT_NONE);
+                       pmap_disconnect(m->phys_page);
 
                if (m->dirty)
                        return(MEMORY_OBJECT_LOCK_RESULT_MUST_CLEAN);
@@ -314,12 +273,9 @@ memory_object_lock_page(
        /*
         *      Handle flushing
         */
-
        if (should_flush) {
                VM_PAGE_FREE(m);
        } else {
-               extern boolean_t vm_page_deactivate_hint;
-
                /*
                 *      XXX Make clean but not flush a paging hint,
                 *      and deactivate the pages.  This is a hack
@@ -339,21 +295,33 @@ memory_object_lock_page(
        return(MEMORY_OBJECT_LOCK_RESULT_DONE);
 }
 
-#define LIST_REQ_PAGEOUT_PAGES(object, data_cnt, action, po) \
+#define LIST_REQ_PAGEOUT_PAGES(object, data_cnt, action, po, ro, ioerr, iosync)    \
 MACRO_BEGIN                                                            \
                                                                        \
-       register int            i;                                      \
-       register vm_page_t      hp;                                     \
-                                                                       \
-       vm_object_unlock(object);                                       \
+        register int            upl_flags;                              \
+       memory_object_t         pager;                                  \
+                                                                       \
+       if ((pager = (object)->pager) != MEMORY_OBJECT_NULL) {          \
+               vm_object_paging_begin(object);                         \
+               vm_object_unlock(object);                               \
                                                                        \
-               (void) memory_object_data_return(object->pager,         \
-               po,                                                     \
-               data_cnt,                                               \
-               (action == MEMORY_OBJECT_LOCK_RESULT_MUST_CLEAN),       \
-               !should_flush);                                         \
+                if (iosync)                                            \
+                        upl_flags = UPL_MSYNC | UPL_IOSYNC;            \
+                else                                                   \
+                        upl_flags = UPL_MSYNC;                         \
+                                                                       \
+               (void) memory_object_data_return(pager,                 \
+                       po,                                             \
+                       data_cnt,                                       \
+                       ro,                                             \
+                       ioerr,                                          \
+                       (action) == MEMORY_OBJECT_LOCK_RESULT_MUST_CLEAN,\
+                       !should_flush,                                  \
+                       upl_flags);                                     \
                                                                        \
-       vm_object_lock(object);                                         \
+               vm_object_lock(object);                                 \
+               vm_object_paging_end(object);                           \
+       }                                                               \
 MACRO_END
 
 /*
@@ -386,13 +354,16 @@ memory_object_lock_request(
        memory_object_control_t         control,
        memory_object_offset_t          offset,
        memory_object_size_t            size,
+       memory_object_offset_t  *       resid_offset,
+       int                     *       io_errno,
        memory_object_return_t          should_return,
        int                             flags,
        vm_prot_t                       prot)
 {
        vm_object_t     object;
-       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,
            "m_o_lock_request, control 0x%X off 0x%X size 0x%X flags %X prot %X\n",
@@ -420,7 +391,7 @@ memory_object_lock_request(
        offset -= object->paging_offset;
 
        (void)vm_object_update(object,
-               offset, size, should_return, flags, prot);
+               offset, size, resid_offset, io_errno, should_return, flags, prot);
 
        vm_object_paging_end(object);
        vm_object_unlock(object);
@@ -510,11 +481,13 @@ boolean_t
 vm_object_sync(
        vm_object_t             object,
        vm_object_offset_t      offset,
-       vm_size_t               size,
+       vm_object_size_t        size,
        boolean_t               should_flush,
-       boolean_t               should_return)
+       boolean_t               should_return,
+       boolean_t               should_iosync)
 {
        boolean_t       rv;
+       int             flags;
 
         XPR(XPR_VM_OBJECT,
             "vm_o_sync, object 0x%X, offset 0x%X size 0x%x flush %d rtn %d\n",
@@ -528,12 +501,19 @@ vm_object_sync(
        vm_object_lock(object);
        vm_object_paging_begin(object);
 
-       rv = vm_object_update(object, offset, size,
+       if (should_flush)
+               flags = MEMORY_OBJECT_DATA_FLUSH;
+       else
+               flags = 0;
+
+       if (should_iosync)
+               flags |= MEMORY_OBJECT_IO_SYNC;
+
+       rv = vm_object_update(object, offset, (vm_object_size_t)size, NULL, NULL,
                (should_return) ?
                        MEMORY_OBJECT_RETURN_ALL :
                        MEMORY_OBJECT_RETURN_NONE,
-               (should_flush) ?
-                       MEMORY_OBJECT_DATA_FLUSH : 0,
+               flags,
                VM_PROT_NO_CHANGE);
 
 
@@ -542,6 +522,150 @@ vm_object_sync(
        return rv;
 }
 
+
+
+
+static int
+vm_object_update_extent(
+        vm_object_t            object,
+        vm_object_offset_t     offset,
+       vm_object_offset_t      offset_end,
+       vm_object_offset_t      *offset_resid,
+       int                     *io_errno,
+        boolean_t              should_flush,
+       memory_object_return_t  should_return,
+        boolean_t              should_iosync, 
+        vm_prot_t              prot)
+{
+        vm_page_t      m;
+        int            retval = 0;
+       vm_size_t       data_cnt = 0;
+       vm_object_offset_t      paging_offset = 0;
+       vm_object_offset_t      last_offset = offset;
+        memory_object_lock_result_t    page_lock_result;
+       memory_object_lock_result_t     pageout_action;
+       
+       pageout_action = MEMORY_OBJECT_LOCK_RESULT_DONE;
+
+       for (;
+            offset < offset_end && object->resident_page_count;
+            offset += PAGE_SIZE_64) {
+
+               /*
+                * Limit the number of pages to be cleaned at once.
+                */
+               if (data_cnt >= PAGE_SIZE * MAX_UPL_TRANSFER) {
+                       LIST_REQ_PAGEOUT_PAGES(object, data_cnt, 
+                                              pageout_action, paging_offset, offset_resid, io_errno, should_iosync);
+                       data_cnt = 0;
+               }
+
+               while ((m = vm_page_lookup(object, offset)) != VM_PAGE_NULL) {
+                       page_lock_result = memory_object_lock_page(m, should_return, should_flush, prot);
+
+                       XPR(XPR_MEMORY_OBJECT,
+                           "m_o_update: lock_page, obj 0x%X offset 0x%X result %d\n",
+                           (integer_t)object, offset, page_lock_result, 0, 0);
+
+                       switch (page_lock_result)
+                       {
+                         case MEMORY_OBJECT_LOCK_RESULT_DONE:
+                           /*
+                            *  End of a cluster of dirty pages.
+                            */
+                           if (data_cnt) {
+                                   LIST_REQ_PAGEOUT_PAGES(object, 
+                                                          data_cnt, pageout_action, 
+                                                          paging_offset, offset_resid, io_errno, should_iosync);
+                                   data_cnt = 0;
+                                   continue;
+                           }
+                           break;
+
+                         case MEMORY_OBJECT_LOCK_RESULT_MUST_BLOCK:
+                           /*
+                            *  Since it is necessary to block,
+                            *  clean any dirty pages now.
+                            */
+                           if (data_cnt) {
+                                   LIST_REQ_PAGEOUT_PAGES(object,
+                                                          data_cnt, pageout_action, 
+                                                          paging_offset, offset_resid, io_errno, should_iosync);
+                                   data_cnt = 0;
+                                   continue;
+                           }
+                           PAGE_SLEEP(object, m, THREAD_UNINT);
+                           continue;
+
+                         case MEMORY_OBJECT_LOCK_RESULT_MUST_CLEAN:
+                         case MEMORY_OBJECT_LOCK_RESULT_MUST_RETURN:
+                           /*
+                            * The clean and return cases are similar.
+                            *
+                            * if this would form a discontiguous block,
+                            * clean the old pages and start anew.
+                            *
+                            * Mark the page busy since we will unlock the
+                            * object if we issue the LIST_REQ_PAGEOUT
+                            */
+                           m->busy = TRUE;
+                           if (data_cnt && 
+                               ((last_offset != offset) || (pageout_action != page_lock_result))) {
+                                   LIST_REQ_PAGEOUT_PAGES(object, 
+                                                          data_cnt, pageout_action, 
+                                                          paging_offset, offset_resid, io_errno, should_iosync);
+                                   data_cnt = 0;
+                           }
+                           m->busy = FALSE;
+
+                           if (m->cleaning) {
+                                   PAGE_SLEEP(object, m, THREAD_UNINT);
+                                   continue;
+                           }
+                           if (data_cnt == 0) {
+                                   pageout_action = page_lock_result;
+                                   paging_offset = offset;
+                           }
+                           data_cnt += PAGE_SIZE;
+                           last_offset = offset + PAGE_SIZE_64;
+
+                           vm_page_lockspin_queues();
+                           /*
+                            * Clean
+                            */
+                           m->list_req_pending = TRUE;
+                           m->cleaning = TRUE;
+
+                           if (should_flush) {
+                                   /*
+                                    * and add additional state
+                                    * for the flush
+                                    */
+                                   m->busy = TRUE;
+                                   m->pageout = TRUE;
+                                   vm_page_wire(m);
+                           }
+                           vm_page_unlock_queues();
+
+                           retval = 1;
+                           break;
+                       }
+                       break;
+               }
+       }
+       /*
+        *      We have completed the scan for applicable pages.
+        *      Clean any pages that have been saved.
+        */
+       if (data_cnt) {
+               LIST_REQ_PAGEOUT_PAGES(object,
+                                      data_cnt, pageout_action, paging_offset, offset_resid, io_errno, should_iosync);
+       }
+       return (retval);
+}
+
+
+
 /*
  *     Routine:        vm_object_update
  *     Description:
@@ -553,24 +677,28 @@ kern_return_t
 vm_object_update(
        register vm_object_t            object,
        register vm_object_offset_t     offset,
-       register vm_size_t              size,
+       register vm_object_size_t       size,
+       register vm_object_offset_t     *resid_offset,
+       int                             *io_errno,
        memory_object_return_t          should_return,
        int                             flags,
-       vm_prot_t                       prot)
+       vm_prot_t                       protection)
 {
-       register vm_page_t      m;
-       vm_page_t               holding_page;
-       vm_size_t               original_size = size;
-       vm_object_offset_t      paging_offset = 0;
-       vm_object_t             copy_object;
-       vm_size_t               data_cnt = 0;
-       vm_object_offset_t      last_offset = offset;
-       memory_object_lock_result_t     page_lock_result;
-       memory_object_lock_result_t     pageout_action;
+        vm_object_t            copy_object = VM_OBJECT_NULL;
        boolean_t               data_returned = FALSE;
        boolean_t               update_cow;
-       boolean_t               should_flush = flags & MEMORY_OBJECT_DATA_FLUSH;
-       boolean_t               pending_pageout = FALSE;
+       boolean_t               should_flush = (flags & MEMORY_OBJECT_DATA_FLUSH) ? TRUE : FALSE;
+       boolean_t               should_iosync = (flags & MEMORY_OBJECT_IO_SYNC) ? TRUE : FALSE;
+       int                     num_of_extents;
+       int                     n;
+#define MAX_EXTENTS    8
+#define EXTENT_SIZE    (1024 * 1024 * 256)
+#define RESIDENT_LIMIT (1024 * 32)
+       struct extent {
+               vm_object_offset_t e_base;
+               vm_object_offset_t e_min;
+               vm_object_offset_t e_max;
+       } extents[MAX_EXTENTS];
 
        /*
         *      To avoid blocking while scanning for pages, save
@@ -594,89 +722,123 @@ vm_object_update(
                                        !(flags & MEMORY_OBJECT_DATA_PURGE)))
                                || (flags & MEMORY_OBJECT_COPY_SYNC);
                        
+       if (update_cow || (flags & (MEMORY_OBJECT_DATA_PURGE | MEMORY_OBJECT_DATA_SYNC))) {
+               int collisions = 0;
+
+               while ((copy_object = object->copy) != VM_OBJECT_NULL) {
+                       /*
+                        * need to do a try here since we're swimming upstream
+                        * against the normal lock ordering... however, we need
+                        * to hold the object stable until we gain control of the
+                        * copy object so we have to be careful how we approach this
+                        */
+                       if (vm_object_lock_try(copy_object)) {
+                              /*
+                               * we 'won' the lock on the copy object...
+                               * no need to hold the object lock any longer...
+                               * take a real reference on the copy object because
+                               * we're going to call vm_fault_page on it which may
+                               * under certain conditions drop the lock and the paging
+                               * reference we're about to take... the reference
+                               * will keep the copy object from going away if that happens
+                               */
+                              vm_object_unlock(object);
+                              vm_object_reference_locked(copy_object);
+                              break;
+                       }
+                       vm_object_unlock(object);
+
+                       collisions++;
+                       mutex_pause(collisions);
 
-       if((((copy_object = object->copy) != NULL) && update_cow) ||
-                                       (flags & MEMORY_OBJECT_DATA_SYNC)) {
-               vm_size_t               i;
-               vm_size_t               copy_size;
-               vm_object_offset_t      copy_offset;
+                       vm_object_lock(object);
+               }
+       }
+       if ((copy_object != VM_OBJECT_NULL && update_cow) || (flags & MEMORY_OBJECT_DATA_SYNC)) {
+               vm_map_size_t           i;
+               vm_map_size_t           copy_size;
+               vm_map_offset_t         copy_offset;
                vm_prot_t               prot;
                vm_page_t               page;
                vm_page_t               top_page;
                kern_return_t           error = 0;
+               struct vm_object_fault_info fault_info;
+
+               if (copy_object != VM_OBJECT_NULL) {
+                       /*
+                        * translate offset with respect to shadow's offset
+                        */
+                       copy_offset = (offset >= copy_object->shadow_offset) ?
+                         (vm_map_offset_t)(offset - copy_object->shadow_offset) :
+                         (vm_map_offset_t) 0;
+
+                       if (copy_offset > copy_object->size)
+                               copy_offset = copy_object->size;
+
+                       /*
+                        * clip size with respect to shadow offset
+                        */
+                       if (offset >= copy_object->shadow_offset) {
+                               copy_size = size;
+                       } else if (size >= copy_object->shadow_offset - offset) {
+                               copy_size = size - (copy_object->shadow_offset - offset);
+                       } else {
+                               copy_size = 0;
+                       }
+                       
+                       if (copy_offset + copy_size > copy_object->size) {
+                               if (copy_object->size >= copy_offset) {
+                                       copy_size = copy_object->size - copy_offset;
+                               } else {
+                                       copy_size = 0;
+                               }
+                       }
+                       copy_size+=copy_offset;
 
-               if(copy_object != NULL) {
-                  /* translate offset with respect to shadow's offset */
-                  copy_offset = (offset >= copy_object->shadow_offset)?
-                       offset - copy_object->shadow_offset :
-                       (vm_object_offset_t) 0;
-                  if(copy_offset > copy_object->size)
-                       copy_offset = copy_object->size;
-
-                  /* clip size with respect to shadow offset */
-                  copy_size = (offset >= copy_object->shadow_offset) ?
-                       size : size - (copy_object->shadow_offset - offset);
-
-                  if(copy_size <= 0) {
-                       copy_size = 0;
-                  } else {
-                       copy_size = ((copy_offset + copy_size) 
-                               <= copy_object->size) ?
-                               copy_size : copy_object->size - copy_offset;
-                  }
-                  /* check for a copy_offset which is beyond the end of */
-                  /* the copy_object */
-                  if(copy_size < 0)
-                       copy_size = 0;
-
-                  copy_size+=offset;
-
-                  vm_object_unlock(object);
-                  vm_object_lock(copy_object);
                } else {
                        copy_object = object;
 
                        copy_size   = offset + size;
                        copy_offset = offset;
                }
+               fault_info.interruptible = THREAD_UNINT;
+               fault_info.behavior  = VM_BEHAVIOR_SEQUENTIAL;
+               fault_info.user_tag  = 0;
+               fault_info.lo_offset = copy_offset;
+               fault_info.hi_offset = copy_size;
+               fault_info.no_cache   = FALSE;
 
                vm_object_paging_begin(copy_object);
-               for (i=copy_offset; i<copy_size; i+=PAGE_SIZE) {
+
+               for (i = copy_offset; i < copy_size; i += PAGE_SIZE) {
        RETRY_COW_OF_LOCK_REQUEST:
+                       fault_info.cluster_size = copy_size - i;
+
                        prot =  VM_PROT_WRITE|VM_PROT_READ;
                        switch (vm_fault_page(copy_object, i, 
                                VM_PROT_WRITE|VM_PROT_READ,
                                FALSE,
-                               THREAD_UNINT,
-                               copy_offset,
-                               copy_offset+copy_size,
-                               VM_BEHAVIOR_SEQUENTIAL,
                                &prot,
                                &page,
                                &top_page,
                                (int *)0,
                                &error,
                                FALSE,
-                               FALSE, NULL, 0)) {
+                               FALSE, &fault_info)) {
 
                        case VM_FAULT_SUCCESS:
-                               if(top_page) {
+                               if (top_page) {
                                        vm_fault_cleanup(
                                                page->object, top_page);
-                                       PAGE_WAKEUP_DONE(page);
-                                       vm_page_lock_queues();
-                                       if (!page->active && !page->inactive)
-                                               vm_page_activate(page);
-                                       vm_page_unlock_queues();
                                        vm_object_lock(copy_object);
                                        vm_object_paging_begin(copy_object);
-                               } else {
-                                       PAGE_WAKEUP_DONE(page);
-                                       vm_page_lock_queues();
-                                       if (!page->active && !page->inactive)
-                                               vm_page_activate(page);
-                                       vm_page_unlock_queues();
                                }
+                               vm_page_lock_queues();
+                               if (!page->active && !page->inactive)
+                                       vm_page_deactivate(page);
+                               vm_page_unlock_queues();
+
+                               PAGE_WAKEUP_DONE(page);
                                break;
                        case VM_FAULT_RETRY:
                                prot =  VM_PROT_WRITE|VM_PROT_READ;
@@ -701,168 +863,166 @@ vm_object_update(
                                vm_object_paging_begin(copy_object);
                                goto RETRY_COW_OF_LOCK_REQUEST;
                        case VM_FAULT_MEMORY_ERROR:
+                               if (object != copy_object)
+                                       vm_object_deallocate(copy_object);
                                vm_object_lock(object);
                                goto BYPASS_COW_COPYIN;
                        }
 
                }
                vm_object_paging_end(copy_object);
-               if(copy_object != object) {
+       }
+       if ((flags & (MEMORY_OBJECT_DATA_SYNC | MEMORY_OBJECT_COPY_SYNC))) {
+               if (copy_object != VM_OBJECT_NULL && copy_object != object) {
                        vm_object_unlock(copy_object);
+                       vm_object_deallocate(copy_object);
                        vm_object_lock(object);
                }
+               return KERN_SUCCESS;
        }
-       if((flags & (MEMORY_OBJECT_DATA_SYNC | MEMORY_OBJECT_COPY_SYNC))) {
-                       return KERN_SUCCESS;
-       }
-       if(((copy_object = object->copy) != NULL) && 
-                                       (flags & MEMORY_OBJECT_DATA_PURGE)) {
-               copy_object->shadow_severed = TRUE;
-               copy_object->shadowed = FALSE;
-               copy_object->shadow = NULL;
-               /* delete the ref the COW was holding on the target object */
-               vm_object_deallocate(object);
+       if (copy_object != VM_OBJECT_NULL && copy_object != object) {
+               if ((flags & MEMORY_OBJECT_DATA_PURGE)) {
+                       copy_object->shadow_severed = TRUE;
+                       copy_object->shadowed = FALSE;
+                       copy_object->shadow = NULL;
+                       /*
+                        * delete the ref the COW was holding on the target object
+                        */
+                       vm_object_deallocate(object);
+               }
+               vm_object_unlock(copy_object);
+               vm_object_deallocate(copy_object);
+               vm_object_lock(object);
        }
 BYPASS_COW_COPYIN:
 
-       for (;
-            size != 0;
-            size -= PAGE_SIZE, offset += PAGE_SIZE_64)
-       {
-           /*
-            *  Limit the number of pages to be cleaned at once.
-            */
-           if (pending_pageout &&
-                   data_cnt >= PAGE_SIZE * DATA_WRITE_MAX)
-           {
-               LIST_REQ_PAGEOUT_PAGES(object, data_cnt, 
-                               pageout_action, paging_offset);
-               data_cnt = 0;
-               pending_pageout = FALSE;
-           }
-
-           while ((m = vm_page_lookup(object, offset)) != VM_PAGE_NULL) {
-               page_lock_result = memory_object_lock_page(m, should_return,
-                                       should_flush, prot);
-
-               XPR(XPR_MEMORY_OBJECT,
-                    "m_o_update: lock_page, obj 0x%X offset 0x%X result %d\n",
-                    (integer_t)object, offset, page_lock_result, 0, 0);
+       /*
+        * when we have a really large range to check relative
+        * to the number of actual resident pages, we'd like
+        * to use the resident page list to drive our checks
+        * however, the object lock will get dropped while processing
+        * the page which means the resident queue can change which
+        * means we can't walk the queue as we process the pages
+        * we also want to do the processing in offset order to allow
+        * 'runs' of pages to be collected if we're being told to 
+        * flush to disk... the resident page queue is NOT ordered.
+        * 
+        * a temporary solution (until we figure out how to deal with
+        * large address spaces more generically) is to pre-flight
+        * the resident page queue (if it's small enough) and develop
+        * a collection of extents (that encompass actual resident pages)
+        * to visit.  This will at least allow us to deal with some of the
+        * more pathological cases in a more efficient manner.  The current
+        * worst case (a single resident page at the end of an extremely large
+        * range) can take minutes to complete for ranges in the terrabyte
+        * category... since this routine is called when truncating a file,
+        * and we currently support files up to 16 Tbytes in size, this
+        * is not a theoretical problem
+        */
 
-               switch (page_lock_result)
-               {
-                   case MEMORY_OBJECT_LOCK_RESULT_DONE:
-                       /*
-                        *      End of a cluster of dirty pages.
-                        */
-                       if(pending_pageout) {
-                               LIST_REQ_PAGEOUT_PAGES(object, 
-                                       data_cnt, pageout_action, 
-                                       paging_offset);
-                               data_cnt = 0;
-                               pending_pageout = FALSE;
-                               continue;
-                       }
-                       break;
+       if ((object->resident_page_count < RESIDENT_LIMIT) && 
+           (atop_64(size) > (unsigned)(object->resident_page_count/(8 * MAX_EXTENTS)))) {
+               vm_page_t               next;
+               vm_object_offset_t      start;
+               vm_object_offset_t      end;
+               vm_object_size_t        e_mask;
+               vm_page_t               m;
 
-                   case MEMORY_OBJECT_LOCK_RESULT_MUST_BLOCK:
-                       /*
-                        *      Since it is necessary to block,
-                        *      clean any dirty pages now.
-                        */
-                       if(pending_pageout) {
-                               LIST_REQ_PAGEOUT_PAGES(object,
-                                       data_cnt, pageout_action, 
-                                       paging_offset);
-                               pending_pageout = FALSE;
-                               data_cnt = 0;
-                               continue;
-                       }
+               start = offset;
+               end   = offset + size;
+               num_of_extents = 0;
+               e_mask = ~((vm_object_size_t)(EXTENT_SIZE - 1));
 
-                       PAGE_SLEEP(object, m, THREAD_UNINT);
-                       continue;
+               m = (vm_page_t) queue_first(&object->memq);
 
-                   case MEMORY_OBJECT_LOCK_RESULT_MUST_CLEAN:
-                   case MEMORY_OBJECT_LOCK_RESULT_MUST_RETURN:
-                       /*
-                        * The clean and return cases are similar.
-                        *
-                        */
+               while (!queue_end(&object->memq, (queue_entry_t) m)) {
+                       next = (vm_page_t) queue_next(&m->listq);
 
-                       /*
-                        * if this would form a discontiguous block,
-                        * clean the old pages and start anew.
-                        *
-                        */
-
-                       /*
-                        * Mark the page busy since we unlock the
-                        * object below.
-                        */
-                       m->busy = TRUE;
-                       if (pending_pageout &&
-                           (last_offset != offset ||
-                            pageout_action != page_lock_result)) {
-                               LIST_REQ_PAGEOUT_PAGES(object, 
-                                               data_cnt, pageout_action, 
-                                               paging_offset);
-                               pending_pageout = FALSE;
-                               data_cnt = 0;
-                       }
-                       m->busy = FALSE;
-                       holding_page = VM_PAGE_NULL;
-                       if(m->cleaning) {
-                               PAGE_SLEEP(object, m, THREAD_UNINT);
-                               continue;
-                       }
-                       if(!pending_pageout) {
-                               pending_pageout = TRUE;
-                               pageout_action = page_lock_result;
-                               paging_offset = offset;
-                       }
-                       if (should_flush) {
-                               vm_page_lock_queues();
-                               m->list_req_pending = TRUE;
-                               m->cleaning = TRUE;
-                               m->busy = TRUE;
-                               m->pageout = TRUE;
-                               vm_page_wire(m);
-                               vm_page_unlock_queues();
-                       } else {
-                               /*
-                                * Clean but do not flush
+                       if ((m->offset >= start) && (m->offset < end)) {
+                               /*
+                                * this is a page we're interested in
+                                * try to fit it into a current extent
                                 */
-                               vm_page_lock_queues();
-                               m->list_req_pending = TRUE;
-                               m->cleaning = TRUE;
-                               vm_page_unlock_queues();
-
+                               for (n = 0; n < num_of_extents; n++) {
+                                       if ((m->offset & e_mask) == extents[n].e_base) {
+                                               /*
+                                                * use (PAGE_SIZE - 1) to determine the
+                                                * max offset so that we don't wrap if
+                                                * we're at the last page of the space
+                                                */
+                                               if (m->offset < extents[n].e_min)
+                                                       extents[n].e_min = m->offset;
+                                               else if ((m->offset + (PAGE_SIZE - 1)) > extents[n].e_max)
+                                                       extents[n].e_max = m->offset + (PAGE_SIZE - 1);
+                                               break;
+                                       }
+                               }
+                               if (n == num_of_extents) {
+                                       /*
+                                        * didn't find a current extent that can encompass
+                                        * this page
+                                        */
+                                       if (n < MAX_EXTENTS) {
+                                               /*
+                                                * if we still have room, 
+                                                * create a new extent
+                                                */
+                                               extents[n].e_base = m->offset & e_mask;
+                                               extents[n].e_min  = m->offset;
+                                               extents[n].e_max  = m->offset + (PAGE_SIZE - 1);
+
+                                               num_of_extents++;
+                                       } else {
+                                               /*
+                                                * no room to create a new extent...
+                                                * fall back to a single extent based
+                                                * on the min and max page offsets 
+                                                * we find in the range we're interested in...
+                                                * first, look through the extent list and
+                                                * develop the overall min and max for the
+                                                * pages we've looked at up to this point
+                                                */                                             
+                                               for (n = 1; n < num_of_extents; n++) {
+                                                       if (extents[n].e_min < extents[0].e_min)
+                                                               extents[0].e_min = extents[n].e_min;
+                                                       if (extents[n].e_max > extents[0].e_max)
+                                                               extents[0].e_max = extents[n].e_max;
+                                               }
+                                               /*
+                                                * now setup to run through the remaining pages
+                                                * to determine the overall min and max
+                                                * offset for the specified range
+                                                */
+                                               extents[0].e_base = 0;
+                                               e_mask = 0;
+                                               num_of_extents = 1;
+
+                                               /*
+                                                * by continuing, we'll reprocess the
+                                                * page that forced us to abandon trying
+                                                * to develop multiple extents
+                                                */
+                                               continue;
+                                       }
+                               }
                        }
-                       vm_object_unlock(object);
-
-
-                       data_cnt += PAGE_SIZE;
-                       last_offset = offset + PAGE_SIZE_64;
-                       data_returned = TRUE;
-
-                       vm_object_lock(object);
-                       break;
+                       m = next;
                }
-               break;
-           }
-       }
+       } else {
+               extents[0].e_min = offset;
+               extents[0].e_max = offset + (size - 1);
 
-       /*
-        *      We have completed the scan for applicable pages.
-        *      Clean any pages that have been saved.
-        */
-       if (pending_pageout) {
-           LIST_REQ_PAGEOUT_PAGES(object,
-                               data_cnt, pageout_action, paging_offset);
+               num_of_extents = 1;
+       }
+       for (n = 0; n < num_of_extents; n++) {
+               if (vm_object_update_extent(object, extents[n].e_min, extents[n].e_max, resid_offset, io_errno,
+                                           should_flush, should_return, should_iosync, protection))
+                       data_returned = TRUE;
        }
        return (data_returned);
 }
 
+
 /*
  *     Routine:        memory_object_synchronize_completed [user interface]
  *
@@ -883,6 +1043,8 @@ memory_object_synchronize_completed(
        vm_object_t                     object;
        msync_req_t                     msr;
 
+       object = memory_object_control_to_vm_object(control);
+
         XPR(XPR_MEMORY_OBJECT,
            "m_o_sync_completed, object 0x%X, offset 0x%X length 0x%X\n",
            (integer_t)object, offset, length, 0, 0);
@@ -891,7 +1053,6 @@ memory_object_synchronize_completed(
         *      Look for bogus arguments
         */
 
-       object = memory_object_control_to_vm_object(control);
        if (object == VM_OBJECT_NULL)
                return (KERN_INVALID_ARGUMENT);
 
@@ -927,7 +1088,6 @@ vm_object_set_attributes_common(
        boolean_t       may_cache,
        memory_object_copy_strategy_t copy_strategy,
        boolean_t       temporary,
-       vm_size_t       cluster_size,
         boolean_t      silent_overwrite,
        boolean_t       advisory_pageout)
 {
@@ -961,17 +1121,6 @@ vm_object_set_attributes_common(
                may_cache = TRUE;
        if (temporary)
                temporary = TRUE;
-       if (cluster_size != 0) {
-               int     pages_per_cluster;
-               pages_per_cluster = atop_32(cluster_size);
-               /*
-                * Cluster size must be integral multiple of page size,
-                * and be a power of 2 number of pages.
-                */
-               if ((cluster_size & (PAGE_SIZE-1)) ||
-                   ((pages_per_cluster-1) & pages_per_cluster))
-                       return KERN_INVALID_ARGUMENT;
-       }
 
        vm_object_lock(object);
 
@@ -985,12 +1134,6 @@ vm_object_set_attributes_common(
        object->temporary = temporary;
        object->silent_overwrite = silent_overwrite;
        object->advisory_pageout = advisory_pageout;
-       if (cluster_size == 0)
-               cluster_size = PAGE_SIZE;
-       object->cluster_size = cluster_size;
-
-       assert(cluster_size >= PAGE_SIZE &&
-              cluster_size % PAGE_SIZE == 0);
 
        /*
         *      Wake up anyone waiting for the ready attribute
@@ -1027,7 +1170,6 @@ memory_object_change_attributes(
        boolean_t                       temporary;
        boolean_t                       may_cache;
        boolean_t                       invalidate;
-       vm_size_t                       cluster_size;
        memory_object_copy_strategy_t   copy_strategy;
        boolean_t                       silent_overwrite;
        boolean_t                       advisory_pageout;
@@ -1046,7 +1188,6 @@ memory_object_change_attributes(
 #if notyet
        invalidate = object->invalidate;
 #endif
-       cluster_size = object->cluster_size;
        vm_object_unlock(object);       
 
        switch (flavor) {
@@ -1099,7 +1240,6 @@ memory_object_change_attributes(
                 perf = (memory_object_perf_info_t) attributes;
 
                may_cache = perf->may_cache;
-               cluster_size = round_page_32(perf->cluster_size);
 
                break;
            }
@@ -1117,7 +1257,6 @@ memory_object_change_attributes(
 
                 may_cache = attr->may_cache;
                 copy_strategy = attr->copy_strategy;
-               cluster_size = page_size;
 
                break;
            }
@@ -1135,7 +1274,6 @@ memory_object_change_attributes(
 
                copy_strategy = attr->copy_strategy;
                 may_cache = attr->may_cache_object;
-               cluster_size = attr->cluster_size;
                temporary = attr->temporary;
 
                break;
@@ -1164,7 +1302,6 @@ memory_object_change_attributes(
                                                     may_cache,
                                                     copy_strategy,
                                                     temporary,
-                                                    cluster_size,
                                                     silent_overwrite,
                                                     advisory_pageout));
 }
@@ -1241,7 +1378,7 @@ memory_object_get_attributes(
                }
 
                perf = (memory_object_perf_info_t) attributes;
-               perf->cluster_size = object->cluster_size;
+               perf->cluster_size = PAGE_SIZE;
                perf->may_cache = object->can_persist;
 
                *count = MEMORY_OBJECT_PERF_INFO_COUNT;
@@ -1276,7 +1413,7 @@ memory_object_get_attributes(
 
                 attr = (memory_object_attr_info_t) attributes;
                attr->copy_strategy = object->copy_strategy;
-               attr->cluster_size = object->cluster_size;
+               attr->cluster_size = PAGE_SIZE;
                attr->may_cache_object = object->can_persist;
                attr->temporary = object->temporary;
 
@@ -1295,6 +1432,141 @@ memory_object_get_attributes(
 }
 
 
+kern_return_t
+memory_object_iopl_request(
+       ipc_port_t              port,
+       memory_object_offset_t  offset,
+       upl_size_t              *upl_size,
+       upl_t                   *upl_ptr,
+       upl_page_info_array_t   user_page_list,
+       unsigned int            *page_list_count,
+       int                     *flags)
+{
+       vm_object_t             object;
+       kern_return_t           ret;
+       int                     caller_flags;
+
+       caller_flags = *flags;
+
+       if (caller_flags & ~UPL_VALID_FLAGS) {
+               /*
+                * For forward compatibility's sake,
+                * reject any unknown flag.
+                */
+               return KERN_INVALID_VALUE;
+       }
+
+       if (ip_kotype(port) == IKOT_NAMED_ENTRY) {
+               vm_named_entry_t        named_entry;
+
+               named_entry = (vm_named_entry_t)port->ip_kobject;
+               /* a few checks to make sure user is obeying rules */
+               if(*upl_size == 0) {
+                       if(offset >= named_entry->size)
+                               return(KERN_INVALID_RIGHT);
+                       *upl_size = named_entry->size - offset;
+               }
+               if(caller_flags & UPL_COPYOUT_FROM) {
+                       if((named_entry->protection & VM_PROT_READ) 
+                                               != VM_PROT_READ) {
+                               return(KERN_INVALID_RIGHT);
+                       }
+               } else {
+                       if((named_entry->protection & 
+                               (VM_PROT_READ | VM_PROT_WRITE)) 
+                               != (VM_PROT_READ | VM_PROT_WRITE)) {
+                               return(KERN_INVALID_RIGHT);
+                       }
+               }
+               if(named_entry->size < (offset + *upl_size))
+                       return(KERN_INVALID_ARGUMENT);
+
+               /* the callers parameter offset is defined to be the */
+               /* offset from beginning of named entry offset in object */
+               offset = offset + named_entry->offset;
+
+               if(named_entry->is_sub_map) 
+                       return (KERN_INVALID_ARGUMENT);
+               
+               named_entry_lock(named_entry);
+
+               if (named_entry->is_pager) {
+                       object = vm_object_enter(named_entry->backing.pager, 
+                                       named_entry->offset + named_entry->size, 
+                                       named_entry->internal, 
+                                       FALSE,
+                                       FALSE);
+                       if (object == VM_OBJECT_NULL) {
+                               named_entry_unlock(named_entry);
+                               return(KERN_INVALID_OBJECT);
+                       }
+
+                       /* JMM - drop reference on pager here? */
+
+                       /* create an extra reference for the named entry */
+                       vm_object_lock(object);
+                       vm_object_reference_locked(object);
+                       named_entry->backing.object = object;
+                       named_entry->is_pager = FALSE;
+                       named_entry_unlock(named_entry);
+
+                       /* wait for object to be ready */
+                       while (!object->pager_ready) {
+                               vm_object_wait(object,
+                                               VM_OBJECT_EVENT_PAGER_READY,
+                                               THREAD_UNINT);
+                               vm_object_lock(object);
+                       }
+                       vm_object_unlock(object);
+               } else {
+                       /* This is the case where we are going to map */
+                       /* an already mapped object.  If the object is */
+                       /* not ready it is internal.  An external     */
+                       /* object cannot be mapped until it is ready  */
+                       /* we can therefore avoid the ready check     */
+                       /* in this case.  */
+                       object = named_entry->backing.object;
+                       vm_object_reference(object);
+                       named_entry_unlock(named_entry);
+               }
+       } else if (ip_kotype(port) == IKOT_MEM_OBJ_CONTROL) {
+               memory_object_control_t control;
+               control = (memory_object_control_t) port;
+               if (control == NULL)
+                       return (KERN_INVALID_ARGUMENT);
+               object = memory_object_control_to_vm_object(control);
+               if (object == VM_OBJECT_NULL)
+                       return (KERN_INVALID_ARGUMENT);
+               vm_object_reference(object);
+       } else {
+               return KERN_INVALID_ARGUMENT;
+       }
+       if (object == VM_OBJECT_NULL)
+               return (KERN_INVALID_ARGUMENT);
+
+       if (!object->private) {
+               if (*upl_size > (MAX_UPL_TRANSFER*PAGE_SIZE))
+                       *upl_size = (MAX_UPL_TRANSFER*PAGE_SIZE);
+               if (object->phys_contiguous) {
+                       *flags = UPL_PHYS_CONTIG;
+               } else {
+                       *flags = 0;
+               }
+       } else {
+               *flags = UPL_DEV_MEMORY | UPL_PHYS_CONTIG;
+       }
+
+       ret = vm_object_iopl_request(object,
+                                    offset,
+                                    *upl_size,
+                                    upl_ptr,
+                                    user_page_list,
+                                    page_list_count,
+                                    caller_flags);
+       vm_object_deallocate(object);
+       return ret;
+}
+
 /*  
  *     Routine:        memory_object_upl_request [interface]
  *     Purpose:
@@ -1308,7 +1580,7 @@ kern_return_t
 memory_object_upl_request(
        memory_object_control_t control,
        memory_object_offset_t  offset,
-       vm_size_t               size,
+       upl_size_t              size,
        upl_t                   *upl_ptr,
        upl_page_info_array_t   user_page_list,
        unsigned int            *page_list_count,
@@ -1344,8 +1616,8 @@ kern_return_t
 memory_object_super_upl_request(
        memory_object_control_t control,
        memory_object_offset_t  offset,
-       vm_size_t               size,
-       vm_size_t               super_cluster,
+       upl_size_t              size,
+       upl_size_t              super_cluster,
        upl_t                   *upl,
        upl_page_info_t         *user_page_list,
        unsigned int            *page_list_count,
@@ -1367,6 +1639,27 @@ memory_object_super_upl_request(
                                           cntrl_flags);
 }
 
+kern_return_t
+memory_object_cluster_size(memory_object_control_t control, memory_object_offset_t *start,
+                          vm_size_t *length, memory_object_fault_info_t fault_info)
+{
+       vm_object_t             object;
+
+       object = memory_object_control_to_vm_object(control);
+
+       if (object == VM_OBJECT_NULL || object->paging_offset > *start)
+               return (KERN_INVALID_ARGUMENT);
+
+       *start -= object->paging_offset;
+
+       vm_object_cluster_size(object, (vm_object_offset_t *)start, length, (vm_object_fault_info_t)fault_info);
+
+       *start += object->paging_offset;
+
+       return (KERN_SUCCESS);
+}
+
+
 int vm_stat_discard_cleared_reply = 0;
 int vm_stat_discard_cleared_unset = 0;
 int vm_stat_discard_cleared_too_late = 0;
@@ -1385,11 +1678,12 @@ kern_return_t
 host_default_memory_manager(
        host_priv_t             host_priv,
        memory_object_default_t *default_manager,
-       vm_size_t               cluster_size)
+       __unused memory_object_cluster_size_t cluster_size)
 {
        memory_object_default_t current_manager;
        memory_object_default_t new_manager;
        memory_object_default_t returned_manager;
+       kern_return_t result = KERN_SUCCESS;
 
        if (host_priv == HOST_PRIV_NULL)
                return(KERN_INVALID_HOST);
@@ -1399,14 +1693,26 @@ host_default_memory_manager(
        new_manager = *default_manager;
        mutex_lock(&memory_manager_default_lock);
        current_manager = memory_manager_default;
+       returned_manager = MEMORY_OBJECT_DEFAULT_NULL;
 
        if (new_manager == MEMORY_OBJECT_DEFAULT_NULL) {
                /*
                 *      Retrieve the current value.
                 */
-               memory_object_default_reference(current_manager);
                returned_manager = current_manager;
+               memory_object_default_reference(returned_manager);
        } else {
+
+               /*
+                *      If this is the first non-null manager, start
+                *      up the internal pager support.
+                */
+               if (current_manager == MEMORY_OBJECT_DEFAULT_NULL) {
+                       result = vm_pageout_internal_start();
+                       if (result != KERN_SUCCESS)
+                               goto out;
+               }
+
                /*
                 *      Retrieve the current value,
                 *      and replace it with the supplied value.
@@ -1414,21 +1720,10 @@ host_default_memory_manager(
                 *      but we have to take a reference on the new
                 *      one.
                 */
-
                returned_manager = current_manager;
                memory_manager_default = new_manager;
                memory_object_default_reference(new_manager);
 
-               if (cluster_size % PAGE_SIZE != 0) {
-#if 0
-                       mutex_unlock(&memory_manager_default_lock);
-                       return KERN_INVALID_ARGUMENT;
-#else
-                       cluster_size = round_page_32(cluster_size);
-#endif
-               }
-               memory_manager_default_cluster = cluster_size;
-
                /*
                 *      In case anyone's been waiting for a memory
                 *      manager to be established, wake them up.
@@ -1436,11 +1731,11 @@ host_default_memory_manager(
 
                thread_wakeup((event_t) &memory_manager_default);
        }
-
+ out:
        mutex_unlock(&memory_manager_default_lock);
 
        *default_manager = returned_manager;
-       return(KERN_SUCCESS);
+       return(result);
 }
 
 /*
@@ -1452,8 +1747,7 @@ host_default_memory_manager(
  */
 
 __private_extern__ memory_object_default_t
-memory_manager_default_reference(
-       vm_size_t       *cluster_size)
+memory_manager_default_reference(void)
 {
        memory_object_default_t current_manager;
 
@@ -1469,7 +1763,6 @@ memory_manager_default_reference(
                current_manager = memory_manager_default;
        }
        memory_object_default_reference(current_manager);
-       *cluster_size = memory_manager_default_cluster;
        mutex_unlock(&memory_manager_default_lock);
 
        return current_manager;
@@ -1512,95 +1805,10 @@ __private_extern__ void
 memory_manager_default_init(void)
 {
        memory_manager_default = MEMORY_OBJECT_DEFAULT_NULL;
-       mutex_init(&memory_manager_default_lock, ETAP_VM_MEMMAN);
+       mutex_init(&memory_manager_default_lock, 0);
 }
 
 
-void
-memory_object_deactivate_pages(
-       vm_object_t             object,
-       vm_object_offset_t      offset,
-       vm_object_size_t        size,
-       boolean_t               kill_page)
-{
-       vm_object_t             orig_object;
-       int pages_moved = 0;
-       int pages_found = 0;
-
-       /*
-        * entered with object lock held, acquire a paging reference to
-        * prevent the memory_object and control ports from
-        * being destroyed.
-        */
-       orig_object = object;
-
-       for (;;) {
-               register vm_page_t      m;
-               vm_object_offset_t      toffset;
-               vm_object_size_t        tsize;
-
-               vm_object_paging_begin(object);
-               vm_page_lock_queues();
-
-               for (tsize = size, toffset = offset; tsize; tsize -= PAGE_SIZE, toffset += PAGE_SIZE) {
-
-                       if ((m = vm_page_lookup(object, toffset)) != VM_PAGE_NULL) {
-
-                               pages_found++;
-
-                               if ((m->wire_count == 0) && (!m->private) && (!m->gobbled) && (!m->busy)) {
-
-                                       m->reference = FALSE;
-                                       pmap_clear_reference(m->phys_page);
-
-                                       if ((kill_page) && (object->internal)) {
-                                               m->precious = FALSE;
-                                               m->dirty = FALSE;
-                                               pmap_clear_modify(m->phys_page);
-                                               vm_external_state_clr(object->existence_map, offset);
-                                       }
-                                       VM_PAGE_QUEUES_REMOVE(m);
-
-                                       if(m->zero_fill) {
-                                               queue_enter_first(
-                                                       &vm_page_queue_zf, 
-                                                       m, vm_page_t, pageq);
-                                       } else {
-                                               queue_enter_first(
-                                                       &vm_page_queue_inactive, 
-                                                       m, vm_page_t, pageq);
-                                       }
-
-                                       m->inactive = TRUE;
-                                       if (!m->fictitious)  
-                                               vm_page_inactive_count++;
-
-                                       pages_moved++;
-                               }
-                       }
-               }
-               vm_page_unlock_queues();
-               vm_object_paging_end(object);
-
-               if (object->shadow) {
-                       vm_object_t     tmp_object;
-
-                       kill_page = 0;
-
-                       offset += object->shadow_offset;
-
-                       tmp_object = object->shadow;
-                       vm_object_lock(tmp_object);
-
-                       if (object != orig_object)
-                               vm_object_unlock(object);
-                       object = tmp_object;
-               } else
-                       break;
-       }
-       if (object != orig_object)
-               vm_object_unlock(object);
-}
 
 /* Allow manipulation of individual page state.  This is actually part of */
 /* the UPL regimen but takes place on the object rather than on a UPL */
@@ -1614,107 +1822,81 @@ memory_object_page_op(
        int                     *flags)
 {
        vm_object_t             object;
-       vm_page_t               dst_page;
-
 
        object = memory_object_control_to_vm_object(control);
        if (object == VM_OBJECT_NULL)
                return (KERN_INVALID_ARGUMENT);
 
-       vm_object_lock(object);
+       return vm_object_page_op(object, offset, ops, phys_entry, flags);
+}
 
-       if(ops & UPL_POP_PHYSICAL) {
-               if(object->phys_contiguous) {
-                       if (phys_entry) {
-                               *phys_entry = (ppnum_t)
-                                       (object->shadow_offset >> 12);
-                       }
-                       vm_object_unlock(object);
-                       return KERN_SUCCESS;
-               } else {
-                       vm_object_unlock(object);
-                       return KERN_INVALID_OBJECT;
-               }
-       }
+/*
+ * memory_object_range_op offers performance enhancement over 
+ * memory_object_page_op for page_op functions which do not require page 
+ * level state to be returned from the call.  Page_op was created to provide 
+ * a low-cost alternative to page manipulation via UPLs when only a single 
+ * page was involved.  The range_op call establishes the ability in the _op 
+ * family of functions to work on multiple pages where the lack of page level
+ * state handling allows the caller to avoid the overhead of the upl structures.
+ */
 
-       while(TRUE) {
-               if(object->phys_contiguous) {
-                       vm_object_unlock(object);
-                       return KERN_INVALID_OBJECT;
-               }
+kern_return_t
+memory_object_range_op(
+       memory_object_control_t control,
+       memory_object_offset_t  offset_beg,
+       memory_object_offset_t  offset_end,
+       int                     ops,
+       int                     *range)
+{
+       vm_object_t             object;
 
-               if((dst_page = vm_page_lookup(object,offset)) == VM_PAGE_NULL) {
-                       vm_object_unlock(object);
-                       return KERN_FAILURE;
-               }
+       object = memory_object_control_to_vm_object(control);
+       if (object == VM_OBJECT_NULL)
+               return (KERN_INVALID_ARGUMENT);
 
-               /* Sync up on getting the busy bit */
-               if((dst_page->busy || dst_page->cleaning) && 
-                          (((ops & UPL_POP_SET) && 
-                          (ops & UPL_POP_BUSY)) || (ops & UPL_POP_DUMP))) {
-                       /* someone else is playing with the page, we will */
-                       /* have to wait */
-                       PAGE_SLEEP(object, dst_page, THREAD_UNINT);
-                       continue;
-               }
+       return vm_object_range_op(object,
+                                 offset_beg,
+                                 offset_end,
+                                 ops,
+                                 range);
+}
 
-               if (ops & UPL_POP_DUMP) {
-                 vm_page_lock_queues();
-                  vm_page_free(dst_page);
-                 vm_page_unlock_queues();
-                 break;
-               }
 
-               if (flags) {
-                       *flags = 0;
+kern_return_t
+memory_object_pages_resident(
+       memory_object_control_t control,
+       boolean_t                       *       has_pages_resident)
+{
+       vm_object_t             object;
 
-                       /* Get the condition of flags before requested ops */
-                       /* are undertaken */
+       *has_pages_resident = FALSE;
 
-                       if(dst_page->dirty) *flags |= UPL_POP_DIRTY;
-                       if(dst_page->pageout) *flags |= UPL_POP_PAGEOUT;
-                       if(dst_page->precious) *flags |= UPL_POP_PRECIOUS;
-                       if(dst_page->absent) *flags |= UPL_POP_ABSENT;
-                       if(dst_page->busy) *flags |= UPL_POP_BUSY;
-               }
-               if (phys_entry)
-                       *phys_entry = dst_page->phys_page;
+       object = memory_object_control_to_vm_object(control);
+       if (object == VM_OBJECT_NULL)
+               return (KERN_INVALID_ARGUMENT);
+
+       if (object->resident_page_count)
+               *has_pages_resident = TRUE;
        
-               /* The caller should have made a call either contingent with */
-               /* or prior to this call to set UPL_POP_BUSY */
-               if(ops & UPL_POP_SET) {
-                       /* The protection granted with this assert will */
-                       /* not be complete.  If the caller violates the */
-                       /* convention and attempts to change page state */
-                       /* without first setting busy we may not see it */
-                       /* because the page may already be busy.  However */
-                       /* if such violations occur we will assert sooner */
-                       /* or later. */
-                       assert(dst_page->busy || (ops & UPL_POP_BUSY));
-                       if (ops & UPL_POP_DIRTY) dst_page->dirty = TRUE;
-                       if (ops & UPL_POP_PAGEOUT) dst_page->pageout = TRUE;
-                       if (ops & UPL_POP_PRECIOUS) dst_page->precious = TRUE;
-                       if (ops & UPL_POP_ABSENT) dst_page->absent = TRUE;
-                       if (ops & UPL_POP_BUSY) dst_page->busy = TRUE;
-               }
+       return (KERN_SUCCESS);
+}
 
-               if(ops & UPL_POP_CLR) {
-                       assert(dst_page->busy);
-                       if (ops & UPL_POP_DIRTY) dst_page->dirty = FALSE;
-                       if (ops & UPL_POP_PAGEOUT) dst_page->pageout = FALSE;
-                       if (ops & UPL_POP_PRECIOUS) dst_page->precious = FALSE;
-                       if (ops & UPL_POP_ABSENT) dst_page->absent = FALSE;
-                       if (ops & UPL_POP_BUSY) {
-                               dst_page->busy = FALSE;
-                               PAGE_WAKEUP(dst_page);
-                       }
-               }
-               break;
-       }
+kern_return_t
+memory_object_signed(
+       memory_object_control_t control,
+       boolean_t               is_signed)
+{
+       vm_object_t     object;
+
+       object = memory_object_control_to_vm_object(control);
+       if (object == VM_OBJECT_NULL)
+               return KERN_INVALID_ARGUMENT;
 
+       vm_object_lock(object);
+       object->code_signed = is_signed;
        vm_object_unlock(object);
+
        return KERN_SUCCESS;
-                               
 }
 
 static zone_t mem_obj_control_zone;
@@ -1736,8 +1918,10 @@ memory_object_control_allocate(
        memory_object_control_t control;
 
        control = (memory_object_control_t)zalloc(mem_obj_control_zone);
-       if (control != MEMORY_OBJECT_CONTROL_NULL)
-               control->object = object;
+       if (control != MEMORY_OBJECT_CONTROL_NULL) {
+               control->moc_object = object;
+               control->moc_ikot = IKOT_MEM_OBJ_CONTROL; /* fake ip_kotype */
+       }
        return (control);
 }
 
@@ -1746,24 +1930,25 @@ memory_object_control_collapse(
        memory_object_control_t control,                       
        vm_object_t             object)
 {                     
-       assert((control->object != VM_OBJECT_NULL) &&
-              (control->object != object));
-       control->object = object;
+       assert((control->moc_object != VM_OBJECT_NULL) &&
+              (control->moc_object != object));
+       control->moc_object = object;
 }
 
 __private_extern__ vm_object_t
 memory_object_control_to_vm_object(
        memory_object_control_t control)
 {
-       if (control == MEMORY_OBJECT_CONTROL_NULL)
+       if (control == MEMORY_OBJECT_CONTROL_NULL ||
+           control->moc_ikot != IKOT_MEM_OBJ_CONTROL)
                return VM_OBJECT_NULL;
 
-       return (control->object);
+       return (control->moc_object);
 }
 
 memory_object_control_t
 convert_port_to_mo_control(
-       mach_port_t     port)
+       __unused mach_port_t    port)
 {
        return MEMORY_OBJECT_CONTROL_NULL;
 }
@@ -1771,14 +1956,14 @@ convert_port_to_mo_control(
 
 mach_port_t
 convert_mo_control_to_port(
-       memory_object_control_t control)
+       __unused memory_object_control_t        control)
 {
        return MACH_PORT_NULL;
 }
 
 void
 memory_object_control_reference(
-       memory_object_control_t control)
+       __unused memory_object_control_t        control)
 {
        return;
 }
@@ -1792,15 +1977,15 @@ void
 memory_object_control_deallocate(
        memory_object_control_t control)
 {
-       zfree(mem_obj_control_zone, (vm_offset_t)control);
+       zfree(mem_obj_control_zone, control);
 }
 
 void
 memory_object_control_disable(
        memory_object_control_t control)
 {
-       assert(control->object != VM_OBJECT_NULL);
-       control->object = VM_OBJECT_NULL;
+       assert(control->moc_object != VM_OBJECT_NULL);
+       control->moc_object = VM_OBJECT_NULL;
 }
 
 void
@@ -1819,7 +2004,7 @@ memory_object_default_deallocate(
 
 memory_object_t
 convert_port_to_memory_object(
-       mach_port_t     port)
+       __unused mach_port_t    port)
 {
        return (MEMORY_OBJECT_NULL);
 }
@@ -1827,54 +2012,26 @@ convert_port_to_memory_object(
 
 mach_port_t
 convert_memory_object_to_port(
-       memory_object_t object)
+       __unused memory_object_t        object)
 {
        return (MACH_PORT_NULL);
 }
 
-#ifdef MACH_BSD
-/* remove after component interface available */
-extern int     vnode_pager_workaround;
-extern int     device_pager_workaround;
-#endif
-
 
 /* Routine memory_object_reference */
 void memory_object_reference(
        memory_object_t memory_object)
 {
-extern void   dp_memory_object_reference(memory_object_t);
-
-#ifdef MACH_BSD
- extern void   vnode_pager_reference(memory_object_t);
- extern void   device_pager_reference(memory_object_t);
-
-               if(memory_object->pager == &vnode_pager_workaround) {
-                       vnode_pager_reference(memory_object);
-               } else if(memory_object->pager == &device_pager_workaround) {
-                       device_pager_reference(memory_object);
-               } else
-#endif
-                       dp_memory_object_reference(memory_object);
+       (memory_object->mo_pager_ops->memory_object_reference)(
+               memory_object);
 }
 
 /* Routine memory_object_deallocate */
 void memory_object_deallocate(
        memory_object_t memory_object)
 {
-extern void   dp_memory_object_deallocate(memory_object_t);
-
-#ifdef MACH_BSD
- extern void   vnode_pager_deallocate(memory_object_t);
- extern void   device_pager_deallocate(memory_object_t);
-
-               if(memory_object->pager == &vnode_pager_workaround) {
-                       vnode_pager_deallocate(memory_object);
-               } else if(memory_object->pager == &device_pager_workaround) {
-                       device_pager_deallocate(memory_object);
-               } else
-#endif
-                       dp_memory_object_deallocate(memory_object);
+       (memory_object->mo_pager_ops->memory_object_deallocate)(
+                memory_object);
 }
 
 
@@ -1883,33 +2040,13 @@ kern_return_t memory_object_init
 (
        memory_object_t memory_object,
        memory_object_control_t memory_control,
-       vm_size_t memory_object_page_size
+       memory_object_cluster_size_t memory_object_page_size
 )
 {
-extern kern_return_t   dp_memory_object_init(memory_object_t,
-                                            memory_object_control_t,
-                                            vm_size_t);
-#ifdef MACH_BSD
-extern kern_return_t   vnode_pager_init(memory_object_t,
-                                       memory_object_control_t,
-                                       vm_size_t);
-extern kern_return_t   device_pager_init(memory_object_t,
-                                       memory_object_control_t,
-                                       vm_size_t);
-
-               if(memory_object->pager == &vnode_pager_workaround) {
-                       return vnode_pager_init(memory_object,
-                               memory_control,
-                               memory_object_page_size);
-               } else if(memory_object->pager == &device_pager_workaround) {
-                       return device_pager_init(memory_object,
-                               memory_control,
-                               memory_object_page_size);
-               } else
-#endif
-                       return dp_memory_object_init(memory_object,
-                               memory_control,
-                               memory_object_page_size);
+       return (memory_object->mo_pager_ops->memory_object_init)(
+               memory_object,
+               memory_control,
+               memory_object_page_size);
 }
 
 /* Routine memory_object_terminate */
@@ -1918,19 +2055,8 @@ kern_return_t memory_object_terminate
        memory_object_t memory_object
 )
 {
-extern kern_return_t dp_memory_object_terminate(memory_object_t);
-
-#ifdef MACH_BSD
-extern kern_return_t vnode_pager_terminate(memory_object_t);
-extern kern_return_t device_pager_terminate(memory_object_t);
-
-       if(memory_object->pager == &vnode_pager_workaround) {
-               return vnode_pager_terminate(memory_object);
-       } else if(memory_object->pager == &device_pager_workaround) {
-               return device_pager_terminate(memory_object);
-       } else
-#endif
-               return dp_memory_object_terminate(memory_object);
+       return (memory_object->mo_pager_ops->memory_object_terminate)(
+               memory_object);
 }
 
 /* Routine memory_object_data_request */
@@ -1938,35 +2064,17 @@ kern_return_t memory_object_data_request
 (
        memory_object_t memory_object,
        memory_object_offset_t offset,
-       vm_size_t length,
-       vm_prot_t desired_access
+       memory_object_cluster_size_t length,
+       vm_prot_t desired_access,
+       memory_object_fault_info_t fault_info
 )
 {
-extern kern_return_t   dp_memory_object_data_request(memory_object_t, 
-                       memory_object_offset_t, vm_size_t, vm_prot_t);
-
-#ifdef MACH_BSD
-extern kern_return_t   vnode_pager_data_request(memory_object_t, 
-                       memory_object_offset_t, vm_size_t, vm_prot_t);
-extern kern_return_t   device_pager_data_request(memory_object_t, 
-                       memory_object_offset_t, vm_size_t, vm_prot_t);
-
-               if (memory_object->pager == &vnode_pager_workaround) {
-                       return vnode_pager_data_request(memory_object, 
-                                     offset, 
-                                     length,
-                                     desired_access);
-               } else if (memory_object->pager == &device_pager_workaround) {
-                       return device_pager_data_request(memory_object, 
-                                     offset, 
-                                     length,
-                                     desired_access);
-               } else
-#endif
-                       return dp_memory_object_data_request(memory_object, 
-                                       offset, 
-                                       length,
-                                       desired_access);
+       return (memory_object->mo_pager_ops->memory_object_data_request)(
+               memory_object,
+               offset, 
+               length,
+               desired_access,
+               fault_info);
 }
 
 /* Routine memory_object_data_return */
@@ -1975,46 +2083,22 @@ kern_return_t memory_object_data_return
        memory_object_t memory_object,
        memory_object_offset_t offset,
        vm_size_t size,
+       memory_object_offset_t *resid_offset,
+       int     *io_error,
        boolean_t dirty,
-       boolean_t kernel_copy
+       boolean_t kernel_copy,
+       int     upl_flags
 )
 {
-  extern kern_return_t dp_memory_object_data_return(memory_object_t,
-                                                   memory_object_offset_t,
-                                                   vm_size_t,
-                                                   boolean_t,
-                                                   boolean_t);
-#ifdef MACH_BSD
-  extern kern_return_t vnode_pager_data_return(memory_object_t,
-                                                   memory_object_offset_t,
-                                                   vm_size_t,
-                                                   boolean_t,
-                                                   boolean_t);
-  extern kern_return_t device_pager_data_return(memory_object_t,
-                                                   memory_object_offset_t,
-                                                   vm_size_t,
-                                                   boolean_t,
-                                                   boolean_t);
-
-       if (memory_object->pager == &vnode_pager_workaround) {
-               return vnode_pager_data_return(memory_object,
-                                      offset,
-                                      size,
-                                      dirty,
-                                      kernel_copy);
-       } else if (memory_object->pager == &device_pager_workaround) {
-               return device_pager_data_return(memory_object,
-                                      offset,
-                                      size,
-                                      dirty,
-                                      kernel_copy);
-       } else
-#endif
-               return dp_memory_object_data_return(memory_object,
-                                      offset,
-                                      size,
-                                      dirty,
-                                      kernel_copy);
+       return (memory_object->mo_pager_ops->memory_object_data_return)(
+               memory_object,
+               offset,
+               size,
+               resid_offset,
+               io_error,
+               dirty,
+               kernel_copy,
+               upl_flags);
 }
 
 /* Routine memory_object_data_initialize */
@@ -2025,31 +2109,10 @@ kern_return_t memory_object_data_initialize
        vm_size_t size
 )
 {
-
-  extern kern_return_t dp_memory_object_data_initialize(memory_object_t,
-                                                   memory_object_offset_t,
-                                                   vm_size_t);
-#ifdef MACH_BSD
-  extern kern_return_t vnode_pager_data_initialize(memory_object_t,
-                                                   memory_object_offset_t,
-                                                   vm_size_t);
-  extern kern_return_t device_pager_data_initialize(memory_object_t,
-                                                   memory_object_offset_t,
-                                                   vm_size_t);
-
-       if (memory_object->pager == &vnode_pager_workaround) {
-               return vnode_pager_data_initialize(memory_object,
-                                      offset,
-                                      size);
-       } else if (memory_object->pager == &device_pager_workaround) {
-               return device_pager_data_initialize(memory_object,
-                                      offset,
-                                      size);
-       } else
-#endif
-               return dp_memory_object_data_initialize(memory_object,
-                                      offset,
-                                      size);
+       return (memory_object->mo_pager_ops->memory_object_data_initialize)(
+               memory_object,
+               offset,
+               size);
 }
 
 /* Routine memory_object_data_unlock */
@@ -2061,37 +2124,11 @@ kern_return_t memory_object_data_unlock
        vm_prot_t desired_access
 )
 {
-  extern kern_return_t dp_memory_object_data_unlock(memory_object_t,
-                                                   memory_object_offset_t,
-                                                   vm_size_t,
-                                                   vm_prot_t);
-#ifdef MACH_BSD
-  extern kern_return_t vnode_pager_data_unlock(memory_object_t,
-                                                   memory_object_offset_t,
-                                                   vm_size_t,
-                                                   vm_prot_t);
-  extern kern_return_t device_pager_data_unlock(memory_object_t,
-                                                   memory_object_offset_t,
-                                                   vm_size_t,
-                                                   vm_prot_t);
-
-       if (memory_object->pager == &vnode_pager_workaround) {
-               return vnode_pager_data_unlock(memory_object,
-                                      offset,
-                                      size,
-                                      desired_access);
-       } else if (memory_object->pager == &device_pager_workaround) {
-               return device_pager_data_unlock(memory_object,
-                                      offset,
-                                      size,
-                                      desired_access);
-       } else
-#endif
-               return dp_memory_object_data_unlock(memory_object,
-                                      offset,
-                                      size,
-                                      desired_access);
-
+       return (memory_object->mo_pager_ops->memory_object_data_unlock)(
+               memory_object,
+               offset,
+               size,
+               desired_access);
 }
 
 /* Routine memory_object_synchronize */
@@ -2103,39 +2140,11 @@ kern_return_t memory_object_synchronize
        vm_sync_t sync_flags
 )
 {
-  extern kern_return_t dp_memory_object_data_synchronize(memory_object_t,
-                                                   memory_object_offset_t,
-                                                   vm_size_t,
-                                                   vm_sync_t);
-#ifdef MACH_BSD
-  extern kern_return_t vnode_pager_data_synchronize(memory_object_t,
-                                                   memory_object_offset_t,
-                                                   vm_size_t,
-                                                   vm_sync_t);
-  extern kern_return_t device_pager_data_synchronize(memory_object_t,
-                                                   memory_object_offset_t,
-                                                   vm_size_t,
-                                                   vm_sync_t);
-
-       if (memory_object->pager == &vnode_pager_workaround) {
-                       return vnode_pager_synchronize(
-                               memory_object,
-                               offset,
-                               size,
-                               sync_flags);
-       } else if (memory_object->pager == &device_pager_workaround) {
-                       return device_pager_synchronize(
-                               memory_object,
-                               offset,
-                               size,
-                               sync_flags);
-       } else
-#endif
-                       return dp_memory_object_synchronize(
-                               memory_object,
-                               offset,
-                               size,
-                               sync_flags);
+       return (memory_object->mo_pager_ops->memory_object_synchronize)(
+               memory_object,
+               offset,
+               size,
+               sync_flags);
 }
 
 /* Routine memory_object_unmap */
@@ -2144,18 +2153,8 @@ kern_return_t memory_object_unmap
        memory_object_t memory_object
 )
 {
-       extern kern_return_t dp_memory_object_unmap(memory_object_t);
-#ifdef MACH_BSD
-       extern kern_return_t vnode_pager_unmap(memory_object_t);
-       extern kern_return_t device_pager_unmap(memory_object_t);
-
-       if (memory_object->pager == &vnode_pager_workaround) {
-                       return vnode_pager_unmap(memory_object);
-       } else if (memory_object->pager == &device_pager_workaround) {
-                       return device_pager_unmap(memory_object);
-       } else
-#endif
-                       return dp_memory_object_unmap(memory_object);
+       return (memory_object->mo_pager_ops->memory_object_unmap)(
+               memory_object);
 }
 
 /* Routine memory_object_create */
@@ -2166,12 +2165,41 @@ kern_return_t memory_object_create
        memory_object_t *new_memory_object
 )
 {
-extern kern_return_t default_pager_memory_object_create(memory_object_default_t,
-                                                       vm_size_t,
-                                                       memory_object_t *);
-
        return default_pager_memory_object_create(default_memory_manager,
                                                  new_memory_object_size,
                                                  new_memory_object);
 }
 
+upl_t
+convert_port_to_upl(
+       ipc_port_t      port)
+{
+       upl_t upl;
+
+       ip_lock(port);
+       if (!ip_active(port) || (ip_kotype(port) != IKOT_UPL)) {
+                       ip_unlock(port);
+                       return (upl_t)NULL;
+       }
+       upl = (upl_t) port->ip_kobject;
+       ip_unlock(port);
+       upl_lock(upl);
+       upl->ref_count+=1;
+       upl_unlock(upl);
+       return upl;
+}
+
+mach_port_t
+convert_upl_to_port(
+       __unused upl_t          upl)
+{
+       return MACH_PORT_NULL;
+}
+
+__private_extern__ void
+upl_no_senders(
+       __unused ipc_port_t                             port,
+       __unused mach_port_mscount_t    mscount)
+{
+       return;
+}