]> git.saurik.com Git - apple/xnu.git/blobdiff - osfmk/vm/vm_object.h
xnu-3789.70.16.tar.gz
[apple/xnu.git] / osfmk / vm / vm_object.h
index 54fc6f9a9d3aa4b42fe5973a2f92ef8e9e99a2aa..daef8471195f24126a2596eb8fc3fa1169852590 100644 (file)
@@ -1,16 +1,19 @@
 /*
- * Copyright (c) 2000 Apple Computer, Inc. All rights reserved.
+ * Copyright (c) 2000-2007 Apple 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@
@@ -63,6 +66,8 @@
 #ifndef        _VM_VM_OBJECT_H_
 #define _VM_VM_OBJECT_H_
 
+#include <debug.h>
+#include <mach_assert.h>
 #include <mach_pagemap.h>
 #include <task_swapper.h>
 
 #include <mach/memory_object_types.h>
 #include <mach/port.h>
 #include <mach/vm_prot.h>
+#include <mach/vm_param.h>
 #include <mach/machine/vm_types.h>
 #include <kern/queue.h>
-#include <kern/lock.h>
+#include <kern/locks.h>
 #include <kern/assert.h>
-#include <kern/ipc_mig.h>
 #include <kern/misc_protos.h>
 #include <kern/macro_help.h>
 #include <ipc/ipc_types.h>
 #include <vm/pmap.h>
 
-#if    MACH_PAGEMAP
 #include <vm/vm_external.h>
-#endif /* MACH_PAGEMAP */
 
-typedef memory_object_control_t        pager_request_t;
-#define        PAGER_REQUEST_NULL      ((pager_request_t) 0)
+#include <vm/vm_options.h>
+#include <vm/vm_page.h>
+
+#if VM_OBJECT_TRACKING
+#include <libkern/OSDebug.h>
+#include <kern/btlog.h>
+extern void vm_object_tracking_init(void);
+extern boolean_t vm_object_tracking_inited;
+extern btlog_t *vm_object_tracking_btlog;
+#define VM_OBJECT_TRACKING_NUM_RECORDS 50000
+#define VM_OBJECT_TRACKING_BTDEPTH 7
+#define VM_OBJECT_TRACKING_OP_CREATED  1
+#define VM_OBJECT_TRACKING_OP_MODIFIED 2
+#define VM_OBJECT_TRACKING_OP_TRUESHARE        3
+#endif /* VM_OBJECT_TRACKING */
+
+struct vm_page;
+struct vm_shared_region_slide_info;
 
 /*
  *     Types defined:
  *
  *     vm_object_t             Virtual memory object.
+ *     vm_object_fault_info_t  Used to determine cluster size.
  */
 
-typedef unsigned long long vm_object_size_t;
+struct vm_object_fault_info {
+       int             interruptible;
+        uint32_t       user_tag;
+        vm_size_t      cluster_size;
+        vm_behavior_t  behavior;
+        vm_map_offset_t        lo_offset;
+       vm_map_offset_t hi_offset;
+       unsigned int
+       /* boolean_t */ no_cache:1,
+       /* boolean_t */ stealth:1,
+       /* boolean_t */ io_sync:1,
+       /* boolean_t */ cs_bypass:1,
+       /* boolean_t */ mark_zf_absent:1,
+       /* boolean_t */ batch_pmap_op:1,
+               __vm_object_fault_info_unused_bits:26;
+       int             pmap_options;
+};
 
 
-struct vm_object {
-       queue_head_t            memq;           /* Resident memory */
-       decl_mutex_data(,       Lock)           /* Synchronization */
+#define        vo_size                         vo_un1.vou_size
+#define vo_cache_pages_to_scan         vo_un1.vou_cache_pages_to_scan
+#define vo_shadow_offset               vo_un2.vou_shadow_offset
+#define vo_cache_ts                    vo_un2.vou_cache_ts
+#define vo_purgeable_owner             vo_un2.vou_purgeable_owner
+#define vo_slide_info                  vo_un2.vou_slide_info
 
-       vm_object_size_t        size;           /* Object size (only valid
-                                                * if internal)
-                                                */
-       vm_object_size_t        frozen_size;    /* How much has been marked
-                                                * copy-on-write (only
-                                                * valid if copy_symmetric)
-                                                */
+struct vm_object {
+       /*
+        * on 64 bit systems we pack the pointers hung off the memq.
+        * those pointers have to be able to point back to the memq.
+        * the packed pointers are required to be on a 64 byte boundary
+        * which means 2 things for the vm_object...  (1) the memq
+        * struct has to be the first element of the structure so that
+        * we can control it's alignment... (2) the vm_object must be
+        * aligned on a 64 byte boundary... for static vm_object's
+        * this is accomplished via the 'aligned' attribute... for
+        * vm_object's in the zone pool, this is accomplished by
+        * rounding the size of the vm_object element to the nearest
+        * 64 byte size before creating the zone.
+        */
+       vm_page_queue_head_t    memq;           /* Resident memory - must be first */
+        lck_rw_t               Lock;           /* Synchronization */
+
+#if DEVELOPMENT || DEBUG
+       thread_t                Lock_owner;
+#endif
+       union {
+               vm_object_size_t  vou_size;     /* Object size (only valid if internal) */
+               int               vou_cache_pages_to_scan;      /* pages yet to be visited in an
+                                                                * external object in cache
+                                                                */
+       } vo_un1;
+
+       struct vm_page          *memq_hint;
        int                     ref_count;      /* Number of references */
-#if    TASK_SWAPPER
-       int                     res_count;      /* Residency references (swap)*/
-#endif /* TASK_SWAPPER */
        unsigned int            resident_page_count;
                                                /* number of resident pages */
+       unsigned int            wired_page_count; /* number of wired pages */
+       unsigned int            reusable_page_count;
 
        struct vm_object        *copy;          /* Object that should receive
                                                 * a copy of my changed pages,
@@ -123,24 +182,47 @@ struct vm_object {
                                                 * copy_call.
                                                 */
        struct vm_object        *shadow;        /* My shadow */
-       vm_object_offset_t      shadow_offset;  /* Offset into shadow */
+
+       union {
+               vm_object_offset_t vou_shadow_offset;   /* Offset into shadow */
+               clock_sec_t     vou_cache_ts;   /* age of an external object
+                                                * present in cache
+                                                */
+               task_t          vou_purgeable_owner;    /* If the purg'a'ble bits below are set 
+                                                        * to volatile/emtpy, this is the task 
+                                                        * that owns this purgeable object.
+                                                        */
+               struct vm_shared_region_slide_info *vou_slide_info;
+       } vo_un2;
 
        memory_object_t         pager;          /* Where to get data */
        vm_object_offset_t      paging_offset;  /* Offset into memory object */
-       pager_request_t         pager_request;  /* Where data comes back */
+       memory_object_control_t pager_control;  /* Where data comes back */
 
        memory_object_copy_strategy_t
                                copy_strategy;  /* How to handle data copy */
 
-       unsigned int            absent_count;   /* The number of pages that
-                                                * have been requested but
-                                                * not filled.  That is, the
-                                                * number of pages for which
-                                                * the "absent" attribute is
-                                                * asserted.
-                                                */
-
-       unsigned int            paging_in_progress;
+#if __LP64__
+       /*
+        * Some user processes (mostly VirtualMachine software) take a large
+        * number of UPLs (via IOMemoryDescriptors) to wire pages in large
+        * VM objects and overflow the 16-bit "activity_in_progress" counter.
+        * Since we never enforced any limit there, let's give them 32 bits
+        * for backwards compatibility's sake.
+        */
+       unsigned int            paging_in_progress:16,
+                               __object1_unused_bits:16;
+       unsigned int            activity_in_progress;
+#else /* __LP64__ */
+       /*
+        * On 32-bit platforms, enlarging "activity_in_progress" would increase
+        * the size of "struct vm_object".  Since we don't know of any actual
+        * overflow of these counters on these platforms, let's keep the
+        * counters as 16-bit integers.
+        */
+       unsigned short          paging_in_progress;
+       unsigned short          activity_in_progress;
+#endif /* __LP64__ */
                                                /* The memory object ports are
                                                 * being used (e.g., for pagein
                                                 * or pageout) -- don't change
@@ -148,6 +230,7 @@ struct vm_object {
                                                 * don't collapse, destroy or
                                                 * terminate)
                                                 */
+
        unsigned int
        /* boolean_t array */   all_wanted:11,  /* Bit array of "want to be
                                                 * awakened" notations.  See
@@ -188,19 +271,13 @@ struct vm_object {
                                                 * a real memory object. */
        /* boolean_t */         alive:1,        /* Not yet terminated */
 
-       /* boolean_t */         lock_in_progress:1,
-                                               /* Is a multi-page lock
-                                                * request in progress?
-                                                */
-       /* boolean_t */         lock_restart:1,
-                                               /* Should lock request in
-                                                * progress restart search?
+       /* boolean_t */         purgable:2,     /* Purgable state.  See
+                                                * VM_PURGABLE_* 
                                                 */
+       /* boolean_t */         purgeable_when_ripe:1, /* Purgeable when a token
+                                                       * becomes ripe.
+                                                       */
        /* boolean_t */         shadowed:1,     /* Shadow may exist */
-       /* boolean_t */         silent_overwrite:1,
-                                               /* Allow full page overwrite
-                                                * without data_request if
-                                                * page is absent */
        /* boolean_t */         advisory_pageout:1,
                                                /* Instead of sending page
                                                 * via OOL, just notify
@@ -252,8 +329,6 @@ struct vm_object {
                                                 * primary caching. (for
                                                 * I/O)
                                                 */
-                                               
-
 
        queue_chain_t           cached_list;    /* Attachment point for the
                                                 * list of objects cached as a
@@ -264,13 +339,15 @@ struct vm_object {
        queue_head_t            msr_q;          /* memory object synchronise
                                                   request queue */
 
+  /*
+   * the following fields are not protected by any locks
+   * they are updated via atomic compare and swap
+   */
        vm_object_offset_t      last_alloc;     /* last allocation offset */
-       vm_object_offset_t      sequential;     /* sequential access size */
-       vm_size_t               cluster_size;   /* size of paging cluster */
-#if    MACH_PAGEMAP
-       vm_external_map_t       existence_map;  /* bitmap of pages written to
-                                                * backing storage */
-#endif /* MACH_PAGEMAP */
+       int                     sequential;     /* sequential access size */
+
+        uint32_t               pages_created;
+        uint32_t               pages_used;
        vm_offset_t             cow_hint;       /* last page present in     */
                                                /* shadow but not in object */
 #if    MACH_ASSERT
@@ -279,20 +356,78 @@ struct vm_object {
                                                 * put in current object
                                                 */
 #endif
-                                       /* hold object lock when altering */
-       unsigned        int                     /* cache WIMG bits         */           
-                       wimg_bits:8,            /* wimg plus some expansion*/
-                       not_in_use:24;
-#ifdef UBC_DEBUG
+       /* hold object lock when altering */
+       unsigned        int
+               wimg_bits:8,            /* cache WIMG bits         */           
+               code_signed:1,          /* pages are signed and should be
+                                          validated; the signatures are stored
+                                          with the pager */
+               hashed:1,               /* object/pager entered in hash */
+               transposed:1,           /* object was transposed with another */
+               mapping_in_progress:1,  /* pager being mapped/unmapped */
+               phantom_isssd:1,
+               volatile_empty:1,
+               volatile_fault:1,
+               all_reusable:1,
+               blocked_access:1,
+               set_cache_attr:1,
+               object_slid:1,
+               purgeable_queue_type:2,
+               purgeable_queue_group:3,
+               io_tracking:1,
+#if CONFIG_SECLUDED_MEMORY
+               eligible_for_secluded:1,
+               can_grab_secluded:1,
+#else /* CONFIG_SECLUDED_MEMORY */
+               __object3_unused_bits:2,
+#endif /* CONFIG_SECLUDED_MEMORY */
+               __object2_unused_bits:5;        /* for expansion */
+
+       uint8_t                 scan_collisions;
+        vm_tag_t               wire_tag;
+       uint8_t                 __object4_unused_bits[2];
+
+#if CONFIG_PHANTOM_CACHE
+       uint32_t                phantom_object_id;
+#endif
+#if CONFIG_IOSCHED || UPL_DEBUG
        queue_head_t            uplq;           /* List of outstanding upls */
-#endif /* UBC_DEBUG */
+#endif
+
+#ifdef VM_PIP_DEBUG
+/*
+ * Keep track of the stack traces for the first holders
+ * of a "paging_in_progress" reference for this VM object.
+ */
+#define VM_PIP_DEBUG_STACK_FRAMES      25      /* depth of each stack trace */
+#define VM_PIP_DEBUG_MAX_REFS          10      /* track that many references */
+       struct __pip_backtrace {
+               void *pip_retaddr[VM_PIP_DEBUG_STACK_FRAMES];
+       } pip_holders[VM_PIP_DEBUG_MAX_REFS];
+#endif /* VM_PIP_DEBUG  */
+
+        queue_chain_t          objq;      /* object queue - currently used for purgable queues */
+
+#if DEBUG
+       void *purgeable_owner_bt[16];
+       task_t vo_purgeable_volatilizer; /* who made it volatile? */
+       void *purgeable_volatilizer_bt[16];
+#endif /* DEBUG */
 };
 
-__private_extern__
+#define VM_OBJECT_PURGEABLE_FAULT_ERROR(object)                                \
+       ((object)->volatile_fault &&                                    \
+        ((object)->purgable == VM_PURGABLE_VOLATILE ||                 \
+         (object)->purgable == VM_PURGABLE_EMPTY))
+
+extern
 vm_object_t    kernel_object;          /* the single kernel object */
 
-__private_extern__
-int            vm_object_absent_max;   /* maximum number of absent pages
+extern
+vm_object_t    compressor_object;      /* the single compressor object */
+
+extern
+unsigned int   vm_object_absent_max;   /* maximum number of absent pages
                                           at a time for each object */
 
 # define       VM_MSYNC_INITIALIZED                    0
@@ -306,27 +441,116 @@ struct msync_req {
        vm_object_offset_t      offset;
        vm_object_size_t        length;
        vm_object_t             object;         /* back pointer */
-       decl_mutex_data(,       msync_req_lock) /* Lock for this structure */
+       decl_lck_mtx_data(,     msync_req_lock) /* Lock for this structure */
 };
 
 typedef struct msync_req       *msync_req_t;
 #define MSYNC_REQ_NULL         ((msync_req_t) 0)
 
+
+extern lck_grp_t               vm_map_lck_grp;
+extern lck_attr_t              vm_map_lck_attr;
+
 /*
  * Macros to allocate and free msync_reqs
  */
 #define msync_req_alloc(msr)                                           \
-       MACRO_BEGIN                                                     \
+    MACRO_BEGIN                                                        \
         (msr) = (msync_req_t)kalloc(sizeof(struct msync_req));         \
-        mutex_init(&(msr)->msync_req_lock, ETAP_VM_MSYNC);             \
-       msr->flag = VM_MSYNC_INITIALIZED;                               \
-        MACRO_END
+        lck_mtx_init(&(msr)->msync_req_lock, &vm_map_lck_grp, &vm_map_lck_attr);               \
+        msr->flag = VM_MSYNC_INITIALIZED;                              \
+    MACRO_END
 
 #define msync_req_free(msr)                                            \
-       (kfree((vm_offset_t)(msr), sizeof(struct msync_req)))
+    MACRO_BEGIN                                                                \
+        lck_mtx_destroy(&(msr)->msync_req_lock, &vm_map_lck_grp);      \
+       kfree((msr), sizeof(struct msync_req));                         \
+    MACRO_END
+
+#define msr_lock(msr)   lck_mtx_lock(&(msr)->msync_req_lock)
+#define msr_unlock(msr) lck_mtx_unlock(&(msr)->msync_req_lock)
+
+#define VM_OBJECT_WIRED(object)                                                \
+    MACRO_BEGIN                                                                \
+    if ((object)->purgable == VM_PURGABLE_DENY)                                \
+    {                                                                  \
+       lck_spin_lock(&vm_objects_wired_lock);                          \
+       assert(!(object)->objq.next);                                   \
+       assert(!(object)->objq.prev);                                   \
+       queue_enter(&vm_objects_wired, (object), vm_object_t, objq);    \
+       lck_spin_unlock(&vm_objects_wired_lock);                        \
+    }                                                                  \
+    MACRO_END
+
+#define VM_OBJECT_UNWIRED(object)                                       \
+    MACRO_BEGIN                                                                 \
+    (object)->wire_tag = VM_KERN_MEMORY_NONE;                           \
+    if (((object)->purgable == VM_PURGABLE_DENY) && (object)->objq.next) \
+    {                                                                   \
+       lck_spin_lock(&vm_objects_wired_lock);                           \
+       queue_remove(&vm_objects_wired, (object), vm_object_t, objq);    \
+       lck_spin_unlock(&vm_objects_wired_lock);                         \
+    }                                                                   \
+    MACRO_END
+
+
+#define OBJECT_LOCK_SHARED     0
+#define OBJECT_LOCK_EXCLUSIVE  1
+
+extern lck_grp_t       vm_object_lck_grp;
+extern lck_grp_attr_t  vm_object_lck_grp_attr;
+extern lck_attr_t      vm_object_lck_attr;
+extern lck_attr_t      kernel_object_lck_attr;
+extern lck_attr_t      compressor_object_lck_attr;
+
+extern vm_object_t     vm_pageout_scan_wants_object;
+
+extern void            vm_object_lock(vm_object_t);
+extern boolean_t       vm_object_lock_try(vm_object_t);
+extern boolean_t       _vm_object_lock_try(vm_object_t);
+extern boolean_t       vm_object_lock_avoid(vm_object_t);
+extern void            vm_object_lock_shared(vm_object_t);
+extern boolean_t       vm_object_lock_try_shared(vm_object_t);
+extern void            vm_object_unlock(vm_object_t);
+extern boolean_t       vm_object_lock_upgrade(vm_object_t);
+
+/*
+ *     Object locking macros
+ */
+
+#define vm_object_lock_init(object)                                    \
+       lck_rw_init(&(object)->Lock, &vm_object_lck_grp,                \
+                   (((object) == kernel_object ||                      \
+                     (object) == vm_submap_object) ?                   \
+                    &kernel_object_lck_attr :                          \
+                    (((object) == compressor_object) ?                 \
+                    &compressor_object_lck_attr :                      \
+                     &vm_object_lck_attr)))
+#define vm_object_lock_destroy(object) lck_rw_destroy(&(object)->Lock, &vm_object_lck_grp)
+
+#define vm_object_lock_try_scan(object)        _vm_object_lock_try(object)
+
+/*
+ * CAUTION: the following vm_object_lock_assert_held*() macros merely
+ * check if anyone is holding the lock, but the holder may not necessarily
+ * be the caller...
+ */
+#if MACH_ASSERT || DEBUG
+#define vm_object_lock_assert_held(object) \
+       lck_rw_assert(&(object)->Lock, LCK_RW_ASSERT_HELD)
+#define vm_object_lock_assert_shared(object) \
+       lck_rw_assert(&(object)->Lock, LCK_RW_ASSERT_SHARED)
+#define vm_object_lock_assert_exclusive(object) \
+       lck_rw_assert(&(object)->Lock, LCK_RW_ASSERT_EXCLUSIVE)
+#define vm_object_lock_assert_notheld(object) \
+       lck_rw_assert(&(object)->Lock, LCK_RW_ASSERT_NOTHELD)
+#else  /* MACH_ASSERT || DEBUG */ 
+#define vm_object_lock_assert_held(object)
+#define vm_object_lock_assert_shared(object)
+#define vm_object_lock_assert_exclusive(object)
+#define vm_object_lock_assert_notheld(object)
+#endif /* MACH_ASSERT || DEBUG */
 
-#define msr_lock(msr)   mutex_lock(&(msr)->msync_req_lock)
-#define msr_unlock(msr) mutex_unlock(&(msr)->msync_req_lock)
 
 /*
  *     Declare procedures that operate on VM objects.
@@ -336,8 +560,14 @@ __private_extern__ void            vm_object_bootstrap(void);
 
 __private_extern__ void                vm_object_init(void);
 
-__private_extern__ vm_object_t vm_object_allocate(
-                                       vm_object_size_t        size);
+__private_extern__ void                vm_object_init_lck_grp(void);
+
+__private_extern__ void                vm_object_reaper_init(void);
+
+__private_extern__ vm_object_t vm_object_allocate(vm_object_size_t size);
+
+__private_extern__ void    _vm_object_allocate(vm_object_size_t size,
+                           vm_object_t object);
 
 #if    TASK_SWAPPER
 
@@ -358,27 +588,39 @@ __private_extern__ void   vm_object_res_deallocate(
 #endif /* TASK_SWAPPER */
 
 #define vm_object_reference_locked(object)             \
-MACRO_BEGIN                                            \
-               vm_object_t RLObject = (object);        \
-               assert((RLObject)->ref_count > 0);      \
-               (RLObject)->ref_count++;                \
-               vm_object_res_reference(RLObject);      \
-MACRO_END
+       MACRO_BEGIN                                     \
+       vm_object_t RLObject = (object);                \
+       vm_object_lock_assert_exclusive(object);        \
+       assert((RLObject)->ref_count > 0);              \
+       (RLObject)->ref_count++;                        \
+       assert((RLObject)->ref_count > 1);              \
+       vm_object_res_reference(RLObject);              \
+       MACRO_END
 
 
-#if    MACH_ASSERT
+#define vm_object_reference_shared(object)                             \
+       MACRO_BEGIN                                                     \
+       vm_object_t RLObject = (object);                                \
+       vm_object_lock_assert_shared(object);                           \
+       assert((RLObject)->ref_count > 0);                              \
+       OSAddAtomic(1, &(RLObject)->ref_count);         \
+       assert((RLObject)->ref_count > 0);                              \
+       /* XXX we would need an atomic version of the following ... */  \
+       vm_object_res_reference(RLObject);                              \
+       MACRO_END
+
 
 __private_extern__ void                vm_object_reference(
                                        vm_object_t     object);
 
-#else  /* MACH_ASSERT */
+#if    !MACH_ASSERT
 
 #define        vm_object_reference(object)                     \
 MACRO_BEGIN                                            \
        vm_object_t RObject = (object);                 \
        if (RObject) {                                  \
-               vm_object_lock(RObject);                \
-               vm_object_reference_locked(RObject);    \
+               vm_object_lock_shared(RObject);         \
+               vm_object_reference_shared(RObject);    \
                vm_object_unlock(RObject);              \
        }                                               \
 MACRO_END
@@ -395,11 +637,20 @@ __private_extern__ kern_return_t vm_object_release_name(
 __private_extern__ void                vm_object_pmap_protect(
                                        vm_object_t             object,
                                        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);
 
+__private_extern__ void                vm_object_pmap_protect_options(
+                                       vm_object_t             object,
+                                       vm_object_offset_t      offset,
+                                       vm_object_size_t        size,
+                                       pmap_t                  pmap,
+                                       vm_map_offset_t         pmap_start,
+                                       vm_prot_t               prot,
+                                       int                     options);
+
 __private_extern__ void                vm_object_page_remove(
                                        vm_object_t             object,
                                        vm_object_offset_t      start,
@@ -409,7 +660,32 @@ __private_extern__ void            vm_object_deactivate_pages(
                                        vm_object_t             object,
                                        vm_object_offset_t      offset,
                                        vm_object_size_t        size,
-                                       boolean_t               kill_page);
+                                       boolean_t               kill_page,
+                                       boolean_t               reusable_page,
+                                       struct pmap             *pmap,
+                                       vm_map_offset_t         pmap_offset);
+
+__private_extern__ void        vm_object_reuse_pages(
+       vm_object_t             object,
+       vm_object_offset_t      start_offset,
+       vm_object_offset_t      end_offset,
+       boolean_t               allow_partial_reuse);
+
+__private_extern__ void                vm_object_purge(
+                                      vm_object_t              object,
+                                      int                      flags);
+
+__private_extern__ kern_return_t vm_object_purgable_control(
+       vm_object_t     object,
+       vm_purgable_t   control,
+       int             *state);
+
+__private_extern__ kern_return_t vm_object_get_page_counts(
+       vm_object_t             object,
+       vm_object_offset_t      offset,
+       vm_object_size_t        size,
+       unsigned int            *resident_page_count,
+       unsigned int            *dirty_page_count);
 
 __private_extern__ boolean_t   vm_object_coalesce(
                                        vm_object_t             prev_object,
@@ -426,7 +702,8 @@ __private_extern__ boolean_t        vm_object_shadow(
 
 __private_extern__ void                vm_object_collapse(
                                        vm_object_t             object,
-                                       vm_object_offset_t      offset);
+                                       vm_object_offset_t      offset,
+                                       boolean_t               can_bypass);
 
 __private_extern__ boolean_t   vm_object_copy_quickly(
                                vm_object_t             *_object,
@@ -447,13 +724,14 @@ __private_extern__ kern_return_t  vm_object_copy_slowly(
                                vm_object_t             src_object,
                                vm_object_offset_t      src_offset,
                                vm_object_size_t        size,
-                               int                     interruptible,
+                               boolean_t               interruptible,
                                vm_object_t             *_result_object);
 
 __private_extern__ vm_object_t vm_object_copy_delayed(
                                vm_object_t             src_object,
                                vm_object_offset_t      src_offset,
-                               vm_object_size_t        size);
+                               vm_object_size_t        size,
+                               boolean_t               src_object_shared);
 
 
 
@@ -464,6 +742,9 @@ __private_extern__ kern_return_t    vm_object_destroy(
 __private_extern__ void                vm_object_pager_create(
                                        vm_object_t     object);
 
+__private_extern__ void                vm_object_compressor_pager_create(
+                                       vm_object_t     object);
+
 __private_extern__ void                vm_object_page_map(
                                vm_object_t     object,
                                vm_object_offset_t      offset,
@@ -475,23 +756,31 @@ __private_extern__ void           vm_object_page_map(
 __private_extern__ kern_return_t vm_object_upl_request(
                                vm_object_t             object, 
                                vm_object_offset_t      offset,
-                               vm_size_t               size,
+                               upl_size_t              size,
                                upl_t                   *upl,
                                upl_page_info_t         *page_info,
                                unsigned int            *count,
-                               int                     flags);
+                               upl_control_flags_t     flags);
+
+__private_extern__ kern_return_t vm_object_transpose(
+                               vm_object_t             object1,
+                               vm_object_t             object2,
+                               vm_object_size_t        transpose_size);
 
 __private_extern__ 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);
 
 __private_extern__ kern_return_t vm_object_update(
                                vm_object_t             object,
                                vm_object_offset_t      offset,
-                               vm_size_t               size, /* should be 64 */
+                               vm_object_size_t        size,
+                               vm_object_offset_t      *error_offset,
+                               int                     *io_errno,
                                memory_object_return_t  should_return,
                                int                     flags,
                                vm_prot_t               prot);
@@ -514,6 +803,79 @@ __private_extern__ vm_object_t     vm_object_enter(
                                        boolean_t               check_named);
 
 
+__private_extern__ void        vm_object_cluster_size(
+                                       vm_object_t             object,
+                                       vm_object_offset_t      *start,
+                                       vm_size_t               *length,
+                                       vm_object_fault_info_t  fault_info,
+                                       uint32_t                *io_streaming);
+
+__private_extern__ kern_return_t vm_object_populate_with_private(
+       vm_object_t             object,
+       vm_object_offset_t      offset,
+       ppnum_t                 phys_page,
+       vm_size_t               size);
+
+__private_extern__ void vm_object_change_wimg_mode(
+       vm_object_t             object,
+       unsigned int            wimg_mode);
+
+extern kern_return_t adjust_vm_object_cache(
+       vm_size_t oval,
+       vm_size_t nval);
+
+extern kern_return_t vm_object_page_op(
+       vm_object_t             object,
+       vm_object_offset_t      offset,
+       int                     ops,
+       ppnum_t                 *phys_entry,
+       int                     *flags);
+
+extern kern_return_t vm_object_range_op(
+       vm_object_t             object,
+       vm_object_offset_t      offset_beg,
+       vm_object_offset_t      offset_end,
+       int                     ops,
+       uint32_t                *range);
+
+
+__private_extern__ void                vm_object_reap_pages(
+                                       vm_object_t object,
+                                       int     reap_type);
+#define REAP_REAP      0
+#define        REAP_TERMINATE  1
+#define REAP_PURGEABLE 2
+#define REAP_DATA_FLUSH        3
+
+#if CONFIG_FREEZE
+
+__private_extern__ void
+vm_object_compressed_freezer_pageout(
+       vm_object_t     object);
+
+__private_extern__ void
+vm_object_compressed_freezer_done(
+       void);
+
+#endif /* CONFIG_FREEZE */
+
+__private_extern__ void
+vm_object_pageout(
+       vm_object_t     object);
+
+#if CONFIG_IOSCHED
+struct io_reprioritize_req {
+       uint64_t        blkno;
+       uint32_t        len;
+       int             priority;
+       struct vnode    *devvp;
+       queue_chain_t   io_reprioritize_list;
+};
+typedef struct io_reprioritize_req *io_reprioritize_req_t;
+
+extern void vm_io_reprioritize_init(void);
+#endif
+
 /*
  *     Event waiting handling
  */
@@ -521,65 +883,195 @@ __private_extern__ vm_object_t   vm_object_enter(
 #define        VM_OBJECT_EVENT_INITIALIZED             0
 #define        VM_OBJECT_EVENT_PAGER_READY             1
 #define        VM_OBJECT_EVENT_PAGING_IN_PROGRESS      2
-#define        VM_OBJECT_EVENT_ABSENT_COUNT            3
+#define        VM_OBJECT_EVENT_MAPPING_IN_PROGRESS     3
 #define        VM_OBJECT_EVENT_LOCK_IN_PROGRESS        4
 #define        VM_OBJECT_EVENT_UNCACHING               5
 #define        VM_OBJECT_EVENT_COPY_CALL               6
 #define        VM_OBJECT_EVENT_CACHING                 7
+#define VM_OBJECT_EVENT_UNBLOCKED              8
+#define VM_OBJECT_EVENT_PAGING_ONLY_IN_PROGRESS        9
+
+#define VM_OBJECT_EVENT_MAX 10 /* 11 bits in "all_wanted", so 0->10 */
+
+static __inline__ wait_result_t
+vm_object_assert_wait(
+       vm_object_t             object,
+       int                     event,
+       wait_interrupt_t        interruptible)
+{
+       wait_result_t wr;
+
+       vm_object_lock_assert_exclusive(object);
+       assert(event >= 0 && event <= VM_OBJECT_EVENT_MAX);
+
+       object->all_wanted |= 1 << event;
+       wr = assert_wait((event_t)((vm_offset_t)object + event),
+                        interruptible);
+       return wr;
+}
+
+static __inline__ wait_result_t
+vm_object_wait(
+       vm_object_t             object,
+       int                     event,
+       wait_interrupt_t        interruptible)
+{
+       wait_result_t wr;
+
+       vm_object_assert_wait(object, event, interruptible);
+       vm_object_unlock(object);
+       wr = thread_block(THREAD_CONTINUE_NULL);
+       return wr;
+}
+
+static __inline__ wait_result_t
+thread_sleep_vm_object(
+       vm_object_t             object,
+       event_t                 event,
+       wait_interrupt_t        interruptible)
+{
+       wait_result_t wr;
+
+#if DEVELOPMENT || DEBUG
+       if (object->Lock_owner != current_thread())
+               panic("thread_sleep_vm_object: now owner - %p\n", object);
+       object->Lock_owner = 0;
+#endif
+        wr = lck_rw_sleep(&object->Lock,
+                         LCK_SLEEP_PROMOTED_PRI,
+                         event,
+                         interruptible);
+#if DEVELOPMENT || DEBUG
+       object->Lock_owner = current_thread();
+#endif
+       return wr;
+}
+
+static __inline__ wait_result_t
+vm_object_sleep(
+       vm_object_t             object,
+       int                     event,
+       wait_interrupt_t        interruptible)
+{
+       wait_result_t wr;
+
+       vm_object_lock_assert_exclusive(object);
+       assert(event >= 0 && event <= VM_OBJECT_EVENT_MAX);
+
+       object->all_wanted |= 1 << event;
+       wr = thread_sleep_vm_object(object,
+                                   (event_t)((vm_offset_t)object + event),
+                                   interruptible);
+       return wr;
+}
+
+static __inline__ void
+vm_object_wakeup(
+       vm_object_t             object,
+       int                     event)
+{
+       vm_object_lock_assert_exclusive(object);
+       assert(event >= 0 && event <= VM_OBJECT_EVENT_MAX);
+
+       if (object->all_wanted & (1 << event))
+               thread_wakeup((event_t)((vm_offset_t)object + event));
+       object->all_wanted &= ~(1 << event);
+}
+
+static __inline__ void
+vm_object_set_wanted(
+       vm_object_t             object,
+       int                     event)
+{
+       vm_object_lock_assert_exclusive(object);
+       assert(event >= 0 && event <= VM_OBJECT_EVENT_MAX);
+
+       object->all_wanted |= (1 << event);
+}
+
+static __inline__ int
+vm_object_wanted(
+       vm_object_t             object,
+       int                     event)
+{
+       vm_object_lock_assert_held(object);
+       assert(event >= 0 && event <= VM_OBJECT_EVENT_MAX);
+
+       return object->all_wanted & (1 << event);
+}
 
-#define        vm_object_assert_wait(object, event, interruptible)             \
-       (((object)->all_wanted |= 1 << (event)),                        \
-        assert_wait((event_t)((vm_offset_t)(object)+(event)),(interruptible)))
-
-#define        vm_object_wait(object, event, interruptible)                    \
-       (vm_object_assert_wait((object),(event),(interruptible)),       \
-       vm_object_unlock(object),                                       \
-       thread_block(THREAD_CONTINUE_NULL))                             \
-
-#define thread_sleep_vm_object(object, event, interruptible)           \
-       thread_sleep_mutex((event_t)(event), &(object)->Lock, (interruptible))
-
-#define vm_object_sleep(object, event, interruptible)                  \
-       (((object)->all_wanted |= 1 << (event)),                        \
-        thread_sleep_vm_object((object),                               \
-               ((vm_offset_t)(object)+(event)), (interruptible)))
-
-#define        vm_object_wakeup(object, event)                                 \
+/*
+ *     Routines implemented as macros
+ */
+#ifdef VM_PIP_DEBUG
+#include <libkern/OSDebug.h>
+#define VM_PIP_DEBUG_BEGIN(object)                                     \
        MACRO_BEGIN                                                     \
-       if ((object)->all_wanted & (1 << (event)))                      \
-               thread_wakeup((event_t)((vm_offset_t)(object) + (event))); \
-       (object)->all_wanted &= ~(1 << (event));                        \
+       int pip = ((object)->paging_in_progress +                       \
+                  (object)->activity_in_progress);                     \
+       if (pip < VM_PIP_DEBUG_MAX_REFS) {                              \
+               (void) OSBacktrace(&(object)->pip_holders[pip].pip_retaddr[0], \
+                                  VM_PIP_DEBUG_STACK_FRAMES);          \
+       }                                                               \
        MACRO_END
+#else  /* VM_PIP_DEBUG */
+#define VM_PIP_DEBUG_BEGIN(object)
+#endif /* VM_PIP_DEBUG */
 
-#define        vm_object_set_wanted(object, event)                             \
+#define                vm_object_activity_begin(object)                        \
        MACRO_BEGIN                                                     \
-       ((object)->all_wanted |= (1 << (event)));                       \
+       vm_object_lock_assert_exclusive((object));                      \
+       VM_PIP_DEBUG_BEGIN((object));                                   \
+       (object)->activity_in_progress++;                               \
+       if ((object)->activity_in_progress == 0) {                      \
+               panic("vm_object_activity_begin(%p): overflow\n", (object));\
+       }                                                               \
        MACRO_END
 
-#define        vm_object_wanted(object, event)                                 \
-       ((object)->all_wanted & (1 << (event)))
-
-/*
- *     Routines implemented as macros
- */
+#define                vm_object_activity_end(object)                          \
+       MACRO_BEGIN                                                     \
+       vm_object_lock_assert_exclusive((object));                      \
+       if ((object)->activity_in_progress == 0) {                      \
+               panic("vm_object_activity_end(%p): underflow\n", (object));\
+       }                                                               \
+       (object)->activity_in_progress--;                               \
+       if ((object)->paging_in_progress == 0 &&                        \
+           (object)->activity_in_progress == 0)                        \
+               vm_object_wakeup((object),                              \
+                                VM_OBJECT_EVENT_PAGING_IN_PROGRESS);   \
+       MACRO_END
 
-#define                vm_object_paging_begin(object)                          \
+#define                vm_object_paging_begin(object)                          \
        MACRO_BEGIN                                                     \
+       vm_object_lock_assert_exclusive((object));                      \
+       VM_PIP_DEBUG_BEGIN((object));                                   \
        (object)->paging_in_progress++;                                 \
+       if ((object)->paging_in_progress == 0) {                        \
+               panic("vm_object_paging_begin(%p): overflow\n", (object));\
+       }                                                               \
        MACRO_END
 
-#define                vm_object_paging_end(object)                            \
+#define                vm_object_paging_end(object)                            \
        MACRO_BEGIN                                                     \
-       assert((object)->paging_in_progress != 0);                      \
-       if (--(object)->paging_in_progress == 0) {                      \
-               vm_object_wakeup(object,                                \
-                       VM_OBJECT_EVENT_PAGING_IN_PROGRESS);            \
+       vm_object_lock_assert_exclusive((object));                      \
+       if ((object)->paging_in_progress == 0) {                        \
+               panic("vm_object_paging_end(%p): underflow\n", (object));\
+       }                                                               \
+       (object)->paging_in_progress--;                                 \
+       if ((object)->paging_in_progress == 0) {                        \
+               vm_object_wakeup((object),                              \
+                                VM_OBJECT_EVENT_PAGING_ONLY_IN_PROGRESS); \
+               if ((object)->activity_in_progress == 0)                \
+                       vm_object_wakeup((object),                      \
+                                        VM_OBJECT_EVENT_PAGING_IN_PROGRESS); \
        }                                                               \
        MACRO_END
 
 #define                vm_object_paging_wait(object, interruptible)            \
        MACRO_BEGIN                                                     \
-       while ((object)->paging_in_progress != 0) {                     \
+       vm_object_lock_assert_exclusive((object));                      \
+       while ((object)->paging_in_progress != 0 ||                     \
+              (object)->activity_in_progress != 0) {                   \
                wait_result_t  _wr;                                     \
                                                                        \
                _wr = vm_object_sleep((object),                         \
@@ -591,28 +1083,60 @@ __private_extern__ vm_object_t   vm_object_enter(
        }                                                               \
        MACRO_END
 
-#define        vm_object_absent_assert_wait(object, interruptible)             \
+#define vm_object_paging_only_wait(object, interruptible)              \
        MACRO_BEGIN                                                     \
-       vm_object_assert_wait(  (object),                               \
-                       VM_OBJECT_EVENT_ABSENT_COUNT,                   \
-                       (interruptible));                               \
+       vm_object_lock_assert_exclusive((object));                      \
+       while ((object)->paging_in_progress != 0) {                     \
+               wait_result_t  _wr;                                     \
+                                                                       \
+               _wr = vm_object_sleep((object),                         \
+                               VM_OBJECT_EVENT_PAGING_ONLY_IN_PROGRESS,\
+                               (interruptible));                       \
+                                                                       \
+               /*XXX if ((interruptible) && (_wr != THREAD_AWAKENED))*/\
+                       /*XXX break; */                                 \
+       }                                                               \
        MACRO_END
 
 
-#define        vm_object_absent_release(object)                                \
+#define vm_object_mapping_begin(object)                                \
+       MACRO_BEGIN                                                     \
+       vm_object_lock_assert_exclusive((object));                      \
+       assert(! (object)->mapping_in_progress);                        \
+       (object)->mapping_in_progress = TRUE;                           \
+       MACRO_END
+
+#define vm_object_mapping_end(object)                                  \
        MACRO_BEGIN                                                     \
-       (object)->absent_count--;                                       \
+       vm_object_lock_assert_exclusive((object));                      \
+       assert((object)->mapping_in_progress);                          \
+       (object)->mapping_in_progress = FALSE;                          \
        vm_object_wakeup((object),                                      \
-                        VM_OBJECT_EVENT_ABSENT_COUNT);                 \
+                        VM_OBJECT_EVENT_MAPPING_IN_PROGRESS);          \
        MACRO_END
 
-/*
- *     Object locking macros
- */
+#define vm_object_mapping_wait(object, interruptible)                  \
+       MACRO_BEGIN                                                     \
+       vm_object_lock_assert_exclusive((object));                      \
+       while ((object)->mapping_in_progress) {                         \
+               wait_result_t   _wr;                                    \
+                                                                       \
+               _wr = vm_object_sleep((object),                         \
+                                     VM_OBJECT_EVENT_MAPPING_IN_PROGRESS, \
+                                     (interruptible));                 \
+               /*XXX if ((interruptible) && (_wr != THREAD_AWAKENED))*/\
+                       /*XXX break; */                                 \
+       }                                                               \
+       assert(!(object)->mapping_in_progress);                         \
+       MACRO_END
+
+
+
+#define vm_object_round_page(x) (((vm_object_offset_t)(x) + PAGE_MASK) & ~((signed)PAGE_MASK))
+#define vm_object_trunc_page(x) ((vm_object_offset_t)(x) & ~((signed)PAGE_MASK))
 
-#define vm_object_lock_init(object)    mutex_init(&(object)->Lock, ETAP_VM_OBJ)
-#define vm_object_lock(object)         mutex_lock(&(object)->Lock)
-#define vm_object_unlock(object)       mutex_unlock(&(object)->Lock)
-#define vm_object_lock_try(object)     mutex_try(&(object)->Lock)
+extern void    vm_object_cache_add(vm_object_t);
+extern void    vm_object_cache_remove(vm_object_t);
+extern int     vm_object_cache_evict(int, int);
 
 #endif /* _VM_VM_OBJECT_H_ */