]> git.saurik.com Git - apple/xnu.git/blobdiff - osfmk/vm/vm_object.h
xnu-4903.241.1.tar.gz
[apple/xnu.git] / osfmk / vm / vm_object.h
index a093bf2f4a313038e37bd1b66227741b855c2878..ccd1de547f36b4e377f6c35f151d451a0f9ed1c1 100644 (file)
@@ -66,6 +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>
 
@@ -77,7 +79,6 @@
 #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/misc_protos.h>
 #include <ipc/ipc_types.h>
 #include <vm/pmap.h>
 
-#if    MACH_PAGEMAP
 #include <vm/vm_external.h>
-#endif /* MACH_PAGEMAP */
+
+#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;
 
@@ -105,25 +120,59 @@ struct vm_object_fault_info {
         vm_behavior_t  behavior;
         vm_map_offset_t        lo_offset;
        vm_map_offset_t hi_offset;
-       boolean_t       no_cache;
+       unsigned int
+       /* boolean_t */ no_cache:1,
+       /* boolean_t */ stealth:1,
+       /* boolean_t */ io_sync:1,
+       /* boolean_t */ cs_bypass:1,
+       /* boolean_t */ pmap_cs_associated:1,
+       /* boolean_t */ mark_zf_absent:1,
+       /* boolean_t */ batch_pmap_op:1,
+               __vm_object_fault_info_unused_bits:25;
+       int             pmap_options;
 };
 
 
+#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_owner                       vo_un2.vou_owner
 
 struct vm_object {
-       queue_head_t            memq;           /* Resident memory */
+       /*
+        * 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 */
 
-       vm_object_size_t        size;           /* Object size (only valid
-                                                * if internal)
-                                                */
+#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
+                                                    use VM_OBJECT_WIRED_PAGE_UPDATE macros to update */
+       unsigned int            reusable_page_count;
 
        struct vm_object        *copy;          /* Object that should receive
                                                 * a copy of my changed pages,
@@ -133,7 +182,17 @@ 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_owner;      /* If the object is purgeable
+                                                * or has a "ledger_tag", this
+                                                * is the task that owns it.
+                                                */
+       } vo_un2;
 
        memory_object_t         pager;          /* Where to get data */
        vm_object_offset_t      paging_offset;  /* Offset into memory object */
@@ -142,7 +201,27 @@ struct vm_object {
        memory_object_copy_strategy_t
                                copy_strategy;  /* How to handle data copy */
 
-       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
@@ -150,6 +229,7 @@ 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
@@ -174,15 +254,6 @@ struct vm_object {
                                                 * therefore, managed by the
                                                 * default memory manger)
                                                 */
-       /* boolean_t */         temporary:1,    /* Permanent objects may be
-                                                * changed externally by the 
-                                                * memory manager, and changes
-                                                * made in memory must be
-                                                * reflected back to the memory
-                                                * manager.  Temporary objects
-                                                * lack both of these
-                                                * characteristics.
-                                                */
        /* boolean_t */         private:1,      /* magic device_pager object,
                                                 * holds private pages only */
        /* boolean_t */         pageout:1,      /* pageout object. contains
@@ -193,17 +264,11 @@ struct vm_object {
        /* boolean_t */         purgable:2,     /* Purgable state.  See
                                                 * VM_PURGABLE_* 
                                                 */
+       /* boolean_t */         purgeable_only_by_kernel:1,
+       /* 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
-                                                * pager that the kernel
-                                                * wants to discard it, page
-                                                * remains in object */
        /* boolean_t */         true_share:1,
                                                /* This object is mapped
                                                 * in more than one place
@@ -243,24 +308,19 @@ struct vm_object {
                                                 * memory rules w.r.t pmap
                                                 * access bits.
                                                 */
-       /* boolean_t */         nophyscache:1;
+       /* boolean_t */         nophyscache:1,
                                                /* When mapped at the 
                                                 * pmap level, don't allow
                                                 * primary caching. (for
                                                 * I/O)
                                                 */
-                                               
-
+       /* boolean_t */         _object5_unused_bits:1;
 
        queue_chain_t           cached_list;    /* Attachment point for the
                                                 * list of objects cached as a
                                                 * result of their can_persist
                                                 * value
                                                 */
-
-       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
@@ -270,29 +330,56 @@ struct vm_object {
 
         uint32_t               pages_created;
         uint32_t               pages_used;
-#if    MACH_PAGEMAP
-       vm_external_map_t       existence_map;  /* bitmap of pages written to
-                                                * backing storage */
-#endif /* MACH_PAGEMAP */
        vm_offset_t             cow_hint;       /* last page present in     */
                                                /* shadow but not in object */
-#if    MACH_ASSERT
-       struct vm_object        *paging_object; /* object which pages to be
-                                                * swapped out are temporary
-                                                * put in current object
-                                                */
-#endif
        /* 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 */
-               not_in_use:23;          /* for expansion */
-
-#ifdef UPL_DEBUG
+               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_is_shared_cache:1,
+               purgeable_queue_type:2,
+               purgeable_queue_group:3,
+               io_tracking:1,
+               no_tag_update: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 */
+#if VM_OBJECT_ACCESS_TRACKING
+               access_tracking:1,
+#else /* VM_OBJECT_ACCESS_TRACKING */
+               __unused_access_tracking:1,
+#endif /* VM_OBJECT_ACCESS_TRACKING */
+               vo_ledger_tag:2,
+               __object2_unused_bits:2;        /* for expansion */
+
+#if VM_OBJECT_ACCESS_TRACKING
+       uint32_t        access_tracking_reads;
+       uint32_t        access_tracking_writes;
+#endif /* VM_OBJECT_ACCESS_TRACKING */
+
+       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 /* UPL_DEBUG */
+#endif
 
 #ifdef VM_PIP_DEBUG
 /*
@@ -306,41 +393,47 @@ struct vm_object {
        } pip_holders[VM_PIP_DEBUG_MAX_REFS];
 #endif /* VM_PIP_DEBUG  */
 
-        queue_chain_t       objq;      /* object queue - currently used for purgable queues */
-};
+        queue_chain_t          objq;      /* object queue - currently used for purgable queues */
+       queue_chain_t           task_objq; /* objects owned by task - protected by task lock */
 
-#define VM_PAGE_REMOVE(page)                                           \
-       MACRO_BEGIN                                                     \
-       vm_page_t __page = (page);                                      \
-       vm_object_t __object = __page->object;                          \
-       if (__page == __object->memq_hint) {                            \
-               vm_page_t       __new_hint;                             \
-               queue_entry_t   __qe;                                   \
-               __qe = queue_next(&__page->listq);                      \
-               if (queue_end(&__object->memq, __qe)) {                 \
-                       __qe = queue_prev(&__page->listq);              \
-                       if (queue_end(&__object->memq, __qe)) {         \
-                               __qe = NULL;                            \
-                       }                                               \
-               }                                                       \
-               __new_hint = (vm_page_t) __qe;                          \
-               __object->memq_hint = __new_hint;                       \
-       }                                                               \
-       queue_remove(&__object->memq, __page, vm_page_t, listq);        \
-       MACRO_END
+#if !VM_TAG_ACTIVE_UPDATE
+       queue_chain_t           wired_objq;
+#endif /* !VM_TAG_ACTIVE_UPDATE */
 
-#define VM_PAGE_INSERT(page, object)                           \
-       MACRO_BEGIN                                             \
-       vm_page_t __page = (page);                              \
-       vm_object_t __object = (object);                        \
-       queue_enter(&__object->memq, __page, vm_page_t, listq); \
-       __object->memq_hint = __page;                           \
-       MACRO_END
+#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__
+/* values for object->vo_ledger_tag */
+#define VM_OBJECT_LEDGER_TAG_NONE      0
+#define VM_OBJECT_LEDGER_TAG_NETWORK   1
+#define VM_OBJECT_LEDGER_TAG_MEDIA     2
+#define VM_OBJECT_LEDGER_TAG_RESERVED  3
+
+#define VM_OBJECT_PURGEABLE_FAULT_ERROR(object)                                \
+       ((object)->volatile_fault &&                                    \
+        ((object)->purgable == VM_PURGABLE_VOLATILE ||                 \
+         (object)->purgable == VM_PURGABLE_EMPTY))
+
+#if VM_OBJECT_ACCESS_TRACKING
+extern uint64_t vm_object_access_tracking_reads;
+extern uint64_t vm_object_access_tracking_writes;
+extern void vm_object_access_tracking(vm_object_t object,
+                                     int *access_tracking,
+                                     uint32_t *access_tracking_reads,
+                                     uint32_t *acess_tracking_writes);
+#endif /* VM_OBJECT_ACCESS_TRACKING */
+
+extern
 vm_object_t    kernel_object;          /* the single kernel object */
 
-__private_extern__
+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 */
 
@@ -348,40 +441,165 @@ unsigned int     vm_object_absent_max;   /* maximum number of absent pages
 # define       VM_MSYNC_SYNCHRONIZING                  1
 # define       VM_MSYNC_DONE                           2
 
-struct msync_req {
-       queue_chain_t           msr_q;          /* object request queue */
-       queue_chain_t           req_q;          /* vm_msync request queue */
-       unsigned int            flag;
-       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 */
-};
 
-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;
+
+#ifndef VM_TAG_ACTIVE_UPDATE
+#error VM_TAG_ACTIVE_UPDATE
+#endif
+
+#if VM_TAG_ACTIVE_UPDATE
+#define VM_OBJECT_WIRED_ENQUEUE(object) panic("VM_OBJECT_WIRED_ENQUEUE")
+#define VM_OBJECT_WIRED_DEQUEUE(object) panic("VM_OBJECT_WIRED_DEQUEUE")
+#else /* VM_TAG_ACTIVE_UPDATE */
+#define VM_OBJECT_WIRED_ENQUEUE(object)                                        \
+       MACRO_BEGIN                                                     \
+       lck_spin_lock(&vm_objects_wired_lock);                          \
+       assert(!(object)->wired_objq.next);                             \
+       assert(!(object)->wired_objq.prev);                             \
+       queue_enter(&vm_objects_wired, (object),                        \
+                   vm_object_t, wired_objq);                           \
+       lck_spin_unlock(&vm_objects_wired_lock);                        \
+       MACRO_END
+#define VM_OBJECT_WIRED_DEQUEUE(object)                                        \
+       MACRO_BEGIN                                                     \
+       if ((object)->wired_objq.next) {                                \
+               lck_spin_lock(&vm_objects_wired_lock);                  \
+               queue_remove(&vm_objects_wired, (object),               \
+                            vm_object_t, wired_objq);                  \
+               lck_spin_unlock(&vm_objects_wired_lock);                \
+       }                                                               \
+       MACRO_END
+#endif /* VM_TAG_ACTIVE_UPDATE */
+
+#define VM_OBJECT_WIRED(object, tag)                                   \
+    MACRO_BEGIN                                                                \
+    assert(VM_KERN_MEMORY_NONE != (tag));                              \
+    assert(VM_KERN_MEMORY_NONE == (object)->wire_tag);                 \
+    (object)->wire_tag = (tag);                                        \
+    if (!VM_TAG_ACTIVE_UPDATE) {                                       \
+       VM_OBJECT_WIRED_ENQUEUE((object));                              \
+    }                                                                  \
+    MACRO_END
+
+#define VM_OBJECT_UNWIRED(object)                                                      \
+    MACRO_BEGIN                                                                                \
+    if (!VM_TAG_ACTIVE_UPDATE) {                                                       \
+           VM_OBJECT_WIRED_DEQUEUE((object));                                          \
+    }                                                                                  \
+    if (VM_KERN_MEMORY_NONE != (object)->wire_tag) {                                   \
+       vm_tag_update_size((object)->wire_tag, -ptoa_64((object)->wired_page_count));   \
+       (object)->wire_tag = VM_KERN_MEMORY_NONE;                                       \
+    }                                                                                  \
+    MACRO_END
+
+// These two macros start & end a C block
+#define VM_OBJECT_WIRED_PAGE_UPDATE_START(object)                                      \
+    MACRO_BEGIN                                                                                \
+    {                                                                                  \
+       int64_t __wireddelta = 0; vm_tag_t __waswired = (object)->wire_tag;
+
+#define VM_OBJECT_WIRED_PAGE_UPDATE_END(object, tag)                                   \
+       if (__wireddelta) {                                                             \
+           boolean_t __overflow __assert_only =                                        \
+           os_add_overflow((object)->wired_page_count, __wireddelta,                   \
+                           &(object)->wired_page_count);                               \
+           assert(!__overflow);                                                        \
+           if (!(object)->pageout && !(object)->no_tag_update) {                       \
+               if (__wireddelta > 0) {                                                 \
+                   assert (VM_KERN_MEMORY_NONE != (tag));                              \
+                   if (VM_KERN_MEMORY_NONE == __waswired) {                            \
+                       VM_OBJECT_WIRED((object), (tag));                               \
+                   }                                                                   \
+                   vm_tag_update_size((object)->wire_tag, ptoa_64(__wireddelta));      \
+               } else if (VM_KERN_MEMORY_NONE != __waswired) {                         \
+                   assert (VM_KERN_MEMORY_NONE != (object)->wire_tag);                 \
+                   vm_tag_update_size((object)->wire_tag, ptoa_64(__wireddelta));      \
+                   if (!(object)->wired_page_count) {                                  \
+                       VM_OBJECT_UNWIRED((object));                                    \
+                   }                                                                   \
+               }                                                                       \
+           }                                                                           \
+       }                                                                               \
+    }                                                                                  \
+    MACRO_END
+
+#define VM_OBJECT_WIRED_PAGE_COUNT(object, delta)               \
+    __wireddelta += delta; \
+
+#define VM_OBJECT_WIRED_PAGE_ADD(object, m)                     \
+    if (!(m)->vmp_private && !(m)->vmp_fictitious) __wireddelta++;
+
+#define VM_OBJECT_WIRED_PAGE_REMOVE(object, m)                  \
+    if (!(m)->vmp_private && !(m)->vmp_fictitious) __wireddelta--;
+
+
+
+#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_yield_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);
 
 /*
- * Macros to allocate and free msync_reqs
+ *     Object locking macros
  */
-#define msync_req_alloc(msr)                                           \
-       MACRO_BEGIN                                                     \
-        (msr) = (msync_req_t)kalloc(sizeof(struct msync_req));         \
-        mutex_init(&(msr)->msync_req_lock, 0);         \
-       msr->flag = VM_MSYNC_INITIALIZED;                               \
-        MACRO_END
 
-#define msync_req_free(msr)                                            \
-       (kfree((msr), sizeof(struct msync_req)))
+#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.
  */
 
-__private_extern__ void                vm_object_bootstrap(void) __attribute__((section("__TEXT, initcode")));
+__private_extern__ void                vm_object_bootstrap(void);
 
 __private_extern__ void                vm_object_init(void);
 
@@ -389,8 +607,7 @@ __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__ 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);
@@ -429,8 +646,8 @@ __private_extern__ void     vm_object_res_deallocate(
        vm_object_t RLObject = (object);                                \
        vm_object_lock_assert_shared(object);                           \
        assert((RLObject)->ref_count > 0);                              \
-       OSAddAtomic(1, (SInt32 *)&(RLObject)->ref_count);               \
-       assert((RLObject)->ref_count > 1);                              \
+       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
@@ -445,8 +662,8 @@ __private_extern__ void             vm_object_reference(
 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
@@ -468,6 +685,15 @@ __private_extern__ void            vm_object_pmap_protect(
                                        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,
@@ -477,16 +703,33 @@ __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__ unsigned int        vm_object_purge(
-                                       vm_object_t             object);
+__private_extern__ uint64_t    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,
                                        vm_object_t             next_object,
@@ -524,7 +767,7 @@ __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(
@@ -542,6 +785,9 @@ __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,
@@ -557,7 +803,8 @@ __private_extern__ kern_return_t vm_object_upl_request(
                                upl_t                   *upl,
                                upl_page_info_t         *page_info,
                                unsigned int            *count,
-                               int                     flags);
+                               upl_control_flags_t     flags,
+                               vm_tag_t            tag);
 
 __private_extern__ kern_return_t vm_object_transpose(
                                vm_object_t             object1,
@@ -592,11 +839,10 @@ __private_extern__ kern_return_t vm_object_lock_request(
 
 
 
-__private_extern__ vm_object_t vm_object_enter(
+__private_extern__ vm_object_t vm_object_memory_object_associate(
                                        memory_object_t         pager,
+                                       vm_object_t             object,
                                        vm_object_size_t        size,
-                                       boolean_t               internal,
-                                       boolean_t               init,
                                        boolean_t               check_named);
 
 
@@ -604,7 +850,8 @@ __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);
+                                       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,
@@ -612,6 +859,10 @@ __private_extern__ kern_return_t vm_object_populate_with_private(
        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);
@@ -628,7 +879,45 @@ extern kern_return_t vm_object_range_op(
        vm_object_offset_t      offset_beg,
        vm_object_offset_t      offset_end,
        int                     ops,
-       int                     *range);
+       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
@@ -637,42 +926,122 @@ extern kern_return_t vm_object_range_op(
 #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_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_assert_wait(object, event, interruptible)             \
-       (((object)->all_wanted |= 1 << (event)),                        \
-        assert_wait((event_t)((vm_offset_t)(object)+(event)),(interruptible)))
+#define VM_OBJECT_EVENT_MAX 10 /* 11 bits in "all_wanted", so 0->10 */
 
-#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)           \
-        lck_rw_sleep(&(object)->Lock, LCK_SLEEP_DEFAULT, (event_t)(event), (interruptible))
+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;
+}
 
-#define vm_object_sleep(object, event, interruptible)                  \
-       (((object)->all_wanted |= 1 << (event)),                        \
-        thread_sleep_vm_object((object),                               \
-               ((vm_offset_t)(object)+(event)), (interruptible)))
+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);
 
-#define        vm_object_wakeup(object, event)                                 \
-       MACRO_BEGIN                                                     \
-       if ((object)->all_wanted & (1 << (event)))                      \
-               thread_wakeup((event_t)((vm_offset_t)(object) + (event))); \
-       (object)->all_wanted &= ~(1 << (event));                        \
-       MACRO_END
+       object->all_wanted |= (1 << event);
+}
 
-#define        vm_object_set_wanted(object, event)                             \
-       MACRO_BEGIN                                                     \
-       ((object)->all_wanted |= (1 << (event)));                       \
-       MACRO_END
+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);
 
-#define        vm_object_wanted(object, event)                                 \
-       ((object)->all_wanted & (1 << (event)))
+       return object->all_wanted & (1 << event);
+}
 
 /*
  *     Routines implemented as macros
@@ -681,8 +1050,9 @@ extern kern_return_t vm_object_range_op(
 #include <libkern/OSDebug.h>
 #define VM_PIP_DEBUG_BEGIN(object)                                     \
        MACRO_BEGIN                                                     \
-       if ((object)->paging_in_progress < VM_PIP_DEBUG_MAX_REFS) {     \
-               int pip = (object)->paging_in_progress;                 \
+       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);          \
        }                                                               \
@@ -691,28 +1061,60 @@ extern kern_return_t vm_object_range_op(
 #define VM_PIP_DEBUG_BEGIN(object)
 #endif /* VM_PIP_DEBUG */
 
-#define                vm_object_paging_begin(object)                          \
+#define                vm_object_activity_begin(object)                        \
+       MACRO_BEGIN                                                     \
+       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_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)                          \
        MACRO_BEGIN                                                     \
        vm_object_lock_assert_exclusive((object));                      \
-       assert((object)->paging_in_progress >= 0);                      \
        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                                                     \
        vm_object_lock_assert_exclusive((object));                      \
-       assert((object)->paging_in_progress > 0);                       \
-       if (--(object)->paging_in_progress == 0) {                      \
-               vm_object_wakeup(object,                                \
-                       VM_OBJECT_EVENT_PAGING_IN_PROGRESS);            \
+       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                                                     \
        vm_object_lock_assert_exclusive((object));                      \
-       while ((object)->paging_in_progress != 0) {                     \
+       while ((object)->paging_in_progress != 0 ||                     \
+              (object)->activity_in_progress != 0) {                   \
                wait_result_t  _wr;                                     \
                                                                        \
                _wr = vm_object_sleep((object),                         \
@@ -724,58 +1126,83 @@ extern kern_return_t vm_object_range_op(
        }                                                               \
        MACRO_END
 
+#define vm_object_paging_only_wait(object, interruptible)              \
+       MACRO_BEGIN                                                     \
+       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 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 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 void            vm_object_lock_shared(vm_object_t);
-extern boolean_t       vm_object_lock_try_shared(vm_object_t);
+#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
 
-/*
- *     Object locking macros
- */
+#define vm_object_mapping_end(object)                                  \
+       MACRO_BEGIN                                                     \
+       vm_object_lock_assert_exclusive((object));                      \
+       assert((object)->mapping_in_progress);                          \
+       (object)->mapping_in_progress = FALSE;                          \
+       vm_object_wakeup((object),                                      \
+                        VM_OBJECT_EVENT_MAPPING_IN_PROGRESS);          \
+       MACRO_END
 
-#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 :                          \
-                    &vm_object_lck_attr))
-#define vm_object_lock_destroy(object) lck_rw_destroy(&(object)->Lock, &vm_object_lck_grp)
+#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_unlock(object)       lck_rw_done(&(object)->Lock)
-#define vm_object_lock_upgrade(object) lck_rw_lock_shared_to_exclusive(&(object)->Lock)
-#define vm_object_lock_try_scan(object)        lck_rw_try_lock_exclusive(&(object)->Lock)
 
-/*
- * 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 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)
-#else /* DEBUG */
-#define vm_object_lock_assert_held(object)
-#define vm_object_lock_assert_shared(object)
-#define vm_object_lock_assert_exclusive(object)
-#endif /* DEBUG */
 
 #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))
 
+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);
+
+#define VM_OBJECT_OWNER_DISOWNED ((task_t) -1)
+#define VM_OBJECT_OWNER(object)                                                \
+       ((((object)->purgable == VM_PURGABLE_DENY &&                    \
+          (object)->vo_ledger_tag == 0) ||                             \
+         (object)->vo_owner == TASK_NULL)                              \
+        ? TASK_NULL    /* not owned */                                 \
+        : (((object)->vo_owner == VM_OBJECT_OWNER_DISOWNED)            \
+           ? kernel_task /* disowned -> kernel */                      \
+           : (object)->vo_owner)) /* explicit owner */                 \
+
+extern void    vm_object_ledger_tag_ledgers(
+       vm_object_t object,
+       int *ledger_idx_volatile,
+       int *ledger_idx_nonvolatile,
+       int *ledger_idx_volatile_compressed,
+       int *ledger_idx_nonvolatile_compressed,
+       boolean_t *do_footprint);
+extern kern_return_t vm_object_ownership_change(
+       vm_object_t object,
+       int ledger_tag,
+       task_t owner,
+       boolean_t task_objq_locked);
+
 #endif /* _VM_VM_OBJECT_H_ */