]> git.saurik.com Git - apple/xnu.git/blobdiff - osfmk/kern/zalloc.c
xnu-1699.32.7.tar.gz
[apple/xnu.git] / osfmk / kern / zalloc.c
index 45de75913111ecc24c5c755f54fa4aafab11d8bb..9d1afa5d6e84a9e430c8c7218737481dd28c24d0 100644 (file)
@@ -1,5 +1,5 @@
 /*
 /*
- * Copyright (c) 2000-2004 Apple Computer, Inc. All rights reserved.
+ * Copyright (c) 2000-2009 Apple Inc. All rights reserved.
  *
  * @APPLE_OSREFERENCE_LICENSE_HEADER_START@
  * 
  *
  * @APPLE_OSREFERENCE_LICENSE_HEADER_START@
  * 
@@ -63,6 +63,7 @@
  *     data blocks for which quick allocation/deallocation is possible.
  */
 #include <zone_debug.h>
  *     data blocks for which quick allocation/deallocation is possible.
  */
 #include <zone_debug.h>
+#include <zone_alias_addr.h>
 #include <norma_vm.h>
 #include <mach_kdb.h>
 
 #include <norma_vm.h>
 #include <mach_kdb.h>
 
@@ -70,6 +71,7 @@
 #include <mach/vm_param.h>
 #include <mach/kern_return.h>
 #include <mach/mach_host_server.h>
 #include <mach/vm_param.h>
 #include <mach/kern_return.h>
 #include <mach/mach_host_server.h>
+#include <mach/task_server.h>
 #include <mach/machine/vm_types.h>
 #include <mach_debug/zone_info.h>
 
 #include <mach/machine/vm_types.h>
 #include <mach_debug/zone_info.h>
 
@@ -78,7 +80,7 @@
 #include <kern/host.h>
 #include <kern/macro_help.h>
 #include <kern/sched.h>
 #include <kern/host.h>
 #include <kern/macro_help.h>
 #include <kern/sched.h>
-#include <kern/lock.h>
+#include <kern/locks.h>
 #include <kern/sched_prim.h>
 #include <kern/misc_protos.h>
 #include <kern/thread_call.h>
 #include <kern/sched_prim.h>
 #include <kern/misc_protos.h>
 #include <kern/thread_call.h>
 
 #include <machine/machparam.h>
 
 
 #include <machine/machparam.h>
 
-#if defined(__ppc__)
-/* for fake zone stat routines */
-#include <ppc/savearea.h>
-#include <ppc/mappings.h>
-#endif
+#include <libkern/OSDebug.h>
+#include <libkern/OSAtomic.h>
+#include <sys/kdebug.h>
 
 
-#if    MACH_ASSERT
-/* Detect use of zone elt after freeing it by two methods:
+/* 
+ * Zone Corruption Debugging
+ *
+ * We provide three methods to detect use of a zone element after it's been freed.  These
+ * checks are enabled by specifying "-zc" and/or "-zp" in the boot-args:
+ *
  * (1) Range-check the free-list "next" ptr for sanity.
  * (2) Store the ptr in two different words, and compare them against
  * (1) Range-check the free-list "next" ptr for sanity.
  * (2) Store the ptr in two different words, and compare them against
- *     each other when re-using the zone elt, to detect modifications;
+ *     each other when re-using the zone element, to detect modifications.
+ * (3) poison the freed memory by overwriting it with 0xdeadbeef.
+ *
+ * The first two checks are fairly light weight and are enabled by specifying "-zc" 
+ * in the boot-args.  If you want more aggressive checking for use-after-free bugs
+ * and you don't mind the additional overhead, then turn on poisoning by adding
+ * "-zp" to the boot-args in addition to "-zc".  If you specify -zp without -zc,
+ * it still poisons the memory when it's freed, but doesn't check if the memory
+ * has been altered later when it's reallocated.
  */
 
  */
 
-#if defined(__alpha)
+boolean_t check_freed_element = FALSE;         /* enabled by -zc in boot-args */
+boolean_t zfree_clear = FALSE;                 /* enabled by -zp in boot-args */
 
 
-#define is_kernel_data_addr(a)                                         \
-  (!(a) || (IS_SYS_VA(a) && !((a) & (sizeof(long)-1))))
-
-#else /* !defined(__alpha) */
+/*
+ * Fake zones for things that want to report via zprint but are not actually zones.
+ */
+struct fake_zone_info {
+       const char* name;
+       void (*init)(int);
+       void (*query)(int *,
+                    vm_size_t *, vm_size_t *, vm_size_t *, vm_size_t *,
+                     uint64_t *, int *, int *, int *);
+};
 
 
-#define is_kernel_data_addr(a)                                         \
-  (!(a) || ((a) >= vm_min_kernel_address && !((a) & 0x3)))
+static struct fake_zone_info fake_zones[] = {
+       {
+               .name = "kernel_stacks",
+               .init = stack_fake_zone_init,
+               .query = stack_fake_zone_info,
+       },
+#if defined(__i386__) || defined (__x86_64__)
+       {
+               .name = "page_tables",
+               .init = pt_fake_zone_init,
+               .query = pt_fake_zone_info,
+       },
+#endif /* i386 */
+       {
+               .name = "kalloc.large",
+               .init = kalloc_fake_zone_init,
+               .query = kalloc_fake_zone_info,
+       },
+};
+unsigned int num_fake_zones = sizeof(fake_zones)/sizeof(fake_zones[0]);
 
 
-#endif /* defined(__alpha) */
+/*
+ * Zone info options
+ */
+boolean_t zinfo_per_task = FALSE;              /* enabled by -zinfop in boot-args */
+#define ZINFO_SLOTS 200                                /* for now */
+#define ZONES_MAX (ZINFO_SLOTS - num_fake_zones - 1)
 
 
-/* Should we set all words of the zone element to an illegal address
- * when it is freed, to help catch usage after freeing?  The down-side
- * is that this obscures the identity of the freed element.
+/* 
+ * Allocation helper macros
  */
  */
-boolean_t zfree_clear = FALSE;
+#define is_kernel_data_addr(a) (!(a) || ((a) >= vm_min_kernel_address && !((a) & 0x3)))
 
 #define ADD_TO_ZONE(zone, element)                                     \
 MACRO_BEGIN                                                            \
 
 #define ADD_TO_ZONE(zone, element)                                     \
 MACRO_BEGIN                                                            \
-               if (zfree_clear)                                        \
-               {   unsigned int i;                                             \
-                   for (i=1;                                           \
-                        i < zone->elem_size/sizeof(vm_offset_t) - 1;   \
-                        i++)                                           \
-                   ((vm_offset_t *)(element))[i] = 0xdeadbeef;         \
-               }                                                       \
-               ((vm_offset_t *)(element))[0] = (zone)->free_elements;  \
-               (zone)->free_elements = (vm_offset_t) (element);        \
-               (zone)->count--;                                        \
-MACRO_END
-
-#define REMOVE_FROM_ZONE(zone, ret, type)                              \
-MACRO_BEGIN                                                            \
-       (ret) = (type) (zone)->free_elements;                           \
-       if ((ret) != (type) 0) {                                        \
-           if (!is_kernel_data_addr(((vm_offset_t *)(ret))[0])) {      \
-               panic("A freed zone element has been modified.\n");     \
-           }                                                           \
-           (zone)->count++;                                            \
-           (zone)->free_elements = *((vm_offset_t *)(ret));            \
+       if (zfree_clear)                                                \
+       {   unsigned int i;                                             \
+           for (i=0;                                                   \
+                i < zone->elem_size/sizeof(uint32_t);                  \
+                i++)                                                   \
+           ((uint32_t *)(element))[i] = 0xdeadbeef;                    \
        }                                                               \
        }                                                               \
-MACRO_END
-#else  /* MACH_ASSERT */
-
-#define ADD_TO_ZONE(zone, element)                                     \
-MACRO_BEGIN                                                            \
-               *((vm_offset_t *)(element)) = (zone)->free_elements;    \
-               (zone)->free_elements = (vm_offset_t) (element);        \
-               (zone)->count--;                                        \
-MACRO_END
-
-#define REMOVE_FROM_ZONE(zone, ret, type)                              \
-MACRO_BEGIN                                                            \
-       (ret) = (type) (zone)->free_elements;                           \
-       if ((ret) != (type) 0) {                                        \
-               (zone)->count++;                                        \
-               (zone)->free_elements = *((vm_offset_t *)(ret));        \
+       *((vm_offset_t *)(element)) = (zone)->free_elements;            \
+       if (check_freed_element) {                                      \
+               if ((zone)->elem_size >= (2 * sizeof(vm_offset_t)))     \
+                       ((vm_offset_t *)(element))[((zone)->elem_size/sizeof(vm_offset_t))-1] = \
+                               (zone)->free_elements;                  \
        }                                                               \
        }                                                               \
+       (zone)->free_elements = (vm_offset_t) (element);                \
+       (zone)->count--;                                                \
 MACRO_END
 
 MACRO_END
 
-#endif /* MACH_ASSERT */
+#define REMOVE_FROM_ZONE(zone, ret, type)                                      \
+MACRO_BEGIN                                                                    \
+       (ret) = (type) (zone)->free_elements;                                   \
+       if ((ret) != (type) 0) {                                                \
+               if (check_freed_element) {                                      \
+                       if (!is_kernel_data_addr(((vm_offset_t *)(ret))[0]) ||  \
+                           ((zone)->elem_size >= (2 * sizeof(vm_offset_t)) &&  \
+                           ((vm_offset_t *)(ret))[((zone)->elem_size/sizeof(vm_offset_t))-1] != \
+                           ((vm_offset_t *)(ret))[0]))                         \
+                               panic("a freed zone element has been modified");\
+                       if (zfree_clear) {                                      \
+                               unsigned int ii;                                \
+                               for (ii = sizeof(vm_offset_t) / sizeof(uint32_t); \
+                                    ii < (zone)->elem_size/sizeof(uint32_t) - sizeof(vm_offset_t) / sizeof(uint32_t); \
+                                        ii++)                                  \
+                                       if (((uint32_t *)(ret))[ii] != (uint32_t)0xdeadbeef) \
+                                               panic("a freed zone element has been modified");\
+                       }                                                       \
+               }                                                               \
+               (zone)->count++;                                                \
+               (zone)->sum_count++;                                            \
+               (zone)->free_elements = *((vm_offset_t *)(ret));                \
+       }                                                                       \
+MACRO_END
 
 #if    ZONE_DEBUG
 #define zone_debug_enabled(z) z->active_zones.next
 
 #if    ZONE_DEBUG
 #define zone_debug_enabled(z) z->active_zones.next
@@ -175,27 +215,47 @@ MACRO_END
 #endif /* ZONE_DEBUG */
 
 /*
 #endif /* ZONE_DEBUG */
 
 /*
- * Support for garbage collection of unused zone pages:
+ * Support for garbage collection of unused zone pages
+ *
+ * The kernel virtually allocates the "zone map" submap of the kernel
+ * map. When an individual zone needs more storage, memory is allocated
+ * out of the zone map, and the two-level "zone_page_table" is
+ * on-demand expanded so that it has entries for those pages.
+ * zone_page_init()/zone_page_alloc() initialize "alloc_count"
+ * to the number of zone elements that occupy the zone page (which may
+ * be a minimum of 1, including if a zone element spans multiple
+ * pages).
+ *
+ * Asynchronously, the zone_gc() logic attempts to walk zone free
+ * lists to see if all the elements on a zone page are free. If
+ * "collect_count" (which it increments during the scan) matches
+ * "alloc_count", the zone page is a candidate for collection and the
+ * physical page is returned to the VM system. During this process, the
+ * first word of the zone page is re-used to maintain a linked list of
+ * to-be-collected zone pages.
  */
  */
+typedef uint32_t zone_page_index_t;
+#define ZONE_PAGE_INDEX_INVALID ((zone_page_index_t)0xFFFFFFFFU)
 
 struct zone_page_table_entry {
 
 struct zone_page_table_entry {
-       struct zone_page_table_entry    *link;
-       short   alloc_count;
-       short   collect_count;
+       volatile        uint16_t        alloc_count;
+       volatile        uint16_t        collect_count;
 };
 
 };
 
+#define        ZONE_PAGE_USED  0
+#define ZONE_PAGE_UNUSED 0xffff
+
 /* Forwards */
 void           zone_page_init(
                                vm_offset_t     addr,
 /* Forwards */
 void           zone_page_init(
                                vm_offset_t     addr,
-                               vm_size_t       size,
-                               int             value);
+                               vm_size_t       size);
 
 void           zone_page_alloc(
                                vm_offset_t     addr,
                                vm_size_t       size);
 
 void           zone_page_free_element(
 
 void           zone_page_alloc(
                                vm_offset_t     addr,
                                vm_size_t       size);
 
 void           zone_page_free_element(
-                               struct zone_page_table_entry    **free_pages,
+                               zone_page_index_t       *free_page_list,
                                vm_offset_t     addr,
                                vm_size_t       size);
 
                                vm_offset_t     addr,
                                vm_size_t       size);
 
@@ -215,6 +275,7 @@ void                zalloc_async(
                                thread_call_param_t     p0,  
                                thread_call_param_t     p1);
 
                                thread_call_param_t     p0,  
                                thread_call_param_t     p1);
 
+void           zone_display_zprint( void );
 
 #if    ZONE_DEBUG && MACH_KDB
 int            zone_count(
 
 #if    ZONE_DEBUG && MACH_KDB
 int            zone_count(
@@ -226,9 +287,12 @@ vm_map_t   zone_map = VM_MAP_NULL;
 
 zone_t         zone_zone = ZONE_NULL;  /* the zone containing other zones */
 
 
 zone_t         zone_zone = ZONE_NULL;  /* the zone containing other zones */
 
+zone_t         zinfo_zone = ZONE_NULL; /* zone of per-task zone info */
+
 /*
  *     The VM system gives us an initial chunk of memory.
  *     It has to be big enough to allocate the zone_zone
 /*
  *     The VM system gives us an initial chunk of memory.
  *     It has to be big enough to allocate the zone_zone
+ *     all the way through the pmap zone.
  */
 
 vm_offset_t    zdata;
  */
 
 vm_offset_t    zdata;
@@ -236,56 +300,70 @@ vm_size_t zdata_size;
 
 #define lock_zone(zone)                                        \
 MACRO_BEGIN                                            \
 
 #define lock_zone(zone)                                        \
 MACRO_BEGIN                                            \
-       mutex_lock(&(zone)->lock);                      \
+       lck_mtx_lock_spin(&(zone)->lock);                       \
 MACRO_END
 
 #define unlock_zone(zone)                              \
 MACRO_BEGIN                                            \
 MACRO_END
 
 #define unlock_zone(zone)                              \
 MACRO_BEGIN                                            \
-       mutex_unlock(&(zone)->lock);                    \
+       lck_mtx_unlock(&(zone)->lock);                  \
 MACRO_END
 
 #define zone_wakeup(zone) thread_wakeup((event_t)(zone))
 #define zone_sleep(zone)                               \
 MACRO_END
 
 #define zone_wakeup(zone) thread_wakeup((event_t)(zone))
 #define zone_sleep(zone)                               \
-       thread_sleep_mutex((event_t)(zone),     \
-                               &(zone)->lock,          \
-                               THREAD_UNINT)
+       (void) lck_mtx_sleep(&(zone)->lock, LCK_SLEEP_SPIN, (event_t)(zone), THREAD_UNINT);
+
 
 #define lock_zone_init(zone)                           \
 MACRO_BEGIN                                            \
 
 #define lock_zone_init(zone)                           \
 MACRO_BEGIN                                            \
-       mutex_init(&zone->lock, 0);     \
+       char _name[32];                                 \
+       (void) snprintf(_name, sizeof (_name), "zone.%s", (zone)->zone_name); \
+       lck_grp_attr_setdefault(&(zone)->lock_grp_attr);                \
+       lck_grp_init(&(zone)->lock_grp, _name, &(zone)->lock_grp_attr); \
+       lck_attr_setdefault(&(zone)->lock_attr);                        \
+       lck_mtx_init_ext(&(zone)->lock, &(zone)->lock_ext,              \
+           &(zone)->lock_grp, &(zone)->lock_attr);                     \
 MACRO_END
 
 MACRO_END
 
-#define lock_try_zone(zone)    mutex_try(&zone->lock)
-
-kern_return_t          zget_space(
-                               vm_offset_t size,
-                               vm_offset_t *result);
-
-decl_simple_lock_data(,zget_space_lock)
-vm_offset_t    zalloc_next_space;
-vm_offset_t    zalloc_end_of_space;
-vm_size_t      zalloc_wasted_space;
+#define lock_try_zone(zone)    lck_mtx_try_lock_spin(&zone->lock)
 
 /*
  *     Garbage collection map information
  */
 
 /*
  *     Garbage collection map information
  */
-struct zone_page_table_entry * zone_page_table;
+#define ZONE_PAGE_TABLE_FIRST_LEVEL_SIZE (32)
+struct zone_page_table_entry * volatile zone_page_table[ZONE_PAGE_TABLE_FIRST_LEVEL_SIZE];
+vm_size_t                      zone_page_table_used_size;
 vm_offset_t                    zone_map_min_address;
 vm_offset_t                    zone_map_max_address;
 unsigned int                   zone_pages;
 vm_offset_t                    zone_map_min_address;
 vm_offset_t                    zone_map_max_address;
 unsigned int                   zone_pages;
+unsigned int                   zone_page_table_second_level_size;                      /* power of 2 */
+unsigned int                   zone_page_table_second_level_shift_amount;
+
+#define zone_page_table_first_level_slot(x)  ((x) >> zone_page_table_second_level_shift_amount)
+#define zone_page_table_second_level_slot(x) ((x) & (zone_page_table_second_level_size - 1))
+
+void   zone_page_table_expand(zone_page_index_t pindex);
+struct zone_page_table_entry *zone_page_table_lookup(zone_page_index_t pindex);
 
 /*
  *     Exclude more than one concurrent garbage collection
  */
 
 /*
  *     Exclude more than one concurrent garbage collection
  */
-decl_mutex_data(,              zone_gc_lock)
+decl_lck_mtx_data(,            zone_gc_lock)
+
+lck_attr_t      zone_lck_attr;
+lck_grp_t       zone_lck_grp;
+lck_grp_attr_t  zone_lck_grp_attr;
+lck_mtx_ext_t   zone_lck_ext;
 
 
+
+#if    !ZONE_ALIAS_ADDR
 #define from_zone_map(addr, size) \
        ((vm_offset_t)(addr) >= zone_map_min_address && \
         ((vm_offset_t)(addr) + size -1) <  zone_map_max_address)
 #define from_zone_map(addr, size) \
        ((vm_offset_t)(addr) >= zone_map_min_address && \
         ((vm_offset_t)(addr) + size -1) <  zone_map_max_address)
-
-#define        ZONE_PAGE_USED  0
-#define ZONE_PAGE_UNUSED -1
-
+#else
+#define from_zone_map(addr, size) \
+       ((vm_offset_t)(zone_virtual_addr((vm_map_address_t)addr)) >= zone_map_min_address && \
+        ((vm_offset_t)(zone_virtual_addr((vm_map_address_t)addr)) + size -1) <  zone_map_max_address)
+#endif
 
 /*
  *     Protects first_zone, last_zone, num_zones,
 
 /*
  *     Protects first_zone, last_zone, num_zones,
@@ -298,9 +376,661 @@ unsigned int              num_zones;
 
 boolean_t zone_gc_allowed = TRUE;
 boolean_t zone_gc_forced = FALSE;
 
 boolean_t zone_gc_allowed = TRUE;
 boolean_t zone_gc_forced = FALSE;
-unsigned zone_gc_last_tick = 0;
-unsigned zone_gc_max_rate = 0;         /* in ticks */
+boolean_t panic_include_zprint = FALSE;
+boolean_t zone_gc_allowed_by_time_throttle = TRUE;
+
+/*
+ * Zone leak debugging code
+ *
+ * When enabled, this code keeps a log to track allocations to a particular zone that have not
+ * yet been freed.  Examining this log will reveal the source of a zone leak.  The log is allocated
+ * only when logging is enabled, so there is no effect on the system when it's turned off.  Logging is
+ * off by default.
+ *
+ * Enable the logging via the boot-args. Add the parameter "zlog=<zone>" to boot-args where <zone>
+ * is the name of the zone you wish to log.  
+ *
+ * This code only tracks one zone, so you need to identify which one is leaking first.
+ * Generally, you'll know you have a leak when you get a "zalloc retry failed 3" panic from the zone
+ * garbage collector.  Note that the zone name printed in the panic message is not necessarily the one
+ * containing the leak.  So do a zprint from gdb and locate the zone with the bloated size.  This
+ * is most likely the problem zone, so set zlog in boot-args to this zone name, reboot and re-run the test.  The
+ * next time it panics with this message, examine the log using the kgmacros zstack, findoldest and countpcs.
+ * See the help in the kgmacros for usage info.
+ *
+ *
+ * Zone corruption logging
+ *
+ * Logging can also be used to help identify the source of a zone corruption.  First, identify the zone
+ * that is being corrupted, then add "-zc zlog=<zone name>" to the boot-args.  When -zc is used in conjunction
+ * with zlog, it changes the logging style to track both allocations and frees to the zone.  So when the
+ * corruption is detected, examining the log will show you the stack traces of the callers who last allocated
+ * and freed any particular element in the zone.  Use the findelem kgmacro with the address of the element that's been
+ * corrupted to examine its history.  This should lead to the source of the corruption.
+ */
+
+static int log_records;        /* size of the log, expressed in number of records */
+
+#define MAX_ZONE_NAME  32      /* max length of a zone name we can take from the boot-args */
+
+static char zone_name_to_log[MAX_ZONE_NAME] = "";      /* the zone name we're logging, if any */
+
+/*
+ * The number of records in the log is configurable via the zrecs parameter in boot-args.  Set this to 
+ * the number of records you want in the log.  For example, "zrecs=1000" sets it to 1000 records.  Note
+ * that the larger the size of the log, the slower the system will run due to linear searching in the log,
+ * but one doesn't generally care about performance when tracking down a leak.  The log is capped at 8000
+ * records since going much larger than this tends to make the system unresponsive and unbootable on small
+ * memory configurations.  The default value is 4000 records.
+ */
+#if    defined(__LP64__)
+#define ZRECORDS_MAX           16000           /* Max records allowed in the log */
+#else
+#define ZRECORDS_MAX           8000            /* Max records allowed in the log */
+#endif
+#define ZRECORDS_DEFAULT       4000            /* default records in log if zrecs is not specificed in boot-args */
+
+/*
+ * Each record in the log contains a pointer to the zone element it refers to, a "time" number that allows
+ * the records to be ordered chronologically, and a small array to hold the pc's from the stack trace.  A
+ * record is added to the log each time a zalloc() is done in the zone_of_interest.  For leak debugging,
+ * the record is cleared when a zfree() is done.  For corruption debugging, the log tracks both allocs and frees.
+ * If the log fills, old records are replaced as if it were a circular buffer.
+ */
+
+struct zrecord {
+        void           *z_element;             /* the element that was zalloc'ed of zfree'ed */
+        uint32_t       z_opcode:1,             /* whether it was a zalloc or zfree */
+                       z_time:31;              /* time index when operation was done */
+        void           *z_pc[MAX_ZTRACE_DEPTH];        /* stack trace of caller */
+};
+
+/*
+ * Opcodes for the z_opcode field:
+ */
+
+#define ZOP_ALLOC      1
+#define ZOP_FREE       0
+
+/*
+ * The allocation log and all the related variables are protected by the zone lock for the zone_of_interest
+ */
+
+static struct zrecord *zrecords;               /* the log itself, dynamically allocated when logging is enabled  */
+static int zcurrent  = 0;                      /* index of the next slot in the log to use */
+static int zrecorded = 0;                      /* number of allocations recorded in the log */
+static unsigned int ztime = 0;                 /* a timestamp of sorts */
+static zone_t  zone_of_interest = NULL;                /* the zone being watched; corresponds to zone_name_to_log */
+
+/*
+ * Decide if we want to log this zone by doing a string compare between a zone name and the name
+ * of the zone to log. Return true if the strings are equal, false otherwise.  Because it's not
+ * possible to include spaces in strings passed in via the boot-args, a period in the logname will
+ * match a space in the zone name.
+ */
+
+static int
+log_this_zone(const char *zonename, const char *logname) 
+{
+       int len;
+       const char *zc = zonename;
+       const char *lc = logname;
+
+       /*
+        * Compare the strings.  We bound the compare by MAX_ZONE_NAME.
+        */
+
+       for (len = 1; len <= MAX_ZONE_NAME; zc++, lc++, len++) {
+
+               /*
+                * If the current characters don't match, check for a space in
+                * in the zone name and a corresponding period in the log name.
+                * If that's not there, then the strings don't match.
+                */
+
+               if (*zc != *lc && !(*zc == ' ' && *lc == '.')) 
+                       break;
+
+               /*
+                * The strings are equal so far.  If we're at the end, then it's a match.
+                */
+
+               if (*zc == '\0')
+                       return TRUE;
+       }
+
+       return FALSE;
+}
+
+
+/*
+ * Test if we want to log this zalloc/zfree event.  We log if this is the zone we're interested in and
+ * the buffer for the records has been allocated.
+ */
+
+#define DO_LOGGING(z)          (zrecords && (z) == zone_of_interest)
+
+extern boolean_t zlog_ready;
+
+#if CONFIG_ZLEAKS
+#pragma mark -
+#pragma mark Zone Leak Detection
+
+/* 
+ * The zone leak detector, abbreviated 'zleak', keeps track of a subset of the currently outstanding
+ * allocations made by the zone allocator.  Every z_sample_factor allocations in each zone, we capture a
+ * backtrace.  Every free, we examine the table and determine if the allocation was being tracked, 
+ * and stop tracking it if it was being tracked.
+ *
+ * We track the allocations in the zallocations hash table, which stores the address that was returned from 
+ * the zone allocator.  Each stored entry in the zallocations table points to an entry in the ztraces table, which
+ * stores the backtrace associated with that allocation.  This provides uniquing for the relatively large
+ * backtraces - we don't store them more than once.
+ *
+ * Data collection begins when the zone map is 50% full, and only occurs for zones that are taking up
+ * a large amount of virtual space.
+ */
+#define ZLEAK_STATE_ENABLED            0x01    /* Zone leak monitoring should be turned on if zone_map fills up. */
+#define ZLEAK_STATE_ACTIVE             0x02    /* We are actively collecting traces. */
+#define ZLEAK_STATE_ACTIVATING                 0x04    /* Some thread is doing setup; others should move along. */
+#define ZLEAK_STATE_FAILED             0x08    /* Attempt to allocate tables failed.  We will not try again. */
+uint32_t       zleak_state = 0;                /* State of collection, as above */
+
+boolean_t      panic_include_ztrace    = FALSE;        /* Enable zleak logging on panic */
+vm_size_t      zleak_global_tracking_threshold;        /* Size of zone map at which to start collecting data */
+vm_size_t      zleak_per_zone_tracking_threshold;      /* Size a zone will have before we will collect data on it */
+unsigned int   z_sample_factor = 1000;                 /* Allocations per sample attempt */
+
+/*
+ * Counters for allocation statistics.
+ */ 
+
+/* Times two active records want to occupy the same spot */
+unsigned int z_alloc_collisions = 0;
+unsigned int z_trace_collisions = 0;
+
+/* Times a new record lands on a spot previously occupied by a freed allocation */
+unsigned int z_alloc_overwrites = 0;
+unsigned int z_trace_overwrites = 0;
+
+/* Times a new alloc or trace is put into the hash table */
+unsigned int z_alloc_recorded  = 0;
+unsigned int z_trace_recorded  = 0;
+
+/* Times zleak_log returned false due to not being able to acquire the lock */
+unsigned int z_total_conflicts = 0;
+
+
+#pragma mark struct zallocation
+/*
+ * Structure for keeping track of an allocation
+ * An allocation bucket is in use if its element is not NULL
+ */
+struct zallocation {
+       uintptr_t               za_element;             /* the element that was zalloc'ed or zfree'ed, NULL if bucket unused */
+       vm_size_t               za_size;                        /* how much memory did this allocation take up? */
+       uint32_t                za_trace_index; /* index into ztraces for backtrace associated with allocation */
+       /* TODO: #if this out */
+       uint32_t                za_hit_count;           /* for determining effectiveness of hash function */
+};
+
+/* Size must be a power of two for the zhash to be able to just mask off bits instead of mod */
+#define ZLEAK_ALLOCATION_MAP_NUM       16384
+#define ZLEAK_TRACE_MAP_NUM            8192
+
+uint32_t zleak_alloc_buckets = ZLEAK_ALLOCATION_MAP_NUM;
+uint32_t zleak_trace_buckets = ZLEAK_TRACE_MAP_NUM;
+
+vm_size_t zleak_max_zonemap_size;
+
+/* Hashmaps of allocations and their corresponding traces */
+static struct zallocation*     zallocations;
+static struct ztrace*          ztraces;
+
+/* not static so that panic can see this, see kern/debug.c */
+struct ztrace*                         top_ztrace;
+
+/* Lock to protect zallocations, ztraces, and top_ztrace from concurrent modification. */
+static lck_mtx_t                       zleak_lock;
+static lck_attr_t                      zleak_lock_attr;
+static lck_grp_t                       zleak_lock_grp;
+static lck_grp_attr_t                  zleak_lock_grp_attr;
+
+/*
+ * Initializes the zone leak monitor.  Called from zone_init()
+ */
+static void 
+zleak_init(vm_size_t max_zonemap_size) 
+{
+       char                    scratch_buf[16];
+       boolean_t               zleak_enable_flag = FALSE;
+
+       zleak_max_zonemap_size = max_zonemap_size;
+       zleak_global_tracking_threshold = max_zonemap_size / 2; 
+       zleak_per_zone_tracking_threshold = zleak_global_tracking_threshold / 8;
+
+       /* -zleakoff (flag to disable zone leak monitor) */
+       if (PE_parse_boot_argn("-zleakoff", scratch_buf, sizeof(scratch_buf))) {
+               zleak_enable_flag = FALSE;
+               printf("zone leak detection disabled\n");
+       } else {
+               zleak_enable_flag = TRUE;
+               printf("zone leak detection enabled\n");
+       }
+       
+       /* zfactor=XXXX (override how often to sample the zone allocator) */
+       if (PE_parse_boot_argn("zfactor", &z_sample_factor, sizeof(z_sample_factor))) {
+               printf("Zone leak factor override:%u\n", z_sample_factor);
+       }
+       
+       /* zleak-allocs=XXXX (override number of buckets in zallocations) */
+       if (PE_parse_boot_argn("zleak-allocs", &zleak_alloc_buckets, sizeof(zleak_alloc_buckets))) {
+               printf("Zone leak alloc buckets override:%u\n", zleak_alloc_buckets);
+               /* uses 'is power of 2' trick: (0x01000 & 0x00FFF == 0) */
+               if (zleak_alloc_buckets == 0 || (zleak_alloc_buckets & (zleak_alloc_buckets-1))) {
+                       printf("Override isn't a power of two, bad things might happen!");
+               }
+       }
+       
+       /* zleak-traces=XXXX (override number of buckets in ztraces) */
+       if (PE_parse_boot_argn("zleak-traces", &zleak_trace_buckets, sizeof(zleak_trace_buckets))) {
+               printf("Zone leak trace buckets override:%u\n", zleak_trace_buckets);
+               /* uses 'is power of 2' trick: (0x01000 & 0x00FFF == 0) */
+               if (zleak_trace_buckets == 0 || (zleak_trace_buckets & (zleak_trace_buckets-1))) {
+                       printf("Override isn't a power of two, bad things might happen!");
+               }
+       }
+       
+       /* allocate the zleak_lock */
+       lck_grp_attr_setdefault(&zleak_lock_grp_attr);
+       lck_grp_init(&zleak_lock_grp, "zleak_lock", &zleak_lock_grp_attr);
+       lck_attr_setdefault(&zleak_lock_attr);
+       lck_mtx_init(&zleak_lock, &zleak_lock_grp, &zleak_lock_attr);
+       
+       if (zleak_enable_flag) {
+               zleak_state = ZLEAK_STATE_ENABLED;
+       }
+}
+
+#if CONFIG_ZLEAKS
+
+/*
+ * Support for kern.zleak.active sysctl - a simplified
+ * simplified version of the zleak_state variable.
+ */
+int
+get_zleak_state(void)
+{
+       if (zleak_state & ZLEAK_STATE_FAILED)
+               return (-1);
+       if (zleak_state & ZLEAK_STATE_ACTIVE)
+               return (1);
+       return (0);
+}
+
+#endif
+
 
 
+kern_return_t
+zleak_activate(void)
+{
+       kern_return_t retval;
+       vm_size_t z_alloc_size = zleak_alloc_buckets * sizeof(struct zallocation);
+       vm_size_t z_trace_size = zleak_trace_buckets * sizeof(struct ztrace);
+       void *allocations_ptr = NULL;
+       void *traces_ptr = NULL;
+
+       /* Only one thread attempts to activate at a time */
+       if (zleak_state & (ZLEAK_STATE_ACTIVE | ZLEAK_STATE_ACTIVATING | ZLEAK_STATE_FAILED)) {
+               return KERN_SUCCESS;
+       }
+
+       /* Indicate that we're doing the setup */
+       lck_mtx_lock_spin(&zleak_lock);
+       if (zleak_state & (ZLEAK_STATE_ACTIVE | ZLEAK_STATE_ACTIVATING | ZLEAK_STATE_FAILED)) {
+               lck_mtx_unlock(&zleak_lock);
+               return KERN_SUCCESS;
+       }
+
+       zleak_state |= ZLEAK_STATE_ACTIVATING;
+       lck_mtx_unlock(&zleak_lock);
+
+       /* Allocate and zero tables */
+       retval = kmem_alloc_kobject(kernel_map, (vm_offset_t*)&allocations_ptr, z_alloc_size);
+       if (retval != KERN_SUCCESS) {
+               goto fail;
+       }
+
+       retval = kmem_alloc_kobject(kernel_map, (vm_offset_t*)&traces_ptr, z_trace_size);
+       if (retval != KERN_SUCCESS) {
+               goto fail;
+       }
+
+       bzero(allocations_ptr, z_alloc_size);
+       bzero(traces_ptr, z_trace_size);
+
+       /* Everything's set.  Install tables, mark active. */
+       zallocations = allocations_ptr;
+       ztraces = traces_ptr;
+
+       /*
+        * Initialize the top_ztrace to the first entry in ztraces, 
+        * so we don't have to check for null in zleak_log
+        */
+       top_ztrace = &ztraces[0];
+
+       /*
+        * Note that we do need a barrier between installing
+        * the tables and setting the active flag, because the zfree()
+        * path accesses the table without a lock if we're active.
+        */
+       lck_mtx_lock_spin(&zleak_lock);
+       zleak_state |= ZLEAK_STATE_ACTIVE;
+       zleak_state &= ~ZLEAK_STATE_ACTIVATING;
+       lck_mtx_unlock(&zleak_lock);
+       
+       return 0;
+
+fail:  
+       /*
+        * If we fail to allocate memory, don't further tax
+        * the system by trying again.
+        */
+       lck_mtx_lock_spin(&zleak_lock);
+       zleak_state |= ZLEAK_STATE_FAILED;
+       zleak_state &= ~ZLEAK_STATE_ACTIVATING;
+       lck_mtx_unlock(&zleak_lock);
+
+       if (allocations_ptr != NULL) {
+               kmem_free(kernel_map, (vm_offset_t)allocations_ptr, z_alloc_size);
+       }
+
+       if (traces_ptr != NULL) {
+               kmem_free(kernel_map, (vm_offset_t)traces_ptr, z_trace_size);
+       }
+
+       return retval;
+}
+
+/*
+ * TODO: What about allocations that never get deallocated, 
+ * especially ones with unique backtraces? Should we wait to record
+ * until after boot has completed?  
+ * (How many persistent zallocs are there?)
+ */
+
+/*
+ * This function records the allocation in the allocations table, 
+ * and stores the associated backtrace in the traces table 
+ * (or just increments the refcount if the trace is already recorded)
+ * If the allocation slot is in use, the old allocation is replaced with the new allocation, and
+ * the associated trace's refcount is decremented.
+ * If the trace slot is in use, it returns.
+ * The refcount is incremented by the amount of memory the allocation consumes.
+ * The return value indicates whether to try again next time.
+ */
+static boolean_t
+zleak_log(uintptr_t* bt,
+                 uintptr_t addr,
+                 uint32_t depth,
+                 vm_size_t allocation_size) 
+{
+       /* Quit if there's someone else modifying the hash tables */
+       if (!lck_mtx_try_lock_spin(&zleak_lock)) {
+               z_total_conflicts++;
+               return FALSE;
+       }
+       
+       struct zallocation* allocation  = &zallocations[hashaddr(addr, zleak_alloc_buckets)];
+       
+       uint32_t trace_index = hashbacktrace(bt, depth, zleak_trace_buckets);
+       struct ztrace* trace = &ztraces[trace_index];
+       
+       allocation->za_hit_count++;
+       trace->zt_hit_count++;
+       
+       /* 
+        * If the allocation bucket we want to be in is occupied, and if the occupier
+        * has the same trace as us, just bail.  
+        */
+       if (allocation->za_element != (uintptr_t) 0 && trace_index == allocation->za_trace_index) {
+               z_alloc_collisions++;
+               
+               lck_mtx_unlock(&zleak_lock);
+               return TRUE;
+       }
+       
+       /* STEP 1: Store the backtrace in the traces array. */
+       /* A size of zero indicates that the trace bucket is free. */
+       
+       if (trace->zt_size > 0 && bcmp(trace->zt_stack, bt, (depth * sizeof(uintptr_t))) != 0 ) {
+               /* 
+                * Different unique trace with same hash!
+                * Just bail - if we're trying to record the leaker, hopefully the other trace will be deallocated
+                * and get out of the way for later chances
+                */
+               trace->zt_collisions++;
+               z_trace_collisions++;
+               
+               lck_mtx_unlock(&zleak_lock);
+               return TRUE;
+       } else if (trace->zt_size > 0) {
+               /* Same trace, already added, so increment refcount */
+               trace->zt_size += allocation_size;
+       } else {
+               /* Found an unused trace bucket, record the trace here! */
+               if (trace->zt_depth != 0) /* if this slot was previously used but not currently in use */
+                       z_trace_overwrites++;
+               
+               z_trace_recorded++;
+               trace->zt_size                  = allocation_size;
+               memcpy(trace->zt_stack, bt, (depth * sizeof(uintptr_t)) );
+               
+               trace->zt_depth         = depth;
+               trace->zt_collisions    = 0;
+       }
+       
+       /* STEP 2: Store the allocation record in the allocations array. */
+       
+       if (allocation->za_element != (uintptr_t) 0) {
+               /* 
+                * Straight up replace any allocation record that was there.  We don't want to do the work
+                * to preserve the allocation entries that were there, because we only record a subset of the 
+                * allocations anyways.
+                */
+               
+               z_alloc_collisions++;
+               
+               struct ztrace* associated_trace = &ztraces[allocation->za_trace_index];
+               /* Knock off old allocation's size, not the new allocation */
+               associated_trace->zt_size -= allocation->za_size;
+       } else if (allocation->za_trace_index != 0) {
+               /* Slot previously used but not currently in use */
+               z_alloc_overwrites++;
+       }
+
+       allocation->za_element          = addr;
+       allocation->za_trace_index      = trace_index;
+       allocation->za_size             = allocation_size;
+       
+       z_alloc_recorded++;
+       
+       if (top_ztrace->zt_size < trace->zt_size)
+               top_ztrace = trace;
+       
+       lck_mtx_unlock(&zleak_lock);
+       return TRUE;
+}
+
+/*
+ * Free the allocation record and release the stacktrace.
+ * This should be as fast as possible because it will be called for every free.
+ */
+static void
+zleak_free(uintptr_t addr,
+                  vm_size_t allocation_size) 
+{
+       if (addr == (uintptr_t) 0)
+               return;
+       
+       struct zallocation* allocation = &zallocations[hashaddr(addr, zleak_alloc_buckets)];
+       
+       /* Double-checked locking: check to find out if we're interested, lock, check to make
+        * sure it hasn't changed, then modify it, and release the lock.
+        */
+       
+       if (allocation->za_element == addr && allocation->za_trace_index < zleak_trace_buckets) {
+               /* if the allocation was the one, grab the lock, check again, then delete it */
+               lck_mtx_lock_spin(&zleak_lock);
+               
+               if (allocation->za_element == addr && allocation->za_trace_index < zleak_trace_buckets) {
+                       struct ztrace *trace;
+
+                       /* allocation_size had better match what was passed into zleak_log - otherwise someone is freeing into the wrong zone! */
+                       if (allocation->za_size != allocation_size) {
+                               panic("Freeing as size %lu memory that was allocated with size %lu\n", 
+                                               (uintptr_t)allocation_size, (uintptr_t)allocation->za_size);
+                       }
+                       
+                       trace = &ztraces[allocation->za_trace_index];
+                       
+                       /* size of 0 indicates trace bucket is unused */
+                       if (trace->zt_size > 0) {
+                               trace->zt_size -= allocation_size;
+                       }
+                       
+                       /* A NULL element means the allocation bucket is unused */
+                       allocation->za_element = 0;
+               }
+               lck_mtx_unlock(&zleak_lock);
+       }
+}
+
+#endif /* CONFIG_ZLEAKS */
+
+/*  These functions outside of CONFIG_ZLEAKS because they are also used in
+ *  mbuf.c for mbuf leak-detection.  This is why they lack the z_ prefix.
+ */
+
+/*
+ * This function captures a backtrace from the current stack and
+ * returns the number of frames captured, limited by max_frames.
+ * It's fast because it does no checking to make sure there isn't bad data.
+ * Since it's only called from threads that we're going to keep executing,
+ * if there's bad data we were going to die eventually.
+ * This seems to work for x86 and X86_64.
+ * ARMTODO: Test it on ARM, I think it will work but I can't test it.  If it works, remove the ifdef.
+ * If this function is inlined, it doesn't record the frame of the function it's inside.
+ * (because there's no stack frame!)
+ */
+uint32_t
+fastbacktrace(uintptr_t* bt, uint32_t max_frames)
+{
+#if defined(__x86_64__) || defined(__i386__)
+       uintptr_t* frameptr = NULL, *frameptr_next = NULL;
+       uintptr_t retaddr = 0;
+       uint32_t frame_index = 0, frames = 0;
+       uintptr_t kstackb, kstackt;
+
+       kstackb = current_thread()->kernel_stack;
+       kstackt = kstackb + kernel_stack_size;
+       /* Load stack frame pointer (EBP on x86) into frameptr */
+       frameptr = __builtin_frame_address(0);
+
+       while (frameptr != NULL && frame_index < max_frames ) {
+               /* Next frame pointer is pointed to by the previous one */
+               frameptr_next = (uintptr_t*) *frameptr;
+
+               /* Bail if we see a zero in the stack frame, that means we've reached the top of the stack */
+                /* That also means the return address is worthless, so don't record it */
+               if (frameptr_next == NULL)
+                       break;
+               /* Verify thread stack bounds */
+               if (((uintptr_t)frameptr_next > kstackt) || ((uintptr_t)frameptr_next < kstackb))
+                       break;
+               /* Pull return address from one spot above the frame pointer */
+               retaddr = *(frameptr + 1);
+
+               /* Store it in the backtrace array */
+               bt[frame_index++] = retaddr;
+
+               frameptr = frameptr_next;
+       }
+
+       /* Save the number of frames captured for return value */
+       frames = frame_index;
+
+       /* Fill in the rest of the backtrace with zeros */
+       while (frame_index < max_frames)
+               bt[frame_index++] = 0;
+
+       return frames;
+#else
+       return OSBacktrace((void*)bt, max_frames);
+#endif
+}
+
+/* "Thomas Wang's 32/64 bit mix functions."  http://www.concentric.net/~Ttwang/tech/inthash.htm */
+uintptr_t
+hash_mix(uintptr_t x)
+{
+#ifndef __LP64__
+       x += ~(x << 15);
+       x ^=  (x >> 10);
+       x +=  (x << 3 );
+       x ^=  (x >> 6 );
+       x += ~(x << 11);
+       x ^=  (x >> 16);
+#else
+       x += ~(x << 32);
+       x ^=  (x >> 22);
+       x += ~(x << 13);
+       x ^=  (x >> 8 );
+       x +=  (x << 3 );
+       x ^=  (x >> 15);
+       x += ~(x << 27);
+       x ^=  (x >> 31);
+#endif
+       return x;
+}
+
+uint32_t
+hashbacktrace(uintptr_t* bt, uint32_t depth, uint32_t max_size)
+{
+
+       uintptr_t hash = 0;
+       uintptr_t mask = max_size - 1;
+
+       while (--depth) {
+               hash += bt[depth];
+       }
+
+       hash = hash_mix(hash) & mask;
+
+       assert(hash < max_size);
+
+       return (uint32_t) hash;
+}
+
+/*
+ *  TODO: Determine how well distributed this is
+ *      max_size must be a power of 2. i.e 0x10000 because 0x10000-1 is 0x0FFFF which is a great bitmask
+ */
+uint32_t
+hashaddr(uintptr_t pt, uint32_t max_size)
+{
+       uintptr_t hash = 0;
+       uintptr_t mask = max_size - 1;
+
+       hash = hash_mix(pt) & mask;
+
+       assert(hash < max_size);
+
+       return (uint32_t) hash;
+}
+
+/* End of all leak-detection code */
+#pragma mark -
 
 /*
  *     zinit initializes a new zone.  The zone data structures themselves
 
 /*
  *     zinit initializes a new zone.  The zone data structures themselves
@@ -317,9 +1047,10 @@ zinit(
        zone_t          z;
 
        if (zone_zone == ZONE_NULL) {
        zone_t          z;
 
        if (zone_zone == ZONE_NULL) {
-               if (zget_space(sizeof(struct zone), (vm_offset_t *)&z)
-                   != KERN_SUCCESS)
-                       return(ZONE_NULL);
+
+               z = (struct zone *)zdata;
+               zdata += sizeof(*z);
+               zdata_size -= sizeof(*z);
        } else
                z = (zone_t) zalloc(zone_zone);
        if (z == ZONE_NULL)
        } else
                z = (zone_t) zalloc(zone_zone);
        if (z == ZONE_NULL)
@@ -344,7 +1075,16 @@ zinit(
         * This size will be used unless
         * the user suggestion is larger AND has less fragmentation
         */
         * This size will be used unless
         * the user suggestion is larger AND has less fragmentation
         */
-       {       vm_size_t best, waste; unsigned int i;
+#if    ZONE_ALIAS_ADDR
+       if ((size < PAGE_SIZE) && (PAGE_SIZE % size <= PAGE_SIZE / 10))
+               alloc = PAGE_SIZE;
+       else
+#endif
+#if    defined(__LP64__)               
+               if (((alloc % size) != 0) || (alloc > PAGE_SIZE * 8))
+#endif
+               {
+               vm_size_t best, waste; unsigned int i;
                best  = PAGE_SIZE;
                waste = best % size;
 
                best  = PAGE_SIZE;
                waste = best % size;
 
@@ -375,6 +1115,7 @@ use_this_allocation:
        z->alloc_size = alloc;
        z->zone_name = name;
        z->count = 0;
        z->alloc_size = alloc;
        z->zone_name = name;
        z->count = 0;
+       z->sum_count = 0LL;
        z->doing_alloc = FALSE;
        z->doing_gc = FALSE;
        z->exhaustible = FALSE;
        z->doing_alloc = FALSE;
        z->doing_gc = FALSE;
        z->exhaustible = FALSE;
@@ -383,26 +1124,159 @@ use_this_allocation:
        z->expandable  = TRUE;
        z->waiting = FALSE;
        z->async_pending = FALSE;
        z->expandable  = TRUE;
        z->waiting = FALSE;
        z->async_pending = FALSE;
+       z->caller_acct = TRUE;
+       z->noencrypt = FALSE;
+       z->no_callout = FALSE;
+       z->async_prio_refill = FALSE;
+       z->prio_refill_watermark = 0;
+       z->zone_replenish_thread = NULL;
+#if CONFIG_ZLEAKS
+       z->num_allocs = 0;
+       z->num_frees = 0;
+       z->zleak_capture = 0;
+       z->zleak_on = FALSE;
+#endif /* CONFIG_ZLEAKS */
+
+#if    ZONE_DEBUG
+       z->active_zones.next = z->active_zones.prev = NULL;     
+       zone_debug_enable(z);
+#endif /* ZONE_DEBUG */
+       lock_zone_init(z);
+
+       /*
+        *      Add the zone to the all-zones list.
+        *      If we are tracking zone info per task, and we have
+        *      already used all the available stat slots, then keep
+        *      using the overflow zone slot.
+        */
+       z->next_zone = ZONE_NULL;
+       thread_call_setup(&z->call_async_alloc, zalloc_async, z);
+       simple_lock(&all_zones_lock);
+       *last_zone = z;
+       last_zone = &z->next_zone;
+       z->index = num_zones;
+       if (zinfo_per_task) {
+               if (num_zones > ZONES_MAX)
+                       z->index = ZONES_MAX;
+       }
+       num_zones++;
+       simple_unlock(&all_zones_lock);
+
+       /*
+        * Check if we should be logging this zone.  If so, remember the zone pointer.
+        */
+
+        if (log_this_zone(z->zone_name, zone_name_to_log)) {
+               zone_of_interest = z;
+       }
+
+       /*
+        * If we want to log a zone, see if we need to allocate buffer space for the log.  Some vm related zones are
+        * zinit'ed before we can do a kmem_alloc, so we have to defer allocation in that case.  zlog_ready is set to
+        * TRUE once enough of the VM system is up and running to allow a kmem_alloc to work.  If we want to log one
+        * of the VM related zones that's set up early on, we will skip allocation of the log until zinit is called again
+        * later on some other zone.  So note we may be allocating a buffer to log a zone other than the one being initialized
+        * right now.
+        */
+
+       if (zone_of_interest != NULL && zrecords == NULL && zlog_ready) {
+               if (kmem_alloc(kernel_map, (vm_offset_t *)&zrecords, log_records * sizeof(struct zrecord)) == KERN_SUCCESS) {
+
+                       /*
+                        * We got the memory for the log.  Zero it out since the code needs this to identify unused records.
+                        * At this point, everything is set up and we're ready to start logging this zone.
+                        */
+       
+                       bzero((void *)zrecords, log_records * sizeof(struct zrecord));
+                       printf("zone: logging started for zone %s (%p)\n", zone_of_interest->zone_name, zone_of_interest);
+
+               } else {
+                       printf("zone: couldn't allocate memory for zrecords, turning off zleak logging\n");
+                       zone_of_interest = NULL;
+               }
+       }
+
+       return(z);
+}
+unsigned       zone_replenish_loops, zone_replenish_wakeups, zone_replenish_wakeups_initiated;
+
+static void zone_replenish_thread(zone_t);
+
+/* High priority VM privileged thread used to asynchronously refill a designated
+ * zone, such as the reserved VM map entry zone.
+ */
+static void zone_replenish_thread(zone_t z) {
+       vm_size_t free_size;
+       current_thread()->options |= TH_OPT_VMPRIV;
+
+       for (;;) {
+               lock_zone(z);
+               assert(z->prio_refill_watermark != 0);
+               while ((free_size = (z->cur_size - (z->count * z->elem_size))) < (z->prio_refill_watermark * z->elem_size)) {
+                       assert(z->doing_alloc == FALSE);
+                       assert(z->async_prio_refill == TRUE);
 
 
-#if    ZONE_DEBUG
-       z->active_zones.next = z->active_zones.prev = 0;        
-       zone_debug_enable(z);
-#endif /* ZONE_DEBUG */
-       lock_zone_init(z);
+                       unlock_zone(z);
+                       int     zflags = KMA_KOBJECT|KMA_NOPAGEWAIT;
+                       vm_offset_t space, alloc_size;
+                       kern_return_t kr;
+                               
+                       if (vm_pool_low())
+                               alloc_size = round_page(z->elem_size);
+                       else
+                               alloc_size = z->alloc_size;
+                               
+                       if (z->noencrypt)
+                               zflags |= KMA_NOENCRYPT;
+                               
+                       kr = kernel_memory_allocate(zone_map, &space, alloc_size, 0, zflags);
+
+                       if (kr == KERN_SUCCESS) {
+#if    ZONE_ALIAS_ADDR
+                               if (alloc_size == PAGE_SIZE)
+                                       space = zone_alias_addr(space);
+#endif
+                               zcram(z, space, alloc_size);
+                       } else if (kr == KERN_RESOURCE_SHORTAGE) {
+                               VM_PAGE_WAIT();
+                       } else if (kr == KERN_NO_SPACE) {
+                               kr = kernel_memory_allocate(kernel_map, &space, alloc_size, 0, zflags);
+                               if (kr == KERN_SUCCESS) {
+#if    ZONE_ALIAS_ADDR
+                                       if (alloc_size == PAGE_SIZE)
+                                               space = zone_alias_addr(space);
+#endif
+                                       zcram(z, space, alloc_size);
+                               } else {
+                                       assert_wait_timeout(&z->zone_replenish_thread, THREAD_UNINT, 1, 100 * NSEC_PER_USEC);
+                                       thread_block(THREAD_CONTINUE_NULL);
+                               }
+                       }
 
 
-       /*
-        *      Add the zone to the all-zones list.
-        */
+                       lock_zone(z);
+                       zone_replenish_loops++;
+               }
 
 
-       z->next_zone = ZONE_NULL;
-       thread_call_setup(&z->call_async_alloc, zalloc_async, z);
-       simple_lock(&all_zones_lock);
-       *last_zone = z;
-       last_zone = &z->next_zone;
-       num_zones++;
-       simple_unlock(&all_zones_lock);
+               unlock_zone(z);
+               assert_wait(&z->zone_replenish_thread, THREAD_UNINT);
+               thread_block(THREAD_CONTINUE_NULL);
+               zone_replenish_wakeups++;
+       }
+}
 
 
-       return(z);
+void
+zone_prio_refill_configure(zone_t z, vm_size_t low_water_mark) {
+       z->prio_refill_watermark = low_water_mark;
+
+       z->async_prio_refill = TRUE;
+       OSMemoryBarrier();
+       kern_return_t tres = kernel_thread_start_priority((thread_continue_t)zone_replenish_thread, z, MAXPRI_KERNEL, &z->zone_replenish_thread);
+
+       if (tres != KERN_SUCCESS) {
+               panic("zone_prio_refill_configure, thread create: 0x%x", tres);
+       }
+
+       thread_deallocate(z->zone_replenish_thread);
 }
 
 /*
 }
 
 /*
@@ -410,12 +1284,12 @@ use_this_allocation:
  */
 void
 zcram(
  */
 void
 zcram(
-       register zone_t         zone,
-       void                    *newaddr,
+       zone_t          zone,
+       vm_offset_t                     newmem,
        vm_size_t               size)
 {
        vm_size_t               size)
 {
-       register vm_size_t      elem_size;
-       vm_offset_t             newmem = (vm_offset_t) newaddr;
+       vm_size_t       elem_size;
+       boolean_t   from_zm = FALSE;
 
        /* Basic sanity checks */
        assert(zone != ZONE_NULL && newmem != (vm_offset_t)0);
 
        /* Basic sanity checks */
        assert(zone != ZONE_NULL && newmem != (vm_offset_t)0);
@@ -424,10 +1298,16 @@ zcram(
 
        elem_size = zone->elem_size;
 
 
        elem_size = zone->elem_size;
 
+       if (from_zone_map(newmem, size))
+               from_zm = TRUE;
+
+       if (from_zm)
+               zone_page_init(newmem, size);
+
        lock_zone(zone);
        while (size >= elem_size) {
                ADD_TO_ZONE(zone, newmem);
        lock_zone(zone);
        while (size >= elem_size) {
                ADD_TO_ZONE(zone, newmem);
-               if (from_zone_map(newmem, elem_size))
+               if (from_zm)
                        zone_page_alloc(newmem, elem_size);
                zone->count++;  /* compensate for ADD_TO_ZONE */
                size -= elem_size;
                        zone_page_alloc(newmem, elem_size);
                zone->count++;  /* compensate for ADD_TO_ZONE */
                size -= elem_size;
@@ -437,86 +1317,6 @@ zcram(
        unlock_zone(zone);
 }
 
        unlock_zone(zone);
 }
 
-/*
- * Contiguous space allocator for non-paged zones. Allocates "size" amount
- * of memory from zone_map.
- */
-
-kern_return_t
-zget_space(
-       vm_offset_t size,
-       vm_offset_t *result)
-{
-       vm_offset_t     new_space = 0;
-       vm_size_t       space_to_add = 0;
-
-       simple_lock(&zget_space_lock);
-       while ((zalloc_next_space + size) > zalloc_end_of_space) {
-               /*
-                *      Add at least one page to allocation area.
-                */
-
-               space_to_add = round_page(size);
-
-               if (new_space == 0) {
-                       kern_return_t retval;
-                       /*
-                        *      Memory cannot be wired down while holding
-                        *      any locks that the pageout daemon might
-                        *      need to free up pages.  [Making the zget_space
-                        *      lock a complex lock does not help in this
-                        *      regard.]
-                        *
-                        *      Unlock and allocate memory.  Because several
-                        *      threads might try to do this at once, don't
-                        *      use the memory before checking for available
-                        *      space again.
-                        */
-
-                       simple_unlock(&zget_space_lock);
-
-                       retval = kernel_memory_allocate(zone_map, &new_space,
-                               space_to_add, 0, KMA_KOBJECT|KMA_NOPAGEWAIT);
-                       if (retval != KERN_SUCCESS)
-                               return(retval);
-                       zone_page_init(new_space, space_to_add,
-                                                       ZONE_PAGE_USED);
-                       simple_lock(&zget_space_lock);
-                       continue;
-               }
-
-               
-               /*
-                *      Memory was allocated in a previous iteration.
-                *
-                *      Check whether the new region is contiguous
-                *      with the old one.
-                */
-
-               if (new_space != zalloc_end_of_space) {
-                       /*
-                        *      Throw away the remainder of the
-                        *      old space, and start a new one.
-                        */
-                       zalloc_wasted_space +=
-                               zalloc_end_of_space - zalloc_next_space;
-                       zalloc_next_space = new_space;
-               }
-
-               zalloc_end_of_space = new_space + space_to_add;
-
-               new_space = 0;
-       }
-       *result = zalloc_next_space;
-       zalloc_next_space += size;              
-       simple_unlock(&zget_space_lock);
-
-       if (new_space != 0)
-               kmem_free(zone_map, new_space, space_to_add);
-
-       return(KERN_SUCCESS);
-}
-
 
 /*
  *     Steal memory for the zone package.  Called from
 
 /*
  *     Steal memory for the zone package.  Called from
@@ -525,14 +1325,15 @@ zget_space(
 void
 zone_steal_memory(void)
 {
 void
 zone_steal_memory(void)
 {
-       zdata_size = round_page(128*sizeof(struct zone));
-       zdata = (vm_offset_t)((char *)pmap_steal_memory(zdata_size) - (char *)0);
+       /* Request enough early memory to get to the pmap zone */
+       zdata_size = 12 * sizeof(struct zone);
+       zdata = (vm_offset_t)pmap_steal_memory(round_page(zdata_size));
 }
 
 
 /*
  * Fill a zone with enough memory to contain at least nelem elements.
 }
 
 
 /*
  * Fill a zone with enough memory to contain at least nelem elements.
- * Memory is obtained with kmem_alloc_wired from the kernel_map.
+ * Memory is obtained with kmem_alloc_kobject from the kernel_map.
  * Return the number of elements actually put into the zone, which may
  * be more than the caller asked for since the memory allocation is
  * rounded up to a full page.
  * Return the number of elements actually put into the zone, which may
  * be more than the caller asked for since the memory allocation is
  * rounded up to a full page.
@@ -552,13 +1353,13 @@ zfill(
                return 0;
        size = nelem * zone->elem_size;
        size = round_page(size);
                return 0;
        size = nelem * zone->elem_size;
        size = round_page(size);
-       kr = kmem_alloc_wired(kernel_map, &memory, size);
+       kr = kmem_alloc_kobject(kernel_map, &memory, size);
        if (kr != KERN_SUCCESS)
                return 0;
 
        zone_change(zone, Z_FOREIGN, TRUE);
        if (kr != KERN_SUCCESS)
                return 0;
 
        zone_change(zone, Z_FOREIGN, TRUE);
-       zcram(zone, (void *)memory, size);
-       nalloc = size / zone->elem_size;
+       zcram(zone, memory, size);
+       nalloc = (int)(size / zone->elem_size);
        assert(nalloc >= nelem);
 
        return nalloc;
        assert(nalloc >= nelem);
 
        return nalloc;
@@ -572,8 +1373,65 @@ zfill(
 void
 zone_bootstrap(void)
 {
 void
 zone_bootstrap(void)
 {
-       vm_size_t zone_zone_size;
-       vm_offset_t zone_zone_space;
+       char temp_buf[16];
+
+#if 6094439
+       /* enable zone checks by default, to try and catch offenders... */
+#if 0
+       /* 7968354: turn "-zc" back off */
+       check_freed_element = TRUE;
+       /* 7995202: turn "-zp" back off */
+       zfree_clear = TRUE;
+#endif
+       
+       /* ... but allow them to be turned off explicitely */
+       if (PE_parse_boot_argn("-no_zc", temp_buf, sizeof (temp_buf))) {
+               check_freed_element = FALSE;
+       }
+       if (PE_parse_boot_argn("-no_zp", temp_buf, sizeof (temp_buf))) {
+               zfree_clear = FALSE;
+       }
+#endif
+
+       /* see if we want freed zone element checking and/or poisoning */
+       if (PE_parse_boot_argn("-zc", temp_buf, sizeof (temp_buf))) {
+               check_freed_element = TRUE;
+       }
+
+       if (PE_parse_boot_argn("-zp", temp_buf, sizeof (temp_buf))) {
+               zfree_clear = TRUE;
+       }
+
+       if (PE_parse_boot_argn("-zinfop", temp_buf, sizeof (temp_buf))) {
+               zinfo_per_task = TRUE;
+       }
+
+       /*
+        * Check for and set up zone leak detection if requested via boot-args.  We recognized two
+        * boot-args:
+        *
+        *      zlog=<zone_to_log>
+        *      zrecs=<num_records_in_log>
+        *
+        * The zlog arg is used to specify the zone name that should be logged, and zrecs is used to
+        * control the size of the log.  If zrecs is not specified, a default value is used.
+        */
+
+       if (PE_parse_boot_argn("zlog", zone_name_to_log, sizeof(zone_name_to_log)) == TRUE) {
+               if (PE_parse_boot_argn("zrecs", &log_records, sizeof(log_records)) == TRUE) {
+
+                       /*
+                        * Don't allow more than ZRECORDS_MAX records even if the user asked for more.
+                        * This prevents accidentally hogging too much kernel memory and making the system
+                        * unusable.
+                        */
+
+                       log_records = MIN(ZRECORDS_MAX, log_records);
+
+               } else {
+                       log_records = ZRECORDS_DEFAULT;
+               }
+       }
 
        simple_lock_init(&all_zones_lock, 0);
 
 
        simple_lock_init(&all_zones_lock, 0);
 
@@ -581,21 +1439,50 @@ zone_bootstrap(void)
        last_zone = &first_zone;
        num_zones = 0;
 
        last_zone = &first_zone;
        num_zones = 0;
 
-       simple_lock_init(&zget_space_lock, 0);
-       zalloc_next_space = zdata;
-       zalloc_end_of_space = zdata + zdata_size;
-       zalloc_wasted_space = 0;
-
        /* assertion: nobody else called zinit before us */
        assert(zone_zone == ZONE_NULL);
        zone_zone = zinit(sizeof(struct zone), 128 * sizeof(struct zone),
                          sizeof(struct zone), "zones");
        zone_change(zone_zone, Z_COLLECT, FALSE);
        /* assertion: nobody else called zinit before us */
        assert(zone_zone == ZONE_NULL);
        zone_zone = zinit(sizeof(struct zone), 128 * sizeof(struct zone),
                          sizeof(struct zone), "zones");
        zone_change(zone_zone, Z_COLLECT, FALSE);
-       zone_zone_size = zalloc_end_of_space - zalloc_next_space;
-       zget_space(zone_zone_size, &zone_zone_space);
-       zcram(zone_zone, (void *)zone_zone_space, zone_zone_size);
+       zone_change(zone_zone, Z_CALLERACCT, FALSE);
+       zone_change(zone_zone, Z_NOENCRYPT, TRUE);
+
+       zcram(zone_zone, zdata, zdata_size);
+
+       /* initialize fake zones and zone info if tracking by task */
+       if (zinfo_per_task) {
+               vm_size_t zisize = sizeof(zinfo_usage_store_t) * ZINFO_SLOTS;
+               unsigned int i;
+
+               for (i = 0; i < num_fake_zones; i++)
+                       fake_zones[i].init(ZINFO_SLOTS - num_fake_zones + i);
+               zinfo_zone = zinit(zisize, zisize * CONFIG_TASK_MAX,
+                                  zisize, "per task zinfo");
+               zone_change(zinfo_zone, Z_CALLERACCT, FALSE);
+       }
+}
+
+void
+zinfo_task_init(task_t task)
+{
+       if (zinfo_per_task) {
+               task->tkm_zinfo = zalloc(zinfo_zone);
+               memset(task->tkm_zinfo, 0, sizeof(zinfo_usage_store_t) * ZINFO_SLOTS);
+       } else {
+               task->tkm_zinfo = NULL;
+       }
 }
 
 }
 
+void
+zinfo_task_free(task_t task)
+{
+       assert(task != kernel_task);
+       if (task->tkm_zinfo != NULL) {
+               zfree(zinfo_zone, task->tkm_zinfo);
+               task->tkm_zinfo = NULL;
+       }
+}
+               
 void
 zone_init(
        vm_size_t max_zonemap_size)
 void
 zone_init(
        vm_size_t max_zonemap_size)
@@ -603,10 +1490,10 @@ zone_init(
        kern_return_t   retval;
        vm_offset_t     zone_min;
        vm_offset_t     zone_max;
        kern_return_t   retval;
        vm_offset_t     zone_min;
        vm_offset_t     zone_max;
-       vm_size_t       zone_table_size;
 
        retval = kmem_suballoc(kernel_map, &zone_min, max_zonemap_size,
 
        retval = kmem_suballoc(kernel_map, &zone_min, max_zonemap_size,
-                                               FALSE, VM_FLAGS_ANYWHERE, &zone_map);
+                              FALSE, VM_FLAGS_ANYWHERE | VM_FLAGS_PERMANENT,
+                              &zone_map);
 
        if (retval != KERN_SUCCESS)
                panic("zone_init: kmem_suballoc failed");
 
        if (retval != KERN_SUCCESS)
                panic("zone_init: kmem_suballoc failed");
@@ -614,19 +1501,104 @@ zone_init(
        /*
         * Setup garbage collection information:
         */
        /*
         * Setup garbage collection information:
         */
-       zone_table_size = atop_32(zone_max - zone_min) * 
-                               sizeof(struct zone_page_table_entry);
-       if (kmem_alloc_wired(zone_map, (vm_offset_t *) &zone_page_table,
-                            zone_table_size) != KERN_SUCCESS)
-               panic("zone_init");
-       zone_min = (vm_offset_t)zone_page_table + round_page(zone_table_size);
-       zone_pages = atop_32(zone_max - zone_min);
        zone_map_min_address = zone_min;
        zone_map_max_address = zone_max;
        zone_map_min_address = zone_min;
        zone_map_max_address = zone_max;
-       mutex_init(&zone_gc_lock, 0);
-       zone_page_init(zone_min, zone_max - zone_min, ZONE_PAGE_UNUSED);
+
+       zone_pages = (unsigned int)atop_kernel(zone_max - zone_min);
+       zone_page_table_used_size = sizeof(zone_page_table);
+
+       zone_page_table_second_level_size = 1;
+       zone_page_table_second_level_shift_amount = 0;
+       
+       /*
+        * Find the power of 2 for the second level that allows
+        * the first level to fit in ZONE_PAGE_TABLE_FIRST_LEVEL_SIZE
+        * slots.
+        */
+       while ((zone_page_table_first_level_slot(zone_pages-1)) >= ZONE_PAGE_TABLE_FIRST_LEVEL_SIZE) {
+               zone_page_table_second_level_size <<= 1;
+               zone_page_table_second_level_shift_amount++;
+       }
+       
+       lck_grp_attr_setdefault(&zone_lck_grp_attr);
+       lck_grp_init(&zone_lck_grp, "zones", &zone_lck_grp_attr);
+       lck_attr_setdefault(&zone_lck_attr);
+       lck_mtx_init_ext(&zone_gc_lock, &zone_lck_ext, &zone_lck_grp, &zone_lck_attr);
+       
+#if CONFIG_ZLEAKS
+       /*
+        * Initialize the zone leak monitor
+        */
+       zleak_init(max_zonemap_size);
+#endif /* CONFIG_ZLEAKS */
+}
+
+void
+zone_page_table_expand(zone_page_index_t pindex)
+{
+       unsigned int first_index;
+       struct zone_page_table_entry * volatile * first_level_ptr;
+
+       assert(pindex < zone_pages);
+
+       first_index = zone_page_table_first_level_slot(pindex);
+       first_level_ptr = &zone_page_table[first_index];
+
+       if (*first_level_ptr == NULL) {
+               /*
+                * We were able to verify the old first-level slot
+                * had NULL, so attempt to populate it.
+                */
+
+               vm_offset_t second_level_array = 0;
+               vm_size_t second_level_size = round_page(zone_page_table_second_level_size * sizeof(struct zone_page_table_entry));
+               zone_page_index_t i;
+               struct zone_page_table_entry *entry_array;
+
+               if (kmem_alloc_kobject(zone_map, &second_level_array,
+                                                          second_level_size) != KERN_SUCCESS) {
+                       panic("zone_page_table_expand");
+               }
+
+               /*
+                * zone_gc() may scan the "zone_page_table" directly,
+                * so make sure any slots have a valid unused state.
+                */
+               entry_array = (struct zone_page_table_entry *)second_level_array;
+               for (i=0; i < zone_page_table_second_level_size; i++) {
+                       entry_array[i].alloc_count = ZONE_PAGE_UNUSED;
+                       entry_array[i].collect_count = 0;
+               }
+
+               if (OSCompareAndSwapPtr(NULL, entry_array, first_level_ptr)) {
+                       /* Old slot was NULL, replaced with expanded level */
+                       OSAddAtomicLong(second_level_size, &zone_page_table_used_size);
+               } else {
+                       /* Old slot was not NULL, someone else expanded first */
+                       kmem_free(zone_map, second_level_array, second_level_size);
+               }
+       } else {
+               /* Old slot was not NULL, already been expanded */
+       }
+}
+
+struct zone_page_table_entry *
+zone_page_table_lookup(zone_page_index_t pindex)
+{
+       unsigned int first_index = zone_page_table_first_level_slot(pindex);
+       struct zone_page_table_entry *second_level = zone_page_table[first_index];
+
+       if (second_level) {
+               return &second_level[zone_page_table_second_level_slot(pindex)];
+       }
+
+       return NULL;
 }
 
 }
 
+extern volatile SInt32 kfree_nop_count;
+
+#pragma mark -
+#pragma mark zalloc_canblock
 
 /*
  *     zalloc returns an element from the specified zone.
 
 /*
  *     zalloc returns an element from the specified zone.
@@ -638,17 +1610,48 @@ zalloc_canblock(
 {
        vm_offset_t     addr;
        kern_return_t retval;
 {
        vm_offset_t     addr;
        kern_return_t retval;
+       uintptr_t       zbt[MAX_ZTRACE_DEPTH];  /* used in zone leak logging and zone leak detection */
+       int             numsaved = 0;
+       int                     i;
+       boolean_t       zone_replenish_wakeup = FALSE;
 
 
-       assert(zone != ZONE_NULL);
+#if CONFIG_ZLEAKS
+       uint32_t        zleak_tracedepth = 0;  /* log this allocation if nonzero */
+#endif /* CONFIG_ZLEAKS */
 
 
+       assert(zone != ZONE_NULL);
+       
        lock_zone(zone);
 
        lock_zone(zone);
 
+       /*
+        * If zone logging is turned on and this is the zone we're tracking, grab a backtrace.
+        */
+       
+       if (DO_LOGGING(zone))
+               numsaved = OSBacktrace((void*) zbt, MAX_ZTRACE_DEPTH);
+       
+#if CONFIG_ZLEAKS
+       /* 
+        * Zone leak detection: capture a backtrace every z_sample_factor
+        * allocations in this zone. 
+        */
+       if (zone->zleak_on && (zone->zleak_capture++ % z_sample_factor == 0)) {
+               zone->zleak_capture = 1;
+               
+               /* Avoid backtracing twice if zone logging is on */
+               if (numsaved == 0 )
+                       zleak_tracedepth = fastbacktrace(zbt, MAX_ZTRACE_DEPTH);
+               else
+                       zleak_tracedepth = numsaved;
+       }
+#endif /* CONFIG_ZLEAKS */
+
        REMOVE_FROM_ZONE(zone, addr, vm_offset_t);
 
        REMOVE_FROM_ZONE(zone, addr, vm_offset_t);
 
-       while ((addr == 0) && canblock && (zone->doing_gc)) {
-               zone->waiting = TRUE;
-               zone_sleep(zone);
-               REMOVE_FROM_ZONE(zone, addr, vm_offset_t);
+       if (zone->async_prio_refill &&
+       ((zone->cur_size - (zone->count * zone->elem_size)) < (zone->prio_refill_watermark * zone->elem_size))) {
+               zone_replenish_wakeup = TRUE;
+               zone_replenish_wakeups_initiated++;
        }
 
        while ((addr == 0) && canblock) {
        }
 
        while ((addr == 0) && canblock) {
@@ -662,8 +1665,20 @@ zalloc_canblock(
                         */
                        zone->waiting = TRUE;
                        zone_sleep(zone);
                         */
                        zone->waiting = TRUE;
                        zone_sleep(zone);
-               }
-               else {
+               } else if (zone->doing_gc) {
+                       /* zone_gc() is running. Since we need an element
+                        * from the free list that is currently being
+                        * collected, set the waiting bit and try to
+                        * interrupt the GC process, and try again
+                        * when we obtain the lock.
+                        */
+                       zone->waiting = TRUE;
+                       zone_sleep(zone);
+               } else {
+                       vm_offset_t space;
+                       vm_size_t alloc_size;
+                       int retry = 0;
+
                        if ((zone->cur_size + zone->elem_size) >
                            zone->max_size) {
                                if (zone->exhaustible)
                        if ((zone->cur_size + zone->elem_size) >
                            zone->max_size) {
                                if (zone->exhaustible)
@@ -689,92 +1704,175 @@ zalloc_canblock(
                        zone->doing_alloc = TRUE;
                        unlock_zone(zone);
 
                        zone->doing_alloc = TRUE;
                        unlock_zone(zone);
 
-                       if (zone->collectable) {
-                               vm_offset_t space;
-                               vm_size_t alloc_size;
-                               boolean_t retry = FALSE;
-
-                               for (;;) {
-
-                                       if (vm_pool_low() || retry == TRUE)
-                                               alloc_size = 
-                                                 round_page(zone->elem_size);
-                                       else
-                                               alloc_size = zone->alloc_size;
-
-                                       retval = kernel_memory_allocate(zone_map,
-                                                                       &space, alloc_size, 0,
-                                                                       KMA_KOBJECT|KMA_NOPAGEWAIT);
-                                       if (retval == KERN_SUCCESS) {
-                                               zone_page_init(space, alloc_size,
-                                                              ZONE_PAGE_USED);
-                                               zcram(zone, (void *)space, alloc_size);
-
-                                               break;
-                                       } else if (retval != KERN_RESOURCE_SHORTAGE) {
-                                               /* would like to cause a zone_gc() */
-                                               if (retry == TRUE)
-                                                       panic("zalloc: \"%s\" (%d elements) retry fail %d", zone->zone_name, zone->count, retval);
-                                               retry = TRUE;
-                                       } else {
-                                               break;
+                       for (;;) {
+                               int     zflags = KMA_KOBJECT|KMA_NOPAGEWAIT;
+                               
+                               if (vm_pool_low() || retry >= 1)
+                                       alloc_size = 
+                                               round_page(zone->elem_size);
+                               else
+                                       alloc_size = zone->alloc_size;
+                               
+                               if (zone->noencrypt)
+                                       zflags |= KMA_NOENCRYPT;
+                               
+                               retval = kernel_memory_allocate(zone_map, &space, alloc_size, 0, zflags);
+                               if (retval == KERN_SUCCESS) {
+#if    ZONE_ALIAS_ADDR
+                                       if (alloc_size == PAGE_SIZE)
+                                               space = zone_alias_addr(space);
+#endif
+                                       
+#if CONFIG_ZLEAKS
+                                       if ((zleak_state & (ZLEAK_STATE_ENABLED | ZLEAK_STATE_ACTIVE)) == ZLEAK_STATE_ENABLED) {
+                                               if (zone_map->size >= zleak_global_tracking_threshold) {
+                                                       kern_return_t kr;
+                                                       
+                                                       kr = zleak_activate();
+                                                       if (kr != KERN_SUCCESS) {
+                                                               printf("Failed to activate live zone leak debugging (%d).\n", kr);
+                                                       }
+                                               }
                                        }
                                        }
-                               }
-                               lock_zone(zone);
-                               zone->doing_alloc = FALSE; 
-                               if (zone->waiting) {
-                                       zone->waiting = FALSE;
-                                       zone_wakeup(zone);
-                               }
-                               REMOVE_FROM_ZONE(zone, addr, vm_offset_t);
-                               if (addr == 0 &&
-                                       retval == KERN_RESOURCE_SHORTAGE) {
-                                       unlock_zone(zone);
                                        
                                        
-                                       VM_PAGE_WAIT();
-                                       lock_zone(zone);
-                               }
-                       } else {
-                               vm_offset_t space;
-                               retval = zget_space(zone->elem_size, &space);
-
-                               lock_zone(zone);
-                               zone->doing_alloc = FALSE; 
-                               if (zone->waiting) {
-                                       zone->waiting = FALSE;
-                                       thread_wakeup((event_t)zone);
-                               }
-                               if (retval == KERN_SUCCESS) {
-                                       zone->count++;
-                                       zone->cur_size += zone->elem_size;
-#if    ZONE_DEBUG
-                                       if (zone_debug_enabled(zone)) {
-                                           enqueue_tail(&zone->active_zones, (queue_entry_t)space);
+                                       if ((zleak_state & ZLEAK_STATE_ACTIVE) && !(zone->zleak_on)) {
+                                               if (zone->cur_size > zleak_per_zone_tracking_threshold) {
+                                                       zone->zleak_on = TRUE;
+                                               }       
                                        }
                                        }
-#endif
-                                       unlock_zone(zone);
-                                       zone_page_alloc(space, zone->elem_size);
-#if    ZONE_DEBUG
-                                       if (zone_debug_enabled(zone))
-                                               space += ZONE_DEBUG_OFFSET;
-#endif
-                                       return((void *)space);
-                               }
-                               if (retval == KERN_RESOURCE_SHORTAGE) {
-                                       unlock_zone(zone);
+#endif /* CONFIG_ZLEAKS */
+                                       
+                                       zcram(zone, space, alloc_size);
+                                       
+                                       break;
+                               } else if (retval != KERN_RESOURCE_SHORTAGE) {
+                                       retry++;
                                        
                                        
-                                       VM_PAGE_WAIT();
-                                       lock_zone(zone);
+                                       if (retry == 2) {
+                                               zone_gc();
+                                               printf("zalloc did gc\n");
+                                               zone_display_zprint();
+                                       }
+                                       if (retry == 3) {
+                                               panic_include_zprint = TRUE;
+#if CONFIG_ZLEAKS
+                                               if ((zleak_state & ZLEAK_STATE_ACTIVE)) {
+                                                       panic_include_ztrace = TRUE;
+                                               }
+#endif /* CONFIG_ZLEAKS */             
+                                               /* TODO: Change this to something more descriptive, perhaps 
+                                                * 'zone_map exhausted' only if we get retval 3 (KERN_NO_SPACE).
+                                                */
+                                               panic("zalloc: \"%s\" (%d elements) retry fail %d, kfree_nop_count: %d", zone->zone_name, zone->count, retval, (int)kfree_nop_count);
+                                       }
                                } else {
                                } else {
-                                       panic("zalloc: \"%s\" (%d elements) zget_space returned %d", zone->zone_name, zone->count, retval);
+                                       break;
                                }
                        }
                                }
                        }
+                       lock_zone(zone);
+                       zone->doing_alloc = FALSE; 
+                       if (zone->waiting) {
+                               zone->waiting = FALSE;
+                               zone_wakeup(zone);
+                       }
+                       REMOVE_FROM_ZONE(zone, addr, vm_offset_t);
+                       if (addr == 0 &&
+                               retval == KERN_RESOURCE_SHORTAGE) {
+                               unlock_zone(zone);
+                               
+                               VM_PAGE_WAIT();
+                               lock_zone(zone);
+                       }
                }
                if (addr == 0)
                        REMOVE_FROM_ZONE(zone, addr, vm_offset_t);
        }
 
                }
                if (addr == 0)
                        REMOVE_FROM_ZONE(zone, addr, vm_offset_t);
        }
 
-       if ((addr == 0) && !canblock && (zone->async_pending == FALSE) && (!vm_pool_low())) {
+#if CONFIG_ZLEAKS
+       /* Zone leak detection:
+        * If we're sampling this allocation, add it to the zleaks hash table. 
+        */
+       if (addr && zleak_tracedepth > 0)  {
+               /* Sampling can fail if another sample is happening at the same time in a different zone. */
+               if (!zleak_log(zbt, addr, zleak_tracedepth, zone->elem_size)) {
+                       /* If it failed, roll back the counter so we sample the next allocation instead. */
+                       zone->zleak_capture = z_sample_factor;
+               }
+       }
+#endif /* CONFIG_ZLEAKS */                     
+                       
+                       
+       /*
+        * See if we should be logging allocations in this zone.  Logging is rarely done except when a leak is
+        * suspected, so this code rarely executes.  We need to do this code while still holding the zone lock
+        * since it protects the various log related data structures.
+        */
+
+       if (DO_LOGGING(zone) && addr) {
+
+               /*
+                * Look for a place to record this new allocation.  We implement two different logging strategies
+                * depending on whether we're looking for the source of a zone leak or a zone corruption.  When looking
+                * for a leak, we want to log as many allocations as possible in order to clearly identify the leaker
+                * among all the records.  So we look for an unused slot in the log and fill that in before overwriting
+                * an old entry.  When looking for a corrution however, it's better to have a chronological log of all
+                * the allocations and frees done in the zone so that the history of operations for a specific zone 
+                * element can be inspected.  So in this case, we treat the log as a circular buffer and overwrite the
+                * oldest entry whenever a new one needs to be added.
+                *
+                * The check_freed_element flag tells us what style of logging to do.  It's set if we're supposed to be
+                * doing corruption style logging (indicated via -zc in the boot-args).
+                */
+
+               if (!check_freed_element && zrecords[zcurrent].z_element && zrecorded < log_records) {
+
+                       /*
+                        * If we get here, we're doing leak style logging and there's still some unused entries in
+                        * the log (since zrecorded is smaller than the size of the log).  Look for an unused slot
+                        * starting at zcurrent and wrap-around if we reach the end of the buffer.  If the buffer
+                        * is already full, we just fall through and overwrite the element indexed by zcurrent.
+                        */
+       
+                      for (i = zcurrent; i < log_records; i++) {
+                               if (zrecords[i].z_element == NULL) {
+                                       zcurrent = i;
+                                       goto empty_slot;
+                               }
+                       }
+
+                       for (i = 0; i < zcurrent; i++) {
+                               if (zrecords[i].z_element == NULL) {
+                                       zcurrent = i;
+                                       goto empty_slot;
+                               }
+                       }
+                }
+       
+               /*
+                * Save a record of this allocation
+                */
+       
+empty_slot:
+                 if (zrecords[zcurrent].z_element == NULL)
+                       zrecorded++;
+       
+                 zrecords[zcurrent].z_element = (void *)addr;
+                 zrecords[zcurrent].z_time = ztime++;
+                 zrecords[zcurrent].z_opcode = ZOP_ALLOC;
+                       
+                 for (i = 0; i < numsaved; i++)
+                       zrecords[zcurrent].z_pc[i] = (void*) zbt[i];
+
+                 for (; i < MAX_ZTRACE_DEPTH; i++)
+                       zrecords[zcurrent].z_pc[i] = 0;
+       
+                 zcurrent++;
+       
+                 if (zcurrent >= log_records)
+                         zcurrent = 0;
+       }
+
+       if ((addr == 0) && !canblock && (zone->async_pending == FALSE) && (zone->no_callout == FALSE) && (zone->exhaustible == FALSE) && (!vm_pool_low())) {
                zone->async_pending = TRUE;
                unlock_zone(zone);
                thread_call_enter(&zone->call_async_alloc);
                zone->async_pending = TRUE;
                unlock_zone(zone);
                thread_call_enter(&zone->call_async_alloc);
@@ -788,9 +1886,33 @@ zalloc_canblock(
                addr += ZONE_DEBUG_OFFSET;
        }
 #endif
                addr += ZONE_DEBUG_OFFSET;
        }
 #endif
+       
+#if CONFIG_ZLEAKS
+       if (addr != 0) {
+               zone->num_allocs++;
+       }
+#endif /* CONFIG_ZLEAKS */
 
        unlock_zone(zone);
 
 
        unlock_zone(zone);
 
+       if (zone_replenish_wakeup)
+               thread_wakeup(&zone->zone_replenish_thread);
+
+       TRACE_MACHLEAKS(ZALLOC_CODE, ZALLOC_CODE_2, zone->elem_size, addr);
+
+       if (addr) {
+               thread_t thr = current_thread();
+               task_t task;
+               zinfo_usage_t zinfo;
+
+               if (zone->caller_acct)
+                       thr->tkm_private.alloc += zone->elem_size;
+               else
+                       thr->tkm_shared.alloc += zone->elem_size;
+
+               if ((task = thr->task) != NULL && (zinfo = task->tkm_zinfo) != NULL)
+                       OSAddAtomic64(zone->elem_size, (int64_t *)&zinfo[zone->index].alloc);
+       }
        return((void *)addr);
 }
 
        return((void *)addr);
 }
 
@@ -830,17 +1952,36 @@ zalloc_async(
  *
  *     This form should be used when you can not block (like when
  *     processing an interrupt).
  *
  *     This form should be used when you can not block (like when
  *     processing an interrupt).
+ *
+ *     XXX: It seems like only vm_page_grab_fictitious_common uses this, and its
+ *  friend vm_page_more_fictitious can block, so it doesn't seem like 
+ *  this is used for interrupts any more....
  */
 void *
 zget(
        register zone_t zone)
 {
        register vm_offset_t    addr;
  */
 void *
 zget(
        register zone_t zone)
 {
        register vm_offset_t    addr;
+       
+#if CONFIG_ZLEAKS
+       uintptr_t       zbt[MAX_ZTRACE_DEPTH];          /* used for zone leak detection */
+       uint32_t        zleak_tracedepth = 0;  /* log this allocation if nonzero */
+#endif /* CONFIG_ZLEAKS */
 
        assert( zone != ZONE_NULL );
 
        if (!lock_try_zone(zone))
                return NULL;
 
        assert( zone != ZONE_NULL );
 
        if (!lock_try_zone(zone))
                return NULL;
+       
+#if CONFIG_ZLEAKS
+       /*
+        * Zone leak detection: capture a backtrace
+        */
+       if (zone->zleak_on && (zone->zleak_capture++ % z_sample_factor == 0)) {
+               zone->zleak_capture = 1;
+               zleak_tracedepth = fastbacktrace(zbt, MAX_ZTRACE_DEPTH);
+       }
+#endif /* CONFIG_ZLEAKS */
 
        REMOVE_FROM_ZONE(zone, addr, vm_offset_t);
 #if    ZONE_DEBUG
 
        REMOVE_FROM_ZONE(zone, addr, vm_offset_t);
 #if    ZONE_DEBUG
@@ -849,6 +1990,24 @@ zget(
                addr += ZONE_DEBUG_OFFSET;
        }
 #endif /* ZONE_DEBUG */
                addr += ZONE_DEBUG_OFFSET;
        }
 #endif /* ZONE_DEBUG */
+       
+#if CONFIG_ZLEAKS
+       /*
+        * Zone leak detection: record the allocation 
+        */
+       if (zone->zleak_on && zleak_tracedepth > 0 && addr) {
+               /* Sampling can fail if another sample is happening at the same time in a different zone. */
+               if (!zleak_log(zbt, addr, zleak_tracedepth, zone->elem_size)) {
+                       /* If it failed, roll back the counter so we sample the next allocation instead. */
+                       zone->zleak_capture = z_sample_factor;
+               }
+       }
+       
+       if (addr != 0) {
+               zone->num_allocs++;
+       }
+#endif /* CONFIG_ZLEAKS */
+       
        unlock_zone(zone);
 
        return((void *) addr);
        unlock_zone(zone);
 
        return((void *) addr);
@@ -867,6 +2026,17 @@ zfree(
        void            *addr)
 {
        vm_offset_t     elem = (vm_offset_t) addr;
        void            *addr)
 {
        vm_offset_t     elem = (vm_offset_t) addr;
+       void            *zbt[MAX_ZTRACE_DEPTH];                 /* only used if zone logging is enabled via boot-args */
+       int             numsaved = 0;
+
+       assert(zone != ZONE_NULL);
+
+       /*
+        * If zone logging is turned on and this is the zone we're tracking, grab a backtrace.
+        */
+
+       if (DO_LOGGING(zone))
+               numsaved = OSBacktrace(&zbt[0], MAX_ZTRACE_DEPTH);
 
 #if MACH_ASSERT
        /* Basic sanity checks */
 
 #if MACH_ASSERT
        /* Basic sanity checks */
@@ -877,17 +2047,74 @@ zfree(
                panic("zfree: freeing to zone_zone breaks zone_gc!");
 #endif
 
                panic("zfree: freeing to zone_zone breaks zone_gc!");
 #endif
 
-       if (zone->collectable && !zone->allows_foreign &&
-           !from_zone_map(elem, zone->elem_size)) {
-#if MACH_ASSERT
-               panic("zfree: non-allocated memory in collectable zone!");
-#endif
-               zone_last_bogus_zone = zone;
-               zone_last_bogus_elem = elem;
-               return;
+       TRACE_MACHLEAKS(ZFREE_CODE, ZFREE_CODE_2, zone->elem_size, (uintptr_t)addr);
+
+       if (zone->collectable && !zone->allows_foreign &&
+           !from_zone_map(elem, zone->elem_size)) {
+#if MACH_ASSERT
+               panic("zfree: non-allocated memory in collectable zone!");
+#endif
+               zone_last_bogus_zone = zone;
+               zone_last_bogus_elem = elem;
+               return;
+       }
+
+       lock_zone(zone);
+
+       /*
+        * See if we're doing logging on this zone.  There are two styles of logging used depending on
+        * whether we're trying to catch a leak or corruption.  See comments above in zalloc for details.
+        */
+
+       if (DO_LOGGING(zone)) {
+               int  i;
+
+               if (check_freed_element) {
+
+                       /*
+                        * We're logging to catch a corruption.  Add a record of this zfree operation
+                        * to log.
+                        */
+
+                       if (zrecords[zcurrent].z_element == NULL)
+                               zrecorded++;
+
+                       zrecords[zcurrent].z_element = (void *)addr;
+                       zrecords[zcurrent].z_time = ztime++;
+                       zrecords[zcurrent].z_opcode = ZOP_FREE;
+
+                       for (i = 0; i < numsaved; i++)
+                               zrecords[zcurrent].z_pc[i] = zbt[i];
+
+                       for (; i < MAX_ZTRACE_DEPTH; i++)
+                               zrecords[zcurrent].z_pc[i] = 0;
+
+                       zcurrent++;
+
+                       if (zcurrent >= log_records)
+                               zcurrent = 0;
+
+               } else {
+
+                       /*
+                        * We're logging to catch a leak. Remove any record we might have for this
+                        * element since it's being freed.  Note that we may not find it if the buffer
+                        * overflowed and that's OK.  Since the log is of a limited size, old records
+                        * get overwritten if there are more zallocs than zfrees.
+                        */
+       
+                       for (i = 0; i < log_records; i++) {
+                               if (zrecords[i].z_element == addr) {
+                                       zrecords[i].z_element = NULL;
+                                       zcurrent = i;
+                                       zrecorded--;
+                                       break;
+                               }
+                       }
+               }
        }
 
        }
 
-       lock_zone(zone);
+
 #if    ZONE_DEBUG
        if (zone_debug_enabled(zone)) {
                queue_t tmp_elem;
 #if    ZONE_DEBUG
        if (zone_debug_enabled(zone)) {
                queue_t tmp_elem;
@@ -904,7 +2131,7 @@ zfree(
                        if (elem != (vm_offset_t)tmp_elem)
                                panic("zfree()ing element from wrong zone");
                }
                        if (elem != (vm_offset_t)tmp_elem)
                                panic("zfree()ing element from wrong zone");
                }
-               remqueue(&zone->active_zones, (queue_t) elem);
+               remqueue((queue_t) elem);
        }
 #endif /* ZONE_DEBUG */
        if (zone_check) {
        }
 #endif /* ZONE_DEBUG */
        if (zone_check) {
@@ -919,7 +2146,23 @@ zfree(
                                panic("zfree");
        }
        ADD_TO_ZONE(zone, elem);
                                panic("zfree");
        }
        ADD_TO_ZONE(zone, elem);
+#if MACH_ASSERT
+       if (zone->count < 0)
+               panic("zfree: count < 0!");
+#endif
+       
+
+#if CONFIG_ZLEAKS
+       zone->num_frees++;
 
 
+       /*
+        * Zone leak detection: un-track the allocation 
+        */
+       if (zone->zleak_on) {
+               zleak_free(elem, zone->elem_size);
+       }
+#endif /* CONFIG_ZLEAKS */
+       
        /*
         * If elements have one or more pages, and memory is low,
         * request to run the garbage collection in the zone  the next 
        /*
         * If elements have one or more pages, and memory is low,
         * request to run the garbage collection in the zone  the next 
@@ -930,6 +2173,20 @@ zfree(
                zone_gc_forced = TRUE;
        }
        unlock_zone(zone);
                zone_gc_forced = TRUE;
        }
        unlock_zone(zone);
+
+       {
+               thread_t thr = current_thread();
+               task_t task;
+               zinfo_usage_t zinfo;
+
+               if (zone->caller_acct)
+                       thr->tkm_private.free += zone->elem_size;
+               else
+                       thr->tkm_shared.free += zone->elem_size;
+               if ((task = thr->task) != NULL && (zinfo = task->tkm_zinfo) != NULL)
+                       OSAddAtomic64(zone->elem_size,
+                                     (int64_t *)&zinfo[zone->index].free);
+       }
 }
 
 
 }
 
 
@@ -946,6 +2203,9 @@ zone_change(
        assert( value == TRUE || value == FALSE );
 
        switch(item){
        assert( value == TRUE || value == FALSE );
 
        switch(item){
+               case Z_NOENCRYPT:
+                       zone->noencrypt = value;
+                       break;
                case Z_EXHAUST:
                        zone->exhaustible = value;
                        break;
                case Z_EXHAUST:
                        zone->exhaustible = value;
                        break;
@@ -958,13 +2218,18 @@ zone_change(
                case Z_FOREIGN:
                        zone->allows_foreign = value;
                        break;
                case Z_FOREIGN:
                        zone->allows_foreign = value;
                        break;
+               case Z_CALLERACCT:
+                       zone->caller_acct = value;
+                       break;
+               case Z_NOCALLOUT:
+                       zone->no_callout = value;
+                       break;
 #if MACH_ASSERT
                default:
                        panic("Zone_change: Wrong Item Type!");
                        /* break; */
 #endif
        }
 #if MACH_ASSERT
                default:
                        panic("Zone_change: Wrong Item Type!");
                        /* break; */
 #endif
        }
-       lock_zone_init(zone);
 }
 
 /*
 }
 
 /*
@@ -980,7 +2245,7 @@ zone_free_count(zone_t zone)
        integer_t free_count;
 
        lock_zone(zone);
        integer_t free_count;
 
        lock_zone(zone);
-       free_count = zone->cur_size/zone->elem_size - zone->count;
+       free_count = (integer_t)(zone->cur_size/zone->elem_size - zone->count);
        unlock_zone(zone);
 
        assert(free_count >= 0);
        unlock_zone(zone);
 
        assert(free_count >= 0);
@@ -1000,10 +2265,9 @@ zprealloc(
         vm_offset_t addr;
 
        if (size != 0) {
         vm_offset_t addr;
 
        if (size != 0) {
-               if (kmem_alloc_wired(zone_map, &addr, size) != KERN_SUCCESS)
+               if (kmem_alloc_kobject(zone_map, &addr, size) != KERN_SUCCESS)
                  panic("zprealloc");
                  panic("zprealloc");
-               zone_page_init(addr, size, ZONE_PAGE_USED);
-               zcram(zone, (void *)addr, size);
+               zcram(zone, addr, size);
        }
 }
 
        }
 }
 
@@ -1017,19 +2281,24 @@ zone_page_collectable(
        vm_size_t       size)
 {
        struct zone_page_table_entry    *zp;
        vm_size_t       size)
 {
        struct zone_page_table_entry    *zp;
-       natural_t i, j;
+       zone_page_index_t i, j;
 
 
+#if    ZONE_ALIAS_ADDR
+       addr = zone_virtual_addr(addr);
+#endif
 #if MACH_ASSERT
        if (!from_zone_map(addr, size))
                panic("zone_page_collectable");
 #endif
 
 #if MACH_ASSERT
        if (!from_zone_map(addr, size))
                panic("zone_page_collectable");
 #endif
 
-       i = atop_32(addr-zone_map_min_address);
-       j = atop_32((addr+size-1) - zone_map_min_address);
+       i = (zone_page_index_t)atop_kernel(addr-zone_map_min_address);
+       j = (zone_page_index_t)atop_kernel((addr+size-1) - zone_map_min_address);
 
 
-       for (zp = zone_page_table + i; i <= j; zp++, i++)
+       for (; i <= j; i++) {
+               zp = zone_page_table_lookup(i);
                if (zp->collect_count == zp->alloc_count)
                        return (TRUE);
                if (zp->collect_count == zp->alloc_count)
                        return (TRUE);
+       }
 
        return (FALSE);
 }
 
        return (FALSE);
 }
@@ -1040,18 +2309,23 @@ zone_page_keep(
        vm_size_t       size)
 {
        struct zone_page_table_entry    *zp;
        vm_size_t       size)
 {
        struct zone_page_table_entry    *zp;
-       natural_t i, j;
+       zone_page_index_t i, j;
 
 
+#if    ZONE_ALIAS_ADDR
+       addr = zone_virtual_addr(addr);
+#endif
 #if MACH_ASSERT
        if (!from_zone_map(addr, size))
                panic("zone_page_keep");
 #endif
 
 #if MACH_ASSERT
        if (!from_zone_map(addr, size))
                panic("zone_page_keep");
 #endif
 
-       i = atop_32(addr-zone_map_min_address);
-       j = atop_32((addr+size-1) - zone_map_min_address);
+       i = (zone_page_index_t)atop_kernel(addr-zone_map_min_address);
+       j = (zone_page_index_t)atop_kernel((addr+size-1) - zone_map_min_address);
 
 
-       for (zp = zone_page_table + i; i <= j; zp++, i++)
+       for (; i <= j; i++) {
+               zp = zone_page_table_lookup(i);
                zp->collect_count = 0;
                zp->collect_count = 0;
+       }
 }
 
 void
 }
 
 void
@@ -1060,39 +2334,51 @@ zone_page_collect(
        vm_size_t       size)
 {
        struct zone_page_table_entry    *zp;
        vm_size_t       size)
 {
        struct zone_page_table_entry    *zp;
-       natural_t i, j;
+       zone_page_index_t i, j;
 
 
+#if    ZONE_ALIAS_ADDR
+       addr = zone_virtual_addr(addr);
+#endif
 #if MACH_ASSERT
        if (!from_zone_map(addr, size))
                panic("zone_page_collect");
 #endif
 
 #if MACH_ASSERT
        if (!from_zone_map(addr, size))
                panic("zone_page_collect");
 #endif
 
-       i = atop_32(addr-zone_map_min_address);
-       j = atop_32((addr+size-1) - zone_map_min_address);
+       i = (zone_page_index_t)atop_kernel(addr-zone_map_min_address);
+       j = (zone_page_index_t)atop_kernel((addr+size-1) - zone_map_min_address);
 
 
-       for (zp = zone_page_table + i; i <= j; zp++, i++)
+       for (; i <= j; i++) {
+               zp = zone_page_table_lookup(i);
                ++zp->collect_count;
                ++zp->collect_count;
+       }
 }
 
 void
 zone_page_init(
        vm_offset_t     addr,
 }
 
 void
 zone_page_init(
        vm_offset_t     addr,
-       vm_size_t       size,
-       int             value)
+       vm_size_t       size)
 {
        struct zone_page_table_entry    *zp;
 {
        struct zone_page_table_entry    *zp;
-       natural_t i, j;
+       zone_page_index_t i, j;
 
 
+#if    ZONE_ALIAS_ADDR
+       addr = zone_virtual_addr(addr);
+#endif
 #if MACH_ASSERT
        if (!from_zone_map(addr, size))
                panic("zone_page_init");
 #endif
 
 #if MACH_ASSERT
        if (!from_zone_map(addr, size))
                panic("zone_page_init");
 #endif
 
-       i = atop_32(addr-zone_map_min_address);
-       j = atop_32((addr+size-1) - zone_map_min_address);
+       i = (zone_page_index_t)atop_kernel(addr-zone_map_min_address);
+       j = (zone_page_index_t)atop_kernel((addr+size-1) - zone_map_min_address);
 
 
-       for (zp = zone_page_table + i; i <= j; zp++, i++) {
-               zp->alloc_count = value;
+       for (; i <= j; i++) {
+               /* make sure entry exists before marking unused */
+               zone_page_table_expand(i);
+
+               zp = zone_page_table_lookup(i);
+               assert(zp);
+               zp->alloc_count = ZONE_PAGE_UNUSED;
                zp->collect_count = 0;
        }
 }
                zp->collect_count = 0;
        }
 }
@@ -1103,54 +2389,73 @@ zone_page_alloc(
        vm_size_t       size)
 {
        struct zone_page_table_entry    *zp;
        vm_size_t       size)
 {
        struct zone_page_table_entry    *zp;
-       natural_t i, j;
+       zone_page_index_t i, j;
 
 
+#if    ZONE_ALIAS_ADDR
+       addr = zone_virtual_addr(addr);
+#endif
 #if MACH_ASSERT
        if (!from_zone_map(addr, size))
                panic("zone_page_alloc");
 #endif
 
 #if MACH_ASSERT
        if (!from_zone_map(addr, size))
                panic("zone_page_alloc");
 #endif
 
-       i = atop_32(addr-zone_map_min_address);
-       j = atop_32((addr+size-1) - zone_map_min_address);
+       i = (zone_page_index_t)atop_kernel(addr-zone_map_min_address);
+       j = (zone_page_index_t)atop_kernel((addr+size-1) - zone_map_min_address);
+
+       for (; i <= j; i++) {
+               zp = zone_page_table_lookup(i);
+               assert(zp);
 
 
-       for (zp = zone_page_table + i; i <= j; zp++, i++) {
                /*
                /*
-                * Set alloc_count to (ZONE_PAGE_USED + 1) if
+                * Set alloc_count to ZONE_PAGE_USED if
                 * it was previously set to ZONE_PAGE_UNUSED.
                 */
                if (zp->alloc_count == ZONE_PAGE_UNUSED)
                 * it was previously set to ZONE_PAGE_UNUSED.
                 */
                if (zp->alloc_count == ZONE_PAGE_UNUSED)
-                       zp->alloc_count = 1;
-               else
-                       ++zp->alloc_count;
+                       zp->alloc_count = ZONE_PAGE_USED;
+
+               ++zp->alloc_count;
        }
 }
 
 void
 zone_page_free_element(
        }
 }
 
 void
 zone_page_free_element(
-       struct zone_page_table_entry    **free_pages,
+       zone_page_index_t       *free_page_list,
        vm_offset_t     addr,
        vm_size_t       size)
 {
        struct zone_page_table_entry    *zp;
        vm_offset_t     addr,
        vm_size_t       size)
 {
        struct zone_page_table_entry    *zp;
-       natural_t i, j;
+       zone_page_index_t i, j;
 
 
+#if    ZONE_ALIAS_ADDR
+       addr = zone_virtual_addr(addr);
+#endif
 #if MACH_ASSERT
        if (!from_zone_map(addr, size))
                panic("zone_page_free_element");
 #endif
 
 #if MACH_ASSERT
        if (!from_zone_map(addr, size))
                panic("zone_page_free_element");
 #endif
 
-       i = atop_32(addr-zone_map_min_address);
-       j = atop_32((addr+size-1) - zone_map_min_address);
+       i = (zone_page_index_t)atop_kernel(addr-zone_map_min_address);
+       j = (zone_page_index_t)atop_kernel((addr+size-1) - zone_map_min_address);
+
+       for (; i <= j; i++) {
+               zp = zone_page_table_lookup(i);
 
 
-       for (zp = zone_page_table + i; i <= j; zp++, i++) {
                if (zp->collect_count > 0)
                        --zp->collect_count;
                if (--zp->alloc_count == 0) {
                if (zp->collect_count > 0)
                        --zp->collect_count;
                if (--zp->alloc_count == 0) {
+                       vm_address_t        free_page_address;
+
                        zp->alloc_count  = ZONE_PAGE_UNUSED;
                        zp->collect_count = 0;
 
                        zp->alloc_count  = ZONE_PAGE_UNUSED;
                        zp->collect_count = 0;
 
-                       zp->link = *free_pages;
-                       *free_pages = zp;
+
+                       /*
+                        * This element was the last one on this page, re-use the page's
+                        * storage for a page freelist
+                        */
+                       free_page_address = zone_map_min_address + PAGE_SIZE * ((vm_size_t)i);
+                       *(zone_page_index_t *)free_page_address = *free_page_list;
+                       *free_page_list = i;
                }
        }
 }
                }
        }
 }
@@ -1162,6 +2467,36 @@ struct zone_free_element {
        struct zone_free_element * next;
 };
 
        struct zone_free_element * next;
 };
 
+/*
+ * Add a linked list of pages starting at base back into the zone
+ * free list. Tail points to the last element on the list.
+ */
+
+#define ADD_LIST_TO_ZONE(zone, base, tail)                             \
+MACRO_BEGIN                                                            \
+       (tail)->next = (void *)((zone)->free_elements);                 \
+       if (check_freed_element) {                                      \
+               if ((zone)->elem_size >= (2 * sizeof(vm_offset_t)))     \
+                       ((vm_offset_t *)(tail))[((zone)->elem_size/sizeof(vm_offset_t))-1] = \
+                                        (zone)->free_elements;         \
+       }                                                               \
+       (zone)->free_elements = (unsigned long)(base);                  \
+MACRO_END
+
+/*
+ * Add an element to the chain pointed to by prev.
+ */
+
+#define ADD_ELEMENT(zone, prev, elem)                                          \
+MACRO_BEGIN                                                            \
+       (prev)->next = (elem);                                          \
+       if (check_freed_element) {                                      \
+               if ((zone)->elem_size >= (2 * sizeof(vm_offset_t)))     \
+                       ((vm_offset_t *)(prev))[((zone)->elem_size/sizeof(vm_offset_t))-1] = \
+                                       (vm_offset_t)(elem);            \
+        }                                                              \
+MACRO_END
+
 struct {
        uint32_t        pgs_freed;
 
 struct {
        uint32_t        pgs_freed;
 
@@ -1183,9 +2518,9 @@ zone_gc(void)
        unsigned int    max_zones;
        zone_t                  z;
        unsigned int    i;
        unsigned int    max_zones;
        zone_t                  z;
        unsigned int    i;
-       struct zone_page_table_entry    *zp, *zone_free_pages;
+       zone_page_index_t zone_free_page_head;
 
 
-       mutex_lock(&zone_gc_lock);
+       lck_mtx_lock(&zone_gc_lock);
 
        simple_lock(&all_zones_lock);
        max_zones = num_zones;
 
        simple_lock(&all_zones_lock);
        max_zones = num_zones;
@@ -1193,11 +2528,15 @@ zone_gc(void)
        simple_unlock(&all_zones_lock);
 
 #if MACH_ASSERT
        simple_unlock(&all_zones_lock);
 
 #if MACH_ASSERT
-       for (i = 0; i < zone_pages; i++)
-               assert(zone_page_table[i].collect_count == 0);
+       for (i = 0; i < zone_pages; i++) {
+               struct zone_page_table_entry    *zp;
+       
+               zp = zone_page_table_lookup(i);
+               assert(!zp || (zp->collect_count == 0));
+       }
 #endif /* MACH_ASSERT */
 
 #endif /* MACH_ASSERT */
 
-       zone_free_pages = NULL;
+       zone_free_page_head = ZONE_PAGE_INDEX_INVALID;
 
        for (i = 0; i < max_zones; i++, z = z->next_zone) {
                unsigned int                            n, m;
 
        for (i = 0; i < max_zones; i++, z = z->next_zone) {
                unsigned int                            n, m;
@@ -1219,9 +2558,11 @@ zone_gc(void)
                 * (i.e we need a whole allocation block's worth of free
                 * elements before we can garbage collect) and
                 * the zone has more than 10 percent of it's elements free
                 * (i.e we need a whole allocation block's worth of free
                 * elements before we can garbage collect) and
                 * the zone has more than 10 percent of it's elements free
+                * or the element size is a multiple of the PAGE_SIZE 
                 */
                 */
-               if (((z->cur_size - z->count * elt_size) <= (2 * z->alloc_size)) ||
-                   ((z->cur_size - z->count * elt_size) <= (z->cur_size / 10))) {
+               if ((elt_size & PAGE_MASK) && 
+                    (((z->cur_size - z->count * elt_size) <= (2 * z->alloc_size)) ||
+                     ((z->cur_size - z->count * elt_size) <= (z->cur_size / 10)))) {
                        unlock_zone(z);         
                        continue;
                }
                        unlock_zone(z);         
                        continue;
                }
@@ -1260,11 +2601,14 @@ zone_gc(void)
                        else {
                                if (keep == NULL)
                                        keep = tail = elt;
                        else {
                                if (keep == NULL)
                                        keep = tail = elt;
-                               else
-                                       tail = tail->next = elt;
+                               else {
+                                       ADD_ELEMENT(z, tail, elt);
+                                       tail = elt;
+                               }
 
 
-                               elt = prev->next = elt->next;
-                               tail->next = NULL;
+                               ADD_ELEMENT(z, prev, elt->next);
+                               elt = elt->next;
+                               ADD_ELEMENT(z, tail, NULL);
                        }
 
                        /*
                        }
 
                        /*
@@ -1273,11 +2617,11 @@ zone_gc(void)
 
                        if (++n >= 50) {
                                if (z->waiting == TRUE) {
 
                        if (++n >= 50) {
                                if (z->waiting == TRUE) {
+                                       /* z->waiting checked without lock held, rechecked below after locking */
                                        lock_zone(z);
 
                                        if (keep != NULL) {
                                        lock_zone(z);
 
                                        if (keep != NULL) {
-                                               tail->next = (void *)z->free_elements;
-                                               z->free_elements = (vm_offset_t) keep;
+                                               ADD_LIST_TO_ZONE(z, keep, tail);
                                                tail = keep = NULL;
                                        } else {
                                                m =0;
                                                tail = keep = NULL;
                                        } else {
                                                m =0;
@@ -1288,9 +2632,8 @@ zone_gc(void)
                                                        elt = elt->next;
                                                }
                                                if (m !=0 ) {
                                                        elt = elt->next;
                                                }
                                                if (m !=0 ) {
-                                                       prev->next = (void *)z->free_elements;
-                                                       z->free_elements = (vm_offset_t) base_elt;
-                                                       base_prev->next = elt;
+                                                       ADD_LIST_TO_ZONE(z, base_elt, prev);
+                                                       ADD_ELEMENT(z, base_prev, elt);
                                                        prev = base_prev;
                                                }
                                        }
                                                        prev = base_prev;
                                                }
                                        }
@@ -1313,8 +2656,12 @@ zone_gc(void)
                if (keep != NULL) {
                        lock_zone(z);
 
                if (keep != NULL) {
                        lock_zone(z);
 
-                       tail->next = (void *)z->free_elements;
-                       z->free_elements = (vm_offset_t) keep;
+                       ADD_LIST_TO_ZONE(z, keep, tail);
+
+                       if (z->waiting) {
+                               z->waiting = FALSE;
+                               zone_wakeup(z);
+                       }
 
                        unlock_zone(z);
                }
 
                        unlock_zone(z);
                }
@@ -1327,16 +2674,24 @@ zone_gc(void)
                 */
 
                size_freed = 0;
                 */
 
                size_freed = 0;
-               prev = (void *)&scan;
                elt = scan;
                n = 0; tail = keep = NULL;
                while (elt != NULL) {
                        if (zone_page_collectable((vm_offset_t)elt, elt_size)) {
                elt = scan;
                n = 0; tail = keep = NULL;
                while (elt != NULL) {
                        if (zone_page_collectable((vm_offset_t)elt, elt_size)) {
+                               struct zone_free_element *next_elt = elt->next;
+
                                size_freed += elt_size;
                                size_freed += elt_size;
-                               zone_page_free_element(&zone_free_pages,
+
+                               /*
+                                * If this is the last allocation on the page(s),
+                                * we may use their storage to maintain the linked
+                                * list of free-able pages. So store elt->next because
+                                * "elt" may be scribbled over.
+                                */
+                               zone_page_free_element(&zone_free_page_head,
                                                                                (vm_offset_t)elt, elt_size);
 
                                                                                (vm_offset_t)elt, elt_size);
 
-                               elt = prev->next = elt->next;
+                               elt = next_elt;
 
                                ++zgc_stats.elems_freed;
                        }
 
                                ++zgc_stats.elems_freed;
                        }
@@ -1345,11 +2700,13 @@ zone_gc(void)
 
                                if (keep == NULL)
                                        keep = tail = elt;
 
                                if (keep == NULL)
                                        keep = tail = elt;
-                               else
-                                       tail = tail->next = elt;
+                               else {
+                                       ADD_ELEMENT(z, tail, elt);
+                                       tail = elt;
+                               }
 
 
-                               elt = prev->next = elt->next;
-                               tail->next = NULL;
+                               elt = elt->next;
+                               ADD_ELEMENT(z, tail, NULL);
 
                                ++zgc_stats.elems_kept;
                        }
 
                                ++zgc_stats.elems_kept;
                        }
@@ -1366,8 +2723,7 @@ zone_gc(void)
                                size_freed = 0;
 
                                if (keep != NULL) {
                                size_freed = 0;
 
                                if (keep != NULL) {
-                                       tail->next = (void *)z->free_elements;
-                                       z->free_elements = (vm_offset_t) keep;
+                                       ADD_LIST_TO_ZONE(z, keep, tail);
                                }
 
                                if (z->waiting) {
                                }
 
                                if (z->waiting) {
@@ -1393,8 +2749,7 @@ zone_gc(void)
                        z->cur_size -= size_freed;
 
                        if (keep != NULL) {
                        z->cur_size -= size_freed;
 
                        if (keep != NULL) {
-                               tail->next = (void *)z->free_elements;
-                               z->free_elements = (vm_offset_t) keep;
+                               ADD_LIST_TO_ZONE(z, keep, tail);
                        }
 
                }
                        }
 
                }
@@ -1411,14 +2766,21 @@ zone_gc(void)
         * Reclaim the pages we are freeing.
         */
 
         * Reclaim the pages we are freeing.
         */
 
-       while ((zp = zone_free_pages) != NULL) {
-               zone_free_pages = zp->link;
-               kmem_free(zone_map, zone_map_min_address + PAGE_SIZE *
-                                                                               (zp - zone_page_table), PAGE_SIZE);
+       while (zone_free_page_head != ZONE_PAGE_INDEX_INVALID) {
+               zone_page_index_t       zind = zone_free_page_head;
+               vm_address_t            free_page_address;
+#if    ZONE_ALIAS_ADDR
+               z = (zone_t)zone_virtual_addr((vm_map_address_t)z);
+#endif
+               /* Use the first word of the page about to be freed to find the next free page */
+               free_page_address = zone_map_min_address + PAGE_SIZE * ((vm_size_t)zind);
+               zone_free_page_head = *(zone_page_index_t *)free_page_address;
+
+               kmem_free(zone_map, free_page_address, PAGE_SIZE);
                ++zgc_stats.pgs_freed;
        }
 
                ++zgc_stats.pgs_freed;
        }
 
-       mutex_unlock(&zone_gc_lock);
+       lck_mtx_unlock(&zone_gc_lock);
 }
 
 /*
 }
 
 /*
@@ -1428,26 +2790,336 @@ zone_gc(void)
  */
 
 void
  */
 
 void
-consider_zone_gc(void)
+consider_zone_gc(boolean_t force)
 {
 {
-       /*
-        *      By default, don't attempt zone GC more frequently
-        *      than once / 1 minutes.
-        */
-
-       if (zone_gc_max_rate == 0)
-               zone_gc_max_rate = (60 << SCHED_TICK_SHIFT) + 1;
 
        if (zone_gc_allowed &&
 
        if (zone_gc_allowed &&
-           ((sched_tick > (zone_gc_last_tick + zone_gc_max_rate)) ||
-            zone_gc_forced)) {
+           (zone_gc_allowed_by_time_throttle ||
+            zone_gc_forced ||
+            force)) {
                zone_gc_forced = FALSE;
                zone_gc_forced = FALSE;
-               zone_gc_last_tick = sched_tick;
+               zone_gc_allowed_by_time_throttle = FALSE; /* reset periodically */
                zone_gc();
        }
 }
 
                zone_gc();
        }
 }
 
+/*
+ *     By default, don't attempt zone GC more frequently
+ *     than once / 1 minutes.
+ */
+void
+compute_zone_gc_throttle(void *arg __unused)
+{
+       zone_gc_allowed_by_time_throttle = TRUE;
+}
+
+
+kern_return_t
+task_zone_info(
+       task_t                  task,
+       mach_zone_name_array_t  *namesp,
+       mach_msg_type_number_t  *namesCntp,
+       task_zone_info_array_t  *infop,
+       mach_msg_type_number_t  *infoCntp)
+{
+       mach_zone_name_t        *names;
+       vm_offset_t             names_addr;
+       vm_size_t               names_size;
+       task_zone_info_t        *info;
+       vm_offset_t             info_addr;
+       vm_size_t               info_size;
+       unsigned int            max_zones, i;
+       zone_t                  z;
+       mach_zone_name_t        *zn;
+       task_zone_info_t        *zi;
+       kern_return_t           kr;
+
+       vm_size_t               used;
+       vm_map_copy_t           copy;
+
+
+       if (task == TASK_NULL)
+               return KERN_INVALID_TASK;
+
+       /*
+        *      We assume that zones aren't freed once allocated.
+        *      We won't pick up any zones that are allocated later.
+        */
+
+       simple_lock(&all_zones_lock);
+       max_zones = (unsigned int)(num_zones + num_fake_zones);
+       z = first_zone;
+       simple_unlock(&all_zones_lock);
+
+       names_size = round_page(max_zones * sizeof *names);
+       kr = kmem_alloc_pageable(ipc_kernel_map,
+                                &names_addr, names_size);
+       if (kr != KERN_SUCCESS)
+               return kr;
+       names = (mach_zone_name_t *) names_addr;
+
+       info_size = round_page(max_zones * sizeof *info);
+       kr = kmem_alloc_pageable(ipc_kernel_map,
+                                &info_addr, info_size);
+       if (kr != KERN_SUCCESS) {
+               kmem_free(ipc_kernel_map,
+                         names_addr, names_size);
+               return kr;
+       }
+
+       info = (task_zone_info_t *) info_addr;
+
+       zn = &names[0];
+       zi = &info[0];
+
+       for (i = 0; i < max_zones - num_fake_zones; i++) {
+               struct zone zcopy;
+
+               assert(z != ZONE_NULL);
+
+               lock_zone(z);
+               zcopy = *z;
+               unlock_zone(z);
+
+               simple_lock(&all_zones_lock);
+               z = z->next_zone;
+               simple_unlock(&all_zones_lock);
+
+               /* assuming here the name data is static */
+               (void) strncpy(zn->mzn_name, zcopy.zone_name,
+                              sizeof zn->mzn_name);
+               zn->mzn_name[sizeof zn->mzn_name - 1] = '\0';
+
+               zi->tzi_count = (uint64_t)zcopy.count;
+               zi->tzi_cur_size = (uint64_t)zcopy.cur_size;
+               zi->tzi_max_size = (uint64_t)zcopy.max_size;
+               zi->tzi_elem_size = (uint64_t)zcopy.elem_size;
+               zi->tzi_alloc_size = (uint64_t)zcopy.alloc_size;
+               zi->tzi_sum_size = zcopy.sum_count * zcopy.elem_size;
+               zi->tzi_exhaustible = (uint64_t)zcopy.exhaustible;
+               zi->tzi_collectable = (uint64_t)zcopy.collectable;
+               zi->tzi_caller_acct = (uint64_t)zcopy.caller_acct;
+               if (task->tkm_zinfo != NULL) {
+                       zi->tzi_task_alloc = task->tkm_zinfo[zcopy.index].alloc;
+                       zi->tzi_task_free = task->tkm_zinfo[zcopy.index].free;
+               } else {
+                       zi->tzi_task_alloc = 0;
+                       zi->tzi_task_free = 0;
+               }
+               zn++;
+               zi++;
+       }
+
+       /*
+        * loop through the fake zones and fill them using the specialized
+        * functions
+        */
+       for (i = 0; i < num_fake_zones; i++) {
+               int count, collectable, exhaustible, caller_acct, index;
+               vm_size_t cur_size, max_size, elem_size, alloc_size;
+               uint64_t sum_size;
+
+               strncpy(zn->mzn_name, fake_zones[i].name, sizeof zn->mzn_name);
+               zn->mzn_name[sizeof zn->mzn_name - 1] = '\0';
+               fake_zones[i].query(&count, &cur_size,
+                                   &max_size, &elem_size,
+                                   &alloc_size, &sum_size,
+                                   &collectable, &exhaustible, &caller_acct);
+               zi->tzi_count = (uint64_t)count;
+               zi->tzi_cur_size = (uint64_t)cur_size;
+               zi->tzi_max_size = (uint64_t)max_size;
+               zi->tzi_elem_size = (uint64_t)elem_size;
+               zi->tzi_alloc_size = (uint64_t)alloc_size;
+               zi->tzi_sum_size = sum_size;
+               zi->tzi_collectable = (uint64_t)collectable;
+               zi->tzi_exhaustible = (uint64_t)exhaustible;
+               zi->tzi_caller_acct = (uint64_t)caller_acct;
+               if (task->tkm_zinfo != NULL) {
+                       index = ZINFO_SLOTS - num_fake_zones + i;
+                       zi->tzi_task_alloc = task->tkm_zinfo[index].alloc;
+                       zi->tzi_task_free = task->tkm_zinfo[index].free;
+               } else {
+                       zi->tzi_task_alloc = 0;
+                       zi->tzi_task_free = 0;
+               }
+               zn++;
+               zi++;
+       }
+
+       used = max_zones * sizeof *names;
+       if (used != names_size)
+               bzero((char *) (names_addr + used), names_size - used);
+
+       kr = vm_map_copyin(ipc_kernel_map, (vm_map_address_t)names_addr,
+                          (vm_map_size_t)names_size, TRUE, &copy);
+       assert(kr == KERN_SUCCESS);
+
+       *namesp = (mach_zone_name_t *) copy;
+       *namesCntp = max_zones;
+
+       used = max_zones * sizeof *info;
+
+       if (used != info_size)
+               bzero((char *) (info_addr + used), info_size - used);
+
+       kr = vm_map_copyin(ipc_kernel_map, (vm_map_address_t)info_addr,
+                          (vm_map_size_t)info_size, TRUE, &copy);
+       assert(kr == KERN_SUCCESS);
+
+       *infop = (task_zone_info_t *) copy;
+       *infoCntp = max_zones;
+
+       return KERN_SUCCESS;
+}
+
+kern_return_t
+mach_zone_info(
+       host_t                  host,
+       mach_zone_name_array_t  *namesp,
+       mach_msg_type_number_t  *namesCntp,
+       mach_zone_info_array_t  *infop,
+       mach_msg_type_number_t  *infoCntp)
+{
+       mach_zone_name_t        *names;
+       vm_offset_t             names_addr;
+       vm_size_t               names_size;
+       mach_zone_info_t        *info;
+       vm_offset_t             info_addr;
+       vm_size_t               info_size;
+       unsigned int            max_zones, i;
+       zone_t                  z;
+       mach_zone_name_t        *zn;
+       mach_zone_info_t        *zi;
+       kern_return_t           kr;
+       
+       vm_size_t               used;
+       vm_map_copy_t           copy;
+
+
+       if (host == HOST_NULL)
+               return KERN_INVALID_HOST;
+
+       num_fake_zones = sizeof fake_zones / sizeof fake_zones[0];
+
+       /*
+        *      We assume that zones aren't freed once allocated.
+        *      We won't pick up any zones that are allocated later.
+        */
+
+       simple_lock(&all_zones_lock);
+       max_zones = (unsigned int)(num_zones + num_fake_zones);
+       z = first_zone;
+       simple_unlock(&all_zones_lock);
+
+       names_size = round_page(max_zones * sizeof *names);
+       kr = kmem_alloc_pageable(ipc_kernel_map,
+                                &names_addr, names_size);
+       if (kr != KERN_SUCCESS)
+               return kr;
+       names = (mach_zone_name_t *) names_addr;
+
+       info_size = round_page(max_zones * sizeof *info);
+       kr = kmem_alloc_pageable(ipc_kernel_map,
+                                &info_addr, info_size);
+       if (kr != KERN_SUCCESS) {
+               kmem_free(ipc_kernel_map,
+                         names_addr, names_size);
+               return kr;
+       }
+
+       info = (mach_zone_info_t *) info_addr;
+
+       zn = &names[0];
+       zi = &info[0];
+
+       for (i = 0; i < max_zones - num_fake_zones; i++) {
+               struct zone zcopy;
+
+               assert(z != ZONE_NULL);
+
+               lock_zone(z);
+               zcopy = *z;
+               unlock_zone(z);
+
+               simple_lock(&all_zones_lock);
+               z = z->next_zone;
+               simple_unlock(&all_zones_lock);
+
+               /* assuming here the name data is static */
+               (void) strncpy(zn->mzn_name, zcopy.zone_name,
+                              sizeof zn->mzn_name);
+               zn->mzn_name[sizeof zn->mzn_name - 1] = '\0';
+
+               zi->mzi_count = (uint64_t)zcopy.count;
+               zi->mzi_cur_size = (uint64_t)zcopy.cur_size;
+               zi->mzi_max_size = (uint64_t)zcopy.max_size;
+               zi->mzi_elem_size = (uint64_t)zcopy.elem_size;
+               zi->mzi_alloc_size = (uint64_t)zcopy.alloc_size;
+               zi->mzi_sum_size = zcopy.sum_count * zcopy.elem_size;
+               zi->mzi_exhaustible = (uint64_t)zcopy.exhaustible;
+               zi->mzi_collectable = (uint64_t)zcopy.collectable;
+               zn++;
+               zi++;
+       }
+
+       /*
+        * loop through the fake zones and fill them using the specialized
+        * functions
+        */
+       for (i = 0; i < num_fake_zones; i++) {
+               int count, collectable, exhaustible, caller_acct;
+               vm_size_t cur_size, max_size, elem_size, alloc_size;
+               uint64_t sum_size;
+
+               strncpy(zn->mzn_name, fake_zones[i].name, sizeof zn->mzn_name);
+               zn->mzn_name[sizeof zn->mzn_name - 1] = '\0';
+               fake_zones[i].query(&count, &cur_size,
+                                   &max_size, &elem_size,
+                                   &alloc_size, &sum_size,
+                                   &collectable, &exhaustible, &caller_acct);
+               zi->mzi_count = (uint64_t)count;
+               zi->mzi_cur_size = (uint64_t)cur_size;
+               zi->mzi_max_size = (uint64_t)max_size;
+               zi->mzi_elem_size = (uint64_t)elem_size;
+               zi->mzi_alloc_size = (uint64_t)alloc_size;
+               zi->mzi_sum_size = sum_size;
+               zi->mzi_collectable = (uint64_t)collectable;
+               zi->mzi_exhaustible = (uint64_t)exhaustible;
+
+               zn++;
+               zi++;
+       }
+
+       used = max_zones * sizeof *names;
+       if (used != names_size)
+               bzero((char *) (names_addr + used), names_size - used);
+
+       kr = vm_map_copyin(ipc_kernel_map, (vm_map_address_t)names_addr,
+                          (vm_map_size_t)names_size, TRUE, &copy);
+       assert(kr == KERN_SUCCESS);
+
+       *namesp = (mach_zone_name_t *) copy;
+       *namesCntp = max_zones;
+
+       used = max_zones * sizeof *info;
+
+       if (used != info_size)
+               bzero((char *) (info_addr + used), info_size - used);
+
+       kr = vm_map_copyin(ipc_kernel_map, (vm_map_address_t)info_addr,
+                          (vm_map_size_t)info_size, TRUE, &copy);
+       assert(kr == KERN_SUCCESS);
+
+       *infop = (mach_zone_info_t *) copy;
+       *infoCntp = max_zones;
+
+       return KERN_SUCCESS;
+}
 
 
+/*
+ * host_zone_info - LEGACY user interface for Mach zone information
+ *                 Should use mach_zone_info() instead!
+ */
 kern_return_t
 host_zone_info(
        host_t                  host,
 kern_return_t
 host_zone_info(
        host_t                  host,
@@ -1468,57 +3140,55 @@ host_zone_info(
        zone_info_t    *zi;
        kern_return_t   kr;
 
        zone_info_t    *zi;
        kern_return_t   kr;
 
+       vm_size_t       used;
+       vm_map_copy_t   copy;
+
+
        if (host == HOST_NULL)
                return KERN_INVALID_HOST;
 
        if (host == HOST_NULL)
                return KERN_INVALID_HOST;
 
+#if defined(__LP64__)
+       if (!thread_is_64bit(current_thread()))
+               return KERN_NOT_SUPPORTED;
+#else
+       if (thread_is_64bit(current_thread()))
+               return KERN_NOT_SUPPORTED;
+#endif
+
+       num_fake_zones = sizeof fake_zones / sizeof fake_zones[0];
+
        /*
         *      We assume that zones aren't freed once allocated.
         *      We won't pick up any zones that are allocated later.
         */
 
        simple_lock(&all_zones_lock);
        /*
         *      We assume that zones aren't freed once allocated.
         *      We won't pick up any zones that are allocated later.
         */
 
        simple_lock(&all_zones_lock);
-#ifdef ppc
-       max_zones = num_zones + 4;
-#else
-       max_zones = num_zones + 3; /* ATN: count the number below!! */
-#endif
+       max_zones = (unsigned int)(num_zones + num_fake_zones);
        z = first_zone;
        simple_unlock(&all_zones_lock);
 
        z = first_zone;
        simple_unlock(&all_zones_lock);
 
-       if (max_zones <= *namesCntp) {
-               /* use in-line memory */
-               names_size = *namesCntp * sizeof *names;
-               names = *namesp;
-       } else {
-               names_size = round_page(max_zones * sizeof *names);
-               kr = kmem_alloc_pageable(ipc_kernel_map,
-                                        &names_addr, names_size);
-               if (kr != KERN_SUCCESS)
-                       return kr;
-               names = (zone_name_t *) names_addr;
-       }
-
-       if (max_zones <= *infoCntp) {
-               /* use in-line memory */
-               info_size = *infoCntp * sizeof *info;
-               info = *infop;
-       } else {
-               info_size = round_page(max_zones * sizeof *info);
-               kr = kmem_alloc_pageable(ipc_kernel_map,
-                                        &info_addr, info_size);
-               if (kr != KERN_SUCCESS) {
-                       if (names != *namesp)
-                               kmem_free(ipc_kernel_map,
-                                         names_addr, names_size);
-                       return kr;
-               }
-
-               info = (zone_info_t *) info_addr;
+       names_size = round_page(max_zones * sizeof *names);
+       kr = kmem_alloc_pageable(ipc_kernel_map,
+                                &names_addr, names_size);
+       if (kr != KERN_SUCCESS)
+               return kr;
+       names = (zone_name_t *) names_addr;
+
+       info_size = round_page(max_zones * sizeof *info);
+       kr = kmem_alloc_pageable(ipc_kernel_map,
+                                &info_addr, info_size);
+       if (kr != KERN_SUCCESS) {
+               kmem_free(ipc_kernel_map,
+                         names_addr, names_size);
+               return kr;
        }
        }
+
+       info = (zone_info_t *) info_addr;
+
        zn = &names[0];
        zi = &info[0];
 
        zn = &names[0];
        zi = &info[0];
 
-       for (i = 0; i < num_zones; i++) {
+       for (i = 0; i < max_zones - num_fake_zones; i++) {
                struct zone zcopy;
 
                assert(z != ZONE_NULL);
                struct zone zcopy;
 
                assert(z != ZONE_NULL);
@@ -1534,6 +3204,7 @@ host_zone_info(
                /* assuming here the name data is static */
                (void) strncpy(zn->zn_name, zcopy.zone_name,
                               sizeof zn->zn_name);
                /* assuming here the name data is static */
                (void) strncpy(zn->zn_name, zcopy.zone_name,
                               sizeof zn->zn_name);
+               zn->zn_name[sizeof zn->zn_name - 1] = '\0';
 
                zi->zi_count = zcopy.count;
                zi->zi_cur_size = zcopy.cur_size;
 
                zi->zi_count = zcopy.count;
                zi->zi_cur_size = zcopy.cur_size;
@@ -1546,74 +3217,88 @@ host_zone_info(
                zn++;
                zi++;
        }
                zn++;
                zi++;
        }
-       strcpy(zn->zn_name, "kernel_stacks");
-       stack_fake_zone_info(&zi->zi_count, &zi->zi_cur_size, &zi->zi_max_size, &zi->zi_elem_size,
-                            &zi->zi_alloc_size, &zi->zi_collectable, &zi->zi_exhaustible);
-       zn++;
-       zi++;
-#ifdef ppc
-       strcpy(zn->zn_name, "save_areas");
-       save_fake_zone_info(&zi->zi_count, &zi->zi_cur_size, &zi->zi_max_size, &zi->zi_elem_size,
-                           &zi->zi_alloc_size, &zi->zi_collectable, &zi->zi_exhaustible);
-       zn++;
-       zi++;
-
-       strcpy(zn->zn_name, "pmap_mappings");
-       mapping_fake_zone_info(&zi->zi_count, &zi->zi_cur_size, &zi->zi_max_size, &zi->zi_elem_size,
-                              &zi->zi_alloc_size, &zi->zi_collectable, &zi->zi_exhaustible);
-       zn++;
-       zi++;
-#endif
 
 
-#ifdef i386
-       strcpy(zn->zn_name, "page_tables");
-       pt_fake_zone_info(&zi->zi_count, &zi->zi_cur_size, &zi->zi_max_size, &zi->zi_elem_size,
-                         &zi->zi_alloc_size, &zi->zi_collectable, &zi->zi_exhaustible);
-       zn++;
-       zi++;
-#endif
+       /*
+        * loop through the fake zones and fill them using the specialized
+        * functions
+        */
+       for (i = 0; i < num_fake_zones; i++) {
+               int caller_acct;
+               uint64_t sum_space;
+               strncpy(zn->zn_name, fake_zones[i].name, sizeof zn->zn_name);
+               zn->zn_name[sizeof zn->zn_name - 1] = '\0';
+               fake_zones[i].query(&zi->zi_count, &zi->zi_cur_size,
+                                   &zi->zi_max_size, &zi->zi_elem_size,
+                                   &zi->zi_alloc_size, &sum_space,
+                                   &zi->zi_collectable, &zi->zi_exhaustible, &caller_acct);
+               zn++;
+               zi++;
+       }
+
+       used = max_zones * sizeof *names;
+       if (used != names_size)
+               bzero((char *) (names_addr + used), names_size - used);
 
 
-       strcpy(zn->zn_name, "kalloc.large");
-       kalloc_fake_zone_info(&zi->zi_count, &zi->zi_cur_size, &zi->zi_max_size, &zi->zi_elem_size,
-                              &zi->zi_alloc_size, &zi->zi_collectable, &zi->zi_exhaustible);
+       kr = vm_map_copyin(ipc_kernel_map, (vm_map_address_t)names_addr,
+                          (vm_map_size_t)names_size, TRUE, &copy);
+       assert(kr == KERN_SUCCESS);
 
 
-       if (names != *namesp) {
-               vm_size_t used;
-               vm_map_copy_t copy;
+       *namesp = (zone_name_t *) copy;
+       *namesCntp = max_zones;
 
 
-               used = max_zones * sizeof *names;
+       used = max_zones * sizeof *info;
+       if (used != info_size)
+               bzero((char *) (info_addr + used), info_size - used);
 
 
-               if (used != names_size)
-                       bzero((char *) (names_addr + used), names_size - used);
+       kr = vm_map_copyin(ipc_kernel_map, (vm_map_address_t)info_addr,
+                          (vm_map_size_t)info_size, TRUE, &copy);
+       assert(kr == KERN_SUCCESS);
 
 
-               kr = vm_map_copyin(ipc_kernel_map, (vm_map_address_t)names_addr,
-                                  (vm_map_size_t)names_size, TRUE, &copy);
-               assert(kr == KERN_SUCCESS);
+       *infop = (zone_info_t *) copy;
+       *infoCntp = max_zones;
 
 
-               *namesp = (zone_name_t *) copy;
-       }
-       *namesCntp = max_zones;
+       return KERN_SUCCESS;
+}
 
 
-       if (info != *infop) {
-               vm_size_t used;
-               vm_map_copy_t copy;
+extern unsigned int stack_total;
+extern unsigned long long stack_allocs;
 
 
-               used = max_zones * sizeof *info;
+#if defined(__i386__) || defined (__x86_64__)
+extern unsigned int inuse_ptepages_count;
+extern long long alloc_ptepages_count;
+#endif
 
 
-               if (used != info_size)
-                       bzero((char *) (info_addr + used), info_size - used);
+void zone_display_zprint()
+{
+       unsigned int    i;
+       zone_t          the_zone;
+
+       if(first_zone!=NULL) {
+               the_zone = first_zone;
+               for (i = 0; i < num_zones; i++) {
+                       if(the_zone->cur_size > (1024*1024)) {
+                               printf("%.20s:\t%lu\n",the_zone->zone_name,(uintptr_t)the_zone->cur_size);
+                       }
 
 
-               kr = vm_map_copyin(ipc_kernel_map, (vm_map_address_t)info_addr,
-                                  (vm_map_size_t)info_size, TRUE, &copy);
-               assert(kr == KERN_SUCCESS);
+                       if(the_zone->next_zone == NULL) {
+                               break;
+                       }
 
 
-               *infop = (zone_info_t *) copy;
+                       the_zone = the_zone->next_zone;
+               }
        }
        }
-       *infoCntp = max_zones;
 
 
-       return KERN_SUCCESS;
+       printf("Kernel Stacks:\t%lu\n",(uintptr_t)(kernel_stack_size * stack_total));
+
+#if defined(__i386__) || defined (__x86_64__)
+       printf("PageTables:\t%lu\n",(uintptr_t)(PAGE_SIZE * inuse_ptepages_count));
+#endif
+
+       printf("Kalloc.Large:\t%lu\n",(uintptr_t)kalloc_large_total);
 }
 
 }
 
+
+
 #if    MACH_KDB
 #include <ddb/db_command.h>
 #include <ddb/db_output.h>
 #if    MACH_KDB
 #include <ddb/db_command.h>
 #include <ddb/db_output.h>
@@ -1652,16 +3337,15 @@ db_print_zone(
                db_printf("C");
        if (zcopy.expandable)
                db_printf("X");
                db_printf("C");
        if (zcopy.expandable)
                db_printf("X");
+       if (zcopy.caller_acct)
+               db_printf("A");
        db_printf("\n");
 }
 
 /*ARGSUSED*/
 void
        db_printf("\n");
 }
 
 /*ARGSUSED*/
 void
-db_show_one_zone(
-        db_expr_t              addr,
-        int                    have_addr,
-        __unused db_expr_t     count,
-        __unused char *        modif)
+db_show_one_zone(db_expr_t addr, boolean_t have_addr,
+                __unused db_expr_t count, __unused char *modif)
 {
        struct zone *z = (zone_t)((char *)0 + addr);
 
 {
        struct zone *z = (zone_t)((char *)0 + addr);
 
@@ -1676,11 +3360,8 @@ db_show_one_zone(
 
 /*ARGSUSED*/
 void
 
 /*ARGSUSED*/
 void
-db_show_all_zones(
-        __unused db_expr_t     addr,
-        int                    have_addr,
-        db_expr_t              count,
-        __unused char *        modif)
+db_show_all_zones(__unused db_expr_t addr, boolean_t have_addr, db_expr_t count,
+                 __unused char *modif)
 {
        zone_t          z;
        unsigned total = 0;
 {
        zone_t          z;
        unsigned total = 0;
@@ -1729,13 +3410,13 @@ db_zone_check_active(
        while (count < zone->count) {
                count++;
                if (tmp_elem == 0) {
        while (count < zone->count) {
                count++;
                if (tmp_elem == 0) {
-                       printf("unexpected zero element, zone=0x%x, count=%d\n",
+                       printf("unexpected zero element, zone=%p, count=%d\n",
                                zone, count);
                        assert(FALSE);
                        break;
                }
                if (queue_end(tmp_elem, &zone->active_zones)) {
                                zone, count);
                        assert(FALSE);
                        break;
                }
                if (queue_end(tmp_elem, &zone->active_zones)) {
-                       printf("unexpected queue_end, zone=0x%x, count=%d\n",
+                       printf("unexpected queue_end, zone=%p, count=%d\n",
                                zone, count);
                        assert(FALSE);
                        break;
                                zone, count);
                        assert(FALSE);
                        break;
@@ -1743,7 +3424,7 @@ db_zone_check_active(
                tmp_elem = queue_next(tmp_elem);
        }
        if (!queue_end(tmp_elem, &zone->active_zones)) {
                tmp_elem = queue_next(tmp_elem);
        }
        if (!queue_end(tmp_elem, &zone->active_zones)) {
-               printf("not at queue_end, zone=0x%x, tmp_elem=0x%x\n",
+               printf("not at queue_end, zone=%p, tmp_elem=%p\n",
                        zone, tmp_elem);
                assert(FALSE);
        }
                        zone, tmp_elem);
                assert(FALSE);
        }
@@ -1757,7 +3438,7 @@ db_zone_print_active(
        queue_t tmp_elem;
 
        if (!zone_debug_enabled(zone)) {
        queue_t tmp_elem;
 
        if (!zone_debug_enabled(zone)) {
-               printf("zone 0x%x debug not enabled\n", zone);
+               printf("zone %p debug not enabled\n", zone);
                return;
        }
        if (!zone_check) {
                return;
        }
        if (!zone_check) {
@@ -1765,11 +3446,11 @@ db_zone_print_active(
                return;
        }
 
                return;
        }
 
-       printf("zone 0x%x, active elements %d\n", zone, zone->count);
+       printf("zone %p, active elements %d\n", zone, zone->count);
        printf("active list:\n");
        tmp_elem = queue_first(&zone->active_zones);
        while (count < zone->count) {
        printf("active list:\n");
        tmp_elem = queue_first(&zone->active_zones);
        while (count < zone->count) {
-               printf("  0x%x", tmp_elem);
+               printf("  %p", tmp_elem);
                count++;
                if ((count % 6) == 0)
                        printf("\n");
                count++;
                if ((count % 6) == 0)
                        printf("\n");
@@ -1784,7 +3465,7 @@ db_zone_print_active(
                tmp_elem = queue_next(tmp_elem);
        }
        if (!queue_end(tmp_elem, &zone->active_zones))
                tmp_elem = queue_next(tmp_elem);
        }
        if (!queue_end(tmp_elem, &zone->active_zones))
-               printf("\nnot at queue_end, tmp_elem=0x%x\n", tmp_elem);
+               printf("\nnot at queue_end, tmp_elem=%p\n", tmp_elem);
        else
                printf("\n");
 }
        else
                printf("\n");
 }
@@ -1799,7 +3480,7 @@ db_zone_print_free(
        vm_offset_t elem;
 
        freecount = zone_free_count(zone);
        vm_offset_t elem;
 
        freecount = zone_free_count(zone);
-       printf("zone 0x%x, free elements %d\n", zone, freecount);
+       printf("zone %p, free elements %d\n", zone, freecount);
        printf("free list:\n");
        elem = zone->free_elements;
        while (count < freecount) {
        printf("free list:\n");
        elem = zone->free_elements;
        while (count < freecount) {
@@ -1835,11 +3516,11 @@ next_element(
        char            *elt = (char *)prev;
 
        if (!zone_debug_enabled(z))
        char            *elt = (char *)prev;
 
        if (!zone_debug_enabled(z))
-               return(0);
+               return(NULL);
        elt -= ZONE_DEBUG_OFFSET;
        elt = (char *) queue_next((queue_t) elt);
        if ((queue_t) elt == &z->active_zones)
        elt -= ZONE_DEBUG_OFFSET;
        elt = (char *) queue_next((queue_t) elt);
        if ((queue_t) elt == &z->active_zones)
-               return(0);
+               return(NULL);
        elt += ZONE_DEBUG_OFFSET;
        return(elt);
 }
        elt += ZONE_DEBUG_OFFSET;
        return(elt);
 }
@@ -1851,9 +3532,9 @@ first_element(
        char            *elt;
 
        if (!zone_debug_enabled(z))
        char            *elt;
 
        if (!zone_debug_enabled(z))
-               return(0);
+               return(NULL);
        if (queue_empty(&z->active_zones))
        if (queue_empty(&z->active_zones))
-               return(0);
+               return(NULL);
        elt = (char *)queue_first(&z->active_zones);
        elt += ZONE_DEBUG_OFFSET;
        return(elt);
        elt = (char *)queue_first(&z->active_zones);
        elt += ZONE_DEBUG_OFFSET;
        return(elt);
@@ -1909,6 +3590,8 @@ zone_debug_disable(
        if (!zone_debug_enabled(z) || zone_in_use(z))
                return;
        z->elem_size -= ZONE_DEBUG_OFFSET;
        if (!zone_debug_enabled(z) || zone_in_use(z))
                return;
        z->elem_size -= ZONE_DEBUG_OFFSET;
-       z->active_zones.next = z->active_zones.prev = 0;        
+       z->active_zones.next = z->active_zones.prev = NULL;
 }
 }
+
+
 #endif /* ZONE_DEBUG */
 #endif /* ZONE_DEBUG */