]> git.saurik.com Git - apple/libc.git/blobdiff - gen/stack_logging_disk.c
Libc-997.1.1.tar.gz
[apple/libc.git] / gen / stack_logging_disk.c
diff --git a/gen/stack_logging_disk.c b/gen/stack_logging_disk.c
deleted file mode 100644 (file)
index 00a0cae..0000000
+++ /dev/null
@@ -1,1727 +0,0 @@
-/*
- * Copyright (c) 2007-2009 Apple Inc. All rights reserved.
- *
- * @APPLE_LICENSE_HEADER_START@
- * 
- * This file contains Original Code and/or Modifications of Original Code
- * as defined in and that are subject to the Apple Public Source License
- * Version 2.0 (the 'License'). You may not use this file except in
- * compliance with the License. Please obtain a copy of the License at
- * http://www.opensource.apple.com/apsl/ and read it before using this
- * file.
- * 
- * The Original Code and all software distributed under the License are
- * distributed on an 'AS IS' basis, WITHOUT WARRANTY OF ANY KIND, EITHER
- * EXPRESS OR IMPLIED, AND APPLE HEREBY DISCLAIMS ALL SUCH WARRANTIES,
- * INCLUDING WITHOUT LIMITATION, ANY WARRANTIES OF MERCHANTABILITY,
- * FITNESS FOR A PARTICULAR PURPOSE, QUIET ENJOYMENT OR NON-INFRINGEMENT.
- * Please see the License for the specific language governing rights and
- * limitations under the License.
- * 
- * @APPLE_LICENSE_HEADER_END@
- */
-
-#include <stdio.h>
-#include <stdlib.h>
-#include <string.h>
-#include <limits.h>
-#include <unistd.h>
-#include <fcntl.h>
-#include <dirent.h>
-#include <libkern/OSAtomic.h>
-#include <mach/mach.h>
-#include <mach/mach_vm.h>
-#include <sys/sysctl.h>
-#include <sys/stat.h>
-#include <sys/mman.h>
-#include <pthread.h>
-#include <paths.h>
-#include <errno.h>
-#include "stack_logging.h"
-#include "malloc_printf.h"
-#include "_simple.h"            // as included by malloc.c, this defines ASL_LEVEL_INFO
-
-#pragma mark -
-#pragma mark Defines
-
-#ifdef TEST_DISK_STACK_LOGGING
-#define _malloc_printf fprintf
-#undef ASL_LEVEL_INFO
-#define ASL_LEVEL_INFO stderr
-#endif
-
-#define STACK_LOGGING_MAX_STACK_SIZE 512
-#define STACK_LOGGING_BLOCK_WRITING_SIZE 8192
-#define STACK_LOGGING_MAX_SIMUL_REMOTE_TASKS_INSPECTED 3
-
-#define BACKTRACE_UNIQUING_DEBUG 0
-
-// The expansion factor controls the shifting up of table size. A factor of 1 will double the size upon expanding,
-// 2 will quadruple the size, etc. Maintaining a 66% fill in an ideal table requires the collision allowance to
-// increase by 3 for every quadrupling of the table size (although this the constant applied to insertion
-// performance O(c*n))
-#define EXPAND_FACTOR 2
-#define COLLISION_GROWTH_RATE 3
-
-// For a uniquing table, the useful node size is slots := floor(table_byte_size / (2 * sizeof(mach_vm_address_t)))
-// Some useful numbers for the initial max collision value (desiring 66% fill):
-// 16K-23K slots -> 16 collisions
-// 24K-31K slots -> 17 collisions
-// 32K-47K slots -> 18 collisions
-// 48K-79K slots -> 19 collisions
-// 80K-96K slots -> 20 collisions
-#define INITIAL_MAX_COLLIDE    19
-#define DEFAULT_UNIQUING_PAGE_SIZE 256
-
-#pragma mark -
-#pragma mark Macros
-
-#define STACK_LOGGING_FLAGS(longlongvar) (uint8_t)((uint64_t)(longlongvar) >> 56)
-#define STACK_LOGGING_OFFSET(longlongvar) ((longlongvar) & 0x00FFFFFFFFFFFFFFull)
-#define STACK_LOGGING_OFFSET_AND_FLAGS(longlongvar, realshortvar) (((uint64_t)(longlongvar) & 0x00FFFFFFFFFFFFFFull) | ((uint64_t)(realshortvar) << 56))
-
-#pragma mark -
-#pragma mark Types
-
-typedef struct {
-       uintptr_t       argument;
-       uintptr_t       address;
-       uint64_t        offset_and_flags; // top 8 bits are actually the flags!
-} stack_logging_index_event;
-
-typedef struct {
-       uint32_t        argument;
-       uint32_t        address;
-       uint64_t        offset_and_flags; // top 8 bits are actually the flags!
-} stack_logging_index_event32;
-
-typedef struct {
-       uint64_t        argument;
-       uint64_t        address;
-       uint64_t        offset_and_flags; // top 8 bits are actually the flags!
-} stack_logging_index_event64;
-
-#pragma pack(push,4)
-typedef struct {
-    uint64_t                                                   numPages; // number of pages of the table
-       uint64_t                                                        numNodes;
-       uint64_t                                                        tableSize;
-       uint64_t                                                        untouchableNodes;
-       mach_vm_address_t                                       table_address;
-       int32_t                                                         max_collide;
-       // 'table_address' is just an always 64-bit version of the pointer-sized 'table' field to remotely read;
-       // it's important that the offset of 'table_address' in the struct does not change between 32 and 64-bit.
-#if BACKTRACE_UNIQUING_DEBUG
-       uint64_t nodesFull;
-       uint64_t backtracesContained;
-#endif
-       mach_vm_address_t                                       *table; // allocated using vm_allocate()
-} backtrace_uniquing_table;
-#pragma pack(pop)
-
-// for storing/looking up allocations that haven't yet be written to disk; consistent size across 32/64-bit processes.
-// It's important that these fields don't change alignment due to the architecture because they may be accessed from an
-// analyzing process with a different arch - hence the pragmas.
-#pragma pack(push,4)
-typedef struct {
-       uint64_t                        start_index_offset;
-       uint32_t                        next_free_index_buffer_offset;
-       mach_vm_address_t       uniquing_table_address;
-       char                            index_buffer[STACK_LOGGING_BLOCK_WRITING_SIZE];
-       backtrace_uniquing_table        *uniquing_table;
-} stack_buffer_shared_memory;
-#pragma pack(pop)
-
-// target process address -> record table (for __mach_stack_logging_get_frames)
-typedef struct {
-       uint64_t                address;
-       uint64_t                index_file_offset;
-} remote_index_node;
-
-// for caching index information client-side:
-typedef struct {
-       size_t                  cache_size;
-       size_t                  cache_node_capacity;
-       uint32_t                collision_allowance;
-       remote_index_node       *table_memory; // this can be malloced; it's on the client side.
-       stack_buffer_shared_memory *shmem; // shared memory
-       stack_buffer_shared_memory snapshot; // memory snapshot of the remote process' shared memory
-       uint32_t                last_pre_written_index_size;
-       uint64_t                last_index_file_offset;
-       backtrace_uniquing_table uniquing_table; // snapshot of the remote process' uniquing table
-} remote_index_cache;
-
-// for reading stack history information from remote processes:
-typedef struct {
-       task_t                  remote_task;
-       pid_t                   remote_pid;
-       int32_t                 task_is_64_bit;
-       int32_t                 in_use_count;
-       FILE                    *index_file_stream;
-       remote_index_cache      *cache;
-} remote_task_file_streams;
-
-#pragma mark -
-#pragma mark Constants/Globals
-
-static OSSpinLock stack_logging_lock = OS_SPINLOCK_INIT;
-
-// support for multi-threaded forks
-extern void __stack_logging_fork_prepare();
-extern void __stack_logging_fork_parent();
-extern void __stack_logging_fork_child();
-extern void __stack_logging_early_finished();
-
-// support for gdb and others checking for stack_logging locks
-__private_extern__ boolean_t __stack_logging_locked();
-
-// single-thread access variables
-static stack_buffer_shared_memory *pre_write_buffers;
-static vm_address_t *stack_buffer;
-static uintptr_t last_logged_malloc_address = 0;
-
-// Constants to define stack logging file path names.
-// Files will get written as /tmp/stack-logs.<pid>.<progname>.XXXXXX.index
-// unless the base directory is specified otherwise with MallocStackLoggingDirectory.
-// In this case, a file /tmp/stack-logs.<pid>.<progname>.XXXXXX.link will also be created.
-static const char *stack_log_file_base_name = "stack-logs.";
-static const char *stack_log_file_suffix = ".index";
-static const char *stack_log_link_suffix = ".link";
-
-static void *stack_log_path_buffers = NULL;
-static char *stack_log_location = NULL;
-static char *stack_log_reference_file = NULL;
-char *__stack_log_file_path__ = NULL;
-static int index_file_descriptor = -1;
-
-// for accessing remote log files
-static remote_task_file_streams remote_fds[STACK_LOGGING_MAX_SIMUL_REMOTE_TASKS_INSPECTED];
-static uint32_t next_remote_task_fd = 0;
-static uint32_t remote_task_fd_count = 0;
-static OSSpinLock remote_fd_list_lock = OS_SPINLOCK_INIT;
-
-// activation variables
-static int logging_use_compaction = 1; // set this to zero to always disable compaction.
-
-// We set malloc_logger to NULL to disable logging, if we encounter errors
-// during file writing
-typedef void (malloc_logger_t)(uint32_t type, uintptr_t arg1, uintptr_t arg2, uintptr_t arg3, uintptr_t result, uint32_t num_hot_frames_to_skip);
-extern malloc_logger_t *malloc_logger;
-
-#pragma mark -
-#pragma mark In-Memory Backtrace Uniquing
-
-static __attribute__((always_inline))
-inline void* 
-allocate_pages(uint64_t memSize)
-{
-    mach_vm_address_t allocatedMem = 0ull;
-       if (mach_vm_allocate(mach_task_self(), &allocatedMem, memSize, VM_FLAGS_ANYWHERE | VM_MAKE_TAG(VM_MEMORY_ANALYSIS_TOOL)) != KERN_SUCCESS) {
-               malloc_printf("allocate_pages(): virtual memory exhaused!\n");
-    }
-    return (void*)(uintptr_t)allocatedMem;
-}
-
-static __attribute__((always_inline))
-inline int 
-deallocate_pages(void* memPointer, uint64_t memSize)
-{
-       return mach_vm_deallocate(mach_task_self(), (mach_vm_address_t)(uintptr_t)memPointer, memSize);
-}
-
-static backtrace_uniquing_table* 
-__create_uniquing_table(void)
-{
-       backtrace_uniquing_table *uniquing_table = (backtrace_uniquing_table*)allocate_pages((uint64_t)round_page(sizeof(backtrace_uniquing_table)));
-       if (!uniquing_table) return NULL;
-       bzero(uniquing_table, sizeof(backtrace_uniquing_table));
-       uniquing_table->numPages = DEFAULT_UNIQUING_PAGE_SIZE;
-       uniquing_table->tableSize = uniquing_table->numPages * vm_page_size;
-       uniquing_table->numNodes = ((uniquing_table->tableSize / (sizeof(mach_vm_address_t) * 2)) >> 1) << 1; // make sure it's even.
-       uniquing_table->table = (mach_vm_address_t*)(uintptr_t)allocate_pages(uniquing_table->tableSize);
-       uniquing_table->table_address = (uintptr_t)uniquing_table->table;
-       uniquing_table->max_collide = INITIAL_MAX_COLLIDE;
-       uniquing_table->untouchableNodes = 0;
-       
-#if BACKTRACE_UNIQUING_DEBUG
-       malloc_printf("create_uniquing_table(): creating. size: %lldKB == %lldMB, numnodes: %lld (%lld untouchable)\n", uniquing_table->tableSize >> 10, uniquing_table->tableSize >> 20, uniquing_table->numNodes, uniquing_table->untouchableNodes);
-       malloc_printf("create_uniquing_table(): table: %p; end: %p\n", uniquing_table->table, (void*)((uintptr_t)uniquing_table->table + (uintptr_t)uniquing_table->tableSize));
-#endif
-       return uniquing_table;
-}
-
-static void
-__destroy_uniquing_table(backtrace_uniquing_table* table)
-{
-       deallocate_pages(table->table, table->tableSize);
-       deallocate_pages(table, sizeof(backtrace_uniquing_table));
-}
-
-static void 
-__expand_uniquing_table(backtrace_uniquing_table *uniquing_table)
-{
-       mach_vm_address_t *oldTable = uniquing_table->table;
-       uint64_t oldsize = uniquing_table->tableSize;
-       uint64_t oldnumnodes = uniquing_table->numNodes;
-       
-       uniquing_table->numPages = uniquing_table->numPages << EXPAND_FACTOR;
-       uniquing_table->tableSize = uniquing_table->numPages * vm_page_size;
-       uniquing_table->numNodes = ((uniquing_table->tableSize / (sizeof(mach_vm_address_t) * 2)) >> 1) << 1; // make sure it's even.
-       mach_vm_address_t *newTable = (mach_vm_address_t*)(uintptr_t)allocate_pages(uniquing_table->tableSize);
-       
-       uniquing_table->table = newTable;
-       uniquing_table->table_address = (uintptr_t)uniquing_table->table;
-       uniquing_table->max_collide = uniquing_table->max_collide + COLLISION_GROWTH_RATE;
-       
-       if (mach_vm_copy(mach_task_self(), (mach_vm_address_t)(uintptr_t)oldTable, oldsize, (mach_vm_address_t)(uintptr_t)newTable) != KERN_SUCCESS) {
-               malloc_printf("expandUniquingTable(): VMCopyFailed\n");
-       }
-       uniquing_table->untouchableNodes = oldnumnodes;
-       
-#if BACKTRACE_UNIQUING_DEBUG
-       malloc_printf("expandUniquingTable(): expanded from nodes full: %lld of: %lld (~%2d%%); to nodes: %lld (inactive = %lld); unique bts: %lld\n", 
-                                 uniquing_table->nodesFull, oldnumnodes, (int)(((uniquing_table->nodesFull * 100.0) / (double)oldnumnodes) + 0.5),
-                                 uniquing_table->numNodes, uniquing_table->untouchableNodes, uniquing_table->backtracesContained);
-       malloc_printf("expandUniquingTable(): allocate: %p; end: %p\n", newTable, (void*)((uintptr_t)newTable + (uintptr_t)(uniquing_table->tableSize)));
-       malloc_printf("expandUniquingTable(): deallocate: %p; end: %p\n", oldTable, (void*)((uintptr_t)oldTable + (uintptr_t)oldsize));
-#endif
-       
-       if (deallocate_pages(oldTable, oldsize) != KERN_SUCCESS) {
-               malloc_printf("expandUniquingTable(): mach_vm_deallocate failed. [%p]\n", uniquing_table->table);
-       }
-}
-
-static int 
-__enter_frames_in_table(backtrace_uniquing_table *uniquing_table, uint64_t *foundIndex, mach_vm_address_t *frames, int32_t count)
-{      
-       // The hash values need to be the same size as the addresses (because we use the value -1), for clarity, define a new type
-       typedef mach_vm_address_t hash_index_t;
-
-       mach_vm_address_t thisPC;
-       hash_index_t hash, uParent = (hash_index_t)(-1ll), modulus = (uniquing_table->numNodes-uniquing_table->untouchableNodes-1);
-       int32_t collisions, lcopy = count, returnVal = 1;
-       hash_index_t hash_multiplier = ((uniquing_table->numNodes - uniquing_table->untouchableNodes)/(uniquing_table->max_collide*2+1));
-       mach_vm_address_t *node;
-       while (--lcopy >= 0) {
-        thisPC = frames[lcopy];
-               
-               // hash = initialHash(uniquing_table, uParent, thisPC);
-               hash = uniquing_table->untouchableNodes + (((uParent << 4) ^ (thisPC >> 2)) % modulus);
-               collisions = uniquing_table->max_collide;
-               
-               while (collisions--) {
-                       node = uniquing_table->table + (hash * 2);
-                       
-                       if (*node == 0 && node[1] == 0) { 
-                               // blank; store this entry!
-                               // Note that we need to test for both head[0] and head[1] as (0, -1) is a valid entry
-                               node[0] = thisPC;
-                               node[1] = uParent;
-                               uParent = hash;
-#if BACKTRACE_UNIQUING_DEBUG
-                               uniquing_table->nodesFull++;
-                               if (lcopy == 0) {
-                                       uniquing_table->backtracesContained++;
-                               }
-#endif
-                               break;
-                       }
-                       if (*node == thisPC && node[1] == uParent) {
-                               // hit! retrieve index and go.
-                               uParent = hash;
-                               break;
-                       }
-                       
-                       hash += collisions * hash_multiplier + 1;
-                       
-                       if (hash >= uniquing_table->numNodes) {
-                               hash -= (uniquing_table->numNodes - uniquing_table->untouchableNodes); // wrap around.
-                       }
-               }
-               
-               if (collisions < 0) {
-                       returnVal = 0;
-                       break;
-               }
-    }
-       
-       if (returnVal) *foundIndex = uParent;
-       
-       return returnVal;
-}
-
-static void 
-__unwind_stack_from_table_index(backtrace_uniquing_table *uniquing_table, uint64_t index_pos, mach_vm_address_t *out_frames_buffer, uint32_t *out_frames_count, uint32_t max_frames)
-{      
-       mach_vm_address_t *node = uniquing_table->table + (index_pos * 2);
-       uint32_t foundFrames = 0;
-       if (index_pos < uniquing_table->numNodes) {
-               while (foundFrames < max_frames) {
-                       out_frames_buffer[foundFrames++] = node[0];
-                       if (node[1] == (mach_vm_address_t)(-1ll)) break;
-                       node = uniquing_table->table + (node[1] * 2);
-               }
-       }
-       
-       *out_frames_count = foundFrames;
-}
-
-#pragma mark -
-#pragma mark Disk Stack Logging
-
-static void delete_log_files(void);    // pre-declare
-static int delete_logging_file(char *log_location);
-
-static void
-append_int(char * filename, pid_t pid, size_t maxLength)
-{      
-       size_t len = strlen(filename);
-       
-       uint32_t count = 0;
-       pid_t value = pid;
-       while (value > 0) {
-               value /= 10;
-               count++;
-       }
-       
-       if (len + count >= maxLength) return; // don't modify the string if it would violate maxLength
-       
-       filename[len + count] = '\0';
-       
-       value = pid;
-       uint32_t i;
-       for (i = 0 ; i < count ; i ++) {
-               filename[len + count - 1 - i] = '0' + value % 10;
-               value /= 10;
-       }
-}
-
-/*
- * <rdar://problem/11128080> if we needed to call confstr during init then setting this
- * flag will postpone stack logging until after Libsystem's initialiser has run.
- */
-static void
-postpone_stack_logging(void)
-{
-       _malloc_printf(ASL_LEVEL_INFO, "stack logging postponed until after initialization.\n");
-       stack_logging_postponed = 1;
-}
-
-/*
- * Check various temporary directory options starting with _PATH_TMP and use confstr.
- * Allocating and releasing target buffer is the caller's responsibility.
- */
-static bool
-get_writeable_temp_dir(char* target)
-{
-       if (!target) return false;
-       if (-1 != access(_PATH_TMP, W_OK)) {
-               strlcpy(target, _PATH_TMP, (size_t)PATH_MAX);
-               return true;
-       }
-       if (getenv("TMPDIR") && (-1 != access(getenv("TMPDIR"), W_OK))) {
-               strlcpy(target, getenv("TMPDIR"), (size_t)PATH_MAX);
-               return true;
-       }
-       if (stack_logging_finished_init) {
-               size_t n = confstr(_CS_DARWIN_USER_TEMP_DIR, target, (size_t) PATH_MAX);
-               if ((n > 0) && (n < PATH_MAX)) return true;
-               n = confstr(_CS_DARWIN_USER_CACHE_DIR, target, (size_t) PATH_MAX);
-               if ((n > 0) && (n < PATH_MAX)) return true;
-       } else {
-               /* <rdar://problem/11128080> Can't call confstr during init, so postpone
-                logging till after */
-               postpone_stack_logging();
-       }
-       /* No writeable tmp directory found. Maybe shd try /private/var/tmp for device here ... */
-       *target = '\0';
-       return false;
-}
-
-/*
- * If successful, returns path to log file that was created, otherwise NULL.
- *
- * The log could be in one of 3 places (in decreasing order of preference)
- *
- * 1)  value of environment variable MallocStackLoggingDirectory
- * 2)  the temp directory /tmp for desktop apps and internal apps on devices, or
- * 3)  the sandbox location + tmp/ in case of 3rd party apps on the device.
- *
- * For 1 and 3, we create a .link file with the path of the file. We prefer to
- * create this file in /tmp, but if we are unable to (device 3rd party case),
- * we create it in the same location as the .index file and issue a message
- * in syslog asking for it to be copied to /tmp to enable tools.
- *
- */
-static char *
-create_log_file(void)
-{
-       pid_t pid = getpid();
-       const char *progname = getprogname();
-       char *created_log_location = NULL;
-
-       if (stack_log_path_buffers == NULL) {
-               /*
-                * on first use, allocate buffers directly from the OS without
-                * using malloc
-                */
-
-               stack_log_path_buffers = allocate_pages((uint64_t)round_page(3*PATH_MAX));
-               if (stack_log_path_buffers == NULL) {
-                       _malloc_printf(ASL_LEVEL_INFO, "unable to allocate memory for path buffers\n");
-                       return NULL;
-               }
-               
-               stack_log_location = &((char *)stack_log_path_buffers)[0*PATH_MAX];
-               stack_log_reference_file = &((char *)stack_log_path_buffers)[1*PATH_MAX];
-               __stack_log_file_path__ = &((char *)stack_log_path_buffers)[2*PATH_MAX];
-       }
-       
-       // WARNING! use of snprintf can induce malloc() calls 
-       bool use_alternate_location = false;
-       char *evn_log_directory = getenv("MallocStackLoggingDirectory");
-       size_t stack_log_len;
-       if (evn_log_directory && *evn_log_directory) {
-               use_alternate_location = true;
-               strlcpy(stack_log_location, evn_log_directory, (size_t)PATH_MAX);
-       }
-       if (!use_alternate_location || (access(stack_log_location, W_OK) == -1)) {
-               if (!get_writeable_temp_dir(stack_log_location)) {
-                       if (!stack_logging_postponed) {
-                               _malloc_printf(ASL_LEVEL_INFO, "No writeable tmp dir\n");
-                       }
-                       return NULL;
-               }
-               if (0 != strcmp(stack_log_location, _PATH_TMP))
-                       use_alternate_location = true;
-       }
-       stack_log_len = strlen(stack_log_location);
-       // add the '/' only if it's not already there.
-       if (stack_log_location[stack_log_len-1] != '/') {
-               strlcat(stack_log_location, "/", (size_t)PATH_MAX);
-               ++stack_log_len;
-       }
-       
-       strlcpy(__stack_log_file_path__, stack_log_location, (size_t)PATH_MAX);
-       
-       strlcat(__stack_log_file_path__, stack_log_file_base_name, (size_t)PATH_MAX);
-       append_int(__stack_log_file_path__, pid, (size_t)PATH_MAX);
-       if (progname && progname[0] != '\0') {
-               strlcat(__stack_log_file_path__, ".", (size_t)PATH_MAX);
-               strlcat(__stack_log_file_path__, progname, (size_t)PATH_MAX);
-       }
-       if (!use_alternate_location) strlcat(__stack_log_file_path__, ".XXXXXX", (size_t)PATH_MAX);
-       strlcat(__stack_log_file_path__, stack_log_file_suffix, (size_t)PATH_MAX);
-       
-       // Securely create the log file.
-       if ((index_file_descriptor = mkstemps(__stack_log_file_path__, (int)strlen(stack_log_file_suffix))) != -1) {
-               _malloc_printf(ASL_LEVEL_INFO, "stack logs being written into %s\n", __stack_log_file_path__);
-               created_log_location = __stack_log_file_path__;
-       } else {
-               _malloc_printf(ASL_LEVEL_INFO, "unable to create stack logs at %s\n", stack_log_location);
-               if (use_alternate_location) delete_logging_file(stack_log_reference_file);
-               stack_log_reference_file[0] = '\0';
-               stack_log_location[0] = '\0';
-               __stack_log_file_path__[0] = '\0';
-               created_log_location = NULL;
-               return created_log_location;
-       }
-       
-       // in the case where the user has specified an alternate location, drop a reference file
-       // in /tmp with the suffix 'stack_log_link_suffix' (".link") and save the path of the
-       // stack logging file there.
-       bool use_alternate_link_location = false;
-       if (use_alternate_location) {
-               strlcpy(stack_log_reference_file, _PATH_TMP, (size_t)PATH_MAX);
-               if (access(stack_log_reference_file, W_OK) == -1) {
-                       strlcpy(stack_log_reference_file, stack_log_location, (size_t)PATH_MAX);
-                       use_alternate_link_location = true;
-               }
-               strlcat(stack_log_reference_file, stack_log_file_base_name, (size_t)PATH_MAX);
-               append_int(stack_log_reference_file, pid, (size_t)PATH_MAX);
-               if (progname && progname[0] != '\0') {
-                       strlcat(stack_log_reference_file, ".", (size_t)PATH_MAX);
-                       strlcat(stack_log_reference_file, progname, (size_t)PATH_MAX);
-               }
-               if (!use_alternate_link_location)
-               strlcat(stack_log_reference_file, ".XXXXXX", (size_t)PATH_MAX);
-               strlcat(stack_log_reference_file, ".XXXXXX", (size_t)PATH_MAX);
-               strlcat(stack_log_reference_file, stack_log_link_suffix, (size_t)PATH_MAX);
-               
-               int link_file_descriptor = mkstemps(stack_log_reference_file, (int)strlen(stack_log_link_suffix));
-               if (link_file_descriptor == -1) {
-                       _malloc_printf(ASL_LEVEL_INFO, "unable to create stack reference file %s at %s\n",
-                               stack_log_reference_file, stack_log_location);
-               } else {
-                       ssize_t written = write(link_file_descriptor, __stack_log_file_path__, strlen(__stack_log_file_path__));
-                       if (written < (ssize_t)strlen(__stack_log_file_path__)) {
-                               _malloc_printf(ASL_LEVEL_INFO, "unable to write to stack reference file %s at %s\n",
-                                       stack_log_reference_file, stack_log_location);
-                       } else {
-                               const char *description_string = "\n(This is a reference file to the stack logs at the path above.)\n";
-                               write(link_file_descriptor, description_string, strlen(description_string));
-               }
-               }
-               close(link_file_descriptor);
-       }
-       if (use_alternate_link_location) {
-               _malloc_printf(ASL_LEVEL_INFO, "Please issue: cp %s %s\n", stack_log_reference_file, _PATH_TMP);
-       }
-       return created_log_location;
-}
-
-// Check to see if the log file is actually a reference to another location
-static int
-log_file_is_reference(char *log_location, char *out_reference_loc_buffer, size_t max_reference_path_size)
-{      
-       if (log_location == NULL || log_location[0] == '\0') return 0;
-       
-       size_t log_len = strlen(log_location);
-       size_t link_suffix_len = strlen(stack_log_link_suffix);
-       if (log_len < link_suffix_len || strncmp(log_location+log_len-link_suffix_len, stack_log_link_suffix, link_suffix_len) != 0) {
-               // not a reference file.
-               return 0;
-       }
-       
-       if (!out_reference_loc_buffer || max_reference_path_size == 0) return 1;
-       
-       FILE *reference_file = fopen(log_location, "r");
-       if (reference_file == NULL) {
-               // if unable to open the file, it may be because another user created it; no need to warn.
-               out_reference_loc_buffer[0] = '\0';
-               return 1;
-       }
-       
-       char *ret = fgets(out_reference_loc_buffer, (int)max_reference_path_size, reference_file);
-       if (!ret) {
-               out_reference_loc_buffer[0] = '\0';
-               _malloc_printf(ASL_LEVEL_INFO, "unable to read from stack logging reference file at %s\n", log_location);
-               return 1;
-       } else {
-               size_t read_line_len = strlen(out_reference_loc_buffer);
-               if (read_line_len >= 1 && out_reference_loc_buffer[read_line_len-1] == '\n') {
-                       out_reference_loc_buffer[read_line_len-1] = '\0';
-               }
-       }
-       
-       fclose(reference_file);
-       
-       return 1;
-}
-
-// This function may be called from either the target process when exiting, or from either the the target process or 
-// a stack log analysis process, when reaping orphaned stack log files.
-// Returns -1 if the files exist and they couldn't be removed, returns 0 otherwise.
-static int
-delete_logging_file(char *log_location)
-{
-       if (log_location == NULL || log_location[0] == '\0') return 0;
-
-       struct stat statbuf;
-       if (unlink(log_location) != 0 && stat(log_location, &statbuf) == 0) {
-               return -1;
-       }
-       return 0;
-}
-
-// This function will be called from atexit() in the target process.
-static void
-delete_log_files(void)
-{
-       if (__stack_log_file_path__ && __stack_log_file_path__[0]) {
-               if (delete_logging_file(__stack_log_file_path__) == 0) {
-                       _malloc_printf(ASL_LEVEL_INFO, "stack logs deleted from %s\n", __stack_log_file_path__);
-                       __stack_log_file_path__[0] = '\0';
-               } else {
-                       _malloc_printf(ASL_LEVEL_INFO, "unable to delete stack logs from %s\n", __stack_log_file_path__);
-               }
-       }
-       if (stack_log_reference_file && stack_log_reference_file[0]) {
-               delete_logging_file(stack_log_reference_file);
-       }
-}
-
-static bool
-is_process_running(pid_t pid)
-{
-       struct kinfo_proc kpt[1];
-       size_t size = sizeof(struct kinfo_proc);
-       int mib[] = {CTL_KERN, KERN_PROC, KERN_PROC_PID, pid};
-
-       sysctl(mib, 4, kpt, &size, NULL, (size_t)0); // size is either 1 or 0 entries when we ask for a single pid
-    
-       return (size==sizeof(struct kinfo_proc));
-}
-
-// The log files can be quite large and aren't too useful after the process that created them no longer exists.
-// Normally they should get removed when the process exits, but if the process crashed the log files might remain.
-// So, reap any stack log files for processes that no longer exist.
-//
-// lf the remove_for_this_pid flag is set, then any log files that already exist for the current process will also be deleted.
-// Those log files are probably the result of this process having been exec'ed from another one (without a fork()).
-// The remove_for_this_pid flag is only set for a target process (one just starting logging);  a stack logging "client"
-// process reaps log files too, but if we're using stack logging on the client process itself, then we don't want to remove
-// its own log files.
-static void
-reap_orphaned_log_files(bool remove_for_this_pid)
-{
-       DIR *dp;
-       struct dirent *entry;
-       char prefix_name[PATH_MAX];
-       char pathname[PATH_MAX];
-       pid_t current_pid = getpid();
-
-       if ((dp = opendir(_PATH_TMP)) == NULL) {
-               return;
-       }
-
-       strlcpy(prefix_name, stack_log_file_base_name, (size_t)PATH_MAX);
-       size_t prefix_length = strlen(prefix_name);
-
-       while ( (entry = readdir(dp)) != NULL ) {
-               if ( entry->d_type != DT_DIR && entry->d_type != DT_LNK && ( strncmp( entry->d_name, prefix_name, prefix_length) == 0 ) ) {
-                       long pid = strtol(&entry->d_name[prefix_length], (char **)NULL, 10);
-                       if ( (! is_process_running((pid_t)pid)) || (remove_for_this_pid && (pid_t)pid == current_pid) ) {
-                               strlcpy(pathname, _PATH_TMP, (size_t)PATH_MAX);
-                               strlcat(pathname, entry->d_name, (size_t)PATH_MAX);
-                               char reference_file_buffer[PATH_MAX];
-                               bool pathname_is_ref_file = false;
-                               if (log_file_is_reference(pathname, reference_file_buffer, (size_t)PATH_MAX) && *reference_file_buffer) {
-                                       pathname_is_ref_file = true;
-                                       if (delete_logging_file(reference_file_buffer) == 0) {
-                                               if (remove_for_this_pid && pid == current_pid) {
-                                                       _malloc_printf(ASL_LEVEL_INFO, "stack logs deleted from %s\n", reference_file_buffer);
-                                               } else {
-                                                       _malloc_printf(ASL_LEVEL_INFO, "process %ld no longer exists, stack logs deleted from %s\n", pid, reference_file_buffer);
-                                               }
-                                       }
-                               }
-                               if (delete_logging_file(pathname) == 0) {
-                                       if (remove_for_this_pid && pid == current_pid) {
-                                               if (!pathname_is_ref_file) _malloc_printf(ASL_LEVEL_INFO, "stack logs deleted from %s\n", pathname);
-                                       } else {
-                                               if (!pathname_is_ref_file) _malloc_printf(ASL_LEVEL_INFO, "process %ld no longer exists, stack logs deleted from %s\n", pid, pathname);
-                                       }
-                                       char shmem_name_string[PATH_MAX];
-                                       strlcpy(shmem_name_string, stack_log_file_base_name, (size_t)PATH_MAX);
-                                       append_int(shmem_name_string, (pid_t)pid, (size_t)PATH_MAX);
-                                       if (pid != current_pid) shm_unlink(shmem_name_string);
-                               }
-                       }
-               }
-       }
-       closedir(dp);
-}      
-
-/*
- * Since there a many errors that could cause stack logging to get disabled, this is a convenience method
- * for disabling any future logging in this process and for informing the user.
- */
-static void
-disable_stack_logging(void)
-{      
-       _malloc_printf(ASL_LEVEL_INFO, "stack logging disabled due to previous errors.\n");
-       stack_logging_enable_logging = 0;
-       malloc_logger = NULL;
-}
-
-/* A wrapper around write() that will try to reopen the index/stack file and
- * write to it if someone closed it underneath us (e.g. the process we just
- * started decide to close all file descriptors except stin/err/out). Some
- * programs like to do that and calling abort() on them is rude.
- */
-static ssize_t
-robust_write(int fd, const void *buf, size_t nbyte) {
-       extern int errno;
-       ssize_t written = write(fd, buf, nbyte);
-       if (written == -1 && errno == EBADF) {
-               char *file_to_reopen = NULL;
-               int *fd_to_reset = NULL;
-
-               // descriptor was closed on us. We need to reopen it
-               if (fd == index_file_descriptor) {
-                       file_to_reopen = __stack_log_file_path__;
-                       fd_to_reset = &index_file_descriptor;
-               } else {
-                       // We don't know about this file. Return (and abort()).
-                       _malloc_printf(ASL_LEVEL_INFO, "Unknown file descriptor; expecting stack logging index file\n");
-                       return -1;
-               }
-
-               // The file *should* already exist. If not, fail.
-               fd = open(file_to_reopen, O_WRONLY | O_APPEND);
-               if (fd < 3) {
-                       // If we somehow got stdin/out/err, we need to relinquish them and
-                       // get another fd.
-                       int fds_to_close[3] = { 0 };
-                       while (fd < 3) {
-                               if (fd == -1) {
-                                       _malloc_printf(ASL_LEVEL_INFO, "unable to re-open stack logging file %s\n", file_to_reopen);
-                                       delete_log_files();
-                                       return -1;
-                               }
-                               fds_to_close[fd] = 1;
-                               fd = dup(fd);
-                       }
-
-                       // We have an fd we like. Close the ones we opened.
-                       if (fds_to_close[0]) close(0);
-                       if (fds_to_close[1]) close(1);
-                       if (fds_to_close[2]) close(2);
-               }
-
-               *fd_to_reset = fd;
-               written = write(fd, buf, nbyte);
-       }
-       return written;
-}
-
-static void
-flush_data(void)
-{
-       ssize_t written; // signed size_t
-       size_t remaining;
-       char * p;
-       
-       if (index_file_descriptor == -1) {
-               if (create_log_file() == NULL) {
-                       return;
-               }
-       }
-       
-       // Write the events before the index so that hopefully the events will be on disk if the index refers to them.
-       p = pre_write_buffers->index_buffer;
-       remaining = (size_t)pre_write_buffers->next_free_index_buffer_offset;
-       while (remaining > 0) {
-               written = robust_write(index_file_descriptor, p, remaining);
-               if (written == -1) {
-                       _malloc_printf(ASL_LEVEL_INFO, "Unable to write to stack logging file %s (%s)\n",
-                               __stack_log_file_path__, strerror(errno));
-                       disable_stack_logging();
-                       return;
-               }
-               p += written;
-               remaining -= written;
-       }
-       
-       pre_write_buffers->start_index_offset += pre_write_buffers->next_free_index_buffer_offset;
-       pre_write_buffers->next_free_index_buffer_offset = 0;
-}
-
-static void
-prepare_to_log_stacks(void)
-{
-       if (!pre_write_buffers) {
-               last_logged_malloc_address = 0ul;
-               logging_use_compaction = (stack_logging_dontcompact ? 0 : logging_use_compaction);
-               
-               // Create a shared memory region to hold the pre-write index and stack buffers. This will allow remote analysis processes to access
-               // these buffers to get logs for even the most recent allocations. The remote process will need to pause this process to assure that
-               // the contents of these buffers don't change while being inspected.
-               char shmem_name_string[PATH_MAX];
-               strlcpy(shmem_name_string, stack_log_file_base_name, (size_t)PATH_MAX);
-               append_int(shmem_name_string, getpid(), (size_t)PATH_MAX);
-               
-               int shmid = shm_open(shmem_name_string, O_RDWR | O_CREAT, S_IRUSR | S_IWUSR);
-               if (shmid < 0) {
-                       // Failed to create shared memory region; turn off stack logging.
-                       _malloc_printf(ASL_LEVEL_INFO, "error while allocating shared memory for disk-based stack logging output buffers\n");
-                       disable_stack_logging();
-                       return;
-               }
-               
-               size_t full_shared_mem_size = sizeof(stack_buffer_shared_memory);
-               ftruncate(shmid, (off_t)full_shared_mem_size);
-               pre_write_buffers = (stack_buffer_shared_memory*)mmap(0, full_shared_mem_size, PROT_READ | PROT_WRITE, MAP_SHARED, shmid, (off_t)0);
-               close(shmid);
-               
-               if (MAP_FAILED == pre_write_buffers) {
-                       _malloc_printf(ASL_LEVEL_INFO, "error mapping in shared memory for disk-based stack logging output buffers\n");
-                       disable_stack_logging();
-                       return;
-               }
-
-               // Store and use the buffer offsets in shared memory so that they can be accessed remotely
-               pre_write_buffers->start_index_offset = 0ull;
-               pre_write_buffers->next_free_index_buffer_offset = 0;
-               
-               // create the backtrace uniquing table
-               pre_write_buffers->uniquing_table = __create_uniquing_table();
-               pre_write_buffers->uniquing_table_address = (mach_vm_address_t)(uintptr_t)pre_write_buffers->uniquing_table;
-               if (!pre_write_buffers->uniquing_table) {
-                       _malloc_printf(ASL_LEVEL_INFO, "error while allocating stack uniquing table\n");
-                       disable_stack_logging();
-                       return;
-               }
-
-               uint64_t stack_buffer_sz = (uint64_t)round_page(sizeof(vm_address_t) * STACK_LOGGING_MAX_STACK_SIZE);
-               stack_buffer = (vm_address_t*)allocate_pages(stack_buffer_sz);
-               if (!stack_buffer) {
-                       _malloc_printf(ASL_LEVEL_INFO, "error while allocating stack trace buffer\n");
-                       disable_stack_logging();
-                       return;
-               }
-               
-               // malloc() can be called by the following, so these need to be done outside the stack_logging_lock but after the buffers have been set up.
-               atexit(delete_log_files);               // atexit() can call malloc()
-               reap_orphaned_log_files(true);          // this calls opendir() which calls malloc()
-
-               // this call ensures that the log files exist; analyzing processes will rely on this assumption.
-               if (create_log_file() == NULL) {
-                       /* postponement support requires cleaning up these structures now */
-                       __destroy_uniquing_table(pre_write_buffers->uniquing_table);
-                       deallocate_pages(stack_buffer, stack_buffer_sz);
-                       stack_buffer = NULL;
-
-                       munmap(pre_write_buffers, full_shared_mem_size);
-                       pre_write_buffers = NULL;
-
-                       if (!stack_logging_postponed) {
-                               disable_stack_logging();
-                       }
-                       return;
-               }
-       }
-}
-
-void
-__disk_stack_logging_log_stack(uint32_t type_flags, uintptr_t zone_ptr, uintptr_t size, uintptr_t ptr_arg, uintptr_t return_val, uint32_t num_hot_to_skip)
-{
-       if (!stack_logging_enable_logging || stack_logging_postponed) return;
-       
-       // check incoming data
-       if (type_flags & stack_logging_type_alloc && type_flags & stack_logging_type_dealloc) {
-               uintptr_t swapper = size;
-               size = ptr_arg;
-               ptr_arg = swapper;
-               if (ptr_arg == return_val) return; // realloc had no effect, skipping
-               
-               if (ptr_arg == 0) { // realloc(NULL, size) same as malloc(size)
-                       type_flags ^= stack_logging_type_dealloc;
-               } else {
-                       // realloc(arg1, arg2) -> result is same as free(arg1); malloc(arg2) -> result
-                       __disk_stack_logging_log_stack(stack_logging_type_dealloc, zone_ptr, ptr_arg, (uintptr_t)0, (uintptr_t)0, num_hot_to_skip + 1);
-                       __disk_stack_logging_log_stack(stack_logging_type_alloc, zone_ptr, size, (uintptr_t)0, return_val, num_hot_to_skip + 1);
-                       return;
-               }
-       }
-       if (type_flags & stack_logging_type_dealloc) {
-               if (size) {
-                       ptr_arg = size;
-                       size = 0;
-               } else return; // free(nil)
-       }
-       if (type_flags & stack_logging_type_alloc && return_val == 0) return; // alloc that failed
-       
-       type_flags &= 0x7;
-       
-       // now actually begin
-       prepare_to_log_stacks();
-
-       // since there could have been a fatal (to stack logging) error such as the log files not being created, check this variable before continuing
-       if (!stack_logging_enable_logging || stack_logging_postponed) return;
-
-       vm_address_t self_thread = (vm_address_t)pthread_self();        // use pthread_self() rather than mach_thread_self() to avoid system call
-       
-       // lock and enter
-       OSSpinLockLock(&stack_logging_lock);
-       
-       if (!stack_logging_enable_logging) {
-               OSSpinLockUnlock(&stack_logging_lock);
-               return;
-       }
-       
-       // compaction
-       if (last_logged_malloc_address && (type_flags & stack_logging_type_dealloc) && STACK_LOGGING_DISGUISE(ptr_arg) == last_logged_malloc_address) {
-               // *waves hand* the last allocation never occurred
-               pre_write_buffers->next_free_index_buffer_offset -= (uint32_t)sizeof(stack_logging_index_event);
-               last_logged_malloc_address = 0ul;
-               
-               OSSpinLockUnlock(&stack_logging_lock);
-               return;
-       }
-
-       // gather stack
-       uint32_t count;
-       thread_stack_pcs(stack_buffer, STACK_LOGGING_MAX_STACK_SIZE-1, &count); // only gather up to STACK_LOGGING_MAX_STACK_SIZE-1 since we append thread id
-       stack_buffer[count++] = self_thread + 1;                // stuffing thread # in the coldest slot.  Add 1 to match what the old stack logging did.
-       num_hot_to_skip += 2;
-       if (count <= num_hot_to_skip) {
-               // Oops!  Didn't get a valid backtrace from thread_stack_pcs().
-               OSSpinLockUnlock(&stack_logging_lock);
-               return;
-       }
-               
-       // unique stack in memory
-       count -= num_hot_to_skip;
-#if __LP64__ 
-       mach_vm_address_t *frames = (mach_vm_address_t*)stack_buffer + num_hot_to_skip;
-#else
-       mach_vm_address_t frames[STACK_LOGGING_MAX_STACK_SIZE];
-       uint32_t i;
-       for (i = 0; i < count; i++) {
-               frames[i] = stack_buffer[i+num_hot_to_skip];
-       }
-#endif
-
-       uint64_t uniqueStackIdentifier = (uint64_t)(-1ll);
-       while (!__enter_frames_in_table(pre_write_buffers->uniquing_table, &uniqueStackIdentifier, frames, (int32_t)count)) {
-               __expand_uniquing_table(pre_write_buffers->uniquing_table);
-    }
-       
-       stack_logging_index_event current_index;
-       if (type_flags & stack_logging_type_alloc) {
-               current_index.address = STACK_LOGGING_DISGUISE(return_val);
-               current_index.argument = size;
-               if (logging_use_compaction) {
-                       last_logged_malloc_address = current_index.address; // disguised
-               }
-       } else {
-               current_index.address = STACK_LOGGING_DISGUISE(ptr_arg);
-               current_index.argument = 0ul;
-               last_logged_malloc_address = 0ul;
-       }
-       current_index.offset_and_flags = STACK_LOGGING_OFFSET_AND_FLAGS(uniqueStackIdentifier, type_flags);
-
-//     the following line is a good debugging tool for logging each allocation event as it happens.
-//     malloc_printf("{0x%lx, %lld}\n", STACK_LOGGING_DISGUISE(current_index.address), uniqueStackIdentifier);
-               
-       // flush the data buffer to disk if necessary
-       if (pre_write_buffers->next_free_index_buffer_offset + sizeof(stack_logging_index_event) >= STACK_LOGGING_BLOCK_WRITING_SIZE) {
-               flush_data();
-       }
-       
-       // store bytes in buffers
-       memcpy(pre_write_buffers->index_buffer+pre_write_buffers->next_free_index_buffer_offset, &current_index, sizeof(stack_logging_index_event));
-       pre_write_buffers->next_free_index_buffer_offset += (uint32_t)sizeof(stack_logging_index_event);
-       
-       OSSpinLockUnlock(&stack_logging_lock);
-}
-
-void 
-__stack_logging_fork_prepare() {
-       OSSpinLockLock(&stack_logging_lock);
-}
-
-void
-__stack_logging_fork_parent() {
-       OSSpinLockUnlock(&stack_logging_lock);
-}
-
-void
-__stack_logging_fork_child() {
-       malloc_logger = NULL;
-       stack_logging_enable_logging = 0;
-       OSSpinLockUnlock(&stack_logging_lock);
-}
-
-void
-__stack_logging_early_finished() {
-       stack_logging_finished_init = 1;
-       stack_logging_postponed = 0;
-}
-
-boolean_t
-__stack_logging_locked()
-{
-       bool acquired_lock = OSSpinLockTry(&stack_logging_lock);
-       if (acquired_lock) OSSpinLockUnlock(&stack_logging_lock);
-       return (acquired_lock ? false : true);
-}
-
-#pragma mark -
-#pragma mark Remote Stack Log Access
-
-#pragma mark - Design notes:
-
-/*
-
-this first one will look through the index, find the "stack_identifier" (i.e. the offset in the log file), and call the third function listed here.
-extern kern_return_t __mach_stack_logging_get_frames(task_t task, mach_vm_address_t address, mach_vm_address_t *stack_frames_buffer, uint32_t max_stack_frames, uint32_t *num_frames);
-    //  Gets the last allocation record about address
-       
-if !address, will load index and iterate through (expensive)
-else will load just index, search for stack, and then use third function here to retrieve. (also expensive)
-extern kern_return_t __mach_stack_logging_enumerate_records(task_t task, mach_vm_address_t address, void enumerator(mach_stack_logging_record_t, void *), void *context);
-    // Applies enumerator to all records involving address sending context as enumerator's second parameter; if !address, applies enumerator to all records 
-
-this function will load the stack file, look for the stack, and follow up to STACK_LOGGING_FORCE_FULL_BACKTRACE_EVERY references to reconstruct.
-extern kern_return_t __mach_stack_logging_frames_for_uniqued_stack(task_t task, uint64_t stack_identifier, mach_vm_address_t *stack_frames_buffer, uint32_t max_stack_frames, uint32_t *count);
-    // Given a uniqued_stack fills stack_frames_buffer
-
-*/
-
-#pragma mark - caching
-
-__attribute__((always_inline)) static inline size_t 
-hash_index(uint64_t address, size_t max_pos) {
-       return (size_t)((address >> 2) % (max_pos-1)); // simplicity rules.
-}
-
-__attribute__((always_inline)) static inline size_t 
-hash_multiplier(size_t capacity, uint32_t allowed_collisions) {
-       return (capacity/(allowed_collisions*2+1));
-}
-
-__attribute__((always_inline)) static inline size_t 
-next_hash(size_t hash, size_t multiplier, size_t capacity, uint32_t collisions) {
-       hash += multiplier * collisions;
-       if (hash >= capacity) hash -= capacity;
-       return hash;
-}
-
-static void
-transfer_node(remote_index_cache *cache, remote_index_node *old_node)
-{
-       uint32_t collisions = 0;
-       size_t pos = hash_index(old_node->address, cache->cache_node_capacity);
-       size_t multiplier = hash_multiplier(cache->cache_node_capacity, cache->collision_allowance);
-       do {
-               if (cache->table_memory[pos].address == old_node->address) { // hit like this shouldn't happen.
-                       fprintf(stderr, "impossible collision! two address==address lists! (transfer_node)\n");
-                       break;
-               } else if (cache->table_memory[pos].address == 0) { // empty
-                       cache->table_memory[pos] = *old_node;
-                       break;
-               } else {
-                       collisions++;
-                       pos = next_hash(pos, multiplier, cache->cache_node_capacity, collisions);
-               }
-       } while (collisions <= cache->collision_allowance);
-       
-       if (collisions > cache->collision_allowance) {
-               fprintf(stderr, "reporting bad hash function! disk stack logging reader %lu bit. (transfer_node)\n", sizeof(void*)*8);
-       }
-}
-
-static void
-expand_cache(remote_index_cache *cache)
-{
-       // keep old stats
-       size_t old_node_capacity = cache->cache_node_capacity;
-       remote_index_node *old_table = cache->table_memory;
-       
-       // double size
-       cache->cache_size <<= 2;
-       cache->cache_node_capacity <<= 2;
-       cache->collision_allowance += 3;
-       cache->table_memory = (void*)calloc(cache->cache_node_capacity, sizeof(remote_index_node));
-               
-       // repopulate (expensive!)
-       size_t i;
-       for (i = 0; i < old_node_capacity; i++) {
-               if (old_table[i].address) {
-                       transfer_node(cache, &old_table[i]);
-               }
-       }       
-       free(old_table);
-//     printf("cache expanded to %0.2f mb (eff: %3.0f%%, capacity: %lu, nodes: %llu, llnodes: %llu)\n", ((float)(cache->cache_size))/(1 << 20), ((float)(cache->cache_node_count)*100.0)/((float)(cache->cache_node_capacity)), cache->cache_node_capacity, cache->cache_node_count, cache->cache_llnode_count);
-}
-
-static void
-insert_node(remote_index_cache *cache, uint64_t address, uint64_t index_file_offset)
-{
-       uint32_t collisions = 0;
-       size_t pos = hash_index(address, cache->cache_node_capacity);
-       size_t multiplier = hash_multiplier(cache->cache_node_capacity, cache->collision_allowance);
-
-       bool inserted = false;
-       while (!inserted) {
-               if (cache->table_memory[pos].address == 0ull || cache->table_memory[pos].address == address) { // hit or empty
-                       cache->table_memory[pos].address = address;
-                       cache->table_memory[pos].index_file_offset = index_file_offset;
-                       inserted = true;
-                       break;
-               }
-               
-               collisions++;
-               pos = next_hash(pos, multiplier, cache->cache_node_capacity, collisions);
-
-               if (collisions > cache->collision_allowance) {
-                       expand_cache(cache);
-                       pos = hash_index(address, cache->cache_node_capacity);
-                       multiplier = hash_multiplier(cache->cache_node_capacity, cache->collision_allowance);
-                       collisions = 0;
-               }
-       }
-       
-}
-
-static void
-update_cache_for_file_streams(remote_task_file_streams *descriptors)
-{
-       remote_index_cache *cache = descriptors->cache;
-       
-       // create from scratch if necessary.
-       if (!cache) {
-               descriptors->cache = cache = (remote_index_cache*)calloc((size_t)1, sizeof(remote_index_cache));
-               cache->cache_node_capacity = 1 << 14;
-               cache->collision_allowance = 17;
-               cache->last_index_file_offset = 0;
-               cache->cache_size = cache->cache_node_capacity*sizeof(remote_index_node);
-               cache->table_memory = (void*)calloc(cache->cache_node_capacity, sizeof(remote_index_node));
-               
-               // now map in the shared memory, if possible
-               char shmem_name_string[PATH_MAX];
-               strlcpy(shmem_name_string, stack_log_file_base_name, (size_t)PATH_MAX);
-               append_int(shmem_name_string, descriptors->remote_pid, (size_t)PATH_MAX);
-               
-               int shmid = shm_open(shmem_name_string, O_RDWR, S_IRUSR | S_IWUSR);
-               if (shmid >= 0) {
-                       cache->shmem = mmap(0, sizeof(stack_buffer_shared_memory), PROT_READ | PROT_WRITE, MAP_SHARED, shmid, (off_t)0);
-                       close(shmid);
-               }
-               
-               if (shmid < 0 || cache->shmem == MAP_FAILED) {
-                       // failed to connect to the shared memory region; warn and continue.
-                       _malloc_printf(ASL_LEVEL_INFO, "warning: unable to connect to remote process' shared memory; allocation histories may not be up-to-date.\n");
-               }
-       }
-       
-       // suspend and see how much updating there is to do. there are three scenarios, listed below
-       bool update_snapshot = false;
-       if (descriptors->remote_task != mach_task_self()) {
-               task_suspend(descriptors->remote_task);
-       }
-               
-       struct stat file_statistics;
-       fstat(fileno(descriptors->index_file_stream), &file_statistics);
-       size_t read_size = (descriptors->task_is_64_bit ? sizeof(stack_logging_index_event64) : sizeof(stack_logging_index_event32));
-       uint64_t read_this_update = 0;
-       
-       // the delta indecies is a complex number; there are three cases:
-       // 1. there is no shared memory (or we can't connect); diff the last_index_file_offset from the filesize.
-       // 2. the only updates have been in shared memory; disk file didn't change at all. delta_indecies should be zero, scan snapshot only.
-       // 3. the updates have flushed to disk, meaning that most likely there is new data on disk that wasn't read from shared memory.
-       //    correct delta_indecies for the pre-scanned amount and read the new data from disk and shmem.
-       uint64_t delta_indecies = (file_statistics.st_size - cache->last_index_file_offset) / read_size;
-       uint32_t last_snapshot_scan_index = 0;
-       if (delta_indecies && cache->shmem) { 
-               // case 3: add cache scanned to known from disk and recalc
-               cache->last_index_file_offset += cache->snapshot.next_free_index_buffer_offset;
-               delta_indecies = (file_statistics.st_size - cache->last_index_file_offset) / read_size;
-               update_snapshot = true;
-       } else if (cache->shmem) {
-               // case 2: set the last snapshot scan count so we don't rescan something we've seen.
-               last_snapshot_scan_index = cache->snapshot.next_free_index_buffer_offset / (uint32_t)read_size;
-       }
-       
-       // no update necessary for the file; check if need a snapshot.
-       if (delta_indecies == 0) {
-               if (cache->shmem && !update_snapshot) {
-                       update_snapshot = (cache->shmem->next_free_index_buffer_offset != cache->snapshot.next_free_index_buffer_offset);
-               }
-       }
-       
-       // if a snapshot is necessary, memcpy from remote frozen process' memory 
-       // note: there were two ways to do this - spin lock or suspend. suspend allows us to
-       // analyze processes even if they were artificially suspended. with a lock, there'd be
-       // worry that the target was suspended with the lock taken.
-       if (update_snapshot) {
-               memcpy(&cache->snapshot, cache->shmem, sizeof(stack_buffer_shared_memory));
-               // also need to update our version of the remote uniquing table
-               vm_address_t local_uniquing_address = 0ul;
-               mach_msg_type_number_t local_uniquing_size = 0;
-               mach_vm_size_t desired_size = round_page(sizeof(backtrace_uniquing_table));
-               kern_return_t err;
-               if ((err = mach_vm_read(descriptors->remote_task, cache->shmem->uniquing_table_address, desired_size, &local_uniquing_address, &local_uniquing_size)) != KERN_SUCCESS 
-                               || local_uniquing_size != desired_size) {
-                       fprintf(stderr, "error while attempting to mach_vm_read remote stack uniquing table (%d): %s\n", err, mach_error_string(err));
-               } else {
-                       // the mach_vm_read was successful, so acquire the uniquing table
-                       
-                       // need to re-read the table, so deallocate the current memory
-                       if (cache->uniquing_table.table) mach_vm_deallocate(mach_task_self(), (mach_vm_address_t)(uintptr_t)(cache->uniquing_table.table), cache->uniquing_table.tableSize);
-               
-                       // the following line gathers the uniquing table structure data, but the actual table memory is invalid since it's a pointer from the
-                       // remote process. this pointer will be mapped shared in a few lines.
-                       cache->uniquing_table = *((backtrace_uniquing_table*)local_uniquing_address);
-                       
-                       vm_address_t local_table_address = 0ul;
-                       mach_msg_type_number_t local_table_size = 0;
-                       
-                       err = mach_vm_read(descriptors->remote_task, cache->uniquing_table.table_address, cache->uniquing_table.tableSize, &local_table_address, &local_table_size);
-                       if (err == KERN_SUCCESS) cache->uniquing_table.table = (mach_vm_address_t*)local_table_address;
-                       else cache->uniquing_table.table = NULL;
-                       
-                       mach_vm_deallocate(mach_task_self(), (mach_vm_address_t)local_uniquing_address, (mach_vm_size_t)local_uniquing_size);
-               }
-       }
-       
-       // resume
-       if (descriptors->remote_task != mach_task_self()) {
-               task_resume(descriptors->remote_task);
-       }
-       
-       if (!update_snapshot && delta_indecies == 0) return; // absolutely no updating needed.
-       
-       FILE *the_index = (descriptors->index_file_stream);
-       
-       // prepare for the read; target process could be 32 or 64 bit.
-
-       stack_logging_index_event32 *target_32_index = NULL;
-       stack_logging_index_event64 *target_64_index = NULL;
-       
-       // perform the update from the file
-       uint32_t i;
-       if (delta_indecies) {
-               char bufferSpace[4096]; // 4 kb
-               target_32_index = (stack_logging_index_event32*)bufferSpace;
-               target_64_index = (stack_logging_index_event64*)bufferSpace;
-               size_t number_slots = (size_t)(4096/read_size);
-       
-               size_t read_count = 0;
-               if (fseeko(the_index, (off_t)(cache->last_index_file_offset), SEEK_SET)) {
-                       fprintf(stderr, "error while attempting to cache information from remote stack index file. (update_cache_for_file_streams)\n");
-               }
-               off_t current_index_position = cache->last_index_file_offset;
-               do {
-                       number_slots = (size_t)MIN(delta_indecies - read_this_update, number_slots);
-                       read_count = fread(bufferSpace, read_size, number_slots, the_index);
-                       if (descriptors->task_is_64_bit) {
-                               for (i = 0; i < read_count; i++) {
-                                       insert_node(cache, STACK_LOGGING_DISGUISE(target_64_index[i].address), (uint64_t)current_index_position);
-                                       read_this_update++;
-                                       current_index_position += read_size;
-                               }
-                       } else {
-                               for (i = 0; i < read_count; i++) {
-                                       insert_node(cache, (mach_vm_address_t)STACK_LOGGING_DISGUISE(target_32_index[i].address), (uint64_t)current_index_position);
-                                       read_this_update++;
-                                       current_index_position += read_size;
-                               }
-                       }
-               } while (read_count);
-
-               if (read_this_update < delta_indecies) {
-                       fprintf(stderr, "insufficient data in remote stack index file; expected more records.\n");
-               }
-               cache->last_index_file_offset += read_this_update * read_size;
-       }
-       
-       if (update_snapshot) {
-               target_32_index = (stack_logging_index_event32*)(cache->snapshot.index_buffer);
-               target_64_index = (stack_logging_index_event64*)(cache->snapshot.index_buffer);
-               
-               uint32_t free_snapshot_scan_index = cache->snapshot.next_free_index_buffer_offset / (uint32_t)read_size;
-               off_t current_index_position = cache->snapshot.start_index_offset;
-               if (descriptors->task_is_64_bit) {
-                       for (i = last_snapshot_scan_index; i < free_snapshot_scan_index; i++) {
-                               insert_node(cache, STACK_LOGGING_DISGUISE(target_64_index[i].address), (uint64_t)(current_index_position + (i * read_size))); 
-                       }
-               } else {
-                       for (i = last_snapshot_scan_index; i < free_snapshot_scan_index; i++) {
-                               insert_node(cache, (mach_vm_address_t)STACK_LOGGING_DISGUISE(target_32_index[i].address), (uint64_t)(current_index_position + (i * read_size)));
-                       }
-               }
-       }
-}
-
-static void
-destroy_cache_for_file_streams(remote_task_file_streams *descriptors)
-{
-       if (descriptors->cache->shmem) {
-               munmap(descriptors->cache->shmem, sizeof(stack_buffer_shared_memory));
-       }
-       free(descriptors->cache->table_memory);
-       free(descriptors->cache);
-       descriptors->cache = NULL;
-}
-
-#pragma mark - internal
-
-// In the stack log analysis process, find the stack logging files for target process <pid>
-// by scanning the temporary directory for directory entries with names of the form "stack-logs.<pid>."
-// If we find such a directory then open the stack logging files in there.
-// We might also have been passed the file path if the client first read it from __stack_log_file_path__
-// global variable in the target task, as will be needed if the .link cannot be put in /tmp.
-static void
-open_log_files(pid_t pid, char* file_path, remote_task_file_streams *this_task_streams)
-{
-       DIR *dp;
-       struct dirent *entry;
-       char prefix_name[PATH_MAX];
-       char pathname[PATH_MAX];
-
-       reap_orphaned_log_files(false);         // reap any left-over log files (for non-existant processes, but not for this analysis process)
-
-       if (file_path != NULL) {
-               this_task_streams->index_file_stream = fopen(file_path, "r");
-               return;
-       }
-
-       if ((dp = opendir(_PATH_TMP)) == NULL) {
-               return;
-       }
-
-       // It's OK to use snprintf in this routine since it should only be called by the clients
-       // of stack logging, and thus calls to malloc are OK.
-       snprintf(prefix_name, (size_t)PATH_MAX, "%s%d.", stack_log_file_base_name, pid);        // make sure to use "%s%d." rather than just "%s%d" to match the whole pid
-       size_t prefix_length = strlen(prefix_name);
-
-       while ( (entry = readdir(dp)) != NULL ) {
-               if ( strncmp( entry->d_name, prefix_name, prefix_length) == 0 ) {
-                       snprintf(pathname, (size_t)PATH_MAX, "%s%s", _PATH_TMP, entry->d_name);
-                       char reference_file[PATH_MAX];
-                       if (log_file_is_reference(pathname, reference_file, (size_t)PATH_MAX)) {
-                               this_task_streams->index_file_stream = fopen(reference_file, "r");
-                       } else {
-                               this_task_streams->index_file_stream = fopen(pathname, "r");
-                       }
-                       
-                       break;
-               }
-       }
-       closedir(dp);
-}      
-
-static remote_task_file_streams*
-retain_file_streams_for_task(task_t task, char* file_path)
-{
-       if (task == MACH_PORT_NULL) return NULL;
-       
-       OSSpinLockLock(&remote_fd_list_lock);
-       
-       // see if they're already in use
-       uint32_t i = 0;
-       for (i = 0; i < remote_task_fd_count; i++) {
-               if (remote_fds[i].remote_task == task) {
-                       remote_fds[i].in_use_count++;
-                       OSSpinLockUnlock(&remote_fd_list_lock);
-                       return &remote_fds[i];
-               }
-       }
-       
-       // open them
-       uint32_t failures = 0;
-       if (remote_task_fd_count == STACK_LOGGING_MAX_SIMUL_REMOTE_TASKS_INSPECTED) {
-               while (remote_fds[next_remote_task_fd].in_use_count > 0) {
-                       next_remote_task_fd++;
-                       if (next_remote_task_fd == STACK_LOGGING_MAX_SIMUL_REMOTE_TASKS_INSPECTED) next_remote_task_fd = 0;
-                       failures++;
-                       if (failures >= STACK_LOGGING_MAX_SIMUL_REMOTE_TASKS_INSPECTED) {
-                               OSSpinLockUnlock(&remote_fd_list_lock);
-                               return NULL;
-                       }
-               }
-               fclose(remote_fds[next_remote_task_fd].index_file_stream);
-               destroy_cache_for_file_streams(&remote_fds[next_remote_task_fd]);
-       }
-       
-       pid_t pid;
-       kern_return_t err = pid_for_task(task, &pid);
-       if (err != KERN_SUCCESS) {
-               OSSpinLockUnlock(&remote_fd_list_lock);
-               return NULL;
-       }
-       
-       remote_task_file_streams *this_task_streams = &remote_fds[next_remote_task_fd];
-       
-       open_log_files(pid, file_path, this_task_streams);
-
-       // check if opens failed
-       if (this_task_streams->index_file_stream == NULL) {
-               if (this_task_streams->index_file_stream) fclose(this_task_streams->index_file_stream);
-               OSSpinLockUnlock(&remote_fd_list_lock);
-               return NULL;
-       }
-       
-       // check if target pid is running 64-bit
-       int mib[] = { CTL_KERN, KERN_PROC, KERN_PROC_PID, pid };
-       struct kinfo_proc processInfo;
-       size_t bufsize = sizeof(processInfo);
-       if (sysctl(mib, (unsigned)(sizeof(mib)/sizeof(int)), &processInfo, &bufsize, NULL, (size_t)0) == 0 && bufsize > 0) {
-               this_task_streams->task_is_64_bit = processInfo.kp_proc.p_flag & P_LP64;
-       } else {
-               this_task_streams->task_is_64_bit = 0;
-       }
-       
-       // otherwise set vars and go
-       this_task_streams->in_use_count = 1;
-       this_task_streams->remote_task = task;
-       this_task_streams->remote_pid = pid;
-       next_remote_task_fd++;
-       if (next_remote_task_fd == STACK_LOGGING_MAX_SIMUL_REMOTE_TASKS_INSPECTED) next_remote_task_fd = 0;
-       remote_task_fd_count = MIN(remote_task_fd_count + 1, STACK_LOGGING_MAX_SIMUL_REMOTE_TASKS_INSPECTED);
-       
-       OSSpinLockUnlock(&remote_fd_list_lock);
-       return this_task_streams;
-}
-
-static void
-release_file_streams_for_task(task_t task)
-{
-       OSSpinLockLock(&remote_fd_list_lock);
-       
-       // decrement in-use count
-       uint32_t i = 0;
-       for (i = 0; i < remote_task_fd_count; i++) {
-               if (remote_fds[i].remote_task == task) {
-                       remote_fds[i].in_use_count--;
-                       break;
-               }
-       }
-       
-       OSSpinLockUnlock(&remote_fd_list_lock);
-}
-
-#pragma mark - extern
-
-//
-// The following is used by client tools like malloc_history and Instruments to pass along the path
-// of the index file as read from the target task's __stack_log_file_path__ variable (set in this file)
-// Eventually, at a suitable point, this additional argument should just be added to the other APIs below.
-//
-kern_return_t
-__mach_stack_logging_set_file_path(task_t task, char* file_path)
-{
-       remote_task_file_streams *remote_fd = retain_file_streams_for_task(task, file_path);
-       if (remote_fd == NULL) {
-               return KERN_FAILURE;
-       }
-       return KERN_SUCCESS;
-}
-
-kern_return_t
-__mach_stack_logging_get_frames(task_t task, mach_vm_address_t address, mach_vm_address_t *stack_frames_buffer, uint32_t max_stack_frames, uint32_t *count)
-{
-       remote_task_file_streams *remote_fd = retain_file_streams_for_task(task, NULL);
-       if (remote_fd == NULL) {
-               return KERN_FAILURE;
-       }
-       
-       update_cache_for_file_streams(remote_fd);
-       
-       uint32_t collisions = 0;
-       size_t hash = hash_index(address, remote_fd->cache->cache_node_capacity);
-       size_t multiplier = hash_multiplier(remote_fd->cache->cache_node_capacity, remote_fd->cache->collision_allowance);
-       uint64_t located_file_position = 0;
-       
-       bool found = false;     
-       do {
-               if (remote_fd->cache->table_memory[hash].address == address) { // hit!
-                       located_file_position = remote_fd->cache->table_memory[hash].index_file_offset;
-                       found = true;
-                       break;
-               } else if (remote_fd->cache->table_memory[hash].address == 0ull) { // failure!
-                       break;
-               }
-               
-               collisions++;
-               hash = next_hash(hash, multiplier, remote_fd->cache->cache_node_capacity, collisions);
-               
-       } while (collisions <= remote_fd->cache->collision_allowance);
-       
-       if (found) {
-               // prepare for the read; target process could be 32 or 64 bit.
-               stack_logging_index_event32 *target_32_index = NULL;
-               stack_logging_index_event64 *target_64_index = NULL;
-               
-               if (located_file_position >= remote_fd->cache->last_index_file_offset) {
-                       // must be in shared memory
-                       if (remote_fd->cache->shmem) {
-                               if (remote_fd->task_is_64_bit) {
-                                       target_64_index = (stack_logging_index_event64*)(remote_fd->cache->snapshot.index_buffer + (located_file_position - remote_fd->cache->snapshot.start_index_offset));
-                                       located_file_position = STACK_LOGGING_OFFSET(target_64_index->offset_and_flags);
-                               } else {
-                                       target_32_index = (stack_logging_index_event32*)(remote_fd->cache->snapshot.index_buffer + (located_file_position - remote_fd->cache->snapshot.start_index_offset));
-                                       located_file_position = STACK_LOGGING_OFFSET(target_32_index->offset_and_flags);                                
-                               }
-                       } else {
-                               found = false;
-                       }
-               
-               } else {
-                       // it's written to disk
-                       char bufferSpace[128];
-                       
-                       size_t read_size = (remote_fd->task_is_64_bit ? sizeof(stack_logging_index_event64) : sizeof(stack_logging_index_event32));
-                       fseeko(remote_fd->index_file_stream, (off_t)located_file_position, SEEK_SET);
-                       size_t read_count = fread(bufferSpace, read_size, (size_t)1, remote_fd->index_file_stream);
-                       if (read_count) {
-                               if (remote_fd->task_is_64_bit) {
-                                       target_64_index = (stack_logging_index_event64*)bufferSpace;
-                                       located_file_position = STACK_LOGGING_OFFSET(target_64_index->offset_and_flags);
-                               } else {
-                                       target_32_index = (stack_logging_index_event32*)bufferSpace;
-                                       located_file_position = STACK_LOGGING_OFFSET(target_32_index->offset_and_flags);
-                               }
-                       } else {
-                               found = false;
-                       }
-               }
-       }
-       
-       release_file_streams_for_task(task);
-       
-       if (!found) {
-               return KERN_FAILURE;
-       }
-       
-       return __mach_stack_logging_frames_for_uniqued_stack(task, located_file_position, stack_frames_buffer, max_stack_frames, count);
-}
-
-
-kern_return_t
-__mach_stack_logging_enumerate_records(task_t task, mach_vm_address_t address, void enumerator(mach_stack_logging_record_t, void *), void *context)
-{
-       remote_task_file_streams *remote_fd = retain_file_streams_for_task(task, NULL);
-       if (remote_fd == NULL) {
-               return KERN_FAILURE;
-       }
-       
-       bool reading_all_addresses = (address == 0 ? true : false);
-       mach_stack_logging_record_t pass_record;
-       kern_return_t err = KERN_SUCCESS;
-       
-       // update (read index file once and only once)
-       update_cache_for_file_streams(remote_fd);
-
-       FILE *the_index = (remote_fd->index_file_stream);
-       
-       // prepare for the read; target process could be 32 or 64 bit.
-       char bufferSpace[2048]; // 2 kb
-       stack_logging_index_event32 *target_32_index = (stack_logging_index_event32*)bufferSpace;
-       stack_logging_index_event64 *target_64_index = (stack_logging_index_event64*)bufferSpace;
-       uint32_t target_addr_32 = (uint32_t)STACK_LOGGING_DISGUISE((uint32_t)address);
-       uint64_t target_addr_64 = STACK_LOGGING_DISGUISE((uint64_t)address);
-       size_t read_size = (remote_fd->task_is_64_bit ? sizeof(stack_logging_index_event64) : sizeof(stack_logging_index_event32));
-       size_t number_slots = (size_t)(2048/read_size);
-       uint64_t total_slots = remote_fd->cache->last_index_file_offset / read_size;
-       
-       // perform the search
-       size_t read_count = 0;
-       int64_t current_file_offset = 0;
-       uint32_t i;
-       do {
-               // at this point, we need to read index events; read them from the file until it's necessary to grab them from the shared memory snapshot
-               // and crop file reading to the point where we last scanned
-               number_slots = (size_t)MIN(number_slots, total_slots);
-               
-               // if out of file to read (as of the time we entered this function), try to use shared memory snapshot
-               if (number_slots == 0) {
-                       if (remote_fd->cache->shmem && remote_fd->cache->snapshot.start_index_offset + remote_fd->cache->snapshot.next_free_index_buffer_offset > (uint64_t)current_file_offset) {
-                               // use shared memory
-                               target_32_index = (stack_logging_index_event32*)remote_fd->cache->snapshot.index_buffer;
-                               target_64_index = (stack_logging_index_event64*)remote_fd->cache->snapshot.index_buffer;
-                               read_count = (uint32_t)(remote_fd->cache->snapshot.start_index_offset + remote_fd->cache->snapshot.next_free_index_buffer_offset - current_file_offset) / read_size;
-                               current_file_offset += read_count * read_size;
-                       } else {
-                               break;
-                       }
-               } else {
-                       // get and save index (enumerator could modify)
-                       fseeko(the_index, current_file_offset, SEEK_SET);
-                       read_count = fread(bufferSpace, read_size, number_slots, the_index);
-                       current_file_offset = ftello(the_index);
-                       total_slots -= read_count;
-               }
-               
-               if (remote_fd->task_is_64_bit) {
-                       for (i = 0; i < read_count; i++) {
-                               if (reading_all_addresses || target_64_index[i].address == target_addr_64) {
-                                       pass_record.address = STACK_LOGGING_DISGUISE(target_64_index[i].address);
-                                       pass_record.argument = target_64_index[i].argument;
-                                       pass_record.stack_identifier = STACK_LOGGING_OFFSET(target_64_index[i].offset_and_flags);
-                                       pass_record.type_flags = STACK_LOGGING_FLAGS(target_64_index[i].offset_and_flags);
-                                       enumerator(pass_record, context);
-                               }
-                       }
-               } else {
-                       for (i = 0; i < read_count; i++) {
-                               if (reading_all_addresses || target_32_index[i].address == target_addr_32) {
-                                       pass_record.address = STACK_LOGGING_DISGUISE(target_32_index[i].address);
-                                       pass_record.argument = target_32_index[i].argument;
-                                       pass_record.stack_identifier = STACK_LOGGING_OFFSET(target_32_index[i].offset_and_flags);
-                                       pass_record.type_flags = STACK_LOGGING_FLAGS(target_32_index[i].offset_and_flags);
-                                       enumerator(pass_record, context);
-                               }
-                       }
-               }
-       } while (read_count);
-       
-       release_file_streams_for_task(task);
-       return err;
-}
-
-
-kern_return_t
-__mach_stack_logging_frames_for_uniqued_stack(task_t task, uint64_t stack_identifier, mach_vm_address_t *stack_frames_buffer, uint32_t max_stack_frames, uint32_t *count)
-{
-       remote_task_file_streams *remote_fd = retain_file_streams_for_task(task, NULL);
-       if (remote_fd == NULL) return KERN_FAILURE;
-
-       __unwind_stack_from_table_index(&remote_fd->cache->uniquing_table, stack_identifier, stack_frames_buffer, count, max_stack_frames);     
-       
-       release_file_streams_for_task(task);
-       
-       if (*count) return KERN_SUCCESS;
-       else return KERN_FAILURE;
-}
-
-
-#ifdef TEST_DISK_STACK_LOGGING
-
-// cc -o stack_logging_disk stack_logging_disk.c -DTEST_DISK_STACK_LOGGING
-
-#include <sys/wait.h>
-
-int
-main()
-{
-  int status;
-  int i;
-  size_t total_globals = 0ul;
-
-  fprintf(stderr, "master test process is %d\n", getpid());
-  fprintf(stderr, "sizeof pre_write_buffers: %lu\n", sizeof(pre_write_buffers)); total_globals += sizeof(pre_write_buffers);
-  fprintf(stderr, "sizeof stack_buffer: %lu\n", sizeof(stack_buffer)); total_globals += sizeof(stack_buffer);
-  fprintf(stderr, "sizeof last_logged_malloc_address: %lu\n", sizeof(last_logged_malloc_address)); total_globals += sizeof(last_logged_malloc_address);
-  fprintf(stderr, "sizeof stack_log_file_base_name: %lu\n", sizeof(stack_log_file_base_name)); total_globals += sizeof(stack_log_file_base_name);
-  fprintf(stderr, "sizeof stack_log_file_suffix: %lu\n", sizeof(stack_log_file_suffix)); total_globals += sizeof(stack_log_file_suffix);
-  fprintf(stderr, "sizeof stack_log_link_suffix: %lu\n", sizeof(stack_log_link_suffix)); total_globals += sizeof(stack_log_link_suffix);
-  fprintf(stderr, "sizeof stack_log_location: %lu\n", (size_t)PATH_MAX); total_globals += (size_t)PATH_MAX;
-  fprintf(stderr, "sizeof stack_log_reference_file: %lu\n", (size_t)PATH_MAX); total_globals += (size_t)PATH_MAX;
-  fprintf(stderr, "sizeof __stack_log_file_path__ (index_file_path): %lu\n", (size_t)PATH_MAX); total_globals += (size_t)PATH_MAX;
-  fprintf(stderr, "sizeof index_file_descriptor: %lu\n", sizeof(index_file_descriptor)); total_globals += sizeof(index_file_descriptor);
-  fprintf(stderr, "sizeof remote_fds: %lu\n", sizeof(remote_fds)); total_globals += sizeof(remote_fds);
-  fprintf(stderr, "sizeof next_remote_task_fd: %lu\n", sizeof(next_remote_task_fd)); total_globals += sizeof(next_remote_task_fd);
-  fprintf(stderr, "sizeof remote_task_fd_count: %lu\n", sizeof(remote_task_fd_count)); total_globals += sizeof(remote_task_fd_count);
-  fprintf(stderr, "sizeof remote_fd_list_lock: %lu\n", sizeof(remote_fd_list_lock)); total_globals += sizeof(remote_fd_list_lock);
-  fprintf(stderr, "sizeof logging_use_compaction: %lu\n", sizeof(logging_use_compaction)); total_globals += sizeof(logging_use_compaction);
-
-  fprintf(stderr, "size of all global data: %lu\n", total_globals);
-
-  create_log_file();
-
-  // create a few child processes and exit them cleanly so their logs should get cleaned up
-  fprintf(stderr, "\ncreating child processes and exiting cleanly\n");
-  for (i = 0; i < 3; i++) {
-    if (fork() == 0) {
-      fprintf(stderr, "\nin child processes %d\n", getpid());
-      create_log_file();
-      fprintf(stderr, "exiting child processes %d\n", getpid());
-      exit(1);
-    }
-    wait(&status);
-  }
-
-  // create a few child processes and abruptly _exit them, leaving their logs around
-  fprintf(stderr, "\ncreating child processes and exiting abruptly, leaving logs around\n");
-  for (i = 0; i < 3; i++) {
-    if (fork() == 0) {
-      fprintf(stderr, "\nin child processes %d\n", getpid());
-      create_log_file();
-      fprintf(stderr, "exiting child processes %d\n", getpid());
-      _exit(1);
-    }
-    wait(&status);
-  }
-
-  // this should reap any remaining logs
-  fprintf(stderr, "\nexiting master test process %d\n", getpid());
-  delete_log_files();
-  return 0;
-}
-
-#endif