Loading...
--- libmalloc/libmalloc-67/src/malloc.c
+++ libmalloc/libmalloc-657.80.3/src/malloc.c
@@ -21,78 +21,29 @@
* @APPLE_LICENSE_HEADER_END@
*/
-#include "magmallocProvider.h"
-
-#include <_simple.h>
-#include <crt_externs.h>
-#include <dlfcn.h>
-#include <errno.h>
-#include <fcntl.h>
-#include <limits.h>
-#include <stdlib.h>
-#include <stdio.h>
-#include <string.h>
-#include <unistd.h>
-#include <libkern/OSAtomic.h>
-#include <mach/mach_vm.h>
-#include <mach/mach_init.h>
-#include <mach/thread_switch.h>
-#include <mach/vm_map.h>
-#include <mach-o/dyld.h>
-#include <os/tsd.h>
-#include <sys/mman.h>
-#include <xlocale.h>
-#include <TargetConditionals.h>
-
-#include "malloc.h"
-#include "malloc_printf.h"
-#include "scalable_malloc.h"
-#include "malloc_internal.h"
-#include "stack_logging.h"
-
-#if TARGET_OS_EMBEDDED || TARGET_IPHONE_SIMULATOR
-// _malloc_printf(ASL_LEVEL_INFO...) on iOS doesn't show up in the Xcode Console log of the device,
+/*
+ * Needed for definition of MALLOC_MSL_LITE_WRAPPED_ZONE_FLAGS
+ */
+#ifndef MALLOC_ENABLE_MSL_LITE_SPI
+#define MALLOC_ENABLE_MSL_LITE_SPI 1
+#endif // MALLOC_ENABLE_MSL_LITE_SPI
+
+#include "internal.h"
+
+#if MALLOC_TARGET_IOS
+// malloc_report(ASL_LEVEL_INFO...) on iOS doesn't show up in the Xcode Console log of the device,
// but ASL_LEVEL_NOTICE does. So raising the log level is helpful.
#undef ASL_LEVEL_INFO
#define ASL_LEVEL_INFO ASL_LEVEL_NOTICE
-#endif
-
-#include <CrashReporterClient.h>
-
-#ifdef __LP64__
-#define CONFIG_NANOZONE 1
-#else
-#define CONFIG_NANOZONE 0
-#endif
-
-/*
- * MALLOC_ABSOLUTE_MAX_SIZE - There are many instances of addition to a
- * user-specified size_t, which can cause overflow (and subsequent crashes)
- * for values near SIZE_T_MAX. Rather than add extra "if" checks everywhere
- * this occurs, it is easier to just set an absolute maximum request size,
- * and immediately return an error if the requested size exceeds this maximum.
- * Of course, values less than this absolute max can fail later if the value
- * is still too large for the available memory. The largest value added
- * seems to be PAGE_SIZE (in the macro round_page()), so to be safe, we set
- * the maximum to be 2 * PAGE_SIZE less than SIZE_T_MAX.
- */
-#define MALLOC_ABSOLUTE_MAX_SIZE (SIZE_T_MAX - (2 * PAGE_SIZE))
-
-#define USE_SLEEP_RATHER_THAN_ABORT 0
-
-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 *__syscall_logger; // use this to set up syscall logging (e.g., vm_allocate, vm_deallocate, mmap, munmap)
-
-extern void __prepare_to_log_stacks(void);
+#endif // MALLOC_TARGET_IOS
+
+#define USE_SLEEP_RATHER_THAN_ABORT 0
static _malloc_lock_s _malloc_lock = _MALLOC_LOCK_INIT;
-#define MALLOC_LOCK() _malloc_lock_lock(&_malloc_lock)
-#define MALLOC_UNLOCK() _malloc_lock_unlock(&_malloc_lock)
-
-static inline void yield(void) {
- thread_switch(MACH_PORT_NULL, SWITCH_OPTION_DEPRESS, 1);
-}
+#define MALLOC_LOCK() _malloc_lock_lock(&_malloc_lock)
+#define MALLOC_TRY_LOCK() _malloc_lock_trylock(&_malloc_lock)
+#define MALLOC_UNLOCK() _malloc_lock_unlock(&_malloc_lock)
+#define MALLOC_REINIT_LOCK() _malloc_lock_init(&_malloc_lock)
/* The following variables are exported for the benefit of performance tools
*
@@ -103,53 +54,187 @@
*/
int32_t malloc_num_zones = 0;
int32_t malloc_num_zones_allocated = 0;
-malloc_zone_t **malloc_zones = 0;
+malloc_zone_t **malloc_zones = (malloc_zone_t **)0xdeaddeaddeaddead;
+
+// TODO: getter and setter rather than direct export so that this can be baked
+// into the cached fastpath checks
malloc_logger_t *malloc_logger = NULL;
+static uint32_t initial_num_zones;
+static malloc_zone_t *initial_scalable_zone;
+#if CONFIG_NANOZONE
+static malloc_zone_t *initial_nano_zone;
+#endif // CONFIG_NANOZONE
+malloc_zone_t *initial_xzone_zone;
+static malloc_zone_t *default_purgeable_zone;
+static bool has_injected_zone0;
+static bool malloc_xzone_enabled = MALLOC_XZONE_ENABLED_DEFAULT;
+
+typedef enum {
+ MALLOC_XZONE_OVERRIDE_DEFAULT,
+ MALLOC_XZONE_OVERRIDE_DISABLED,
+ MALLOC_XZONE_OVERRIDE_ENABLED,
+} malloc_xzone_override_t;
+
+static malloc_xzone_override_t malloc_xzone_enabled_override;
+static malloc_xzone_override_t malloc_xzone_nano_override;
+static malloc_xzone_override_t malloc_nano_on_xzone_override;
+
+#if CONFIG_CHECK_PLATFORM_BINARY
+static bool malloc_xzone_allow_non_platform;
+#endif // CONFIG_CHECK_PLATFORM_BINARY
+
unsigned malloc_debug_flags = 0;
+bool malloc_tracing_enabled = false;
+bool malloc_space_efficient_enabled = false;
+bool malloc_medium_space_efficient_enabled = true;
+
+bool malloc_sanitizer_enabled = false;
+
+bool malloc_interposition_compat = false;
+
+
+#if CONFIG_MALLOC_PROCESS_IDENTITY
+malloc_process_identity_t malloc_process_identity = MALLOC_PROCESS_NONE;
+#endif // CONFIG_MALLOC_PROCESS_IDENTITY
unsigned malloc_check_start = 0; // 0 means don't check
unsigned malloc_check_counter = 0;
unsigned malloc_check_each = 1000;
-/* global flag to suppress ASL logging e.g. for syslogd */
-int _malloc_no_asl_log = 0;
-
static int malloc_check_sleep = 100; // default 100 second sleep
-static int malloc_check_abort = 0; // default is to sleep, not abort
-
-static int malloc_debug_file = STDERR_FILENO;
-static boolean_t _malloc_is_initialized = FALSE;
-
-static const char Malloc_Facility[] = "com.apple.Libsystem.malloc";
+static int malloc_check_abort = 0; // default is to sleep, not abort
+static bool malloc_simple_stack_logging = false;
+#define MALLOC_SIMPLE_STACK_LOGGING_FLAGS \
+ (ASL_LEVEL_NOTICE | MALLOC_REPORT_NOPREFIX | MALLOC_REPORT_BACKTRACE | MALLOC_REPORT_NOWRITE)
+
+bool malloc_slowpath = false;
+
+static struct _malloc_msl_symbols msl = {};
+
+static struct malloc_sanitizer_poison *sanitizer = NULL;
+
+#if CONFIG_SANITIZER && TARGET_OS_OSX
+// Fallback pointers for memory poisoning functions from upstream ASan runtime
+static struct {
+ void (*memory_poison)(uintptr_t ptr, size_t sz);
+ void (*memory_unpoison)(uintptr_t ptr, size_t sz);
+} sanitizer_fallback_ptrs;
+
+// Fallback structure for heap allocation functions from our ASan runtime
+static struct malloc_sanitizer_poison sanitizer_fallback = {0};
+
+static void malloc_sanitizer_fallback_allocate_poison(uintptr_t ptr, size_t leftrz_sz, size_t alloc_sz, size_t rightrz_sz);
+static void malloc_sanitizer_fallback_deallocate_poison(uintptr_t ptr, size_t sz);
+#endif
+/* These masks are exported for libdispatch to register with (see "internal.h") */
+const unsigned long malloc_memorypressure_mask_default_4libdispatch = MALLOC_MEMORYPRESSURE_MASK_DEFAULT;
+const unsigned long malloc_memorypressure_mask_msl_4libdispatch = MALLOC_MEMORYPRESSURE_MASK_MSL;
+
+MALLOC_NOEXPORT malloc_zone_t* lite_zone = NULL;
/*
* Counters that coordinate zone destruction (in malloc_zone_unregister) with
* find_registered_zone (here abbreviated as FRZ).
*/
-static int counterAlice = 0, counterBob = 0;
-static int *pFRZCounterLive= &counterAlice, *pFRZCounterDrain = &counterBob;
-
-#define MALLOC_LOG_TYPE_ALLOCATE stack_logging_type_alloc
-#define MALLOC_LOG_TYPE_DEALLOCATE stack_logging_type_dealloc
-#define MALLOC_LOG_TYPE_HAS_ZONE stack_logging_flag_zone
-#define MALLOC_LOG_TYPE_CLEARED stack_logging_flag_cleared
+static int32_t volatile counterAlice = 0, counterBob = 0;
+static int32_t volatile * volatile pFRZCounterLive = &counterAlice;
+static int32_t volatile * volatile pFRZCounterDrain = &counterBob;
+
+unsigned int _os_cpu_number_override = -1;
+
+static inline malloc_zone_t *inline_malloc_default_zone(void) __attribute__((always_inline));
+
+#define MALLOC_LOG_TYPE_ALLOCATE stack_logging_type_alloc
+#define MALLOC_LOG_TYPE_DEALLOCATE stack_logging_type_dealloc
+#define MALLOC_LOG_TYPE_HAS_ZONE stack_logging_flag_zone
+#define MALLOC_LOG_TYPE_CLEARED stack_logging_flag_cleared
#define DEFAULT_MALLOC_ZONE_STRING "DefaultMallocZone"
+#define DEFAULT_SCALABLE_ZONE_STRING "DefaultScalableMallocZone"
#define DEFAULT_PUREGEABLE_ZONE_STRING "DefaultPurgeableMallocZone"
-
-boolean_t malloc_engaged_nano(void);
-#if CONFIG_NANOZONE
-extern boolean_t _malloc_engaged_nano;
-malloc_zone_t *create_nano_zone(size_t initial_size, malloc_zone_t *helper_zone, unsigned debug_flags);
-void nano_forked_zone(malloc_zone_t *zone);
#define MALLOC_HELPER_ZONE_STRING "MallocHelperZone"
-#endif
+
+MALLOC_NOEXPORT
+unsigned int phys_ncpus;
+
+MALLOC_NOEXPORT
+unsigned int logical_ncpus;
+
+#if CONFIG_MAGAZINE_PER_CLUSTER
+MALLOC_NOEXPORT
+unsigned int ncpuclusters;
+#endif // CONFIG_MAGAZINE_PER_CLUSTER
+
+MALLOC_NOEXPORT
+unsigned int hyper_shift;
+
+MALLOC_NOEXPORT
+size_t malloc_absolute_max_size;
+
+// Boot argument for max magazine control
+static const char max_magazines_boot_arg[] = "malloc_max_magazines";
+
+static const char large_expanded_cache_threshold_boot_arg[] = "malloc_large_expanded_cache_threshold";
+
+
+MALLOC_NOEXPORT
+unsigned malloc_zero_on_free_sample_period = 0; // 0 means don't sample
+
+static const char zero_on_free_sample_period_boot_arg[] = "malloc_zero_on_free_sample_period";
+
+MALLOC_NOEXPORT
+malloc_zero_policy_t malloc_zero_policy = MALLOC_ZERO_POLICY_DEFAULT;
+
+static const char zero_on_free_enabled_boot_arg[] = "malloc_zero_on_free_enabled";
+
+
+#if CONFIG_MEDIUM_ALLOCATOR
+static const char medium_enabled_boot_arg[] = "malloc_medium_zone";
+static const char max_medium_magazines_boot_arg[] = "malloc_max_medium_magazines";
+static const char medium_activation_threshold_boot_arg[] = "malloc_medium_activation_threshold";
+static const char medium_space_efficient_boot_arg[] = "malloc_medium_space_efficient";
+static const char medium_madvise_dram_scale_divisor_boot_arg[] = "malloc_medium_madvise_dram_scale_divisor";
+#endif // CONFIG_MEDIUM_ALLOCATOR
+
+static
+bool malloc_report_config = false;
/********* Utilities ************/
-__attribute__((visibility("hidden"))) uint64_t malloc_entropy[2] = {0, 0};
+static bool _malloc_entropy_initialized;
+
+#if !TARGET_OS_DRIVERKIT
+#include <dlfcn.h>
+
+typedef void * (*dlopen_t) (const char * __path, int __mode);
+typedef void * (*dlsym_t) (void * __handle, const char * __symbol);
+
+static dlopen_t LIBMALLOC_FUNCTION_PTRAUTH(_dlopen) = NULL;
+static dlsym_t LIBMALLOC_FUNCTION_PTRAUTH(_dlsym) = NULL;
+#else
+#define _dlopen(...) NULL
+#define _dlsym(...) NULL
+#endif // TARGET_OS_DRIVERKIT
+
+void
+malloc_slowpath_update(void)
+{
+ bool slowpath = has_injected_zone0 ||
+ malloc_num_zones == 0 ||
+ malloc_check_start ||
+ lite_zone ||
+ malloc_tracing_enabled ||
+ malloc_simple_stack_logging ||
+ (malloc_debug_flags & MALLOC_DO_SCRIBBLE) != 0 ||
+ malloc_interposition_compat;
+
+ if (malloc_slowpath != slowpath) {
+ malloc_slowpath = slowpath;
+ }
+}
void __malloc_init(const char *apple[]);
+static void _malloc_initialize(const char *apple[], const char *bootargs);
static int
__entropy_from_kernel(const char *str)
@@ -160,87 +245,903 @@
/* Skip over key to the first value */
str = strchr(str, '=');
- if (str == NULL)
+ if (str == NULL) {
return 0;
+ }
str++;
- while (str && idx < sizeof(malloc_entropy)/sizeof(malloc_entropy[0])) {
+ while (str && idx < sizeof(malloc_entropy) / sizeof(malloc_entropy[0])) {
strlcpy(tmp, str, 20);
p = strchr(tmp, ',');
- if (p) *p = '\0';
+ if (p) {
+ *p = '\0';
+ }
val = strtoull_l(tmp, NULL, 0, NULL);
malloc_entropy[idx] = (uint64_t)val;
idx++;
- if ((str = strchr(str, ',')) != NULL)
+ if ((str = strchr(str, ',')) != NULL) {
str++;
+ }
}
return idx;
}
-/* TODO: Remove __malloc_entropy_setup, it was left to avoid rev-lock */
-void
-__malloc_entropy_setup(const char *apple[]) {
- __malloc_init(apple);
-}
-
-/* TODO: Investigate adding _malloc_initialize() into this libSystem initializer */
+#if TARGET_OS_OSX && defined(__x86_64__)
+static uint64_t
+__is_translated(void)
+{
+ return (*(uint64_t*)_COMM_PAGE_CPU_CAPABILITIES64) & kIsTranslated;
+}
+#endif /* TARGET_OS_OSX */
+
+#define LIBMALLOC_EXPERIMENT_FACTORS_KEY "MallocExperiment="
+#define LIBMALLOC_EXPERIMENT_DISABLE_MEDIUM (1ULL)
+#define LIBMALLOC_DEFERRED_RECLAIM_ENABLE "MallocDeferredReclaim=1"
+static void
+__malloc_init_experiments(const char *str)
+{
+ uint64_t experiment_factors = 0;
+ str = strchr(str, '=');
+ if (str) {
+ experiment_factors = strtoull_l(str + 1, NULL, 16, NULL);
+ }
+ switch (experiment_factors) {
+ case LIBMALLOC_EXPERIMENT_DISABLE_MEDIUM:
+ magazine_medium_enabled = false;
+ break;
+ }
+}
+
+static void
+__malloc_init_from_bootargs(const char *bootargs)
+{
+ // The maximum number of magazines can be set either via a
+ // boot argument or from the environment. Get the boot argument value
+ // here and store it. We can't bounds check it until we have phys_ncpus,
+ // which happens later in _malloc_initialize(), along with handling
+ // of the environment value setting.
+ char value_buf[256];
+ const char *flag = malloc_common_value_for_key_copy(bootargs,
+ max_magazines_boot_arg, value_buf, sizeof(value_buf));
+ if (flag) {
+ const char *endp;
+ long value = malloc_common_convert_to_long(flag, &endp);
+ if (!*endp && value >= 0) {
+ max_magazines = (unsigned int)value;
+ } else {
+ malloc_report(ASL_LEVEL_ERR,
+ "malloc_max_magazines must be positive - ignored.\n");
+ }
+ }
+
+ flag = malloc_common_value_for_key_copy(bootargs,
+ large_expanded_cache_threshold_boot_arg, value_buf, sizeof(value_buf));
+ if (flag) {
+ const char *endp;
+ long value = malloc_common_convert_to_long(flag, &endp);
+ if (!*endp && value >= 0) {
+ magazine_large_expanded_cache_threshold = (unsigned int)value;
+ } else {
+ malloc_report(ASL_LEVEL_ERR,
+ "malloc_large_expanded_cache_threshold must be positive - ignored.\n");
+ }
+ }
+
+ flag = malloc_common_value_for_key_copy(bootargs,
+ zero_on_free_enabled_boot_arg, value_buf, sizeof(value_buf));
+ if (flag) {
+ const char *endp;
+ long value = malloc_common_convert_to_long(flag, &endp);
+ if (!*endp && (value == 0 || value == 1)) {
+ malloc_zero_policy = value ? MALLOC_ZERO_ON_FREE : MALLOC_ZERO_NONE;
+ } else {
+ malloc_report(ASL_LEVEL_ERR,
+ "malloc_zero_on_free_enabled must be 0 or 1 - ignored.\n");
+ }
+ }
+
+ flag = malloc_common_value_for_key_copy(bootargs,
+ zero_on_free_sample_period_boot_arg, value_buf, sizeof(value_buf));
+ if (flag) {
+ const char *endp;
+ long value = malloc_common_convert_to_long(flag, &endp);
+ if (!*endp && value >= 0) {
+ malloc_zero_on_free_sample_period = (unsigned int)value;
+ } else {
+ malloc_report(ASL_LEVEL_ERR,
+ "malloc_zero_on_free_sample_period must be positive - ignored.\n");
+ }
+ }
+
+
+#if CONFIG_MEDIUM_ALLOCATOR
+#if TARGET_OS_OSX
+#if defined(__x86_64__)
+ if (__is_translated()) {
+ magazine_medium_active_threshold = 0;
+ }
+#elif defined(__arm64__)
+ magazine_medium_active_threshold = 0;
+#endif
+#endif /* TARGET_OS_OSX */
+
+ flag = malloc_common_value_for_key_copy(bootargs, medium_enabled_boot_arg,
+ value_buf, sizeof(value_buf));
+ if (flag) {
+ const char *endp;
+ long value = malloc_common_convert_to_long(flag, &endp);
+ if (!*endp) {
+ magazine_medium_enabled = (value != 0);
+ }
+ }
+
+ flag = malloc_common_value_for_key_copy(bootargs,
+ medium_activation_threshold_boot_arg, value_buf, sizeof(value_buf));
+ if (flag) {
+ const char *endp;
+ long value = malloc_common_convert_to_long(flag, &endp);
+ if (!*endp && value >= 0) {
+ magazine_medium_active_threshold = (uint64_t)value;
+ } else {
+ malloc_report(ASL_LEVEL_ERR,
+ "malloc_medium_activation_threshold must be positive - ignored.\n");
+ }
+ }
+
+ flag = malloc_common_value_for_key_copy(bootargs,
+ max_medium_magazines_boot_arg, value_buf, sizeof(value_buf));
+ if (flag) {
+ const char *endp;
+ long value = malloc_common_convert_to_long(flag, &endp);
+ if (!*endp && value >= 0) {
+ max_medium_magazines = (int)value;
+ } else {
+ malloc_report(ASL_LEVEL_ERR,
+ "malloc_max_medium_magazines must be positive - ignored.\n");
+ }
+ }
+
+ flag = malloc_common_value_for_key_copy(bootargs,
+ medium_space_efficient_boot_arg, value_buf, sizeof(value_buf));
+ if (flag) {
+ const char *endp;
+ long value = malloc_common_convert_to_long(flag, &endp);
+ if (!*endp) {
+ malloc_medium_space_efficient_enabled = (value != 0);
+ }
+ }
+
+ flag = malloc_common_value_for_key_copy(bootargs, medium_madvise_dram_scale_divisor_boot_arg, value_buf, sizeof(value_buf));
+ if (flag) {
+ const char *endp;
+ long value = malloc_common_convert_to_long(flag, &endp);
+ if (!*endp && value >= 0) {
+ magazine_medium_madvise_dram_scale_divisor = (uint64_t)value;
+ } else {
+ malloc_report(ASL_LEVEL_ERR,
+ "malloc_medium_madvise_dram_scale_divisor must be positive - ignored.\n");
+ }
+ }
+#endif // CONFIG_MEDIUM_ALLOCATOR
+}
+
+#if CONFIG_MALLOC_PROCESS_IDENTITY
+
+static void
+_malloc_check_process_identity(const char *apple[])
+{
+ static const struct {
+ const char *name;
+ malloc_process_identity_t identity;
+ } name_identity_mapping[] = {
+#if TARGET_OS_SIMULATOR
+ { "launchd_sim", MALLOC_PROCESS_LAUNCHD, },
+ { "launchd_sim.development", MALLOC_PROCESS_LAUNCHD, },
+ { "launchd_sim.debug", MALLOC_PROCESS_LAUNCHD, },
+#else // TARGET_OS_SIMULATOR
+ { "launchd", MALLOC_PROCESS_LAUNCHD, },
+ { "launchd.development", MALLOC_PROCESS_LAUNCHD, },
+ { "launchd.debug", MALLOC_PROCESS_LAUNCHD, },
+ { "launchd.testing", MALLOC_PROCESS_LAUNCHD, },
+#endif // TARGET_OS_SIMULATOR
+ { "logd", MALLOC_PROCESS_LOGD, },
+ { "notifyd", MALLOC_PROCESS_NOTIFYD, },
+
+ { "mediaparserd", MALLOC_PROCESS_MEDIAPARSERD, },
+ { "videocodecd", MALLOC_PROCESS_VIDEOCODECD, },
+ { "mediaplaybackd", MALLOC_PROCESS_MEDIAPLAYBACKD },
+ { "audiomxd", MALLOC_PROCESS_AUDIOMXD, },
+ { "avconferenced", MALLOC_PROCESS_AVCONFERENCED, },
+ { "mediaserverd", MALLOC_PROCESS_MEDIASERVERD, },
+ { "cameracaptured", MALLOC_PROCESS_CAMERACAPTURED, },
+
+ { "MessagesBlastDoorService", MALLOC_PROCESS_BLASTDOOR_MESSAGES, },
+ { "MessagesAirlockService", MALLOC_PROCESS_BLASTDOOR_MESSAGES, },
+ { "IDSBlastDoorService", MALLOC_PROCESS_BLASTDOOR_IDS, },
+ { "IMDPersistenceAgent", MALLOC_PROCESS_IMDPERSISTENCEAGENT },
+ { "imagent", MALLOC_PROCESS_IMAGENT, },
+
+ { "ThumbnailExtensionSecure", MALLOC_PROCESS_QUICKLOOK_THUMBNAIL_SECURE, },
+ { "com.apple.quicklook.extension.previewUI", MALLOC_PROCESS_QUICKLOOK_PREVIEW, },
+ { "QuickLookUIExtension", MALLOC_PROCESS_QUICKLOOK_PREVIEW, },
+ { "ThumbnailExtension", MALLOC_PROCESS_QUICKLOOK_THUMBNAIL },
+#if TARGET_OS_OSX
+ // Load-bearing: not already MallocSpaceEfficient
+ { "QuickLookUIService", MALLOC_PROCESS_QUICKLOOK_UISERVICE },
+ // Already MallocSpaceEfficient, but valuable to identify as
+ // security-relevant for other special treatment
+ { "ThumbnailExtension_macOS", MALLOC_PROCESS_QUICKLOOK_THUMBNAIL },
+ { "QuickLookSatellite", MALLOC_PROCESS_QUICKLOOK_MACOS },
+ { "quicklookd", MALLOC_PROCESS_QUICKLOOK_MACOS },
+ { "com.apple.quicklook.ThumbnailsAgent", MALLOC_PROCESS_QUICKLOOK_MACOS },
+ { "ExternalQuickLookSatellite-arm64", MALLOC_PROCESS_QUICKLOOK_MACOS },
+ { "ExternalQuickLookSatellite-x86_64", MALLOC_PROCESS_QUICKLOOK_MACOS },
+#endif // TARGET_OS_OSX
+
+ { "MobileSafari", MALLOC_PROCESS_BROWSER, },
+ { "com.apple.WebKit.Networking", MALLOC_PROCESS_BROWSER, },
+ { "com.apple.WebKit.GPU", MALLOC_PROCESS_BROWSER, },
+ { "com.apple.WebKit.WebContent", MALLOC_PROCESS_BROWSER, },
+ { "com.apple.WebKit.WebContent.CaptivePortal", MALLOC_PROCESS_BROWSER, },
+ { "MTLCompilerService", MALLOC_PROCESS_MTLCOMPILERSERVICE },
+
+#if TARGET_OS_OSX
+ { "Safari", MALLOC_PROCESS_SAFARI, },
+ { "com.apple.Safari.CredentialExtractionHelper", MALLOC_PROCESS_SAFARI_SUPPORT, },
+ { "com.apple.Safari.History", MALLOC_PROCESS_SAFARI_SUPPORT, },
+ { "com.apple.Safari.SandboxBroker", MALLOC_PROCESS_SAFARI_SUPPORT, },
+ { "com.apple.Safari.SearchHelper", MALLOC_PROCESS_SAFARI_SUPPORT, },
+ { "com.apple.SafariFoundation.CredentialProviderExtensionHelper", MALLOC_PROCESS_SAFARI_SUPPORT, },
+ { "com.apple.SafariPlatformSupport.Helper", MALLOC_PROCESS_SAFARI_SUPPORT, },
+ { "com.apple.SafariServices.ExtensionHelper", MALLOC_PROCESS_SAFARI_SUPPORT, },
+ { "com.apple.SafariServices", MALLOC_PROCESS_SAFARI_SUPPORT, },
+ { "VTDecoderXPCService", MALLOC_PROCESS_VTDECODERXPCSERVICE, },
+#endif // TARGET_OS_OSX
+
+ { "callservicesd", MALLOC_PROCESS_CALLSERVICESD, },
+ { "maild", MALLOC_PROCESS_MAILD, },
+ { "mDNSResponder", MALLOC_PROCESS_MDNSRESPONDER, },
+ { "ASVAssetViewer", MALLOC_PROCESS_ASVASSETVIEWER },
+ { "identityservicesd", MALLOC_PROCESS_IDENTITYSERVICESD, },
+ { "wifid", MALLOC_PROCESS_WIFID, },
+ { "fmfd", MALLOC_PROCESS_FMFD, },
+ { "searchpartyd", MALLOC_PROCESS_SEARCHPARTYD, },
+ { "vmd", MALLOC_PROCESS_VMD, },
+ { "CommCenter", MALLOC_PROCESS_COMMCENTER, },
+ { "wifip2pd", MALLOC_PROCESS_WIFIP2PD, },
+ { "wifianalyticsd", MALLOC_PROCESS_WIFIANALYTICSD, },
+
+ { "mds_stores", MALLOC_PROCESS_MDS_STORES },
+
+ { "AegirPoster", MALLOC_PROCESS_AEGIRPOSTER, },
+ { "CollectionsPoster", MALLOC_PROCESS_COLLECTIONSPOSTER, },
+ };
+
+ if (getpid() == 1) {
+ malloc_process_identity = MALLOC_PROCESS_LAUNCHD;
+ return;
+ }
+
+ const char *flag = _simple_getenv(apple, "HardenedRuntime");
+ if (flag) {
+ unsigned long long value = strtoull(flag, NULL, 0);
+ if (value) {
+ // reproduced from xnu
+ enum {
+ MallocBrowserHostEntitlementMask = 0x01,
+ MallocBrowserGPUEntitlementMask = 0x02,
+ MallocBrowserNetworkEntitlementMask = 0x04,
+ MallocBrowserWebContentEntitlementMask = 0x08,
+ };
+
+ long enablement_mask = MallocBrowserHostEntitlementMask |
+ MallocBrowserGPUEntitlementMask |
+ MallocBrowserNetworkEntitlementMask |
+ MallocBrowserWebContentEntitlementMask;
+ if (value & enablement_mask) {
+ malloc_process_identity = MALLOC_PROCESS_BROWSER;
+ return;
+ }
+ }
+ }
+
+ const char *name = getprogname();
+ for (size_t i = 0; i < countof(name_identity_mapping); i++) {
+ if (!strcmp(name, name_identity_mapping[i].name)) {
+ malloc_process_identity = name_identity_mapping[i].identity;
+ break;
+ }
+ }
+}
+
+#if !CONFIG_FEATUREFLAGS_SIMPLE
+#error "must have feature flags"
+#endif
+
+static bool
+_malloc_check_secure_allocator_process_enablement(void)
+{
+ // launchd is special because the feature flag check can't work for it
+ if (malloc_process_identity == MALLOC_PROCESS_LAUNCHD) {
+ return MALLOC_SECURE_ALLOCATOR_LAUNCHD_ENABLED_DEFAULT;
+ }
+
+#define ENABLEMENT_CASE(id, enable_status) \
+ case MALLOC_PROCESS_##id: \
+ return (enable_status);
+#define ENABLEMENT_CASE_FF(id, name, darwin_default, simulator_default) \
+ case MALLOC_PROCESS_##id: \
+ return malloc_secure_feature_enabled(SecureAllocator_process_##name, \
+ (darwin_default), (simulator_default))
+
+
+ switch (malloc_process_identity) {
+ ENABLEMENT_CASE(LOGD, true);
+ ENABLEMENT_CASE(NOTIFYD, true);
+
+ ENABLEMENT_CASE(MEDIAPARSERD, true);
+ ENABLEMENT_CASE(VIDEOCODECD, true);
+ ENABLEMENT_CASE(MEDIAPLAYBACKD, true);
+ ENABLEMENT_CASE(AVCONFERENCED, true);
+ ENABLEMENT_CASE(MEDIASERVERD, true);
+ ENABLEMENT_CASE(AUDIOMXD, true);
+ ENABLEMENT_CASE(CAMERACAPTURED, true);
+
+ ENABLEMENT_CASE(BLASTDOOR_MESSAGES, true);
+ ENABLEMENT_CASE(BLASTDOOR_IDS, true);
+ ENABLEMENT_CASE(IMDPERSISTENCEAGENT, true);
+ ENABLEMENT_CASE(IMAGENT, true);
+
+#if TARGET_OS_OSX
+ ENABLEMENT_CASE(QUICKLOOK_MACOS, true);
+ ENABLEMENT_CASE(QUICKLOOK_UISERVICE, true);
+#endif
+ ENABLEMENT_CASE(QUICKLOOK_THUMBNAIL_SECURE, true);
+ ENABLEMENT_CASE(QUICKLOOK_PREVIEW, true);
+ ENABLEMENT_CASE(QUICKLOOK_THUMBNAIL, true);
+
+#if TARGET_OS_OSX
+ ENABLEMENT_CASE_FF(MTLCOMPILERSERVICE, MTLCompilerService, false, false);
+#else
+ ENABLEMENT_CASE(MTLCOMPILERSERVICE, true);
+#endif
+
+ ENABLEMENT_CASE(CALLSERVICESD, true);
+ ENABLEMENT_CASE(MAILD, true);
+ ENABLEMENT_CASE(MDNSRESPONDER, true);
+ ENABLEMENT_CASE(ASVASSETVIEWER, true);
+ ENABLEMENT_CASE(IDENTITYSERVICESD, true);
+ ENABLEMENT_CASE(WIFID, true);
+ ENABLEMENT_CASE(FMFD, true);
+ ENABLEMENT_CASE(SEARCHPARTYD, true);
+ ENABLEMENT_CASE(VMD, true);
+ ENABLEMENT_CASE(WIFIP2PD, true);
+ ENABLEMENT_CASE(WIFIANALYTICSD, true);
+
+ ENABLEMENT_CASE(COMMCENTER, true);
+
+#if TARGET_OS_SIMULATOR
+ ENABLEMENT_CASE(BROWSER, false);
+#else
+ ENABLEMENT_CASE(BROWSER, true);
+#endif // TARGET_OS_SIMULATOR
+
+#if TARGET_OS_OSX
+ ENABLEMENT_CASE(SAFARI, true);
+ ENABLEMENT_CASE(SAFARI_SUPPORT, true);
+#endif
+
+ ENABLEMENT_CASE_FF(AEGIRPOSTER, aegirposter, false, false);
+ ENABLEMENT_CASE_FF(COLLECTIONSPOSTER, CollectionsPoster, false, false);
+
+#if TARGET_OS_OSX
+ ENABLEMENT_CASE(VTDECODERXPCSERVICE, true);
+#endif
+
+ ENABLEMENT_CASE(MDS_STORES, true);
+
+ default:
+ return false;
+ }
+}
+
+#endif // CONFIG_MALLOC_PROCESS_IDENTITY
+
+static void
+_malloc_init_featureflags(void)
+{
+#if CONFIG_FEATUREFLAGS_SIMPLE
+ bool zero_on_free_feature_enabled = os_feature_enabled_simple(libmalloc,
+ ZeroOnFree, MALLOC_ZERO_POLICY_DEFAULT == MALLOC_ZERO_ON_FREE);
+ bool policy_is_zero_on_free = (malloc_zero_policy == MALLOC_ZERO_ON_FREE);
+ if (zero_on_free_feature_enabled != policy_is_zero_on_free) {
+ malloc_zero_policy = zero_on_free_feature_enabled ?
+ MALLOC_ZERO_ON_FREE : MALLOC_ZERO_NONE;
+ }
+
+ bool secure_allocator = false;
+#if CONFIG_MALLOC_PROCESS_IDENTITY
+ if (malloc_process_identity != MALLOC_PROCESS_NONE) {
+ secure_allocator = _malloc_check_secure_allocator_process_enablement();
+ } else
+#endif // CONFIG_MALLOC_PROCESS_IDENTITY
+ {
+#if MALLOC_TARGET_IOS_ONLY
+ secure_allocator = malloc_secure_feature_enabled(
+ SecureAllocator_SystemWide, true, true);
+#else
+ secure_allocator = malloc_secure_feature_enabled(
+ SecureAllocator_SystemWide, false, false);
+#endif // MALLOC_TARGET_IOS_ONLY
+ }
+
+#if TARGET_OS_OSX && !TARGET_CPU_ARM64
+ if (!os_feature_enabled_simple(libmalloc, SecureAllocator_Intel, false)) {
+ secure_allocator = false;
+ }
+#endif // TARGET_OS_OSX && !TARGET_CPU_ARM64
+
+#if TARGET_OS_OSX
+ bool allow_non_platform = os_feature_enabled_simple(libmalloc,
+ SecureAllocator_NonPlatform, false);
+ if (malloc_xzone_allow_non_platform != allow_non_platform) {
+ malloc_xzone_allow_non_platform = allow_non_platform;
+ }
+#endif // TARGET_OS_OSX
+
+ if (secure_allocator != malloc_xzone_enabled) {
+ malloc_xzone_enabled = secure_allocator;
+ }
+#endif // CONFIG_FEATUREFLAGS_SIMPLE
+}
+
+extern malloc_zone_t *force_asan_init_if_present(void)
+ asm("_malloc_default_zone");
+
void
__malloc_init(const char *apple[])
{
+ // We could try to be clever and cater for arbitrary length bootarg
+ // strings, but it's probably not worth it, especially as we would need
+ // to temporarily allocate at least a page of memory to read the bootargs
+ // into.
+ char bootargs[1024] = { '\0' };
+ bool allow_bootargs = true;
+#if CONFIG_FEATUREFLAGS_SIMPLE
+ // os_feature_enabled_simple() doesn't work in launchd, but launchd should
+ // be able to read boot-args
+ if (getpid() != 1) {
+ allow_bootargs &= os_feature_enabled_simple(libmalloc, EnableBootArgs, false);
+ }
+#endif
+#if defined(_COMM_PAGE_DEV_FIRM)
+ allow_bootargs &= !!*((uint32_t *)_COMM_PAGE_DEV_FIRM);
+#endif // _COMM_PAGE_DEV_FIRM
+
+ size_t len = sizeof(bootargs) - 1;
+ if (allow_bootargs &&
+ !sysctlbyname("kern.bootargs", bootargs, &len, NULL, 0) &&
+ len > 0) {
+ bootargs[len + 1] = '\0';
+ }
+
+ // Cache the calculation of this "constant", which unfortunately depends on
+ // runtime values of vm_kernel_page_size and vm_page_size
+ malloc_absolute_max_size = _MALLOC_ABSOLUTE_MAX_SIZE;
+
+#if CONFIG_CHECK_PLATFORM_BINARY
+ bool is_platform_binary = _malloc_is_platform_binary();
+ if (malloc_is_platform_binary != is_platform_binary) {
+ malloc_is_platform_binary = is_platform_binary;
+ }
+#endif
+
+#if CONFIG_CHECK_SECURITY_POLICY
+ bool allow_internal_security = _malloc_allow_internal_security_policy();
+ if (allow_internal_security != malloc_internal_security_policy) {
+ malloc_internal_security_policy = allow_internal_security;
+ }
+#endif
+
+#if CONFIG_MALLOC_PROCESS_IDENTITY
+ _malloc_check_process_identity(apple);
+#endif
+
+ _malloc_init_featureflags();
+
const char **p;
-#if CONFIG_NANOZONE
- _malloc_engaged_nano = 0;
- for (p = apple; p && *p; p++) {
- if (0 == strncmp(*p, "MallocNanoZone=1", strlen("MallocNanoZone=1"))) {
- // _malloc_printf(ASL_LEVEL_INFO, "MallocNanoZone=1\n");
- _malloc_engaged_nano = 1;
- break;
- }
- }
-#endif
-
+ const char *malloc_experiments = NULL;
for (p = apple; p && *p; p++) {
if (strstr(*p, "malloc_entropy") == *p) {
int count = __entropy_from_kernel(*p);
- bzero((void*)*p, strlen(*p));
-
- if (sizeof(malloc_entropy)/sizeof(malloc_entropy[0]) == count) {
- return;
+ bzero((void *)*p, strlen(*p));
+
+ if (sizeof(malloc_entropy) / sizeof(malloc_entropy[0]) == count) {
+ _malloc_entropy_initialized = true;
}
- break;
- }
- }
-
- malloc_entropy[0] = ((uint64_t)arc4random()) << 32 | ((uint64_t)arc4random());
- malloc_entropy[1] = ((uint64_t)arc4random()) << 32 | ((uint64_t)arc4random());
- return;
-}
-
-static inline malloc_zone_t * find_registered_zone(const void *, size_t *) __attribute__((always_inline));
+ }
+ if (strstr(*p, LIBMALLOC_EXPERIMENT_FACTORS_KEY) == *p) {
+ malloc_experiments = *p;
+ }
+#if CONFIG_DEFERRED_RECLAIM
+ if (strstr(*p, LIBMALLOC_DEFERRED_RECLAIM_ENABLE) == *p) {
+ // Turn on the large cache which will place
+ // its free entries in the deferred reclaim buffer
+ large_cache_enabled = 1;
+ }
+#endif /* CONFIG_DEFERRED_RECLAIM */
+ }
+ if (!_malloc_entropy_initialized) {
+ getentropy((void*)malloc_entropy, sizeof(malloc_entropy));
+ _malloc_entropy_initialized = true;
+ }
+
+ if (malloc_experiments) {
+ __malloc_init_experiments(malloc_experiments);
+ }
+ __malloc_init_from_bootargs(bootargs);
+ mvm_aslr_init();
+
+ /*
+ * This really is a renamed call to malloc_default_zone() which is
+ * interposable and interposed by asan, so that we trigger the lazy
+ * initialization of asan _BEFORE_ _malloc_initialize().
+ *
+ * If we do it after, then _malloc_initialize() then ASAN will replace
+ * the system allocator too late and bad things happen.
+ */
+ force_asan_init_if_present();
+
+ _malloc_initialize(apple, bootargs);
+}
+
+static void register_pgm_zone(bool internal_diagnostics);
+static void stack_logging_early_finished(const struct _malloc_late_init *funcs);
+
+// WARNING: The passed _malloc_late_init is a stack variable in
+// libSystem_initializer(). We must not hold on to it.
+//
+// WARNING 2: By the time this is called, malloc() is already used
+// a bunch of times (from libobjc, libxpc, ...).
+void
+__malloc_late_init(const struct _malloc_late_init *mli)
+{
+ register_pgm_zone(mli->internal_diagnostics);
+ stack_logging_early_finished(mli);
+ initial_num_zones = malloc_num_zones;
+
+#if CONFIG_SANITIZER
+ if (malloc_sanitizer_enabled) {
+ sanitizer_reset_environment();
+
+#if TARGET_OS_OSX
+ // Fetch the memory poisoning functions from the upstream ASan runtime
+ sanitizer_fallback_ptrs.memory_poison = _dlsym(RTLD_MAIN_ONLY, "__asan_poison_memory_region");
+ sanitizer_fallback_ptrs.memory_unpoison = _dlsym(RTLD_MAIN_ONLY, "__asan_unpoison_memory_region");
+
+ // Prefer the heap allocation functions from our ASan runtime, if available
+ void *heap_alloc = _dlsym(RTLD_MAIN_ONLY, "__asan_poison_heap_memory_alloc"),
+ *heap_dealloc = _dlsym(RTLD_MAIN_ONLY, "__asan_poison_heap_memory_free"),
+ *heap_internal_poison = _dlsym(RTLD_MAIN_ONLY, "__asan_poison_heap_memory_internal");
+ if (heap_alloc && heap_dealloc && heap_internal_poison) {
+ sanitizer_fallback.heap_allocate_poison = heap_alloc;
+ sanitizer_fallback.heap_deallocate_poison = heap_dealloc;
+ sanitizer_fallback.heap_internal_poison = heap_internal_poison;
+ } else {
+ sanitizer_fallback.heap_allocate_poison = &malloc_sanitizer_fallback_allocate_poison;
+ sanitizer_fallback.heap_deallocate_poison = &malloc_sanitizer_fallback_deallocate_poison;
+ sanitizer_fallback.heap_internal_poison = &malloc_sanitizer_fallback_deallocate_poison;
+ }
+#endif
+ }
+#endif
+}
+
+MALLOC_ALWAYS_INLINE
static inline malloc_zone_t *
-find_registered_zone(const void *ptr, size_t *returned_size) {
+runtime_default_zone(void) {
+ return (lite_zone) ? lite_zone : inline_malloc_default_zone();
+}
+
+static size_t
+default_zone_size(malloc_zone_t *zone, const void *ptr)
+{
+ zone = runtime_default_zone();
+
+ return zone->size(zone, ptr);
+}
+
+static void *
+default_zone_malloc(malloc_zone_t *zone, size_t size)
+{
+ zone = runtime_default_zone();
+
+ return zone->malloc(zone, size);
+}
+
+static void *
+default_zone_calloc(malloc_zone_t *zone, size_t num_items, size_t size)
+{
+ zone = runtime_default_zone();
+
+ return zone->calloc(zone, num_items, size);
+}
+
+static void *
+default_zone_valloc(malloc_zone_t *zone, size_t size)
+{
+ zone = runtime_default_zone();
+
+ return zone->valloc(zone, size);
+}
+
+static void
+default_zone_free(malloc_zone_t *zone, void *ptr)
+{
+ zone = runtime_default_zone();
+
+ return zone->free(zone, ptr);
+}
+
+static void *
+default_zone_realloc(malloc_zone_t *zone, void *ptr, size_t new_size)
+{
+ zone = runtime_default_zone();
+
+ return zone->realloc(zone, ptr, new_size);
+}
+
+static void
+default_zone_destroy(malloc_zone_t *zone)
+{
+ zone = runtime_default_zone();
+
+ return zone->destroy(zone);
+}
+
+static unsigned
+default_zone_batch_malloc(malloc_zone_t *zone, size_t size, void **results, unsigned count)
+{
+ zone = runtime_default_zone();
+
+ return zone->batch_malloc(zone, size, results, count);
+}
+
+static void
+default_zone_batch_free(malloc_zone_t *zone, void **to_be_freed, unsigned count)
+{
+ zone = runtime_default_zone();
+
+ return zone->batch_free(zone, to_be_freed, count);
+}
+
+static void *
+default_zone_memalign(malloc_zone_t *zone, size_t alignment, size_t size)
+{
+ zone = runtime_default_zone();
+
+ return zone->memalign(zone, alignment, size);
+}
+
+static void
+default_zone_free_definite_size(malloc_zone_t *zone, void *ptr, size_t size)
+{
+ zone = runtime_default_zone();
+
+ return zone->free_definite_size(zone, ptr, size);
+}
+
+static size_t
+default_zone_pressure_relief(malloc_zone_t *zone, size_t goal)
+{
+ zone = runtime_default_zone();
+
+ return zone->pressure_relief(zone, goal);
+}
+
+static boolean_t
+default_zone_malloc_claimed_address(malloc_zone_t *zone, void *ptr)
+{
+ zone = runtime_default_zone();
+
+ return malloc_zone_claimed_address(zone, ptr);
+}
+
+static kern_return_t
+default_zone_ptr_in_use_enumerator(task_t task,
+ void *context,
+ unsigned type_mask,
+ vm_address_t zone_address,
+ memory_reader_t reader,
+ vm_range_recorder_t recorder)
+{
+ malloc_zone_t *zone = runtime_default_zone();
+
+ return zone->introspect->enumerator(task, context, type_mask, (vm_address_t) zone, reader, recorder);
+}
+
+static size_t
+default_zone_good_size(malloc_zone_t *zone, size_t size)
+{
+ zone = runtime_default_zone();
+
+ return zone->introspect->good_size(zone, size);
+}
+
+static boolean_t
+default_zone_check(malloc_zone_t *zone)
+{
+ zone = runtime_default_zone();
+
+ return zone->introspect->check(zone);
+}
+
+static void
+default_zone_print(malloc_zone_t *zone, boolean_t verbose)
+{
+ zone = runtime_default_zone();
+
+ return (void)zone->introspect->print(zone, verbose);
+}
+
+static void
+default_zone_log(malloc_zone_t *zone, void *log_address)
+{
+ zone = runtime_default_zone();
+
+ return zone->introspect->log(zone, log_address);
+}
+
+static void
+default_zone_force_lock(malloc_zone_t *zone)
+{
+ zone = runtime_default_zone();
+
+ return zone->introspect->force_lock(zone);
+}
+
+static void
+default_zone_force_unlock(malloc_zone_t *zone)
+{
+ zone = runtime_default_zone();
+
+ return zone->introspect->force_unlock(zone);
+}
+
+static void
+default_zone_statistics(malloc_zone_t *zone, malloc_statistics_t *stats)
+{
+ zone = runtime_default_zone();
+
+ return zone->introspect->statistics(zone, stats);
+}
+
+static boolean_t
+default_zone_locked(malloc_zone_t *zone)
+{
+ zone = runtime_default_zone();
+
+ return zone->introspect->zone_locked(zone);
+}
+
+static void
+default_zone_reinit_lock(malloc_zone_t *zone)
+{
+ zone = runtime_default_zone();
+
+ return zone->introspect->reinit_lock(zone);
+}
+
+static struct malloc_introspection_t default_zone_introspect = {
+ default_zone_ptr_in_use_enumerator,
+ default_zone_good_size,
+ default_zone_check,
+ default_zone_print,
+ default_zone_log,
+ default_zone_force_lock,
+ default_zone_force_unlock,
+ default_zone_statistics,
+ default_zone_locked,
+ NULL,
+ NULL,
+ NULL,
+ NULL,
+ default_zone_reinit_lock
+};
+
+typedef struct {
+ malloc_zone_t malloc_zone;
+ uint8_t pad[PAGE_MAX_SIZE - sizeof(malloc_zone_t)];
+} virtual_default_zone_t;
+
+static virtual_default_zone_t virtual_default_zone
+__attribute__((section("__DATA,__v_zone")))
+__attribute__((aligned(PAGE_MAX_SIZE))) = {
+ NULL,
+ NULL,
+ default_zone_size,
+ default_zone_malloc,
+ default_zone_calloc,
+ default_zone_valloc,
+ default_zone_free,
+ default_zone_realloc,
+ default_zone_destroy,
+ DEFAULT_MALLOC_ZONE_STRING,
+ default_zone_batch_malloc,
+ default_zone_batch_free,
+ &default_zone_introspect,
+ 10,
+ default_zone_memalign,
+ default_zone_free_definite_size,
+ default_zone_pressure_relief,
+ default_zone_malloc_claimed_address,
+};
+
+MALLOC_NOEXPORT malloc_zone_t *default_zone = &virtual_default_zone.malloc_zone;
+
+MALLOC_NOEXPORT
+/*static*/ boolean_t
+has_default_zone0(void)
+{
+ return !has_injected_zone0;
+}
+
+static inline malloc_zone_t *_find_registered_zone(const void *, size_t *, bool) __attribute__((always_inline));
+static inline malloc_zone_t *
+_find_registered_zone(const void *ptr, size_t *returned_size, bool known_non_default)
+{
// Returns a zone which contains ptr, else NULL
if (0 == malloc_num_zones) {
- if (returned_size) *returned_size = 0;
+ if (returned_size) {
+ *returned_size = 0;
+ }
return NULL;
}
- // The default zone is registered in malloc_zones[0]. There's no danger that it will ever be unregistered.
- // So don't advance the FRZ counter yet.
- malloc_zone_t *zone = malloc_zones[0];
- size_t size = zone->size(zone, ptr);
- if (size) { // Claimed by this zone?
- if (returned_size) *returned_size = size;
- return zone;
- }
-
- int *pFRZCounter = pFRZCounterLive; // Capture pointer to the counter of the moment
- __sync_fetch_and_add(pFRZCounter, 1); // Advance this counter -- our thread is in FRZ
-
- unsigned index;
- int32_t limit = *(volatile int32_t *)&malloc_num_zones;
- malloc_zone_t **zones = &malloc_zones[1];
+ // first look in the lite zone
+ if (lite_zone) {
+ malloc_zone_t *zone = lite_zone;
+ size_t size = zone->size(zone, ptr);
+ if (size) { // Claimed by this zone?
+ if (returned_size) {
+ *returned_size = size;
+ }
+ // Return the virtual default zone instead of the lite zone - see <rdar://problem/24994311>
+ return default_zone;
+ }
+ }
+
+ malloc_zone_t *zone;
+ size_t size;
+
+ // We assume that the initial zones will never be unregistered concurrently while this code is running so we can have
+ // a fast path without synchronization. Callers who really do unregister these (to install their own default zone) need
+ // to ensure they establish their zone setup during initialization and before entering a multi-threaded environment.
+ for (uint32_t i = known_non_default ? 1 : 0; i < initial_num_zones; i++) {
+ zone = malloc_zones[i];
+ size = zone->size(zone, ptr);
+
+ if (size) { // Claimed by this zone?
+ if (returned_size) {
+ *returned_size = size;
+ }
+
+ // Asan and others replace the zone at position 0 with their own zone.
+ // In that case just return that zone as they need this information.
+ // Otherwise return the virtual default zone, not the actual zone in position 0.
+ if (i == 0 && has_default_zone0()) {
+ return default_zone;
+ }
+
+ return zone;
+ }
+ }
+
+ int32_t volatile *pFRZCounter = pFRZCounterLive; // Capture pointer to the counter of the moment
+ OSAtomicIncrement32Barrier(pFRZCounter); // Advance this counter -- our thread is in FRZ
+
+ int32_t limit = *(int32_t volatile *)&malloc_num_zones;
// From this point on, FRZ is accessing the malloc_zones[] array without locking
// in order to avoid contention on common operations (such as non-default-zone free()).
@@ -257,23 +1158,34 @@
// are still valid). It also ensures that all the pointers in the zones array are
// valid until it returns, so that a stale value in limit is not dangerous.
- for (index = 1; index < limit; ++index, ++zones) {
- zone = *zones;
+ for (uint32_t i = initial_num_zones; i < limit; i++) {
+ zone = malloc_zones[i];
size = zone->size(zone, ptr);
if (size) { // Claimed by this zone?
- if (returned_size) *returned_size = size;
- __sync_fetch_and_sub(pFRZCounter, 1); // our thread is leaving FRZ
- return zone;
+ goto out;
}
}
// Unclaimed by any zone.
- if (returned_size) *returned_size = 0;
- __sync_fetch_and_sub(pFRZCounter, 1); // our thread is leaving FRZ
- return NULL;
-}
-
-__attribute__((visibility("hidden"))) __attribute__((noinline)) void
-malloc_error_break(void) {
+ zone = NULL;
+ size = 0;
+out:
+ if (returned_size) {
+ *returned_size = size;
+ }
+ OSAtomicDecrement32Barrier(pFRZCounter); // our thread is leaving FRZ
+ return zone;
+}
+
+malloc_zone_t *
+find_registered_zone(const void *ptr, size_t *returned_size,
+ bool known_non_default)
+{
+ return _find_registered_zone(ptr, returned_size, known_non_default);
+}
+
+void
+malloc_error_break(void)
+{
// Provides a non-inlined place for various malloc error procedures to call
// that will be called after an error message appears. It does not make
// sense for developers to call this function, so it is marked
@@ -281,10 +1193,9 @@
MAGMALLOC_MALLOCERRORBREAK(); // DTrace USDT probe
}
-extern boolean_t __stack_logging_locked();
-
-__attribute__((visibility("hidden"))) __attribute__((noinline)) __attribute__((used)) int
-malloc_gdb_po_unsafe(void) {
+int
+malloc_gdb_po_unsafe(void)
+{
// In order to implement "po" other data formatters in gdb, the debugger
// calls functions that call malloc. The debugger will only run one thread
// of the program in this case, so if another thread is holding a zone lock,
@@ -294,8 +1205,9 @@
// lock. If trylock succeeds, unlock it, otherwise return "locked". Returns
// 0 == safe, 1 == locked/unsafe.
- if (__stack_logging_locked())
+ if (msl.stack_logging_locked && msl.stack_logging_locked()) {
return 1;
+ }
malloc_zone_t **zones = malloc_zones;
unsigned i, e = malloc_num_zones;
@@ -304,11 +1216,13 @@
malloc_zone_t *zone = zones[i];
// Version must be >= 5 to look at the new introspection field.
- if (zone->version < 5)
+ if (zone->version < 5) {
continue;
-
- if (zone->introspect->zone_locked && zone->introspect->zone_locked(zone))
+ }
+
+ if (zone->introspect->zone_locked && zone->introspect->zone_locked(zone)) {
return 1;
+ }
}
return 0;
}
@@ -317,18 +1231,20 @@
static void set_flags_from_environment(void);
-static void
-malloc_zone_register_while_locked(malloc_zone_t *zone) {
+MALLOC_NOEXPORT void
+malloc_zone_register_while_locked(malloc_zone_t *zone, bool make_default)
+{
size_t protect_size;
unsigned i;
/* scan the list of zones, to see if this zone is already registered. If
* so, print an error message and return. */
- for (i = 0; i != malloc_num_zones; ++i)
+ for (i = 0; i != malloc_num_zones; ++i) {
if (zone == malloc_zones[i]) {
- _malloc_printf(ASL_LEVEL_ERR, "Attempted to register zone more than once: %p\n", zone);
+ malloc_report(ASL_LEVEL_ERR, "Attempted to register zone more than once: %p\n", zone);
return;
}
+ }
if (malloc_num_zones == malloc_num_zones_allocated) {
size_t malloc_zones_size = malloc_num_zones * sizeof(malloc_zone_t *);
@@ -339,25 +1255,25 @@
vm_addr = vm_page_size;
kern_return_t kr = mach_vm_allocate(mach_task_self(), &vm_addr, alloc_size, alloc_flags);
if (kr) {
- _malloc_printf(ASL_LEVEL_ERR, "malloc_zone_register allocation failed: %d\n", kr);
+ malloc_report(ASL_LEVEL_ERR, "malloc_zone_register allocation failed: %d\n", kr);
return;
}
- malloc_zone_t **new_zones = (uintptr_t)vm_addr;
+ malloc_zone_t **new_zones = (malloc_zone_t **)vm_addr;
/* If there were previously allocated malloc zones, we need to copy them
* out of the previous array and into the new zones array */
if (malloc_zones) {
memcpy(new_zones, malloc_zones, malloc_zones_size);
- vm_addr = malloc_zones;
+ vm_addr = (mach_vm_address_t)malloc_zones;
mach_vm_size_t dealloc_size = round_page(malloc_zones_size);
mach_vm_deallocate(mach_task_self(), vm_addr, dealloc_size);
}
/* Update the malloc_zones pointer, which we leak if it was previously
* allocated, and the number of zones allocated */
- protect_size = alloc_size;
+ protect_size = (size_t)alloc_size;
malloc_zones = new_zones;
- malloc_num_zones_allocated = alloc_size / sizeof(malloc_zone_t *);
+ malloc_num_zones_allocated = (int32_t)(alloc_size / sizeof(malloc_zone_t *));
} else {
/* If we don't need to reallocate zones, we need to briefly change the
* page protection the malloc zones to allow writes */
@@ -365,180 +1281,353 @@
mprotect(malloc_zones, protect_size, PROT_READ | PROT_WRITE);
}
+ if (make_default) {
+ memmove(&malloc_zones[1], &malloc_zones[0], malloc_num_zones * sizeof(malloc_zone_t *));
+ malloc_zones[0] = zone;
+ } else {
+ malloc_zones[malloc_num_zones] = zone;
+ }
+
/* <rdar://problem/12871662> This store-increment needs to be visible in the correct
* order to any threads in find_registered_zone, such that if the incremented value
* in malloc_num_zones is visible then the pointer write before it must also be visible.
*
* While we could be slightly more efficent here with atomic ops the cleanest way to
- * ensure the proper store-release operation is performed is to use __sync... to update
- * malloc_num_zones.
+ * ensure the proper store-release operation is performed is to use OSAtomic*Barrier
+ * to update malloc_num_zones.
*/
- malloc_zones[malloc_num_zones] = zone;
- __sync_fetch_and_add(&malloc_num_zones, 1);
+ OSAtomicIncrement32Barrier(&malloc_num_zones);
/* Finally, now that the zone is registered, disallow write access to the
* malloc_zones array */
mprotect(malloc_zones, protect_size, PROT_READ);
- //_malloc_printf(ASL_LEVEL_INFO, "Registered malloc_zone %p in malloc_zones %p [%u zones, %u bytes]\n", zone, malloc_zones, malloc_num_zones, protect_size);
-}
-
+
+ // ASan and others interpose on mprotect() to ensure their zone is always at position 0.
+ if (make_default && malloc_zones[0] != zone) {
+ has_injected_zone0 = true;
+ malloc_slowpath_update();
+ }
+
+ //malloc_report(ASL_LEVEL_INFO, "Registered malloc_zone %p in malloc_zones %p [%u zones, %u bytes]\n", zone, malloc_zones,
+ // malloc_num_zones, protect_size);
+}
+
+// This used to be called lazyily because it is using
+// dyld_process_is_restricted() before dyld_init() has run.
+//
+// However this function is safe to use, we keep this function separate
+// if we ever need to have a 2-stage init in the future.
static void
-_malloc_initialize(void) {
- MALLOC_LOCK();
- if (!_malloc_is_initialized) {
- unsigned n;
- malloc_zone_t *zone;
-
- _malloc_is_initialized = TRUE;
-
- set_flags_from_environment(); // will only set flags up to two times
- n = malloc_num_zones;
-
+_malloc_initialize(const char *apple[], const char *bootargs)
+{
+ phys_ncpus = *(uint8_t *)(uintptr_t)_COMM_PAGE_PHYSICAL_CPUS;
+ logical_ncpus = *(uint8_t *)(uintptr_t)_COMM_PAGE_LOGICAL_CPUS;
+#if CONFIG_MAGAZINE_PER_CLUSTER
+ {
+ ncpuclusters = *(uint8_t *)(uintptr_t)_COMM_PAGE_CPU_CLUSTERS;
+ }
+#endif // CONFIG_MAGAZINE_PER_CLUSTER
+
+ if (0 != (logical_ncpus % phys_ncpus)) {
+ MALLOC_REPORT_FATAL_ERROR(logical_ncpus % phys_ncpus,
+ "logical_ncpus %% phys_ncpus != 0\n");
+ }
+
+ switch (logical_ncpus / phys_ncpus) {
+ case 1:
+ hyper_shift = 0;
+ break;
+ case 2:
+ hyper_shift = 1;
+ break;
+ case 4:
+ hyper_shift = 2;
+ break;
+ default:
+ MALLOC_REPORT_FATAL_ERROR(logical_ncpus / phys_ncpus, "logical_ncpus / phys_ncpus not 1, 2, or 4");
+ }
+
+ // max_magazines may already be set from a boot argument. Make sure that it
+ // is bounded by the number of CPUs.
+ if (max_magazines) {
+ max_magazines = MIN(max_magazines, logical_ncpus);
+ } else {
+ max_magazines = logical_ncpus;
+ }
+
+ // similiarly, cap medium magazines at logical_ncpus but don't cap it by
+ // the max magazines if it has been set explicitly
+ if (max_medium_magazines) {
+ max_medium_magazines = MIN(max_medium_magazines, logical_ncpus);
+ } else {
+ max_medium_magazines = max_magazines;
+ }
+
+ // Don't enable for pre-AMP iOS hardware, which we identify as "iOS, plain
+ // arm64"
+#if CONFIG_MAGAZINE_PER_CLUSTER && \
+ (MALLOC_TARGET_IOS_ONLY || MALLOC_TARGET_DK_IOS) && \
+ defined(__arm64__) && !defined(__arm64e__)
+ if (ncpuclusters == 1) {
+#if CONFIG_FEATUREFLAGS_SIMPLE
+ // Not with the other feature flag checks because ncpuclusters needs to
+ // be initialized first
+ bool secure_allocator_single_cluster = os_feature_enabled_simple(
+ libmalloc, SecureAllocator_SingleCluster, false);
+#else // CONFIG_FEATUREFLAGS_SIMPLE
+ bool secure_allocator_single_cluster = false;
+#endif // CONFIG_FEATUREFLAGS_SIMPLE
+ if (malloc_xzone_enabled && !secure_allocator_single_cluster) {
+ malloc_xzone_enabled = false;
+ }
+ }
+#endif
+
+ _malloc_detect_interposition();
+
+
+ set_flags_from_environment();
+
+
+#if CONFIG_SANITIZER
+ malloc_sanitizer_enabled = sanitizer_should_enable();
+#endif
+
#if CONFIG_NANOZONE
- malloc_zone_t *helper_zone = create_scalable_zone(0, malloc_debug_flags);
- zone = create_nano_zone(0, helper_zone, malloc_debug_flags);
-
- if (zone) {
- malloc_zone_register_while_locked(zone);
- malloc_zone_register_while_locked(helper_zone);
-
- // Must call malloc_set_zone_name() *after* helper and nano are hooked together.
- malloc_set_zone_name(zone, DEFAULT_MALLOC_ZONE_STRING);
+ // TODO: envp should be passed down from Libsystem
+ const char **envp = (const char **)*_NSGetEnviron();
+
+ // Disable nano when:
+ // - sanitizer is enabled, to avoid speculative out-of-bounds
+ // use-after-free reads that nano/nanov2 performs, OR
+ // - MallocScribble is enabled, which nanov2 does not implement
+ // - MallocCheckZeroOnFreeCorruption sampling is enabled, which nanov2 does
+ // not implement
+ if (!malloc_sanitizer_enabled &&
+ !(malloc_debug_flags & MALLOC_DO_SCRIBBLE) &&
+ !malloc_zero_on_free_sample_period) {
+ nano_common_init(envp, apple, bootargs);
+ }
+#endif // CONFIG_NANOZONE
+
+ bool nano_on_xzone = false;
+
+
+ if (!initial_xzone_zone || nano_on_xzone) {
+ if (!initial_xzone_zone) {
+ initial_scalable_zone = create_scalable_zone(0, malloc_debug_flags);
+ malloc_set_zone_name(initial_scalable_zone, DEFAULT_MALLOC_ZONE_STRING);
+ malloc_zone_register_while_locked(initial_scalable_zone, /*make_default=*/true);
+ }
+
+#if CONFIG_NANOZONE
+ nano_common_configure();
+
+ malloc_zone_t *helper_zone = initial_xzone_zone ?: initial_scalable_zone;
+
+ if (_malloc_engaged_nano == NANO_V2) {
+ if (malloc_report_config) {
+ bool nano_on_xzone_enabled = (helper_zone == initial_xzone_zone);
+ malloc_report(ASL_LEVEL_INFO, "NanoV2 Config:\n"
+ "\tNano On Xzone: %d\n",
+ nano_on_xzone_enabled);
+ }
+ initial_nano_zone = nanov2_create_zone(helper_zone, malloc_debug_flags);
+ }
+
+ if (initial_nano_zone) {
+ malloc_set_zone_name(initial_nano_zone, DEFAULT_MALLOC_ZONE_STRING);
malloc_set_zone_name(helper_zone, MALLOC_HELPER_ZONE_STRING);
+ malloc_zone_register_while_locked(initial_nano_zone, /*make_default=*/true);
+ }
+#endif
+ }
+
+#if CONFIG_SANITIZER
+ if (malloc_sanitizer_enabled) {
+ malloc_zone_t *wrapped_zone = malloc_zones[0];
+ malloc_zone_t *sanitizer_zone = sanitizer_create_zone(wrapped_zone);
+ malloc_zone_register_while_locked(sanitizer_zone, /*make_default=*/true);
+ }
+#endif
+
+ // Initialize slowpath check after we've registered zones so that
+ // malloc_num_zones check is meaningful
+ malloc_slowpath_update();
+
+ initial_num_zones = malloc_num_zones;
+
+#if CONFIG_DEFERRED_RECLAIM
+ kern_return_t vmdr_kr = KERN_SUCCESS;
+ if (large_cache_enabled) {
+ if (initial_xzone_zone) {
+ // xzone_malloc will own the deferred_reclaim buffer
+ large_cache_enabled = false;
} else {
- zone = helper_zone;
- malloc_zone_register_while_locked(zone);
- malloc_set_zone_name(zone, DEFAULT_MALLOC_ZONE_STRING);
- }
-#else
- zone = create_scalable_zone(0, malloc_debug_flags);
- malloc_zone_register_while_locked(zone);
- malloc_set_zone_name(zone, DEFAULT_MALLOC_ZONE_STRING);
-#endif
-
- if (n != 0) { // make the default first, for efficiency
- unsigned protect_size = malloc_num_zones_allocated * sizeof(malloc_zone_t *);
- malloc_zone_t *hold = malloc_zones[0];
-
- if(hold->zone_name && strcmp(hold->zone_name, DEFAULT_MALLOC_ZONE_STRING) == 0) {
- malloc_set_zone_name(hold, NULL);
+ vmdr_kr = mvm_deferred_reclaim_init();
+ if (vmdr_kr != KERN_SUCCESS) {
+ large_cache_enabled = false;
+ malloc_report(ASL_LEVEL_ERR, "Unable to set up reclaim buffer (%d) - disabling large cache\n", vmdr_kr);
}
-
- mprotect(malloc_zones, protect_size, PROT_READ | PROT_WRITE);
- malloc_zones[0] = malloc_zones[n];
- malloc_zones[n] = hold;
- mprotect(malloc_zones, protect_size, PROT_READ);
- }
- // _malloc_printf(ASL_LEVEL_INFO, "%d registered zones\n", malloc_num_zones);
- // _malloc_printf(ASL_LEVEL_INFO, "malloc_zones is at %p; malloc_num_zones is at %p\n", (unsigned)&malloc_zones, (unsigned)&malloc_num_zones);
- }
- MALLOC_UNLOCK();
-}
-
-static inline malloc_zone_t *inline_malloc_default_zone(void) __attribute__((always_inline));
+ }
+ }
+#endif /* CONFIG_DEFERRED_RECLAIM */
+
+ if (malloc_report_config && initial_scalable_zone) {
+ bool scribble = !!(malloc_debug_flags & MALLOC_DO_SCRIBBLE);
+ malloc_report(ASL_LEVEL_INFO, "Magazine Config:\n"
+ "\tMax Magazines: %d\n"
+ "\tMedium Enabled: %d\n"
+ "\tAggressive Madvise: %d\n"
+#if CONFIG_DEFERRED_RECLAIM
+ "\tLarge Cache: %d%s\n"
+#endif // CONFIG_DEFERRED_RECLAIM
+ "\tScribble: %d\n",
+ max_magazines, magazine_medium_enabled,
+ aggressive_madvise_enabled,
+#if CONFIG_DEFERRED_RECLAIM
+ vmdr_kr ?: large_cache_enabled, vmdr_kr ? " (ERROR)" : "",
+#endif // CONFIG_DEFERRED_RECLAIM
+ scribble);
+ }
+
+#if CONFIG_MEDIUM_ALLOCATOR
+ uint64_t memsize = platform_hw_memsize();
+ if (memsize >= 8 * magazine_medium_madvise_dram_scale_divisor) {
+ magazine_medium_madvise_window_scale_factor = 8;
+ } else if (memsize >= 4 * magazine_medium_madvise_dram_scale_divisor) {
+ magazine_medium_madvise_window_scale_factor = 4;
+ } else if (memsize >= 2 * magazine_medium_madvise_dram_scale_divisor) {
+ magazine_medium_madvise_window_scale_factor = 2;
+ } else {
+ magazine_medium_madvise_window_scale_factor = 1;
+ }
+#endif /* CONFIG_MEDIUM_ALLOCATOR */
+
+ // malloc_report(ASL_LEVEL_INFO, "%d registered zones\n", malloc_num_zones);
+ // malloc_report(ASL_LEVEL_INFO, "malloc_zones is at %p; malloc_num_zones is at %p\n", (unsigned)&malloc_zones,
+ // (unsigned)&malloc_num_zones);
+}
+
+static bool
+enable_pgm(unsigned flags)
+{
+ bool other_debug_tool = has_injected_zone0 || malloc_sanitizer_enabled;
+ // To avoid allocations in the lite helper zone that don't have msl data at
+ // the end of the allocation, don't enable PGM on the lite helper zone
+ bool zone_is_msl = flags & MALLOC_MSL_LITE_WRAPPED_ZONE_FLAGS;
+ return !other_debug_tool && !zone_is_msl && pgm_should_enable();
+}
+
+static void
+register_pgm_zone(bool internal_diagnostics)
+{
+ pgm_init_config(internal_diagnostics);
+ if (enable_pgm(0)) {
+ malloc_zone_t *wrapped_zone = malloc_zones[0];
+ malloc_zone_t *pgm_zone = pgm_create_zone(wrapped_zone);
+ malloc_zone_register_while_locked(pgm_zone, /*make_default=*/true);
+ }
+}
+
static inline malloc_zone_t *
-inline_malloc_default_zone(void) {
- if (!_malloc_is_initialized) _malloc_initialize();
- // _malloc_printf(ASL_LEVEL_INFO, "In inline_malloc_default_zone with %d %d\n", malloc_num_zones, malloc_has_debug_zone);
+inline_malloc_default_zone(void)
+{
+ // malloc_report(ASL_LEVEL_INFO, "In inline_malloc_default_zone with %d %d\n", malloc_num_zones, malloc_has_debug_zone);
return malloc_zones[0];
}
malloc_zone_t *
-malloc_default_zone(void) {
- return inline_malloc_default_zone();
-}
-
-static inline malloc_zone_t *inline_malloc_default_scalable_zone(void) __attribute__((always_inline));
-static inline malloc_zone_t *
-inline_malloc_default_scalable_zone(void) {
- unsigned index;
-
- if (!_malloc_is_initialized) _malloc_initialize();
- // _malloc_printf(ASL_LEVEL_INFO, "In inline_malloc_default_scalable_zone with %d %d\n", malloc_num_zones, malloc_has_debug_zone);
-
- MALLOC_LOCK();
+malloc_default_zone(void)
+{
+ return default_zone;
+}
+
+static void *
+legacy_zeroing_large_malloc(malloc_zone_t *zone, size_t size)
+{
+ if (size > LEGACY_ZEROING_THRESHOLD) {
+ // Leopard and earlier returned a ZFOD range, so clear to zero always,
+ // ham-handedly touching in each page
+ return default_zone_calloc(zone, 1, size);
+ } else {
+ return default_zone_malloc(zone, size);
+ }
+}
+
+static void *
+legacy_zeroing_large_valloc(malloc_zone_t *zone, size_t size)
+{
+ void *p = default_zone_valloc(zone, size);
+
+ // Leopard and earlier returned a ZFOD range, so ...
+ memset(p, 0, size); // Clear to zero always, ham-handedly touching in each page
+ return p;
+}
+
+void
+zeroify_scalable_zone(malloc_zone_t *zone)
+{
+ // <rdar://problem/27190324> this checkfix should replace the default zone's
+ // allocation routines with the zeroing versions. Instead of getting in hot
+ // water with the wrong zone, ensure that we're mutating the zone we expect.
+ //
+ // Additionally, the default_zone is no longer PROT_READ, so the two mprotect
+ // calls that were here are no longer needed.
+ if (zone == default_zone) {
+ zone->malloc = (void *)legacy_zeroing_large_malloc;
+ zone->valloc = (void *)legacy_zeroing_large_valloc;
+ }
+}
+
+/*
+ * Returns the version of the Nano allocator that's in use, or 0 if not.
+ */
+int
+malloc_engaged_nano(void)
+{
#if CONFIG_NANOZONE
- for (index = 0; index < malloc_num_zones; ++index) {
- malloc_zone_t *z = malloc_zones[index];
-
- if(z->zone_name && strcmp(z->zone_name, MALLOC_HELPER_ZONE_STRING) == 0) {
- MALLOC_UNLOCK();
- return z;
- }
- }
-#endif
- for (index = 0; index < malloc_num_zones; ++index) {
- malloc_zone_t *z = malloc_zones[index];
-
- if(z->zone_name && strcmp(z->zone_name, DEFAULT_MALLOC_ZONE_STRING) == 0) {
- MALLOC_UNLOCK();
- return z;
- }
- }
- MALLOC_UNLOCK();
-
- malloc_printf("*** malloc_default_scalable_zone() failed to find 'DefaultMallocZone'\n");
- return NULL; // FIXME: abort() instead?
-}
-
-/*
- * malloc_engaged_nano() is for the benefit of libdispatch, which calls here just once.
- */
-boolean_t malloc_engaged_nano(void)
-{
-#if CONFIG_NANOZONE
- return _malloc_engaged_nano;
+ return (initial_nano_zone || initial_xzone_zone) ? _malloc_engaged_nano : 0;
#else
return 0;
#endif
}
-malloc_zone_t *
-malloc_default_purgeable_zone(void) {
- static malloc_zone_t *dpz;
-
- if (!dpz) {
- //
- // PR_7288598: Must pass a *scalable* zone (szone) as the helper for create_purgeable_zone().
- // Take care that the zone so obtained is not subject to interposing.
- //
- malloc_zone_t *tmp = create_purgeable_zone(0, inline_malloc_default_scalable_zone(), malloc_debug_flags);
- malloc_zone_register(tmp);
- malloc_set_zone_name(tmp, DEFAULT_PUREGEABLE_ZONE_STRING);
- if (!__sync_bool_compare_and_swap(&dpz, NULL, tmp))
- malloc_destroy_zone(tmp);
- }
- return dpz;
+int
+malloc_engaged_secure_allocator(void)
+{
+ return !!initial_xzone_zone;
}
static void
-set_flags_from_environment(void) {
- const char *flag;
- int fd;
- char **env = * _NSGetEnviron();
- char **p;
- char *c;
-
- if (malloc_debug_file != STDERR_FILENO) {
- close(malloc_debug_file);
- malloc_debug_file = STDERR_FILENO;
- }
-#if __LP64__
- malloc_debug_flags = SCALABLE_MALLOC_ABORT_ON_CORRUPTION; // Set always on 64-bit processes
+set_flags_from_environment(void)
+{
+ const char *flag;
+ const char **env = (const char **)*_NSGetEnviron();
+ const char **p;
+ const char *c;
+
+#if defined(__LP64__)
+ malloc_debug_flags = MALLOC_ABORT_ON_CORRUPTION; // Set always on 64-bit processes
#else
- int libSystemVersion = NSVersionOfLinkTimeLibrary("System");
- if ((-1 != libSystemVersion) && ((libSystemVersion >> 16) < 126) /* Lion or greater */)
+ int libSystemVersion = NSVersionOfLinkTimeLibrary("System");
+ if ((-1 != libSystemVersion) && ((libSystemVersion >> 16) < 126) /* Lion or greater */) {
malloc_debug_flags = 0;
- else
- malloc_debug_flags = SCALABLE_MALLOC_ABORT_ON_CORRUPTION;
+ } else {
+ malloc_debug_flags = MALLOC_ABORT_ON_CORRUPTION;
+ }
#endif
- stack_logging_enable_logging = 0;
- stack_logging_dontcompact = 0;
- malloc_logger = NULL;
- malloc_check_start = 0;
- malloc_check_each = 1000;
- malloc_check_abort = 0;
- malloc_check_sleep = 100;
+
+#if TARGET_OS_OSX
+ // rdar://99288027
+ if (!dyld_program_sdk_at_least(dyld_platform_version_macOS_13_0)) {
+ if (malloc_zero_policy == MALLOC_ZERO_ON_FREE) {
+ malloc_zero_policy = MALLOC_ZERO_ON_ALLOC;
+ }
+ }
+#else // TARGET_OS_OSX
+#endif // TARGET_OS_OSX
+
/*
* Given that all environment variables start with "Malloc" we optimize by scanning quickly
* first the environment, therefore avoiding repeated calls to getenv().
@@ -546,159 +1635,431 @@
* our behaviour.
*/
for (p = env; (c = *p) != NULL; ++p) {
+#if RDAR_48993662
+ if (!strncmp(c, "Malloc", 6) || !strncmp(c, "_Malloc", 6)) {
+#else // RDAR_48993662
if (!strncmp(c, "Malloc", 6)) {
- if (issetugid())
+#endif // RDAR_48993662
+ if (issetugid()) {
return;
+ }
break;
}
}
- if (c == NULL)
+
+ /*
+ * Deny certain flags for entitled processes rdar://problem/13521742
+ * MallocLogFile & MallocCorruptionAbort
+ * as these provide the ability to turn *off* aborting in error cases.
+ */
+ bool restricted = dyld_process_is_restricted();
+ malloc_print_configure(restricted);
+
+ if (c == NULL) {
return;
- flag = getenv("MallocLogFile");
+ }
+
+ flag = getenv("MallocGuardEdges");
if (flag) {
- fd = open(flag, O_WRONLY|O_APPEND|O_CREAT, 0644);
- if (fd >= 0) {
- malloc_debug_file = fd;
- fcntl(fd, F_SETFD, 0); // clear close-on-exec flag XXX why?
+ if (!strcmp(flag, "all")) {
+ // "MallocGuardEdges=all" adds guard page(s) for every region.
+ // Do not do this on 32-bit platforms because there is insufficient
+ // address space. These pages are always protected.
+#if MALLOC_TARGET_64BIT
+ malloc_debug_flags |= MALLOC_GUARD_ALL | MALLOC_ADD_GUARD_PAGE_FLAGS;
+ malloc_debug_flags &= ~(MALLOC_DONT_PROTECT_PRELUDE|MALLOC_DONT_PROTECT_POSTLUDE);
+ malloc_report(ASL_LEVEL_INFO, "adding guard pages to all regions\n");
+#endif // MALLOC_TARGET_64BIT
} else {
- malloc_printf("Could not open %s, using stderr\n", flag);
- }
- }
- if (getenv("MallocGuardEdges")) {
- malloc_debug_flags |= SCALABLE_MALLOC_ADD_GUARD_PAGES;
- _malloc_printf(ASL_LEVEL_INFO, "protecting edges\n");
- if (getenv("MallocDoNotProtectPrelude")) {
- malloc_debug_flags |= SCALABLE_MALLOC_DONT_PROTECT_PRELUDE;
- _malloc_printf(ASL_LEVEL_INFO, "... but not protecting prelude guard page\n");
- }
- if (getenv("MallocDoNotProtectPostlude")) {
- malloc_debug_flags |= SCALABLE_MALLOC_DONT_PROTECT_POSTLUDE;
- _malloc_printf(ASL_LEVEL_INFO, "... but not protecting postlude guard page\n");
- }
- }
- flag = getenv("MallocStackLogging");
- if (!flag) {
- flag = getenv("MallocStackLoggingNoCompact");
- stack_logging_dontcompact = 1;
- }
- if (flag) {
- // Set up stack logging as early as possible to catch all ensuing VM allocations,
- // including those from _malloc_printf and malloc zone setup. Make sure to set
- // __syscall_logger after this, because prepare_to_log_stacks() itself makes VM
- // allocations that we aren't prepared to log yet.
- __prepare_to_log_stacks();
- if (strcmp(flag,"malloc") == 0) {
- malloc_logger = __disk_stack_logging_log_stack;
- _malloc_printf(ASL_LEVEL_INFO, "recording malloc (but not VM allocation) stacks to disk using standard recorder\n");
- } else if (strcmp(flag,"vm") == 0) {
- __syscall_logger = __disk_stack_logging_log_stack;
- _malloc_printf(ASL_LEVEL_INFO, "recording VM allocation (but not malloc) stacks to disk using standard recorder\n");
- } else {
- malloc_logger = __disk_stack_logging_log_stack;
- __syscall_logger = __disk_stack_logging_log_stack;
- _malloc_printf(ASL_LEVEL_INFO, "recording malloc and VM allocation stacks to disk using standard recorder\n");
- }
- stack_logging_enable_logging = 1;
- if (stack_logging_dontcompact) {
- if (malloc_logger == __disk_stack_logging_log_stack) {
- _malloc_printf(ASL_LEVEL_INFO, "stack logging compaction turned off; size of log files on disk can increase rapidly\n");
- } else {
- _malloc_printf(ASL_LEVEL_INFO, "stack logging compaction turned off; VM can increase rapidly\n");
+ malloc_debug_flags |= MALLOC_ADD_GUARD_PAGE_FLAGS;
+ malloc_debug_flags &= ~MALLOC_GUARD_ALL;
+ malloc_report(ASL_LEVEL_INFO, "adding guard pages for large allocator blocks\n");
+ if (getenv("MallocDoNotProtectPrelude")) {
+ malloc_debug_flags |= MALLOC_DONT_PROTECT_PRELUDE;
+ malloc_report(ASL_LEVEL_INFO, "... but not protecting prelude guard page\n");
}
- }
- }
+ if (getenv("MallocDoNotProtectPostlude")) {
+ malloc_debug_flags |= MALLOC_DONT_PROTECT_POSTLUDE;
+ malloc_report(ASL_LEVEL_INFO, "... but not protecting postlude guard page\n");
+ }
+ }
+ }
+
if (getenv("MallocScribble")) {
- malloc_debug_flags |= SCALABLE_MALLOC_DO_SCRIBBLE;
- _malloc_printf(ASL_LEVEL_INFO, "enabling scribbling to detect mods to free blocks\n");
+ malloc_debug_flags |= MALLOC_DO_SCRIBBLE;
+ malloc_report(ASL_LEVEL_INFO, "enabling scribbling to detect mods to free blocks\n");
}
if (getenv("MallocErrorAbort")) {
- malloc_debug_flags |= SCALABLE_MALLOC_ABORT_ON_ERROR;
- _malloc_printf(ASL_LEVEL_INFO, "enabling abort() on bad malloc or free\n");
- }
-#if CONFIG_NANOZONE
- /* Explicit overrides from the environment */
- if ((flag = getenv("MallocNanoZone"))) {
- if (flag[0] == '1') {
- _malloc_engaged_nano = 1;
- } else if (flag[0] == '0') {
- _malloc_engaged_nano = 0;
- }
- }
-#endif /* CONFIG_NANOZONE */
-
-#if __LP64__
- /* initialization above forces SCALABLE_MALLOC_ABORT_ON_CORRUPTION of 64-bit processes */
+ malloc_debug_flags |= MALLOC_ABORT_ON_ERROR;
+ malloc_report(ASL_LEVEL_INFO, "enabling abort() on bad malloc or free\n");
+ }
+ if (getenv("MallocTracing")) {
+ malloc_tracing_enabled = true;
+ }
+ if (getenv("MallocSimpleStackLogging")) {
+ malloc_simple_stack_logging = true;
+ }
+ if (getenv("MallocReportConfig")) {
+ malloc_report_config = true;
+ }
+
+#if defined(__LP64__)
+/* initialization above forces MALLOC_ABORT_ON_CORRUPTION of 64-bit processes */
#else
flag = getenv("MallocCorruptionAbort");
- if (flag && (flag[0] == '0')) { // Set from an environment variable in 32-bit processes
- malloc_debug_flags &= ~SCALABLE_MALLOC_ABORT_ON_CORRUPTION;
+ if (!restricted && flag && (flag[0] == '0')) { // Set from an environment variable in 32-bit processes
+ malloc_debug_flags &= ~MALLOC_ABORT_ON_CORRUPTION;
} else if (flag) {
- malloc_debug_flags |= SCALABLE_MALLOC_ABORT_ON_CORRUPTION;
+ malloc_debug_flags |= MALLOC_ABORT_ON_CORRUPTION;
}
#endif
flag = getenv("MallocCheckHeapStart");
if (flag) {
- malloc_check_start = strtoul(flag, NULL, 0);
- if (malloc_check_start == 0) malloc_check_start = 1;
- if (malloc_check_start == -1) malloc_check_start = 1;
+ malloc_check_start = (unsigned)strtoul(flag, NULL, 0);
+ if (malloc_check_start == 0) {
+ malloc_check_start = 1;
+ }
+ if (malloc_check_start == -1) {
+ malloc_check_start = 1;
+ }
flag = getenv("MallocCheckHeapEach");
if (flag) {
- malloc_check_each = strtoul(flag, NULL, 0);
- if (malloc_check_each == 0) malloc_check_each = 1;
- if (malloc_check_each == -1) malloc_check_each = 1;
- }
- _malloc_printf(ASL_LEVEL_INFO, "checks heap after %dth operation and each %d operations\n", malloc_check_start, malloc_check_each);
+ malloc_check_each = (unsigned)strtoul(flag, NULL, 0);
+ if (malloc_check_each == 0) {
+ malloc_check_each = 1;
+ }
+ if (malloc_check_each == -1) {
+ malloc_check_each = 1;
+ }
+ }
+ malloc_report(ASL_LEVEL_INFO, "checks heap after operation #%d and each %d operations\n", malloc_check_start, malloc_check_each);
flag = getenv("MallocCheckHeapAbort");
- if (flag)
- malloc_check_abort = strtol(flag, NULL, 0);
- if (malloc_check_abort)
- _malloc_printf(ASL_LEVEL_INFO, "will abort on heap corruption\n");
- else {
+ if (flag) {
+ malloc_check_abort = (unsigned)strtol(flag, NULL, 0);
+ }
+ if (malloc_check_abort) {
+ malloc_report(ASL_LEVEL_INFO, "will abort on heap corruption\n");
+ } else {
flag = getenv("MallocCheckHeapSleep");
- if (flag)
- malloc_check_sleep = strtol(flag, NULL, 0);
- if (malloc_check_sleep > 0)
- _malloc_printf(ASL_LEVEL_INFO, "will sleep for %d seconds on heap corruption\n", malloc_check_sleep);
- else if (malloc_check_sleep < 0)
- _malloc_printf(ASL_LEVEL_INFO, "will sleep once for %d seconds on heap corruption\n", -malloc_check_sleep);
- else
- _malloc_printf(ASL_LEVEL_INFO, "no sleep on heap corruption\n");
- }
- }
+ if (flag) {
+ malloc_check_sleep = (unsigned)strtol(flag, NULL, 0);
+ }
+ if (malloc_check_sleep > 0) {
+ malloc_report(ASL_LEVEL_INFO, "will sleep for %d seconds on heap corruption\n", malloc_check_sleep);
+ } else if (malloc_check_sleep < 0) {
+ malloc_report(ASL_LEVEL_INFO, "will sleep once for %d seconds on heap corruption\n", -malloc_check_sleep);
+ } else {
+ malloc_report(ASL_LEVEL_INFO, "no sleep on heap corruption\n");
+ }
+ }
+ }
+
+ flag = getenv("MallocMaxMagazines");
+#if RDAR_48993662
+ if (!flag) {
+ flag = getenv("_MallocMaxMagazines");
+ }
+#endif // RDAR_48993662
+ if (flag) {
+
+ int value = (int)strtol(flag, NULL, 0);
+ if (value == 0) {
+ malloc_report(ASL_LEVEL_INFO, "Maximum magazines defaulted to %d\n", max_magazines);
+#if CONFIG_MAGAZINE_PER_CLUSTER
+ } else if (value == UINT16_MAX) {
+ {
+ max_magazines = ncpuclusters;
+ malloc_report(ASL_LEVEL_INFO,
+ "Maximum magazines limited to ncpuclusters (%d)\n",
+ max_magazines);
+ }
+#endif // CONFIG_MAGAZINE_PER_CLUSTER
+ } else if (value < 0) {
+ malloc_report(ASL_LEVEL_ERR, "Maximum magazines must be positive - ignored.\n");
+ } else if (value > logical_ncpus) {
+ max_magazines = logical_ncpus;
+ malloc_report(ASL_LEVEL_INFO, "Maximum magazines limited to number of logical CPUs (%d)\n", max_magazines);
+ } else {
+ max_magazines = value;
+ malloc_report(ASL_LEVEL_INFO, "Maximum magazines set to %d\n", max_magazines);
+ }
+ }
+
+ flag = getenv("MallocLargeExpandedCacheThreshold");
+ if (flag) {
+ uint64_t value = (uint64_t)strtoull(flag, NULL, 0);
+ if (value == 0) {
+ malloc_report(ASL_LEVEL_INFO, "Large expanded cache threshold defaulted to %lly\n", magazine_large_expanded_cache_threshold);
+ } else if (value < 0) {
+ malloc_report(ASL_LEVEL_ERR, "MallocLargeExpandedCacheThreshold must be positive - ignored.\n");
+ } else {
+ magazine_large_expanded_cache_threshold = value;
+ malloc_report(ASL_LEVEL_INFO, "Large expanded cache threshold set to %lly\n", magazine_large_expanded_cache_threshold);
+ }
+ }
+
+ flag = getenv("MallocLargeDisableASLR");
+ if (flag) {
+ uint64_t value = (uint64_t)strtoull(flag, NULL, 0);
+ if (value == 0) {
+ malloc_report(ASL_LEVEL_INFO, "Enabling ASLR slide on large allocations\n");
+ malloc_debug_flags &= ~DISABLE_LARGE_ASLR;
+ } else if (value != 0) {
+ malloc_report(ASL_LEVEL_INFO, "Disabling ASLR slide on large allocations\n");
+ malloc_debug_flags |= DISABLE_LARGE_ASLR;
+ }
+ }
+
+#if CONFIG_AGGRESSIVE_MADVISE || CONFIG_LARGE_CACHE
+ // convenience flag to configure policies usually associated with memory-constrained platforms (iOS)
+ // that trade some amount of time efficiency for space efficiency
+ flag = getenv("MallocSpaceEfficient");
+ if (flag) {
+ const char *endp;
+ long value = malloc_common_convert_to_long(flag, &endp);
+ if (!*endp && endp != flag && (value == 0 || value == 1)) {
+#if CONFIG_AGGRESSIVE_MADVISE
+ aggressive_madvise_enabled = (value == 1);
+#endif // CONFIG_AGGRESSIVE_MADVISE
+#if CONFIG_LARGE_CACHE
+ // Disable the large cache in space efficient mode
+ if (value != 0){
+ large_cache_enabled = 0;
+ }
+#endif // CONFIG_LARGE_CACHE
+ malloc_space_efficient_enabled = (value == 1);
+ // consider disabling medium magazine if aggressive madvise is not sufficient
+ } else {
+ malloc_report(ASL_LEVEL_ERR, "MallocSpaceEfficient must be 0 or 1.\n");
+ }
+ }
+#endif // CONFIG_AGGRESSIVE_MADVISE || CONFIG_LARGE_CACHE
+
+#if CONFIG_MEDIUM_ALLOCATOR
+ flag = getenv("MallocMediumZone");
+ if (flag) {
+ int value = (unsigned)strtol(flag, NULL, 0);
+ if (value == 0) {
+ magazine_medium_enabled = false;
+ } else if (value == 1) {
+ magazine_medium_enabled = true;
+ }
+ }
+
+ flag = getenv("MallocMediumActivationThreshold");
+ if (flag) {
+ uint64_t value = (uint64_t)strtoull(flag, NULL, 0);
+ if (value == 0) {
+ malloc_report(ASL_LEVEL_INFO, "Medium activation threshold defaulted to %lly\n", magazine_medium_active_threshold);
+ } else if (value < 0) {
+ malloc_report(ASL_LEVEL_ERR, "MallocMediumActivationThreshold must be positive - ignored.\n");
+ } else {
+ magazine_medium_active_threshold = value;
+ malloc_report(ASL_LEVEL_INFO, "Medium activation threshold set to %lly\n", magazine_medium_active_threshold);
+ }
+ }
+
+ flag = getenv("MallocMediumSpaceEfficient");
+ if (flag) {
+ uint64_t value = (uint64_t)strtoull(flag, NULL, 0);
+ if (value == 0) {
+ malloc_medium_space_efficient_enabled = false;
+ } else if (value == 1) {
+ malloc_medium_space_efficient_enabled = true;
+ }
+ }
+
+ if (malloc_medium_space_efficient_enabled && malloc_space_efficient_enabled) {
+ // Bring down MallocMaxMediumMagazines to only a single magazine in
+ // space-efficent processes but do this before the envvar so that it
+ // can still be overridden at the command line.
+ max_medium_magazines = 1;
+ }
+
+ flag = getenv("MallocMaxMediumMagazines");
+#if RDAR_48993662
+ if (!flag) {
+ flag = getenv("_MallocMaxMediumMagazines");
+ }
+#endif // RDAR_48993662
+ if (flag) {
+ int value = (unsigned)strtol(flag, NULL, 0);
+ if (value == 0) {
+ malloc_report(ASL_LEVEL_INFO, "Maximum medium magazines defaulted to %d\n", max_magazines);
+ } else if (value < 0) {
+ malloc_report(ASL_LEVEL_ERR, "Maximum medium magazines must be positive - ignored.\n");
+ } else if (value > logical_ncpus) {
+ max_medium_magazines = logical_ncpus;
+ malloc_report(ASL_LEVEL_INFO, "Maximum medium magazines limited to number of logical CPUs (%d)\n", max_medium_magazines);
+ } else {
+ max_medium_magazines = value;
+ malloc_report(ASL_LEVEL_INFO, "Maximum medium magazines set to %d\n", max_medium_magazines);
+ }
+ }
+#endif // CONFIG_MEDIUM_ALLOCATOR
+
+#if CONFIG_AGGRESSIVE_MADVISE
+ flag = getenv("MallocAggressiveMadvise");
+ if (flag) {
+ const char *endp;
+ long value = malloc_common_convert_to_long(flag, &endp);
+ if (!*endp && endp != flag && (value == 0 || value == 1)) {
+ aggressive_madvise_enabled = (value == 1);
+ } else {
+ malloc_report(ASL_LEVEL_ERR, "MallocAggressiveMadvise must be 0 or 1.\n");
+ }
+ }
+#endif // CONFIG_AGGRESSIVE_MADVISE
+
+#if CONFIG_LARGE_CACHE
+ flag = getenv("MallocLargeCache");
+ if (flag) {
+ const char *endp;
+ long value = malloc_common_convert_to_long(flag, &endp);
+ if (!*endp && endp != flag && (value == 0 || value == 1)) {
+ large_cache_enabled = (value == 1);
+ } else {
+ malloc_report(ASL_LEVEL_ERR, "MallocLargeCache must be 0 or 1.\n");
+ }
+ }
+#endif // CONFIG_LARGE_CACHE
+
+#if CONFIG_RECIRC_DEPOT
+ flag = getenv("MallocRecircRetainedRegions");
+ if (flag) {
+ int value = (int)strtol(flag, NULL, 0);
+ if (value > 0) {
+ recirc_retained_regions = value;
+ } else {
+ malloc_report(ASL_LEVEL_ERR, "MallocRecircRetainedRegions must be positive - ignored.\n");
+ }
+ }
+#endif // CONFIG_RECIRC_DEPOT
+
+ flag = getenv("MallocZeroOnFree");
+ if (flag) {
+ const char *endp;
+ long value = malloc_common_convert_to_long(flag, &endp);
+ if (!*endp && endp != flag && (value == 0 || value == 1)) {
+ malloc_zero_policy = value ? MALLOC_ZERO_ON_FREE : MALLOC_ZERO_NONE;
+ } else {
+ malloc_report(ASL_LEVEL_ERR, "MallocZeroOnFree must be 0 or 1.\n");
+ }
+ }
+
+ flag = getenv("MallocZeroOnAlloc");
+ if (flag) {
+ const char *endp;
+ long value = malloc_common_convert_to_long(flag, &endp);
+ if (!*endp && endp != flag && (value == 0 || value == 1)) {
+ malloc_zero_policy = value ? MALLOC_ZERO_ON_ALLOC : MALLOC_ZERO_NONE;
+ } else {
+ malloc_report(ASL_LEVEL_ERR, "MallocZeroOnAlloc must be 0 or 1.\n");
+ }
+ }
+
+ flag = getenv("MallocCheckZeroOnFreeCorruption");
+ if (flag) {
+ int value = (int)strtol(flag, NULL, 0);
+ if (value > 0) {
+ malloc_zero_on_free_sample_period = value;
+ } else {
+ malloc_report(ASL_LEVEL_ERR,
+ "malloc_zero_on_free_sample_period must be positive - ignored.\n");
+ }
+ }
+
+ flag = getenv("MallocSecureAllocator");
+ if (flag && malloc_internal_security_policy) {
+ const char *endp;
+ long value = malloc_common_convert_to_long(flag, &endp);
+ if (!*endp && endp != flag && (value == 0 || value == 1)) {
+ malloc_xzone_enabled = value;
+ malloc_xzone_enabled_override = value ?
+ MALLOC_XZONE_OVERRIDE_ENABLED :
+ MALLOC_XZONE_OVERRIDE_DISABLED;
+ } else {
+ malloc_report(ASL_LEVEL_ERR, "MallocSecureAllocator must be 0 or 1.\n");
+ }
+ }
+
+ flag = getenv("MallocSecureAllocatorNano");
+ if (flag) {
+ const char *endp;
+ long value = malloc_common_convert_to_long(flag, &endp);
+ if (!*endp && endp != flag && (value == 0 || value == 1)) {
+ malloc_xzone_nano_override = value ? MALLOC_XZONE_OVERRIDE_ENABLED :
+ MALLOC_XZONE_OVERRIDE_DISABLED;
+ } else {
+ malloc_report(ASL_LEVEL_ERR, "MallocSecureAllocatorNano must be 0 or 1.\n");
+ }
+ }
+
+ flag = getenv("MallocNanoOnXzone");
+ if (flag) {
+ const char *endp;
+ long value = malloc_common_convert_to_long(flag, &endp);
+ if (!*endp && endp != flag && (value == 0 || value == 1)) {
+ malloc_nano_on_xzone_override = value ? MALLOC_XZONE_OVERRIDE_ENABLED :
+ MALLOC_XZONE_OVERRIDE_DISABLED;
+ } else {
+ malloc_report(ASL_LEVEL_ERR, "MallocNanoOnXzone must be 0 or 1.\n");
+ }
+ }
+
+
+
if (getenv("MallocHelp")) {
- _malloc_printf(ASL_LEVEL_INFO,
- "environment variables that can be set for debug:\n"
- "- MallocLogFile <f> to create/append messages to file <f> instead of stderr\n"
- "- MallocGuardEdges to add 2 guard pages for each large block\n"
- "- MallocDoNotProtectPrelude to disable protection (when previous flag set)\n"
- "- MallocDoNotProtectPostlude to disable protection (when previous flag set)\n"
- "- MallocStackLogging to record all stacks. Tools like leaks can then be applied\n"
- "- MallocStackLoggingNoCompact to record all stacks. Needed for malloc_history\n"
- "- MallocStackLoggingDirectory to set location of stack logs, which can grow large; default is /tmp\n"
- "- MallocScribble to detect writing on free blocks and missing initializers:\n"
- " 0x55 is written upon free and 0xaa is written on allocation\n"
- "- MallocCheckHeapStart <n> to start checking the heap after <n> operations\n"
- "- MallocCheckHeapEach <s> to repeat the checking of the heap after <s> operations\n"
- "- MallocCheckHeapSleep <t> to sleep <t> seconds on heap corruption\n"
- "- MallocCheckHeapAbort <b> to abort on heap corruption if <b> is non-zero\n"
- "- MallocCorruptionAbort to abort on malloc errors, but not on out of memory for 32-bit processes\n"
- " MallocCorruptionAbort is always set on 64-bit processes\n"
- "- MallocErrorAbort to abort on any malloc error, including out of memory\n"
- "- MallocHelp - this help!\n");
- }
-}
+ malloc_report(ASL_LEVEL_INFO,
+ "environment variables that can be set for debug:\n"
+ "- MallocLogFile <f> to create/append messages to file <f> instead of stderr\n"
+ "- MallocGuardEdges to add 2 guard pages for each large block\n"
+ "- MallocDoNotProtectPrelude to disable protection (when previous flag set)\n"
+ "- MallocDoNotProtectPostlude to disable protection (when previous flag set)\n"
+ "- MallocStackLogging to record all stacks. Tools like leaks can then be applied\n"
+ "- MallocStackLoggingNoCompact to record all stacks. Needed for malloc_history\n"
+ "- MallocStackLoggingDirectory to set location of stack logs, which can grow large; default is /tmp\n"
+ "- MallocScribble to detect writing on free blocks and missing initializers:\n"
+ " 0x55 is written upon free and 0xaa is written on allocation\n"
+ "- MallocCheckHeapStart <n> to start checking the heap after <n> operations\n"
+ "- MallocCheckHeapEach <s> to repeat the checking of the heap after <s> operations\n"
+ "- MallocCheckHeapSleep <t> to sleep <t> seconds on heap corruption\n"
+ "- MallocCheckHeapAbort <b> to abort on heap corruption if <b> is non-zero\n"
+ "- MallocCorruptionAbort to abort on malloc errors, but not on out of memory for 32-bit processes\n"
+ " MallocCorruptionAbort is always set on 64-bit processes\n"
+ "- MallocErrorAbort to abort on any malloc error, including out of memory\n"\
+ "- MallocTracing to emit kdebug trace points on malloc entry points\n"\
+ "- MallocZeroOnFree to enable or disable zero-on-free behavior (for debugging only)\n"\
+ "- MallocCheckZeroOnFreeCorruption to enable zero-on-free corruption detection\n"\
+ "- MallocHelp - this help!\n");
+ }
+}
+
malloc_zone_t *
malloc_create_zone(vm_size_t start_size, unsigned flags)
{
- malloc_zone_t *zone;
+ malloc_zone_t *zone = NULL;
/* start_size doesn't actually appear to be used, but we test anyway. */
- if (start_size > MALLOC_ABSOLUTE_MAX_SIZE) {
+ if (start_size > malloc_absolute_max_size) {
return NULL;
}
- if (!_malloc_is_initialized) _malloc_initialize();
- zone = create_scalable_zone(start_size, flags | malloc_debug_flags);
+
+
+ if (!zone) {
+ zone = create_scalable_zone(start_size, flags | malloc_debug_flags);
+ }
+
+ if (enable_pgm(flags)) {
+ malloc_zone_t *pgm_zone = pgm_create_zone(zone);
+ MALLOC_LOCK();
+ malloc_zone_register_while_locked(pgm_zone, false);
+ malloc_zone_register_while_locked(zone, false);
+ MALLOC_UNLOCK();
+ return pgm_zone;
+ }
+
malloc_zone_register(zone);
return zone;
}
@@ -710,234 +2071,598 @@
void
malloc_create_legacy_default_zone(void)
{
- malloc_zone_t *zone;
- int i;
-
- if (!_malloc_is_initialized) _malloc_initialize();
+ malloc_zone_t *zone;
+
zone = create_legacy_scalable_zone(0, malloc_debug_flags);
MALLOC_LOCK();
- malloc_zone_register_while_locked(zone);
//
// Establish the legacy scalable zone just created as the default zone.
//
malloc_zone_t *hold = malloc_zones[0];
- if(hold->zone_name && strcmp(hold->zone_name, DEFAULT_MALLOC_ZONE_STRING) == 0) {
+ if (hold->zone_name && strcmp(hold->zone_name, DEFAULT_MALLOC_ZONE_STRING) == 0) {
malloc_set_zone_name(hold, NULL);
}
malloc_set_zone_name(zone, DEFAULT_MALLOC_ZONE_STRING);
- unsigned protect_size = malloc_num_zones_allocated * sizeof(malloc_zone_t *);
- mprotect(malloc_zones, protect_size, PROT_READ | PROT_WRITE);
-
- // assert(zone == malloc_zones[malloc_num_zones - 1];
- for (i = malloc_num_zones - 1; i > 0; --i) {
- malloc_zones[i] = malloc_zones[i - 1];
- }
- malloc_zones[0] = zone;
-
- mprotect(malloc_zones, protect_size, PROT_READ);
+ malloc_zone_register_while_locked(zone, /*make_default=*/true);
MALLOC_UNLOCK();
}
void
-malloc_destroy_zone(malloc_zone_t *zone) {
+malloc_destroy_zone(malloc_zone_t *zone)
+{
malloc_set_zone_name(zone, NULL); // Deallocate zone name wherever it may reside PR_7701095
malloc_zone_unregister(zone);
zone->destroy(zone);
}
+static vm_address_t *frames = NULL;
+static unsigned num_frames;
+
+MALLOC_NOINLINE
+void
+malloc_zone_check_fail(const char *msg, const char *fmt, ...)
+{
+ _SIMPLE_STRING b = _simple_salloc();
+ if (b) {
+ _simple_sprintf(b, "*** MallocCheckHeap: FAILED check at operation #%d\n", malloc_check_counter - 1);
+ } else {
+ malloc_report(MALLOC_REPORT_NOLOG, "*** MallocCheckHeap: FAILED check at operation #%d\n", malloc_check_counter - 1);
+ }
+ if (frames) {
+ unsigned index = 1;
+ if (b) {
+ _simple_sappend(b, "Stack for last operation where the malloc check succeeded: ");
+ while (index < num_frames)
+ _simple_sprintf(b, "%p ", (void*)frames[index++]);
+ malloc_report(MALLOC_REPORT_NOLOG, "%s\n(Use 'atos' for a symbolic stack)\n", _simple_string(b));
+ } else {
+ /*
+ * Should only get here if vm_allocate() can't get a single page of
+ * memory, implying _simple_asl_log() would also fail. So we just
+ * print to the file descriptor.
+ */
+ malloc_report(MALLOC_REPORT_NOLOG, "Stack for last operation where the malloc check succeeded: ");
+ while (index < num_frames) {
+ malloc_report(MALLOC_REPORT_NOLOG, "%p ", (void *)frames[index++]);
+ }
+ malloc_report(MALLOC_REPORT_NOLOG, "\n(Use 'atos' for a symbolic stack)\n");
+ }
+ }
+ if (malloc_check_each > 1) {
+ unsigned recomm_each = (malloc_check_each > 10) ? malloc_check_each / 10 : 1;
+ unsigned recomm_start =
+ (malloc_check_counter > malloc_check_each + 1) ? malloc_check_counter - 1 - malloc_check_each : 1;
+ malloc_report(MALLOC_REPORT_NOLOG,
+ "*** Recommend using 'setenv MallocCheckHeapStart %d; setenv MallocCheckHeapEach %d' to narrow down failure\n",
+ recomm_start, recomm_each);
+ }
+
+ if (b) {
+ _simple_sfree(b);
+ }
+
+ // Use malloc_vreport() to:
+ // * report the error
+ // * call malloc_error_break() for a breakpoint
+ // * sleep or stop for debug
+ // * set the crash message and crash if malloc_check_abort is set.
+ unsigned sleep_time = 0;
+ uint32_t report_flags = ASL_LEVEL_ERR | MALLOC_REPORT_DEBUG | MALLOC_REPORT_NOLOG;
+ if (malloc_check_abort) {
+ report_flags |= MALLOC_REPORT_CRASH;
+ } else {
+ if (malloc_check_sleep > 0) {
+ malloc_report(ASL_LEVEL_NOTICE, "*** Will sleep for %d seconds to leave time to attach\n", malloc_check_sleep);
+ sleep_time = malloc_check_sleep;
+ } else if (malloc_check_sleep < 0) {
+ malloc_report(ASL_LEVEL_NOTICE, "*** Will sleep once for %d seconds to leave time to attach\n", -malloc_check_sleep);
+ sleep_time = -malloc_check_sleep;
+ malloc_check_sleep = 0;
+ }
+ }
+ va_list ap;
+ va_start(ap, fmt);
+ malloc_vreport(report_flags, sleep_time, msg, NULL, fmt, ap);
+ va_end(ap);
+}
+
/********* Block creation and manipulation ************/
+__attribute__((cold, noinline))
static void
-internal_check(void) {
- static vm_address_t *frames = NULL;
- static unsigned num_frames;
+internal_check(void)
+{
+ if (malloc_check_counter++ < malloc_check_start) {
+ return;
+ }
if (malloc_zone_check(NULL)) {
- if (!frames) vm_allocate(mach_task_self(), (void *)&frames, vm_page_size, 1);
- thread_stack_pcs(frames, vm_page_size/sizeof(vm_address_t) - 1, &num_frames);
+ if (!frames) {
+ vm_allocate(mach_task_self(), (void *)&frames, vm_page_size, 1);
+ }
+ thread_stack_pcs(frames, (unsigned)(vm_page_size / sizeof(vm_address_t) - 1), &num_frames);
+ }
+ malloc_check_start += malloc_check_each;
+}
+
+MALLOC_NOINLINE
+static void *
+_malloc_zone_malloc_instrumented_or_legacy(malloc_zone_t *zone, size_t size,
+ malloc_zone_options_t mzo)
+{
+ uint64_t type_id = malloc_get_tsd_type_id();
+#if MALLOC_TARGET_64BIT
+ bool clear_type = false;
+ if (!type_id) {
+ malloc_type_descriptor_t fallback =
+ malloc_callsite_fallback_type_descriptor();
+ malloc_set_tsd_type_descriptor(fallback);
+ type_id = fallback.type_id;
+ clear_type = true;
+ }
+#endif // MALLOC_TARGET_64BIT
+ MALLOC_TRACE(TRACE_malloc | DBG_FUNC_START, (uintptr_t)zone, size, type_id,
+ 0);
+
+ void *ptr = NULL;
+
+ if (malloc_check_start) {
+ internal_check();
+ }
+ if (size > malloc_absolute_max_size) {
+ goto out;
+ }
+
+ ptr = zone->malloc(zone, size);
+
+ if (os_unlikely(malloc_logger)) {
+ malloc_logger(MALLOC_LOG_TYPE_ALLOCATE | MALLOC_LOG_TYPE_HAS_ZONE,
+ (uintptr_t)zone, (uintptr_t)size, 0, (uintptr_t)ptr, 0);
+ }
+
+ if (os_unlikely(malloc_simple_stack_logging)) {
+ malloc_report(MALLOC_SIMPLE_STACK_LOGGING_FLAGS, "malloc (%p/%llu): ",
+ ptr, (unsigned long long)size);
+ }
+
+ MALLOC_TRACE(TRACE_malloc | DBG_FUNC_END, (uintptr_t)zone, size,
+ (uintptr_t)ptr, type_id);
+out:
+#if MALLOC_TARGET_64BIT
+ if (clear_type) {
+ malloc_set_tsd_type_descriptor(MALLOC_TYPE_DESCRIPTOR_NONE);
+ }
+#endif // MALLOC_TARGET_64BIT
+ if (os_unlikely(ptr == NULL)) {
+ malloc_set_errno_fast(mzo, ENOMEM);
+ }
+ return ptr;
+}
+
+void *
+_malloc_zone_malloc(malloc_zone_t *zone, size_t size, malloc_zone_options_t mzo)
+{
+ if (zone == default_zone && !lite_zone) {
+ // Eagerly resolve the virtual default zone to make the zone version
+ // check accurate
+ zone = malloc_zones[0];
+ }
+
+ if (os_unlikely(malloc_slowpath || malloc_logger || zone->version < 13)) {
+ return _malloc_zone_malloc_instrumented_or_legacy(zone, size, mzo);
+ }
+
+ if (os_unlikely(size > malloc_absolute_max_size)) {
+ malloc_set_errno_fast(mzo, ENOMEM);
+ return NULL;
+ }
+
+ if (zone->version >= 16) {
+ return zone->malloc_type_malloc(zone, size,
+ malloc_callsite_fallback_type_id());
+ }
+
+ // zone versions >= 13 set errno on failure so we can tail-call
+ return zone->malloc(zone, size);
+}
+
+MALLOC_NOINLINE
+void *
+malloc_zone_malloc(malloc_zone_t *zone, size_t size)
+{
+ return _malloc_zone_malloc(zone, size, MZ_NONE);
+}
+
+MALLOC_NOINLINE
+static void *
+_malloc_zone_calloc_instrumented_or_legacy(malloc_zone_t *zone,
+ size_t num_items, size_t size, malloc_zone_options_t mzo)
+{
+ uint64_t type_id = malloc_get_tsd_type_id();
+#if MALLOC_TARGET_64BIT
+ bool clear_type = false;
+ if (!type_id) {
+ malloc_type_descriptor_t fallback =
+ malloc_callsite_fallback_type_descriptor();
+ malloc_set_tsd_type_descriptor(fallback);
+ type_id = fallback.type_id;
+ clear_type = true;
+ }
+#endif // MALLOC_TARGET_64BIT
+ MALLOC_TRACE(TRACE_calloc | DBG_FUNC_START, (uintptr_t)zone, num_items,
+ size, type_id);
+
+ void *ptr;
+ if (malloc_check_start) {
+ internal_check();
+ }
+
+ ptr = zone->calloc(zone, num_items, size);
+
+ if (os_unlikely(malloc_logger)) {
+ uint32_t logger_type = MALLOC_LOG_TYPE_ALLOCATE |
+ MALLOC_LOG_TYPE_HAS_ZONE | MALLOC_LOG_TYPE_CLEARED;
+ malloc_logger(logger_type, (uintptr_t)zone,
+ (uintptr_t)(num_items * size), 0, (uintptr_t)ptr, 0);
+ }
+
+ if (os_unlikely(malloc_simple_stack_logging)) {
+ malloc_report(MALLOC_SIMPLE_STACK_LOGGING_FLAGS,
+ "calloc (%p/%llu*%llu): ", ptr, (unsigned long long)num_items,
+ (unsigned long long)size);
+ }
+
+ MALLOC_TRACE(TRACE_calloc | DBG_FUNC_END, (uintptr_t)zone, num_items, size,
+ (uintptr_t)ptr);
+#if MALLOC_TARGET_64BIT
+ if (clear_type) {
+ malloc_set_tsd_type_descriptor(MALLOC_TYPE_DESCRIPTOR_NONE);
+ }
+#endif // MALLOC_TARGET_64BIT
+ if (os_unlikely(ptr == NULL)) {
+ malloc_set_errno_fast(mzo, ENOMEM);
+ }
+ return ptr;
+}
+
+MALLOC_NOINLINE
+void *
+_malloc_zone_calloc(malloc_zone_t *zone, size_t num_items, size_t size,
+ malloc_zone_options_t mzo)
+{
+ if (zone == default_zone && !lite_zone) {
+ // Eagerly resolve the virtual default zone to make the zone version
+ // check accurate
+ zone = malloc_zones[0];
+ }
+
+ if (os_unlikely(malloc_slowpath || malloc_logger || zone->version < 13)) {
+ return _malloc_zone_calloc_instrumented_or_legacy(zone, num_items, size, mzo);
+ }
+
+ if (zone->version >= 16) {
+ return zone->malloc_type_calloc(zone, num_items, size,
+ malloc_callsite_fallback_type_id());
+ }
+
+ // zone versions >= 13 set errno on failure so we can tail-call
+ return zone->calloc(zone, num_items, size);
+}
+
+MALLOC_NOINLINE
+void *
+malloc_zone_calloc(malloc_zone_t *zone, size_t num_items, size_t size)
+{
+ return _malloc_zone_calloc(zone, num_items, size, MZ_NONE);
+}
+
+MALLOC_NOINLINE
+void *
+_malloc_zone_valloc(malloc_zone_t *zone, size_t size, malloc_zone_options_t mzo)
+{
+ uint64_t type_id = malloc_get_tsd_type_id();
+#if MALLOC_TARGET_64BIT
+ bool clear_type = false;
+ if (!type_id) {
+ malloc_type_descriptor_t fallback =
+ malloc_callsite_fallback_type_descriptor();
+ malloc_set_tsd_type_descriptor(fallback);
+ type_id = fallback.type_id;
+ clear_type = true;
+ }
+#endif // MALLOC_TARGET_64BIT
+ MALLOC_TRACE(TRACE_valloc | DBG_FUNC_START, (uintptr_t)zone, size, type_id,
+ 0);
+
+ void *ptr = NULL;
+ if (malloc_check_start) {
+ internal_check();
+ }
+ if (size > malloc_absolute_max_size) {
+ goto out;
+ }
+
+ ptr = zone->valloc(zone, size);
+
+ if (os_unlikely(malloc_logger)) {
+ malloc_logger(MALLOC_LOG_TYPE_ALLOCATE | MALLOC_LOG_TYPE_HAS_ZONE,
+ (uintptr_t)zone, (uintptr_t)size, 0, (uintptr_t)ptr, 0);
+ }
+
+ if (os_unlikely(malloc_simple_stack_logging)) {
+ malloc_report(MALLOC_SIMPLE_STACK_LOGGING_FLAGS, "valloc (%p/%llu): ",
+ ptr, (unsigned long long)size);
+ }
+
+ MALLOC_TRACE(TRACE_valloc | DBG_FUNC_END, (uintptr_t)zone, size,
+ (uintptr_t)ptr, type_id);
+out:
+#if MALLOC_TARGET_64BIT
+ if (clear_type) {
+ malloc_set_tsd_type_descriptor(MALLOC_TYPE_DESCRIPTOR_NONE);
+ }
+#endif // MALLOC_TARGET_64BIT
+
+ if (os_unlikely(ptr == NULL)) {
+ malloc_set_errno_fast(mzo, ENOMEM);
+ }
+ return ptr;
+}
+
+MALLOC_NOINLINE
+void *
+malloc_zone_valloc(malloc_zone_t *zone, size_t size)
+{
+ return _malloc_zone_valloc(zone, size, MZ_NONE);
+}
+
+// We have this function so code within libmalloc can call it without going
+// through the (potentially interposed) dyld symbol stub
+void *
+_malloc_zone_realloc(malloc_zone_t *zone, void *ptr, size_t size,
+ malloc_type_descriptor_t type_desc)
+{
+ uint64_t type_id = malloc_get_tsd_type_id();
+#if MALLOC_TARGET_64BIT
+ bool clear_type = false;
+ if (!type_id) {
+ // A type descriptor in the TSD takes precendence over one passed as a
+ // parameter - the one in the TSD will be a real one set by e.g.
+ // _malloc_type_realloc_outlined(), whereas the parameter will be a
+ // callsite-derived fallback
+ malloc_set_tsd_type_descriptor(type_desc);
+ type_id = type_desc.type_id;
+ clear_type = true;
+ }
+#endif // MALLOC_TARGET_64BIT
+ MALLOC_TRACE(TRACE_realloc | DBG_FUNC_START, (uintptr_t)zone,
+ (uintptr_t)ptr, size, type_id);
+
+ void *new_ptr = NULL;
+ if (malloc_check_start) {
+ internal_check();
+ }
+ if (size > malloc_absolute_max_size) {
+ goto out;
+ }
+
+ new_ptr = zone->realloc(zone, ptr, size);
+
+ if (os_unlikely(malloc_logger)) {
+ uint32_t logger_type = MALLOC_LOG_TYPE_ALLOCATE |
+ MALLOC_LOG_TYPE_DEALLOCATE | MALLOC_LOG_TYPE_HAS_ZONE;
+ malloc_logger(logger_type, (uintptr_t)zone, (uintptr_t)ptr,
+ (uintptr_t)size, (uintptr_t)new_ptr, 0);
+ }
+
+ if (os_unlikely(malloc_simple_stack_logging)) {
+ malloc_report(MALLOC_SIMPLE_STACK_LOGGING_FLAGS,
+ "realloc (%p->%p/%llu): ", ptr, new_ptr,
+ (unsigned long long)size);
+ }
+
+ MALLOC_TRACE(TRACE_realloc | DBG_FUNC_END, (uintptr_t)zone, (uintptr_t)ptr,
+ size, (uintptr_t)new_ptr);
+out:
+#if MALLOC_TARGET_64BIT
+ if (clear_type) {
+ malloc_set_tsd_type_descriptor(MALLOC_TYPE_DESCRIPTOR_NONE);
+ }
+#endif // MALLOC_TARGET_64BIT
+ return new_ptr;
+}
+
+MALLOC_NOINLINE
+void *
+malloc_zone_realloc(malloc_zone_t *zone, void *ptr, size_t size)
+{
+ return _malloc_zone_realloc(zone, ptr, size,
+ malloc_callsite_fallback_type_descriptor());
+}
+
+MALLOC_NOINLINE
+void
+malloc_zone_free(malloc_zone_t *zone, void *ptr)
+{
+ MALLOC_TRACE(TRACE_free, (uintptr_t)zone, (uintptr_t)ptr, (ptr) ? *(uintptr_t*)ptr : 0, 0);
+
+ if (os_unlikely(malloc_logger)) {
+ malloc_logger(MALLOC_LOG_TYPE_DEALLOCATE | MALLOC_LOG_TYPE_HAS_ZONE, (uintptr_t)zone, (uintptr_t)ptr, 0, 0, 0);
+ }
+ if (os_unlikely(malloc_simple_stack_logging)) {
+ malloc_report(MALLOC_SIMPLE_STACK_LOGGING_FLAGS, "malloc_zone_free (%p): ", ptr);
+ }
+ if (malloc_check_start) {
+ internal_check();
+ }
+
+ zone->free(zone, ptr);
+}
+
+static void
+malloc_zone_free_definite_size(malloc_zone_t *zone, void *ptr, size_t size)
+{
+ MALLOC_TRACE(TRACE_free, (uintptr_t)zone, (uintptr_t)ptr, size, (ptr && size) ? *(uintptr_t*)ptr : 0);
+
+ if (os_unlikely(malloc_logger)) {
+ malloc_logger(MALLOC_LOG_TYPE_DEALLOCATE | MALLOC_LOG_TYPE_HAS_ZONE, (uintptr_t)zone, (uintptr_t)ptr, 0, 0, 0);
+ }
+ if (os_unlikely(malloc_simple_stack_logging)) {
+ malloc_report(MALLOC_SIMPLE_STACK_LOGGING_FLAGS, "free (%p/%llu): ",
+ ptr, (unsigned long long)size);
+ }
+ if (malloc_check_start) {
+ internal_check();
+ }
+
+ zone->free_definite_size(zone, ptr, size);
+}
+
+malloc_zone_t *
+malloc_zone_from_ptr(const void *ptr)
+{
+ if (!ptr) {
+ return NULL;
} else {
- _SIMPLE_STRING b = _simple_salloc();
- if (b)
- _simple_sprintf(b, "*** MallocCheckHeap: FAILED check at %dth operation\n", malloc_check_counter-1);
- else
- _malloc_printf(MALLOC_PRINTF_NOLOG, "*** MallocCheckHeap: FAILED check at %dth operation\n", malloc_check_counter-1);
- malloc_printf("*** MallocCheckHeap: FAILED check at %dth operation\n", malloc_check_counter-1);
- if (frames) {
- unsigned index = 1;
- if (b) {
- _simple_sappend(b, "Stack for last operation where the malloc check succeeded: ");
- while (index < num_frames) _simple_sprintf(b, "%p ", frames[index++]);
- malloc_printf("%s\n(Use 'atos' for a symbolic stack)\n", _simple_string(b));
- } else {
- /*
- * Should only get here if vm_allocate() can't get a single page of
- * memory, implying _simple_asl_log() would also fail. So we just
- * print to the file descriptor.
- */
- _malloc_printf(MALLOC_PRINTF_NOLOG, "Stack for last operation where the malloc check succeeded: ");
- while (index < num_frames) _malloc_printf(MALLOC_PRINTF_NOLOG, "%p ", frames[index++]);
- _malloc_printf(MALLOC_PRINTF_NOLOG, "\n(Use 'atos' for a symbolic stack)\n");
- }
- }
- if (malloc_check_each > 1) {
- unsigned recomm_each = (malloc_check_each > 10) ? malloc_check_each/10 : 1;
- unsigned recomm_start = (malloc_check_counter > malloc_check_each+1) ? malloc_check_counter-1-malloc_check_each : 1;
- malloc_printf("*** Recommend using 'setenv MallocCheckHeapStart %d; setenv MallocCheckHeapEach %d' to narrow down failure\n", recomm_start, recomm_each);
- }
- if (malloc_check_abort) {
- CRSetCrashLogMessage(b ? _simple_string(b) : "*** MallocCheckHeap: FAILED check");
- abort();
- } else if (b)
- _simple_sfree(b);
- if (malloc_check_sleep > 0) {
- _malloc_printf(ASL_LEVEL_NOTICE, "*** Sleeping for %d seconds to leave time to attach\n",
- malloc_check_sleep);
- sleep(malloc_check_sleep);
- } else if (malloc_check_sleep < 0) {
- _malloc_printf(ASL_LEVEL_NOTICE, "*** Sleeping once for %d seconds to leave time to attach\n",
- -malloc_check_sleep);
- sleep(-malloc_check_sleep);
- malloc_check_sleep = 0;
- }
- }
- malloc_check_start += malloc_check_each;
-}
-
+ return _find_registered_zone(ptr, NULL, false);
+ }
+}
+
+MALLOC_NOINLINE
void *
-malloc_zone_malloc(malloc_zone_t *zone, size_t size) {
- void *ptr;
- if (malloc_check_start && (malloc_check_counter++ >= malloc_check_start)) {
+_malloc_zone_memalign(malloc_zone_t *zone, size_t alignment, size_t size,
+ malloc_zone_options_t mzo, malloc_type_descriptor_t type_desc)
+{
+ uint64_t type_id = malloc_get_tsd_type_id();
+#if MALLOC_TARGET_64BIT
+ bool clear_type = false;
+ if (!type_id) {
+ // A type descriptor in the TSD takes precendence over one passed as a
+ // parameter - the one in the TSD will be a real one set by e.g.
+ // _malloc_type_aligned_alloc_outlined(), whereas the parameter will
+ // usually be a callsite-derived fallback
+ malloc_set_tsd_type_descriptor(type_desc);
+ type_id = type_desc.type_id;
+ clear_type = true;
+ }
+#endif // MALLOC_TARGET_64BIT
+ MALLOC_TRACE(TRACE_memalign | DBG_FUNC_START, (uintptr_t)zone, alignment,
+ size, type_id);
+
+ void *ptr = NULL;
+ int err = ENOMEM;
+ // Version must be >= 5 to look at the new memalign field.
+ if (zone->version < 5) {
+ goto out;
+ }
+ if (malloc_check_start) {
internal_check();
}
- if (size > MALLOC_ABSOLUTE_MAX_SIZE) {
- return NULL;
- }
- ptr = zone->malloc(zone, size);
- if (malloc_logger)
- malloc_logger(MALLOC_LOG_TYPE_ALLOCATE | MALLOC_LOG_TYPE_HAS_ZONE, (uintptr_t)zone, (uintptr_t)size, 0, (uintptr_t)ptr, 0);
+ if (size > malloc_absolute_max_size) {
+ goto out;
+ }
+ // excludes 0 == alignment
+ // relies on sizeof(void *) being a power of two.
+ if (alignment < sizeof(void *) ||
+ 0 != (alignment & (alignment - 1))) {
+ err = EINVAL;
+ goto out;
+ }
+ // C11 aligned_alloc requires size to be a multiple of alignment, but
+ // posix_memalign does not
+ if ((mzo & MZ_C11) && (size & (alignment - 1)) != 0) {
+ err = EINVAL;
+ goto out;
+ }
+
+ if (!(zone->memalign)) {
+ goto out;
+ }
+ ptr = zone->memalign(zone, alignment, size);
+
+ if (os_unlikely(malloc_logger)) {
+ malloc_logger(MALLOC_LOG_TYPE_ALLOCATE | MALLOC_LOG_TYPE_HAS_ZONE,
+ (uintptr_t)zone, (uintptr_t)size, 0, (uintptr_t)ptr, 0);
+ }
+ if (os_unlikely(malloc_simple_stack_logging)) {
+ malloc_report(MALLOC_SIMPLE_STACK_LOGGING_FLAGS,
+ "memalign (%p/%llu,%llu): ", ptr, (unsigned long long)alignment,
+ (unsigned long long)size);
+ }
+
+ MALLOC_TRACE(TRACE_memalign | DBG_FUNC_END, (uintptr_t)zone, alignment,
+ size, (uintptr_t)ptr);
+
+out:
+#if MALLOC_TARGET_64BIT
+ if (clear_type) {
+ malloc_set_tsd_type_descriptor(MALLOC_TYPE_DESCRIPTOR_NONE);
+ }
+#endif // MALLOC_TARGET_64BIT
+
+ if (os_unlikely(ptr == NULL)) {
+ if (mzo & MZ_POSIX) {
+ malloc_set_errno_fast(mzo, err);
+ }
+ }
return ptr;
}
+MALLOC_NOINLINE
void *
-malloc_zone_calloc(malloc_zone_t *zone, size_t num_items, size_t size) {
- void *ptr;
- if (malloc_check_start && (malloc_check_counter++ >= malloc_check_start)) {
+malloc_zone_memalign(malloc_zone_t *zone, size_t alignment, size_t size)
+{
+ return _malloc_zone_memalign(zone, alignment, size, MZ_NONE,
+ malloc_callsite_fallback_type_descriptor());
+}
+
+boolean_t
+malloc_zone_claimed_address(malloc_zone_t *zone, void *ptr)
+{
+ if (!ptr) {
+ // NULL is not a member of any zone.
+ return false;
+ }
+
+ if (malloc_check_start) {
internal_check();
}
- if (size > MALLOC_ABSOLUTE_MAX_SIZE) {
- return NULL;
- }
- ptr = zone->calloc(zone, num_items, size);
- if (malloc_logger)
- malloc_logger(MALLOC_LOG_TYPE_ALLOCATE | MALLOC_LOG_TYPE_HAS_ZONE | MALLOC_LOG_TYPE_CLEARED, (uintptr_t)zone, (uintptr_t)(num_items * size), 0,
- (uintptr_t)ptr, 0);
- return ptr;
-}
-
-void *
-malloc_zone_valloc(malloc_zone_t *zone, size_t size) {
- void *ptr;
- if (malloc_check_start && (malloc_check_counter++ >= malloc_check_start)) {
- internal_check();
- }
- if (size > MALLOC_ABSOLUTE_MAX_SIZE) {
- return NULL;
- }
- ptr = zone->valloc(zone, size);
- if (malloc_logger)
- malloc_logger(MALLOC_LOG_TYPE_ALLOCATE | MALLOC_LOG_TYPE_HAS_ZONE, (uintptr_t)zone, (uintptr_t)size, 0, (uintptr_t)ptr, 0);
- return ptr;
-}
-
-void *
-malloc_zone_realloc(malloc_zone_t *zone, void *ptr, size_t size) {
- void *new_ptr;
- if (malloc_check_start && (malloc_check_counter++ >= malloc_check_start)) {
- internal_check();
- }
- if (size > MALLOC_ABSOLUTE_MAX_SIZE) {
- return NULL;
- }
- new_ptr = zone->realloc(zone, ptr, size);
- if (malloc_logger)
- malloc_logger(MALLOC_LOG_TYPE_ALLOCATE | MALLOC_LOG_TYPE_DEALLOCATE | MALLOC_LOG_TYPE_HAS_ZONE, (uintptr_t)zone, (uintptr_t)ptr, (uintptr_t)size,
- (uintptr_t)new_ptr, 0);
- return new_ptr;
-}
+
+ if (zone->version < 10 || !zone->claimed_address) {
+ // For zones that have not implemented claimed_address, we always have
+ // to return true to avoid a false negative.
+ return true;
+ }
+
+ return zone->claimed_address(zone, ptr);
+}
+
+/********* Functions for zone implementors ************/
void
-malloc_zone_free(malloc_zone_t *zone, void *ptr) {
- if (malloc_logger)
- malloc_logger(MALLOC_LOG_TYPE_DEALLOCATE | MALLOC_LOG_TYPE_HAS_ZONE, (uintptr_t)zone, (uintptr_t)ptr, 0, 0, 0);
- if (malloc_check_start && (malloc_check_counter++ >= malloc_check_start)) {
- internal_check();
- }
- zone->free(zone, ptr);
-}
-
-static void
-malloc_zone_free_definite_size(malloc_zone_t *zone, void *ptr, size_t size) {
- if (malloc_logger)
- malloc_logger(MALLOC_LOG_TYPE_DEALLOCATE | MALLOC_LOG_TYPE_HAS_ZONE, (uintptr_t)zone, (uintptr_t)ptr, 0, 0, 0);
- if (malloc_check_start && (malloc_check_counter++ >= malloc_check_start)) {
- internal_check();
- }
- zone->free_definite_size(zone, ptr, size);
-}
-
-malloc_zone_t *
-malloc_zone_from_ptr(const void *ptr) {
- if (!ptr)
- return NULL;
- else
- return find_registered_zone(ptr, NULL);
-}
-
-void *
-malloc_zone_memalign(malloc_zone_t *zone, size_t alignment, size_t size) {
- void *ptr;
- if (zone->version < 5) // Version must be >= 5 to look at the new memalign field.
- return NULL;
- if (!(zone->memalign))
- return NULL;
- if (malloc_check_start && (malloc_check_counter++ >= malloc_check_start)) {
- internal_check();
- }
- if (size > MALLOC_ABSOLUTE_MAX_SIZE) {
- return NULL;
- }
- if (alignment < sizeof( void *) || // excludes 0 == alignment
- 0 != (alignment & (alignment - 1))) { // relies on sizeof(void *) being a power of two.
- return NULL;
- }
- ptr = zone->memalign(zone, alignment, size);
- if (malloc_logger)
- malloc_logger(MALLOC_LOG_TYPE_ALLOCATE | MALLOC_LOG_TYPE_HAS_ZONE, (uintptr_t)zone, (uintptr_t)size, 0, (uintptr_t)ptr, 0);
- return ptr;
-}
-
-/********* Functions for zone implementors ************/
+malloc_zone_register(malloc_zone_t *zone)
+{
+ MALLOC_LOCK();
+ malloc_zone_register_while_locked(zone, false);
+ MALLOC_UNLOCK();
+}
void
-malloc_zone_register(malloc_zone_t *zone) {
- MALLOC_LOCK();
- malloc_zone_register_while_locked(zone);
- MALLOC_UNLOCK();
-}
-
-void
-malloc_zone_unregister(malloc_zone_t *z) {
- unsigned index;
-
- if (malloc_num_zones == 0)
+malloc_zone_unregister(malloc_zone_t *z)
+{
+ unsigned index;
+
+ if (malloc_num_zones == 0) {
return;
+ }
MALLOC_LOCK();
for (index = 0; index < malloc_num_zones; ++index) {
- if (z != malloc_zones[index])
+ if (z != malloc_zones[index]) {
continue;
+ }
// Modify the page to be allow write access, so that we can update the
// malloc_zones array.
@@ -952,155 +2677,177 @@
mprotect(malloc_zones, protect_size, PROT_READ);
+ // MAX(num_zones, 1) retains the fast path in find_registered_zone() for zone 0 even
+ // if it is a custom zone, e.g., ASan and user zones.
+ initial_num_zones = MAX(MIN(malloc_num_zones, initial_num_zones), 1);
+
// Exchange the roles of the FRZ counters. The counter that has captured the number of threads presently
- // executing *inside* find_regiatered_zone is swapped with the counter drained to zero last time through.
+ // executing *inside* find_registered_zone is swapped with the counter drained to zero last time through.
// The former is then allowed to drain to zero while this thread yields.
- int *p = pFRZCounterLive;
+ int32_t volatile *p = pFRZCounterLive;
pFRZCounterLive = pFRZCounterDrain;
pFRZCounterDrain = p;
- __sync_synchronize(); // Full memory barrier
-
- while (0 != *pFRZCounterDrain) { yield(); }
+ OSMemoryBarrier(); // Full memory barrier
+
+ while (0 != *pFRZCounterDrain) {
+ yield();
+ }
MALLOC_UNLOCK();
return;
}
MALLOC_UNLOCK();
- malloc_printf("*** malloc_zone_unregister() failed for %p\n", z);
+ malloc_report(ASL_LEVEL_ERR, "*** malloc_zone_unregister() failed for %p\n", z);
}
void
-malloc_set_zone_name(malloc_zone_t *z, const char *name) {
- char *newName;
-
- mprotect(z, sizeof(malloc_zone_t), PROT_READ | PROT_WRITE);
+malloc_set_zone_name(malloc_zone_t *z, const char *name)
+{
+ // TODO: save and restore permissions generally
+ bool mprotect_zone = true;
+ if (mprotect_zone) {
+ mprotect(z, sizeof(malloc_zone_t), PROT_READ | PROT_WRITE);
+ }
if (z->zone_name) {
- free((char *)z->zone_name);
+ malloc_zone_t *old_zone = _find_registered_zone(z->zone_name, NULL, false);
+ if (old_zone) {
+ malloc_zone_free(old_zone, (char *)z->zone_name);
+ }
z->zone_name = NULL;
}
if (name) {
size_t buflen = strlen(name) + 1;
- newName = malloc_zone_malloc(z, buflen);
- if (newName) {
- strlcpy(newName, name, buflen);
- z->zone_name = (const char *)newName;
+ if (_dyld_is_memory_immutable(name, buflen)) {
+ z->zone_name = name;
} else {
- z->zone_name = NULL;
- }
- }
- mprotect(z, sizeof(malloc_zone_t), PROT_READ);
+ char *name_copy = _malloc_zone_malloc(z, buflen, MZ_NONE);
+ if (name_copy) {
+ strcpy(name_copy, name);
+ z->zone_name = name_copy;
+ }
+ }
+
+ malloc_zone_t *wrapped_zone = get_wrapped_zone(z);
+ if (wrapped_zone) {
+ // <name>-<wrapper_label>-<suffix>\0 // Wrapped zone name format
+ // ^ ^ ^ // 2 dashes and \0 -> +3
+ const char *wrapper_label = get_wrapper_zone_label(z);
+ const char *suffix = "Wrapped";
+ size_t buflen = strlen(name) + strlen(wrapper_label) + strlen(suffix) + 3;
+ char *wz_name = _malloc_zone_malloc(wrapped_zone, buflen, MZ_NONE);
+ if (wz_name) {
+ // snprintf() may allocate (not safe to use from libmalloc) and
+ // _simple_sprintf()/_simple_salloc() call vm_allocate() which is
+ // undesirable for such a simple API as malloc_set_zone_name()
+ strcpy(wz_name, name);
+ strcat(wz_name, "-");
+ strcat(wz_name, wrapper_label);
+ strcat(wz_name, "-");
+ strcat(wz_name, suffix);
+ malloc_set_zone_name(wrapped_zone, wz_name);
+ malloc_zone_free(wrapped_zone, wz_name);
+ }
+ }
+ }
+ if (mprotect_zone) {
+ mprotect(z, sizeof(malloc_zone_t), PROT_READ);
+ }
}
const char *
-malloc_get_zone_name(malloc_zone_t *zone) {
+malloc_get_zone_name(malloc_zone_t *zone)
+{
return zone->zone_name;
}
-/*
- * XXX malloc_printf now uses _simple_*printf. It only deals with a
- * subset of printf format specifiers, but it doesn't call malloc.
- */
-
-__attribute__((visibility("hidden")))
void
-_malloc_vprintf(int flags, const char *format, va_list ap)
-{
- _SIMPLE_STRING b;
-
- if (_malloc_no_asl_log || (flags & MALLOC_PRINTF_NOLOG) || (b = _simple_salloc()) == NULL) {
- if (!(flags & MALLOC_PRINTF_NOPREFIX)) {
- void *self = _os_tsd_get_direct(__TSD_THREAD_SELF);
- _simple_dprintf(malloc_debug_file, "%s(%d,%p) malloc: ", getprogname(), getpid(), self);
- }
- _simple_vdprintf(malloc_debug_file, format, ap);
+find_zone_and_free(void *ptr, bool known_non_default)
+{
+ malloc_zone_t *zone;
+ size_t size;
+ if (!ptr) {
return;
}
- if (!(flags & MALLOC_PRINTF_NOPREFIX)) {
- void *self = _os_tsd_get_direct(__TSD_THREAD_SELF);
- _simple_sprintf(b, "%s(%d,%p) malloc: ", getprogname(), getpid(), self);
- }
- _simple_vsprintf(b, format, ap);
- _simple_put(b, malloc_debug_file);
- _simple_asl_log(flags & MALLOC_PRINTF_LEVEL_MASK, Malloc_Facility, _simple_string(b));
- _simple_sfree(b);
-}
-
-__attribute__((visibility("hidden")))
+
+ zone = _find_registered_zone(ptr, &size, known_non_default);
+ if (!zone) {
+ int flags = MALLOC_REPORT_DEBUG | MALLOC_REPORT_NOLOG;
+ if ((malloc_debug_flags & (MALLOC_ABORT_ON_CORRUPTION | MALLOC_ABORT_ON_ERROR))) {
+ flags = MALLOC_REPORT_CRASH | MALLOC_REPORT_NOLOG;
+ }
+ malloc_report(flags,
+ "*** error for object %p: pointer being freed was not allocated\n", ptr);
+ } else if (zone->version >= 6 && zone->free_definite_size) {
+ malloc_zone_free_definite_size(zone, ptr, size);
+ } else {
+ malloc_zone_free(zone, ptr);
+ }
+}
+
+
+/********* Generic ANSI callouts ************/
+
+MALLOC_NOINLINE
+void *
+malloc(size_t size)
+{
+ return _malloc_zone_malloc(default_zone, size, MZ_POSIX);
+}
+
+MALLOC_NOINLINE
+void *
+aligned_alloc(size_t alignment, size_t size)
+{
+ return _malloc_zone_memalign(default_zone, alignment, size,
+ MZ_POSIX | MZ_C11, malloc_callsite_fallback_type_descriptor());
+}
+
+MALLOC_NOINLINE
+void *
+calloc(size_t num_items, size_t size)
+{
+ return _malloc_zone_calloc(default_zone, num_items, size, MZ_POSIX);
+}
+
+// We have this function so code within libmalloc can call it without going
+// through the (potentially interposed) dyld symbol stub
void
-_malloc_printf(int flags, const char *format, ...)
-{
- va_list ap;
-
- va_start(ap, format);
- _malloc_vprintf(flags, format, ap);
- va_end(ap);
-}
-
+_free(void *ptr)
+{
+ if (!ptr) {
+ return;
+ }
+
+ malloc_zone_t *zone0 = malloc_zones[0];
+ if (os_unlikely(malloc_slowpath ||
+ malloc_logger ||
+ zone0->version < 13)) {
+ find_zone_and_free(ptr, false);
+ return;
+ }
+
+ if (zone0->try_free_default) {
+ zone0->try_free_default(zone0, ptr);
+ } else {
+ find_zone_and_free(ptr, false);
+ }
+}
+
+MALLOC_NOINLINE
void
-malloc_printf(const char *format, ...)
-{
- va_list ap;
-
- va_start(ap, format);
- _malloc_vprintf(ASL_LEVEL_ERR, format, ap);
- va_end(ap);
-}
-
-/********* Generic ANSI callouts ************/
-
+free(void *ptr)
+{
+ return _free(ptr);
+}
+
+// We have this function so code within libmalloc can call it without going
+// through the (potentially interposed) dyld symbol stub
void *
-malloc(size_t size) {
- void *retval;
- retval = malloc_zone_malloc(inline_malloc_default_zone(), size);
- if (retval == NULL) {
- errno = ENOMEM;
- }
- return retval;
-}
-
-void *
-calloc(size_t num_items, size_t size) {
- void *retval;
- retval = malloc_zone_calloc(inline_malloc_default_zone(), num_items, size);
- if (retval == NULL) {
- errno = ENOMEM;
- }
- return retval;
-}
-
-void
-free(void *ptr) {
- malloc_zone_t *zone;
- size_t size;
- if (!ptr)
- return;
- zone = find_registered_zone(ptr, &size);
- if (!zone) {
- malloc_printf("*** error for object %p: pointer being freed was not allocated\n"
- "*** set a breakpoint in malloc_error_break to debug\n", ptr);
- malloc_error_break();
- if ((malloc_debug_flags & (SCALABLE_MALLOC_ABORT_ON_CORRUPTION|SCALABLE_MALLOC_ABORT_ON_ERROR))) {
- _SIMPLE_STRING b = _simple_salloc();
- if (b) {
- _simple_sprintf(b, "*** error for object %p: pointer being freed was not allocated\n", ptr);
- CRSetCrashLogMessage(_simple_string(b));
- } else {
- CRSetCrashLogMessage("*** error: pointer being freed was not allocated\n");
- }
- abort();
- }
- } else if (zone->version >= 6 && zone->free_definite_size)
- malloc_zone_free_definite_size(zone, ptr, size);
- else
- malloc_zone_free(zone, ptr);
-}
-
-void *
-realloc(void *in_ptr, size_t new_size) {
- void *retval = NULL;
- void *old_ptr;
- malloc_zone_t *zone;
+_realloc(void *in_ptr, size_t new_size)
+{
+ void *retval = NULL;
+ malloc_zone_t *zone;
// SUSv3: "If size is 0 and ptr is not a null pointer, the object
// pointed to is freed. If the space cannot be allocated, the object
@@ -1110,67 +2857,102 @@
// malloc_zone_malloc with zero size, which matches "If ptr is a null
// pointer, realloc() shall be equivalent to malloc() for the specified
// size." So we only free the original memory if the allocation succeeds.
- old_ptr = (new_size == 0) ? NULL : in_ptr;
- if (!old_ptr) {
- retval = malloc_zone_malloc(inline_malloc_default_zone(), new_size);
+ //
+ // When in_ptr is NULL, we want to ensure that the fallback type descriptor
+ // is good. We can ensure that by tail-calling from here, so that the
+ // callsite information is accurate.
+ //
+ // We don't really care about the quality of the type descriptor
+ // for the new_size == 0 allocation, so it's fine for it to always be based
+ // on the callsite in this function.
+ //
+ // Note: the fact that we allocate from the default zone in the
+ // new_size == 0 case regardless of the zone in_ptr belongs to is arguably a
+ // bug.
+ if (!in_ptr) {
+ return _malloc_zone_malloc(default_zone, new_size, MZ_POSIX);
+ } else if (new_size == 0) {
+ retval = _malloc_zone_malloc(default_zone, new_size, MZ_NONE);
} else {
- zone = find_registered_zone(old_ptr, NULL);
+ zone = _find_registered_zone(in_ptr, NULL, false);
if (!zone) {
- malloc_printf("*** error for object %p: pointer being realloc'd was not allocated\n"
- "*** set a breakpoint in malloc_error_break to debug\n", old_ptr);
- malloc_error_break();
- if ((malloc_debug_flags & (SCALABLE_MALLOC_ABORT_ON_CORRUPTION|SCALABLE_MALLOC_ABORT_ON_ERROR))) {
- _SIMPLE_STRING b = _simple_salloc();
- if (b) {
- _simple_sprintf(b, "*** error for object %p: pointer being realloc'd was not allocated\n", old_ptr);
- CRSetCrashLogMessage(_simple_string(b));
- } else {
- CRSetCrashLogMessage("*** error: pointer being realloc'd was not allocated\n");
- }
- abort();
+ int flags = MALLOC_REPORT_DEBUG | MALLOC_REPORT_NOLOG;
+ const int abort_flags =
+ (MALLOC_ABORT_ON_CORRUPTION | MALLOC_ABORT_ON_ERROR);
+ if (malloc_debug_flags & abort_flags) {
+ flags = MALLOC_REPORT_CRASH | MALLOC_REPORT_NOLOG;
}
+ malloc_report(flags, "*** error for object %p: "
+ "pointer being realloc'd was not allocated\n", in_ptr);
} else {
- retval = malloc_zone_realloc(zone, old_ptr, new_size);
- }
- }
+ retval = _malloc_zone_realloc(zone, in_ptr, new_size,
+ malloc_callsite_fallback_type_descriptor());
+ }
+ }
+
if (retval == NULL) {
- errno = ENOMEM;
+ malloc_set_errno_fast(MZ_POSIX, ENOMEM);
} else if (new_size == 0) {
+ _free(in_ptr);
+ }
+ return retval;
+}
+
+MALLOC_NOINLINE
+void *
+realloc(void *in_ptr, size_t new_size)
+{
+ return _realloc(in_ptr, new_size);
+}
+
+MALLOC_NOINLINE
+void *
+reallocf(void *in_ptr, size_t new_size)
+{
+ void *ptr = realloc(in_ptr, new_size);
+
+ if (!ptr && in_ptr && new_size != 0) {
+ // Save and restore `errno`, because `realloc` will set it to ENOMEM
+ // on allocation failure, but it could be overwritten if `free` calls
+ // into a library function that also modifies `errno`
+ errno_t error = errno;
free(in_ptr);
- }
- return retval;
-}
-
+ errno = error;
+ }
+
+ return ptr;
+}
+
+MALLOC_NOINLINE
void *
-valloc(size_t size) {
- void *retval;
- malloc_zone_t *zone = inline_malloc_default_zone();
- retval = malloc_zone_valloc(zone, size);
- if (retval == NULL) {
- errno = ENOMEM;
- }
- return retval;
+valloc(size_t size)
+{
+ return _malloc_zone_valloc(default_zone, size, MZ_POSIX);
}
extern void
-vfree(void *ptr) {
- free(ptr);
+vfree(void *ptr)
+{
+ _free(ptr);
}
size_t
-malloc_size(const void *ptr) {
- size_t size = 0;
-
- if (!ptr)
+malloc_size(const void *ptr)
+{
+ size_t size = 0;
+
+ if (!ptr) {
return size;
-
- (void)find_registered_zone(ptr, &size);
+ }
+
+ (void)_find_registered_zone(ptr, &size, false);
return size;
}
size_t
-malloc_good_size (size_t size) {
- malloc_zone_t *zone = inline_malloc_default_zone();
+malloc_good_size(size_t size)
+{
+ malloc_zone_t *zone = default_zone;
return zone->introspect->good_size(zone, size);
}
@@ -1190,21 +2972,24 @@
* There is insufficient memory available with the requested alignment.
*/
+// We have this function so code within libmalloc can call it without going
+// through the (potentially interposed) dyld symbol stub
int
-posix_memalign(void **memptr, size_t alignment, size_t size)
-{
- void *retval;
+_posix_memalign(void **memptr, size_t alignment, size_t size)
+{
+ void *retval;
/* POSIX is silent on NULL == memptr !?! */
- retval = malloc_zone_memalign(inline_malloc_default_zone(), alignment, size);
+ retval = _malloc_zone_memalign(default_zone, alignment, size, MZ_NONE,
+ malloc_callsite_fallback_type_descriptor());
if (retval == NULL) {
// To avoid testing the alignment constraints redundantly, we'll rely on the
// test made in malloc_zone_memalign to vet each request. Only if that test fails
// and returns NULL, do we arrive here to detect the bogus alignment and give the
// required EINVAL return.
- if (alignment < sizeof( void *) || // excludes 0 == alignment
- 0 != (alignment & (alignment - 1))) { // relies on sizeof(void *) being a power of two.
+ if (alignment < sizeof(void *) || // excludes 0 == alignment
+ 0 != (alignment & (alignment - 1))) { // relies on sizeof(void *) being a power of two.
return EINVAL;
}
return ENOMEM;
@@ -1214,10 +2999,211 @@
}
}
+MALLOC_NOINLINE
+int
+posix_memalign(void **memptr, size_t alignment, size_t size)
+{
+ return _posix_memalign(memptr, alignment, size);
+}
+
+boolean_t
+malloc_claimed_address(void *ptr)
+{
+ // We need to check with each registered zone whether it claims "ptr".
+ // Use logic similar to that in find_registered_zone().
+ if (malloc_num_zones == 0) {
+ return false;
+ }
+
+ // Start with the lite zone, if it's in use.
+ if (lite_zone && malloc_zone_claimed_address(lite_zone, ptr)) {
+ return true;
+ }
+
+ // Next, try the initial zones.
+ for (uint32_t i = 0; i < initial_num_zones; i++) {
+ if (malloc_zone_claimed_address(malloc_zones[i], ptr)) {
+ return true;
+ }
+ }
+
+ // Try all the other zones. Increment the FRZ barrier so that we can
+ // walk the zones array without a lock (see find_registered_zone() for
+ // the details).
+ int32_t volatile *pFRZCounter = pFRZCounterLive;
+ OSAtomicIncrement32Barrier(pFRZCounter);
+
+ int32_t limit = *(int32_t volatile *)&malloc_num_zones;
+ boolean_t result = false;
+ for (uint32_t i = initial_num_zones; i < limit; i++) {
+ malloc_zone_t *zone = malloc_zones[i];
+ if (malloc_zone_claimed_address(zone, ptr)) {
+ result = true;
+ break;
+ }
+ }
+
+ OSAtomicDecrement32Barrier(pFRZCounter);
+ return result;
+}
+
+void *
+reallocarray(void * in_ptr, size_t nmemb, size_t size){
+ size_t alloc_size;
+ if (os_mul_overflow(nmemb, size, &alloc_size)){
+ malloc_set_errno_fast(MZ_POSIX, ENOMEM);
+ return NULL;
+ }
+ return realloc(in_ptr, alloc_size);
+}
+
+void *
+reallocarrayf(void * in_ptr, size_t nmemb, size_t size){
+ size_t alloc_size;
+ if (os_mul_overflow(nmemb, size, &alloc_size)){
+ malloc_set_errno_fast(MZ_POSIX, ENOMEM);
+ return NULL;
+ }
+ return reallocf(in_ptr, alloc_size);
+}
+
+void *
+_malloc_zone_malloc_with_options_np_outlined(malloc_zone_t *zone, size_t align,
+ size_t size, malloc_options_np_t options)
+{
+ void *ptr = NULL;
+
+ if (zone == NULL || zone == default_zone) {
+ zone = runtime_default_zone();
+ }
+
+#if MALLOC_TARGET_64BIT
+ uint64_t type_id = malloc_get_tsd_type_id();
+ bool clear_type = false;
+ if (!type_id) {
+ malloc_type_descriptor_t fallback =
+ malloc_callsite_fallback_type_descriptor();
+ malloc_set_tsd_type_descriptor(fallback);
+ type_id = fallback.type_id;
+ clear_type = true;
+ }
+#endif // MALLOC_TARGET_64BIT
+
+ if (malloc_interposition_compat || (zone->version < 15) ||
+ !zone->malloc_with_options) {
+ // There's no reasonable way to have the fallback callsite type
+ // descriptor work here. That's okay, as it's uncommon and SPI, so its
+ // callers should be built with TMO.
+ if (align) {
+ ptr = malloc_zone_memalign(zone, align, size);
+ if (ptr && (options & MALLOC_NP_OPTION_CLEAR)) {
+ memset(ptr, 0, size);
+ }
+ } else if (options & MALLOC_NP_OPTION_CLEAR) {
+ ptr = malloc_zone_calloc(zone, 1, size);
+ } else {
+ ptr = malloc_zone_malloc(zone, size);
+ }
+ } else {
+ MALLOC_TRACE(TRACE_malloc_options | DBG_FUNC_START, (uintptr_t)zone,
+ align, size, 0);
+ ptr = zone->malloc_with_options(zone, align, size, options);
+
+ if (os_unlikely(malloc_logger)) {
+ uint32_t flags = MALLOC_LOG_TYPE_ALLOCATE | MALLOC_LOG_TYPE_HAS_ZONE;
+ if (options & MALLOC_NP_OPTION_CLEAR) {
+ flags |= MALLOC_LOG_TYPE_CLEARED;
+ }
+ malloc_logger(flags, (uintptr_t)zone, (uintptr_t)size,
+ 0, (uintptr_t)ptr, 0);
+ }
+ if (os_unlikely(malloc_simple_stack_logging)) {
+ malloc_report(MALLOC_SIMPLE_STACK_LOGGING_FLAGS,
+ "malloc_with_options (%p/%llu,%llu): ", ptr,
+ (unsigned long long)align, (unsigned long long)size);
+ }
+ MALLOC_TRACE(TRACE_malloc_options | DBG_FUNC_END,
+ (uintptr_t)zone, align, size, (uintptr_t)ptr);
+ }
+
+#if MALLOC_TARGET_64BIT
+ if (clear_type) {
+ malloc_set_tsd_type_descriptor(MALLOC_TYPE_DESCRIPTOR_NONE);
+ }
+#endif
+
+ return ptr;
+}
+
+void *
+malloc_zone_malloc_with_options_np(malloc_zone_t *zone, size_t align,
+ size_t size, malloc_options_np_t options)
+{
+ if (os_unlikely((align != 0) && (!powerof2(align) ||
+ ((size & (align-1)) != 0)))) { // equivalent to (size % align != 0)
+ return NULL;
+ }
+
+ if (os_unlikely(malloc_logger || malloc_slowpath)) {
+ return _malloc_zone_malloc_with_options_np_outlined(zone, align, size,
+ options);
+ }
+
+ if (zone == NULL || zone == default_zone) {
+ zone = malloc_zones[0];
+ }
+
+ if (zone->version >= 16 && zone->malloc_type_malloc_with_options) {
+ return zone->malloc_type_malloc_with_options(zone, align, size,
+ options, malloc_callsite_fallback_type_id());
+ }
+
+ return _malloc_zone_malloc_with_options_np_outlined(zone, align, size,
+ options);
+}
+
+/********* Purgeable zone ************/
+
+static void
+_malloc_create_purgeable_zone(void * __unused ctx)
+{
+
+ //
+ // PR_7288598: Must pass a *scalable* zone (szone) as the helper for create_purgeable_zone().
+ // Take care that the zone so obtained is not subject to interposing.
+ //
+ if (!initial_scalable_zone) {
+ // If the process didn't initially create a scalable zone because xzone
+ // malloc is enabled, we should create one now.
+ //
+ // TODO: xzone-backed purgeable zone
+ initial_scalable_zone = create_scalable_zone(0, malloc_debug_flags);
+ malloc_set_zone_name(initial_scalable_zone,
+ DEFAULT_SCALABLE_ZONE_STRING);
+ malloc_zone_register(initial_scalable_zone);
+ }
+
+ default_purgeable_zone = create_purgeable_zone(0, initial_scalable_zone,
+ malloc_debug_flags);
+ malloc_zone_register(default_purgeable_zone);
+ malloc_set_zone_name(default_purgeable_zone,
+ DEFAULT_PUREGEABLE_ZONE_STRING);
+}
+
+malloc_zone_t *
+malloc_default_purgeable_zone(void)
+{
+ static os_once_t pred;
+ os_once(&pred, NULL, _malloc_create_purgeable_zone);
+ return default_purgeable_zone;
+}
+
static malloc_zone_t *
-find_registered_purgeable_zone(void *ptr) {
- if (!ptr)
+find_registered_purgeable_zone(void *ptr)
+{
+ if (!ptr) {
return NULL;
+ }
/*
* Look for a zone which contains ptr. If that zone does not have the purgeable malloc flag
@@ -1226,25 +3212,29 @@
* and only search those.
*/
size_t size = 0;
- malloc_zone_t *zone = find_registered_zone(ptr, &size);
+ malloc_zone_t *zone = _find_registered_zone(ptr, &size, false);
/* FIXME: would really like a zone->introspect->flags->purgeable check, but haven't determined
* binary compatibility impact of changing the introspect struct yet. */
- if (!zone)
+ if (!zone) {
return NULL;
+ }
/* Check to make sure pointer is page aligned and size is multiple of page size */
- if ((size < vm_page_size) || ((size % vm_page_size) != 0))
+ if ((size < vm_page_size) || ((size % vm_page_size) != 0)) {
return NULL;
+ }
return zone;
}
void
-malloc_make_purgeable(void *ptr) {
+malloc_make_purgeable(void *ptr)
+{
malloc_zone_t *zone = find_registered_purgeable_zone(ptr);
- if (!zone)
+ if (!zone) {
return;
+ }
int state = VM_PURGABLE_VOLATILE;
vm_purgable_control(mach_task_self(), (vm_address_t)ptr, VM_PURGABLE_SET_STATE, &state);
@@ -1254,21 +3244,59 @@
/* Returns true if ptr is valid. Ignore the return value from vm_purgeable_control and only report
* state. */
int
-malloc_make_nonpurgeable(void *ptr) {
+malloc_make_nonpurgeable(void *ptr)
+{
malloc_zone_t *zone = find_registered_purgeable_zone(ptr);
- if (!zone)
+ if (!zone) {
return 0;
+ }
int state = VM_PURGABLE_NONVOLATILE;
vm_purgable_control(mach_task_self(), (vm_address_t)ptr, VM_PURGABLE_SET_STATE, &state);
- if (state == VM_PURGABLE_EMPTY)
+ if (state == VM_PURGABLE_EMPTY) {
return EFAULT;
+ }
return 0;
}
-size_t malloc_zone_pressure_relief(malloc_zone_t *zone, size_t goal)
+/********* Memory events ************/
+
+void
+malloc_enter_process_memory_limit_warn_mode(void)
+{
+ // <rdar://problem/25063714>
+}
+
+// Note that malloc_memory_event_handler is not thread-safe, and we are relying on the callers of this for synchronization
+void
+malloc_memory_event_handler(unsigned long event)
+{
+#if CONFIG_MADVISE_PRESSURE_RELIEF || (CONFIG_LARGE_CACHE && !CONFIG_DEFERRED_RECLAIM)
+ if (event & MALLOC_MEMORYSTATUS_MASK_PRESSURE_RELIEF) {
+ malloc_zone_pressure_relief(0, 0);
+ }
+#endif /* CONFIG_MADVISE_PRESSURE_RELIEF || (CONFIG_LARGE_CACHE && !CONFIG_DEFERRED_RECLAIM) */
+
+ if ((event & NOTE_MEMORYSTATUS_MSL_STATUS) != 0 && (event & ~NOTE_MEMORYSTATUS_MSL_STATUS) == 0) {
+ malloc_register_stack_logger();
+ }
+
+#if ENABLE_MEMORY_RESOURCE_EXCEPTION_HANDLING
+ if (event & MALLOC_MEMORYSTATUS_MASK_RESOURCE_EXCEPTION_HANDLING) {
+ malloc_register_stack_logger();
+ }
+#endif /* ENABLE_MEMORY_RESOURCE_EXCEPTION_HANDLING */
+
+ if (msl.handle_memory_event) {
+ // Let MSL see the event.
+ msl.handle_memory_event(event);
+ }
+}
+
+size_t
+malloc_zone_pressure_relief(malloc_zone_t *zone, size_t goal)
{
if (!zone) {
unsigned index = 0;
@@ -1278,25 +3306,30 @@
MALLOC_LOCK();
while (index < malloc_num_zones) {
zone = malloc_zones[index++];
- if (zone->version < 8)
+ if (zone->version < 8) {
continue;
- if (NULL == zone->pressure_relief)
+ }
+ if (NULL == zone->pressure_relief) {
continue;
- if (0 == goal) /* Greedy */
+ }
+ if (0 == goal) { /* Greedy */
total += zone->pressure_relief(zone, 0);
- else if (goal > total)
+ } else if (goal > total) {
total += zone->pressure_relief(zone, goal - total);
- else /* total >= goal */
+ } else { /* total >= goal */
break;
+ }
}
MALLOC_UNLOCK();
return total;
} else {
// Assumes zone is not destroyed for the duration of this call
- if (zone->version < 8)
+ if (zone->version < 8) {
return 0;
- if (NULL == zone->pressure_relief)
+ }
+ if (NULL == zone->pressure_relief) {
return 0;
+ }
return zone->pressure_relief(zone, goal);
}
}
@@ -1304,18 +3337,21 @@
/********* Batch methods ************/
unsigned
-malloc_zone_batch_malloc(malloc_zone_t *zone, size_t size, void **results, unsigned num_requested) {
- unsigned (*batch_malloc)(malloc_zone_t *, size_t, void **, unsigned) = zone-> batch_malloc;
- if (! batch_malloc)
+malloc_zone_batch_malloc(malloc_zone_t *zone, size_t size, void **results, unsigned num_requested)
+{
+ if (!zone->batch_malloc) {
return 0;
- if (malloc_check_start && (malloc_check_counter++ >= malloc_check_start)) {
+ }
+ if (malloc_check_start) {
internal_check();
}
- unsigned batched = batch_malloc(zone, size, results, num_requested);
- if (malloc_logger) {
- unsigned index = 0;
+ unsigned batched = zone->batch_malloc(zone, size, results, num_requested);
+
+ if (os_unlikely(malloc_logger)) {
+ unsigned index = 0;
while (index < batched) {
- malloc_logger(MALLOC_LOG_TYPE_ALLOCATE | MALLOC_LOG_TYPE_HAS_ZONE, (uintptr_t)zone, (uintptr_t)size, 0, (uintptr_t)results[index], 0);
+ malloc_logger(MALLOC_LOG_TYPE_ALLOCATE | MALLOC_LOG_TYPE_HAS_ZONE, (uintptr_t)zone, (uintptr_t)size, 0,
+ (uintptr_t)results[index], 0);
index++;
}
}
@@ -1323,60 +3359,139 @@
}
void
-malloc_zone_batch_free(malloc_zone_t *zone, void **to_be_freed, unsigned num) {
- if (malloc_check_start && (malloc_check_counter++ >= malloc_check_start)) {
+malloc_zone_batch_free(malloc_zone_t *zone, void **to_be_freed, unsigned num)
+{
+ if (malloc_check_start) {
internal_check();
}
- if (malloc_logger) {
- unsigned index = 0;
+ if (os_unlikely(malloc_logger)) {
+ unsigned index = 0;
while (index < num) {
- malloc_logger(MALLOC_LOG_TYPE_DEALLOCATE | MALLOC_LOG_TYPE_HAS_ZONE, (uintptr_t)zone, (uintptr_t)to_be_freed[index], 0, 0, 0);
+ malloc_logger(
+ MALLOC_LOG_TYPE_DEALLOCATE | MALLOC_LOG_TYPE_HAS_ZONE, (uintptr_t)zone, (uintptr_t)to_be_freed[index], 0, 0, 0);
index++;
}
}
- void (*batch_free)(malloc_zone_t *, void **, unsigned) = zone-> batch_free;
- if (batch_free) {
- batch_free(zone, to_be_freed, num);
+
+ if (zone->batch_free) {
+ zone->batch_free(zone, to_be_freed, num);
} else {
- void (*free_fun)(malloc_zone_t *, void *) = zone->free;
+ void (*free_fun)(malloc_zone_t *, void *) = zone->free;
+
while (num--) {
- void *ptr = *to_be_freed++;
+ void *ptr = *to_be_freed++;
free_fun(zone, ptr);
}
}
}
+/********* Functions for sanitization ************/
+
+#if TARGET_OS_OSX
+
+// Shadow memory helpers
+#define SHADOW_MEMORY_BASE (0x0000200000000000ull)
+#define SHADOW_GRANULARITY_BITS 3u
+#define SHADOW_GRANULARITY (1u << SHADOW_GRANULARITY_BITS)
+#define PTR_TO_SHADOW(p) (void *)((((uintptr_t)ptr) >> SHADOW_GRANULARITY_BITS) + SHADOW_MEMORY_BASE)
+#define SIZE_TO_SHADOW_SIZE(p) ((p) >> SHADOW_GRANULARITY_BITS)
+#define SHADOW_POISON_HEAP_LEFT_RZ 0xfa
+#define SHADOW_POISON_HEAP_RIGHT_RZ 0xfb
+#define SHADOW_POISON_HEAP_FREED 0xfd
+
+static void
+malloc_sanitizer_fallback_allocate_poison(uintptr_t ptr, size_t leftrz_sz, size_t alloc_sz, size_t rightrz_sz)
+{
+ uintptr_t alloc_ptr, rightrz_ptr;
+ MALLOC_ASSERT(!os_add_overflow(ptr, leftrz_sz, &alloc_ptr) &&
+ !os_add_overflow(alloc_ptr, alloc_sz, &rightrz_ptr));
+
+ // Tag the ASan shadow as ASAN_POISON_USER/ASAN_VALID/ASAN_POISON_USER
+ if (sanitizer_fallback_ptrs.memory_poison && sanitizer_fallback_ptrs.memory_unpoison) {
+ (*sanitizer_fallback_ptrs.memory_poison)(ptr, leftrz_sz);
+ (*sanitizer_fallback_ptrs.memory_unpoison)(alloc_ptr, alloc_sz);
+ (*sanitizer_fallback_ptrs.memory_poison)(rightrz_ptr, rightrz_sz);
+ } else {
+ // This code should only be executed by Rosetta on macOS
+ MALLOC_ASSERT(!(leftrz_sz % SHADOW_GRANULARITY) &&
+ !((alloc_sz + rightrz_sz) % SHADOW_GRANULARITY));
+
+ const uint8_t partial = alloc_sz & (SHADOW_GRANULARITY - 1);
+ // Partial poison is not supported, leave it unpoisoned
+ const size_t left = SIZE_TO_SHADOW_SIZE(leftrz_sz),
+ right = SIZE_TO_SHADOW_SIZE(rightrz_sz),
+ alloc = SIZE_TO_SHADOW_SIZE(alloc_sz) + !!partial;
+ uint8_t *shadow = PTR_TO_SHADOW(ptr);
+ memset(shadow, SHADOW_POISON_HEAP_LEFT_RZ, left);
+ shadow += left;
+ bzero(shadow + left, alloc);
+ shadow += alloc;
+ memset(shadow, SHADOW_POISON_HEAP_RIGHT_RZ, right);
+ }
+}
+
+static void
+malloc_sanitizer_fallback_deallocate_poison(uintptr_t ptr, size_t sz)
+{
+ // Tag the ASan shadow as ASAN_POISON_USER
+ if (sanitizer_fallback_ptrs.memory_poison) {
+ (*sanitizer_fallback_ptrs.memory_poison)(ptr, sz);
+ } else {
+ memset(PTR_TO_SHADOW(ptr), SHADOW_POISON_HEAP_FREED, SIZE_TO_SHADOW_SIZE(sz));
+ }
+}
+
+#endif // TARGET_OS_OSX
+
+bool malloc_sanitizer_is_enabled(void)
+{
+ return malloc_sanitizer_enabled;
+}
+
+const struct malloc_sanitizer_poison *
+malloc_sanitizer_get_functions(void)
+{
+#if TARGET_OS_OSX
+ return sanitizer ? sanitizer : &sanitizer_fallback;
+#else
+ return sanitizer;
+#endif
+}
+
+void
+malloc_sanitizer_set_functions(struct malloc_sanitizer_poison *s)
+{
+ sanitizer = s;
+}
+
/********* Functions for performance tools ************/
-static kern_return_t
-_malloc_default_reader(task_t task, vm_address_t address, vm_size_t size, void **ptr) {
- *ptr = (void *)address;
- return 0;
-}
-
kern_return_t
-malloc_get_all_zones(task_t task, memory_reader_t reader, vm_address_t **addresses, unsigned *count) {
- // Note that the 2 following addresses are not correct if the address of the target is different from your own. This notably occurs if the address of System.framework is slid (e.g. different than at B & I )
- vm_address_t remote_malloc_zones = (vm_address_t)&malloc_zones;
- vm_address_t remote_malloc_num_zones = (vm_address_t)&malloc_num_zones;
- kern_return_t err;
- vm_address_t zones_address;
- vm_address_t *zones_address_ref;
- unsigned num_zones;
- unsigned *num_zones_ref;
- if (!reader) reader = _malloc_default_reader;
+malloc_get_all_zones(task_t task, memory_reader_t reader, vm_address_t **addresses, unsigned *count)
+{
+ // Note that the 2 following addresses are not correct if the address of the target is different from your own. This notably
+ // occurs if the address of System.framework is slid (e.g. different than at B & I )
+ vm_address_t remote_malloc_zones = (vm_address_t)&malloc_zones;
+ vm_address_t remote_malloc_num_zones = (vm_address_t)&malloc_num_zones;
+ kern_return_t err;
+ vm_address_t zones_address;
+ vm_address_t *zones_address_ref;
+ unsigned num_zones;
+ unsigned *num_zones_ref;
+ reader = reader_or_in_memory_fallback(reader, task);
+
// printf("Read malloc_zones at address %p should be %p\n", &malloc_zones, malloc_zones);
err = reader(task, remote_malloc_zones, sizeof(void *), (void **)&zones_address_ref);
// printf("Read malloc_zones[%p]=%p\n", remote_malloc_zones, *zones_address_ref);
if (err) {
- malloc_printf("*** malloc_get_all_zones: error reading zones_address at %p\n", (unsigned)remote_malloc_zones);
+ malloc_report(ASL_LEVEL_ERR, "*** malloc_get_all_zones: error reading zones_address at %p\n", (void *)remote_malloc_zones);
return err;
}
zones_address = *zones_address_ref;
// printf("Reading num_zones at address %p\n", remote_malloc_num_zones);
err = reader(task, remote_malloc_num_zones, sizeof(unsigned), (void **)&num_zones_ref);
if (err) {
- malloc_printf("*** malloc_get_all_zones: error reading num_zones at %p\n", (unsigned)remote_malloc_num_zones);
+ malloc_report(ASL_LEVEL_ERR, "*** malloc_get_all_zones: error reading num_zones at %p\n", (void *)remote_malloc_num_zones);
return err;
}
num_zones = *num_zones_ref;
@@ -1385,7 +3500,7 @@
// printf("malloc_get_all_zones succesfully found %d zones\n", num_zones);
err = reader(task, zones_address, sizeof(malloc_zone_t *) * num_zones, (void **)addresses);
if (err) {
- malloc_printf("*** malloc_get_all_zones: error reading zones at %p\n", &zones_address);
+ malloc_report(ASL_LEVEL_ERR, "*** malloc_get_all_zones: error reading zones at %p\n", &zones_address);
return err;
}
// printf("malloc_get_all_zones succesfully read %d zones\n", num_zones);
@@ -1395,9 +3510,12 @@
/********* Debug helpers ************/
void
-malloc_zone_print_ptr_info(void *ptr) {
- malloc_zone_t *zone;
- if (!ptr) return;
+malloc_zone_print_ptr_info(void *ptr)
+{
+ malloc_zone_t *zone;
+ if (!ptr) {
+ return;
+ }
zone = malloc_zone_from_ptr(ptr);
if (zone) {
printf("ptr %p in registered zone %p\n", ptr, zone);
@@ -1407,13 +3525,16 @@
}
boolean_t
-malloc_zone_check(malloc_zone_t *zone) {
- boolean_t ok = 1;
+malloc_zone_check(malloc_zone_t *zone)
+{
+ boolean_t ok = 1;
if (!zone) {
- unsigned index = 0;
+ unsigned index = 0;
while (index < malloc_num_zones) {
zone = malloc_zones[index++];
- if (!zone->introspect->check(zone)) ok = 0;
+ if (!zone->introspect->check(zone)) {
+ ok = 0;
+ }
}
} else {
ok = zone->introspect->check(zone);
@@ -1422,9 +3543,10 @@
}
void
-malloc_zone_print(malloc_zone_t *zone, boolean_t verbose) {
+malloc_zone_print(malloc_zone_t *zone, boolean_t verbose)
+{
if (!zone) {
- unsigned index = 0;
+ unsigned index = 0;
while (index < malloc_num_zones) {
zone = malloc_zones[index++];
zone->introspect->print(zone, verbose);
@@ -1435,13 +3557,14 @@
}
void
-malloc_zone_statistics(malloc_zone_t *zone, malloc_statistics_t *stats) {
+malloc_zone_statistics(malloc_zone_t *zone, malloc_statistics_t *stats)
+{
if (!zone) {
memset(stats, 0, sizeof(*stats));
- unsigned index = 0;
+ unsigned index = 0;
while (index < malloc_num_zones) {
zone = malloc_zones[index++];
- malloc_statistics_t this_stats;
+ malloc_statistics_t this_stats;
zone->introspect->statistics(zone, &this_stats);
stats->blocks_in_use += this_stats.blocks_in_use;
stats->size_in_use += this_stats.size_in_use;
@@ -1454,9 +3577,10 @@
}
void
-malloc_zone_log(malloc_zone_t *zone, void *address) {
+malloc_zone_log(malloc_zone_t *zone, void *address)
+{
if (!zone) {
- unsigned index = 0;
+ unsigned index = 0;
while (index < malloc_num_zones) {
zone = malloc_zones[index++];
zone->introspect->log(zone, address);
@@ -1468,80 +3592,142 @@
/********* Misc other entry points ************/
+void
+mag_set_thread_index(unsigned int index)
+{
+ _os_cpu_number_override = index;
+#if CONFIG_NANOZONE
+ nano_common_cpu_number_override_set();
+#endif // CONFIG_NANOZONE
+}
+
static void
-DefaultMallocError(int x) {
+DefaultMallocError(int x)
+{
#if USE_SLEEP_RATHER_THAN_ABORT
- malloc_printf("*** error %d\n", x);
+ malloc_report(ASL_LEVEL_ERR, "*** error %d\n", x);
sleep(3600);
#else
_SIMPLE_STRING b = _simple_salloc();
if (b) {
_simple_sprintf(b, "*** error %d", x);
- malloc_printf("%s\n", _simple_string(b));
- CRSetCrashLogMessage(_simple_string(b));
+ malloc_report(MALLOC_REPORT_NOLOG, "%s\n", _simple_string(b));
+ _os_set_crash_log_message_dynamic(_simple_string(b));
} else {
- _malloc_printf(MALLOC_PRINTF_NOLOG, "*** error %d", x);
- CRSetCrashLogMessage("*** DefaultMallocError called");
+ malloc_report(MALLOC_REPORT_NOLOG, "*** error %d\n", x);
+ _os_set_crash_log_message("*** DefaultMallocError called");
}
abort();
#endif
}
-void (*
- malloc_error(void (*func)(int)))(int) {
+void (*malloc_error(void (*func)(int)))(int)
+{
return DefaultMallocError;
}
-/* Stack logging fork-handling prototypes */
-extern void __stack_logging_fork_prepare();
-extern void __stack_logging_fork_parent();
-extern void __stack_logging_fork_child();
-extern void __stack_logging_early_finished();
-
static void
-_malloc_lock_all(void (*callout)(void)) {
+_malloc_lock_all(void (*callout)(void))
+{
unsigned index = 0;
MALLOC_LOCK();
+#if CONFIG_EARLY_MALLOC
+ mfm_lock();
+#endif
while (index < malloc_num_zones) {
malloc_zone_t *zone = malloc_zones[index++];
zone->introspect->force_lock(zone);
}
- callout();
+#if CONFIG_XZONE_MALLOC
+ // All of the xzone malloc zones share some global state, and that global
+ // state needs to be locked after all the zone specific state has been
+ // locked, to prevent deadlocks
+ if (initial_xzone_zone) {
+ xzm_force_lock_global_state(initial_xzone_zone);
+ }
+#endif // CONFIG_XZONE_MALLOC
+ if (callout) {
+ callout();
+ }
}
static void
-_malloc_unlock_all(void (*callout)(void)) {
+_malloc_unlock_all(void (*callout)(void))
+{
+#if CONFIG_XZONE_MALLOC
+ if (initial_xzone_zone) {
+ xzm_force_unlock_global_state(initial_xzone_zone);
+ }
+#endif // CONFIG_XZONE_MALLOC
unsigned index = 0;
- callout();
+ if (callout) {
+ callout();
+ }
while (index < malloc_num_zones) {
malloc_zone_t *zone = malloc_zones[index++];
zone->introspect->force_unlock(zone);
}
+#if CONFIG_EARLY_MALLOC
+ mfm_unlock();
+#endif
MALLOC_UNLOCK();
}
+
+static void
+_malloc_reinit_lock_all(void (*callout)(void))
+{
+#if CONFIG_XZONE_MALLOC
+ if (initial_xzone_zone) {
+ xzm_force_reinit_lock_global_state(initial_xzone_zone);
+ }
+#endif // CONFIG_XZONE_MALLOC
+ unsigned index = 0;
+ if (callout) {
+ callout();
+ }
+ while (index < malloc_num_zones) {
+ malloc_zone_t *zone = malloc_zones[index++];
+ if (zone->version < 9) { // Version must be >= 9 to look at reinit_lock
+ zone->introspect->force_unlock(zone);
+ } else {
+ zone->introspect->reinit_lock(zone);
+ }
+ }
+#if CONFIG_EARLY_MALLOC
+ mfm_reinit_lock();
+#endif
+ MALLOC_REINIT_LOCK();
+}
+
// Called prior to fork() to guarantee that malloc is not in any critical
// sections during the fork(); prevent any locks from being held by non-
// surviving threads after the fork.
void
-_malloc_fork_prepare(void) {
- return _malloc_lock_all(&__stack_logging_fork_prepare);
+_malloc_fork_prepare(void)
+{
+ return _malloc_lock_all(msl.fork_prepare);
}
// Called in the parent process after fork() to resume normal operation.
void
-_malloc_fork_parent(void) {
- return _malloc_unlock_all(&__stack_logging_fork_parent);
+_malloc_fork_parent(void)
+{
+ return _malloc_unlock_all(msl.fork_parent);
}
// Called in the child process after fork() to resume normal operation.
void
-_malloc_fork_child(void) {
+_malloc_fork_child(void)
+{
#if CONFIG_NANOZONE
- if (_malloc_is_initialized && _malloc_engaged_nano)
- nano_forked_zone(inline_malloc_default_zone());
+ if (_malloc_entropy_initialized) {
+ if (initial_nano_zone) {
+ nanov2_forked_zone((nanozonev2_t *)initial_nano_zone);
+ }
+ }
#endif
- return _malloc_unlock_all(&__stack_logging_fork_child);
+ return _malloc_reinit_lock_all(msl.fork_child);
}
/*
@@ -1562,42 +3748,50 @@
m.chunks_used = s.blocks_in_use;
m.bytes_used = s.size_in_use;
m.chunks_free = 0;
- m.bytes_free = m.bytes_total - m.bytes_used; /* isn't this somewhat obvious? */
-
- return(m);
+ m.bytes_free = m.bytes_total - m.bytes_used; /* isn't this somewhat obvious? */
+
+ return (m);
}
boolean_t
malloc_zone_enable_discharge_checking(malloc_zone_t *zone)
{
- if (zone->version < 7) // Version must be >= 7 to look at the new discharge checking fields.
+ if (zone->version < 7) { // Version must be >= 7 to look at the new discharge checking fields.
return FALSE;
- if (NULL == zone->introspect->enable_discharge_checking)
+ }
+ if (NULL == zone->introspect->enable_discharge_checking) {
return FALSE;
+ }
return zone->introspect->enable_discharge_checking(zone);
}
void
malloc_zone_disable_discharge_checking(malloc_zone_t *zone)
{
- if (zone->version < 7) // Version must be >= 7 to look at the new discharge checking fields.
+ if (zone->version < 7) { // Version must be >= 7 to look at the new discharge checking fields.
return;
- if (NULL == zone->introspect->disable_discharge_checking)
+ }
+ if (NULL == zone->introspect->disable_discharge_checking) {
return;
+ }
zone->introspect->disable_discharge_checking(zone);
}
void
malloc_zone_discharge(malloc_zone_t *zone, void *memory)
{
- if (NULL == zone)
+ if (NULL == zone) {
zone = malloc_zone_from_ptr(memory);
- if (NULL == zone)
+ }
+ if (NULL == zone) {
return;
- if (zone->version < 7) // Version must be >= 7 to look at the new discharge checking fields.
+ }
+ if (zone->version < 7) { // Version must be >= 7 to look at the new discharge checking fields.
return;
- if (NULL == zone->introspect->discharge)
+ }
+ if (NULL == zone->introspect->discharge) {
return;
+ }
zone->introspect->discharge(zone, memory);
}
@@ -1605,54 +3799,246 @@
malloc_zone_enumerate_discharged_pointers(malloc_zone_t *zone, void (^report_discharged)(void *memory, void *info))
{
if (!zone) {
- unsigned index = 0;
+ unsigned index = 0;
while (index < malloc_num_zones) {
zone = malloc_zones[index++];
- if (zone->version < 7)
+ if (zone->version < 7) {
continue;
- if (NULL == zone->introspect->enumerate_discharged_pointers)
+ }
+ if (NULL == zone->introspect->enumerate_discharged_pointers) {
continue;
+ }
zone->introspect->enumerate_discharged_pointers(zone, report_discharged);
}
} else {
- if (zone->version < 7)
+ if (zone->version < 7) {
return;
- if (NULL == zone->introspect->enumerate_discharged_pointers)
+ }
+ if (NULL == zone->introspect->enumerate_discharged_pointers) {
return;
+ }
zone->introspect->enumerate_discharged_pointers(zone, report_discharged);
}
}
+void
+malloc_zero_on_free_disable(void)
+{
+ malloc_zero_policy = MALLOC_ZERO_NONE;
+}
+
+bool
+malloc_variant_is_debug_4test(void)
+{
+#ifdef DEBUG
+ return true;
+#else
+ return false;
+#endif
+}
+
/***************** OBSOLETE ENTRY POINTS ********************/
+
+#ifndef PHASE_OUT_OLD_MALLOC
+#define PHASE_OUT_OLD_MALLOC 0
+#endif
#if PHASE_OUT_OLD_MALLOC
#error PHASE OUT THE FOLLOWING FUNCTIONS
#endif
void
-set_malloc_singlethreaded(boolean_t single) {
+set_malloc_singlethreaded(boolean_t single)
+{
static boolean_t warned = 0;
if (!warned) {
#if PHASE_OUT_OLD_MALLOC
- malloc_printf("*** OBSOLETE: set_malloc_singlethreaded(%d)\n", single);
+ malloc_report(ASL_LEVEL_ERR, "*** OBSOLETE: set_malloc_singlethreaded(%d)\n", single);
#endif
warned = 1;
}
}
void
-malloc_singlethreaded(void) {
+malloc_singlethreaded(void)
+{
static boolean_t warned = 0;
if (!warned) {
- malloc_printf("*** OBSOLETE: malloc_singlethreaded()\n");
+ malloc_report(ASL_LEVEL_ERR, "*** OBSOLETE: malloc_singlethreaded()\n");
warned = 1;
}
}
int
-malloc_debug(int level) {
- malloc_printf("*** OBSOLETE: malloc_debug()\n");
+malloc_debug(int level)
+{
+ malloc_report(ASL_LEVEL_ERR, "*** OBSOLETE: malloc_debug()\n");
return 0;
}
+#pragma mark -
+#pragma mark Malloc Thread Options
+
+typedef union {
+ malloc_thread_options_t options;
+ void *storage;
+} th_opts_t;
+
+MALLOC_STATIC_ASSERT(sizeof(th_opts_t) == sizeof(void *), "Options fit into pointer bits");
+
+malloc_thread_options_t
+malloc_get_thread_options(void)
+{
+ th_opts_t x = {.storage = _pthread_getspecific_direct(__TSD_MALLOC_THREAD_OPTIONS)};
+ return x.options;
+}
+
+void
+malloc_set_thread_options(malloc_thread_options_t opts)
+{
+ // Canonicalize options
+ if (opts.DisableExpensiveDebuggingOptions) {
+ opts.DisableProbabilisticGuardMalloc = true;
+ opts.DisableMallocStackLogging = true;
+ }
+
+ pgm_thread_set_disabled(opts.DisableProbabilisticGuardMalloc);
+
+ th_opts_t x = {.options = opts};
+ _pthread_setspecific_direct(__TSD_MALLOC_THREAD_OPTIONS, x.storage);
+}
+
+#pragma mark -
+#pragma mark Malloc Stack Logging
+
+static bool _malloc_register_stack_logger(bool at_startup);
+
+/* this is called from libsystem during initialization. */
+static void
+stack_logging_early_finished(const struct _malloc_late_init *funcs)
+{
+#if !TARGET_OS_DRIVERKIT
+ _dlopen = funcs->dlopen;
+ _dlsym = funcs->dlsym;
+#endif
+ if (funcs->version >= 2 && funcs->msl) {
+ msl = *funcs->msl;
+ }
+ const char **env = (const char**) *_NSGetEnviron();
+ for (const char **e = env; *e; e++) {
+ if (0==strncmp(*e, "MallocStackLogging", 18)) {
+ _malloc_register_stack_logger(true);
+ if (msl.set_flags_from_environment) {
+ // rdar://125495815 - re-fetch env, as it may have moved
+ msl.set_flags_from_environment((const char **)*_NSGetEnviron());
+ }
+ break;
+ }
+ }
+ if (msl.initialize) {
+ msl.initialize();
+ }
+}
+
+
+static os_once_t _register_msl_dylib_pred;
+
+static void
+register_msl_dylib(void *dylib)
+{
+#if TARGET_OS_DRIVERKIT
+ set_msl_lite_hooks(msl.copy_msl_lite_hooks);
+#else
+ if (!dylib) {
+ return;
+ }
+ msl.handle_memory_event = _dlsym(dylib, "msl_handle_memory_event");
+ msl.stack_logging_locked = _dlsym(dylib, "msl_stack_logging_locked");
+ msl.fork_prepare = _dlsym(dylib, "msl_fork_prepare");
+ msl.fork_child = _dlsym(dylib, "msl_fork_child");
+ msl.fork_parent = _dlsym(dylib, "msl_fork_parent");
+ msl.turn_on_stack_logging = _dlsym(dylib, "msl_turn_on_stack_logging");
+ msl.turn_off_stack_logging = _dlsym(dylib, "msl_turn_off_stack_logging");
+ msl.set_flags_from_environment = _dlsym(dylib, "msl_set_flags_from_environment");
+ msl.initialize = _dlsym(dylib, "msl_initialize");
+
+ void (*msl_copy_msl_lite_hooks) (struct _malloc_msl_lite_hooks_s *hooksp, size_t size);
+ msl_copy_msl_lite_hooks = _dlsym(dylib, "msl_copy_msl_lite_hooks");
+ if (msl_copy_msl_lite_hooks) {
+ set_msl_lite_hooks(msl_copy_msl_lite_hooks);
+ }
+#endif
+}
+
+MALLOC_EXPORT
+boolean_t
+malloc_register_stack_logger(void)
+{
+ return _malloc_register_stack_logger(false);
+}
+
+static bool
+_malloc_register_stack_logger(bool at_startup)
+{
+ void *dylib = NULL;
+ if (malloc_sanitizer_enabled && !at_startup) {
+ return false;
+ }
+#if !TARGET_OS_DRIVERKIT
+ if (msl.handle_memory_event != NULL) {
+ return true;
+ }
+ if (_dlopen == NULL) {
+ return false;
+ }
+ dylib = _dlopen("/System/Library/PrivateFrameworks/MallocStackLogging.framework/MallocStackLogging", RTLD_GLOBAL);
+ if (dylib == NULL) {
+ return false;
+ }
+#endif
+ os_once(&_register_msl_dylib_pred, dylib, register_msl_dylib);
+ if (!msl.handle_memory_event) {
+ malloc_report(ASL_LEVEL_WARNING, "failed to load MallocStackLogging.framework\n");
+ return false;
+ }
+ return true;
+}
+
+/* Symbolication.framework looks up this symbol by name inside libsystem_malloc.dylib. */
+uint64_t __mach_stack_logging_shared_memory_address = 0;
+
+
+#pragma mark -
+#pragma mark Malloc Stack Logging - Legacy stubs
+
+/*
+ * legacy API for MallocStackLogging.
+ *
+ * TODO, deprecate this, move clients off it and delete it. Clients should move
+ * to MallocStackLogging.framework for these APIs.
+ */
+
+MALLOC_EXPORT
+boolean_t
+turn_on_stack_logging(stack_logging_mode_type mode)
+{
+ malloc_register_stack_logger();
+ if (!msl.turn_on_stack_logging) {
+ return false;
+ }
+ return msl.turn_on_stack_logging(mode);
+}
+
+MALLOC_EXPORT
+void turn_off_stack_logging(void)
+{
+ malloc_register_stack_logger();
+ if (msl.turn_off_stack_logging) {
+ msl.turn_off_stack_logging();
+ }
+}
+
+/* WeChat references this, only god knows why. This symbol does nothing. */
+int stack_logging_enable_logging = 0;
+
/* vim: set noet:ts=4:sw=4:cindent: */