Loading...
--- libmalloc/libmalloc-283.40.1/src/nano_malloc.c
+++ libmalloc/libmalloc-53.30.1/src/nano_malloc.c
@@ -21,48 +21,326 @@
* @APPLE_LICENSE_HEADER_END@
*/
-#include "internal.h"
-
-/* nano_malloc for 64bit ABI */
-#if CONFIG_NANOZONE
+#ifdef __LP64__ /* nano_malloc for 64bit ABI */
+#define NDEBUG 1
+#define NANO_FREE_DEQUEUE_DILIGENCE 1 /* Check for corrupt free list */
+
+#include <_simple.h>
+#include <assert.h>
+#include <stddef.h>
+#include <stdint.h>
+#include <stdlib.h>
+#include <unistd.h>
+#include <limits.h>
+#include <errno.h>
+#include <TargetConditionals.h>
+
+#include <sys/types.h>
+#include <sys/mman.h>
+#include <sys/param.h>
+
+#include <mach/mach.h>
+#include <mach/mach_vm.h>
+
+#include <libkern/OSAtomic.h>
+#include <mach-o/dyld_priv.h> /* for _dyld_get_image_slide() */
+#include <crt_externs.h> /* for _NSGetMachExecuteHeader() */
+
+#include <os/tsd.h>
+
+#if defined(__x86_64__)
+#define __APPLE_API_PRIVATE
+#include <machine/cpu_capabilities.h>
+#define _COMM_PAGE_VERSION_REQD 9
+#undef __APPLE_API_PRIVATE
+#else
+Unknown Architecture
+#endif
+
+#include "scalable_malloc.h"
+#include "malloc_internal.h"
+#include "malloc_printf.h"
+
+#include <CrashReporterClient.h>
+
+#include "bitarray.h"
+
+#ifndef VM_MEMORY_MALLOC_NANO /* Until osfmk/mach/vm_statistics.h is updated in xnu */
+#define VM_MEMORY_MALLOC_NANO 11
+#endif
+
+extern uint64_t malloc_entropy[2];
+/********************* DEFINITIONS ************************/
+
+#define INLINE __inline__
+#define ALWAYSINLINE __attribute__((always_inline))
+#define NOINLINE __attribute__((noinline))
+
+#if defined(__x86_64__)
+#define CACHE_LINE 64
+#define CACHE_ALIGN __attribute__ ((aligned (64) ))
+#else
+#define CACHE_ALIGN /* TBD for other platforms */
+#endif
+
+#define NANO_MAG_INDEX(nz) (_os_cpu_number() >> nz->hyper_shift)
+
+#define SCRIBBLE_BYTE 0xaa /* allocated scribble */
+#define SCRABBLE_BYTE 0x55 /* free()'d scribble */
+#define SCRUBBLE_BYTE 0xdd /* madvise(..., MADV_FREE) scriblle */
+
+#define MAX_RECORDER_BUFFER 256
+
+/************* nanozone address field layout ******************/
+
+#if defined(__x86_64)
+#define NANO_SIGNATURE_BITS 20
+#define NANOZONE_SIGNATURE 0x00006ULL // 0x00006nnnnnnnnnnn the address range devoted to us.
+#define NANO_MAG_BITS 5
+#define NANO_BAND_BITS 18
+#define NANO_SLOT_BITS 4
+#define NANO_OFFSET_BITS 17
+
+#else
+#error Unknown Architecture
+#endif
+
+#if defined(__BIG_ENDIAN__)
+struct nano_blk_addr_s {
+ uint64_t
+nano_signature:NANO_SIGNATURE_BITS, // 0x00006nnnnnnnnnnn the address range devoted to us.
+nano_mag_index:NANO_MAG_BITS, // the core that allocated this block
+nano_band:NANO_BAND_BITS,
+nano_slot:NANO_SLOT_BITS, // bucket of homogenous quanta-multiple blocks
+nano_offset:NANO_OFFSET_BITS; // locates the block
+};
+#else
+// least significant bits declared first
+struct nano_blk_addr_s {
+ uint64_t
+nano_offset:NANO_OFFSET_BITS, // locates the block
+nano_slot:NANO_SLOT_BITS, // bucket of homogenous quanta-multiple blocks
+nano_band:NANO_BAND_BITS,
+nano_mag_index:NANO_MAG_BITS, // the core that allocated this block
+nano_signature:NANO_SIGNATURE_BITS; // 0x00006nnnnnnnnnnn the address range devoted to us.
+};
+#endif
+
+typedef union {
+ uint64_t addr;
+ struct nano_blk_addr_s fields;
+} nano_blk_addr_t;
+
+/* Are we using the nano allocator? Set by the initializer. */
+__attribute__((visibility("hidden")))
+boolean_t _malloc_engaged_nano;
+
+#define NANO_MAX_SIZE 256 /* Buckets sized {16, 32, 48, 64, 80, 96, 112, ...} */
+#define SHIFT_NANO_QUANTUM 4
+#define NANO_REGIME_QUANTA_SIZE (1 << SHIFT_NANO_QUANTUM) // 16
+#define NANO_QUANTA_MASK 0xFULL // NANO_REGIME_QUANTA_SIZE - 1
+
+#define SLOT_IN_BAND_SIZE (1 << NANO_OFFSET_BITS)
+#define SLOT_KEY_LIMIT (1 << NANO_SLOT_BITS) /* Must track nano_slot width */
+#define BAND_SIZE (1 << (NANO_SLOT_BITS + NANO_OFFSET_BITS)) /* == Number of bytes covered by a page table entry */
+#define NANO_MAG_SIZE (1 << NANO_MAG_BITS)
+#define NANO_SLOT_SIZE (1 << NANO_SLOT_BITS)
+
+/****************************** zone itself ***********************************/
+
+/*
+ * Note that objects whose adddress are held in pointers here must be pursued
+ * individually in the nano_in_use_enumeration() routines.
+ */
+
+typedef struct chained_block_s {
+ uintptr_t double_free_guard;
+ struct chained_block_s *next;
+} *chained_block_t;
+
+typedef struct nano_meta_s {
+ OSQueueHead slot_LIFO CACHE_ALIGN;
+ unsigned int slot_madvised_log_page_count;
+ volatile uintptr_t slot_current_base_addr;
+ volatile uintptr_t slot_limit_addr;
+ volatile size_t slot_objects_mapped;
+ volatile size_t slot_objects_skipped;
+ bitarray_t slot_madvised_pages;
+ volatile uintptr_t slot_bump_addr CACHE_ALIGN; // position on cache line distinct from that of slot_LIFO
+ volatile boolean_t slot_exhausted;
+ unsigned int slot_bytes;
+ unsigned int slot_objects;
+} *nano_meta_admin_t;
+
+typedef struct nanozone_s { // vm_allocate()'d, so page-aligned to begin with.
+ malloc_zone_t basic_zone; // first page will be given read-only protection
+ uint8_t pad[PAGE_MAX_SIZE - sizeof(malloc_zone_t)];
+
+ // remainder of structure is R/W (contains no function pointers)
+ // page-aligned
+ struct nano_meta_s meta_data[NANO_MAG_SIZE][NANO_SLOT_SIZE]; // max: NANO_MAG_SIZE cores x NANO_SLOT_SIZE slots for nano blocks {16 .. 256}
+ _malloc_lock_s band_resupply_lock[NANO_MAG_SIZE];
+ uintptr_t band_max_mapped_baseaddr[NANO_MAG_SIZE];
+ size_t core_mapped_size[NANO_MAG_SIZE];
+
+ unsigned debug_flags;
+ unsigned our_signature;
+ unsigned phys_ncpus;
+ unsigned logical_ncpus;
+ unsigned hyper_shift;
+
+ /* security cookie */
+ uintptr_t cookie;
+
+ /*
+ * The nano zone constructed by create_nano_zone() would like to hand off tiny, small, and large
+ * allocations to the default scalable zone. Record the latter as the "helper" zone here.
+ */
+ malloc_zone_t *helper_zone;
+} nanozone_t;
+
+#define SZONE_PAGED_SIZE ((sizeof(nanozone_t) + vm_page_size - 1) & ~ (vm_page_size - 1))
/********************* PROTOTYPES ***********************/
+extern void malloc_error_break(void);
+
+// msg prints after fmt, ...
+static NOINLINE void nanozone_error(nanozone_t *nanozone, int is_corruption, const char *msg, const void *ptr, const char *fmt, ...)
+__printflike(5, 6);
static void nano_statistics(nanozone_t *nanozone, malloc_statistics_t *stats);
/********************* VERY LOW LEVEL UTILITIES ************************/
// msg prints after fmt, ...
-static MALLOC_ALWAYS_INLINE unsigned int
-nano_mag_index(const nanozone_t *nanozone)
-{
- if (os_likely(_os_cpu_number_override == -1)) {
- return (_os_cpu_number() >> hyper_shift) % nano_common_max_magazines;
- }
- return (_os_cpu_number_override >> hyper_shift) % nano_common_max_magazines;
-}
-
-#if NANO_PREALLOCATE_BAND_VM
-static boolean_t
-nano_preallocate_band_vm(void)
-{
- nano_blk_addr_t u;
- uintptr_t s, e;
-
- u.fields.nano_signature = NANOZONE_SIGNATURE;
- u.fields.nano_mag_index = 0;
- u.fields.nano_band = 0;
- u.fields.nano_slot = 0;
- u.fields.nano_offset = 0;
- s = u.addr; // start of first possible band
-
- u.fields.nano_mag_index = (1 << NANO_MAG_BITS) - 1;
- u.fields.nano_band = (1 << NANO_BAND_BITS) - 1;
- e = u.addr + BAND_SIZE; // end of last possible band
-
- return nano_common_allocate_vm_space(s, e - s);
-}
-#endif
+static NOINLINE void
+nanozone_error(nanozone_t *nanozone, int is_corruption, const char *msg, const void *ptr, const char *fmt, ...)
+{
+ va_list ap;
+ _SIMPLE_STRING b = _simple_salloc();
+
+ if (b) {
+ if (fmt) {
+ va_start(ap, fmt);
+ _simple_vsprintf(b, fmt, ap);
+ va_end(ap);
+ }
+ if (ptr) {
+ _simple_sprintf(b, "*** error for object %p: %s\n", ptr, msg);
+ } else {
+ _simple_sprintf(b, "*** error: %s\n", msg);
+ }
+ malloc_printf("%s*** set a breakpoint in malloc_error_break to debug\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.
+ */
+ if (fmt) {
+ va_start(ap, fmt);
+ _malloc_vprintf(MALLOC_PRINTF_NOLOG, fmt, ap);
+ va_end(ap);
+ }
+ if (ptr) {
+ _malloc_printf(MALLOC_PRINTF_NOLOG, "*** error for object %p: %s\n", ptr, msg);
+ } else {
+ _malloc_printf(MALLOC_PRINTF_NOLOG, "*** error: %s\n", msg);
+ }
+ _malloc_printf(MALLOC_PRINTF_NOLOG, "*** set a breakpoint in malloc_error_break to debug\n");
+ }
+ malloc_error_break();
+
+ // Call abort() if this is a memory corruption error and the abort on
+ // corruption flag is set, or if any error should abort.
+ if ((is_corruption && (nanozone->debug_flags & SCALABLE_MALLOC_ABORT_ON_CORRUPTION)) ||
+ (nanozone->debug_flags & SCALABLE_MALLOC_ABORT_ON_ERROR)) {
+ CRSetCrashLogMessage(b ? _simple_string(b) : msg);
+ abort();
+ } else if (b) {
+ _simple_sfree(b);
+ }
+}
+
+static void
+protect(void *address, size_t size, unsigned protection, unsigned debug_flags)
+{
+ kern_return_t err;
+
+ if (!(debug_flags & SCALABLE_MALLOC_DONT_PROTECT_PRELUDE)) {
+ err = mprotect((void *)((uintptr_t)address - vm_page_size), vm_page_size, protection);
+ if (err) {
+ malloc_printf("*** can't protect(%p) region for prelude guard page at %p\n",
+ protection,(uintptr_t)address - (1 << vm_page_shift));
+ }
+ }
+ if (!(debug_flags & SCALABLE_MALLOC_DONT_PROTECT_POSTLUDE)) {
+ err = mprotect((void *)((uintptr_t)address + size), vm_page_size, protection);
+ if (err) {
+ malloc_printf("*** can't protect(%p) region for postlude guard page at %p\n",
+ protection, (uintptr_t)address + size);
+ }
+ }
+}
+
+static void *
+allocate_based_pages(nanozone_t *nanozone, size_t size, unsigned char align, unsigned debug_flags, int vm_page_label, void *base_addr)
+{
+ boolean_t add_guard_pages = debug_flags & SCALABLE_MALLOC_ADD_GUARD_PAGES;
+ mach_vm_address_t vm_addr;
+ uintptr_t addr;
+ mach_vm_size_t allocation_size = round_page(size);
+ mach_vm_offset_t allocation_mask = ((mach_vm_offset_t)1 << align) - 1;
+ int alloc_flags = VM_FLAGS_ANYWHERE | VM_MAKE_TAG(vm_page_label);
+ kern_return_t kr;
+
+ if (!allocation_size) allocation_size = vm_page_size;
+ if (add_guard_pages) allocation_size += 2 * vm_page_size;
+ if (allocation_size < size) // size_t arithmetic wrapped!
+ return NULL;
+
+ vm_addr = round_page((mach_vm_address_t)base_addr);
+ if (!vm_addr) vm_addr = vm_page_size;
+ kr = mach_vm_map(mach_task_self(), &vm_addr, allocation_size,
+ allocation_mask, alloc_flags, MEMORY_OBJECT_NULL, 0, FALSE,
+ VM_PROT_DEFAULT, VM_PROT_ALL, VM_INHERIT_DEFAULT);
+ if (kr) {
+ nanozone_error(nanozone, 0, "can't allocate pages", NULL,
+ "*** mach_vm_map(size=%lu) failed (error code=%d)\n",
+ size, kr);
+ return NULL;
+ }
+ addr = (uintptr_t)vm_addr;
+
+ if (add_guard_pages) {
+ addr += vm_page_size;
+ protect((void *)addr, size, PROT_NONE, debug_flags);
+ }
+ return (void *)addr;
+}
+
+static void *
+allocate_pages(nanozone_t *nanozone, size_t size, unsigned char align, unsigned debug_flags, int vm_page_label)
+{
+ return allocate_based_pages(nanozone, size, align, debug_flags, vm_page_label, 0);
+}
+
+static void
+deallocate_pages(nanozone_t *nanozone, void *addr, size_t size, unsigned debug_flags)
+{
+ boolean_t add_guard_pages = debug_flags & SCALABLE_MALLOC_ADD_GUARD_PAGES;
+ mach_vm_address_t vm_addr = (mach_vm_address_t)addr;
+ mach_vm_size_t allocation_size = size;
+ kern_return_t kr;
+
+ if (add_guard_pages) {
+ vm_addr -= vm_page_size;
+ allocation_size += 2 * vm_page_size;
+ }
+ kr = mach_vm_deallocate(mach_task_self(), vm_addr, allocation_size);
+ if (kr && nanozone)
+ nanozone_error(nanozone, 0, "Can't deallocate_pages at", addr, NULL);
+}
/*
* We maintain separate free lists for each (quantized) size. The literature
@@ -70,7 +348,7 @@
*/
static boolean_t
-segregated_band_grow(nanozone_t *nanozone, nano_meta_admin_t pMeta, size_t slot_bytes, unsigned int mag_index)
+segregated_band_grow(nanozone_t *nanozone, nano_meta_admin_t pMeta, unsigned int slot_bytes, unsigned int mag_index)
{
nano_blk_addr_t u; // the compiler holds this in a register
uintptr_t p, s;
@@ -85,37 +363,33 @@
u.fields.nano_offset = 0;
p = u.addr;
- pMeta->slot_bytes = (unsigned int)slot_bytes;
+ pMeta->slot_bytes = slot_bytes;
pMeta->slot_objects = SLOT_IN_BAND_SIZE / slot_bytes;
} else {
p = pMeta->slot_current_base_addr + BAND_SIZE; // Growing, so stride ahead by BAND_SIZE
u.addr = (uint64_t)p;
- if (0 == u.fields.nano_band) { // Did the band index wrap?
+ if (0 == u.fields.nano_band) // Did the band index wrap?
return FALSE;
- }
assert(slot_bytes == pMeta->slot_bytes);
}
pMeta->slot_current_base_addr = p;
mach_vm_address_t vm_addr = p & ~((uintptr_t)(BAND_SIZE - 1)); // Address of the (2MB) band covering this (128KB) slot
- if (nanozone->band_max_mapped_baseaddr[mag_index] < vm_addr) {
-#if !NANO_PREALLOCATE_BAND_VM
- // Obtain the next band to cover this slot
- kern_return_t kr = mach_vm_map(mach_task_self(), &vm_addr, BAND_SIZE, 0, VM_MAKE_TAG(VM_MEMORY_MALLOC_NANO),
- MEMORY_OBJECT_NULL, 0, FALSE, VM_PROT_DEFAULT, VM_PROT_ALL, VM_INHERIT_DEFAULT);
-
- void *q = (void *)vm_addr;
- if (kr || q != (void *)(p & ~((uintptr_t)(BAND_SIZE - 1)))) { // Must get exactly what we asked for
- if (!kr) {
- mach_vm_deallocate(mach_task_self(), vm_addr, BAND_SIZE);
- }
- return FALSE;
- }
-#endif
- nanozone->band_max_mapped_baseaddr[mag_index] = vm_addr;
- }
+
+ if (nanozone->band_max_mapped_baseaddr[mag_index] < vm_addr) {
+ // Obtain the next band to cover this slot
+ kern_return_t kr = mach_vm_map(mach_task_self(), &vm_addr, BAND_SIZE,
+ 0, VM_MAKE_TAG(VM_MEMORY_MALLOC_NANO), MEMORY_OBJECT_NULL, 0, FALSE,
+ VM_PROT_DEFAULT, VM_PROT_ALL, VM_INHERIT_DEFAULT);
+
+ void *q = (uintptr_t)vm_addr;
+ if (kr || q != (void *)(p & ~((uintptr_t)(BAND_SIZE - 1)))) // Must get exactly what we asked for
+ return FALSE;
+
+ nanozone->band_max_mapped_baseaddr[mag_index] = vm_addr;
+ }
// Randomize the starting allocation from this slot (introduces 11 to 14 bits of entropy)
if (0 == pMeta->slot_objects_mapped) { // First encounter?
@@ -136,9 +410,9 @@
s = u.addr; // Base for this core.
// Set the high water mark for this CPU's entire magazine, if this resupply raised it.
- watermark = nanozone->core_mapped_size[mag_index];
- hiwater = MAX(watermark, p - s + SLOT_IN_BAND_SIZE);
- nanozone->core_mapped_size[mag_index] = hiwater;
+ watermark = nanozone->core_mapped_size[mag_index];
+ hiwater = MAX( watermark, p - s + SLOT_IN_BAND_SIZE );
+ nanozone->core_mapped_size[mag_index] = hiwater;
return TRUE;
}
@@ -152,20 +426,19 @@
return a / b;
}
-static MALLOC_INLINE void *
-segregated_next_block(nanozone_t *nanozone, nano_meta_admin_t pMeta, size_t slot_bytes, unsigned int mag_index)
+static INLINE void *
+segregated_next_block(nanozone_t *nanozone, nano_meta_admin_t pMeta, unsigned int slot_bytes, unsigned int mag_index)
{
while (1) {
uintptr_t theLimit = pMeta->slot_limit_addr; // Capture the slot limit that bounds slot_bump_addr right now
uintptr_t b = OSAtomicAdd64Barrier(slot_bytes, (volatile int64_t *)&(pMeta->slot_bump_addr));
b -= slot_bytes; // Atomic op returned addr of *next* free block. Subtract to get addr for *this* allocation.
- if (b < theLimit) { // Did we stay within the bound of the present slot allocation?
+ if (b < theLimit) { // Did we stay within the bound of the present slot allocation?
return (void *)b; // Yep, so the slot_bump_addr this thread incremented is good to go
} else {
if (pMeta->slot_exhausted) { // exhausted all the bands availble for this slot?
- pMeta->slot_bump_addr = theLimit;
- return 0; // We're toast
+ return 0; // We're toast
} else {
// One thread will grow the heap, others will see its been grown and retry allocation
_malloc_lock_lock(&nanozone->band_resupply_lock[mag_index]);
@@ -181,7 +454,6 @@
continue; // ... the slot has been successfully grown by us. Now try again.
} else {
pMeta->slot_exhausted = TRUE;
- pMeta->slot_bump_addr = theLimit;
_malloc_lock_unlock(&nanozone->band_resupply_lock[mag_index]);
return 0;
}
@@ -190,62 +462,58 @@
}
}
-static MALLOC_INLINE size_t
-segregated_size_to_fit(nanozone_t *nanozone, size_t size, size_t *pKey)
-{
- size_t k, slot_bytes;
-
- if (0 == size) {
+static INLINE unsigned int
+segregated_size_to_fit(nanozone_t *nanozone, size_t size, unsigned int *pKey)
+{
+ unsigned int k, slot_bytes;
+
+ if (0 == size)
size = NANO_REGIME_QUANTA_SIZE; // Historical behavior
- }
+
k = (size + NANO_REGIME_QUANTA_SIZE - 1) >> SHIFT_NANO_QUANTUM; // round up and shift for number of quanta
- slot_bytes = k << SHIFT_NANO_QUANTUM; // multiply by power of two quanta size
- *pKey = k - 1; // Zero-based!
+ slot_bytes = k << SHIFT_NANO_QUANTUM; // multiply by power of two quanta size
+ *pKey = k - 1; // Zero-based!
return slot_bytes;
}
-static MALLOC_INLINE index_t
+static INLINE index_t
offset_to_index(nanozone_t *nanozone, nano_meta_admin_t pMeta, uintptr_t offset)
{
unsigned int slot_bytes = pMeta->slot_bytes;
unsigned int slot_objects = pMeta->slot_objects; // SLOT_IN_BAND_SIZE / slot_bytes;
- unsigned int rem;
- unsigned long quo = divrem(offset, BAND_SIZE, &rem);
-
- assert(0 == rem % slot_bytes || pMeta->slot_exhausted);
- return (index_t)((quo * slot_objects) + (rem / slot_bytes));
-}
-
-static MALLOC_INLINE uintptr_t
+ unsigned int rem;
+ unsigned long quo = divrem(offset, BAND_SIZE, &rem);
+
+ assert(0 == rem%slot_bytes);
+ return (quo * slot_objects) + (rem / slot_bytes);
+}
+
+static INLINE uintptr_t
index_to_offset(nanozone_t *nanozone, nano_meta_admin_t pMeta, index_t i)
{
unsigned int slot_bytes = pMeta->slot_bytes;
unsigned int slot_objects = pMeta->slot_objects; // SLOT_IN_BAND_SIZE / slot_bytes;
- unsigned int rem;
- unsigned long quo = divrem(i, slot_objects, &rem);
-
+ unsigned int rem;
+ unsigned long quo = divrem(i, slot_objects, &rem);
+
return (quo * BAND_SIZE) + (rem * slot_bytes);
}
static kern_return_t
-segregated_in_use_enumerator(task_t task,
- void *context,
- unsigned type_mask,
- nanozone_t *nanozone,
- memory_reader_t reader,
- vm_range_recorder_t recorder)
-{
- unsigned int mag_index, slot_key;
- vm_range_t ptr_range;
- vm_range_t buffer[MAX_RECORDER_BUFFER];
- kern_return_t err;
- unsigned count = 0;
-
- for (mag_index = 0; mag_index < nano_common_max_magazines; mag_index++) {
- uintptr_t clone_magazine; // magazine base for ourselves
- nano_blk_addr_t p; // slot base for remote
- uintptr_t clone_slot_base; // slot base for ourselves (tracks with "p")
+segregated_in_use_enumerator(task_t task, void *context, unsigned type_mask, nanozone_t *nanozone,
+ memory_reader_t reader, vm_range_recorder_t recorder)
+{
+ unsigned int mag_index, slot_key;
+ vm_range_t ptr_range;
+ vm_range_t buffer[MAX_RECORDER_BUFFER];
+ kern_return_t err;
+ unsigned count = 0;
+
+ for (mag_index = 0; mag_index < nanozone->phys_ncpus; mag_index++) {
+ uintptr_t clone_magazine; // magazine base for ourselves
+ nano_blk_addr_t p; // slot base for remote
+ uintptr_t clone_slot_base; // slot base for ourselves (tracks with "p")
// Establish p as base address for slot 0 in remote
p.fields.nano_signature = NANOZONE_SIGNATURE;
@@ -270,15 +538,15 @@
clone_slot_base = clone_magazine = 0; // and won't be used in this loop
}
- for (slot_key = 0; slot_key < SLOT_KEY_LIMIT; p.addr += SLOT_IN_BAND_SIZE, // Advance to next slot base for remote
- clone_slot_base += SLOT_IN_BAND_SIZE, // Advance to next slot base for ourselves
- slot_key++) {
+ for (slot_key = 0; slot_key < SLOT_KEY_LIMIT;
+ p.addr += SLOT_IN_BAND_SIZE, // Advance to next slot base for remote
+ clone_slot_base += SLOT_IN_BAND_SIZE, // Advance to next slot base for ourselves
+ slot_key++) {
nano_meta_admin_t pMeta = &(nanozone->meta_data[mag_index][slot_key]);
size_t slot_objects_mapped = pMeta->slot_objects_mapped; // capture this volatile count
- if (0 == slot_objects_mapped) { // Nothing allocated in this magazine for this slot?
+ if (0 == slot_objects_mapped) // Nothing allocated in this magazine for this slot?
continue;
- }
if (type_mask & MALLOC_ADMIN_REGION_RANGE_TYPE) {
/* do NOTHING as there is no distinct admin region */
@@ -286,7 +554,7 @@
if (type_mask & (MALLOC_PTR_REGION_RANGE_TYPE | MALLOC_ADMIN_REGION_RANGE_TYPE)) {
nano_blk_addr_t q = p;
- uintptr_t skip_adj = index_to_offset(nanozone, pMeta, (index_t)pMeta->slot_objects_skipped);
+ uintptr_t skip_adj = index_to_offset(nanozone, pMeta, pMeta->slot_objects_skipped);
while (q.addr < pMeta->slot_limit_addr) {
ptr_range.address = q.addr + skip_adj;
@@ -300,82 +568,17 @@
if (type_mask & MALLOC_PTR_IN_USE_RANGE_TYPE) {
nano_blk_addr_t q = p;
uintptr_t slot_band, clone_slot_band_base = clone_slot_base;
- uintptr_t skip_adj = index_to_offset(nanozone, pMeta, (index_t)pMeta->slot_objects_skipped);
-
- // Copy the bitarray_t denoting madvise()'d pages (if any) into *this* task's address space
- bitarray_t madv_page_bitarray;
- int log_page_count;
-
- if (pMeta->slot_madvised_pages) {
- log_page_count = pMeta->slot_madvised_log_page_count;
- err = reader(task, (vm_address_t)(pMeta->slot_madvised_pages), bitarray_size(log_page_count),
- (void **)&madv_page_bitarray);
- if (err) {
+ uintptr_t skip_adj = index_to_offset(nanozone, pMeta, pMeta->slot_objects_skipped);
+
+ while (q.addr < pMeta->slot_limit_addr) {
+ // read slot in each remote band. Lands in some random location.
+ size_t len = MIN(pMeta->slot_bump_addr - q.addr, SLOT_IN_BAND_SIZE);
+ err = reader(task, (vm_address_t)(q.addr + skip_adj), len - skip_adj, (void **)&slot_band);
+ if (err)
return err;
- }
- } else {
- madv_page_bitarray = NULL;
- log_page_count = 0;
- }
-
- while (q.addr < pMeta->slot_limit_addr) {
- // read slot in each remote band. Lands in some random location. Do not read
- // parts of the slot that are in madvised pages.
- if (!madv_page_bitarray) {
- // Nothing madvised yet - read everything in one go.
- size_t len = MIN(pMeta->slot_bump_addr - q.addr, SLOT_IN_BAND_SIZE) - skip_adj;
- err = reader(task, (vm_address_t)(q.addr + skip_adj), len, (void **)&slot_band);
- if (err) {
- return err;
- }
-
- // Place the data just read in the correct position relative to the local magazine.
- memcpy((void *)(clone_slot_band_base + skip_adj), (void *)slot_band, len);
- } else {
- // We madvised at least one page. Read only the pages that
- // have not been madvised. If bitarray_t had operations
- // like "get next bit set after a given bit" and "find
- // next unset bit after a given bit", we could do this more
- // efficiently but given that it doesn't, we have to walk
- // through each page individually. In practice this is not
- // much of an issue because this code is only used by
- // sampling tools and the additional time required is not
- // really noticeable.
- size_t len = MIN(pMeta->slot_bump_addr - q.addr, SLOT_IN_BAND_SIZE) - skip_adj;
- vm_address_t start_addr = (vm_address_t)(q.addr + skip_adj);
- vm_address_t end_addr = (vm_address_t)(start_addr + len);
- void *target_addr = (void *)(clone_slot_band_base + skip_adj);
- for (vm_address_t addr = start_addr; addr < end_addr;) {
- vm_address_t next_page_addr = trunc_page_kernel(addr + vm_kernel_page_size);
- size_t read_size = MIN(len, next_page_addr - addr);
-
- boolean_t madvised = false;
- nano_blk_addr_t r;
- r.addr = addr;
- index_t pgnum = ((((unsigned)r.fields.nano_band) << NANO_OFFSET_BITS) | ((unsigned)r.fields.nano_offset)) >>
- vm_kernel_page_shift;
- unsigned int log_page_count = pMeta->slot_madvised_log_page_count;
- madvised = (pgnum < (1 << log_page_count)) &&
- bitarray_get(madv_page_bitarray, log_page_count, pgnum);
- if (!madvised) {
- // This is not an madvised page - grab the data.
- err = reader(task, addr, read_size, (void **)&slot_band);
- if (err) {
- return err;
- }
-
- // Place the data just read in the correct position relative to the local magazine.
- memcpy(target_addr, (void *)slot_band, read_size);
- } else {
- // This is an madvised page - there should be nothing in here that's
- // on the freelist, so just write garbage to the target memory.
- memset(target_addr, (char)0xee, read_size);
- }
- addr = next_page_addr;
- target_addr += read_size;
- len -= read_size;
- }
- }
+
+ // Place the data just read in the correct position relative to the local magazine.
+ memcpy((void *)(clone_slot_band_base + skip_adj), (void *)slot_band, len - skip_adj);
// Simultaneously advance pointers in remote and ourselves to the next band.
q.addr += BAND_SIZE;
@@ -387,58 +590,65 @@
int log_size = 64 - __builtin_clzl(slot_objects_mapped);
bitarray_t slot_bitarray = bitarray_create(log_size);
- if (!slot_bitarray) {
+ if (!slot_bitarray)
return errno;
- }
chained_block_t t;
- unsigned stoploss = (unsigned)slot_objects_mapped;
- while ((t = OSAtomicDequeue(
- &(pMeta->slot_LIFO), offsetof(struct chained_block_s, next) + (clone_slot_base - p.addr)))) {
+ unsigned stoploss = slot_objects_mapped;
+ while ((t = OSAtomicDequeue( &(pMeta->slot_LIFO), offsetof(struct chained_block_s,next) + (clone_slot_base - p.addr)))) {
if (0 == stoploss) {
- malloc_report(ASL_LEVEL_ERR, "Free list walk in segregated_in_use_enumerator exceeded object count.\n");
+ malloc_printf("Free list walk in segregated_in_use_enumerator exceeded object count.");
break;
}
stoploss--;
- uintptr_t offset = ((uintptr_t)t - p.addr); // offset from beginning of slot, task-independent
- index_t block_index = offset_to_index(nanozone, pMeta, offset);
-
- if (block_index < slot_objects_mapped) {
+ uintptr_t offset = ((uintptr_t)t - p.addr); // offset from beginning of slot, task-independent
+ index_t block_index = offset_to_index(nanozone, pMeta, offset);
+
+ if (block_index < slot_objects_mapped)
bitarray_set(slot_bitarray, log_size, block_index);
- }
}
// N.B. pMeta->slot_LIFO in *this* task is now drained (remote free list has *not* been disturbed)
-
+ // Copy the bitarray_t denoting madvise()'d pages (if any) into *this* task's address space
+ bitarray_t madv_page_bitarray;
+ int log_page_count;
+
+ if (pMeta->slot_madvised_pages) {
+ log_page_count = pMeta->slot_madvised_log_page_count;
+ err = reader(task, (vm_address_t)(pMeta->slot_madvised_pages), bitarray_size(log_page_count), (void **)&madv_page_bitarray);
+ if (err)
+ return err;
+ } else {
+ madv_page_bitarray = NULL;
+ log_page_count = 0;
+ }
+
// Enumerate all the block indices issued to date, and report those not on the free list
index_t i;
- for (i = (index_t)pMeta->slot_objects_skipped; i < slot_objects_mapped; ++i) {
+ for (i = pMeta->slot_objects_skipped; i < slot_objects_mapped; ++i) {
uintptr_t block_offset = index_to_offset(nanozone, pMeta, i);
- if (p.addr + block_offset >= pMeta->slot_bump_addr) {
+ if (p.addr + block_offset >= pMeta->slot_bump_addr)
break;
- }
// blocks falling on madvise()'d pages are free! So not enumerated.
if (madv_page_bitarray) {
nano_blk_addr_t q;
index_t pgnum, pgnum_end;
-
+
q.addr = p.addr + block_offset;
- pgnum = ((((unsigned)q.fields.nano_band) << NANO_OFFSET_BITS) | ((unsigned)q.fields.nano_offset)) >>
- vm_kernel_page_shift;
+ pgnum = ((((unsigned)q.fields.nano_band) << NANO_OFFSET_BITS) | ((unsigned)q.fields.nano_offset)) >> vm_page_shift;
q.addr += pMeta->slot_bytes - 1;
- pgnum_end = ((((unsigned)q.fields.nano_band) << NANO_OFFSET_BITS) | ((unsigned)q.fields.nano_offset)) >>
- vm_kernel_page_shift;
-
- if (pgnum < (1 << log_page_count)) { // bounds check for bitarray_get()'s that follow
+ pgnum_end = ((((unsigned)q.fields.nano_band) << NANO_OFFSET_BITS) | ((unsigned)q.fields.nano_offset)) >> vm_page_shift;
+
+ if (pgnum < (1 << log_page_count)) {// bounds check for bitarray_get()'s that follow
if (bitarray_get(madv_page_bitarray, log_page_count, pgnum) ||
- bitarray_get(madv_page_bitarray, log_page_count, pgnum_end)) {
+ bitarray_get(madv_page_bitarray, log_page_count, pgnum_end)) {
continue;
}
}
}
-
+
if (!bitarray_get(slot_bitarray, log_size, i)) {
buffer[count].address = p.addr + block_offset;
buffer[count].size = (slot_key + 1) << SHIFT_NANO_QUANTUM;
@@ -473,7 +683,153 @@
* the extent of the nano allocation regime -- (0, 256].
*/
-static MALLOC_INLINE MALLOC_UNUSED boolean_t
+static INLINE size_t
+__nano_vet_and_size(nanozone_t *nanozone, const void *ptr)
+{
+ // Extracts the size of the block in bytes. Checks for a plausible ptr.
+ nano_blk_addr_t p; // the compiler holds this in a register
+ nano_meta_admin_t pMeta;
+
+ p.addr = (uint64_t)ptr; // Begin the dissection of ptr
+
+ if (nanozone->phys_ncpus <= p.fields.nano_mag_index)
+ return 0;
+
+ if (p.fields.nano_offset & NANO_QUANTA_MASK) // stray low-order bits?
+ return 0;
+
+ pMeta = &(nanozone->meta_data[p.fields.nano_mag_index][p.fields.nano_slot]);
+ if ((void *)(pMeta->slot_bump_addr) <= ptr)
+ return 0; // Beyond what's ever been allocated!
+
+ if ((p.fields.nano_offset % pMeta->slot_bytes) != 0)
+ return 0; // Not an exact multiple of the block size for this slot
+
+ return pMeta->slot_bytes;
+}
+
+static INLINE size_t
+_nano_vet_and_size_of_live(nanozone_t *nanozone, const void *ptr)
+{
+ size_t size = __nano_vet_and_size(nanozone, ptr);
+ if (size && ((((chained_block_t)ptr)->double_free_guard ^ nanozone->cookie) != 0xBADDC0DEDEADBEADULL))
+ return size; // Common case: not on a free list, hence live. Return its size.
+ else
+ // ptr is either on a free list (its got the correct canary) in which case return zero, OR
+ // the caller has stored the canary value in the double_free_guard slot entirely by coincidence
+ // and the block is a live allocation. The latter is very unlikely (1 in 2^64) so just return 0.
+ return 0;
+}
+
+static INLINE size_t
+_nano_vet_and_size_of_free(nanozone_t *nanozone, const void *ptr)
+{
+ size_t size = __nano_vet_and_size(nanozone, ptr);
+ if (size && ((((chained_block_t)ptr)->double_free_guard ^ nanozone->cookie) == 0xBADDC0DEDEADBEADULL))
+ return size;
+ else
+ return 0;
+}
+
+static void *
+_nano_malloc_check_clear(nanozone_t *nanozone, size_t size, boolean_t cleared_requested)
+{
+ void *ptr;
+ unsigned int slot_key;
+ unsigned int slot_bytes = segregated_size_to_fit(nanozone, size, &slot_key); // Note slot_key is set here
+ unsigned int mag_index = NANO_MAG_INDEX(nanozone);
+
+ nano_meta_admin_t pMeta = &(nanozone->meta_data[mag_index][slot_key]);
+
+ ptr = OSAtomicDequeue( &(pMeta->slot_LIFO), offsetof(struct chained_block_s,next));
+ if (ptr) {
+#if NANO_FREE_DEQUEUE_DILIGENCE
+ size_t gotSize;
+ nano_blk_addr_t p; // the compiler holds this in a register
+
+ p.addr = (uint64_t)ptr; // Begin the dissection of ptr
+ if (nanozone->our_signature != p.fields.nano_signature) {
+ nanozone_error(nanozone, 1,
+ "Invalid signature for pointer dequeued from free list", ptr, NULL);
+ }
+
+ if (mag_index != p.fields.nano_mag_index) {
+ nanozone_error(nanozone, 1,
+ "Mismatched magazine for pointer dequeued from free list", ptr, NULL);
+ }
+
+ gotSize = _nano_vet_and_size_of_free(nanozone, ptr);
+ if (0 == gotSize) {
+ nanozone_error(nanozone, 1,
+ "Invalid pointer dequeued from free list", ptr, NULL);
+ }
+ if (gotSize != slot_bytes) {
+ nanozone_error(nanozone, 1,
+ "Mismatched size for pointer dequeued from free list", ptr, NULL);
+ }
+
+ if ((((chained_block_t)ptr)->double_free_guard ^ nanozone->cookie) != 0xBADDC0DEDEADBEADULL) {
+ nanozone_error(nanozone, 1,
+ "Heap corruption detected, free list canary is damaged", ptr, NULL);
+ }
+#if defined(DEBUG)
+ void *next = (void *) (((chained_block_t)ptr)->next);
+ if (next) {
+ p.addr = (uint64_t)next; // Begin the dissection of next
+ if (nanozone->our_signature != p.fields.nano_signature) {
+ nanozone_error(nanozone, 1,
+ "Invalid next signature for pointer dequeued from free list (showing ptr, next)",
+ ptr, ", %p", next);
+ }
+
+ if (mag_index != p.fields.nano_mag_index) {
+ nanozone_error(nanozone, 1,
+ "Mismatched next magazine for pointer dequeued from free list (showing ptr, next)",
+ ptr, ", %p", next);
+ }
+
+ gotSize = _nano_vet_and_size_of_free(nanozone, next);
+ if (0 == gotSize) {
+ nanozone_error(nanozone, 1,
+ "Invalid next for pointer dequeued from free list (showing ptr, next)",
+ ptr, ", %p", next);
+ }
+ if (gotSize != slot_bytes) {
+ nanozone_error(nanozone, 1,
+ "Mismatched next size for pointer dequeued from free list (showing ptr, next)",
+ ptr, ", %p", next);
+ }
+ }
+#endif /* DEBUG */
+#endif /* NANO_FREE_DEQUEUE_DILIGENCE */
+
+ ((chained_block_t)ptr)->double_free_guard = 0;
+ ((chained_block_t)ptr)->next = NULL; // clear out next pointer to protect free list
+ } else {
+ ptr = segregated_next_block(nanozone, pMeta, slot_bytes, mag_index);
+ }
+
+ if (cleared_requested && ptr)
+ memset(ptr, 0, slot_bytes); // TODO: Needs a memory barrier after memset to ensure zeroes land first?
+
+ return ptr;
+}
+
+static void *
+_nano_malloc_check_scribble(nanozone_t *nanozone, size_t size)
+{
+ void *ptr = _nano_malloc_check_clear(nanozone, size, 0);
+
+ /*
+ * Scribble on allocated memory when requested.
+ */
+ if ((nanozone->debug_flags & SCALABLE_MALLOC_DO_SCRIBBLE) && ptr && size)
+ memset(ptr, SCRIBBLE_BYTE, _nano_vet_and_size_of_live(nanozone, ptr));
+
+ return ptr;
+}
+
+static INLINE boolean_t
_nano_block_inuse_p(nanozone_t *nanozone, const void *ptr)
{
nano_blk_addr_t p; // happily, the compiler holds this in a register
@@ -485,21 +841,22 @@
pMeta = &(nanozone->meta_data[p.fields.nano_mag_index][p.fields.nano_slot]);
+ if ((void *)(pMeta->slot_bump_addr) <= ptr)
+ return FALSE; // Beyond what's ever been allocated, so trivially not in use.
+
// pop elements off the free list all the while looking for ptr.
- unsigned stoploss = (unsigned)pMeta->slot_objects_mapped;
- while ((t = OSAtomicDequeue(&(pMeta->slot_LIFO), offsetof(struct chained_block_s, next)))) {
+ unsigned stoploss = pMeta->slot_objects_mapped;
+ while ((t = OSAtomicDequeue( &(pMeta->slot_LIFO), offsetof(struct chained_block_s,next)))) {
if (0 == stoploss) {
- malloc_zone_error(nanozone->debug_flags, true,
- "Free list walk for slot %p in _nano_block_inuse_p exceeded object count.\n",
- (void *)&(pMeta->slot_LIFO));
+ nanozone_error(nanozone, 1, "Free list walk in _nano_block_inuse_p exceeded object count.",
+ (void *)&(pMeta->slot_LIFO), NULL);
}
stoploss--;
- if (NULL == head) {
+ if (NULL == head)
head = t;
- } else {
+ else
tail->next = t;
- }
tail = t;
if (ptr == t) {
@@ -507,277 +864,71 @@
break;
}
}
- if (tail) {
+ if (tail)
tail->next = NULL;
- }
// push the free list extracted above back onto the LIFO, all at once
- if (head) {
- OSAtomicEnqueue(&(pMeta->slot_LIFO), head, (uintptr_t)tail - (uintptr_t)head + offsetof(struct chained_block_s, next));
- }
+ if (head)
+ OSAtomicEnqueue( &(pMeta->slot_LIFO), head, (uintptr_t)tail - (uintptr_t)head + offsetof(struct chained_block_s,next));
return inuse;
}
-static MALLOC_INLINE size_t
-__nano_vet_and_size_inner(nanozone_t *nanozone, const void *ptr, boolean_t inner)
-{
- // Extracts the size of the block in bytes. Checks for a plausible ptr.
- nano_blk_addr_t p; // the compiler holds this in a register
- nano_meta_admin_t pMeta;
-
- p.addr = (uint64_t)ptr; // Begin the dissection of ptr
-
- if (NANOZONE_SIGNATURE != p.fields.nano_signature) {
- return 0;
- }
-
- if (nano_common_max_magazines <= p.fields.nano_mag_index) {
- return 0;
- }
-
- if (!inner && p.fields.nano_offset & NANO_QUANTA_MASK) { // stray low-order bits?
- return 0;
- }
-
- pMeta = &(nanozone->meta_data[p.fields.nano_mag_index][p.fields.nano_slot]);
- if ((void *)(pMeta->slot_bump_addr) <= ptr) {
- return 0; // Beyond what's ever been allocated!
- }
- if (!inner && ((p.fields.nano_offset % pMeta->slot_bytes) != 0)) {
- return 0; // Not an exact multiple of the block size for this slot
- }
- return pMeta->slot_bytes;
-}
-
-
-static MALLOC_INLINE size_t
-__nano_vet_and_size(nanozone_t *nanozone, const void *ptr)
-{
- return __nano_vet_and_size_inner(nanozone, ptr, false);
-}
-
-static MALLOC_ALWAYS_INLINE boolean_t
-_nano_block_has_canary_value(nanozone_t *nanozone, const void *ptr)
-{
- return (((chained_block_t)ptr)->double_free_guard ^ nanozone->cookie)
- == (uintptr_t)ptr;
-}
-
-static MALLOC_ALWAYS_INLINE void
-_nano_block_set_canary_value(nanozone_t *nanozone, const void *ptr)
-{
- ((chained_block_t)ptr)->double_free_guard =
- ((uintptr_t)ptr) ^ nanozone->cookie;
-}
-
-static MALLOC_INLINE size_t
-_nano_vet_and_size_of_live(nanozone_t *nanozone, const void *ptr)
-{
- size_t size = __nano_vet_and_size(nanozone, ptr);
-
- if (0 == size) { // ptr fails sanity check?
- return 0;
- }
-
- // We have the invariant: If ptr is on a free list, then ptr->double_free_guard is the canary.
- // So if ptr->double_free_guard is NOT the canary, then ptr is not on a free list, hence is live.
- if (!_nano_block_has_canary_value(nanozone, ptr)) {
- return size; // Common case: not on a free list, hence live. Return its size.
- } else {
- // confirm that ptr is live despite ptr->double_free_guard having the canary value
- if (_nano_block_inuse_p(nanozone, ptr)) {
- return size; // live block that exhibits canary
- } else {
- return 0; // ptr wasn't live after all (likely a double free)
- }
- }
-}
-
-static MALLOC_INLINE size_t
-_nano_vet_and_size_of_free(nanozone_t *nanozone, const void *ptr)
-{
- size_t size = __nano_vet_and_size(nanozone, ptr);
-
- if (0 == size) { // ptr fails sanity check?
- return 0;
- }
-
- // ptr was just dequed from a free list, so ptr->double_free_guard must have the canary value.
- if (_nano_block_has_canary_value(nanozone, ptr)) {
- return size; // return the size of this well formed free block.
- } else {
- return 0; // Broken invariant: If ptr is on a free list, then ptr->double_free_guard is the canary. (likely use after free)
- }
-}
-
-static void *
-_nano_malloc_check_clear(nanozone_t *nanozone, size_t size, boolean_t cleared_requested)
-{
- MALLOC_TRACE(TRACE_nano_malloc, (uintptr_t)nanozone, size, cleared_requested, 0);
-
- void *ptr;
- size_t slot_key;
- size_t slot_bytes = segregated_size_to_fit(nanozone, size, &slot_key); // Note slot_key is set here
- mag_index_t mag_index = nano_mag_index(nanozone);
-
- nano_meta_admin_t pMeta = &(nanozone->meta_data[mag_index][slot_key]);
-
- ptr = OSAtomicDequeue(&(pMeta->slot_LIFO), offsetof(struct chained_block_s, next));
- if (ptr) {
- unsigned debug_flags = nanozone->debug_flags;
-#if NANO_FREE_DEQUEUE_DILIGENCE
- size_t gotSize;
- nano_blk_addr_t p; // the compiler holds this in a register
-
- p.addr = (uint64_t)ptr; // Begin the dissection of ptr
- if (NANOZONE_SIGNATURE != p.fields.nano_signature) {
- malloc_zone_error(debug_flags, true,
- "Invalid signature for pointer %p dequeued from free list\n",
- ptr);
- }
-
- if (mag_index != p.fields.nano_mag_index) {
- malloc_zone_error(debug_flags, true,
- "Mismatched magazine for pointer %p dequeued from free list\n",
- ptr);
- }
-
- gotSize = _nano_vet_and_size_of_free(nanozone, ptr);
- if (0 == gotSize) {
- malloc_zone_error(debug_flags, true,
- "Invalid pointer %p dequeued from free list\n", ptr);
- }
- if (gotSize != slot_bytes) {
- malloc_zone_error(debug_flags, true,
- "Mismatched size for pointer %p dequeued from free list\n",
- ptr);
- }
-
- if (!_nano_block_has_canary_value(nanozone, ptr)) {
- malloc_zone_error(debug_flags, true,
- "Heap corruption detected, free list canary is damaged for %p\n"
- "*** Incorrect guard value: %lu\n", ptr,
- ((chained_block_t)ptr)->double_free_guard);
- }
-
-#if defined(DEBUG)
- void *next = (void *)(((chained_block_t)ptr)->next);
- if (next) {
- p.addr = (uint64_t)next; // Begin the dissection of next
- if (NANOZONE_SIGNATURE != p.fields.nano_signature) {
- malloc_zone_error(debug_flags, true,
- "Invalid next signature for pointer %p dequeued from free "
- "list, next = %p\n", ptr, "next");
- }
-
- if (mag_index != p.fields.nano_mag_index) {
- malloc_zone_error(debug_flags, true,
- "Mismatched next magazine for pointer %p dequeued from "
- "free list, next = %p\n", ptr, next);
- }
-
- gotSize = _nano_vet_and_size_of_free(nanozone, next);
- if (0 == gotSize) {
- malloc_zone_error(debug_flags, true,
- "Invalid next for pointer %p dequeued from free list, "
- "next = %p\n", ptr, next);
- }
- if (gotSize != slot_bytes) {
- malloc_zone_error(debug_flags, true,
- "Mismatched next size for pointer %p dequeued from free "
- "list, next = %p\n", ptr, next);
- }
- }
-#endif /* DEBUG */
-#endif /* NANO_FREE_DEQUEUE_DILIGENCE */
-
- ((chained_block_t)ptr)->double_free_guard = 0;
- ((chained_block_t)ptr)->next = NULL; // clear out next pointer to protect free list
- } else {
- ptr = segregated_next_block(nanozone, pMeta, slot_bytes, mag_index);
- }
-
- if (cleared_requested && ptr) {
- memset(ptr, 0, slot_bytes); // TODO: Needs a memory barrier after memset to ensure zeroes land first?
- }
- return ptr;
-}
-
-static void *
-_nano_malloc_check_scribble(nanozone_t *nanozone, size_t size)
-{
- void *ptr = _nano_malloc_check_clear(nanozone, size, 0);
-
- /*
- * Scribble on allocated memory when requested.
- */
- if ((nanozone->debug_flags & MALLOC_DO_SCRIBBLE) && ptr && size) {
- memset(ptr, SCRIBBLE_BYTE, _nano_vet_and_size_of_live(nanozone, ptr));
- }
-
- return ptr;
-}
-
-static MALLOC_INLINE size_t
+static INLINE size_t
_nano_size(nanozone_t *nanozone, const void *ptr)
{
- return _nano_vet_and_size_of_live(nanozone, ptr);
-}
-
-static MALLOC_INLINE size_t
+ return _nano_vet_and_size_of_live(nanozone, ptr);
+}
+
+static INLINE size_t
_nano_good_size(nanozone_t *nanozone, size_t size)
{
- return (size <= NANO_REGIME_QUANTA_SIZE) ? NANO_REGIME_QUANTA_SIZE
- : (((size + NANO_REGIME_QUANTA_SIZE - 1) >> SHIFT_NANO_QUANTUM) << SHIFT_NANO_QUANTUM);
-}
-
-static MALLOC_INLINE void _nano_free_trusted_size_check_scribble(nanozone_t *nanozone,
- void *ptr,
- size_t trusted_size,
- boolean_t do_scribble) MALLOC_ALWAYS_INLINE;
-
-static MALLOC_INLINE void
+ return (size <= NANO_REGIME_QUANTA_SIZE) ?
+ NANO_REGIME_QUANTA_SIZE :
+ (((size + NANO_REGIME_QUANTA_SIZE - 1) >> SHIFT_NANO_QUANTUM) << SHIFT_NANO_QUANTUM);
+}
+
+static INLINE void _nano_free_trusted_size_check_scribble(nanozone_t *nanozone, void *ptr, size_t trusted_size, boolean_t do_scribble) ALWAYSINLINE;
+
+static INLINE void
_nano_free_trusted_size_check_scribble(nanozone_t *nanozone, void *ptr, size_t trusted_size, boolean_t do_scribble)
{
if (trusted_size) {
nano_blk_addr_t p; // happily, the compiler holds this in a register
nano_meta_admin_t pMeta;
- if (do_scribble) {
+ if (do_scribble)
(void)memset(ptr, SCRABBLE_BYTE, trusted_size);
- }
- _nano_block_set_canary_value(nanozone, ptr);
+ ((chained_block_t)ptr)->double_free_guard = (0xBADDC0DEDEADBEADULL ^ nanozone->cookie);
p.addr = (uint64_t)ptr; // place ptr on the dissecting table
pMeta = &(nanozone->meta_data[p.fields.nano_mag_index][p.fields.nano_slot]);
- OSAtomicEnqueue(&(pMeta->slot_LIFO), ptr, offsetof(struct chained_block_s, next));
+ OSAtomicEnqueue( &(pMeta->slot_LIFO), ptr, offsetof(struct chained_block_s,next));
} else {
- malloc_zone_error(nanozone->debug_flags, true,
- "Freeing unallocated pointer %p\n", ptr);
- }
-}
-
-static MALLOC_INLINE void _nano_free_check_scribble(nanozone_t *nanozone, void *ptr, boolean_t do_scribble) MALLOC_ALWAYS_INLINE;
-
-static MALLOC_INLINE void
+ nanozone_error(nanozone, 1, "Freeing unallocated pointer", ptr, NULL);
+ }
+}
+
+static INLINE void _nano_free_check_scribble(nanozone_t *nanozone, void *ptr, boolean_t do_scribble) ALWAYSINLINE;
+
+static INLINE void
_nano_free_check_scribble(nanozone_t *nanozone, void *ptr, boolean_t do_scribble)
{
- _nano_free_trusted_size_check_scribble(nanozone, ptr, _nano_vet_and_size_of_live(nanozone, ptr), do_scribble);
-}
-
-static MALLOC_INLINE void *
+ _nano_free_trusted_size_check_scribble(nanozone, ptr, _nano_vet_and_size_of_live(nanozone, ptr), do_scribble);
+}
+
+static INLINE void *
_nano_realloc(nanozone_t *nanozone, void *ptr, size_t new_size)
{
- size_t old_size, new_good_size, valid_size;
- void *new_ptr;
+ size_t old_size, new_good_size, valid_size;
+ void *new_ptr;
if (FALSE && NULL == ptr) { // ptr has our_signature so can't be NULL, but if it were Posix sez ...
// If ptr is a null pointer, realloc() shall be equivalent to malloc() for the specified size.
return _nano_malloc_check_scribble(nanozone, new_size);
} else if (0 == new_size) {
// If size is 0 and ptr is not a null pointer, the object pointed to is freed.
- _nano_free_check_scribble(nanozone, ptr, (nanozone->debug_flags & MALLOC_DO_SCRIBBLE));
+ _nano_free_check_scribble(nanozone, ptr, (nanozone->debug_flags & SCALABLE_MALLOC_DO_SCRIBBLE));
// If size is 0, either a null pointer or a unique pointer that can be successfully passed
// to free() shall be returned.
return _nano_malloc_check_scribble(nanozone, 1);
@@ -785,8 +936,7 @@
old_size = _nano_vet_and_size_of_live(nanozone, ptr);
if (!old_size) {
- malloc_zone_error(nanozone->debug_flags, true,
- "pointer %p being reallocated was not allocated\n", ptr);
+ nanozone_error(nanozone, 1, "pointer being reallocated was not allocated", ptr, NULL);
return NULL;
}
@@ -797,9 +947,8 @@
/* Serious shrinkage (more than half). FALL THROUGH to alloc/copy/free. */
} else {
/* Let's hang on to what we got. */
- if (nanozone->debug_flags & MALLOC_DO_SCRIBBLE) {
+ if (nanozone->debug_flags & SCALABLE_MALLOC_DO_SCRIBBLE)
memset(ptr + new_size, SCRIBBLE_BYTE, old_size - new_size);
- }
return ptr;
}
@@ -807,23 +956,21 @@
* Allocate a new buffer and copy.
*/
new_ptr = _nano_malloc_check_scribble(nanozone, new_good_size);
- if (new_ptr == NULL) {
+ if (new_ptr == NULL)
return NULL;
- }
valid_size = MIN(old_size, new_good_size);
memcpy(new_ptr, ptr, valid_size);
- _nano_free_check_scribble(nanozone, ptr, (nanozone->debug_flags & MALLOC_DO_SCRIBBLE));
+ _nano_free_check_scribble(nanozone, ptr, (nanozone->debug_flags & SCALABLE_MALLOC_DO_SCRIBBLE));
return new_ptr;
}
-static MALLOC_INLINE void
+static INLINE void
_nano_destroy(nanozone_t *nanozone)
{
/* Now destroy the separate nanozone region */
- nano_common_deallocate_pages((void *)nanozone, NANOZONE_PAGED_SIZE,
- nanozone->debug_flags);
+ deallocate_pages(nanozone, (void *)nanozone, SZONE_PAGED_SIZE, 0);
}
/****************** nanozone dispatch **********************/
@@ -860,9 +1007,8 @@
/*
* Scribble on allocated memory.
*/
- if (size) {
+ if (size)
memset(ptr, SCRIBBLE_BYTE, _nano_vet_and_size_of_live(nanozone, ptr));
- }
return ptr;
} else {
@@ -876,10 +1022,17 @@
static void *
nano_calloc(nanozone_t *nanozone, size_t num_items, size_t size)
{
- size_t total_bytes;
-
- if (calloc_get_size(num_items, size, 0, &total_bytes)) {
- return NULL;
+ size_t total_bytes = num_items * size;
+
+ // Check for overflow of integer multiplication
+ if (num_items > 1) {
+ /* size_t is uint64_t */
+ if ((num_items | size) & 0xffffffff00000000ul) {
+ // num_items or size equals or exceeds sqrt(2^64) == 2^32, appeal to wider arithmetic
+ __uint128_t product = ((__uint128_t)num_items) * ((__uint128_t)size);
+ if ((uint64_t)(product >> 64)) // compiles to test on upper register of register pair
+ return NULL;
+ }
}
if (total_bytes <= NANO_MAX_SIZE) {
@@ -908,24 +1061,22 @@
return zone->valloc(zone, size);
}
-static MALLOC_INLINE void
-__nano_free_definite_size(nanozone_t *nanozone, void *ptr, size_t size, boolean_t do_scribble) MALLOC_ALWAYS_INLINE;
-
-static MALLOC_INLINE void
+static INLINE void __nano_free_definite_size(nanozone_t *nanozone, void *ptr, size_t size, boolean_t do_scribble) ALWAYSINLINE;
+
+static INLINE void
__nano_free_definite_size(nanozone_t *nanozone, void *ptr, size_t size, boolean_t do_scribble)
{
nano_blk_addr_t p; // happily, the compiler holds this in a register
p.addr = (uint64_t)ptr; // place ptr on the dissecting table
- if (NANOZONE_SIGNATURE == p.fields.nano_signature) {
- if (size == ((p.fields.nano_slot + 1) << SHIFT_NANO_QUANTUM)) { // "Trust but verify."
- _nano_free_trusted_size_check_scribble(nanozone, ptr, size, do_scribble);
- return;
- } else {
- malloc_zone_error(nanozone->debug_flags, true,
- "Freeing pointer %p whose size was misdeclared\n", ptr);
- }
+ if (nanozone->our_signature == p.fields.nano_signature) {
+ if (size == ((p.fields.nano_slot + 1) << SHIFT_NANO_QUANTUM)) { // "Trust but verify."
+ _nano_free_trusted_size_check_scribble(nanozone, ptr, size, do_scribble);
+ return;
} else {
+ nanozone_error(nanozone, 1, "Freeing pointer whose size was misdeclared", ptr, NULL);
+ }
+ } else {
malloc_zone_t *zone = (malloc_zone_t *)(nanozone->helper_zone);
zone->free_definite_size(zone, ptr, size);
return;
@@ -945,25 +1096,19 @@
__nano_free_definite_size(nanozone, ptr, size, 1);
}
-static MALLOC_INLINE void __nano_free(nanozone_t *nanozone, void *ptr, boolean_t do_scribble) MALLOC_ALWAYS_INLINE;
-
-static MALLOC_INLINE void
+static INLINE void __nano_free(nanozone_t *nanozone, void *ptr, boolean_t do_scribble) ALWAYSINLINE;
+
+static INLINE void
__nano_free(nanozone_t *nanozone, void *ptr, boolean_t do_scribble)
{
- MALLOC_TRACE(TRACE_nano_free, (uintptr_t)nanozone, (uintptr_t)ptr, do_scribble, 0);
-
- if (!ptr) {
+ nano_blk_addr_t p; // happily, the compiler holds this in a register
+
+ if (!ptr)
return; // Protect against malloc_zone_free() passing NULL.
- }
-
- // <rdar://problem/26481467> exhausting a slot may result in a pointer with
- // the nanozone prefix being given to nano_free via malloc_zone_free. Calling
- // vet_and_size here, instead of in _nano_free_check_scribble means we can
- // early-out into the helper_zone if it turns out nano does not own this ptr.
- size_t sz = _nano_vet_and_size_of_live(nanozone, ptr);
-
- if (sz) {
- _nano_free_trusted_size_check_scribble(nanozone, ptr, sz, do_scribble);
+
+ p.addr = (uint64_t)ptr; // place ptr on the dissecting table
+ if (nanozone->our_signature == p.fields.nano_signature) {
+ _nano_free_check_scribble(nanozone, ptr, do_scribble);
return;
} else {
malloc_zone_t *zone = (malloc_zone_t *)(nanozone->helper_zone);
@@ -982,17 +1127,13 @@
static void
nano_forked_free(nanozone_t *nanozone, void *ptr)
{
- if (!ptr) {
+ nano_blk_addr_t p; // happily, the compiler holds this in a register
+
+ if (!ptr)
return; // Protect against malloc_zone_free() passing NULL.
- }
-
- // <rdar://problem/26481467> exhausting a slot may result in a pointer with
- // the nanozone prefix being given to nano_free via malloc_zone_free. Calling
- // vet_and_size here, instead of in _nano_free_check_scribble means we can
- // early-out into the helper_zone if it turns out nano does not own this ptr.
- size_t sz = _nano_vet_and_size_of_live(nanozone, ptr);
-
- if (sz) {
+
+ p.addr = (uint64_t)ptr; // place ptr on the dissecting table
+ if (nanozone->our_signature == p.fields.nano_signature) {
/* NOTHING. Drop it on the floor as nanozone metadata could be fouled by fork. */
return;
} else {
@@ -1022,7 +1163,7 @@
p.addr = (uint64_t)ptr; // place ptr on the dissecting table
- if (NANOZONE_SIGNATURE == p.fields.nano_signature) { // Our signature?
+ if (nanozone->our_signature == p.fields.nano_signature) { // Our signature?
return _nano_size(nanozone, ptr);
} else {
malloc_zone_t *zone = (malloc_zone_t *)(nanozone->helper_zone);
@@ -1034,42 +1175,49 @@
static void *
nano_realloc(nanozone_t *nanozone, void *ptr, size_t new_size)
{
- // could occur through malloc_zone_realloc() path
- if (!ptr) {
- // If ptr is a null pointer, realloc() shall be equivalent to malloc() for the specified size.
+ nano_blk_addr_t p; // happily, the compiler holds this in a register
+
+ p.addr = (uint64_t)ptr; // place ptr on the dissecting table
+
+ if (NULL == ptr) { // could occur through malloc_zone_realloc() path
+ // If ptr is a null pointer, realloc() shall be equivalent to malloc() for the specified size.
return nano_malloc(nanozone, new_size);
- }
-
- size_t old_size = _nano_vet_and_size_of_live(nanozone, ptr);
- if (!old_size) {
- // not-nano pointer, hand down to helper zone
- malloc_zone_t *zone = (malloc_zone_t *)(nanozone->helper_zone);
- return zone->realloc(zone, ptr, new_size);
- } else {
- if (new_size <= NANO_MAX_SIZE) {
- // nano to nano?
+ } else if (nanozone->our_signature == p.fields.nano_signature) { // Our signature?
+ if (new_size <= NANO_MAX_SIZE) { // nano to nano?
void *q = _nano_realloc(nanozone, ptr, new_size);
if (q) {
return q;
- } else {
- // nano exhausted
+ } else { // nano exhausted
/* FALLTHROUGH to helper zone copying case */
}
}
+ // nano to larger-than-nano (or FALLTHROUGH from just above)
+ size_t old_size = _nano_vet_and_size_of_live(nanozone, ptr);
+
+ if (!old_size) {
+ nanozone_error(nanozone, 1, "pointer being reallocated was not allocated", ptr, NULL);
+ return NULL;
+ } else {
+ malloc_zone_t *zone = (malloc_zone_t *)(nanozone->helper_zone);
+ void *new_ptr = zone->malloc(zone, new_size);
+
+ if (new_ptr) {
+ size_t valid_size = MIN(old_size, new_size);
+ memcpy(new_ptr, ptr, valid_size);
+ _nano_free_check_scribble(nanozone, ptr, (nanozone->debug_flags & SCALABLE_MALLOC_DO_SCRIBBLE));
+ return new_ptr;
+ } else {
+ /* Original ptr is left intact */
+ return NULL;
+ }
+ /* NOTREACHED */
+ }
+ } else {
+ // other-than-nano (not necessarily larger! possibly NULL!) to whatever
malloc_zone_t *zone = (malloc_zone_t *)(nanozone->helper_zone);
- void *new_ptr = zone->malloc(zone, new_size);
-
- if (new_ptr) {
- size_t valid_size = MIN(old_size, new_size);
- memcpy(new_ptr, ptr, valid_size);
- _nano_free_check_scribble(nanozone, ptr, (nanozone->debug_flags & MALLOC_DO_SCRIBBLE));
- return new_ptr;
- } else {
- /* Original ptr is left intact */
- return NULL;
- }
- /* NOTREACHED */
+
+ return zone->realloc(zone, ptr, new_size);
}
/* NOTREACHED */
}
@@ -1077,40 +1225,48 @@
static void *
nano_forked_realloc(nanozone_t *nanozone, void *ptr, size_t new_size)
{
- // could occur through malloc_zone_realloc() path
- if (!ptr) {
- // If ptr is a null pointer, realloc() shall be equivalent to malloc() for the specified size.
+ nano_blk_addr_t p; // happily, the compiler holds this in a register
+
+ p.addr = (uint64_t)ptr; // place ptr on the dissecting table
+
+ if (NULL == ptr) { // could occur through malloc_zone_realloc() path
+ // If ptr is a null pointer, realloc() shall be equivalent to malloc() for the specified size.
return nano_forked_malloc(nanozone, new_size);
- }
-
- size_t old_size = _nano_vet_and_size_of_live(nanozone, ptr);
- if (!old_size) {
- // not-nano pointer, hand down to helper zone
+ } else if (nanozone->our_signature == p.fields.nano_signature) { // Our signature?
+ if (0 == new_size) {
+ // If size is 0 and ptr is not a null pointer, the object pointed to is freed.
+ // However as nanozone metadata could be fouled by fork, we'll intentionally leak it.
+
+ // If size is 0, either a null pointer or a unique pointer that can be successfully passed
+ // to free() shall be returned.
+ return nano_forked_malloc(nanozone, 1);
+ }
+
+ size_t old_size = _nano_vet_and_size_of_live(nanozone, ptr);
+
+ if (!old_size) {
+ nanozone_error(nanozone, 1, "pointer being reallocated was not allocated", ptr, NULL);
+ return NULL;
+ } else {
+ malloc_zone_t *zone = (malloc_zone_t *)(nanozone->helper_zone);
+ void *new_ptr = zone->malloc(zone, new_size);
+
+ if (new_ptr) {
+ size_t valid_size = MIN(old_size, new_size);
+ memcpy(new_ptr, ptr, valid_size);
+ /* Original pointer is intentionally leaked as nanozone metadata could be fouled by fork. */
+ return new_ptr;
+ } else {
+ /* Original ptr is left intact */
+ return NULL;
+ }
+ /* NOTREACHED */
+ }
+ } else {
+ // other-than-nano (not necessarily larger! possibly NULL!) to whatever
malloc_zone_t *zone = (malloc_zone_t *)(nanozone->helper_zone);
+
return zone->realloc(zone, ptr, new_size);
- } else {
- if (0 == new_size) {
- // If size is 0 and ptr is not a null pointer, the object pointed to is freed.
- // However as nanozone metadata could be fouled by fork, we'll intentionally leak it.
-
- // If size is 0, either a null pointer or a unique pointer that can be successfully passed
- // to free() shall be returned.
- return nano_forked_malloc(nanozone, 1);
- }
-
- malloc_zone_t *zone = (malloc_zone_t *)(nanozone->helper_zone);
- void *new_ptr = zone->malloc(zone, new_size);
-
- if (new_ptr) {
- size_t valid_size = MIN(old_size, new_size);
- memcpy(new_ptr, ptr, valid_size);
- /* Original pointer is intentionally leaked as nanozone metadata could be fouled by fork. */
- return new_ptr;
- } else {
- /* Original ptr is left intact */
- return NULL;
- }
- /* NOTREACHED */
}
/* NOTREACHED */
}
@@ -1127,14 +1283,13 @@
static unsigned
nano_batch_malloc(nanozone_t *nanozone, size_t size, void **results, unsigned count)
{
- unsigned found = 0;
+ unsigned found = 0;
if (size <= NANO_MAX_SIZE) {
while (found < count) {
void *ptr = _nano_malloc_check_clear(nanozone, size, 0);
- if (!ptr) {
+ if (!ptr)
break;
- }
*results++ = ptr;
found++;
@@ -1160,38 +1315,34 @@
static void
nano_batch_free(nanozone_t *nanozone, void **to_be_freed, unsigned count)
{
- void *ptr;
+ void *ptr;
// frees all the pointers in to_be_freed
// note that to_be_freed may be overwritten during the process
- if (!count) {
+ if (!count)
return;
- }
while (count--) {
ptr = to_be_freed[count];
- if (ptr) {
+ if (ptr)
nano_free(nanozone, ptr);
- }
}
}
static void
nano_forked_batch_free(nanozone_t *nanozone, void **to_be_freed, unsigned count)
{
- void *ptr;
+ void *ptr;
// frees all the pointers in to_be_freed
// note that to_be_freed may be overwritten during the process
- if (!count) {
+ if (!count)
return;
- }
while (count--) {
ptr = to_be_freed[count];
- if (ptr) {
+ if (ptr)
nano_forked_free(nanozone, ptr);
- }
}
}
@@ -1202,35 +1353,14 @@
return zone->memalign(zone, alignment, size);
}
-static boolean_t
-nano_claimed_address(nanozone_t *nanozone, void *ptr)
-{
- nano_blk_addr_t p;
- p.addr = (uint64_t)ptr;
- if (NANOZONE_SIGNATURE != p.fields.nano_signature) {
- // Not a nano address - let the helper zone handle it.
- malloc_zone_t *helper_zone = nanozone->helper_zone;
- return malloc_zone_claimed_address(helper_zone, ptr);
- }
- return __nano_vet_and_size_inner(nanozone, ptr, true) != 0;
-}
-
-static boolean_t
-nano_forked_claimed_address(struct _malloc_zone_t *zone, void *ptr)
-{
- // This does not operate after fork - default to true to avoid
- // false negatives.
- return true;
-}
-
static size_t
nano_try_madvise(nanozone_t *nanozone, size_t goal)
{
- unsigned int mag_index, slot_key;
- size_t bytes_toward_goal = 0;
-
- for (mag_index = 0; mag_index < nano_common_max_magazines; mag_index++) {
- nano_blk_addr_t p;
+ unsigned int mag_index, slot_key;
+ size_t bytes_toward_goal = 0;
+
+ for (mag_index = 0; mag_index < nanozone->phys_ncpus; mag_index++) {
+ nano_blk_addr_t p;
// Establish p as base address for band 0, slot 0, offset 0
p.fields.nano_signature = NANOZONE_SIGNATURE;
@@ -1239,11 +1369,13 @@
p.fields.nano_slot = 0;
p.fields.nano_offset = 0;
- for (slot_key = 0; slot_key < SLOT_KEY_LIMIT; p.addr += SLOT_IN_BAND_SIZE, // Advance to next slot base
- slot_key++) {
- // malloc_report(ASL_LEVEL_WARNING,"nano_try_madvise examining slot base %p\n", p.addr);
+ for (slot_key = 0; slot_key < SLOT_KEY_LIMIT;
+ p.addr += SLOT_IN_BAND_SIZE, // Advance to next slot base
+ slot_key++) {
+
+ // _malloc_printf(ASL_LEVEL_WARNING,"nano_try_madvise examining slot base %p\n", p.addr);
nano_meta_admin_t pMeta = &(nanozone->meta_data[mag_index][slot_key]);
- uintptr_t slot_bump_addr = pMeta->slot_bump_addr; // capture this volatile pointer
+ uintptr_t slot_bump_addr = pMeta->slot_bump_addr; // capture this volatile pointer
size_t slot_objects_mapped = pMeta->slot_objects_mapped; // capture this volatile count
if (0 == slot_objects_mapped) { // Nothing allocated in this magazine for this slot?
@@ -1254,52 +1386,47 @@
bitarray_t slot_bitarray = bitarray_create(log_size);
unsigned int slot_bytes = pMeta->slot_bytes;
- int log_page_count = 64 - __builtin_clzl((slot_objects_mapped * slot_bytes) / vm_kernel_page_size);
+ int log_page_count = 64 - __builtin_clzl((slot_objects_mapped * slot_bytes) / vm_page_size);
log_page_count = 1 + MAX(0, log_page_count);
bitarray_t page_bitarray = bitarray_create(log_page_count);
- // malloc_report(ASL_LEVEL_WARNING,"slot_bitarray: %db page_bitarray: %db\n", bitarray_size(log_size),
- // bitarray_size(log_page_count));
+ // _malloc_printf(ASL_LEVEL_WARNING,"slot_bitarray: %db page_bitarray: %db\n", bitarray_size(log_size), bitarray_size(log_page_count));
if (!slot_bitarray) {
- malloc_report(ASL_LEVEL_ERR, "bitarray_create(%d) in nano_try_madvise returned errno=%d.\n", log_size, errno);
- free(page_bitarray);
+ malloc_printf("bitarray_create(%d) in nano_try_madvise returned errno=%d.", log_size, errno);
return bytes_toward_goal;
}
if (!page_bitarray) {
- malloc_report(ASL_LEVEL_ERR, "bitarray_create(%d) in nano_try_madvise returned errno=%d.\n", log_page_count, errno);
+ malloc_printf("bitarray_create(%d) in nano_try_madvise returned errno=%d.", log_page_count, errno);
free(slot_bitarray);
return bytes_toward_goal;
}
chained_block_t head = NULL, tail = NULL, t;
- unsigned stoploss = (unsigned)slot_objects_mapped;
- while ((t = OSAtomicDequeue(&(pMeta->slot_LIFO), offsetof(struct chained_block_s, next)))) {
+ unsigned stoploss = slot_objects_mapped;
+ while ((t = OSAtomicDequeue( &(pMeta->slot_LIFO), offsetof(struct chained_block_s,next)))) {
if (0 == stoploss) {
- malloc_report(ASL_LEVEL_ERR, "Free list walk in nano_try_madvise exceeded object count.\n");
+ malloc_printf("Free list walk in nano_try_madvise exceeded object count.");
break;
}
stoploss--;
- uintptr_t offset = ((uintptr_t)t - p.addr); // offset from beginning of slot
- index_t block_index = offset_to_index(nanozone, pMeta, offset);
+ uintptr_t offset = ((uintptr_t)t - p.addr); // offset from beginning of slot
+ index_t block_index = offset_to_index(nanozone, pMeta, offset);
// build a simple linked list of the free blocks we're able to obtain
- if (NULL == head) {
+ if (NULL == head)
head = t;
- } else {
+ else
tail->next = t;
- }
tail = t;
// take note in a bitarray_t of each free block we're able to obtain (allows fast lookup below)
- if (block_index < slot_objects_mapped) {
+ if (block_index < slot_objects_mapped)
bitarray_set(slot_bitarray, log_size, block_index);
- }
}
- if (tail) {
+ if (tail)
tail->next = NULL;
- }
if (NULL == head) {
free(slot_bitarray);
@@ -1310,37 +1437,33 @@
index_t i;
nano_blk_addr_t q;
size_t pgnum;
- for (i = (index_t)pMeta->slot_objects_skipped; i < slot_objects_mapped; ++i) {
+ for (i = pMeta->slot_objects_skipped; i < slot_objects_mapped; ++i) {
uintptr_t block_offset = index_to_offset(nanozone, pMeta, i);
- if (p.addr + block_offset >= slot_bump_addr) {
+ if (p.addr + block_offset >= slot_bump_addr)
break;
- }
if (!bitarray_get(slot_bitarray, log_size, i)) { // is block i allocated or already on an madvise'd page?
// Mark the page(s) it resides on as live
q.addr = p.addr + block_offset;
- pgnum = ((((unsigned)q.fields.nano_band) << NANO_OFFSET_BITS) | ((unsigned)q.fields.nano_offset)) >>
- vm_kernel_page_shift;
- bitarray_set(page_bitarray, log_page_count, (index_t)pgnum);
+ pgnum = ((((unsigned)q.fields.nano_band) << NANO_OFFSET_BITS) | ((unsigned)q.fields.nano_offset)) >> vm_page_shift;
+ bitarray_set(page_bitarray, log_page_count, pgnum);
q.addr += slot_bytes - 1;
- pgnum = ((((unsigned)q.fields.nano_band) << NANO_OFFSET_BITS) | ((unsigned)q.fields.nano_offset)) >>
- vm_kernel_page_shift;
- bitarray_set(page_bitarray, log_page_count, (index_t)pgnum);
+ pgnum = ((((unsigned)q.fields.nano_band) << NANO_OFFSET_BITS) | ((unsigned)q.fields.nano_offset)) >> vm_page_shift;
+ bitarray_set(page_bitarray, log_page_count, pgnum);
}
}
free(slot_bitarray);
- q.addr = p.addr + index_to_offset(nanozone, pMeta, (index_t)pMeta->slot_objects_skipped);
- index_t pgstart =
- ((((unsigned)q.fields.nano_band) << NANO_OFFSET_BITS) | ((unsigned)q.fields.nano_offset)) >> vm_kernel_page_shift;
+ q.addr = p.addr + index_to_offset(nanozone, pMeta, pMeta->slot_objects_skipped);
+ index_t pgstart = ((((unsigned)q.fields.nano_band) << NANO_OFFSET_BITS) | ((unsigned)q.fields.nano_offset)) >> vm_page_shift;
q.addr = slot_bump_addr - slot_bytes;
- pgnum = ((((unsigned)q.fields.nano_band) << NANO_OFFSET_BITS) | ((unsigned)q.fields.nano_offset)) >> vm_kernel_page_shift;
-
- // malloc_report(ASL_LEVEL_WARNING,"Examining %d pages. Slot base %p.\n", pgnum - pgstart + 1, p.addr);
+ pgnum = ((((unsigned)q.fields.nano_band) << NANO_OFFSET_BITS) | ((unsigned)q.fields.nano_offset)) >> vm_page_shift;
+
+ // _malloc_printf(ASL_LEVEL_WARNING,"Examining %d pages. Slot base %p.\n", pgnum - pgstart + 1, p.addr);
if (pMeta->slot_madvised_pages) {
if (pMeta->slot_madvised_log_page_count < log_page_count) {
@@ -1360,11 +1483,11 @@
bitarray_t will_madvise_pages = bitarray_create(log_page_count);
int num_advised = 0;
-
+
for (i = pgstart; i < pgnum; ++i) {
if ((i < (1 << log_page_count)) && // bounds check for the bitarray_get()'s that follow.
- !bitarray_get(pMeta->slot_madvised_pages, log_page_count, i) && // already madvise'd?
- !bitarray_get(page_bitarray, log_page_count, i)) // no live allocations?
+ !bitarray_get(pMeta->slot_madvised_pages, log_page_count, i) && // already madvise'd?
+ !bitarray_get(page_bitarray, log_page_count, i)) // no live allocations?
{
num_advised++;
bitarray_set(will_madvise_pages, log_page_count, i);
@@ -1374,81 +1497,72 @@
if (num_advised) {
chained_block_t new_head = NULL, new_tail = NULL;
- // malloc_report(ASL_LEVEL_WARNING,"Constructing residual free list starting at %p num_advised %d\n", head,
- // num_advised);
+ // _malloc_printf(ASL_LEVEL_WARNING,"Constructing residual free list starting at %p num_advised %d\n", head, num_advised);
t = head;
while (t) {
q.addr = (uintptr_t)t;
- index_t pgnum_start =
- ((((unsigned)q.fields.nano_band) << NANO_OFFSET_BITS) | ((unsigned)q.fields.nano_offset)) >>
- vm_kernel_page_shift;
+ index_t pgnum_start = ((((unsigned)q.fields.nano_band) << NANO_OFFSET_BITS) | ((unsigned)q.fields.nano_offset)) >> vm_page_shift;
q.addr += slot_bytes - 1;
- index_t pgnum_end =
- ((((unsigned)q.fields.nano_band) << NANO_OFFSET_BITS) | ((unsigned)q.fields.nano_offset)) >>
- vm_kernel_page_shift;
+ index_t pgnum_end = ((((unsigned)q.fields.nano_band) << NANO_OFFSET_BITS) | ((unsigned)q.fields.nano_offset)) >> vm_page_shift;
// bounds check for the bitarray_get()'s that follow. If the pgnum is beyond the
// capacity of the will_madvise_pages just restore the block to the free list.
if (pgnum_start >= (1 << log_page_count)) {
- if (NULL == new_head) {
+ if (NULL == new_head)
new_head = t;
- } else {
+ else
new_tail->next = t;
- }
new_tail = t;
}
// If the block nowhere lies on an madvise()'d page restore it to the slot free list.
else if (!bitarray_get(will_madvise_pages, log_page_count, pgnum_start) &&
!bitarray_get(will_madvise_pages, log_page_count, pgnum_end)) {
- if (NULL == new_head) {
+ if (NULL == new_head)
new_head = t;
- } else {
+ else
new_tail->next = t;
- }
new_tail = t;
}
t = t->next;
}
- if (new_tail) {
+ if (new_tail)
new_tail->next = NULL;
- }
// push the free list extracted above back onto the LIFO, all at once
- if (new_head) {
- OSAtomicEnqueue(&(pMeta->slot_LIFO), new_head,
- (uintptr_t)new_tail - (uintptr_t)new_head + offsetof(struct chained_block_s, next));
- }
+ if (new_head)
+ OSAtomicEnqueue( &(pMeta->slot_LIFO), new_head,
+ (uintptr_t)new_tail - (uintptr_t)new_head + offsetof(struct chained_block_s,next));
} else {
- // malloc_report(ASL_LEVEL_WARNING,"Reinstating free list since no pages were madvised (%d).\n", num_advised);
- if (head) {
- OSAtomicEnqueue(&(pMeta->slot_LIFO), head,
- (uintptr_t)tail - (uintptr_t)head + offsetof(struct chained_block_s, next));
- }
+ // _malloc_printf(ASL_LEVEL_WARNING,"Reinstating free list since no pages were madvised (%d).\n", num_advised);
+ if (head)
+ OSAtomicEnqueue( &(pMeta->slot_LIFO), head,
+ (uintptr_t)tail - (uintptr_t)head + offsetof(struct chained_block_s,next));
}
for (i = pgstart; i < pgnum; ++i) {
if ((i < (1 << log_page_count)) && bitarray_get(will_madvise_pages, log_page_count, i)) {
q = p;
- q.fields.nano_band = (i << vm_kernel_page_shift) >> NANO_OFFSET_BITS;
- q.fields.nano_offset = (i << vm_kernel_page_shift) & ((1 << NANO_OFFSET_BITS) - 1);
- // malloc_report(ASL_LEVEL_WARNING,"Entire page non-live: %d. Slot base %p, madvising %p\n", i, p.addr,
- // q.addr);
-
- if (nanozone->debug_flags & MALLOC_DO_SCRIBBLE) {
- memset((void *)q.addr, SCRUBBLE_BYTE, vm_kernel_page_size);
- }
-
- if (-1 == madvise((void *)q.addr, vm_kernel_page_size, MADV_FREE_REUSABLE))
+ q.fields.nano_band = (i << vm_page_shift) >> NANO_OFFSET_BITS;
+ q.fields.nano_offset = (i << vm_page_shift) & ((1 << NANO_OFFSET_BITS) - 1);
+ // _malloc_printf(ASL_LEVEL_WARNING,"Entire page non-live: %d. Slot base %p, madvising %p\n", i, p.addr, q.addr);
+
+ if (nanozone->debug_flags & SCALABLE_MALLOC_DO_SCRIBBLE)
+ memset((void *)q.addr, SCRUBBLE_BYTE, vm_page_size);
+#if TARGET_OS_EMBEDDED
+ if (-1 == madvise((void *)q.addr, vm_page_size, MADV_FREE))
+#else
+ if (-1 == madvise((void *)q.addr, vm_page_size, MADV_FREE_REUSABLE))
+#endif
{
- /* -1 return: VM map entry change makes this unfit for reuse. Something evil lurks. */
+ /* -1 return: VM map entry change makes this unfit for reuse. Something evil lurks. */
#if DEBUG_MADVISE
- nanozone_error(nanozone, 0, "madvise(..., MADV_FREE_REUSABLE) failed", (void *)cwq.addrpgLo,
- "length=%d\n", vm_page_size);
+ nanozone_error(nanozone, 0, "madvise(..., MADV_FREE_REUSABLE) failed",
+ (void *)cwq.addrpgLo, "length=%d\n", vm_page_size);
#endif
} else {
- bytes_toward_goal += vm_kernel_page_size;
- bitarray_set(pMeta->slot_madvised_pages, log_page_count, i);
+ bytes_toward_goal += vm_page_size;
+ bitarray_set(pMeta->slot_madvised_pages, log_page_count, i);
}
}
}
@@ -1460,9 +1574,8 @@
pMeta->slot_madvised_log_page_count = 0;
}
- if (goal && bytes_toward_goal >= goal) {
+ if (goal && bytes_toward_goal >= goal)
return bytes_toward_goal;
- }
}
}
}
@@ -1472,60 +1585,49 @@
static size_t
nano_pressure_relief(nanozone_t *nanozone, size_t goal)
{
- MAGMALLOC_PRESSURERELIEFBEGIN((void *)nanozone, nanozone->basic_zone.zone_name, (int)goal);
- MALLOC_TRACE(TRACE_nano_memory_pressure | DBG_FUNC_START, (uint64_t)nanozone, goal, 0, 0);
-
- size_t total = nano_try_madvise(nanozone, goal);
-
- MAGMALLOC_PRESSURERELIEFEND((void *)nanozone, nanozone->basic_zone.zone_name, (int)goal, (int)total);
- MALLOC_TRACE(TRACE_nano_memory_pressure | DBG_FUNC_END, (uint64_t)nanozone, goal, total, 0);
-
- return total;
+ return nano_try_madvise(nanozone, goal);
}
/**************** introspection methods *********************/
static kern_return_t
-nano_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)
-{
- nanozone_t *nanozone;
- kern_return_t err;
- struct nanozone_s zone_copy;
-
- if (!reader) {
- reader = nano_common_default_reader;
- }
+nanozone_default_reader(task_t task, vm_address_t address, vm_size_t size, void **ptr)
+{
+ *ptr = (void *)address;
+ return 0;
+}
+
+static kern_return_t
+nano_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)
+{
+ nanozone_t *nanozone;
+ kern_return_t err;
+
+ if (!reader) reader = nanozone_default_reader;
err = reader(task, zone_address, sizeof(nanozone_t), (void **)&nanozone);
- if (err) {
- return err;
- }
- memcpy(&zone_copy, nanozone, sizeof(zone_copy));
-
- err = segregated_in_use_enumerator(task, context, type_mask, &zone_copy, reader, recorder);
+ if (err) return err;
+
+ err = segregated_in_use_enumerator(task, context, type_mask, nanozone, reader, recorder);
return err;
}
static size_t
nano_good_size(nanozone_t *nanozone, size_t size)
{
- if (size <= NANO_MAX_SIZE) {
- return _nano_common_good_size(size);
- } else {
+ if (size <= NANO_MAX_SIZE)
+ return _nano_good_size(nanozone, size);
+ else {
malloc_zone_t *zone = (malloc_zone_t *)(nanozone->helper_zone);
return zone->introspect->good_size(zone, size);
}
}
// TODO sanity checks
-static unsigned nanozone_check_counter = 0;
-static unsigned nanozone_check_start = 0;
-static unsigned nanozone_check_modulo = 1;
+unsigned nanozone_check_counter = 0;
+unsigned nanozone_check_start = 0;
+unsigned nanozone_check_modulo = 1;
static boolean_t
nano_check_all(nanozone_t *nanozone, const char *function)
@@ -1536,17 +1638,14 @@
static boolean_t
nanozone_check(nanozone_t *nanozone)
{
- if ((++nanozone_check_counter % 10000) == 0) {
- malloc_report(ASL_LEVEL_NOTICE, "at nanozone_check counter=%d\n", nanozone_check_counter);
- }
-
- if (nanozone_check_counter < nanozone_check_start) {
+ if ((++nanozone_check_counter % 10000) == 0)
+ _malloc_printf(ASL_LEVEL_NOTICE, "at nanozone_check counter=%d\n", nanozone_check_counter);
+
+ if (nanozone_check_counter < nanozone_check_start)
return 1;
- }
-
- if (nanozone_check_counter % nanozone_check_modulo) {
+
+ if (nanozone_check_counter % nanozone_check_modulo)
return 1;
- }
return nano_check_all(nanozone, "");
}
@@ -1557,32 +1656,28 @@
chained_block_t head = NULL, tail = NULL, t;
unsigned count = 0;
- unsigned stoploss = (unsigned)pMeta->slot_objects_mapped;
- while ((t = OSAtomicDequeue(&(pMeta->slot_LIFO), offsetof(struct chained_block_s, next)))) {
+ unsigned stoploss = pMeta->slot_objects_mapped;
+ while ((t = OSAtomicDequeue( &(pMeta->slot_LIFO), offsetof(struct chained_block_s,next)))) {
if (0 == stoploss) {
- malloc_zone_error(nanozone->debug_flags, true,
- "Free list walk in count_free exceeded object count.\n",
- (void *)&(pMeta->slot_LIFO), NULL);
+ nanozone_error(nanozone, 1, "Free list walk in count_free exceeded object count.",
+ (void *)&(pMeta->slot_LIFO), NULL);
}
stoploss--;
- if (NULL == head) {
+ if (NULL == head)
head = t;
- } else {
+ else
tail->next = t;
- }
tail = t;
count++;
}
- if (tail) {
+ if (tail)
tail->next = NULL;
- }
// push the free list extracted above back onto the LIFO, all at once
- if (head) {
- OSAtomicEnqueue(&(pMeta->slot_LIFO), head, (uintptr_t)tail - (uintptr_t)head + offsetof(struct chained_block_s, next));
- }
+ if (head)
+ OSAtomicEnqueue( &(pMeta->slot_LIFO), head, (uintptr_t)tail - (uintptr_t)head + offsetof(struct chained_block_s,next));
return count;
}
@@ -1590,17 +1685,16 @@
static void
nano_print(nanozone_t *nanozone, boolean_t verbose)
{
- unsigned int mag_index, slot_key;
+ unsigned int mag_index, slot_key;
malloc_statistics_t stats;
nano_statistics(nanozone, &stats);
- malloc_report(MALLOC_REPORT_NOLOG | MALLOC_REPORT_NOPREFIX,
- "Nanozone %p: inUse=%d(%lluKB) touched=%lluKB allocated=%lluMB\n",
- nanozone, stats.blocks_in_use, (uint64_t)stats.size_in_use >> 10,
- (uint64_t)stats.max_size_in_use >> 10, (uint64_t)stats.size_allocated >> 20);
-
- for (mag_index = 0; mag_index < nano_common_max_magazines; mag_index++) {
- nano_blk_addr_t p;
+ _malloc_printf(MALLOC_PRINTF_NOLOG | MALLOC_PRINTF_NOPREFIX,
+ "Nanozone %p: inUse=%d(%dKB) touched=%dKB allocated=%dMB\n",
+ nanozone, stats.blocks_in_use, stats.size_in_use>>10, stats.max_size_in_use>>10, stats.size_allocated>>20);
+
+ for (mag_index = 0; mag_index < nanozone->phys_ncpus; mag_index++) {
+ nano_blk_addr_t p;
// Establish p as base address for band 0, slot 0, offset 0
p.fields.nano_signature = NANOZONE_SIGNATURE;
@@ -1609,20 +1703,22 @@
p.fields.nano_slot = 0;
p.fields.nano_offset = 0;
- for (slot_key = 0; slot_key < SLOT_KEY_LIMIT; p.addr += SLOT_IN_BAND_SIZE, // Advance to next slot base
- slot_key++) {
+ for (slot_key = 0; slot_key < SLOT_KEY_LIMIT;
+ p.addr += SLOT_IN_BAND_SIZE, // Advance to next slot base
+ slot_key++) {
+
nano_meta_admin_t pMeta = &(nanozone->meta_data[mag_index][slot_key]);
- uintptr_t slot_bump_addr = pMeta->slot_bump_addr; // capture this volatile pointer
+ uintptr_t slot_bump_addr = pMeta->slot_bump_addr; // capture this volatile pointer
size_t slot_objects_mapped = pMeta->slot_objects_mapped; // capture this volatile count
if (0 == slot_objects_mapped) { // Nothing allocated in this magazine for this slot?
- malloc_report(MALLOC_REPORT_NOLOG | MALLOC_REPORT_NOPREFIX, "Magazine %2d(%3d) Unrealized\n", mag_index,
- (slot_key + 1) << SHIFT_NANO_QUANTUM);
+ _malloc_printf(MALLOC_PRINTF_NOLOG | MALLOC_PRINTF_NOPREFIX,
+ "Magazine %2d(%3d) Unrealized\n",mag_index, (slot_key + 1) << SHIFT_NANO_QUANTUM);
continue;
}
uintptr_t offset = (0 == slot_bump_addr ? 0 : slot_bump_addr - p.addr);
- unsigned blocks_touched = offset_to_index(nanozone, pMeta, offset) - (unsigned)pMeta->slot_objects_skipped;
+ unsigned blocks_touched = offset_to_index(nanozone, pMeta, offset) - pMeta->slot_objects_skipped;
unsigned blocks_now_free = count_free(nanozone, pMeta);
unsigned blocks_in_use = blocks_touched - blocks_now_free;
@@ -1630,10 +1726,10 @@
size_t size_in_use = ((slot_key + 1) << SHIFT_NANO_QUANTUM) * blocks_in_use;
size_t size_allocated = ((offset / BAND_SIZE) + 1) * SLOT_IN_BAND_SIZE;
- malloc_report(MALLOC_REPORT_NOLOG | MALLOC_REPORT_NOPREFIX,
- "Magazine %2d(%3d) [%p, %3lluKB] \t Allocations in use=%4d \t Bytes in use=%llub \t Untouched=%lluKB\n", mag_index,
- (slot_key + 1) << SHIFT_NANO_QUANTUM, (void *)p.addr, (uint64_t)(size_allocated >> 10), blocks_in_use, (uint64_t)size_in_use,
- (uint64_t)((size_allocated - size_hiwater) >> 10));
+ _malloc_printf(MALLOC_PRINTF_NOLOG | MALLOC_PRINTF_NOPREFIX,
+ "Magazine %2d(%3d) [%p, %3dKB] \t Allocations in use=%4d \t Bytes in use=%db \t Untouched=%dKB\n",
+ mag_index, (slot_key + 1) << SHIFT_NANO_QUANTUM, p,
+ (size_allocated>>10), blocks_in_use, size_in_use, (size_allocated - size_hiwater)>>10);
if (!verbose) {
continue;
@@ -1643,70 +1739,63 @@
bitarray_t slot_bitarray = bitarray_create(log_size);
if (!slot_bitarray) {
- malloc_report(ASL_LEVEL_ERR, "bitarray_create(%d) in nano_print returned errno=%d.\n", log_size, errno);
+ malloc_printf("bitarray_create(%d) in nano_print returned errno=%d.", log_size, errno);
return;
}
chained_block_t head = NULL, tail = NULL, t;
- unsigned stoploss = (unsigned)slot_objects_mapped;
- while ((t = OSAtomicDequeue(&(pMeta->slot_LIFO), offsetof(struct chained_block_s, next)))) {
+ unsigned stoploss = slot_objects_mapped;
+ while ((t = OSAtomicDequeue( &(pMeta->slot_LIFO), offsetof(struct chained_block_s,next)))) {
if (0 == stoploss) {
- malloc_report(ASL_LEVEL_ERR, "Free list walk in nano_print exceeded object count.\n");
+ malloc_printf("Free list walk in nano_print exceeded object count.");
break;
}
stoploss--;
- uintptr_t offset = ((uintptr_t)t - p.addr); // offset from beginning of slot
- index_t block_index = offset_to_index(nanozone, pMeta, offset);
-
- if (NULL == head) {
+ uintptr_t offset = ((uintptr_t)t - p.addr); // offset from beginning of slot
+ index_t block_index = offset_to_index(nanozone, pMeta, offset);
+
+ if (NULL == head)
head = t;
- } else {
+ else
tail->next = t;
- }
tail = t;
- if (block_index < slot_objects_mapped) {
+ if (block_index < slot_objects_mapped)
bitarray_set(slot_bitarray, log_size, block_index);
- }
}
- if (tail) {
+ if (tail)
tail->next = NULL;
- }
index_t i;
- for (i = 0; i < slot_objects_mapped; ++i) {
+ for (i = 0; i < slot_objects_mapped; ++i) {
nano_blk_addr_t q;
size_t pgnum;
uintptr_t block_offset = index_to_offset(nanozone, pMeta, i);
- if (p.addr + block_offset >= slot_bump_addr) {
+ if (p.addr + block_offset >= slot_bump_addr)
break;
- }
q.addr = p.addr + block_offset;
- pgnum = ((((unsigned)q.fields.nano_band) << NANO_OFFSET_BITS) | ((unsigned)q.fields.nano_offset)) >>
- vm_kernel_page_shift;
+ pgnum = ((((unsigned)q.fields.nano_band) << NANO_OFFSET_BITS) | ((unsigned)q.fields.nano_offset)) >> vm_page_shift;
if (i < pMeta->slot_objects_skipped) {
- malloc_report(MALLOC_REPORT_NOLOG | MALLOC_REPORT_NOPREFIX, "_");
+ _malloc_printf(MALLOC_PRINTF_NOLOG | MALLOC_PRINTF_NOPREFIX,"_");
} else if (bitarray_get(slot_bitarray, log_size, i)) {
- malloc_report(MALLOC_REPORT_NOLOG | MALLOC_REPORT_NOPREFIX, "F");
- } else if (pMeta->slot_madvised_pages && (pgnum < (1 << pMeta->slot_madvised_log_page_count)) &&
- bitarray_get(pMeta->slot_madvised_pages, pMeta->slot_madvised_log_page_count, (index_t)pgnum)) {
- malloc_report(MALLOC_REPORT_NOLOG | MALLOC_REPORT_NOPREFIX, "M");
+ _malloc_printf(MALLOC_PRINTF_NOLOG | MALLOC_PRINTF_NOPREFIX,"F");
+ } else if (pMeta->slot_madvised_pages && (pgnum < ( 1 << pMeta->slot_madvised_log_page_count)) &&
+ bitarray_get(pMeta->slot_madvised_pages, pMeta->slot_madvised_log_page_count, pgnum)) {
+ _malloc_printf(MALLOC_PRINTF_NOLOG | MALLOC_PRINTF_NOPREFIX,"M");
} else {
- malloc_report(MALLOC_REPORT_NOLOG | MALLOC_REPORT_NOPREFIX, ".");
+ _malloc_printf(MALLOC_PRINTF_NOLOG | MALLOC_PRINTF_NOPREFIX,".");
}
}
- malloc_report(MALLOC_REPORT_NOLOG | MALLOC_REPORT_NOPREFIX, "\n");
+ _malloc_printf(MALLOC_PRINTF_NOLOG | MALLOC_PRINTF_NOPREFIX,"\n");
free(slot_bitarray);
// push the free list extracted above back onto the LIFO, all at once
- if (head) {
- OSAtomicEnqueue(
- &(pMeta->slot_LIFO), head, (uintptr_t)tail - (uintptr_t)head + offsetof(struct chained_block_s, next));
- }
+ if (head)
+ OSAtomicEnqueue( &(pMeta->slot_LIFO), head, (uintptr_t)tail - (uintptr_t)head + offsetof(struct chained_block_s,next));
}
}
}
@@ -1723,8 +1812,8 @@
{
int i;
- for (i = 0; i < nano_common_max_magazines; ++i) {
- _malloc_lock_lock(&nanozone->band_resupply_lock[i]);
+ for (i = 0; i < nanozone->phys_ncpus; ++i) {
+ _malloc_lock_lock(&nanozone->band_resupply_lock[i]);
}
}
@@ -1733,30 +1822,20 @@
{
int i;
- for (i = 0; i < nano_common_max_magazines; ++i) {
- _malloc_lock_unlock(&nanozone->band_resupply_lock[i]);
- }
-}
-
-static void
-nano_reinit_lock(nanozone_t *nanozone)
-{
- int i;
-
- for (i = 0; i < nano_common_max_magazines; ++i) {
- _malloc_lock_init(&nanozone->band_resupply_lock[i]);
+ for (i = 0; i < nanozone->phys_ncpus; ++i) {
+ _malloc_lock_unlock(&nanozone->band_resupply_lock[i]);
}
}
static void
nano_statistics(nanozone_t *nanozone, malloc_statistics_t *stats)
{
- int i, j;
+ int i,j;
bzero(stats, sizeof(*stats));
- for (i = 0; i < nano_common_max_magazines; ++i) {
- nano_blk_addr_t p;
+ for (i = 0; i < nanozone->phys_ncpus; ++i) {
+ nano_blk_addr_t p;
// Establish p as base address for slot 0 in this CPU magazine
p.fields.nano_signature = NANOZONE_SIGNATURE;
@@ -1765,15 +1844,16 @@
p.fields.nano_slot = 0;
p.fields.nano_offset = 0;
- for (j = 0; j < NANO_SLOT_SIZE; p.addr += SLOT_IN_BAND_SIZE, // Advance to next slot base
- ++j) {
+ for (j = 0; j < NANO_SLOT_SIZE;
+ p.addr += SLOT_IN_BAND_SIZE, // Advance to next slot base
+ ++j) {
nano_meta_admin_t pMeta = &nanozone->meta_data[i][j];
uintptr_t offset = pMeta->slot_bump_addr - p.addr;
if (0 == pMeta->slot_current_base_addr) { // Nothing allocated in this magazine for this slot?
continue;
} else {
- unsigned blocks_touched = offset_to_index(nanozone, pMeta, offset) - (unsigned)pMeta->slot_objects_skipped;
+ unsigned blocks_touched = offset_to_index(nanozone, pMeta, offset) - pMeta->slot_objects_skipped;
unsigned blocks_now_free = count_free(nanozone, pMeta);
unsigned blocks_in_use = blocks_touched - blocks_now_free;
@@ -1796,11 +1876,11 @@
{
int i;
- for (i = 0; i < nano_common_max_magazines; ++i) {
- if (_malloc_lock_trylock(&nanozone->band_resupply_lock[i])) {
- _malloc_lock_unlock(&nanozone->band_resupply_lock[i]);
- return TRUE;
- }
+ for (i = 0; i < nanozone->phys_ncpus; ++i) {
+ if (_malloc_lock_trylock(&nanozone->band_resupply_lock[i])) {
+ _malloc_lock_unlock(&nanozone->band_resupply_lock[i]);
+ return TRUE;
+ }
}
return FALSE;
}
@@ -1814,12 +1894,19 @@
}
static const struct malloc_introspection_t nano_introspect = {
- (void *)nano_ptr_in_use_enumerator, (void *)nano_good_size, (void *)nanozone_check, (void *)nano_print, nano_log,
- (void *)nano_force_lock, (void *)nano_force_unlock, (void *)nano_statistics, (void *)nano_locked, NULL, NULL, NULL,
- NULL, /* Zone enumeration version 7 and forward. */
- (void *)nano_reinit_lock, // reinit_lock version 9 and foward
+ (void *)nano_ptr_in_use_enumerator,
+ (void *)nano_good_size,
+ (void *)nanozone_check,
+ (void *)nano_print,
+ nano_log,
+ (void *)nano_force_lock,
+ (void *)nano_force_unlock,
+ (void *)nano_statistics,
+ (void *)nano_locked,
+ NULL, NULL, NULL, NULL, /* Zone enumeration version 7 and forward. */
}; // marked as const to spare the DATA section
+__attribute__((visibility("hidden")))
void
nano_forked_zone(nanozone_t *nanozone)
{
@@ -1843,114 +1930,122 @@
nanozone->basic_zone.batch_malloc = (void *)nano_forked_batch_malloc;
nanozone->basic_zone.batch_free = (void *)nano_forked_batch_free;
nanozone->basic_zone.introspect = (struct malloc_introspection_t *)&nano_introspect; /* Unchanged. */
- nanozone->basic_zone.memalign = (void *)nano_memalign; /* Unchanged. */
+ nanozone->basic_zone.memalign = (void *)nano_memalign; /* Unchanged. */
nanozone->basic_zone.free_definite_size = (void *)nano_forked_free_definite_size;
- nanozone->basic_zone.claimed_address = nano_forked_claimed_address;
mprotect(nanozone, sizeof(nanozone->basic_zone), PROT_READ);
-}
-
+
+}
+
+__attribute__((visibility("hidden")))
malloc_zone_t *
-nano_create_zone(malloc_zone_t *helper_zone, unsigned debug_flags)
-{
- nanozone_t *nanozone;
+create_nano_zone(size_t initial_size, malloc_zone_t *helper_zone, unsigned debug_flags)
+{
+ nanozone_t *nanozone;
int i, j;
- /* Note: It is important that nano_create_zone resets _malloc_engaged_nano
- * if it is unable to enable the nanozone (and chooses not to abort). As
- * several functions rely on _malloc_engaged_nano to determine if they
- * should manipulate the nanozone, and these should not run if we failed
- * to create the zone.
- */
- MALLOC_ASSERT(_malloc_engaged_nano == NANO_V1);
+ if (!_malloc_engaged_nano) return NULL;
+
+#if defined(__x86_64__)
+ if (_COMM_PAGE_VERSION_REQD > (*((uint16_t *)_COMM_PAGE_VERSION))) {
+ malloc_printf("*** FATAL ERROR - comm page version mismatch.\n");
+ exit(-1);
+ }
+#endif
/* get memory for the zone. */
- nanozone = nano_common_allocate_based_pages(NANOZONE_PAGED_SIZE, 0, 0, VM_MEMORY_MALLOC, 0);
- if (!nanozone) {
- _malloc_engaged_nano = NANO_NONE;
+ nanozone = allocate_pages(NULL, SZONE_PAGED_SIZE, 0, 0, VM_MEMORY_MALLOC);
+ if (!nanozone)
return NULL;
- }
/* set up the basic_zone portion of the nanozone structure */
- nanozone->basic_zone.version = 10;
+ nanozone->basic_zone.version = 8;
nanozone->basic_zone.size = (void *)nano_size;
- nanozone->basic_zone.malloc = (debug_flags & MALLOC_DO_SCRIBBLE) ? (void *)nano_malloc_scribble : (void *)nano_malloc;
+ nanozone->basic_zone.malloc = (debug_flags & SCALABLE_MALLOC_DO_SCRIBBLE) ? (void *)nano_malloc_scribble : (void *)nano_malloc;
nanozone->basic_zone.calloc = (void *)nano_calloc;
nanozone->basic_zone.valloc = (void *)nano_valloc;
- nanozone->basic_zone.free = (debug_flags & MALLOC_DO_SCRIBBLE) ? (void *)nano_free_scribble : (void *)nano_free;
+ nanozone->basic_zone.free = (debug_flags & SCALABLE_MALLOC_DO_SCRIBBLE) ? (void *)nano_free_scribble : (void *)nano_free;
nanozone->basic_zone.realloc = (void *)nano_realloc;
nanozone->basic_zone.destroy = (void *)nano_destroy;
nanozone->basic_zone.batch_malloc = (void *)nano_batch_malloc;
nanozone->basic_zone.batch_free = (void *)nano_batch_free;
nanozone->basic_zone.introspect = (struct malloc_introspection_t *)&nano_introspect;
nanozone->basic_zone.memalign = (void *)nano_memalign;
- nanozone->basic_zone.free_definite_size = (debug_flags & MALLOC_DO_SCRIBBLE) ? (void *)nano_free_definite_size_scribble
- : (void *)nano_free_definite_size;
-
+ nanozone->basic_zone.free_definite_size = (debug_flags & SCALABLE_MALLOC_DO_SCRIBBLE) ?
+ (void *)nano_free_definite_size_scribble : (void *)nano_free_definite_size;
+
nanozone->basic_zone.pressure_relief = (void *)nano_pressure_relief;
- nanozone->basic_zone.claimed_address = (void *)nano_claimed_address;
-
+
nanozone->basic_zone.reserved1 = 0; /* Set to zero once and for all as required by CFAllocator. */
nanozone->basic_zone.reserved2 = 0; /* Set to zero once and for all as required by CFAllocator. */
-
+
mprotect(nanozone, sizeof(nanozone->basic_zone), PROT_READ); /* Prevent overwriting the function pointers in basic_zone. */
-
- /* Nano zone does not support MALLOC_ADD_GUARD_PAGES. */
- if (debug_flags & MALLOC_ADD_GUARD_PAGES) {
- malloc_report(ASL_LEVEL_INFO, "nano zone does not support guard pages\n");
- debug_flags &= ~MALLOC_ADD_GUARD_PAGES;
- }
-
+
/* set up the remainder of the nanozone structure */
nanozone->debug_flags = debug_flags;
-
- if (phys_ncpus > sizeof(nanozone->core_mapped_size) /
- sizeof(nanozone->core_mapped_size[0])) {
- MALLOC_REPORT_FATAL_ERROR(phys_ncpus,
- "nanozone abandoned because NCPUS > max magazines.\n");
- }
-
+ nanozone->our_signature = NANOZONE_SIGNATURE;
+
+ /* Query the number of configured processors. */
+#if defined(__x86_64__)
+ nanozone->phys_ncpus = *(uint8_t *)(uintptr_t)_COMM_PAGE_PHYSICAL_CPUS;
+ nanozone->logical_ncpus = *(uint8_t *)(uintptr_t)_COMM_PAGE_LOGICAL_CPUS;
+#else
+#error Unknown architecture
+#endif
+
+ if (nanozone->phys_ncpus > sizeof(nanozone->core_mapped_size)/sizeof(nanozone->core_mapped_size[0])) {
+ _malloc_printf(ASL_LEVEL_NOTICE, "nano zone abandoned because NCPUS mismatch.\n");
+ return NULL;
+ }
+
+ if (0 != (nanozone->logical_ncpus % nanozone->phys_ncpus)) {
+ malloc_printf("*** FATAL ERROR - logical_ncpus % phys_ncpus != 0.\n");
+ exit(-1);
+ }
+
+ switch (nanozone->logical_ncpus/nanozone->phys_ncpus) {
+ case 1:
+ nanozone->hyper_shift = 0;
+ break;
+ case 2:
+ nanozone->hyper_shift = 1;
+ break;
+ case 4:
+ nanozone->hyper_shift = 2;
+ break;
+ default:
+ malloc_printf("*** FATAL ERROR - logical_ncpus / phys_ncpus not 1, 2, or 4.\n");
+ exit(-1);
+ }
+
/* Initialize slot queue heads and resupply locks. */
OSQueueHead q0 = OS_ATOMIC_QUEUE_INIT;
- for (i = 0; i < nano_common_max_magazines; ++i) {
- _malloc_lock_init(&nanozone->band_resupply_lock[i]);
-
+ for (i = 0; i < nanozone->phys_ncpus; ++i) {
+ _malloc_lock_init(&nanozone->band_resupply_lock[i]);
+
for (j = 0; j < NANO_SLOT_SIZE; ++j) {
nanozone->meta_data[i][j].slot_LIFO = q0;
}
}
-
+
/* Initialize the security token. */
+ if (0 == _dyld_get_image_slide((const struct mach_header*)_NSGetMachExecuteHeader())) {
+ // zero slide when ASLR has been disabled by boot-arg. Eliminate cloaking.
+ malloc_entropy[0] = 0;
+ malloc_entropy[1] = 0;
+ }
nanozone->cookie = (uintptr_t)malloc_entropy[0] & 0x0000ffffffff0000ULL; // scramble central 32bits with this cookie
-
+
+ /* Nano zone does not support SCALABLE_MALLOC_ADD_GUARD_PAGES. */
+ if (nanozone->debug_flags & SCALABLE_MALLOC_ADD_GUARD_PAGES) {
+ _malloc_printf(ASL_LEVEL_INFO, "nano zone does not support guard pages\n");
+ nanozone->debug_flags &= ~SCALABLE_MALLOC_ADD_GUARD_PAGES;
+ }
+
nanozone->helper_zone = helper_zone;
-
+
return (malloc_zone_t *)nanozone;
}
-
-void
-nano_init(const char *envp[], const char *apple[],
- const char *bootargs MALLOC_UNUSED)
-{
-#if NANO_PREALLOCATE_BAND_VM
- // Unconditionally preallocate the VA space set aside for nano malloc to
- // reserve it in all configurations. rdar://problem/33392283
- boolean_t preallocated = nano_preallocate_band_vm();
- if (!preallocated) {
- malloc_report(ASL_LEVEL_NOTICE, "nano zone abandoned due to inability to preallocate reserved vm space.\n");
- _malloc_engaged_nano = NANO_NONE;
- }
-#endif
-}
-
-// Second phase of initialization, called during _malloc_initialize(), after
-// environment variables have been read and processed.
-void
-nano_configure()
-{
- // Nothing to do.
-}
-
-#endif // CONFIG_NANOZONE
+#endif /* defined(__LP64__) */
/* vim: set noet:ts=4:sw=4:cindent: */