#define inline inline __attribute__((always_inline))
#endif
-uintptr_t __brk(uintptr_t);
void *__mmap(void *, size_t, int, int, int, off_t);
int __munmap(void *, size_t);
void *__mremap(void *, size_t, size_t, int, ...);
};
struct bin {
- int lock[2];
+ volatile int lock[2];
struct chunk *head;
struct chunk *tail;
};
static struct {
- uintptr_t brk;
- size_t *heap;
- uint64_t binmap;
+ volatile uint64_t binmap;
struct bin bins[64];
- int brk_lock[2];
- int free_lock[2];
+ volatile int free_lock[2];
} mal;
static inline void lock(volatile int *lk)
{
- if (!libc.threads_minus_1) return;
- while(a_swap(lk, 1)) __wait(lk, lk+1, 1, 1);
+ if (libc.threads_minus_1)
+ while(a_swap(lk, 1)) __wait(lk, lk+1, 1, 1);
}
static inline void unlock(volatile int *lk)
{
- if (!libc.threads_minus_1) return;
- a_store(lk, 0);
- if (lk[1]) __wake(lk, 1, 1);
+ if (lk[0]) {
+ a_store(lk, 0);
+ if (lk[1]) __wake(lk, 1, 1);
+ }
}
static inline void lock_bin(int i)
{
- if (libc.threads_minus_1)
- lock(mal.bins[i].lock);
+ lock(mal.bins[i].lock);
if (!mal.bins[i].head)
mal.bins[i].head = mal.bins[i].tail = BIN_TO_CHUNK(i);
}
static inline void unlock_bin(int i)
{
- if (!libc.threads_minus_1) return;
unlock(mal.bins[i].lock);
}
#endif
}
+static const unsigned char bin_tab[60] = {
+ 32,33,34,35,36,36,37,37,38,38,39,39,
+ 40,40,40,40,41,41,41,41,42,42,42,42,43,43,43,43,
+ 44,44,44,44,44,44,44,44,45,45,45,45,45,45,45,45,
+ 46,46,46,46,46,46,46,46,47,47,47,47,47,47,47,47,
+};
+
static int bin_index(size_t x)
{
x = x / SIZE_ALIGN - 1;
if (x <= 32) return x;
+ if (x < 512) return bin_tab[x/8-4];
if (x > 0x1c00) return 63;
- return ((union { float v; uint32_t r; }){(int)x}.r>>21) - 496;
+ return bin_tab[x/128-4] + 16;
}
static int bin_index_up(size_t x)
{
x = x / SIZE_ALIGN - 1;
if (x <= 32) return x;
- return ((union { float v; uint32_t r; }){(int)x}.r+0x1fffff>>21) - 496;
+ x--;
+ if (x < 512) return bin_tab[x/8-4] + 1;
+ return bin_tab[x/128-4] + 17;
}
#if 0
}
#endif
+void *__expand_heap(size_t *);
+
static struct chunk *expand_heap(size_t n)
{
+ static int heap_lock[2];
+ static void *end;
+ void *p;
struct chunk *w;
- uintptr_t new;
-
- lock(mal.brk_lock);
- if (n > SIZE_MAX - mal.brk - 2*PAGE_SIZE) goto fail;
- new = mal.brk + n + SIZE_ALIGN + PAGE_SIZE - 1 & -PAGE_SIZE;
- n = new - mal.brk;
+ /* The argument n already accounts for the caller's chunk
+ * overhead needs, but if the heap can't be extended in-place,
+ * we need room for an extra zero-sized sentinel chunk. */
+ n += SIZE_ALIGN;
- if (__brk(new) != new) goto fail;
+ lock(heap_lock);
- w = MEM_TO_CHUNK(new);
- w->psize = n | C_INUSE;
- w->csize = 0 | C_INUSE;
-
- w = MEM_TO_CHUNK(mal.brk);
- w->csize = n | C_INUSE;
- mal.brk = new;
-
- unlock(mal.brk_lock);
-
- return w;
-fail:
- unlock(mal.brk_lock);
- return 0;
-}
-
-static int init_malloc(size_t n)
-{
- static int init, waiters;
- int state;
- struct chunk *c;
-
- if (init == 2) return 0;
-
- while ((state=a_swap(&init, 1)) == 1)
- __wait(&init, &waiters, 1, 1);
- if (state) {
- a_store(&init, 2);
+ p = __expand_heap(&n);
+ if (!p) {
+ unlock(heap_lock);
return 0;
}
- mal.brk = __brk(0);
-#ifdef SHARED
- mal.brk = mal.brk + PAGE_SIZE-1 & -PAGE_SIZE;
-#endif
- mal.brk = mal.brk + 2*SIZE_ALIGN-1 & -SIZE_ALIGN;
+ /* If not just expanding existing space, we need to make a
+ * new sentinel chunk below the allocated space. */
+ if (p != end) {
+ /* Valid/safe because of the prologue increment. */
+ n -= SIZE_ALIGN;
+ p = (char *)p + SIZE_ALIGN;
+ w = MEM_TO_CHUNK(p);
+ w->psize = 0 | C_INUSE;
+ }
- c = expand_heap(n);
+ /* Record new heap end and fill in footer. */
+ end = (char *)p + n;
+ w = MEM_TO_CHUNK(end);
+ w->psize = n | C_INUSE;
+ w->csize = 0 | C_INUSE;
- if (!c) {
- a_store(&init, 0);
- if (waiters) __wake(&init, 1, 1);
- return -1;
- }
+ /* Fill in header, which may be new or may be replacing a
+ * zero-size sentinel header at the old end-of-heap. */
+ w = MEM_TO_CHUNK(p);
+ w->csize = n | C_INUSE;
- mal.heap = (void *)c;
- c->psize = 0 | C_INUSE;
- free(CHUNK_TO_MEM(c));
+ unlock(heap_lock);
- a_store(&init, 2);
- if (waiters) __wake(&init, -1, 1);
- return 1;
+ return w;
}
static int adjust_size(size_t *n)
for (;;) {
uint64_t mask = mal.binmap & -(1ULL<<i);
if (!mask) {
- if (init_malloc(n) > 0) continue;
c = expand_heap(n);
if (!c) return 0;
if (alloc_rev(c)) {
j = first_set(mask);
lock_bin(j);
c = mal.bins[j].head;
- if (c != BIN_TO_CHUNK(j) && j == bin_index(c->csize)) {
+ if (c != BIN_TO_CHUNK(j)) {
if (!pretrim(c, n, i, j)) unbin(c, j);
unlock_bin(j);
break;
return CHUNK_TO_MEM(c);
}
+void *__malloc0(size_t n)
+{
+ void *p = malloc(n);
+ if (p && !IS_MMAPPED(MEM_TO_CHUNK(p))) {
+ size_t *z;
+ n = (n + sizeof *z - 1)/sizeof *z;
+ for (z=p; n; n--, z++) if (*z) *z=0;
+ }
+ return p;
+}
+
void *realloc(void *p, size_t n)
{
struct chunk *self, *next;
if (oldlen == newlen) return p;
base = __mremap(base, oldlen, newlen, MREMAP_MAYMOVE);
if (base == (void *)-1)
- return newlen < oldlen ? p : 0;
+ goto copy_realloc;
self = (void *)(base + extra);
self->csize = newlen - extra;
return CHUNK_TO_MEM(self);
return CHUNK_TO_MEM(self);
}
+copy_realloc:
/* As a last resort, allocate a new chunk and copy to it. */
new = malloc(n-OVERHEAD);
if (!new) return 0;
void free(void *p)
{
- struct chunk *self = MEM_TO_CHUNK(p);
- struct chunk *next;
+ struct chunk *self, *next;
size_t final_size, new_size, size;
int reclaim=0;
int i;
if (!p) return;
+ self = MEM_TO_CHUNK(p);
+
if (IS_MMAPPED(self)) {
size_t extra = self->psize;
char *base = (char *)self - extra;
if (next->psize != self->csize) a_crash();
for (;;) {
- /* Replace middle of large chunks with fresh zero pages */
- if (reclaim && (self->psize & next->csize & C_INUSE)) {
- uintptr_t a = (uintptr_t)self + SIZE_ALIGN+PAGE_SIZE-1 & -PAGE_SIZE;
- uintptr_t b = (uintptr_t)next - SIZE_ALIGN & -PAGE_SIZE;
-#if 1
- __madvise((void *)a, b-a, MADV_DONTNEED);
-#else
- __mmap((void *)a, b-a, PROT_READ|PROT_WRITE,
- MAP_PRIVATE|MAP_ANONYMOUS|MAP_FIXED, -1, 0);
-#endif
- }
-
if (self->psize & next->csize & C_INUSE) {
self->csize = final_size | C_INUSE;
next->psize = final_size | C_INUSE;
}
}
+ if (!(mal.binmap & 1ULL<<i))
+ a_or_64(&mal.binmap, 1ULL<<i);
+
self->csize = final_size;
next->psize = final_size;
unlock(mal.free_lock);
self->next->prev = self;
self->prev->next = self;
- if (!(mal.binmap & 1ULL<<i))
- a_or_64(&mal.binmap, 1ULL<<i);
+ /* Replace middle of large chunks with fresh zero pages */
+ if (reclaim) {
+ uintptr_t a = (uintptr_t)self + SIZE_ALIGN+PAGE_SIZE-1 & -PAGE_SIZE;
+ uintptr_t b = (uintptr_t)next - SIZE_ALIGN & -PAGE_SIZE;
+#if 1
+ __madvise((void *)a, b-a, MADV_DONTNEED);
+#else
+ __mmap((void *)a, b-a, PROT_READ|PROT_WRITE,
+ MAP_PRIVATE|MAP_ANONYMOUS|MAP_FIXED, -1, 0);
+#endif
+ }
unlock_bin(i);
}