summaryrefslogtreecommitdiff
path: root/src/malloc/malloc.c
blob: b90636cc67fda06dac201526d43a42f5e040971c (plain) (blame)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
#define _GNU_SOURCE
#include <stdlib.h>
#include <string.h>
#include <limits.h>
#include <stdint.h>
#include <errno.h>
#include <sys/mman.h>
#include "libc.h"
#include "atomic.h"
#include "pthread_impl.h"

#if defined(__GNUC__) && defined(__PIC__)
#define inline inline __attribute__((always_inline))
#endif

void *__mmap(void *, size_t, int, int, int, off_t);
int __munmap(void *, size_t);
void *__mremap(void *, size_t, size_t, int, ...);
int __madvise(void *, size_t, int);

struct chunk {
	size_t psize, csize;
	struct chunk *next, *prev;
};

struct bin {
	volatile int lock[2];
	struct chunk *head;
	struct chunk *tail;
};

static struct {
	volatile uint64_t binmap;
	struct bin bins[64];
	volatile int free_lock[2];
} mal;


#define SIZE_ALIGN (4*sizeof(size_t))
#define SIZE_MASK (-SIZE_ALIGN)
#define OVERHEAD (2*sizeof(size_t))
#define MMAP_THRESHOLD (0x1c00*SIZE_ALIGN)
#define DONTCARE 16
#define RECLAIM 163840

#define CHUNK_SIZE(c) ((c)->csize & -2)
#define CHUNK_PSIZE(c) ((c)->psize & -2)
#define PREV_CHUNK(c) ((struct chunk *)((char *)(c) - CHUNK_PSIZE(c)))
#define NEXT_CHUNK(c) ((struct chunk *)((char *)(c) + CHUNK_SIZE(c)))
#define MEM_TO_CHUNK(p) (struct chunk *)((char *)(p) - OVERHEAD)
#define CHUNK_TO_MEM(c) (void *)((char *)(c) + OVERHEAD)
#define BIN_TO_CHUNK(i) (MEM_TO_CHUNK(&mal.bins[i].head))

#define C_INUSE  ((size_t)1)

#define IS_MMAPPED(c) !((c)->csize & (C_INUSE))


/* Synchronization tools */

static inline void lock(volatile int *lk)
{
	if (libc.threads_minus_1)
		while(a_swap(lk, 1)) __wait(lk, lk+1, 1, 1);
}

static inline void unlock(volatile int *lk)
{
	if (lk[0]) {
		a_store(lk, 0);
		if (lk[1]) __wake(lk, 1, 1);
	}
}

static inline void lock_bin(int i)
{
	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)
{
	unlock(mal.bins[i].lock);
}

static int first_set(uint64_t x)
{
#if 1
	return a_ctz_64(x);
#else
	static const char debruijn64[64] = {
		0, 1, 2, 53, 3, 7, 54, 27, 4, 38, 41, 8, 34, 55, 48, 28,
		62, 5, 39, 46, 44, 42, 22, 9, 24, 35, 59, 56, 49, 18, 29, 11,
		63, 52, 6, 26, 37, 40, 33, 47, 61, 45, 43, 21, 23, 58, 17, 10,
		51, 25, 36, 32, 60, 20, 57, 16, 50, 31, 19, 15, 30, 14, 13, 12
	};
	static const char debruijn32[32] = {
		0, 1, 23, 2, 29, 24, 19, 3, 30, 27, 25, 11, 20, 8, 4, 13,
		31, 22, 28, 18, 26, 10, 7, 12, 21, 17, 9, 6, 16, 5, 15, 14
	};
	if (sizeof(long) < 8) {
		uint32_t y = x;
		if (!y) {
			y = x>>32;
			return 32 + debruijn32[(y&-y)*0x076be629 >> 27];
		}
		return debruijn32[(y&-y)*0x076be629 >> 27];
	}
	return debruijn64[(x&-x)*0x022fdd63cc95386dull >> 58];
#endif
}

static int bin_index(size_t x)
{
	x = x / SIZE_ALIGN - 1;
	if (x <= 32) return x;
	if (x > 0x1c00) return 63;
	return ((union { float v; uint32_t r; }){(int)x}.r>>21) - 496;
}

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;
}

#if 0
void __dump_heap(int x)
{
	struct chunk *c;
	int i;
	for (c = (void *)mal.heap; CHUNK_SIZE(c); c = NEXT_CHUNK(c))
		fprintf(stderr, "base %p size %zu (%d) flags %d/%d\n",
			c, CHUNK_SIZE(c), bin_index(CHUNK_SIZE(c)),
			c->csize & 15,
			NEXT_CHUNK(c)->psize & 15);
	for (i=0; i<64; i++) {
		if (mal.bins[i].head != BIN_TO_CHUNK(i) && mal.bins[i].head) {
			fprintf(stderr, "bin %d: %p\n", i, mal.bins[i].head);
			if (!(mal.binmap & 1ULL<<i))
				fprintf(stderr, "missing from binmap!\n");
		} else if (mal.binmap & 1ULL<<i)
			fprintf(stderr, "binmap wrongly contains %d!\n", i);
	}
}
#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;

	/* 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;

	lock(heap_lock);

	p = __expand_heap(&n);
	if (!p) {
		unlock(heap_lock);
		return 0;
	}

	/* 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;
	}

	/* 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;

	/* 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;

	unlock(heap_lock);

	return w;
}

static int adjust_size(size_t *n)
{
	/* Result of pointer difference must fit in ptrdiff_t. */
	if (*n-1 > PTRDIFF_MAX - SIZE_ALIGN - PAGE_SIZE) {
		if (*n) {
			errno = ENOMEM;
			return -1;
		} else {
			*n = SIZE_ALIGN;
			return 0;
		}
	}
	*n = (*n + OVERHEAD + SIZE_ALIGN - 1) & SIZE_MASK;
	return 0;
}

static void unbin(struct chunk *c, int i)
{
	if (c->prev == c->next)
		a_and_64(&mal.binmap, ~(1ULL<<i));
	c->prev->next = c->next;
	c->next->prev = c->prev;
	c->csize |= C_INUSE;
	NEXT_CHUNK(c)->psize |= C_INUSE;
}

static int alloc_fwd(struct chunk *c)
{
	int i;
	size_t k;
	while (!((k=c->csize) & C_INUSE)) {
		i = bin_index(k);
		lock_bin(i);
		if (c->csize == k) {
			unbin(c, i);
			unlock_bin(i);
			return 1;
		}
		unlock_bin(i);
	}
	return 0;
}

static int alloc_rev(struct chunk *c)
{
	int i;
	size_t k;
	while (!((k=c->psize) & C_INUSE)) {
		i = bin_index(k);
		lock_bin(i);
		if (c->psize == k) {
			unbin(PREV_CHUNK(c), i);
			unlock_bin(i);
			return 1;
		}
		unlock_bin(i);
	}
	return 0;
}


/* pretrim - trims a chunk _prior_ to removing it from its bin.
 * Must be called with i as the ideal bin for size n, j the bin
 * for the _free_ chunk self, and bin j locked. */
static int pretrim(struct chunk *self, size_t n, int i, int j)
{
	size_t n1;
	struct chunk *next, *split;

	/* We cannot pretrim if it would require re-binning. */
	if (j < 40) return 0;
	if (j < i+3) {
		if (j != 63) return 0;
		n1 = CHUNK_SIZE(self);
		if (n1-n <= MMAP_THRESHOLD) return 0;
	} else {
		n1 = CHUNK_SIZE(self);
	}
	if (bin_index(n1-n) != j) return 0;

	next = NEXT_CHUNK(self);
	split = (void *)((char *)self + n);

	split->prev = self->prev;
	split->next = self->next;
	split->prev->next = split;
	split->next->prev = split;
	split->psize = n | C_INUSE;
	split->csize = n1-n;
	next->psize = n1-n;
	self->csize = n | C_INUSE;
	return 1;
}

static void trim(struct chunk *self, size_t n)
{
	size_t n1 = CHUNK_SIZE(self);
	struct chunk *next, *split;

	if (n >= n1 - DONTCARE) return;

	next = NEXT_CHUNK(self);
	split = (void *)((char *)self + n);

	split->psize = n | C_INUSE;
	split->csize = n1-n | C_INUSE;
	next->psize = n1-n | C_INUSE;
	self->csize = n | C_INUSE;

	free(CHUNK_TO_MEM(split));
}

void *malloc(size_t n)
{
	struct chunk *c;
	int i, j;

	if (adjust_size(&n) < 0) return 0;

	if (n > MMAP_THRESHOLD) {
		size_t len = n + OVERHEAD + PAGE_SIZE - 1 & -PAGE_SIZE;
		char *base = __mmap(0, len, PROT_READ|PROT_WRITE,
			MAP_PRIVATE|MAP_ANONYMOUS, -1, 0);
		if (base == (void *)-1) return 0;
		c = (void *)(base + SIZE_ALIGN - OVERHEAD);
		c->csize = len - (SIZE_ALIGN - OVERHEAD);
		c->psize = SIZE_ALIGN - OVERHEAD;
		return CHUNK_TO_MEM(c);
	}

	i = bin_index_up(n);
	for (;;) {
		uint64_t mask = mal.binmap & -(1ULL<<i);
		if (!mask) {
			c = expand_heap(n);
			if (!c) return 0;
			if (alloc_rev(c)) {
				struct chunk *x = c;
				c = PREV_CHUNK(c);
				NEXT_CHUNK(x)->psize = c->csize =
					x->csize + CHUNK_SIZE(c);
			}
			break;
		}
		j = first_set(mask);
		lock_bin(j);
		c = mal.bins[j].head;
		if (c != BIN_TO_CHUNK(j)) {
			if (!pretrim(c, n, i, j)) unbin(c, j);
			unlock_bin(j);
			break;
		}
		unlock_bin(j);
	}

	/* Now patch up in case we over-allocated */
	trim(c, n);

	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;
	size_t n0, n1;
	void *new;

	if (!p) return malloc(n);

	if (adjust_size(&n) < 0) return 0;

	self = MEM_TO_CHUNK(p);
	n1 = n0 = CHUNK_SIZE(self);

	if (IS_MMAPPED(self)) {
		size_t extra = self->psize;
		char *base = (char *)self - extra;
		size_t oldlen = n0 + extra;
		size_t newlen = n + extra;
		/* Crash on realloc of freed chunk */
		if (extra & 1) a_crash();
		if (newlen < PAGE_SIZE && (new = malloc(n))) {
			memcpy(new, p, n-OVERHEAD);
			free(p);
			return new;
		}
		newlen = (newlen + PAGE_SIZE-1) & -PAGE_SIZE;
		if (oldlen == newlen) return p;
		base = __mremap(base, oldlen, newlen, MREMAP_MAYMOVE);
		if (base == (void *)-1)
			return newlen < oldlen ? p : 0;
		self = (void *)(base + extra);
		self->csize = newlen - extra;
		return CHUNK_TO_MEM(self);
	}

	next = NEXT_CHUNK(self);

	/* Crash on corrupted footer (likely from buffer overflow) */
	if (next->psize != self->csize) a_crash();

	/* Merge adjacent chunks if we need more space. This is not
	 * a waste of time even if we fail to get enough space, because our
	 * subsequent call to free would otherwise have to do the merge. */
	if (n > n1 && alloc_fwd(next)) {
		n1 += CHUNK_SIZE(next);
		next = NEXT_CHUNK(next);
	}
	/* FIXME: find what's wrong here and reenable it..? */
	if (0 && n > n1 && alloc_rev(self)) {
		self = PREV_CHUNK(self);
		n1 += CHUNK_SIZE(self);
	}
	self->csize = n1 | C_INUSE;
	next->psize = n1 | C_INUSE;

	/* If we got enough space, split off the excess and return */
	if (n <= n1) {
		//memmove(CHUNK_TO_MEM(self), p, n0-OVERHEAD);
		trim(self, n);
		return CHUNK_TO_MEM(self);
	}

	/* As a last resort, allocate a new chunk and copy to it. */
	new = malloc(n-OVERHEAD);
	if (!new) return 0;
	memcpy(new, p, n0-OVERHEAD);
	free(CHUNK_TO_MEM(self));
	return new;
}

void free(void *p)
{
	struct chunk *self = MEM_TO_CHUNK(p);
	struct chunk *next;
	size_t final_size, new_size, size;
	int reclaim=0;
	int i;

	if (!p) return;

	if (IS_MMAPPED(self)) {
		size_t extra = self->psize;
		char *base = (char *)self - extra;
		size_t len = CHUNK_SIZE(self) + extra;
		/* Crash on double free */
		if (extra & 1) a_crash();
		__munmap(base, len);
		return;
	}

	final_size = new_size = CHUNK_SIZE(self);
	next = NEXT_CHUNK(self);

	/* Crash on corrupted footer (likely from buffer overflow) */
	if (next->psize != self->csize) a_crash();

	for (;;) {
		if (self->psize & next->csize & C_INUSE) {
			self->csize = final_size | C_INUSE;
			next->psize = final_size | C_INUSE;
			i = bin_index(final_size);
			lock_bin(i);
			lock(mal.free_lock);
			if (self->psize & next->csize & C_INUSE)
				break;
			unlock(mal.free_lock);
			unlock_bin(i);
		}

		if (alloc_rev(self)) {
			self = PREV_CHUNK(self);
			size = CHUNK_SIZE(self);
			final_size += size;
			if (new_size+size > RECLAIM && (new_size+size^size) > size)
				reclaim = 1;
		}

		if (alloc_fwd(next)) {
			size = CHUNK_SIZE(next);
			final_size += size;
			if (new_size+size > RECLAIM && (new_size+size^size) > size)
				reclaim = 1;
			next = NEXT_CHUNK(next);
		}
	}

	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 = BIN_TO_CHUNK(i);
	self->prev = mal.bins[i].tail;
	self->next->prev = self;
	self->prev->next = self;

	/* 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);
}