Zero-copy FTP/HTTP Daemon compatible with all POSIX systems
| 1 | /* |
| 2 | MIT License |
| 3 | |
| 4 | Copyright (c) 2026 Seregon |
| 5 | |
| 6 | Permission is hereby granted, free of charge, to any person obtaining a copy |
| 7 | of this software and associated documentation files (the "Software"), to deal |
| 8 | in the Software without restriction, including without limitation the rights |
| 9 | to use, copy, modify, merge, publish, distribute, sublicense, and/or sell |
| 10 | copies of the Software, and to permit persons to whom the Software is |
| 11 | furnished to do so, subject to the following conditions: |
| 12 | |
| 13 | The above copyright notice and this permission notice shall be included in all |
| 14 | copies or substantial portions of the Software. |
| 15 | |
| 16 | THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR |
| 17 | IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, |
| 18 | FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE |
| 19 | AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER |
| 20 | LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, |
| 21 | OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE |
| 22 | SOFTWARE. |
| 23 | */ |
| 24 | /** |
| 25 | * @file pal_alloc.h |
| 26 | * @brief Unified memory allocation abstraction (malloc/free) |
| 27 | * |
| 28 | * @author Seregon |
| 29 | * @version 1.0.0 |
| 30 | * |
| 31 | * PLATFORMS: FreeBSD (PS4/PS5 kqueue), Linux (epoll) |
| 32 | * DESIGN: Single-threaded, non-blocking I/O |
| 33 | * |
| 34 | */ |
| 35 | #include "pal_alloc.h" |
| 36 | |
| 37 | #include <stdatomic.h> |
| 38 | #include <stdalign.h> |
| 39 | #include <string.h> |
| 40 | #include <stdlib.h> |
| 41 | |
| 42 | #ifndef PAL_ALLOC_DEFAULT_SIZE |
| 43 | #define PAL_ALLOC_DEFAULT_SIZE (8U * 1024U * 1024U) |
| 44 | #endif |
| 45 | |
| 46 | #ifndef PAL_ALLOC_MIN_ORDER |
| 47 | #define PAL_ALLOC_MIN_ORDER 5U |
| 48 | #endif |
| 49 | |
| 50 | #ifndef PAL_ALLOC_MAX_ORDER |
| 51 | #define PAL_ALLOC_MAX_ORDER 26U |
| 52 | #endif |
| 53 | |
| 54 | #define PAL_ALLOC_MAGIC 0xA11C0A7U |
| 55 | |
| 56 | #ifndef PAL_ALLOC_HARD_FAIL |
| 57 | #if defined(FTP_DEBUG) && (FTP_DEBUG != 0) |
| 58 | #define PAL_ALLOC_HARD_FAIL 1 |
| 59 | #else |
| 60 | #define PAL_ALLOC_HARD_FAIL 0 |
| 61 | #endif |
| 62 | #endif |
| 63 | |
| 64 | typedef struct { |
| 65 | uint32_t magic; |
| 66 | uint16_t order; |
| 67 | uint16_t used; |
| 68 | uint64_t pad; |
| 69 | } pal_alloc_hdr_t; |
| 70 | |
| 71 | typedef struct pal_alloc_free_node { |
| 72 | pal_alloc_hdr_t hdr; |
| 73 | struct pal_alloc_free_node *next; |
| 74 | } pal_alloc_free_node_t; |
| 75 | |
| 76 | static pal_allocator_t g_alloc = { |
| 77 | .base = NULL, |
| 78 | .size = 0U, |
| 79 | .max_order = 0U, |
| 80 | .lock = ATOMIC_FLAG_INIT, |
| 81 | .initialized = ATOMIC_VAR_INIT(0), |
| 82 | .alloc_calls = ATOMIC_VAR_INIT(0), |
| 83 | .free_calls = ATOMIC_VAR_INIT(0), |
| 84 | .calloc_calls = ATOMIC_VAR_INIT(0), |
| 85 | .realloc_calls = ATOMIC_VAR_INIT(0), |
| 86 | .aligned_calls = ATOMIC_VAR_INIT(0), |
| 87 | .failures = ATOMIC_VAR_INIT(0), |
| 88 | .bytes_in_use = ATOMIC_VAR_INIT(0), |
| 89 | .bytes_peak = ATOMIC_VAR_INIT(0), |
| 90 | .free_lists = {0}, |
| 91 | }; |
| 92 | |
| 93 | static _Alignas(4096) uint8_t g_alloc_default_buf[PAL_ALLOC_DEFAULT_SIZE]; |
| 94 | |
| 95 | #if defined(__clang__) || defined(__GNUC__) |
| 96 | #define PAL_ASSUME_ALIGNED(p, a) __builtin_assume_aligned((p), (a)) |
| 97 | #else |
| 98 | #define PAL_ASSUME_ALIGNED(p, a) (p) |
| 99 | #endif |
| 100 | |
| 101 | static void pal_alloc_lock(pal_allocator_t *a) |
| 102 | { |
| 103 | while (atomic_flag_test_and_set_explicit(&a->lock, memory_order_acquire)) { |
| 104 | } |
| 105 | } |
| 106 | |
| 107 | static void pal_alloc_unlock(pal_allocator_t *a) |
| 108 | { |
| 109 | atomic_flag_clear_explicit(&a->lock, memory_order_release); |
| 110 | } |
| 111 | |
| 112 | #if PAL_ALLOC_HARD_FAIL |
| 113 | static void pal_alloc_fail_fast(void) |
| 114 | { |
| 115 | #if defined(__clang__) || defined(__GNUC__) |
| 116 | __builtin_trap(); |
| 117 | #else |
| 118 | abort(); |
| 119 | #endif |
| 120 | } |
| 121 | #endif |
| 122 | |
| 123 | static uint32_t floor_log2_u64(uint64_t v) |
| 124 | { |
| 125 | uint32_t r = 0U; |
| 126 | while (v >>= 1U) { |
| 127 | r++; |
| 128 | } |
| 129 | return r; |
| 130 | } |
| 131 | |
| 132 | static uint32_t ceil_log2_u64(uint64_t v) |
| 133 | { |
| 134 | if (v <= 1U) { |
| 135 | return 0U; |
| 136 | } |
| 137 | uint32_t f = floor_log2_u64(v - 1U); |
| 138 | return f + 1U; |
| 139 | } |
| 140 | |
| 141 | static pal_alloc_free_node_t *list_pop(pal_allocator_t *a, uint32_t order) |
| 142 | { |
| 143 | uint32_t idx = order - PAL_ALLOC_MIN_ORDER; |
| 144 | pal_alloc_free_node_t *n = (pal_alloc_free_node_t *)a->free_lists[idx]; |
| 145 | if (n != NULL) { |
| 146 | a->free_lists[idx] = (void *)n->next; |
| 147 | n->next = NULL; |
| 148 | } |
| 149 | return n; |
| 150 | } |
| 151 | |
| 152 | static void list_push(pal_allocator_t *a, uint32_t order, pal_alloc_free_node_t *n) |
| 153 | { |
| 154 | uint32_t idx = order - PAL_ALLOC_MIN_ORDER; |
| 155 | n->next = (pal_alloc_free_node_t *)a->free_lists[idx]; |
| 156 | a->free_lists[idx] = (void *)n; |
| 157 | } |
| 158 | |
| 159 | static int list_remove(pal_allocator_t *a, uint32_t order, pal_alloc_free_node_t *target) |
| 160 | { |
| 161 | uint32_t idx = order - PAL_ALLOC_MIN_ORDER; |
| 162 | pal_alloc_free_node_t *prev = NULL; |
| 163 | pal_alloc_free_node_t *cur = (pal_alloc_free_node_t *)a->free_lists[idx]; |
| 164 | while (cur != NULL) { |
| 165 | if (cur == target) { |
| 166 | if (prev != NULL) { |
| 167 | prev->next = cur->next; |
| 168 | } else { |
| 169 | a->free_lists[idx] = (void *)cur->next; |
| 170 | } |
| 171 | cur->next = NULL; |
| 172 | return 1; |
| 173 | } |
| 174 | prev = cur; |
| 175 | cur = cur->next; |
| 176 | } |
| 177 | return 0; |
| 178 | } |
| 179 | |
| 180 | static void prefault_pages(uint8_t *base, size_t size) |
| 181 | { |
| 182 | if (base == NULL || size == 0U) { |
| 183 | return; |
| 184 | } |
| 185 | for (size_t i = 0U; i < size; i += 4096U) { |
| 186 | base[i] = 0U; |
| 187 | } |
| 188 | base[size - 1U] = 0U; |
| 189 | } |
| 190 | |
| 191 | static int ptr_in_arena(pal_allocator_t *a, const void *p) |
| 192 | { |
| 193 | if (p == NULL || a == NULL || a->base == NULL || a->size == 0U) { |
| 194 | return 0; |
| 195 | } |
| 196 | const uint8_t *bp = (const uint8_t *)p; |
| 197 | return (bp >= a->base) && (bp < (a->base + a->size)); |
| 198 | } |
| 199 | |
| 200 | void pal_allocator_get_stats(pal_allocator_t *a, pal_alloc_stats_t *out) |
| 201 | { |
| 202 | if (out == NULL) { |
| 203 | return; |
| 204 | } |
| 205 | memset(out, 0, sizeof(*out)); |
| 206 | if (a == NULL || atomic_load(&a->initialized) == 0) { |
| 207 | return; |
| 208 | } |
| 209 | out->alloc_calls = atomic_load(&a->alloc_calls); |
| 210 | out->free_calls = atomic_load(&a->free_calls); |
| 211 | out->calloc_calls = atomic_load(&a->calloc_calls); |
| 212 | out->realloc_calls = atomic_load(&a->realloc_calls); |
| 213 | out->aligned_calls = atomic_load(&a->aligned_calls); |
| 214 | out->failures = atomic_load(&a->failures); |
| 215 | out->bytes_in_use = atomic_load(&a->bytes_in_use); |
| 216 | out->bytes_peak = atomic_load(&a->bytes_peak); |
| 217 | } |
| 218 | |
| 219 | void pal_allocator_reset_stats(pal_allocator_t *a) |
| 220 | { |
| 221 | if (a == NULL) { |
| 222 | return; |
| 223 | } |
| 224 | atomic_store(&a->alloc_calls, 0U); |
| 225 | atomic_store(&a->free_calls, 0U); |
| 226 | atomic_store(&a->calloc_calls, 0U); |
| 227 | atomic_store(&a->realloc_calls, 0U); |
| 228 | atomic_store(&a->aligned_calls, 0U); |
| 229 | atomic_store(&a->failures, 0U); |
| 230 | atomic_store(&a->bytes_in_use, 0U); |
| 231 | atomic_store(&a->bytes_peak, 0U); |
| 232 | } |
| 233 | |
| 234 | int pal_allocator_init(pal_allocator_t *a, void *buffer, size_t size) |
| 235 | { |
| 236 | if (a == NULL || buffer == NULL) { |
| 237 | return -1; |
| 238 | } |
| 239 | if (size < (size_t)(1U << PAL_ALLOC_MIN_ORDER)) { |
| 240 | return -1; |
| 241 | } |
| 242 | |
| 243 | uintptr_t p = (uintptr_t)buffer; |
| 244 | uintptr_t aligned = (p + 15U) & ~(uintptr_t)15U; |
| 245 | size_t adj = (size_t)(aligned - p); |
| 246 | if (adj >= size) { |
| 247 | return -1; |
| 248 | } |
| 249 | uint8_t *base = (uint8_t *)aligned; |
| 250 | size_t usable = size - adj; |
| 251 | |
| 252 | uint32_t max_order = floor_log2_u64((uint64_t)usable); |
| 253 | if (max_order > PAL_ALLOC_MAX_ORDER) { |
| 254 | max_order = PAL_ALLOC_MAX_ORDER; |
| 255 | } |
| 256 | if (max_order < PAL_ALLOC_MIN_ORDER) { |
| 257 | return -1; |
| 258 | } |
| 259 | |
| 260 | size_t arena_size = (size_t)1U << max_order; |
| 261 | |
| 262 | pal_alloc_lock(a); |
| 263 | a->base = base; |
| 264 | a->size = arena_size; |
| 265 | a->max_order = max_order; |
| 266 | for (size_t i = 0U; i < (sizeof(a->free_lists) / sizeof(a->free_lists[0])); i++) { |
| 267 | a->free_lists[i] = NULL; |
| 268 | } |
| 269 | pal_allocator_reset_stats(a); |
| 270 | |
| 271 | prefault_pages(base, arena_size); |
| 272 | |
| 273 | pal_alloc_free_node_t *root = |
| 274 | (pal_alloc_free_node_t *)PAL_ASSUME_ALIGNED(base, alignof(pal_alloc_free_node_t)); |
| 275 | root->hdr.magic = PAL_ALLOC_MAGIC; |
| 276 | root->hdr.order = (uint16_t)max_order; |
| 277 | root->hdr.used = 0U; |
| 278 | root->hdr.pad = 0U; |
| 279 | root->next = NULL; |
| 280 | list_push(a, max_order, root); |
| 281 | |
| 282 | atomic_store(&a->initialized, 1); |
| 283 | pal_alloc_unlock(a); |
| 284 | |
| 285 | return 0; |
| 286 | } |
| 287 | |
| 288 | int pal_alloc_init(void *buffer, size_t size) |
| 289 | { |
| 290 | return pal_allocator_init(&g_alloc, buffer, size); |
| 291 | } |
| 292 | |
| 293 | int pal_alloc_init_default(void) |
| 294 | { |
| 295 | if (atomic_load(&g_alloc.initialized) != 0) { |
| 296 | return 0; |
| 297 | } |
| 298 | return pal_alloc_init(g_alloc_default_buf, sizeof(g_alloc_default_buf)); |
| 299 | } |
| 300 | |
| 301 | size_t pal_alloc_arena_size(void) |
| 302 | { |
| 303 | if (atomic_load(&g_alloc.initialized) == 0) { |
| 304 | return 0U; |
| 305 | } |
| 306 | return g_alloc.size; |
| 307 | } |
| 308 | |
| 309 | size_t pal_alloc_arena_free_approx(void) |
| 310 | { |
| 311 | if (atomic_load(&g_alloc.initialized) == 0) { |
| 312 | return 0U; |
| 313 | } |
| 314 | |
| 315 | pal_alloc_lock(&g_alloc); |
| 316 | size_t free_total = 0U; |
| 317 | for (uint32_t order = PAL_ALLOC_MIN_ORDER; order <= g_alloc.max_order; order++) { |
| 318 | uint32_t idx = order - PAL_ALLOC_MIN_ORDER; |
| 319 | pal_alloc_free_node_t *n = (pal_alloc_free_node_t *)g_alloc.free_lists[idx]; |
| 320 | while (n != NULL) { |
| 321 | free_total += (size_t)1U << order; |
| 322 | n = n->next; |
| 323 | } |
| 324 | } |
| 325 | pal_alloc_unlock(&g_alloc); |
| 326 | return free_total; |
| 327 | } |
| 328 | |
| 329 | void pal_alloc_get_stats(pal_alloc_stats_t *out) |
| 330 | { |
| 331 | pal_allocator_get_stats(&g_alloc, out); |
| 332 | } |
| 333 | |
| 334 | void pal_alloc_reset_stats(void) |
| 335 | { |
| 336 | pal_allocator_reset_stats(&g_alloc); |
| 337 | } |
| 338 | |
| 339 | static void bump_peak(pal_allocator_t *a, uint64_t in_use) |
| 340 | { |
| 341 | uint64_t peak = atomic_load(&a->bytes_peak); |
| 342 | while (in_use > peak) { |
| 343 | if (atomic_compare_exchange_weak(&a->bytes_peak, &peak, in_use)) { |
| 344 | break; |
| 345 | } |
| 346 | } |
| 347 | } |
| 348 | |
| 349 | static void *alloc_locked(pal_allocator_t *a, size_t size) |
| 350 | { |
| 351 | if (size == 0U) { |
| 352 | size = 1U; |
| 353 | } |
| 354 | |
| 355 | size_t need = size + sizeof(pal_alloc_hdr_t); |
| 356 | if (need > a->size) { |
| 357 | return NULL; |
| 358 | } |
| 359 | |
| 360 | uint32_t order = ceil_log2_u64((uint64_t)need); |
| 361 | if (order < PAL_ALLOC_MIN_ORDER) { |
| 362 | order = PAL_ALLOC_MIN_ORDER; |
| 363 | } |
| 364 | if (order > a->max_order) { |
| 365 | return NULL; |
| 366 | } |
| 367 | |
| 368 | uint32_t cur = order; |
| 369 | pal_alloc_free_node_t *node = NULL; |
| 370 | |
| 371 | while (cur <= a->max_order) { |
| 372 | node = list_pop(a, cur); |
| 373 | if (node != NULL) { |
| 374 | break; |
| 375 | } |
| 376 | cur++; |
| 377 | } |
| 378 | |
| 379 | if (node == NULL) { |
| 380 | return NULL; |
| 381 | } |
| 382 | |
| 383 | while (cur > order) { |
| 384 | cur--; |
| 385 | size_t half = (size_t)1U << cur; |
| 386 | uint8_t *b = (uint8_t *)node; |
| 387 | pal_alloc_free_node_t *buddy = |
| 388 | (pal_alloc_free_node_t *)PAL_ASSUME_ALIGNED((b + half), alignof(pal_alloc_free_node_t)); |
| 389 | |
| 390 | buddy->hdr.magic = PAL_ALLOC_MAGIC; |
| 391 | buddy->hdr.order = (uint16_t)cur; |
| 392 | buddy->hdr.used = 0U; |
| 393 | buddy->hdr.pad = 0U; |
| 394 | buddy->next = NULL; |
| 395 | list_push(a, cur, buddy); |
| 396 | |
| 397 | node->hdr.order = (uint16_t)cur; |
| 398 | } |
| 399 | |
| 400 | node->hdr.magic = PAL_ALLOC_MAGIC; |
| 401 | node->hdr.order = (uint16_t)order; |
| 402 | node->hdr.used = 1U; |
| 403 | node->hdr.pad = 0U; |
| 404 | |
| 405 | return (void *)((uint8_t *)node + sizeof(pal_alloc_hdr_t)); |
| 406 | } |
| 407 | |
| 408 | void *pal_allocator_malloc(pal_allocator_t *a, size_t size) |
| 409 | { |
| 410 | if (a == NULL) { |
| 411 | return NULL; |
| 412 | } |
| 413 | if (atomic_load(&a->initialized) == 0) { |
| 414 | return NULL; |
| 415 | } |
| 416 | |
| 417 | pal_alloc_lock(a); |
| 418 | void *p = alloc_locked(a, size); |
| 419 | if (p != NULL) { |
| 420 | atomic_fetch_add(&a->alloc_calls, 1U); |
| 421 | pal_alloc_hdr_t *hdr = |
| 422 | (pal_alloc_hdr_t *)PAL_ASSUME_ALIGNED(((uint8_t *)p - sizeof(pal_alloc_hdr_t)), |
| 423 | alignof(pal_alloc_hdr_t)); |
| 424 | uint64_t bytes = (uint64_t)((size_t)1U << (uint32_t)hdr->order); |
| 425 | uint64_t in_use = atomic_fetch_add(&a->bytes_in_use, bytes) + bytes; |
| 426 | bump_peak(a, in_use); |
| 427 | } else { |
| 428 | atomic_fetch_add(&a->failures, 1U); |
| 429 | } |
| 430 | pal_alloc_unlock(a); |
| 431 | return p; |
| 432 | } |
| 433 | |
| 434 | static void free_locked(pal_allocator_t *a, void *ptr) |
| 435 | { |
| 436 | if (ptr == NULL) { |
| 437 | return; |
| 438 | } |
| 439 | |
| 440 | uint8_t *p = (uint8_t *)ptr; |
| 441 | pal_alloc_hdr_t *hdr = |
| 442 | (pal_alloc_hdr_t *)PAL_ASSUME_ALIGNED((p - sizeof(pal_alloc_hdr_t)), alignof(pal_alloc_hdr_t)); |
| 443 | if (hdr->magic != PAL_ALLOC_MAGIC || hdr->used == 0U) { |
| 444 | if (((uintptr_t)ptr % alignof(void *)) == 0U) { |
| 445 | void *raw = *((void **)ptr - 1); |
| 446 | if (ptr_in_arena(a, raw)) { |
| 447 | uint8_t *rp = (uint8_t *)raw; |
| 448 | pal_alloc_hdr_t *rh = |
| 449 | (pal_alloc_hdr_t *)PAL_ASSUME_ALIGNED((rp - sizeof(pal_alloc_hdr_t)), |
| 450 | alignof(pal_alloc_hdr_t)); |
| 451 | if ((rh->magic == PAL_ALLOC_MAGIC) && (rh->used != 0U)) { |
| 452 | hdr = rh; |
| 453 | } else { |
| 454 | #if PAL_ALLOC_HARD_FAIL |
| 455 | pal_alloc_fail_fast(); |
| 456 | #endif |
| 457 | return; |
| 458 | } |
| 459 | } else { |
| 460 | #if PAL_ALLOC_HARD_FAIL |
| 461 | pal_alloc_fail_fast(); |
| 462 | #endif |
| 463 | return; |
| 464 | } |
| 465 | } else { |
| 466 | #if PAL_ALLOC_HARD_FAIL |
| 467 | pal_alloc_fail_fast(); |
| 468 | #endif |
| 469 | return; |
| 470 | } |
| 471 | } |
| 472 | |
| 473 | uint32_t order = (uint32_t)hdr->order; |
| 474 | if (order < PAL_ALLOC_MIN_ORDER || order > a->max_order) { |
| 475 | #if PAL_ALLOC_HARD_FAIL |
| 476 | pal_alloc_fail_fast(); |
| 477 | #endif |
| 478 | return; |
| 479 | } |
| 480 | |
| 481 | uint64_t bytes = (uint64_t)((size_t)1U << order); |
| 482 | atomic_fetch_add(&a->free_calls, 1U); |
| 483 | atomic_fetch_sub(&a->bytes_in_use, bytes); |
| 484 | |
| 485 | hdr->used = 0U; |
| 486 | pal_alloc_free_node_t *node = (pal_alloc_free_node_t *)hdr; |
| 487 | node->next = NULL; |
| 488 | |
| 489 | size_t block_size = (size_t)1U << order; |
| 490 | size_t offset = (size_t)((uint8_t *)node - a->base); |
| 491 | |
| 492 | while (order < a->max_order) { |
| 493 | size_t buddy_offset = offset ^ block_size; |
| 494 | if (buddy_offset >= a->size) { |
| 495 | break; |
| 496 | } |
| 497 | |
| 498 | pal_alloc_free_node_t *buddy = |
| 499 | (pal_alloc_free_node_t *)PAL_ASSUME_ALIGNED((a->base + buddy_offset), |
| 500 | alignof(pal_alloc_free_node_t)); |
| 501 | if (buddy->hdr.magic != PAL_ALLOC_MAGIC || buddy->hdr.used != 0U) { |
| 502 | break; |
| 503 | } |
| 504 | if ((uint32_t)buddy->hdr.order != order) { |
| 505 | break; |
| 506 | } |
| 507 | |
| 508 | if (list_remove(a, order, buddy) == 0) { |
| 509 | break; |
| 510 | } |
| 511 | |
| 512 | if (buddy_offset < offset) { |
| 513 | node = buddy; |
| 514 | offset = buddy_offset; |
| 515 | } |
| 516 | |
| 517 | order++; |
| 518 | block_size <<= 1U; |
| 519 | node->hdr.order = (uint16_t)order; |
| 520 | node->hdr.used = 0U; |
| 521 | node->hdr.pad = 0U; |
| 522 | node->next = NULL; |
| 523 | } |
| 524 | |
| 525 | list_push(a, order, node); |
| 526 | } |
| 527 | |
| 528 | void pal_allocator_free(pal_allocator_t *a, void *ptr) |
| 529 | { |
| 530 | if (ptr == NULL) { |
| 531 | return; |
| 532 | } |
| 533 | if (a == NULL || atomic_load(&a->initialized) == 0) { |
| 534 | return; |
| 535 | } |
| 536 | pal_alloc_lock(a); |
| 537 | free_locked(a, ptr); |
| 538 | pal_alloc_unlock(a); |
| 539 | } |
| 540 | |
| 541 | void *pal_malloc(size_t size) |
| 542 | { |
| 543 | if (atomic_load(&g_alloc.initialized) == 0) { |
| 544 | if (pal_alloc_init_default() != 0) { |
| 545 | return NULL; |
| 546 | } |
| 547 | } |
| 548 | return pal_allocator_malloc(&g_alloc, size); |
| 549 | } |
| 550 | |
| 551 | void pal_free(void *ptr) |
| 552 | { |
| 553 | if (atomic_load(&g_alloc.initialized) == 0) { |
| 554 | return; |
| 555 | } |
| 556 | pal_allocator_free(&g_alloc, ptr); |
| 557 | } |
| 558 | |
| 559 | void *pal_allocator_calloc(pal_allocator_t *a, size_t nmemb, size_t size) |
| 560 | { |
| 561 | if (a == NULL) { |
| 562 | return NULL; |
| 563 | } |
| 564 | atomic_fetch_add(&a->calloc_calls, 1U); |
| 565 | |
| 566 | if (nmemb == 0U || size == 0U) { |
| 567 | return pal_allocator_malloc(a, 0U); |
| 568 | } |
| 569 | if (nmemb > (SIZE_MAX / size)) { |
| 570 | return NULL; |
| 571 | } |
| 572 | size_t total = nmemb * size; |
| 573 | void *p = pal_allocator_malloc(a, total); |
| 574 | if (p != NULL) { |
| 575 | memset(p, 0, total); |
| 576 | } |
| 577 | return p; |
| 578 | } |
| 579 | |
| 580 | void *pal_calloc(size_t nmemb, size_t size) |
| 581 | { |
| 582 | if (atomic_load(&g_alloc.initialized) == 0) { |
| 583 | if (pal_alloc_init_default() != 0) { |
| 584 | return NULL; |
| 585 | } |
| 586 | } |
| 587 | return pal_allocator_calloc(&g_alloc, nmemb, size); |
| 588 | } |
| 589 | |
| 590 | void *pal_allocator_realloc(pal_allocator_t *a, void *ptr, size_t size) |
| 591 | { |
| 592 | if (a == NULL) { |
| 593 | return NULL; |
| 594 | } |
| 595 | atomic_fetch_add(&a->realloc_calls, 1U); |
| 596 | |
| 597 | if (ptr == NULL) { |
| 598 | return pal_allocator_malloc(a, size); |
| 599 | } |
| 600 | if (size == 0U) { |
| 601 | pal_allocator_free(a, ptr); |
| 602 | return NULL; |
| 603 | } |
| 604 | if (atomic_load(&a->initialized) == 0) { |
| 605 | return NULL; |
| 606 | } |
| 607 | |
| 608 | uint8_t *p = (uint8_t *)ptr; |
| 609 | pal_alloc_hdr_t *hdr = |
| 610 | (pal_alloc_hdr_t *)PAL_ASSUME_ALIGNED((p - sizeof(pal_alloc_hdr_t)), alignof(pal_alloc_hdr_t)); |
| 611 | if (hdr->magic != PAL_ALLOC_MAGIC || hdr->used == 0U) { |
| 612 | #if PAL_ALLOC_HARD_FAIL |
| 613 | pal_alloc_fail_fast(); |
| 614 | #endif |
| 615 | return NULL; |
| 616 | } |
| 617 | |
| 618 | uint32_t order = (uint32_t)hdr->order; |
| 619 | size_t cap = ((size_t)1U << order) - sizeof(pal_alloc_hdr_t); |
| 620 | if (size <= cap) { |
| 621 | return ptr; |
| 622 | } |
| 623 | |
| 624 | void *n = pal_allocator_malloc(a, size); |
| 625 | if (n == NULL) { |
| 626 | return NULL; |
| 627 | } |
| 628 | memcpy(n, ptr, cap); |
| 629 | pal_allocator_free(a, ptr); |
| 630 | return n; |
| 631 | } |
| 632 | |
| 633 | void *pal_realloc(void *ptr, size_t size) |
| 634 | { |
| 635 | if (atomic_load(&g_alloc.initialized) == 0) { |
| 636 | if (pal_alloc_init_default() != 0) { |
| 637 | return NULL; |
| 638 | } |
| 639 | } |
| 640 | return pal_allocator_realloc(&g_alloc, ptr, size); |
| 641 | } |
| 642 | |
| 643 | void *pal_allocator_aligned_alloc(pal_allocator_t *a, size_t alignment, size_t size) |
| 644 | { |
| 645 | if (a == NULL) { |
| 646 | return NULL; |
| 647 | } |
| 648 | atomic_fetch_add(&a->aligned_calls, 1U); |
| 649 | |
| 650 | if (alignment < sizeof(void *)) { |
| 651 | alignment = sizeof(void *); |
| 652 | } |
| 653 | if ((alignment & (alignment - 1U)) != 0U) { |
| 654 | return NULL; |
| 655 | } |
| 656 | |
| 657 | size_t need = size + alignment + sizeof(void *); |
| 658 | void *raw = pal_allocator_malloc(a, need); |
| 659 | if (raw == NULL) { |
| 660 | return NULL; |
| 661 | } |
| 662 | |
| 663 | uintptr_t start = (uintptr_t)raw + sizeof(void *); |
| 664 | uintptr_t aligned = (start + (alignment - 1U)) & ~(uintptr_t)(alignment - 1U); |
| 665 | void **slot = (void **)(aligned - sizeof(void *)); |
| 666 | *slot = raw; |
| 667 | return (void *)aligned; |
| 668 | } |
| 669 | |
| 670 | void *pal_aligned_alloc(size_t alignment, size_t size) |
| 671 | { |
| 672 | if (atomic_load(&g_alloc.initialized) == 0) { |
| 673 | if (pal_alloc_init_default() != 0) { |
| 674 | return NULL; |
| 675 | } |
| 676 | } |
| 677 | return pal_allocator_aligned_alloc(&g_alloc, alignment, size); |
| 678 | } |
| 679 | |
| 680 | int pal_allocator_posix_memalign(pal_allocator_t *a, void **memptr, size_t alignment, size_t size) |
| 681 | { |
| 682 | if (memptr == NULL) { |
| 683 | return 22; |
| 684 | } |
| 685 | void *p = pal_allocator_aligned_alloc(a, alignment, size); |
| 686 | if (p == NULL) { |
| 687 | *memptr = NULL; |
| 688 | return 12; |
| 689 | } |
| 690 | *memptr = p; |
| 691 | return 0; |
| 692 | } |
| 693 | |
| 694 | int pal_posix_memalign(void **memptr, size_t alignment, size_t size) |
| 695 | { |
| 696 | if (atomic_load(&g_alloc.initialized) == 0) { |
| 697 | if (pal_alloc_init_default() != 0) { |
| 698 | return 12; |
| 699 | } |
| 700 | } |
| 701 | return pal_allocator_posix_memalign(&g_alloc, memptr, alignment, size); |
| 702 | } |
| 703 |