1 /* 2 * Bitmap Module 3 * 4 * Stolen from linux/src/lib/bitmap.c 5 * 6 * Copyright (C) 2010 Corentin Chary 7 * 8 * This source code is licensed under the GNU General Public License, 9 * Version 2. 10 */ 11 12 #include "qemu/osdep.h" 13 #include "qemu/bitops.h" 14 #include "qemu/bitmap.h" 15 #include "qemu/atomic.h" 16 17 /* 18 * bitmaps provide an array of bits, implemented using an 19 * array of unsigned longs. The number of valid bits in a 20 * given bitmap does _not_ need to be an exact multiple of 21 * BITS_PER_LONG. 22 * 23 * The possible unused bits in the last, partially used word 24 * of a bitmap are 'don't care'. The implementation makes 25 * no particular effort to keep them zero. It ensures that 26 * their value will not affect the results of any operation. 27 * The bitmap operations that return Boolean (bitmap_empty, 28 * for example) or scalar (bitmap_weight, for example) results 29 * carefully filter out these unused bits from impacting their 30 * results. 31 * 32 * These operations actually hold to a slightly stronger rule: 33 * if you don't input any bitmaps to these ops that have some 34 * unused bits set, then they won't output any set unused bits 35 * in output bitmaps. 36 * 37 * The byte ordering of bitmaps is more natural on little 38 * endian architectures. 39 */ 40 41 int slow_bitmap_empty(const unsigned long *bitmap, long bits) 42 { 43 long k, lim = bits/BITS_PER_LONG; 44 45 for (k = 0; k < lim; ++k) { 46 if (bitmap[k]) { 47 return 0; 48 } 49 } 50 if (bits % BITS_PER_LONG) { 51 if (bitmap[k] & BITMAP_LAST_WORD_MASK(bits)) { 52 return 0; 53 } 54 } 55 56 return 1; 57 } 58 59 int slow_bitmap_full(const unsigned long *bitmap, long bits) 60 { 61 long k, lim = bits/BITS_PER_LONG; 62 63 for (k = 0; k < lim; ++k) { 64 if (~bitmap[k]) { 65 return 0; 66 } 67 } 68 69 if (bits % BITS_PER_LONG) { 70 if (~bitmap[k] & BITMAP_LAST_WORD_MASK(bits)) { 71 return 0; 72 } 73 } 74 75 return 1; 76 } 77 78 int slow_bitmap_equal(const unsigned long *bitmap1, 79 const unsigned long *bitmap2, long bits) 80 { 81 long k, lim = bits/BITS_PER_LONG; 82 83 for (k = 0; k < lim; ++k) { 84 if (bitmap1[k] != bitmap2[k]) { 85 return 0; 86 } 87 } 88 89 if (bits % BITS_PER_LONG) { 90 if ((bitmap1[k] ^ bitmap2[k]) & BITMAP_LAST_WORD_MASK(bits)) { 91 return 0; 92 } 93 } 94 95 return 1; 96 } 97 98 void slow_bitmap_complement(unsigned long *dst, const unsigned long *src, 99 long bits) 100 { 101 long k, lim = bits/BITS_PER_LONG; 102 103 for (k = 0; k < lim; ++k) { 104 dst[k] = ~src[k]; 105 } 106 107 if (bits % BITS_PER_LONG) { 108 dst[k] = ~src[k] & BITMAP_LAST_WORD_MASK(bits); 109 } 110 } 111 112 int slow_bitmap_and(unsigned long *dst, const unsigned long *bitmap1, 113 const unsigned long *bitmap2, long bits) 114 { 115 long k; 116 long nr = BITS_TO_LONGS(bits); 117 unsigned long result = 0; 118 119 for (k = 0; k < nr; k++) { 120 result |= (dst[k] = bitmap1[k] & bitmap2[k]); 121 } 122 return result != 0; 123 } 124 125 void slow_bitmap_or(unsigned long *dst, const unsigned long *bitmap1, 126 const unsigned long *bitmap2, long bits) 127 { 128 long k; 129 long nr = BITS_TO_LONGS(bits); 130 131 for (k = 0; k < nr; k++) { 132 dst[k] = bitmap1[k] | bitmap2[k]; 133 } 134 } 135 136 void slow_bitmap_xor(unsigned long *dst, const unsigned long *bitmap1, 137 const unsigned long *bitmap2, long bits) 138 { 139 long k; 140 long nr = BITS_TO_LONGS(bits); 141 142 for (k = 0; k < nr; k++) { 143 dst[k] = bitmap1[k] ^ bitmap2[k]; 144 } 145 } 146 147 int slow_bitmap_andnot(unsigned long *dst, const unsigned long *bitmap1, 148 const unsigned long *bitmap2, long bits) 149 { 150 long k; 151 long nr = BITS_TO_LONGS(bits); 152 unsigned long result = 0; 153 154 for (k = 0; k < nr; k++) { 155 result |= (dst[k] = bitmap1[k] & ~bitmap2[k]); 156 } 157 return result != 0; 158 } 159 160 void bitmap_set(unsigned long *map, long start, long nr) 161 { 162 unsigned long *p = map + BIT_WORD(start); 163 const long size = start + nr; 164 int bits_to_set = BITS_PER_LONG - (start % BITS_PER_LONG); 165 unsigned long mask_to_set = BITMAP_FIRST_WORD_MASK(start); 166 167 assert(start >= 0 && nr >= 0); 168 169 while (nr - bits_to_set >= 0) { 170 *p |= mask_to_set; 171 nr -= bits_to_set; 172 bits_to_set = BITS_PER_LONG; 173 mask_to_set = ~0UL; 174 p++; 175 } 176 if (nr) { 177 mask_to_set &= BITMAP_LAST_WORD_MASK(size); 178 *p |= mask_to_set; 179 } 180 } 181 182 void bitmap_set_atomic(unsigned long *map, long start, long nr) 183 { 184 unsigned long *p = map + BIT_WORD(start); 185 const long size = start + nr; 186 int bits_to_set = BITS_PER_LONG - (start % BITS_PER_LONG); 187 unsigned long mask_to_set = BITMAP_FIRST_WORD_MASK(start); 188 189 assert(start >= 0 && nr >= 0); 190 191 /* First word */ 192 if (nr - bits_to_set > 0) { 193 qatomic_or(p, mask_to_set); 194 nr -= bits_to_set; 195 bits_to_set = BITS_PER_LONG; 196 mask_to_set = ~0UL; 197 p++; 198 } 199 200 /* Full words */ 201 if (bits_to_set == BITS_PER_LONG) { 202 while (nr >= BITS_PER_LONG) { 203 *p = ~0UL; 204 nr -= BITS_PER_LONG; 205 p++; 206 } 207 } 208 209 /* Last word */ 210 if (nr) { 211 mask_to_set &= BITMAP_LAST_WORD_MASK(size); 212 qatomic_or(p, mask_to_set); 213 } else { 214 /* If we avoided the full barrier in qatomic_or(), issue a 215 * barrier to account for the assignments in the while loop. 216 */ 217 smp_mb(); 218 } 219 } 220 221 void bitmap_clear(unsigned long *map, long start, long nr) 222 { 223 unsigned long *p = map + BIT_WORD(start); 224 const long size = start + nr; 225 int bits_to_clear = BITS_PER_LONG - (start % BITS_PER_LONG); 226 unsigned long mask_to_clear = BITMAP_FIRST_WORD_MASK(start); 227 228 assert(start >= 0 && nr >= 0); 229 230 while (nr - bits_to_clear >= 0) { 231 *p &= ~mask_to_clear; 232 nr -= bits_to_clear; 233 bits_to_clear = BITS_PER_LONG; 234 mask_to_clear = ~0UL; 235 p++; 236 } 237 if (nr) { 238 mask_to_clear &= BITMAP_LAST_WORD_MASK(size); 239 *p &= ~mask_to_clear; 240 } 241 } 242 243 bool bitmap_test_and_clear_atomic(unsigned long *map, long start, long nr) 244 { 245 unsigned long *p = map + BIT_WORD(start); 246 const long size = start + nr; 247 int bits_to_clear = BITS_PER_LONG - (start % BITS_PER_LONG); 248 unsigned long mask_to_clear = BITMAP_FIRST_WORD_MASK(start); 249 unsigned long dirty = 0; 250 unsigned long old_bits; 251 252 assert(start >= 0 && nr >= 0); 253 254 /* First word */ 255 if (nr - bits_to_clear > 0) { 256 old_bits = qatomic_fetch_and(p, ~mask_to_clear); 257 dirty |= old_bits & mask_to_clear; 258 nr -= bits_to_clear; 259 bits_to_clear = BITS_PER_LONG; 260 mask_to_clear = ~0UL; 261 p++; 262 } 263 264 /* Full words */ 265 if (bits_to_clear == BITS_PER_LONG) { 266 while (nr >= BITS_PER_LONG) { 267 if (*p) { 268 old_bits = qatomic_xchg(p, 0); 269 dirty |= old_bits; 270 } 271 nr -= BITS_PER_LONG; 272 p++; 273 } 274 } 275 276 /* Last word */ 277 if (nr) { 278 mask_to_clear &= BITMAP_LAST_WORD_MASK(size); 279 old_bits = qatomic_fetch_and(p, ~mask_to_clear); 280 dirty |= old_bits & mask_to_clear; 281 } else { 282 if (!dirty) { 283 smp_mb(); 284 } 285 } 286 287 return dirty != 0; 288 } 289 290 void bitmap_copy_and_clear_atomic(unsigned long *dst, unsigned long *src, 291 long nr) 292 { 293 while (nr > 0) { 294 *dst = qatomic_xchg(src, 0); 295 dst++; 296 src++; 297 nr -= BITS_PER_LONG; 298 } 299 } 300 301 #define ALIGN_MASK(x,mask) (((x)+(mask))&~(mask)) 302 303 /** 304 * bitmap_find_next_zero_area - find a contiguous aligned zero area 305 * @map: The address to base the search on 306 * @size: The bitmap size in bits 307 * @start: The bitnumber to start searching at 308 * @nr: The number of zeroed bits we're looking for 309 * @align_mask: Alignment mask for zero area 310 * 311 * The @align_mask should be one less than a power of 2; the effect is that 312 * the bit offset of all zero areas this function finds is multiples of that 313 * power of 2. A @align_mask of 0 means no alignment is required. 314 */ 315 unsigned long bitmap_find_next_zero_area(unsigned long *map, 316 unsigned long size, 317 unsigned long start, 318 unsigned long nr, 319 unsigned long align_mask) 320 { 321 unsigned long index, end, i; 322 again: 323 index = find_next_zero_bit(map, size, start); 324 325 /* Align allocation */ 326 index = ALIGN_MASK(index, align_mask); 327 328 end = index + nr; 329 if (end > size) { 330 return end; 331 } 332 i = find_next_bit(map, end, index); 333 if (i < end) { 334 start = i + 1; 335 goto again; 336 } 337 return index; 338 } 339 340 int slow_bitmap_intersects(const unsigned long *bitmap1, 341 const unsigned long *bitmap2, long bits) 342 { 343 long k, lim = bits/BITS_PER_LONG; 344 345 for (k = 0; k < lim; ++k) { 346 if (bitmap1[k] & bitmap2[k]) { 347 return 1; 348 } 349 } 350 351 if (bits % BITS_PER_LONG) { 352 if ((bitmap1[k] & bitmap2[k]) & BITMAP_LAST_WORD_MASK(bits)) { 353 return 1; 354 } 355 } 356 return 0; 357 } 358 359 long slow_bitmap_count_one(const unsigned long *bitmap, long nbits) 360 { 361 long k, lim = nbits / BITS_PER_LONG, result = 0; 362 363 for (k = 0; k < lim; k++) { 364 result += ctpopl(bitmap[k]); 365 } 366 367 if (nbits % BITS_PER_LONG) { 368 result += ctpopl(bitmap[k] & BITMAP_LAST_WORD_MASK(nbits)); 369 } 370 371 return result; 372 } 373 374 static void bitmap_to_from_le(unsigned long *dst, 375 const unsigned long *src, long nbits) 376 { 377 long len = BITS_TO_LONGS(nbits); 378 379 #ifdef HOST_WORDS_BIGENDIAN 380 long index; 381 382 for (index = 0; index < len; index++) { 383 # if HOST_LONG_BITS == 64 384 dst[index] = bswap64(src[index]); 385 # else 386 dst[index] = bswap32(src[index]); 387 # endif 388 } 389 #else 390 memcpy(dst, src, len * sizeof(unsigned long)); 391 #endif 392 } 393 394 void bitmap_from_le(unsigned long *dst, const unsigned long *src, 395 long nbits) 396 { 397 bitmap_to_from_le(dst, src, nbits); 398 } 399 400 void bitmap_to_le(unsigned long *dst, const unsigned long *src, 401 long nbits) 402 { 403 bitmap_to_from_le(dst, src, nbits); 404 } 405 406 /* 407 * Copy "src" bitmap with a positive offset and put it into the "dst" 408 * bitmap. The caller needs to make sure the bitmap size of "src" 409 * is bigger than (shift + nbits). 410 */ 411 void bitmap_copy_with_src_offset(unsigned long *dst, const unsigned long *src, 412 unsigned long shift, unsigned long nbits) 413 { 414 unsigned long left_mask, right_mask, last_mask; 415 416 /* Proper shift src pointer to the first word to copy from */ 417 src += BIT_WORD(shift); 418 shift %= BITS_PER_LONG; 419 420 if (!shift) { 421 /* Fast path */ 422 bitmap_copy(dst, src, nbits); 423 return; 424 } 425 426 right_mask = (1ul << shift) - 1; 427 left_mask = ~right_mask; 428 429 while (nbits >= BITS_PER_LONG) { 430 *dst = (*src & left_mask) >> shift; 431 *dst |= (src[1] & right_mask) << (BITS_PER_LONG - shift); 432 dst++; 433 src++; 434 nbits -= BITS_PER_LONG; 435 } 436 437 if (nbits > BITS_PER_LONG - shift) { 438 *dst = (*src & left_mask) >> shift; 439 nbits -= BITS_PER_LONG - shift; 440 last_mask = (1ul << nbits) - 1; 441 *dst |= (src[1] & last_mask) << (BITS_PER_LONG - shift); 442 } else if (nbits) { 443 last_mask = (1ul << nbits) - 1; 444 *dst = (*src >> shift) & last_mask; 445 } 446 } 447 448 /* 449 * Copy "src" bitmap into the "dst" bitmap with an offset in the 450 * "dst". The caller needs to make sure the bitmap size of "dst" is 451 * bigger than (shift + nbits). 452 */ 453 void bitmap_copy_with_dst_offset(unsigned long *dst, const unsigned long *src, 454 unsigned long shift, unsigned long nbits) 455 { 456 unsigned long left_mask, right_mask, last_mask; 457 458 /* Proper shift dst pointer to the first word to copy from */ 459 dst += BIT_WORD(shift); 460 shift %= BITS_PER_LONG; 461 462 if (!shift) { 463 /* Fast path */ 464 bitmap_copy(dst, src, nbits); 465 return; 466 } 467 468 right_mask = (1ul << (BITS_PER_LONG - shift)) - 1; 469 left_mask = ~right_mask; 470 471 *dst &= (1ul << shift) - 1; 472 while (nbits >= BITS_PER_LONG) { 473 *dst |= (*src & right_mask) << shift; 474 dst[1] = (*src & left_mask) >> (BITS_PER_LONG - shift); 475 dst++; 476 src++; 477 nbits -= BITS_PER_LONG; 478 } 479 480 if (nbits > BITS_PER_LONG - shift) { 481 *dst |= (*src & right_mask) << shift; 482 nbits -= BITS_PER_LONG - shift; 483 last_mask = ((1ul << nbits) - 1) << (BITS_PER_LONG - shift); 484 dst[1] = (*src & last_mask) >> (BITS_PER_LONG - shift); 485 } else if (nbits) { 486 last_mask = (1ul << nbits) - 1; 487 *dst |= (*src & last_mask) << shift; 488 } 489 } 490