1 /* 2 * L2/refcount table cache for the QCOW2 format 3 * 4 * Copyright (c) 2010 Kevin Wolf <kwolf@redhat.com> 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 14 * all 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 19 * THE 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 22 * THE SOFTWARE. 23 */ 24 25 /* Needed for CONFIG_MADVISE */ 26 #include "config-host.h" 27 28 #if defined(CONFIG_MADVISE) || defined(CONFIG_POSIX_MADVISE) 29 #include <sys/mman.h> 30 #endif 31 32 #include "block/block_int.h" 33 #include "qemu-common.h" 34 #include "qemu/osdep.h" 35 #include "qcow2.h" 36 #include "trace.h" 37 38 typedef struct Qcow2CachedTable { 39 int64_t offset; 40 uint64_t lru_counter; 41 int ref; 42 bool dirty; 43 } Qcow2CachedTable; 44 45 struct Qcow2Cache { 46 Qcow2CachedTable *entries; 47 struct Qcow2Cache *depends; 48 int size; 49 bool depends_on_flush; 50 void *table_array; 51 uint64_t lru_counter; 52 uint64_t cache_clean_lru_counter; 53 }; 54 55 static inline void *qcow2_cache_get_table_addr(BlockDriverState *bs, 56 Qcow2Cache *c, int table) 57 { 58 BDRVQcowState *s = bs->opaque; 59 return (uint8_t *) c->table_array + (size_t) table * s->cluster_size; 60 } 61 62 static inline int qcow2_cache_get_table_idx(BlockDriverState *bs, 63 Qcow2Cache *c, void *table) 64 { 65 BDRVQcowState *s = bs->opaque; 66 ptrdiff_t table_offset = (uint8_t *) table - (uint8_t *) c->table_array; 67 int idx = table_offset / s->cluster_size; 68 assert(idx >= 0 && idx < c->size && table_offset % s->cluster_size == 0); 69 return idx; 70 } 71 72 static void qcow2_cache_table_release(BlockDriverState *bs, Qcow2Cache *c, 73 int i, int num_tables) 74 { 75 #if QEMU_MADV_DONTNEED != QEMU_MADV_INVALID 76 BDRVQcowState *s = bs->opaque; 77 void *t = qcow2_cache_get_table_addr(bs, c, i); 78 int align = getpagesize(); 79 size_t mem_size = (size_t) s->cluster_size * num_tables; 80 size_t offset = QEMU_ALIGN_UP((uintptr_t) t, align) - (uintptr_t) t; 81 size_t length = QEMU_ALIGN_DOWN(mem_size - offset, align); 82 if (length > 0) { 83 qemu_madvise((uint8_t *) t + offset, length, QEMU_MADV_DONTNEED); 84 } 85 #endif 86 } 87 88 static inline bool can_clean_entry(Qcow2Cache *c, int i) 89 { 90 Qcow2CachedTable *t = &c->entries[i]; 91 return t->ref == 0 && !t->dirty && t->offset != 0 && 92 t->lru_counter <= c->cache_clean_lru_counter; 93 } 94 95 void qcow2_cache_clean_unused(BlockDriverState *bs, Qcow2Cache *c) 96 { 97 int i = 0; 98 while (i < c->size) { 99 int to_clean = 0; 100 101 /* Skip the entries that we don't need to clean */ 102 while (i < c->size && !can_clean_entry(c, i)) { 103 i++; 104 } 105 106 /* And count how many we can clean in a row */ 107 while (i < c->size && can_clean_entry(c, i)) { 108 c->entries[i].offset = 0; 109 c->entries[i].lru_counter = 0; 110 i++; 111 to_clean++; 112 } 113 114 if (to_clean > 0) { 115 qcow2_cache_table_release(bs, c, i - to_clean, to_clean); 116 } 117 } 118 119 c->cache_clean_lru_counter = c->lru_counter; 120 } 121 122 Qcow2Cache *qcow2_cache_create(BlockDriverState *bs, int num_tables) 123 { 124 BDRVQcowState *s = bs->opaque; 125 Qcow2Cache *c; 126 127 c = g_new0(Qcow2Cache, 1); 128 c->size = num_tables; 129 c->entries = g_try_new0(Qcow2CachedTable, num_tables); 130 c->table_array = qemu_try_blockalign(bs->file, 131 (size_t) num_tables * s->cluster_size); 132 133 if (!c->entries || !c->table_array) { 134 qemu_vfree(c->table_array); 135 g_free(c->entries); 136 g_free(c); 137 c = NULL; 138 } 139 140 return c; 141 } 142 143 int qcow2_cache_destroy(BlockDriverState *bs, Qcow2Cache *c) 144 { 145 int i; 146 147 for (i = 0; i < c->size; i++) { 148 assert(c->entries[i].ref == 0); 149 } 150 151 qemu_vfree(c->table_array); 152 g_free(c->entries); 153 g_free(c); 154 155 return 0; 156 } 157 158 static int qcow2_cache_flush_dependency(BlockDriverState *bs, Qcow2Cache *c) 159 { 160 int ret; 161 162 ret = qcow2_cache_flush(bs, c->depends); 163 if (ret < 0) { 164 return ret; 165 } 166 167 c->depends = NULL; 168 c->depends_on_flush = false; 169 170 return 0; 171 } 172 173 static int qcow2_cache_entry_flush(BlockDriverState *bs, Qcow2Cache *c, int i) 174 { 175 BDRVQcowState *s = bs->opaque; 176 int ret = 0; 177 178 if (!c->entries[i].dirty || !c->entries[i].offset) { 179 return 0; 180 } 181 182 trace_qcow2_cache_entry_flush(qemu_coroutine_self(), 183 c == s->l2_table_cache, i); 184 185 if (c->depends) { 186 ret = qcow2_cache_flush_dependency(bs, c); 187 } else if (c->depends_on_flush) { 188 ret = bdrv_flush(bs->file); 189 if (ret >= 0) { 190 c->depends_on_flush = false; 191 } 192 } 193 194 if (ret < 0) { 195 return ret; 196 } 197 198 if (c == s->refcount_block_cache) { 199 ret = qcow2_pre_write_overlap_check(bs, QCOW2_OL_REFCOUNT_BLOCK, 200 c->entries[i].offset, s->cluster_size); 201 } else if (c == s->l2_table_cache) { 202 ret = qcow2_pre_write_overlap_check(bs, QCOW2_OL_ACTIVE_L2, 203 c->entries[i].offset, s->cluster_size); 204 } else { 205 ret = qcow2_pre_write_overlap_check(bs, 0, 206 c->entries[i].offset, s->cluster_size); 207 } 208 209 if (ret < 0) { 210 return ret; 211 } 212 213 if (c == s->refcount_block_cache) { 214 BLKDBG_EVENT(bs->file, BLKDBG_REFBLOCK_UPDATE_PART); 215 } else if (c == s->l2_table_cache) { 216 BLKDBG_EVENT(bs->file, BLKDBG_L2_UPDATE); 217 } 218 219 ret = bdrv_pwrite(bs->file, c->entries[i].offset, 220 qcow2_cache_get_table_addr(bs, c, i), s->cluster_size); 221 if (ret < 0) { 222 return ret; 223 } 224 225 c->entries[i].dirty = false; 226 227 return 0; 228 } 229 230 int qcow2_cache_flush(BlockDriverState *bs, Qcow2Cache *c) 231 { 232 BDRVQcowState *s = bs->opaque; 233 int result = 0; 234 int ret; 235 int i; 236 237 trace_qcow2_cache_flush(qemu_coroutine_self(), c == s->l2_table_cache); 238 239 for (i = 0; i < c->size; i++) { 240 ret = qcow2_cache_entry_flush(bs, c, i); 241 if (ret < 0 && result != -ENOSPC) { 242 result = ret; 243 } 244 } 245 246 if (result == 0) { 247 ret = bdrv_flush(bs->file); 248 if (ret < 0) { 249 result = ret; 250 } 251 } 252 253 return result; 254 } 255 256 int qcow2_cache_set_dependency(BlockDriverState *bs, Qcow2Cache *c, 257 Qcow2Cache *dependency) 258 { 259 int ret; 260 261 if (dependency->depends) { 262 ret = qcow2_cache_flush_dependency(bs, dependency); 263 if (ret < 0) { 264 return ret; 265 } 266 } 267 268 if (c->depends && (c->depends != dependency)) { 269 ret = qcow2_cache_flush_dependency(bs, c); 270 if (ret < 0) { 271 return ret; 272 } 273 } 274 275 c->depends = dependency; 276 return 0; 277 } 278 279 void qcow2_cache_depends_on_flush(Qcow2Cache *c) 280 { 281 c->depends_on_flush = true; 282 } 283 284 int qcow2_cache_empty(BlockDriverState *bs, Qcow2Cache *c) 285 { 286 int ret, i; 287 288 ret = qcow2_cache_flush(bs, c); 289 if (ret < 0) { 290 return ret; 291 } 292 293 for (i = 0; i < c->size; i++) { 294 assert(c->entries[i].ref == 0); 295 c->entries[i].offset = 0; 296 c->entries[i].lru_counter = 0; 297 } 298 299 qcow2_cache_table_release(bs, c, 0, c->size); 300 301 c->lru_counter = 0; 302 303 return 0; 304 } 305 306 static int qcow2_cache_do_get(BlockDriverState *bs, Qcow2Cache *c, 307 uint64_t offset, void **table, bool read_from_disk) 308 { 309 BDRVQcowState *s = bs->opaque; 310 int i; 311 int ret; 312 int lookup_index; 313 uint64_t min_lru_counter = UINT64_MAX; 314 int min_lru_index = -1; 315 316 trace_qcow2_cache_get(qemu_coroutine_self(), c == s->l2_table_cache, 317 offset, read_from_disk); 318 319 /* Check if the table is already cached */ 320 i = lookup_index = (offset / s->cluster_size * 4) % c->size; 321 do { 322 const Qcow2CachedTable *t = &c->entries[i]; 323 if (t->offset == offset) { 324 goto found; 325 } 326 if (t->ref == 0 && t->lru_counter < min_lru_counter) { 327 min_lru_counter = t->lru_counter; 328 min_lru_index = i; 329 } 330 if (++i == c->size) { 331 i = 0; 332 } 333 } while (i != lookup_index); 334 335 if (min_lru_index == -1) { 336 /* This can't happen in current synchronous code, but leave the check 337 * here as a reminder for whoever starts using AIO with the cache */ 338 abort(); 339 } 340 341 /* Cache miss: write a table back and replace it */ 342 i = min_lru_index; 343 trace_qcow2_cache_get_replace_entry(qemu_coroutine_self(), 344 c == s->l2_table_cache, i); 345 346 ret = qcow2_cache_entry_flush(bs, c, i); 347 if (ret < 0) { 348 return ret; 349 } 350 351 trace_qcow2_cache_get_read(qemu_coroutine_self(), 352 c == s->l2_table_cache, i); 353 c->entries[i].offset = 0; 354 if (read_from_disk) { 355 if (c == s->l2_table_cache) { 356 BLKDBG_EVENT(bs->file, BLKDBG_L2_LOAD); 357 } 358 359 ret = bdrv_pread(bs->file, offset, qcow2_cache_get_table_addr(bs, c, i), 360 s->cluster_size); 361 if (ret < 0) { 362 return ret; 363 } 364 } 365 366 c->entries[i].offset = offset; 367 368 /* And return the right table */ 369 found: 370 c->entries[i].ref++; 371 *table = qcow2_cache_get_table_addr(bs, c, i); 372 373 trace_qcow2_cache_get_done(qemu_coroutine_self(), 374 c == s->l2_table_cache, i); 375 376 return 0; 377 } 378 379 int qcow2_cache_get(BlockDriverState *bs, Qcow2Cache *c, uint64_t offset, 380 void **table) 381 { 382 return qcow2_cache_do_get(bs, c, offset, table, true); 383 } 384 385 int qcow2_cache_get_empty(BlockDriverState *bs, Qcow2Cache *c, uint64_t offset, 386 void **table) 387 { 388 return qcow2_cache_do_get(bs, c, offset, table, false); 389 } 390 391 void qcow2_cache_put(BlockDriverState *bs, Qcow2Cache *c, void **table) 392 { 393 int i = qcow2_cache_get_table_idx(bs, c, *table); 394 395 c->entries[i].ref--; 396 *table = NULL; 397 398 if (c->entries[i].ref == 0) { 399 c->entries[i].lru_counter = ++c->lru_counter; 400 } 401 402 assert(c->entries[i].ref >= 0); 403 } 404 405 void qcow2_cache_entry_mark_dirty(BlockDriverState *bs, Qcow2Cache *c, 406 void *table) 407 { 408 int i = qcow2_cache_get_table_idx(bs, c, table); 409 assert(c->entries[i].offset != 0); 410 c->entries[i].dirty = true; 411 } 412