1 /* 2 * Block driver for the QCOW version 2 format 3 * 4 * Copyright (c) 2004-2006 Fabrice Bellard 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 #ifndef BLOCK_QCOW2_H 26 #define BLOCK_QCOW2_H 27 28 #include "qemu/aes.h" 29 #include "block/coroutine.h" 30 31 //#define DEBUG_ALLOC 32 //#define DEBUG_ALLOC2 33 //#define DEBUG_EXT 34 35 #define QCOW_MAGIC (('Q' << 24) | ('F' << 16) | ('I' << 8) | 0xfb) 36 37 #define QCOW_CRYPT_NONE 0 38 #define QCOW_CRYPT_AES 1 39 40 #define QCOW_MAX_CRYPT_CLUSTERS 32 41 42 /* indicate that the refcount of the referenced cluster is exactly one. */ 43 #define QCOW_OFLAG_COPIED (1LL << 63) 44 /* indicate that the cluster is compressed (they never have the copied flag) */ 45 #define QCOW_OFLAG_COMPRESSED (1LL << 62) 46 /* The cluster reads as all zeros */ 47 #define QCOW_OFLAG_ZERO (1LL << 0) 48 49 #define REFCOUNT_SHIFT 1 /* refcount size is 2 bytes */ 50 51 #define MIN_CLUSTER_BITS 9 52 #define MAX_CLUSTER_BITS 21 53 54 #define L2_CACHE_SIZE 16 55 56 /* Must be at least 4 to cover all cases of refcount table growth */ 57 #define REFCOUNT_CACHE_SIZE 4 58 59 #define DEFAULT_CLUSTER_SIZE 65536 60 61 62 #define QCOW2_OPT_LAZY_REFCOUNTS "lazy_refcounts" 63 64 typedef struct QCowHeader { 65 uint32_t magic; 66 uint32_t version; 67 uint64_t backing_file_offset; 68 uint32_t backing_file_size; 69 uint32_t cluster_bits; 70 uint64_t size; /* in bytes */ 71 uint32_t crypt_method; 72 uint32_t l1_size; /* XXX: save number of clusters instead ? */ 73 uint64_t l1_table_offset; 74 uint64_t refcount_table_offset; 75 uint32_t refcount_table_clusters; 76 uint32_t nb_snapshots; 77 uint64_t snapshots_offset; 78 79 /* The following fields are only valid for version >= 3 */ 80 uint64_t incompatible_features; 81 uint64_t compatible_features; 82 uint64_t autoclear_features; 83 84 uint32_t refcount_order; 85 uint32_t header_length; 86 } QCowHeader; 87 88 typedef struct QCowSnapshot { 89 uint64_t l1_table_offset; 90 uint32_t l1_size; 91 char *id_str; 92 char *name; 93 uint64_t disk_size; 94 uint64_t vm_state_size; 95 uint32_t date_sec; 96 uint32_t date_nsec; 97 uint64_t vm_clock_nsec; 98 } QCowSnapshot; 99 100 struct Qcow2Cache; 101 typedef struct Qcow2Cache Qcow2Cache; 102 103 typedef struct Qcow2UnknownHeaderExtension { 104 uint32_t magic; 105 uint32_t len; 106 QLIST_ENTRY(Qcow2UnknownHeaderExtension) next; 107 uint8_t data[]; 108 } Qcow2UnknownHeaderExtension; 109 110 enum { 111 QCOW2_FEAT_TYPE_INCOMPATIBLE = 0, 112 QCOW2_FEAT_TYPE_COMPATIBLE = 1, 113 QCOW2_FEAT_TYPE_AUTOCLEAR = 2, 114 }; 115 116 /* Incompatible feature bits */ 117 enum { 118 QCOW2_INCOMPAT_DIRTY_BITNR = 0, 119 QCOW2_INCOMPAT_DIRTY = 1 << QCOW2_INCOMPAT_DIRTY_BITNR, 120 121 QCOW2_INCOMPAT_MASK = QCOW2_INCOMPAT_DIRTY, 122 }; 123 124 /* Compatible feature bits */ 125 enum { 126 QCOW2_COMPAT_LAZY_REFCOUNTS_BITNR = 0, 127 QCOW2_COMPAT_LAZY_REFCOUNTS = 1 << QCOW2_COMPAT_LAZY_REFCOUNTS_BITNR, 128 129 QCOW2_COMPAT_FEAT_MASK = QCOW2_COMPAT_LAZY_REFCOUNTS, 130 }; 131 132 typedef struct Qcow2Feature { 133 uint8_t type; 134 uint8_t bit; 135 char name[46]; 136 } QEMU_PACKED Qcow2Feature; 137 138 typedef struct BDRVQcowState { 139 int cluster_bits; 140 int cluster_size; 141 int cluster_sectors; 142 int l2_bits; 143 int l2_size; 144 int l1_size; 145 int l1_vm_state_index; 146 int csize_shift; 147 int csize_mask; 148 uint64_t cluster_offset_mask; 149 uint64_t l1_table_offset; 150 uint64_t *l1_table; 151 152 Qcow2Cache* l2_table_cache; 153 Qcow2Cache* refcount_block_cache; 154 155 uint8_t *cluster_cache; 156 uint8_t *cluster_data; 157 uint64_t cluster_cache_offset; 158 QLIST_HEAD(QCowClusterAlloc, QCowL2Meta) cluster_allocs; 159 160 uint64_t *refcount_table; 161 uint64_t refcount_table_offset; 162 uint32_t refcount_table_size; 163 int64_t free_cluster_index; 164 int64_t free_byte_offset; 165 166 CoMutex lock; 167 168 uint32_t crypt_method; /* current crypt method, 0 if no key yet */ 169 uint32_t crypt_method_header; 170 AES_KEY aes_encrypt_key; 171 AES_KEY aes_decrypt_key; 172 uint64_t snapshots_offset; 173 int snapshots_size; 174 int nb_snapshots; 175 QCowSnapshot *snapshots; 176 177 int flags; 178 int qcow_version; 179 bool use_lazy_refcounts; 180 181 uint64_t incompatible_features; 182 uint64_t compatible_features; 183 uint64_t autoclear_features; 184 185 size_t unknown_header_fields_size; 186 void* unknown_header_fields; 187 QLIST_HEAD(, Qcow2UnknownHeaderExtension) unknown_header_ext; 188 } BDRVQcowState; 189 190 /* XXX: use std qcow open function ? */ 191 typedef struct QCowCreateState { 192 int cluster_size; 193 int cluster_bits; 194 uint16_t *refcount_block; 195 uint64_t *refcount_table; 196 int64_t l1_table_offset; 197 int64_t refcount_table_offset; 198 int64_t refcount_block_offset; 199 } QCowCreateState; 200 201 struct QCowAIOCB; 202 203 typedef struct Qcow2COWRegion { 204 /** 205 * Offset of the COW region in bytes from the start of the first cluster 206 * touched by the request. 207 */ 208 uint64_t offset; 209 210 /** Number of sectors to copy */ 211 int nb_sectors; 212 } Qcow2COWRegion; 213 214 /** 215 * Describes an in-flight (part of a) write request that writes to clusters 216 * that are not referenced in their L2 table yet. 217 */ 218 typedef struct QCowL2Meta 219 { 220 /** Guest offset of the first newly allocated cluster */ 221 uint64_t offset; 222 223 /** Host offset of the first newly allocated cluster */ 224 uint64_t alloc_offset; 225 226 /** 227 * Number of sectors from the start of the first allocated cluster to 228 * the end of the (possibly shortened) request 229 */ 230 int nb_available; 231 232 /** Number of newly allocated clusters */ 233 int nb_clusters; 234 235 /** 236 * Requests that overlap with this allocation and wait to be restarted 237 * when the allocating request has completed. 238 */ 239 CoQueue dependent_requests; 240 241 /** 242 * The COW Region between the start of the first allocated cluster and the 243 * area the guest actually writes to. 244 */ 245 Qcow2COWRegion cow_start; 246 247 /** 248 * The COW Region between the area the guest actually writes to and the 249 * end of the last allocated cluster. 250 */ 251 Qcow2COWRegion cow_end; 252 253 /** Pointer to next L2Meta of the same write request */ 254 struct QCowL2Meta *next; 255 256 QLIST_ENTRY(QCowL2Meta) next_in_flight; 257 } QCowL2Meta; 258 259 enum { 260 QCOW2_CLUSTER_UNALLOCATED, 261 QCOW2_CLUSTER_NORMAL, 262 QCOW2_CLUSTER_COMPRESSED, 263 QCOW2_CLUSTER_ZERO 264 }; 265 266 #define L1E_OFFSET_MASK 0x00ffffffffffff00ULL 267 #define L2E_OFFSET_MASK 0x00ffffffffffff00ULL 268 #define L2E_COMPRESSED_OFFSET_SIZE_MASK 0x3fffffffffffffffULL 269 270 #define REFT_OFFSET_MASK 0xffffffffffffff00ULL 271 272 static inline int64_t start_of_cluster(BDRVQcowState *s, int64_t offset) 273 { 274 return offset & ~(s->cluster_size - 1); 275 } 276 277 static inline int64_t offset_into_cluster(BDRVQcowState *s, int64_t offset) 278 { 279 return offset & (s->cluster_size - 1); 280 } 281 282 static inline int size_to_clusters(BDRVQcowState *s, int64_t size) 283 { 284 return (size + (s->cluster_size - 1)) >> s->cluster_bits; 285 } 286 287 static inline int size_to_l1(BDRVQcowState *s, int64_t size) 288 { 289 int shift = s->cluster_bits + s->l2_bits; 290 return (size + (1ULL << shift) - 1) >> shift; 291 } 292 293 static inline int offset_to_l2_index(BDRVQcowState *s, int64_t offset) 294 { 295 return (offset >> s->cluster_bits) & (s->l2_size - 1); 296 } 297 298 static inline int64_t align_offset(int64_t offset, int n) 299 { 300 offset = (offset + n - 1) & ~(n - 1); 301 return offset; 302 } 303 304 static inline int qcow2_get_cluster_type(uint64_t l2_entry) 305 { 306 if (l2_entry & QCOW_OFLAG_COMPRESSED) { 307 return QCOW2_CLUSTER_COMPRESSED; 308 } else if (l2_entry & QCOW_OFLAG_ZERO) { 309 return QCOW2_CLUSTER_ZERO; 310 } else if (!(l2_entry & L2E_OFFSET_MASK)) { 311 return QCOW2_CLUSTER_UNALLOCATED; 312 } else { 313 return QCOW2_CLUSTER_NORMAL; 314 } 315 } 316 317 /* Check whether refcounts are eager or lazy */ 318 static inline bool qcow2_need_accurate_refcounts(BDRVQcowState *s) 319 { 320 return !(s->incompatible_features & QCOW2_INCOMPAT_DIRTY); 321 } 322 323 static inline uint64_t l2meta_cow_start(QCowL2Meta *m) 324 { 325 return m->offset + m->cow_start.offset; 326 } 327 328 static inline uint64_t l2meta_cow_end(QCowL2Meta *m) 329 { 330 return m->offset + m->cow_end.offset 331 + (m->cow_end.nb_sectors << BDRV_SECTOR_BITS); 332 } 333 334 // FIXME Need qcow2_ prefix to global functions 335 336 /* qcow2.c functions */ 337 int qcow2_backing_read1(BlockDriverState *bs, QEMUIOVector *qiov, 338 int64_t sector_num, int nb_sectors); 339 340 int qcow2_mark_dirty(BlockDriverState *bs); 341 int qcow2_update_header(BlockDriverState *bs); 342 343 /* qcow2-refcount.c functions */ 344 int qcow2_refcount_init(BlockDriverState *bs); 345 void qcow2_refcount_close(BlockDriverState *bs); 346 347 int64_t qcow2_alloc_clusters(BlockDriverState *bs, int64_t size); 348 int qcow2_alloc_clusters_at(BlockDriverState *bs, uint64_t offset, 349 int nb_clusters); 350 int64_t qcow2_alloc_bytes(BlockDriverState *bs, int size); 351 void qcow2_free_clusters(BlockDriverState *bs, 352 int64_t offset, int64_t size); 353 void qcow2_free_any_clusters(BlockDriverState *bs, 354 uint64_t cluster_offset, int nb_clusters); 355 356 int qcow2_update_snapshot_refcount(BlockDriverState *bs, 357 int64_t l1_table_offset, int l1_size, int addend); 358 359 int qcow2_check_refcounts(BlockDriverState *bs, BdrvCheckResult *res, 360 BdrvCheckMode fix); 361 362 /* qcow2-cluster.c functions */ 363 int qcow2_grow_l1_table(BlockDriverState *bs, int min_size, bool exact_size); 364 void qcow2_l2_cache_reset(BlockDriverState *bs); 365 int qcow2_decompress_cluster(BlockDriverState *bs, uint64_t cluster_offset); 366 void qcow2_encrypt_sectors(BDRVQcowState *s, int64_t sector_num, 367 uint8_t *out_buf, const uint8_t *in_buf, 368 int nb_sectors, int enc, 369 const AES_KEY *key); 370 371 int qcow2_get_cluster_offset(BlockDriverState *bs, uint64_t offset, 372 int *num, uint64_t *cluster_offset); 373 int qcow2_alloc_cluster_offset(BlockDriverState *bs, uint64_t offset, 374 int n_start, int n_end, int *num, uint64_t *host_offset, QCowL2Meta **m); 375 uint64_t qcow2_alloc_compressed_cluster_offset(BlockDriverState *bs, 376 uint64_t offset, 377 int compressed_size); 378 379 int qcow2_alloc_cluster_link_l2(BlockDriverState *bs, QCowL2Meta *m); 380 int qcow2_discard_clusters(BlockDriverState *bs, uint64_t offset, 381 int nb_sectors); 382 int qcow2_zero_clusters(BlockDriverState *bs, uint64_t offset, int nb_sectors); 383 384 /* qcow2-snapshot.c functions */ 385 int qcow2_snapshot_create(BlockDriverState *bs, QEMUSnapshotInfo *sn_info); 386 int qcow2_snapshot_goto(BlockDriverState *bs, const char *snapshot_id); 387 int qcow2_snapshot_delete(BlockDriverState *bs, const char *snapshot_id); 388 int qcow2_snapshot_list(BlockDriverState *bs, QEMUSnapshotInfo **psn_tab); 389 int qcow2_snapshot_load_tmp(BlockDriverState *bs, const char *snapshot_name); 390 391 void qcow2_free_snapshots(BlockDriverState *bs); 392 int qcow2_read_snapshots(BlockDriverState *bs); 393 394 /* qcow2-cache.c functions */ 395 Qcow2Cache *qcow2_cache_create(BlockDriverState *bs, int num_tables); 396 int qcow2_cache_destroy(BlockDriverState* bs, Qcow2Cache *c); 397 398 void qcow2_cache_entry_mark_dirty(Qcow2Cache *c, void *table); 399 int qcow2_cache_flush(BlockDriverState *bs, Qcow2Cache *c); 400 int qcow2_cache_set_dependency(BlockDriverState *bs, Qcow2Cache *c, 401 Qcow2Cache *dependency); 402 void qcow2_cache_depends_on_flush(Qcow2Cache *c); 403 404 int qcow2_cache_get(BlockDriverState *bs, Qcow2Cache *c, uint64_t offset, 405 void **table); 406 int qcow2_cache_get_empty(BlockDriverState *bs, Qcow2Cache *c, uint64_t offset, 407 void **table); 408 int qcow2_cache_put(BlockDriverState *bs, Qcow2Cache *c, void **table); 409 410 #endif 411