]> git.proxmox.com Git - qemu.git/blame - block-qcow2.c
Add a scatter-gather list type and accessors (Avi Kivity)
[qemu.git] / block-qcow2.c
CommitLineData
585f8587
FB
1/*
2 * Block driver for the QCOW version 2 format
5fafdf24 3 *
585f8587 4 * Copyright (c) 2004-2006 Fabrice Bellard
5fafdf24 5 *
585f8587
FB
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 */
faf07963 24#include "qemu-common.h"
585f8587
FB
25#include "block_int.h"
26#include <zlib.h>
27#include "aes.h"
28#include <assert.h>
29
30/*
31 Differences with QCOW:
32
33 - Support for multiple incremental snapshots.
34 - Memory management by reference counts.
35 - Clusters which have a reference count of one have the bit
36 QCOW_OFLAG_COPIED to optimize write performance.
5fafdf24 37 - Size of compressed clusters is stored in sectors to reduce bit usage
585f8587
FB
38 in the cluster offsets.
39 - Support for storing additional data (such as the VM state) in the
3b46e624 40 snapshots.
585f8587
FB
41 - If a backing store is used, the cluster size is not constrained
42 (could be backported to QCOW).
43 - L2 tables have always a size of one cluster.
44*/
45
46//#define DEBUG_ALLOC
47//#define DEBUG_ALLOC2
5fafdf24 48
585f8587
FB
49#define QCOW_MAGIC (('Q' << 24) | ('F' << 16) | ('I' << 8) | 0xfb)
50#define QCOW_VERSION 2
51
52#define QCOW_CRYPT_NONE 0
53#define QCOW_CRYPT_AES 1
54
095a9c58
AL
55#define QCOW_MAX_CRYPT_CLUSTERS 32
56
585f8587
FB
57/* indicate that the refcount of the referenced cluster is exactly one. */
58#define QCOW_OFLAG_COPIED (1LL << 63)
59/* indicate that the cluster is compressed (they never have the copied flag) */
60#define QCOW_OFLAG_COMPRESSED (1LL << 62)
61
62#define REFCOUNT_SHIFT 1 /* refcount size is 2 bytes */
63
585f8587
FB
64typedef 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} QCowHeader;
79
80typedef struct __attribute__((packed)) QCowSnapshotHeader {
81 /* header is 8 byte aligned */
82 uint64_t l1_table_offset;
83
84 uint32_t l1_size;
85 uint16_t id_str_size;
86 uint16_t name_size;
87
88 uint32_t date_sec;
89 uint32_t date_nsec;
90
91 uint64_t vm_clock_nsec;
92
93 uint32_t vm_state_size;
94 uint32_t extra_data_size; /* for extension */
95 /* extra data follows */
96 /* id_str follows */
97 /* name follows */
98} QCowSnapshotHeader;
99
100#define L2_CACHE_SIZE 16
101
102typedef struct QCowSnapshot {
103 uint64_t l1_table_offset;
104 uint32_t l1_size;
105 char *id_str;
106 char *name;
107 uint32_t vm_state_size;
108 uint32_t date_sec;
109 uint32_t date_nsec;
110 uint64_t vm_clock_nsec;
111} QCowSnapshot;
112
113typedef struct BDRVQcowState {
114 BlockDriverState *hd;
115 int cluster_bits;
116 int cluster_size;
117 int cluster_sectors;
118 int l2_bits;
119 int l2_size;
120 int l1_size;
121 int l1_vm_state_index;
122 int csize_shift;
123 int csize_mask;
124 uint64_t cluster_offset_mask;
125 uint64_t l1_table_offset;
126 uint64_t *l1_table;
127 uint64_t *l2_cache;
128 uint64_t l2_cache_offsets[L2_CACHE_SIZE];
129 uint32_t l2_cache_counts[L2_CACHE_SIZE];
130 uint8_t *cluster_cache;
131 uint8_t *cluster_data;
132 uint64_t cluster_cache_offset;
133
134 uint64_t *refcount_table;
135 uint64_t refcount_table_offset;
136 uint32_t refcount_table_size;
137 uint64_t refcount_block_cache_offset;
138 uint16_t *refcount_block_cache;
139 int64_t free_cluster_index;
140 int64_t free_byte_offset;
141
142 uint32_t crypt_method; /* current crypt method, 0 if no key yet */
143 uint32_t crypt_method_header;
144 AES_KEY aes_encrypt_key;
145 AES_KEY aes_decrypt_key;
5d4cbd78
AL
146
147 int64_t highest_alloc; /* highest cluester allocated (in clusters) */
1987530f 148 int64_t nc_free; /* num of free clusters below highest_alloc */
5d4cbd78 149
585f8587
FB
150 uint64_t snapshots_offset;
151 int snapshots_size;
152 int nb_snapshots;
153 QCowSnapshot *snapshots;
154} BDRVQcowState;
155
156static int decompress_cluster(BDRVQcowState *s, uint64_t cluster_offset);
5fafdf24 157static int qcow_read(BlockDriverState *bs, int64_t sector_num,
585f8587
FB
158 uint8_t *buf, int nb_sectors);
159static int qcow_read_snapshots(BlockDriverState *bs);
160static void qcow_free_snapshots(BlockDriverState *bs);
161static int refcount_init(BlockDriverState *bs);
162static void refcount_close(BlockDriverState *bs);
163static int get_refcount(BlockDriverState *bs, int64_t cluster_index);
5fafdf24 164static int update_cluster_refcount(BlockDriverState *bs,
585f8587
FB
165 int64_t cluster_index,
166 int addend);
5fafdf24
TS
167static void update_refcount(BlockDriverState *bs,
168 int64_t offset, int64_t length,
585f8587
FB
169 int addend);
170static int64_t alloc_clusters(BlockDriverState *bs, int64_t size);
171static int64_t alloc_bytes(BlockDriverState *bs, int size);
5fafdf24 172static void free_clusters(BlockDriverState *bs,
585f8587
FB
173 int64_t offset, int64_t size);
174#ifdef DEBUG_ALLOC
175static void check_refcounts(BlockDriverState *bs);
176#endif
1987530f 177static void scan_refcount(BlockDriverState *bs, int64_t *high, int64_t *free);
5d4cbd78 178
585f8587
FB
179
180static int qcow_probe(const uint8_t *buf, int buf_size, const char *filename)
181{
182 const QCowHeader *cow_header = (const void *)buf;
3b46e624 183
585f8587
FB
184 if (buf_size >= sizeof(QCowHeader) &&
185 be32_to_cpu(cow_header->magic) == QCOW_MAGIC &&
5fafdf24 186 be32_to_cpu(cow_header->version) == QCOW_VERSION)
585f8587
FB
187 return 100;
188 else
189 return 0;
190}
191
192static int qcow_open(BlockDriverState *bs, const char *filename, int flags)
193{
194 BDRVQcowState *s = bs->opaque;
195 int len, i, shift, ret;
196 QCowHeader header;
197
4dc822d7
AL
198 /* Performance is terrible right now with cache=writethrough due mainly
199 * to reference count updates. If the user does not explicitly specify
200 * a caching type, force to writeback caching.
201 */
202 if ((flags & BDRV_O_CACHE_DEF)) {
203 flags |= BDRV_O_CACHE_WB;
204 flags &= ~BDRV_O_CACHE_DEF;
205 }
b5eff355 206 ret = bdrv_file_open(&s->hd, filename, flags);
585f8587
FB
207 if (ret < 0)
208 return ret;
209 if (bdrv_pread(s->hd, 0, &header, sizeof(header)) != sizeof(header))
210 goto fail;
211 be32_to_cpus(&header.magic);
212 be32_to_cpus(&header.version);
213 be64_to_cpus(&header.backing_file_offset);
214 be32_to_cpus(&header.backing_file_size);
215 be64_to_cpus(&header.size);
216 be32_to_cpus(&header.cluster_bits);
217 be32_to_cpus(&header.crypt_method);
218 be64_to_cpus(&header.l1_table_offset);
219 be32_to_cpus(&header.l1_size);
220 be64_to_cpus(&header.refcount_table_offset);
221 be32_to_cpus(&header.refcount_table_clusters);
222 be64_to_cpus(&header.snapshots_offset);
223 be32_to_cpus(&header.nb_snapshots);
3b46e624 224
585f8587
FB
225 if (header.magic != QCOW_MAGIC || header.version != QCOW_VERSION)
226 goto fail;
5fafdf24
TS
227 if (header.size <= 1 ||
228 header.cluster_bits < 9 ||
585f8587
FB
229 header.cluster_bits > 16)
230 goto fail;
231 if (header.crypt_method > QCOW_CRYPT_AES)
232 goto fail;
233 s->crypt_method_header = header.crypt_method;
234 if (s->crypt_method_header)
235 bs->encrypted = 1;
236 s->cluster_bits = header.cluster_bits;
237 s->cluster_size = 1 << s->cluster_bits;
238 s->cluster_sectors = 1 << (s->cluster_bits - 9);
239 s->l2_bits = s->cluster_bits - 3; /* L2 is always one cluster */
240 s->l2_size = 1 << s->l2_bits;
241 bs->total_sectors = header.size / 512;
242 s->csize_shift = (62 - (s->cluster_bits - 8));
243 s->csize_mask = (1 << (s->cluster_bits - 8)) - 1;
244 s->cluster_offset_mask = (1LL << s->csize_shift) - 1;
245 s->refcount_table_offset = header.refcount_table_offset;
5fafdf24 246 s->refcount_table_size =
585f8587
FB
247 header.refcount_table_clusters << (s->cluster_bits - 3);
248
249 s->snapshots_offset = header.snapshots_offset;
250 s->nb_snapshots = header.nb_snapshots;
251
252 /* read the level 1 table */
253 s->l1_size = header.l1_size;
254 shift = s->cluster_bits + s->l2_bits;
255 s->l1_vm_state_index = (header.size + (1LL << shift) - 1) >> shift;
256 /* the L1 table must contain at least enough entries to put
257 header.size bytes */
258 if (s->l1_size < s->l1_vm_state_index)
259 goto fail;
260 s->l1_table_offset = header.l1_table_offset;
261 s->l1_table = qemu_malloc(s->l1_size * sizeof(uint64_t));
262 if (!s->l1_table)
263 goto fail;
5fafdf24 264 if (bdrv_pread(s->hd, s->l1_table_offset, s->l1_table, s->l1_size * sizeof(uint64_t)) !=
585f8587
FB
265 s->l1_size * sizeof(uint64_t))
266 goto fail;
267 for(i = 0;i < s->l1_size; i++) {
268 be64_to_cpus(&s->l1_table[i]);
269 }
270 /* alloc L2 cache */
271 s->l2_cache = qemu_malloc(s->l2_size * L2_CACHE_SIZE * sizeof(uint64_t));
272 if (!s->l2_cache)
273 goto fail;
274 s->cluster_cache = qemu_malloc(s->cluster_size);
275 if (!s->cluster_cache)
276 goto fail;
277 /* one more sector for decompressed data alignment */
095a9c58
AL
278 s->cluster_data = qemu_malloc(QCOW_MAX_CRYPT_CLUSTERS * s->cluster_size
279 + 512);
585f8587
FB
280 if (!s->cluster_data)
281 goto fail;
282 s->cluster_cache_offset = -1;
3b46e624 283
585f8587
FB
284 if (refcount_init(bs) < 0)
285 goto fail;
286
1987530f 287 scan_refcount(bs, &s->highest_alloc, &s->nc_free);
5d4cbd78 288
585f8587
FB
289 /* read the backing file name */
290 if (header.backing_file_offset != 0) {
291 len = header.backing_file_size;
292 if (len > 1023)
293 len = 1023;
294 if (bdrv_pread(s->hd, header.backing_file_offset, bs->backing_file, len) != len)
295 goto fail;
296 bs->backing_file[len] = '\0';
297 }
298 if (qcow_read_snapshots(bs) < 0)
299 goto fail;
300
301#ifdef DEBUG_ALLOC
302 check_refcounts(bs);
303#endif
304 return 0;
305
306 fail:
307 qcow_free_snapshots(bs);
308 refcount_close(bs);
309 qemu_free(s->l1_table);
310 qemu_free(s->l2_cache);
311 qemu_free(s->cluster_cache);
312 qemu_free(s->cluster_data);
313 bdrv_delete(s->hd);
314 return -1;
315}
316
317static int qcow_set_key(BlockDriverState *bs, const char *key)
318{
319 BDRVQcowState *s = bs->opaque;
320 uint8_t keybuf[16];
321 int len, i;
3b46e624 322
585f8587
FB
323 memset(keybuf, 0, 16);
324 len = strlen(key);
325 if (len > 16)
326 len = 16;
327 /* XXX: we could compress the chars to 7 bits to increase
328 entropy */
329 for(i = 0;i < len;i++) {
330 keybuf[i] = key[i];
331 }
332 s->crypt_method = s->crypt_method_header;
333
334 if (AES_set_encrypt_key(keybuf, 128, &s->aes_encrypt_key) != 0)
335 return -1;
336 if (AES_set_decrypt_key(keybuf, 128, &s->aes_decrypt_key) != 0)
337 return -1;
338#if 0
339 /* test */
340 {
341 uint8_t in[16];
342 uint8_t out[16];
343 uint8_t tmp[16];
344 for(i=0;i<16;i++)
345 in[i] = i;
346 AES_encrypt(in, tmp, &s->aes_encrypt_key);
347 AES_decrypt(tmp, out, &s->aes_decrypt_key);
348 for(i = 0; i < 16; i++)
349 printf(" %02x", tmp[i]);
350 printf("\n");
351 for(i = 0; i < 16; i++)
352 printf(" %02x", out[i]);
353 printf("\n");
354 }
355#endif
356 return 0;
357}
358
359/* The crypt function is compatible with the linux cryptoloop
360 algorithm for < 4 GB images. NOTE: out_buf == in_buf is
361 supported */
362static void encrypt_sectors(BDRVQcowState *s, int64_t sector_num,
363 uint8_t *out_buf, const uint8_t *in_buf,
364 int nb_sectors, int enc,
365 const AES_KEY *key)
366{
367 union {
368 uint64_t ll[2];
369 uint8_t b[16];
370 } ivec;
371 int i;
372
373 for(i = 0; i < nb_sectors; i++) {
374 ivec.ll[0] = cpu_to_le64(sector_num);
375 ivec.ll[1] = 0;
5fafdf24 376 AES_cbc_encrypt(in_buf, out_buf, 512, key,
585f8587
FB
377 ivec.b, enc);
378 sector_num++;
379 in_buf += 512;
380 out_buf += 512;
381 }
382}
383
384static int copy_sectors(BlockDriverState *bs, uint64_t start_sect,
385 uint64_t cluster_offset, int n_start, int n_end)
386{
387 BDRVQcowState *s = bs->opaque;
388 int n, ret;
389
390 n = n_end - n_start;
391 if (n <= 0)
392 return 0;
393 ret = qcow_read(bs, start_sect + n_start, s->cluster_data, n);
394 if (ret < 0)
395 return ret;
396 if (s->crypt_method) {
5fafdf24
TS
397 encrypt_sectors(s, start_sect + n_start,
398 s->cluster_data,
585f8587
FB
399 s->cluster_data, n, 1,
400 &s->aes_encrypt_key);
401 }
5fafdf24 402 ret = bdrv_write(s->hd, (cluster_offset >> 9) + n_start,
585f8587
FB
403 s->cluster_data, n);
404 if (ret < 0)
405 return ret;
406 return 0;
407}
408
409static void l2_cache_reset(BlockDriverState *bs)
410{
411 BDRVQcowState *s = bs->opaque;
412
413 memset(s->l2_cache, 0, s->l2_size * L2_CACHE_SIZE * sizeof(uint64_t));
414 memset(s->l2_cache_offsets, 0, L2_CACHE_SIZE * sizeof(uint64_t));
415 memset(s->l2_cache_counts, 0, L2_CACHE_SIZE * sizeof(uint32_t));
416}
417
418static inline int l2_cache_new_entry(BlockDriverState *bs)
419{
420 BDRVQcowState *s = bs->opaque;
421 uint32_t min_count;
422 int min_index, i;
423
424 /* find a new entry in the least used one */
425 min_index = 0;
426 min_count = 0xffffffff;
427 for(i = 0; i < L2_CACHE_SIZE; i++) {
428 if (s->l2_cache_counts[i] < min_count) {
429 min_count = s->l2_cache_counts[i];
430 min_index = i;
431 }
432 }
433 return min_index;
434}
435
436static int64_t align_offset(int64_t offset, int n)
437{
438 offset = (offset + n - 1) & ~(n - 1);
439 return offset;
440}
441
442static int grow_l1_table(BlockDriverState *bs, int min_size)
443{
444 BDRVQcowState *s = bs->opaque;
445 int new_l1_size, new_l1_size2, ret, i;
446 uint64_t *new_l1_table;
447 uint64_t new_l1_table_offset;
643e5399 448 uint8_t data[12];
585f8587
FB
449
450 new_l1_size = s->l1_size;
451 if (min_size <= new_l1_size)
452 return 0;
453 while (min_size > new_l1_size) {
454 new_l1_size = (new_l1_size * 3 + 1) / 2;
455 }
456#ifdef DEBUG_ALLOC2
457 printf("grow l1_table from %d to %d\n", s->l1_size, new_l1_size);
458#endif
459
460 new_l1_size2 = sizeof(uint64_t) * new_l1_size;
461 new_l1_table = qemu_mallocz(new_l1_size2);
462 if (!new_l1_table)
463 return -ENOMEM;
464 memcpy(new_l1_table, s->l1_table, s->l1_size * sizeof(uint64_t));
465
466 /* write new table (align to cluster) */
467 new_l1_table_offset = alloc_clusters(bs, new_l1_size2);
3b46e624 468
585f8587
FB
469 for(i = 0; i < s->l1_size; i++)
470 new_l1_table[i] = cpu_to_be64(new_l1_table[i]);
471 ret = bdrv_pwrite(s->hd, new_l1_table_offset, new_l1_table, new_l1_size2);
472 if (ret != new_l1_size2)
473 goto fail;
474 for(i = 0; i < s->l1_size; i++)
475 new_l1_table[i] = be64_to_cpu(new_l1_table[i]);
3b46e624 476
585f8587 477 /* set new table */
643e5399
AL
478 cpu_to_be32w((uint32_t*)data, new_l1_size);
479 cpu_to_be64w((uint64_t*)(data + 4), new_l1_table_offset);
480 if (bdrv_pwrite(s->hd, offsetof(QCowHeader, l1_size), data,
481 sizeof(data)) != sizeof(data))
585f8587
FB
482 goto fail;
483 qemu_free(s->l1_table);
484 free_clusters(bs, s->l1_table_offset, s->l1_size * sizeof(uint64_t));
485 s->l1_table_offset = new_l1_table_offset;
486 s->l1_table = new_l1_table;
487 s->l1_size = new_l1_size;
488 return 0;
489 fail:
490 qemu_free(s->l1_table);
491 return -EIO;
492}
493
108534b9
AL
494/*
495 * seek_l2_table
585f8587 496 *
108534b9
AL
497 * seek l2_offset in the l2_cache table
498 * if not found, return NULL,
499 * if found,
500 * increments the l2 cache hit count of the entry,
501 * if counter overflow, divide by two all counters
502 * return the pointer to the l2 cache entry
585f8587 503 *
108534b9
AL
504 */
505
506static uint64_t *seek_l2_table(BDRVQcowState *s, uint64_t l2_offset)
507{
508 int i, j;
509
510 for(i = 0; i < L2_CACHE_SIZE; i++) {
511 if (l2_offset == s->l2_cache_offsets[i]) {
512 /* increment the hit count */
513 if (++s->l2_cache_counts[i] == 0xffffffff) {
514 for(j = 0; j < L2_CACHE_SIZE; j++) {
515 s->l2_cache_counts[j] >>= 1;
516 }
517 }
518 return s->l2_cache + (i << s->l2_bits);
519 }
520 }
521 return NULL;
522}
523
524/*
525 * l2_load
526 *
527 * Loads a L2 table into memory. If the table is in the cache, the cache
528 * is used; otherwise the L2 table is loaded from the image file.
529 *
530 * Returns a pointer to the L2 table on success, or NULL if the read from
531 * the image file failed.
532 */
533
534static uint64_t *l2_load(BlockDriverState *bs, uint64_t l2_offset)
535{
536 BDRVQcowState *s = bs->opaque;
537 int min_index;
538 uint64_t *l2_table;
539
540 /* seek if the table for the given offset is in the cache */
541
542 l2_table = seek_l2_table(s, l2_offset);
543 if (l2_table != NULL)
544 return l2_table;
545
546 /* not found: load a new entry in the least used one */
547
548 min_index = l2_cache_new_entry(bs);
549 l2_table = s->l2_cache + (min_index << s->l2_bits);
550 if (bdrv_pread(s->hd, l2_offset, l2_table, s->l2_size * sizeof(uint64_t)) !=
551 s->l2_size * sizeof(uint64_t))
552 return NULL;
553 s->l2_cache_offsets[min_index] = l2_offset;
554 s->l2_cache_counts[min_index] = 1;
555
556 return l2_table;
557}
558
559/*
560 * l2_allocate
585f8587 561 *
108534b9
AL
562 * Allocate a new l2 entry in the file. If l1_index points to an already
563 * used entry in the L2 table (i.e. we are doing a copy on write for the L2
564 * table) copy the contents of the old L2 table into the newly allocated one.
565 * Otherwise the new table is initialized with zeros.
585f8587 566 *
585f8587 567 */
108534b9
AL
568
569static uint64_t *l2_allocate(BlockDriverState *bs, int l1_index)
570{
571 BDRVQcowState *s = bs->opaque;
572 int min_index;
573 uint64_t old_l2_offset, tmp;
574 uint64_t *l2_table, l2_offset;
575
576 old_l2_offset = s->l1_table[l1_index];
577
578 /* allocate a new l2 entry */
579
580 l2_offset = alloc_clusters(bs, s->l2_size * sizeof(uint64_t));
581
582 /* update the L1 entry */
583
584 s->l1_table[l1_index] = l2_offset | QCOW_OFLAG_COPIED;
585
586 tmp = cpu_to_be64(l2_offset | QCOW_OFLAG_COPIED);
587 if (bdrv_pwrite(s->hd, s->l1_table_offset + l1_index * sizeof(tmp),
588 &tmp, sizeof(tmp)) != sizeof(tmp))
589 return NULL;
590
591 /* allocate a new entry in the l2 cache */
592
593 min_index = l2_cache_new_entry(bs);
594 l2_table = s->l2_cache + (min_index << s->l2_bits);
595
596 if (old_l2_offset == 0) {
597 /* if there was no old l2 table, clear the new table */
598 memset(l2_table, 0, s->l2_size * sizeof(uint64_t));
599 } else {
600 /* if there was an old l2 table, read it from the disk */
601 if (bdrv_pread(s->hd, old_l2_offset,
602 l2_table, s->l2_size * sizeof(uint64_t)) !=
603 s->l2_size * sizeof(uint64_t))
604 return NULL;
605 }
606 /* write the l2 table to the file */
607 if (bdrv_pwrite(s->hd, l2_offset,
608 l2_table, s->l2_size * sizeof(uint64_t)) !=
609 s->l2_size * sizeof(uint64_t))
610 return NULL;
611
612 /* update the l2 cache entry */
613
614 s->l2_cache_offsets[min_index] = l2_offset;
615 s->l2_cache_counts[min_index] = 1;
616
617 return l2_table;
618}
619
6db6c638
AL
620static int size_to_clusters(BDRVQcowState *s, int64_t size)
621{
622 return (size + (s->cluster_size - 1)) >> s->cluster_bits;
623}
624
625static int count_contiguous_clusters(uint64_t nb_clusters, int cluster_size,
ff4b91c2 626 uint64_t *l2_table, uint64_t start, uint64_t mask)
6db6c638
AL
627{
628 int i;
629 uint64_t offset = be64_to_cpu(l2_table[0]) & ~mask;
630
ab5ccbd6
AL
631 if (!offset)
632 return 0;
633
ff4b91c2 634 for (i = start; i < start + nb_clusters; i++)
6db6c638
AL
635 if (offset + i * cluster_size != (be64_to_cpu(l2_table[i]) & ~mask))
636 break;
637
ff4b91c2 638 return (i - start);
6db6c638
AL
639}
640
641static int count_contiguous_free_clusters(uint64_t nb_clusters, uint64_t *l2_table)
642{
643 int i = 0;
644
645 while(nb_clusters-- && l2_table[i] == 0)
646 i++;
647
648 return i;
649}
650
05203524
AL
651/*
652 * get_cluster_offset
653 *
654 * For a given offset of the disk image, return cluster offset in
655 * qcow2 file.
656 *
095a9c58
AL
657 * on entry, *num is the number of contiguous clusters we'd like to
658 * access following offset.
659 *
660 * on exit, *num is the number of contiguous clusters we can read.
661 *
05203524
AL
662 * Return 1, if the offset is found
663 * Return 0, otherwise.
664 *
665 */
666
095a9c58
AL
667static uint64_t get_cluster_offset(BlockDriverState *bs,
668 uint64_t offset, int *num)
05203524
AL
669{
670 BDRVQcowState *s = bs->opaque;
671 int l1_index, l2_index;
6db6c638
AL
672 uint64_t l2_offset, *l2_table, cluster_offset;
673 int l1_bits, c;
674 int index_in_cluster, nb_available, nb_needed, nb_clusters;
095a9c58
AL
675
676 index_in_cluster = (offset >> 9) & (s->cluster_sectors - 1);
677 nb_needed = *num + index_in_cluster;
678
679 l1_bits = s->l2_bits + s->cluster_bits;
680
681 /* compute how many bytes there are between the offset and
6db6c638 682 * the end of the l1 entry
095a9c58
AL
683 */
684
685 nb_available = (1 << l1_bits) - (offset & ((1 << l1_bits) - 1));
686
687 /* compute the number of available sectors */
688
689 nb_available = (nb_available >> 9) + index_in_cluster;
690
691 cluster_offset = 0;
05203524
AL
692
693 /* seek the the l2 offset in the l1 table */
694
095a9c58 695 l1_index = offset >> l1_bits;
05203524 696 if (l1_index >= s->l1_size)
095a9c58 697 goto out;
05203524
AL
698
699 l2_offset = s->l1_table[l1_index];
700
701 /* seek the l2 table of the given l2 offset */
702
703 if (!l2_offset)
095a9c58 704 goto out;
05203524
AL
705
706 /* load the l2 table in memory */
707
708 l2_offset &= ~QCOW_OFLAG_COPIED;
709 l2_table = l2_load(bs, l2_offset);
710 if (l2_table == NULL)
768706a5 711 return 0;
05203524
AL
712
713 /* find the cluster offset for the given disk offset */
714
715 l2_index = (offset >> s->cluster_bits) & (s->l2_size - 1);
716 cluster_offset = be64_to_cpu(l2_table[l2_index]);
6db6c638 717 nb_clusters = size_to_clusters(s, nb_needed << 9);
095a9c58
AL
718
719 if (!cluster_offset) {
6db6c638
AL
720 /* how many empty clusters ? */
721 c = count_contiguous_free_clusters(nb_clusters, &l2_table[l2_index]);
095a9c58 722 } else {
6db6c638
AL
723 /* how many allocated clusters ? */
724 c = count_contiguous_clusters(nb_clusters, s->cluster_size,
ff4b91c2 725 &l2_table[l2_index], 0, QCOW_OFLAG_COPIED);
6db6c638 726 }
095a9c58 727
6db6c638 728 nb_available = (c * s->cluster_sectors);
095a9c58
AL
729out:
730 if (nb_available > nb_needed)
731 nb_available = nb_needed;
732
733 *num = nb_available - index_in_cluster;
734
6db6c638 735 return cluster_offset & ~QCOW_OFLAG_COPIED;
05203524
AL
736}
737
738/*
52d893ec 739 * free_any_clusters
05203524 740 *
52d893ec 741 * free clusters according to its type: compressed or not
05203524 742 *
52d893ec
AL
743 */
744
745static void free_any_clusters(BlockDriverState *bs,
095a9c58 746 uint64_t cluster_offset, int nb_clusters)
52d893ec
AL
747{
748 BDRVQcowState *s = bs->opaque;
749
52d893ec
AL
750 /* free the cluster */
751
752 if (cluster_offset & QCOW_OFLAG_COMPRESSED) {
753 int nb_csectors;
754 nb_csectors = ((cluster_offset >> s->csize_shift) &
755 s->csize_mask) + 1;
756 free_clusters(bs, (cluster_offset & s->cluster_offset_mask) & ~511,
757 nb_csectors * 512);
758 return;
759 }
760
095a9c58
AL
761 free_clusters(bs, cluster_offset, nb_clusters << s->cluster_bits);
762
763 return;
52d893ec
AL
764}
765
766/*
767 * get_cluster_table
05203524 768 *
52d893ec
AL
769 * for a given disk offset, load (and allocate if needed)
770 * the l2 table.
771 *
772 * the l2 table offset in the qcow2 file and the cluster index
773 * in the l2 table are given to the caller.
05203524
AL
774 *
775 */
776
52d893ec
AL
777static int get_cluster_table(BlockDriverState *bs, uint64_t offset,
778 uint64_t **new_l2_table,
779 uint64_t *new_l2_offset,
780 int *new_l2_index)
585f8587
FB
781{
782 BDRVQcowState *s = bs->opaque;
108534b9 783 int l1_index, l2_index, ret;
52d893ec 784 uint64_t l2_offset, *l2_table;
108534b9
AL
785
786 /* seek the the l2 offset in the l1 table */
3b46e624 787
585f8587
FB
788 l1_index = offset >> (s->l2_bits + s->cluster_bits);
789 if (l1_index >= s->l1_size) {
108534b9
AL
790 ret = grow_l1_table(bs, l1_index + 1);
791 if (ret < 0)
585f8587
FB
792 return 0;
793 }
794 l2_offset = s->l1_table[l1_index];
108534b9
AL
795
796 /* seek the l2 table of the given l2 offset */
797
05203524
AL
798 if (l2_offset & QCOW_OFLAG_COPIED) {
799 /* load the l2 table in memory */
800 l2_offset &= ~QCOW_OFLAG_COPIED;
801 l2_table = l2_load(bs, l2_offset);
802 if (l2_table == NULL)
585f8587 803 return 0;
05203524
AL
804 } else {
805 if (l2_offset)
806 free_clusters(bs, l2_offset, s->l2_size * sizeof(uint64_t));
108534b9
AL
807 l2_table = l2_allocate(bs, l1_index);
808 if (l2_table == NULL)
585f8587 809 return 0;
108534b9 810 l2_offset = s->l1_table[l1_index] & ~QCOW_OFLAG_COPIED;
585f8587 811 }
108534b9
AL
812
813 /* find the cluster offset for the given disk offset */
814
585f8587 815 l2_index = (offset >> s->cluster_bits) & (s->l2_size - 1);
05203524 816
52d893ec
AL
817 *new_l2_table = l2_table;
818 *new_l2_offset = l2_offset;
819 *new_l2_index = l2_index;
820
821 return 1;
822}
823
824/*
825 * alloc_compressed_cluster_offset
826 *
827 * For a given offset of the disk image, return cluster offset in
828 * qcow2 file.
829 *
830 * If the offset is not found, allocate a new compressed cluster.
831 *
832 * Return the cluster offset if successful,
833 * Return 0, otherwise.
834 *
835 */
836
837static uint64_t alloc_compressed_cluster_offset(BlockDriverState *bs,
838 uint64_t offset,
839 int compressed_size)
840{
841 BDRVQcowState *s = bs->opaque;
842 int l2_index, ret;
843 uint64_t l2_offset, *l2_table, cluster_offset;
844 int nb_csectors;
845
846 ret = get_cluster_table(bs, offset, &l2_table, &l2_offset, &l2_index);
847 if (ret == 0)
848 return 0;
849
850 cluster_offset = be64_to_cpu(l2_table[l2_index]);
05203524
AL
851 if (cluster_offset & QCOW_OFLAG_COPIED)
852 return cluster_offset & ~QCOW_OFLAG_COPIED;
853
095a9c58
AL
854 if (cluster_offset)
855 free_any_clusters(bs, cluster_offset, 1);
108534b9 856
52d893ec
AL
857 cluster_offset = alloc_bytes(bs, compressed_size);
858 nb_csectors = ((cluster_offset + compressed_size - 1) >> 9) -
859 (cluster_offset >> 9);
05203524 860
52d893ec
AL
861 cluster_offset |= QCOW_OFLAG_COMPRESSED |
862 ((uint64_t)nb_csectors << s->csize_shift);
05203524 863
52d893ec 864 /* update L2 table */
05203524 865
52d893ec 866 /* compressed clusters never have the copied flag */
05203524 867
52d893ec
AL
868 l2_table[l2_index] = cpu_to_be64(cluster_offset);
869 if (bdrv_pwrite(s->hd,
870 l2_offset + l2_index * sizeof(uint64_t),
871 l2_table + l2_index,
872 sizeof(uint64_t)) != sizeof(uint64_t))
873 return 0;
05203524 874
52d893ec
AL
875 return cluster_offset;
876}
05203524 877
e976c6a1
AL
878typedef struct QCowL2Meta
879{
880 uint64_t offset;
881 int n_start;
882 int nb_available;
883 int nb_clusters;
884} QCowL2Meta;
885
886static int alloc_cluster_link_l2(BlockDriverState *bs, uint64_t cluster_offset,
887 QCowL2Meta *m)
888{
889 BDRVQcowState *s = bs->opaque;
890 int i, j = 0, l2_index, ret;
891 uint64_t *old_cluster, start_sect, l2_offset, *l2_table;
892
893 if (m->nb_clusters == 0)
894 return 0;
895
896 if (!(old_cluster = qemu_malloc(m->nb_clusters * sizeof(uint64_t))))
897 return -ENOMEM;
898
899 /* copy content of unmodified sectors */
900 start_sect = (m->offset & ~(s->cluster_size - 1)) >> 9;
901 if (m->n_start) {
902 ret = copy_sectors(bs, start_sect, cluster_offset, 0, m->n_start);
903 if (ret < 0)
904 goto err;
905 }
906
907 if (m->nb_available & (s->cluster_sectors - 1)) {
908 uint64_t end = m->nb_available & ~(uint64_t)(s->cluster_sectors - 1);
909 ret = copy_sectors(bs, start_sect + end, cluster_offset + (end << 9),
910 m->nb_available - end, s->cluster_sectors);
911 if (ret < 0)
912 goto err;
913 }
914
915 ret = -EIO;
916 /* update L2 table */
917 if (!get_cluster_table(bs, m->offset, &l2_table, &l2_offset, &l2_index))
918 goto err;
919
920 for (i = 0; i < m->nb_clusters; i++) {
921 if(l2_table[l2_index + i] != 0)
922 old_cluster[j++] = l2_table[l2_index + i];
923
924 l2_table[l2_index + i] = cpu_to_be64((cluster_offset +
925 (i << s->cluster_bits)) | QCOW_OFLAG_COPIED);
926 }
927
928 if (bdrv_pwrite(s->hd, l2_offset + l2_index * sizeof(uint64_t),
929 l2_table + l2_index, m->nb_clusters * sizeof(uint64_t)) !=
930 m->nb_clusters * sizeof(uint64_t))
931 goto err;
932
933 for (i = 0; i < j; i++)
934 free_any_clusters(bs, old_cluster[i], 1);
935
936 ret = 0;
937err:
938 qemu_free(old_cluster);
939 return ret;
940 }
941
52d893ec
AL
942/*
943 * alloc_cluster_offset
944 *
945 * For a given offset of the disk image, return cluster offset in
946 * qcow2 file.
947 *
948 * If the offset is not found, allocate a new cluster.
949 *
950 * Return the cluster offset if successful,
951 * Return 0, otherwise.
952 *
953 */
954
955static uint64_t alloc_cluster_offset(BlockDriverState *bs,
956 uint64_t offset,
095a9c58 957 int n_start, int n_end,
e976c6a1 958 int *num, QCowL2Meta *m)
52d893ec
AL
959{
960 BDRVQcowState *s = bs->opaque;
961 int l2_index, ret;
962 uint64_t l2_offset, *l2_table, cluster_offset;
e976c6a1 963 int nb_clusters, i = 0;
52d893ec
AL
964
965 ret = get_cluster_table(bs, offset, &l2_table, &l2_offset, &l2_index);
966 if (ret == 0)
967 return 0;
968
6db6c638
AL
969 nb_clusters = size_to_clusters(s, n_end << 9);
970
e976c6a1 971 nb_clusters = MIN(nb_clusters, s->l2_size - l2_index);
095a9c58 972
52d893ec 973 cluster_offset = be64_to_cpu(l2_table[l2_index]);
52d893ec 974
095a9c58
AL
975 /* We keep all QCOW_OFLAG_COPIED clusters */
976
977 if (cluster_offset & QCOW_OFLAG_COPIED) {
6db6c638 978 nb_clusters = count_contiguous_clusters(nb_clusters, s->cluster_size,
ff4b91c2 979 &l2_table[l2_index], 0, 0);
095a9c58 980
095a9c58 981 cluster_offset &= ~QCOW_OFLAG_COPIED;
e976c6a1 982 m->nb_clusters = 0;
095a9c58
AL
983
984 goto out;
985 }
986
987 /* for the moment, multiple compressed clusters are not managed */
988
989 if (cluster_offset & QCOW_OFLAG_COMPRESSED)
990 nb_clusters = 1;
991
bc352085 992 /* how many available clusters ? */
095a9c58 993
bc352085 994 while (i < nb_clusters) {
ab5ccbd6 995 i += count_contiguous_clusters(nb_clusters - i, s->cluster_size,
ff4b91c2 996 &l2_table[l2_index], i, 0);
ab5ccbd6
AL
997
998 if(be64_to_cpu(l2_table[l2_index + i]))
999 break;
1000
6db6c638
AL
1001 i += count_contiguous_free_clusters(nb_clusters - i,
1002 &l2_table[l2_index + i]);
095a9c58 1003
6db6c638 1004 cluster_offset = be64_to_cpu(l2_table[l2_index + i]);
095a9c58 1005
6db6c638 1006 if ((cluster_offset & QCOW_OFLAG_COPIED) ||
bc352085 1007 (cluster_offset & QCOW_OFLAG_COMPRESSED))
6db6c638 1008 break;
095a9c58 1009 }
bc352085 1010 nb_clusters = i;
05203524
AL
1011
1012 /* allocate a new cluster */
1013
095a9c58 1014 cluster_offset = alloc_clusters(bs, nb_clusters * s->cluster_size);
05203524 1015
e976c6a1
AL
1016 /* save info needed for meta data update */
1017 m->offset = offset;
1018 m->n_start = n_start;
1019 m->nb_clusters = nb_clusters;
05203524 1020
095a9c58 1021out:
e976c6a1
AL
1022 m->nb_available = MIN(nb_clusters << (s->cluster_bits - 9), n_end);
1023
1024 *num = m->nb_available - n_start;
095a9c58 1025
585f8587
FB
1026 return cluster_offset;
1027}
1028
5fafdf24 1029static int qcow_is_allocated(BlockDriverState *bs, int64_t sector_num,
585f8587
FB
1030 int nb_sectors, int *pnum)
1031{
585f8587
FB
1032 uint64_t cluster_offset;
1033
095a9c58
AL
1034 *pnum = nb_sectors;
1035 cluster_offset = get_cluster_offset(bs, sector_num << 9, pnum);
1036
585f8587
FB
1037 return (cluster_offset != 0);
1038}
1039
1040static int decompress_buffer(uint8_t *out_buf, int out_buf_size,
1041 const uint8_t *buf, int buf_size)
1042{
1043 z_stream strm1, *strm = &strm1;
1044 int ret, out_len;
1045
1046 memset(strm, 0, sizeof(*strm));
1047
1048 strm->next_in = (uint8_t *)buf;
1049 strm->avail_in = buf_size;
1050 strm->next_out = out_buf;
1051 strm->avail_out = out_buf_size;
1052
1053 ret = inflateInit2(strm, -12);
1054 if (ret != Z_OK)
1055 return -1;
1056 ret = inflate(strm, Z_FINISH);
1057 out_len = strm->next_out - out_buf;
1058 if ((ret != Z_STREAM_END && ret != Z_BUF_ERROR) ||
1059 out_len != out_buf_size) {
1060 inflateEnd(strm);
1061 return -1;
1062 }
1063 inflateEnd(strm);
1064 return 0;
1065}
3b46e624 1066
585f8587
FB
1067static int decompress_cluster(BDRVQcowState *s, uint64_t cluster_offset)
1068{
1069 int ret, csize, nb_csectors, sector_offset;
1070 uint64_t coffset;
1071
1072 coffset = cluster_offset & s->cluster_offset_mask;
1073 if (s->cluster_cache_offset != coffset) {
1074 nb_csectors = ((cluster_offset >> s->csize_shift) & s->csize_mask) + 1;
1075 sector_offset = coffset & 511;
1076 csize = nb_csectors * 512 - sector_offset;
1077 ret = bdrv_read(s->hd, coffset >> 9, s->cluster_data, nb_csectors);
1078 if (ret < 0) {
1079 return -1;
1080 }
1081 if (decompress_buffer(s->cluster_cache, s->cluster_size,
1082 s->cluster_data + sector_offset, csize) < 0) {
1083 return -1;
1084 }
1085 s->cluster_cache_offset = coffset;
1086 }
1087 return 0;
1088}
1089
a9465922 1090/* handle reading after the end of the backing file */
5fafdf24 1091static int backing_read1(BlockDriverState *bs,
a9465922
FB
1092 int64_t sector_num, uint8_t *buf, int nb_sectors)
1093{
1094 int n1;
1095 if ((sector_num + nb_sectors) <= bs->total_sectors)
1096 return nb_sectors;
1097 if (sector_num >= bs->total_sectors)
1098 n1 = 0;
1099 else
1100 n1 = bs->total_sectors - sector_num;
1101 memset(buf + n1 * 512, 0, 512 * (nb_sectors - n1));
1102 return n1;
1103}
1104
5fafdf24 1105static int qcow_read(BlockDriverState *bs, int64_t sector_num,
585f8587
FB
1106 uint8_t *buf, int nb_sectors)
1107{
1108 BDRVQcowState *s = bs->opaque;
a9465922 1109 int ret, index_in_cluster, n, n1;
585f8587 1110 uint64_t cluster_offset;
3b46e624 1111
585f8587 1112 while (nb_sectors > 0) {
095a9c58
AL
1113 n = nb_sectors;
1114 cluster_offset = get_cluster_offset(bs, sector_num << 9, &n);
585f8587 1115 index_in_cluster = sector_num & (s->cluster_sectors - 1);
585f8587
FB
1116 if (!cluster_offset) {
1117 if (bs->backing_hd) {
1118 /* read from the base image */
a9465922
FB
1119 n1 = backing_read1(bs->backing_hd, sector_num, buf, n);
1120 if (n1 > 0) {
1121 ret = bdrv_read(bs->backing_hd, sector_num, buf, n1);
1122 if (ret < 0)
1123 return -1;
1124 }
585f8587
FB
1125 } else {
1126 memset(buf, 0, 512 * n);
1127 }
1128 } else if (cluster_offset & QCOW_OFLAG_COMPRESSED) {
1129 if (decompress_cluster(s, cluster_offset) < 0)
1130 return -1;
1131 memcpy(buf, s->cluster_cache + index_in_cluster * 512, 512 * n);
1132 } else {
1133 ret = bdrv_pread(s->hd, cluster_offset + index_in_cluster * 512, buf, n * 512);
5fafdf24 1134 if (ret != n * 512)
585f8587
FB
1135 return -1;
1136 if (s->crypt_method) {
5fafdf24 1137 encrypt_sectors(s, sector_num, buf, buf, n, 0,
585f8587
FB
1138 &s->aes_decrypt_key);
1139 }
1140 }
1141 nb_sectors -= n;
1142 sector_num += n;
1143 buf += n * 512;
1144 }
1145 return 0;
1146}
1147
5fafdf24 1148static int qcow_write(BlockDriverState *bs, int64_t sector_num,
585f8587
FB
1149 const uint8_t *buf, int nb_sectors)
1150{
1151 BDRVQcowState *s = bs->opaque;
1152 int ret, index_in_cluster, n;
1153 uint64_t cluster_offset;
095a9c58 1154 int n_end;
e976c6a1 1155 QCowL2Meta l2meta;
3b46e624 1156
585f8587
FB
1157 while (nb_sectors > 0) {
1158 index_in_cluster = sector_num & (s->cluster_sectors - 1);
095a9c58
AL
1159 n_end = index_in_cluster + nb_sectors;
1160 if (s->crypt_method &&
1161 n_end > QCOW_MAX_CRYPT_CLUSTERS * s->cluster_sectors)
1162 n_end = QCOW_MAX_CRYPT_CLUSTERS * s->cluster_sectors;
52d893ec 1163 cluster_offset = alloc_cluster_offset(bs, sector_num << 9,
05203524 1164 index_in_cluster,
e976c6a1 1165 n_end, &n, &l2meta);
585f8587
FB
1166 if (!cluster_offset)
1167 return -1;
1168 if (s->crypt_method) {
1169 encrypt_sectors(s, sector_num, s->cluster_data, buf, n, 1,
1170 &s->aes_encrypt_key);
5fafdf24 1171 ret = bdrv_pwrite(s->hd, cluster_offset + index_in_cluster * 512,
585f8587
FB
1172 s->cluster_data, n * 512);
1173 } else {
1174 ret = bdrv_pwrite(s->hd, cluster_offset + index_in_cluster * 512, buf, n * 512);
1175 }
e976c6a1
AL
1176 if (ret != n * 512 || alloc_cluster_link_l2(bs, cluster_offset, &l2meta) < 0) {
1177 free_any_clusters(bs, cluster_offset, l2meta.nb_clusters);
585f8587 1178 return -1;
e976c6a1 1179 }
585f8587
FB
1180 nb_sectors -= n;
1181 sector_num += n;
1182 buf += n * 512;
1183 }
1184 s->cluster_cache_offset = -1; /* disable compressed cache */
1185 return 0;
1186}
1187
ce1a14dc
PB
1188typedef struct QCowAIOCB {
1189 BlockDriverAIOCB common;
585f8587
FB
1190 int64_t sector_num;
1191 uint8_t *buf;
1192 int nb_sectors;
1193 int n;
1194 uint64_t cluster_offset;
5fafdf24 1195 uint8_t *cluster_data;
585f8587 1196 BlockDriverAIOCB *hd_aiocb;
1490791f 1197 QEMUBH *bh;
e976c6a1 1198 QCowL2Meta l2meta;
585f8587
FB
1199} QCowAIOCB;
1200
1490791f
AL
1201static void qcow_aio_read_cb(void *opaque, int ret);
1202static void qcow_aio_read_bh(void *opaque)
1203{
1204 QCowAIOCB *acb = opaque;
1205 qemu_bh_delete(acb->bh);
1206 acb->bh = NULL;
1207 qcow_aio_read_cb(opaque, 0);
1208}
1209
a32ef786
AL
1210static int qcow_schedule_bh(QEMUBHFunc *cb, QCowAIOCB *acb)
1211{
1212 if (acb->bh)
1213 return -EIO;
1214
1215 acb->bh = qemu_bh_new(cb, acb);
1216 if (!acb->bh)
1217 return -EIO;
1218
1219 qemu_bh_schedule(acb->bh);
1220
1221 return 0;
1222}
1223
585f8587
FB
1224static void qcow_aio_read_cb(void *opaque, int ret)
1225{
ce1a14dc
PB
1226 QCowAIOCB *acb = opaque;
1227 BlockDriverState *bs = acb->common.bs;
585f8587 1228 BDRVQcowState *s = bs->opaque;
a9465922 1229 int index_in_cluster, n1;
585f8587 1230
ce1a14dc 1231 acb->hd_aiocb = NULL;
585f8587 1232 if (ret < 0) {
ac674887 1233fail:
ce1a14dc
PB
1234 acb->common.cb(acb->common.opaque, ret);
1235 qemu_aio_release(acb);
585f8587
FB
1236 return;
1237 }
1238
585f8587 1239 /* post process the read buffer */
ce1a14dc 1240 if (!acb->cluster_offset) {
585f8587 1241 /* nothing to do */
ce1a14dc 1242 } else if (acb->cluster_offset & QCOW_OFLAG_COMPRESSED) {
585f8587
FB
1243 /* nothing to do */
1244 } else {
1245 if (s->crypt_method) {
5fafdf24
TS
1246 encrypt_sectors(s, acb->sector_num, acb->buf, acb->buf,
1247 acb->n, 0,
585f8587
FB
1248 &s->aes_decrypt_key);
1249 }
1250 }
1251
ce1a14dc
PB
1252 acb->nb_sectors -= acb->n;
1253 acb->sector_num += acb->n;
1254 acb->buf += acb->n * 512;
585f8587 1255
ce1a14dc 1256 if (acb->nb_sectors == 0) {
585f8587 1257 /* request completed */
ce1a14dc
PB
1258 acb->common.cb(acb->common.opaque, 0);
1259 qemu_aio_release(acb);
585f8587
FB
1260 return;
1261 }
3b46e624 1262
585f8587 1263 /* prepare next AIO request */
095a9c58
AL
1264 acb->n = acb->nb_sectors;
1265 acb->cluster_offset = get_cluster_offset(bs, acb->sector_num << 9, &acb->n);
ce1a14dc 1266 index_in_cluster = acb->sector_num & (s->cluster_sectors - 1);
ce1a14dc
PB
1267
1268 if (!acb->cluster_offset) {
585f8587
FB
1269 if (bs->backing_hd) {
1270 /* read from the base image */
5fafdf24 1271 n1 = backing_read1(bs->backing_hd, acb->sector_num,
ce1a14dc 1272 acb->buf, acb->n);
a9465922 1273 if (n1 > 0) {
5fafdf24 1274 acb->hd_aiocb = bdrv_aio_read(bs->backing_hd, acb->sector_num,
ce1a14dc
PB
1275 acb->buf, acb->n, qcow_aio_read_cb, acb);
1276 if (acb->hd_aiocb == NULL)
a9465922
FB
1277 goto fail;
1278 } else {
a32ef786
AL
1279 ret = qcow_schedule_bh(qcow_aio_read_bh, acb);
1280 if (ret < 0)
ac674887 1281 goto fail;
a9465922 1282 }
585f8587
FB
1283 } else {
1284 /* Note: in this case, no need to wait */
ce1a14dc 1285 memset(acb->buf, 0, 512 * acb->n);
a32ef786
AL
1286 ret = qcow_schedule_bh(qcow_aio_read_bh, acb);
1287 if (ret < 0)
ac674887 1288 goto fail;
585f8587 1289 }
ce1a14dc 1290 } else if (acb->cluster_offset & QCOW_OFLAG_COMPRESSED) {
585f8587 1291 /* add AIO support for compressed blocks ? */
ce1a14dc 1292 if (decompress_cluster(s, acb->cluster_offset) < 0)
585f8587 1293 goto fail;
5fafdf24 1294 memcpy(acb->buf,
ce1a14dc 1295 s->cluster_cache + index_in_cluster * 512, 512 * acb->n);
a32ef786
AL
1296 ret = qcow_schedule_bh(qcow_aio_read_bh, acb);
1297 if (ret < 0)
ac674887 1298 goto fail;
585f8587 1299 } else {
ce1a14dc 1300 if ((acb->cluster_offset & 511) != 0) {
585f8587
FB
1301 ret = -EIO;
1302 goto fail;
1303 }
ce1a14dc 1304 acb->hd_aiocb = bdrv_aio_read(s->hd,
5fafdf24 1305 (acb->cluster_offset >> 9) + index_in_cluster,
ce1a14dc
PB
1306 acb->buf, acb->n, qcow_aio_read_cb, acb);
1307 if (acb->hd_aiocb == NULL)
585f8587
FB
1308 goto fail;
1309 }
1310}
1311
ce1a14dc
PB
1312static QCowAIOCB *qcow_aio_setup(BlockDriverState *bs,
1313 int64_t sector_num, uint8_t *buf, int nb_sectors,
1314 BlockDriverCompletionFunc *cb, void *opaque)
585f8587 1315{
ce1a14dc
PB
1316 QCowAIOCB *acb;
1317
1318 acb = qemu_aio_get(bs, cb, opaque);
1319 if (!acb)
1320 return NULL;
1321 acb->hd_aiocb = NULL;
1322 acb->sector_num = sector_num;
1323 acb->buf = buf;
1324 acb->nb_sectors = nb_sectors;
1325 acb->n = 0;
1326 acb->cluster_offset = 0;
e976c6a1 1327 acb->l2meta.nb_clusters = 0;
ce1a14dc
PB
1328 return acb;
1329}
1330
1331static BlockDriverAIOCB *qcow_aio_read(BlockDriverState *bs,
1332 int64_t sector_num, uint8_t *buf, int nb_sectors,
1333 BlockDriverCompletionFunc *cb, void *opaque)
1334{
1335 QCowAIOCB *acb;
1336
1337 acb = qcow_aio_setup(bs, sector_num, buf, nb_sectors, cb, opaque);
1338 if (!acb)
1339 return NULL;
585f8587
FB
1340
1341 qcow_aio_read_cb(acb, 0);
ce1a14dc 1342 return &acb->common;
585f8587
FB
1343}
1344
1345static void qcow_aio_write_cb(void *opaque, int ret)
1346{
ce1a14dc
PB
1347 QCowAIOCB *acb = opaque;
1348 BlockDriverState *bs = acb->common.bs;
585f8587 1349 BDRVQcowState *s = bs->opaque;
585f8587 1350 int index_in_cluster;
585f8587 1351 const uint8_t *src_buf;
095a9c58 1352 int n_end;
ce1a14dc
PB
1353
1354 acb->hd_aiocb = NULL;
1355
585f8587
FB
1356 if (ret < 0) {
1357 fail:
ce1a14dc
PB
1358 acb->common.cb(acb->common.opaque, ret);
1359 qemu_aio_release(acb);
585f8587
FB
1360 return;
1361 }
1362
e976c6a1
AL
1363 if (alloc_cluster_link_l2(bs, acb->cluster_offset, &acb->l2meta) < 0) {
1364 free_any_clusters(bs, acb->cluster_offset, acb->l2meta.nb_clusters);
1365 goto fail;
1366 }
1367
ce1a14dc
PB
1368 acb->nb_sectors -= acb->n;
1369 acb->sector_num += acb->n;
1370 acb->buf += acb->n * 512;
585f8587 1371
ce1a14dc 1372 if (acb->nb_sectors == 0) {
585f8587 1373 /* request completed */
ce1a14dc
PB
1374 acb->common.cb(acb->common.opaque, 0);
1375 qemu_aio_release(acb);
585f8587
FB
1376 return;
1377 }
3b46e624 1378
ce1a14dc 1379 index_in_cluster = acb->sector_num & (s->cluster_sectors - 1);
095a9c58
AL
1380 n_end = index_in_cluster + acb->nb_sectors;
1381 if (s->crypt_method &&
1382 n_end > QCOW_MAX_CRYPT_CLUSTERS * s->cluster_sectors)
1383 n_end = QCOW_MAX_CRYPT_CLUSTERS * s->cluster_sectors;
1384
e976c6a1 1385 acb->cluster_offset = alloc_cluster_offset(bs, acb->sector_num << 9,
05203524 1386 index_in_cluster,
e976c6a1
AL
1387 n_end, &acb->n, &acb->l2meta);
1388 if (!acb->cluster_offset || (acb->cluster_offset & 511) != 0) {
585f8587
FB
1389 ret = -EIO;
1390 goto fail;
1391 }
1392 if (s->crypt_method) {
ce1a14dc 1393 if (!acb->cluster_data) {
095a9c58
AL
1394 acb->cluster_data = qemu_mallocz(QCOW_MAX_CRYPT_CLUSTERS *
1395 s->cluster_size);
ce1a14dc 1396 if (!acb->cluster_data) {
585f8587
FB
1397 ret = -ENOMEM;
1398 goto fail;
1399 }
1400 }
5fafdf24 1401 encrypt_sectors(s, acb->sector_num, acb->cluster_data, acb->buf,
ce1a14dc
PB
1402 acb->n, 1, &s->aes_encrypt_key);
1403 src_buf = acb->cluster_data;
585f8587 1404 } else {
ce1a14dc 1405 src_buf = acb->buf;
585f8587 1406 }
ce1a14dc 1407 acb->hd_aiocb = bdrv_aio_write(s->hd,
e976c6a1 1408 (acb->cluster_offset >> 9) + index_in_cluster,
5fafdf24 1409 src_buf, acb->n,
ce1a14dc
PB
1410 qcow_aio_write_cb, acb);
1411 if (acb->hd_aiocb == NULL)
585f8587
FB
1412 goto fail;
1413}
1414
ce1a14dc
PB
1415static BlockDriverAIOCB *qcow_aio_write(BlockDriverState *bs,
1416 int64_t sector_num, const uint8_t *buf, int nb_sectors,
1417 BlockDriverCompletionFunc *cb, void *opaque)
585f8587 1418{
585f8587 1419 BDRVQcowState *s = bs->opaque;
ce1a14dc 1420 QCowAIOCB *acb;
3b46e624 1421
585f8587
FB
1422 s->cluster_cache_offset = -1; /* disable compressed cache */
1423
ce1a14dc
PB
1424 acb = qcow_aio_setup(bs, sector_num, (uint8_t*)buf, nb_sectors, cb, opaque);
1425 if (!acb)
1426 return NULL;
3b46e624 1427
585f8587 1428 qcow_aio_write_cb(acb, 0);
ce1a14dc 1429 return &acb->common;
585f8587
FB
1430}
1431
ce1a14dc 1432static void qcow_aio_cancel(BlockDriverAIOCB *blockacb)
585f8587 1433{
ce1a14dc
PB
1434 QCowAIOCB *acb = (QCowAIOCB *)blockacb;
1435 if (acb->hd_aiocb)
1436 bdrv_aio_cancel(acb->hd_aiocb);
1437 qemu_aio_release(acb);
585f8587
FB
1438}
1439
1440static void qcow_close(BlockDriverState *bs)
1441{
1442 BDRVQcowState *s = bs->opaque;
1443 qemu_free(s->l1_table);
1444 qemu_free(s->l2_cache);
1445 qemu_free(s->cluster_cache);
1446 qemu_free(s->cluster_data);
1447 refcount_close(bs);
1448 bdrv_delete(s->hd);
1449}
1450
1451/* XXX: use std qcow open function ? */
1452typedef struct QCowCreateState {
1453 int cluster_size;
1454 int cluster_bits;
1455 uint16_t *refcount_block;
1456 uint64_t *refcount_table;
1457 int64_t l1_table_offset;
1458 int64_t refcount_table_offset;
1459 int64_t refcount_block_offset;
1460} QCowCreateState;
1461
1462static void create_refcount_update(QCowCreateState *s,
1463 int64_t offset, int64_t size)
1464{
1465 int refcount;
1466 int64_t start, last, cluster_offset;
1467 uint16_t *p;
1468
1469 start = offset & ~(s->cluster_size - 1);
1470 last = (offset + size - 1) & ~(s->cluster_size - 1);
5fafdf24 1471 for(cluster_offset = start; cluster_offset <= last;
585f8587
FB
1472 cluster_offset += s->cluster_size) {
1473 p = &s->refcount_block[cluster_offset >> s->cluster_bits];
1474 refcount = be16_to_cpu(*p);
1475 refcount++;
1476 *p = cpu_to_be16(refcount);
1477 }
1478}
1479
1480static int qcow_create(const char *filename, int64_t total_size,
1481 const char *backing_file, int flags)
1482{
1483 int fd, header_size, backing_filename_len, l1_size, i, shift, l2_bits;
1484 QCowHeader header;
1485 uint64_t tmp, offset;
1486 QCowCreateState s1, *s = &s1;
3b46e624 1487
585f8587
FB
1488 memset(s, 0, sizeof(*s));
1489
1490 fd = open(filename, O_WRONLY | O_CREAT | O_TRUNC | O_BINARY, 0644);
1491 if (fd < 0)
1492 return -1;
1493 memset(&header, 0, sizeof(header));
1494 header.magic = cpu_to_be32(QCOW_MAGIC);
1495 header.version = cpu_to_be32(QCOW_VERSION);
1496 header.size = cpu_to_be64(total_size * 512);
1497 header_size = sizeof(header);
1498 backing_filename_len = 0;
1499 if (backing_file) {
1500 header.backing_file_offset = cpu_to_be64(header_size);
1501 backing_filename_len = strlen(backing_file);
1502 header.backing_file_size = cpu_to_be32(backing_filename_len);
1503 header_size += backing_filename_len;
1504 }
1505 s->cluster_bits = 12; /* 4 KB clusters */
1506 s->cluster_size = 1 << s->cluster_bits;
1507 header.cluster_bits = cpu_to_be32(s->cluster_bits);
1508 header_size = (header_size + 7) & ~7;
ec36ba14 1509 if (flags & BLOCK_FLAG_ENCRYPT) {
585f8587
FB
1510 header.crypt_method = cpu_to_be32(QCOW_CRYPT_AES);
1511 } else {
1512 header.crypt_method = cpu_to_be32(QCOW_CRYPT_NONE);
1513 }
1514 l2_bits = s->cluster_bits - 3;
1515 shift = s->cluster_bits + l2_bits;
1516 l1_size = (((total_size * 512) + (1LL << shift) - 1) >> shift);
1517 offset = align_offset(header_size, s->cluster_size);
1518 s->l1_table_offset = offset;
1519 header.l1_table_offset = cpu_to_be64(s->l1_table_offset);
1520 header.l1_size = cpu_to_be32(l1_size);
15e6690a 1521 offset += align_offset(l1_size * sizeof(uint64_t), s->cluster_size);
585f8587
FB
1522
1523 s->refcount_table = qemu_mallocz(s->cluster_size);
1524 if (!s->refcount_table)
1525 goto fail;
1526 s->refcount_block = qemu_mallocz(s->cluster_size);
1527 if (!s->refcount_block)
1528 goto fail;
3b46e624 1529
585f8587
FB
1530 s->refcount_table_offset = offset;
1531 header.refcount_table_offset = cpu_to_be64(offset);
1532 header.refcount_table_clusters = cpu_to_be32(1);
1533 offset += s->cluster_size;
1534
1535 s->refcount_table[0] = cpu_to_be64(offset);
1536 s->refcount_block_offset = offset;
1537 offset += s->cluster_size;
1538
1539 /* update refcounts */
1540 create_refcount_update(s, 0, header_size);
15e6690a 1541 create_refcount_update(s, s->l1_table_offset, l1_size * sizeof(uint64_t));
585f8587
FB
1542 create_refcount_update(s, s->refcount_table_offset, s->cluster_size);
1543 create_refcount_update(s, s->refcount_block_offset, s->cluster_size);
3b46e624 1544
585f8587
FB
1545 /* write all the data */
1546 write(fd, &header, sizeof(header));
1547 if (backing_file) {
1548 write(fd, backing_file, backing_filename_len);
1549 }
1550 lseek(fd, s->l1_table_offset, SEEK_SET);
1551 tmp = 0;
1552 for(i = 0;i < l1_size; i++) {
1553 write(fd, &tmp, sizeof(tmp));
1554 }
1555 lseek(fd, s->refcount_table_offset, SEEK_SET);
1556 write(fd, s->refcount_table, s->cluster_size);
3b46e624 1557
585f8587
FB
1558 lseek(fd, s->refcount_block_offset, SEEK_SET);
1559 write(fd, s->refcount_block, s->cluster_size);
1560
1561 qemu_free(s->refcount_table);
1562 qemu_free(s->refcount_block);
1563 close(fd);
1564 return 0;
1565 fail:
1566 qemu_free(s->refcount_table);
1567 qemu_free(s->refcount_block);
1568 close(fd);
1569 return -ENOMEM;
1570}
1571
1572static int qcow_make_empty(BlockDriverState *bs)
1573{
1574#if 0
1575 /* XXX: not correct */
1576 BDRVQcowState *s = bs->opaque;
1577 uint32_t l1_length = s->l1_size * sizeof(uint64_t);
1578 int ret;
1579
1580 memset(s->l1_table, 0, l1_length);
1581 if (bdrv_pwrite(s->hd, s->l1_table_offset, s->l1_table, l1_length) < 0)
ac674887 1582 return -1;
585f8587
FB
1583 ret = bdrv_truncate(s->hd, s->l1_table_offset + l1_length);
1584 if (ret < 0)
1585 return ret;
3b46e624 1586
585f8587
FB
1587 l2_cache_reset(bs);
1588#endif
1589 return 0;
1590}
1591
1592/* XXX: put compressed sectors first, then all the cluster aligned
1593 tables to avoid losing bytes in alignment */
5fafdf24 1594static int qcow_write_compressed(BlockDriverState *bs, int64_t sector_num,
585f8587
FB
1595 const uint8_t *buf, int nb_sectors)
1596{
1597 BDRVQcowState *s = bs->opaque;
1598 z_stream strm;
1599 int ret, out_len;
1600 uint8_t *out_buf;
1601 uint64_t cluster_offset;
1602
1603 if (nb_sectors == 0) {
1604 /* align end of file to a sector boundary to ease reading with
1605 sector based I/Os */
1606 cluster_offset = bdrv_getlength(s->hd);
1607 cluster_offset = (cluster_offset + 511) & ~511;
1608 bdrv_truncate(s->hd, cluster_offset);
1609 return 0;
1610 }
1611
1612 if (nb_sectors != s->cluster_sectors)
1613 return -EINVAL;
1614
1615 out_buf = qemu_malloc(s->cluster_size + (s->cluster_size / 1000) + 128);
1616 if (!out_buf)
1617 return -ENOMEM;
1618
1619 /* best compression, small window, no zlib header */
1620 memset(&strm, 0, sizeof(strm));
1621 ret = deflateInit2(&strm, Z_DEFAULT_COMPRESSION,
5fafdf24 1622 Z_DEFLATED, -12,
585f8587
FB
1623 9, Z_DEFAULT_STRATEGY);
1624 if (ret != 0) {
1625 qemu_free(out_buf);
1626 return -1;
1627 }
1628
1629 strm.avail_in = s->cluster_size;
1630 strm.next_in = (uint8_t *)buf;
1631 strm.avail_out = s->cluster_size;
1632 strm.next_out = out_buf;
1633
1634 ret = deflate(&strm, Z_FINISH);
1635 if (ret != Z_STREAM_END && ret != Z_OK) {
1636 qemu_free(out_buf);
1637 deflateEnd(&strm);
1638 return -1;
1639 }
1640 out_len = strm.next_out - out_buf;
1641
1642 deflateEnd(&strm);
1643
1644 if (ret != Z_STREAM_END || out_len >= s->cluster_size) {
1645 /* could not compress: write normal cluster */
1646 qcow_write(bs, sector_num, buf, s->cluster_sectors);
1647 } else {
52d893ec
AL
1648 cluster_offset = alloc_compressed_cluster_offset(bs, sector_num << 9,
1649 out_len);
1650 if (!cluster_offset)
1651 return -1;
585f8587
FB
1652 cluster_offset &= s->cluster_offset_mask;
1653 if (bdrv_pwrite(s->hd, cluster_offset, out_buf, out_len) != out_len) {
1654 qemu_free(out_buf);
1655 return -1;
1656 }
1657 }
3b46e624 1658
585f8587
FB
1659 qemu_free(out_buf);
1660 return 0;
1661}
1662
1663static void qcow_flush(BlockDriverState *bs)
1664{
1665 BDRVQcowState *s = bs->opaque;
1666 bdrv_flush(s->hd);
1667}
1668
1669static int qcow_get_info(BlockDriverState *bs, BlockDriverInfo *bdi)
1670{
1671 BDRVQcowState *s = bs->opaque;
1672 bdi->cluster_size = s->cluster_size;
5fafdf24 1673 bdi->vm_state_offset = (int64_t)s->l1_vm_state_index <<
585f8587 1674 (s->cluster_bits + s->l2_bits);
c4218205 1675 bdi->highest_alloc = s->highest_alloc << s->cluster_bits;
1987530f 1676 bdi->num_free_bytes = s->nc_free << s->cluster_bits;
585f8587
FB
1677 return 0;
1678}
1679
1680/*********************************************************/
1681/* snapshot support */
1682
1683/* update the refcounts of snapshots and the copied flag */
5fafdf24 1684static int update_snapshot_refcount(BlockDriverState *bs,
585f8587
FB
1685 int64_t l1_table_offset,
1686 int l1_size,
1687 int addend)
1688{
1689 BDRVQcowState *s = bs->opaque;
1690 uint64_t *l1_table, *l2_table, l2_offset, offset, l1_size2, l1_allocated;
1691 int64_t old_offset, old_l2_offset;
1692 int l2_size, i, j, l1_modified, l2_modified, nb_csectors, refcount;
3b46e624 1693
585f8587
FB
1694 l2_cache_reset(bs);
1695
1696 l2_table = NULL;
1697 l1_table = NULL;
1698 l1_size2 = l1_size * sizeof(uint64_t);
1699 l1_allocated = 0;
1700 if (l1_table_offset != s->l1_table_offset) {
1701 l1_table = qemu_malloc(l1_size2);
1702 if (!l1_table)
1703 goto fail;
1704 l1_allocated = 1;
5fafdf24 1705 if (bdrv_pread(s->hd, l1_table_offset,
585f8587
FB
1706 l1_table, l1_size2) != l1_size2)
1707 goto fail;
1708 for(i = 0;i < l1_size; i++)
1709 be64_to_cpus(&l1_table[i]);
1710 } else {
1711 assert(l1_size == s->l1_size);
1712 l1_table = s->l1_table;
1713 l1_allocated = 0;
1714 }
3b46e624 1715
585f8587
FB
1716 l2_size = s->l2_size * sizeof(uint64_t);
1717 l2_table = qemu_malloc(l2_size);
1718 if (!l2_table)
1719 goto fail;
1720 l1_modified = 0;
1721 for(i = 0; i < l1_size; i++) {
1722 l2_offset = l1_table[i];
1723 if (l2_offset) {
1724 old_l2_offset = l2_offset;
1725 l2_offset &= ~QCOW_OFLAG_COPIED;
1726 l2_modified = 0;
1727 if (bdrv_pread(s->hd, l2_offset, l2_table, l2_size) != l2_size)
1728 goto fail;
1729 for(j = 0; j < s->l2_size; j++) {
1730 offset = be64_to_cpu(l2_table[j]);
1731 if (offset != 0) {
1732 old_offset = offset;
1733 offset &= ~QCOW_OFLAG_COPIED;
1734 if (offset & QCOW_OFLAG_COMPRESSED) {
5fafdf24 1735 nb_csectors = ((offset >> s->csize_shift) &
585f8587
FB
1736 s->csize_mask) + 1;
1737 if (addend != 0)
1738 update_refcount(bs, (offset & s->cluster_offset_mask) & ~511,
1739 nb_csectors * 512, addend);
1740 /* compressed clusters are never modified */
5fafdf24 1741 refcount = 2;
585f8587
FB
1742 } else {
1743 if (addend != 0) {
1744 refcount = update_cluster_refcount(bs, offset >> s->cluster_bits, addend);
1745 } else {
1746 refcount = get_refcount(bs, offset >> s->cluster_bits);
1747 }
1748 }
1749
1750 if (refcount == 1) {
1751 offset |= QCOW_OFLAG_COPIED;
1752 }
1753 if (offset != old_offset) {
1754 l2_table[j] = cpu_to_be64(offset);
1755 l2_modified = 1;
1756 }
1757 }
1758 }
1759 if (l2_modified) {
5fafdf24 1760 if (bdrv_pwrite(s->hd,
585f8587
FB
1761 l2_offset, l2_table, l2_size) != l2_size)
1762 goto fail;
1763 }
1764
1765 if (addend != 0) {
1766 refcount = update_cluster_refcount(bs, l2_offset >> s->cluster_bits, addend);
1767 } else {
1768 refcount = get_refcount(bs, l2_offset >> s->cluster_bits);
1769 }
1770 if (refcount == 1) {
1771 l2_offset |= QCOW_OFLAG_COPIED;
1772 }
1773 if (l2_offset != old_l2_offset) {
1774 l1_table[i] = l2_offset;
1775 l1_modified = 1;
1776 }
1777 }
1778 }
1779 if (l1_modified) {
1780 for(i = 0; i < l1_size; i++)
1781 cpu_to_be64s(&l1_table[i]);
5fafdf24 1782 if (bdrv_pwrite(s->hd, l1_table_offset, l1_table,
585f8587
FB
1783 l1_size2) != l1_size2)
1784 goto fail;
1785 for(i = 0; i < l1_size; i++)
1786 be64_to_cpus(&l1_table[i]);
1787 }
1788 if (l1_allocated)
1789 qemu_free(l1_table);
1790 qemu_free(l2_table);
1791 return 0;
1792 fail:
1793 if (l1_allocated)
1794 qemu_free(l1_table);
1795 qemu_free(l2_table);
1796 return -EIO;
1797}
1798
1799static void qcow_free_snapshots(BlockDriverState *bs)
1800{
1801 BDRVQcowState *s = bs->opaque;
1802 int i;
1803
1804 for(i = 0; i < s->nb_snapshots; i++) {
1805 qemu_free(s->snapshots[i].name);
1806 qemu_free(s->snapshots[i].id_str);
1807 }
1808 qemu_free(s->snapshots);
1809 s->snapshots = NULL;
1810 s->nb_snapshots = 0;
1811}
1812
1813static int qcow_read_snapshots(BlockDriverState *bs)
1814{
1815 BDRVQcowState *s = bs->opaque;
1816 QCowSnapshotHeader h;
1817 QCowSnapshot *sn;
1818 int i, id_str_size, name_size;
1819 int64_t offset;
1820 uint32_t extra_data_size;
1821
63c75dcd 1822 if (!s->nb_snapshots) {
1823 s->snapshots = NULL;
1824 s->snapshots_size = 0;
1825 return 0;
1826 }
1827
585f8587
FB
1828 offset = s->snapshots_offset;
1829 s->snapshots = qemu_mallocz(s->nb_snapshots * sizeof(QCowSnapshot));
1830 if (!s->snapshots)
1831 goto fail;
1832 for(i = 0; i < s->nb_snapshots; i++) {
1833 offset = align_offset(offset, 8);
1834 if (bdrv_pread(s->hd, offset, &h, sizeof(h)) != sizeof(h))
1835 goto fail;
1836 offset += sizeof(h);
1837 sn = s->snapshots + i;
1838 sn->l1_table_offset = be64_to_cpu(h.l1_table_offset);
1839 sn->l1_size = be32_to_cpu(h.l1_size);
1840 sn->vm_state_size = be32_to_cpu(h.vm_state_size);
1841 sn->date_sec = be32_to_cpu(h.date_sec);
1842 sn->date_nsec = be32_to_cpu(h.date_nsec);
1843 sn->vm_clock_nsec = be64_to_cpu(h.vm_clock_nsec);
1844 extra_data_size = be32_to_cpu(h.extra_data_size);
1845
1846 id_str_size = be16_to_cpu(h.id_str_size);
1847 name_size = be16_to_cpu(h.name_size);
1848
1849 offset += extra_data_size;
1850
1851 sn->id_str = qemu_malloc(id_str_size + 1);
1852 if (!sn->id_str)
1853 goto fail;
1854 if (bdrv_pread(s->hd, offset, sn->id_str, id_str_size) != id_str_size)
1855 goto fail;
1856 offset += id_str_size;
1857 sn->id_str[id_str_size] = '\0';
1858
1859 sn->name = qemu_malloc(name_size + 1);
1860 if (!sn->name)
1861 goto fail;
1862 if (bdrv_pread(s->hd, offset, sn->name, name_size) != name_size)
1863 goto fail;
1864 offset += name_size;
1865 sn->name[name_size] = '\0';
1866 }
1867 s->snapshots_size = offset - s->snapshots_offset;
1868 return 0;
1869 fail:
1870 qcow_free_snapshots(bs);
1871 return -1;
1872}
1873
1874/* add at the end of the file a new list of snapshots */
1875static int qcow_write_snapshots(BlockDriverState *bs)
1876{
1877 BDRVQcowState *s = bs->opaque;
1878 QCowSnapshot *sn;
1879 QCowSnapshotHeader h;
1880 int i, name_size, id_str_size, snapshots_size;
1881 uint64_t data64;
1882 uint32_t data32;
1883 int64_t offset, snapshots_offset;
1884
1885 /* compute the size of the snapshots */
1886 offset = 0;
1887 for(i = 0; i < s->nb_snapshots; i++) {
1888 sn = s->snapshots + i;
1889 offset = align_offset(offset, 8);
1890 offset += sizeof(h);
1891 offset += strlen(sn->id_str);
1892 offset += strlen(sn->name);
1893 }
1894 snapshots_size = offset;
1895
1896 snapshots_offset = alloc_clusters(bs, snapshots_size);
1897 offset = snapshots_offset;
3b46e624 1898
585f8587
FB
1899 for(i = 0; i < s->nb_snapshots; i++) {
1900 sn = s->snapshots + i;
1901 memset(&h, 0, sizeof(h));
1902 h.l1_table_offset = cpu_to_be64(sn->l1_table_offset);
1903 h.l1_size = cpu_to_be32(sn->l1_size);
1904 h.vm_state_size = cpu_to_be32(sn->vm_state_size);
1905 h.date_sec = cpu_to_be32(sn->date_sec);
1906 h.date_nsec = cpu_to_be32(sn->date_nsec);
1907 h.vm_clock_nsec = cpu_to_be64(sn->vm_clock_nsec);
3b46e624 1908
585f8587
FB
1909 id_str_size = strlen(sn->id_str);
1910 name_size = strlen(sn->name);
1911 h.id_str_size = cpu_to_be16(id_str_size);
1912 h.name_size = cpu_to_be16(name_size);
1913 offset = align_offset(offset, 8);
1914 if (bdrv_pwrite(s->hd, offset, &h, sizeof(h)) != sizeof(h))
1915 goto fail;
1916 offset += sizeof(h);
1917 if (bdrv_pwrite(s->hd, offset, sn->id_str, id_str_size) != id_str_size)
1918 goto fail;
1919 offset += id_str_size;
1920 if (bdrv_pwrite(s->hd, offset, sn->name, name_size) != name_size)
1921 goto fail;
1922 offset += name_size;
1923 }
1924
1925 /* update the various header fields */
1926 data64 = cpu_to_be64(snapshots_offset);
1927 if (bdrv_pwrite(s->hd, offsetof(QCowHeader, snapshots_offset),
1928 &data64, sizeof(data64)) != sizeof(data64))
1929 goto fail;
1930 data32 = cpu_to_be32(s->nb_snapshots);
1931 if (bdrv_pwrite(s->hd, offsetof(QCowHeader, nb_snapshots),
1932 &data32, sizeof(data32)) != sizeof(data32))
1933 goto fail;
1934
1935 /* free the old snapshot table */
1936 free_clusters(bs, s->snapshots_offset, s->snapshots_size);
1937 s->snapshots_offset = snapshots_offset;
1938 s->snapshots_size = snapshots_size;
1939 return 0;
1940 fail:
1941 return -1;
1942}
1943
1944static void find_new_snapshot_id(BlockDriverState *bs,
1945 char *id_str, int id_str_size)
1946{
1947 BDRVQcowState *s = bs->opaque;
1948 QCowSnapshot *sn;
1949 int i, id, id_max = 0;
1950
1951 for(i = 0; i < s->nb_snapshots; i++) {
1952 sn = s->snapshots + i;
1953 id = strtoul(sn->id_str, NULL, 10);
1954 if (id > id_max)
1955 id_max = id;
1956 }
1957 snprintf(id_str, id_str_size, "%d", id_max + 1);
1958}
1959
1960static int find_snapshot_by_id(BlockDriverState *bs, const char *id_str)
1961{
1962 BDRVQcowState *s = bs->opaque;
1963 int i;
1964
1965 for(i = 0; i < s->nb_snapshots; i++) {
1966 if (!strcmp(s->snapshots[i].id_str, id_str))
1967 return i;
1968 }
1969 return -1;
1970}
1971
1972static int find_snapshot_by_id_or_name(BlockDriverState *bs, const char *name)
1973{
1974 BDRVQcowState *s = bs->opaque;
1975 int i, ret;
3b46e624 1976
585f8587
FB
1977 ret = find_snapshot_by_id(bs, name);
1978 if (ret >= 0)
1979 return ret;
1980 for(i = 0; i < s->nb_snapshots; i++) {
1981 if (!strcmp(s->snapshots[i].name, name))
1982 return i;
1983 }
1984 return -1;
1985}
1986
1987/* if no id is provided, a new one is constructed */
5fafdf24 1988static int qcow_snapshot_create(BlockDriverState *bs,
585f8587
FB
1989 QEMUSnapshotInfo *sn_info)
1990{
1991 BDRVQcowState *s = bs->opaque;
1992 QCowSnapshot *snapshots1, sn1, *sn = &sn1;
1993 int i, ret;
1994 uint64_t *l1_table = NULL;
3b46e624 1995
585f8587
FB
1996 memset(sn, 0, sizeof(*sn));
1997
1998 if (sn_info->id_str[0] == '\0') {
1999 /* compute a new id */
2000 find_new_snapshot_id(bs, sn_info->id_str, sizeof(sn_info->id_str));
2001 }
2002
2003 /* check that the ID is unique */
2004 if (find_snapshot_by_id(bs, sn_info->id_str) >= 0)
2005 return -ENOENT;
2006
2007 sn->id_str = qemu_strdup(sn_info->id_str);
2008 if (!sn->id_str)
2009 goto fail;
2010 sn->name = qemu_strdup(sn_info->name);
2011 if (!sn->name)
2012 goto fail;
2013 sn->vm_state_size = sn_info->vm_state_size;
2014 sn->date_sec = sn_info->date_sec;
2015 sn->date_nsec = sn_info->date_nsec;
2016 sn->vm_clock_nsec = sn_info->vm_clock_nsec;
2017
2018 ret = update_snapshot_refcount(bs, s->l1_table_offset, s->l1_size, 1);
2019 if (ret < 0)
2020 goto fail;
2021
2022 /* create the L1 table of the snapshot */
2023 sn->l1_table_offset = alloc_clusters(bs, s->l1_size * sizeof(uint64_t));
2024 sn->l1_size = s->l1_size;
2025
2026 l1_table = qemu_malloc(s->l1_size * sizeof(uint64_t));
2027 if (!l1_table)
2028 goto fail;
2029 for(i = 0; i < s->l1_size; i++) {
2030 l1_table[i] = cpu_to_be64(s->l1_table[i]);
2031 }
2032 if (bdrv_pwrite(s->hd, sn->l1_table_offset,
5fafdf24 2033 l1_table, s->l1_size * sizeof(uint64_t)) !=
585f8587
FB
2034 (s->l1_size * sizeof(uint64_t)))
2035 goto fail;
2036 qemu_free(l1_table);
2037 l1_table = NULL;
2038
2039 snapshots1 = qemu_malloc((s->nb_snapshots + 1) * sizeof(QCowSnapshot));
2040 if (!snapshots1)
2041 goto fail;
63c75dcd 2042 if (s->snapshots) {
2043 memcpy(snapshots1, s->snapshots, s->nb_snapshots * sizeof(QCowSnapshot));
2044 qemu_free(s->snapshots);
2045 }
585f8587
FB
2046 s->snapshots = snapshots1;
2047 s->snapshots[s->nb_snapshots++] = *sn;
2048
2049 if (qcow_write_snapshots(bs) < 0)
2050 goto fail;
2051#ifdef DEBUG_ALLOC
2052 check_refcounts(bs);
2053#endif
2054 return 0;
2055 fail:
2056 qemu_free(sn->name);
2057 qemu_free(l1_table);
2058 return -1;
2059}
2060
2061/* copy the snapshot 'snapshot_name' into the current disk image */
5fafdf24 2062static int qcow_snapshot_goto(BlockDriverState *bs,
585f8587
FB
2063 const char *snapshot_id)
2064{
2065 BDRVQcowState *s = bs->opaque;
2066 QCowSnapshot *sn;
2067 int i, snapshot_index, l1_size2;
2068
2069 snapshot_index = find_snapshot_by_id_or_name(bs, snapshot_id);
2070 if (snapshot_index < 0)
2071 return -ENOENT;
2072 sn = &s->snapshots[snapshot_index];
2073
2074 if (update_snapshot_refcount(bs, s->l1_table_offset, s->l1_size, -1) < 0)
2075 goto fail;
2076
2077 if (grow_l1_table(bs, sn->l1_size) < 0)
2078 goto fail;
2079
2080 s->l1_size = sn->l1_size;
2081 l1_size2 = s->l1_size * sizeof(uint64_t);
2082 /* copy the snapshot l1 table to the current l1 table */
5fafdf24 2083 if (bdrv_pread(s->hd, sn->l1_table_offset,
585f8587
FB
2084 s->l1_table, l1_size2) != l1_size2)
2085 goto fail;
2086 if (bdrv_pwrite(s->hd, s->l1_table_offset,
2087 s->l1_table, l1_size2) != l1_size2)
2088 goto fail;
2089 for(i = 0;i < s->l1_size; i++) {
2090 be64_to_cpus(&s->l1_table[i]);
2091 }
2092
2093 if (update_snapshot_refcount(bs, s->l1_table_offset, s->l1_size, 1) < 0)
2094 goto fail;
2095
2096#ifdef DEBUG_ALLOC
2097 check_refcounts(bs);
2098#endif
2099 return 0;
2100 fail:
2101 return -EIO;
2102}
2103
2104static int qcow_snapshot_delete(BlockDriverState *bs, const char *snapshot_id)
2105{
2106 BDRVQcowState *s = bs->opaque;
2107 QCowSnapshot *sn;
2108 int snapshot_index, ret;
3b46e624 2109
585f8587
FB
2110 snapshot_index = find_snapshot_by_id_or_name(bs, snapshot_id);
2111 if (snapshot_index < 0)
2112 return -ENOENT;
2113 sn = &s->snapshots[snapshot_index];
2114
2115 ret = update_snapshot_refcount(bs, sn->l1_table_offset, sn->l1_size, -1);
2116 if (ret < 0)
2117 return ret;
2118 /* must update the copied flag on the current cluster offsets */
2119 ret = update_snapshot_refcount(bs, s->l1_table_offset, s->l1_size, 0);
2120 if (ret < 0)
2121 return ret;
2122 free_clusters(bs, sn->l1_table_offset, sn->l1_size * sizeof(uint64_t));
2123
2124 qemu_free(sn->id_str);
2125 qemu_free(sn->name);
2126 memmove(sn, sn + 1, (s->nb_snapshots - snapshot_index - 1) * sizeof(*sn));
2127 s->nb_snapshots--;
2128 ret = qcow_write_snapshots(bs);
2129 if (ret < 0) {
2130 /* XXX: restore snapshot if error ? */
2131 return ret;
2132 }
2133#ifdef DEBUG_ALLOC
2134 check_refcounts(bs);
2135#endif
2136 return 0;
2137}
2138
5fafdf24 2139static int qcow_snapshot_list(BlockDriverState *bs,
585f8587
FB
2140 QEMUSnapshotInfo **psn_tab)
2141{
2142 BDRVQcowState *s = bs->opaque;
2143 QEMUSnapshotInfo *sn_tab, *sn_info;
2144 QCowSnapshot *sn;
2145 int i;
2146
2147 sn_tab = qemu_mallocz(s->nb_snapshots * sizeof(QEMUSnapshotInfo));
2148 if (!sn_tab)
2149 goto fail;
2150 for(i = 0; i < s->nb_snapshots; i++) {
2151 sn_info = sn_tab + i;
2152 sn = s->snapshots + i;
2153 pstrcpy(sn_info->id_str, sizeof(sn_info->id_str),
2154 sn->id_str);
2155 pstrcpy(sn_info->name, sizeof(sn_info->name),
2156 sn->name);
2157 sn_info->vm_state_size = sn->vm_state_size;
2158 sn_info->date_sec = sn->date_sec;
2159 sn_info->date_nsec = sn->date_nsec;
2160 sn_info->vm_clock_nsec = sn->vm_clock_nsec;
2161 }
2162 *psn_tab = sn_tab;
2163 return s->nb_snapshots;
2164 fail:
2165 qemu_free(sn_tab);
2166 *psn_tab = NULL;
2167 return -ENOMEM;
2168}
2169
2170/*********************************************************/
2171/* refcount handling */
2172
2173static int refcount_init(BlockDriverState *bs)
2174{
2175 BDRVQcowState *s = bs->opaque;
2176 int ret, refcount_table_size2, i;
3b46e624 2177
585f8587
FB
2178 s->refcount_block_cache = qemu_malloc(s->cluster_size);
2179 if (!s->refcount_block_cache)
2180 goto fail;
2181 refcount_table_size2 = s->refcount_table_size * sizeof(uint64_t);
2182 s->refcount_table = qemu_malloc(refcount_table_size2);
2183 if (!s->refcount_table)
2184 goto fail;
2185 if (s->refcount_table_size > 0) {
2186 ret = bdrv_pread(s->hd, s->refcount_table_offset,
2187 s->refcount_table, refcount_table_size2);
2188 if (ret != refcount_table_size2)
2189 goto fail;
2190 for(i = 0; i < s->refcount_table_size; i++)
2191 be64_to_cpus(&s->refcount_table[i]);
2192 }
2193 return 0;
2194 fail:
2195 return -ENOMEM;
2196}
2197
2198static void refcount_close(BlockDriverState *bs)
2199{
2200 BDRVQcowState *s = bs->opaque;
2201 qemu_free(s->refcount_block_cache);
2202 qemu_free(s->refcount_table);
2203}
2204
2205
5fafdf24 2206static int load_refcount_block(BlockDriverState *bs,
585f8587
FB
2207 int64_t refcount_block_offset)
2208{
2209 BDRVQcowState *s = bs->opaque;
2210 int ret;
5fafdf24 2211 ret = bdrv_pread(s->hd, refcount_block_offset, s->refcount_block_cache,
585f8587
FB
2212 s->cluster_size);
2213 if (ret != s->cluster_size)
2214 return -EIO;
2215 s->refcount_block_cache_offset = refcount_block_offset;
2216 return 0;
2217}
2218
1987530f 2219static void scan_refcount(BlockDriverState *bs, int64_t *high, int64_t *free)
5d4cbd78
AL
2220{
2221 BDRVQcowState *s = bs->opaque;
1987530f
AL
2222 int64_t refcnt_index, cluster_index, cluster_end, h = 0, f = 0;
2223 int64_t tail = 0; /* do not count last consecutive free entries */
5d4cbd78
AL
2224
2225 for (refcnt_index=0; refcnt_index < s->refcount_table_size; refcnt_index++){
2226 if (s->refcount_table[refcnt_index] == 0) {
1987530f
AL
2227 f += 1 << (s->cluster_bits - REFCOUNT_SHIFT);
2228 tail += 1 << (s->cluster_bits - REFCOUNT_SHIFT);
5d4cbd78
AL
2229 continue;
2230 }
2231 cluster_index = refcnt_index << (s->cluster_bits - REFCOUNT_SHIFT);
2232 cluster_end = (refcnt_index + 1) << (s->cluster_bits - REFCOUNT_SHIFT);
2233 for ( ; cluster_index < cluster_end; cluster_index++) {
1987530f
AL
2234 if (get_refcount(bs, cluster_index) == 0) {
2235 f++;
2236 tail++;
2237 }
2238 else {
5d4cbd78 2239 h = cluster_index;
1987530f
AL
2240 tail = 0;
2241 }
5d4cbd78
AL
2242 }
2243 }
2244
1987530f
AL
2245 f -= tail;
2246 if (free)
2247 *free = f;
5d4cbd78
AL
2248 if (high)
2249 *high = (h+1);
2250}
2251
585f8587
FB
2252static int get_refcount(BlockDriverState *bs, int64_t cluster_index)
2253{
2254 BDRVQcowState *s = bs->opaque;
2255 int refcount_table_index, block_index;
2256 int64_t refcount_block_offset;
2257
2258 refcount_table_index = cluster_index >> (s->cluster_bits - REFCOUNT_SHIFT);
2259 if (refcount_table_index >= s->refcount_table_size)
2260 return 0;
2261 refcount_block_offset = s->refcount_table[refcount_table_index];
2262 if (!refcount_block_offset)
2263 return 0;
2264 if (refcount_block_offset != s->refcount_block_cache_offset) {
2265 /* better than nothing: return allocated if read error */
2266 if (load_refcount_block(bs, refcount_block_offset) < 0)
2267 return 1;
2268 }
5fafdf24 2269 block_index = cluster_index &
585f8587
FB
2270 ((1 << (s->cluster_bits - REFCOUNT_SHIFT)) - 1);
2271 return be16_to_cpu(s->refcount_block_cache[block_index]);
2272}
2273
2274/* return < 0 if error */
2275static int64_t alloc_clusters_noref(BlockDriverState *bs, int64_t size)
2276{
2277 BDRVQcowState *s = bs->opaque;
2278 int i, nb_clusters;
2279
6db6c638
AL
2280 nb_clusters = size_to_clusters(s, size);
2281retry:
2282 for(i = 0; i < nb_clusters; i++) {
2283 int64_t i = s->free_cluster_index++;
2284 if (get_refcount(bs, i) != 0)
2285 goto retry;
2286 }
585f8587 2287#ifdef DEBUG_ALLOC2
6db6c638
AL
2288 printf("alloc_clusters: size=%lld -> %lld\n",
2289 size,
2290 (s->free_cluster_index - nb_clusters) << s->cluster_bits);
585f8587 2291#endif
5d4cbd78 2292
1987530f
AL
2293 if (s->highest_alloc < s->free_cluster_index) {
2294 s->nc_free += (s->free_cluster_index - s->highest_alloc);
5d4cbd78 2295 s->highest_alloc = s->free_cluster_index;
1987530f 2296 }
5d4cbd78 2297
6db6c638 2298 return (s->free_cluster_index - nb_clusters) << s->cluster_bits;
585f8587
FB
2299}
2300
2301static int64_t alloc_clusters(BlockDriverState *bs, int64_t size)
2302{
2303 int64_t offset;
2304
2305 offset = alloc_clusters_noref(bs, size);
2306 update_refcount(bs, offset, size, 1);
2307 return offset;
2308}
2309
2310/* only used to allocate compressed sectors. We try to allocate
2311 contiguous sectors. size must be <= cluster_size */
2312static int64_t alloc_bytes(BlockDriverState *bs, int size)
2313{
2314 BDRVQcowState *s = bs->opaque;
2315 int64_t offset, cluster_offset;
2316 int free_in_cluster;
3b46e624 2317
585f8587
FB
2318 assert(size > 0 && size <= s->cluster_size);
2319 if (s->free_byte_offset == 0) {
2320 s->free_byte_offset = alloc_clusters(bs, s->cluster_size);
2321 }
2322 redo:
5fafdf24 2323 free_in_cluster = s->cluster_size -
585f8587
FB
2324 (s->free_byte_offset & (s->cluster_size - 1));
2325 if (size <= free_in_cluster) {
2326 /* enough space in current cluster */
2327 offset = s->free_byte_offset;
2328 s->free_byte_offset += size;
2329 free_in_cluster -= size;
2330 if (free_in_cluster == 0)
2331 s->free_byte_offset = 0;
2332 if ((offset & (s->cluster_size - 1)) != 0)
2333 update_cluster_refcount(bs, offset >> s->cluster_bits, 1);
2334 } else {
2335 offset = alloc_clusters(bs, s->cluster_size);
2336 cluster_offset = s->free_byte_offset & ~(s->cluster_size - 1);
2337 if ((cluster_offset + s->cluster_size) == offset) {
2338 /* we are lucky: contiguous data */
2339 offset = s->free_byte_offset;
2340 update_cluster_refcount(bs, offset >> s->cluster_bits, 1);
2341 s->free_byte_offset += size;
2342 } else {
2343 s->free_byte_offset = offset;
2344 goto redo;
2345 }
2346 }
2347 return offset;
2348}
2349
5fafdf24 2350static void free_clusters(BlockDriverState *bs,
585f8587
FB
2351 int64_t offset, int64_t size)
2352{
2353 update_refcount(bs, offset, size, -1);
2354}
2355
2356static int grow_refcount_table(BlockDriverState *bs, int min_size)
2357{
2358 BDRVQcowState *s = bs->opaque;
2359 int new_table_size, new_table_size2, refcount_table_clusters, i, ret;
2360 uint64_t *new_table;
2361 int64_t table_offset;
643e5399 2362 uint8_t data[12];
23be50f1
TS
2363 int old_table_size;
2364 int64_t old_table_offset;
585f8587
FB
2365
2366 if (min_size <= s->refcount_table_size)
2367 return 0;
2368 /* compute new table size */
2369 refcount_table_clusters = s->refcount_table_size >> (s->cluster_bits - 3);
2370 for(;;) {
2371 if (refcount_table_clusters == 0) {
2372 refcount_table_clusters = 1;
2373 } else {
2374 refcount_table_clusters = (refcount_table_clusters * 3 + 1) / 2;
2375 }
2376 new_table_size = refcount_table_clusters << (s->cluster_bits - 3);
2377 if (min_size <= new_table_size)
2378 break;
2379 }
15e6690a
FB
2380#ifdef DEBUG_ALLOC2
2381 printf("grow_refcount_table from %d to %d\n",
2382 s->refcount_table_size,
2383 new_table_size);
2384#endif
585f8587
FB
2385 new_table_size2 = new_table_size * sizeof(uint64_t);
2386 new_table = qemu_mallocz(new_table_size2);
2387 if (!new_table)
2388 return -ENOMEM;
5fafdf24 2389 memcpy(new_table, s->refcount_table,
585f8587
FB
2390 s->refcount_table_size * sizeof(uint64_t));
2391 for(i = 0; i < s->refcount_table_size; i++)
2392 cpu_to_be64s(&new_table[i]);
2393 /* Note: we cannot update the refcount now to avoid recursion */
2394 table_offset = alloc_clusters_noref(bs, new_table_size2);
2395 ret = bdrv_pwrite(s->hd, table_offset, new_table, new_table_size2);
5fafdf24 2396 if (ret != new_table_size2)
585f8587
FB
2397 goto fail;
2398 for(i = 0; i < s->refcount_table_size; i++)
2399 be64_to_cpus(&new_table[i]);
2400
643e5399
AL
2401 cpu_to_be64w((uint64_t*)data, table_offset);
2402 cpu_to_be32w((uint32_t*)(data + 8), refcount_table_clusters);
585f8587 2403 if (bdrv_pwrite(s->hd, offsetof(QCowHeader, refcount_table_offset),
643e5399 2404 data, sizeof(data)) != sizeof(data))
585f8587
FB
2405 goto fail;
2406 qemu_free(s->refcount_table);
23be50f1
TS
2407 old_table_offset = s->refcount_table_offset;
2408 old_table_size = s->refcount_table_size;
585f8587
FB
2409 s->refcount_table = new_table;
2410 s->refcount_table_size = new_table_size;
a4080ece 2411 s->refcount_table_offset = table_offset;
585f8587
FB
2412
2413 update_refcount(bs, table_offset, new_table_size2, 1);
23be50f1 2414 free_clusters(bs, old_table_offset, old_table_size * sizeof(uint64_t));
585f8587
FB
2415 return 0;
2416 fail:
2417 free_clusters(bs, table_offset, new_table_size2);
2418 qemu_free(new_table);
2419 return -EIO;
2420}
2421
2422/* addend must be 1 or -1 */
2423/* XXX: cache several refcount block clusters ? */
5fafdf24 2424static int update_cluster_refcount(BlockDriverState *bs,
585f8587
FB
2425 int64_t cluster_index,
2426 int addend)
2427{
2428 BDRVQcowState *s = bs->opaque;
2429 int64_t offset, refcount_block_offset;
2430 int ret, refcount_table_index, block_index, refcount;
2431 uint64_t data64;
2432
2433 refcount_table_index = cluster_index >> (s->cluster_bits - REFCOUNT_SHIFT);
2434 if (refcount_table_index >= s->refcount_table_size) {
2435 if (addend < 0)
2436 return -EINVAL;
2437 ret = grow_refcount_table(bs, refcount_table_index + 1);
2438 if (ret < 0)
2439 return ret;
2440 }
2441 refcount_block_offset = s->refcount_table[refcount_table_index];
2442 if (!refcount_block_offset) {
2443 if (addend < 0)
2444 return -EINVAL;
2445 /* create a new refcount block */
2446 /* Note: we cannot update the refcount now to avoid recursion */
2447 offset = alloc_clusters_noref(bs, s->cluster_size);
2448 memset(s->refcount_block_cache, 0, s->cluster_size);
2449 ret = bdrv_pwrite(s->hd, offset, s->refcount_block_cache, s->cluster_size);
2450 if (ret != s->cluster_size)
2451 return -EINVAL;
2452 s->refcount_table[refcount_table_index] = offset;
2453 data64 = cpu_to_be64(offset);
5fafdf24
TS
2454 ret = bdrv_pwrite(s->hd, s->refcount_table_offset +
2455 refcount_table_index * sizeof(uint64_t),
585f8587
FB
2456 &data64, sizeof(data64));
2457 if (ret != sizeof(data64))
2458 return -EINVAL;
2459
2460 refcount_block_offset = offset;
2461 s->refcount_block_cache_offset = offset;
2462 update_refcount(bs, offset, s->cluster_size, 1);
2463 } else {
2464 if (refcount_block_offset != s->refcount_block_cache_offset) {
2465 if (load_refcount_block(bs, refcount_block_offset) < 0)
2466 return -EIO;
2467 }
2468 }
2469 /* we can update the count and save it */
5fafdf24 2470 block_index = cluster_index &
585f8587
FB
2471 ((1 << (s->cluster_bits - REFCOUNT_SHIFT)) - 1);
2472 refcount = be16_to_cpu(s->refcount_block_cache[block_index]);
1987530f
AL
2473
2474 if (refcount == 1 && addend == -1)
2475 s->nc_free += 1;
2476 else if (refcount == 0 && addend == 1)
2477 s->nc_free -= 1;
2478
585f8587
FB
2479 refcount += addend;
2480 if (refcount < 0 || refcount > 0xffff)
2481 return -EINVAL;
2482 if (refcount == 0 && cluster_index < s->free_cluster_index) {
2483 s->free_cluster_index = cluster_index;
2484 }
2485 s->refcount_block_cache[block_index] = cpu_to_be16(refcount);
5fafdf24
TS
2486 if (bdrv_pwrite(s->hd,
2487 refcount_block_offset + (block_index << REFCOUNT_SHIFT),
585f8587
FB
2488 &s->refcount_block_cache[block_index], 2) != 2)
2489 return -EIO;
2490 return refcount;
2491}
2492
5fafdf24
TS
2493static void update_refcount(BlockDriverState *bs,
2494 int64_t offset, int64_t length,
585f8587
FB
2495 int addend)
2496{
2497 BDRVQcowState *s = bs->opaque;
2498 int64_t start, last, cluster_offset;
2499
2500#ifdef DEBUG_ALLOC2
5fafdf24 2501 printf("update_refcount: offset=%lld size=%lld addend=%d\n",
585f8587
FB
2502 offset, length, addend);
2503#endif
2504 if (length <= 0)
2505 return;
2506 start = offset & ~(s->cluster_size - 1);
2507 last = (offset + length - 1) & ~(s->cluster_size - 1);
5fafdf24 2508 for(cluster_offset = start; cluster_offset <= last;
585f8587
FB
2509 cluster_offset += s->cluster_size) {
2510 update_cluster_refcount(bs, cluster_offset >> s->cluster_bits, addend);
2511 }
2512}
2513
2514#ifdef DEBUG_ALLOC
5fafdf24
TS
2515static void inc_refcounts(BlockDriverState *bs,
2516 uint16_t *refcount_table,
585f8587
FB
2517 int refcount_table_size,
2518 int64_t offset, int64_t size)
2519{
2520 BDRVQcowState *s = bs->opaque;
2521 int64_t start, last, cluster_offset;
2522 int k;
3b46e624 2523
585f8587
FB
2524 if (size <= 0)
2525 return;
2526
2527 start = offset & ~(s->cluster_size - 1);
2528 last = (offset + size - 1) & ~(s->cluster_size - 1);
5fafdf24 2529 for(cluster_offset = start; cluster_offset <= last;
585f8587
FB
2530 cluster_offset += s->cluster_size) {
2531 k = cluster_offset >> s->cluster_bits;
2532 if (k < 0 || k >= refcount_table_size) {
2533 printf("ERROR: invalid cluster offset=0x%llx\n", cluster_offset);
2534 } else {
2535 if (++refcount_table[k] == 0) {
2536 printf("ERROR: overflow cluster offset=0x%llx\n", cluster_offset);
2537 }
2538 }
2539 }
2540}
2541
5fafdf24
TS
2542static int check_refcounts_l1(BlockDriverState *bs,
2543 uint16_t *refcount_table,
585f8587
FB
2544 int refcount_table_size,
2545 int64_t l1_table_offset, int l1_size,
2546 int check_copied)
2547{
2548 BDRVQcowState *s = bs->opaque;
2549 uint64_t *l1_table, *l2_table, l2_offset, offset, l1_size2;
2550 int l2_size, i, j, nb_csectors, refcount;
2551
2552 l2_table = NULL;
2553 l1_size2 = l1_size * sizeof(uint64_t);
2554
2555 inc_refcounts(bs, refcount_table, refcount_table_size,
2556 l1_table_offset, l1_size2);
2557
2558 l1_table = qemu_malloc(l1_size2);
2559 if (!l1_table)
2560 goto fail;
5fafdf24 2561 if (bdrv_pread(s->hd, l1_table_offset,
585f8587
FB
2562 l1_table, l1_size2) != l1_size2)
2563 goto fail;
2564 for(i = 0;i < l1_size; i++)
2565 be64_to_cpus(&l1_table[i]);
3b46e624 2566
585f8587
FB
2567 l2_size = s->l2_size * sizeof(uint64_t);
2568 l2_table = qemu_malloc(l2_size);
2569 if (!l2_table)
2570 goto fail;
2571 for(i = 0; i < l1_size; i++) {
2572 l2_offset = l1_table[i];
2573 if (l2_offset) {
2574 if (check_copied) {
2575 refcount = get_refcount(bs, (l2_offset & ~QCOW_OFLAG_COPIED) >> s->cluster_bits);
2576 if ((refcount == 1) != ((l2_offset & QCOW_OFLAG_COPIED) != 0)) {
2577 printf("ERROR OFLAG_COPIED: l2_offset=%llx refcount=%d\n",
2578 l2_offset, refcount);
2579 }
2580 }
2581 l2_offset &= ~QCOW_OFLAG_COPIED;
2582 if (bdrv_pread(s->hd, l2_offset, l2_table, l2_size) != l2_size)
2583 goto fail;
2584 for(j = 0; j < s->l2_size; j++) {
2585 offset = be64_to_cpu(l2_table[j]);
2586 if (offset != 0) {
2587 if (offset & QCOW_OFLAG_COMPRESSED) {
2588 if (offset & QCOW_OFLAG_COPIED) {
2589 printf("ERROR: cluster %lld: copied flag must never be set for compressed clusters\n",
2590 offset >> s->cluster_bits);
2591 offset &= ~QCOW_OFLAG_COPIED;
2592 }
5fafdf24 2593 nb_csectors = ((offset >> s->csize_shift) &
585f8587
FB
2594 s->csize_mask) + 1;
2595 offset &= s->cluster_offset_mask;
5fafdf24 2596 inc_refcounts(bs, refcount_table,
585f8587
FB
2597 refcount_table_size,
2598 offset & ~511, nb_csectors * 512);
2599 } else {
2600 if (check_copied) {
2601 refcount = get_refcount(bs, (offset & ~QCOW_OFLAG_COPIED) >> s->cluster_bits);
2602 if ((refcount == 1) != ((offset & QCOW_OFLAG_COPIED) != 0)) {
2603 printf("ERROR OFLAG_COPIED: offset=%llx refcount=%d\n",
2604 offset, refcount);
2605 }
2606 }
2607 offset &= ~QCOW_OFLAG_COPIED;
5fafdf24 2608 inc_refcounts(bs, refcount_table,
585f8587
FB
2609 refcount_table_size,
2610 offset, s->cluster_size);
2611 }
2612 }
2613 }
5fafdf24 2614 inc_refcounts(bs, refcount_table,
585f8587
FB
2615 refcount_table_size,
2616 l2_offset,
2617 s->cluster_size);
2618 }
2619 }
2620 qemu_free(l1_table);
2621 qemu_free(l2_table);
2622 return 0;
2623 fail:
2624 printf("ERROR: I/O error in check_refcounts_l1\n");
2625 qemu_free(l1_table);
2626 qemu_free(l2_table);
2627 return -EIO;
2628}
2629
2630static void check_refcounts(BlockDriverState *bs)
2631{
2632 BDRVQcowState *s = bs->opaque;
2633 int64_t size;
2634 int nb_clusters, refcount1, refcount2, i;
2635 QCowSnapshot *sn;
2636 uint16_t *refcount_table;
2637
2638 size = bdrv_getlength(s->hd);
6db6c638 2639 nb_clusters = size_to_clusters(s, size);
585f8587 2640 refcount_table = qemu_mallocz(nb_clusters * sizeof(uint16_t));
15e6690a 2641
585f8587
FB
2642 /* header */
2643 inc_refcounts(bs, refcount_table, nb_clusters,
2644 0, s->cluster_size);
3b46e624 2645
585f8587
FB
2646 check_refcounts_l1(bs, refcount_table, nb_clusters,
2647 s->l1_table_offset, s->l1_size, 1);
2648
2649 /* snapshots */
2650 for(i = 0; i < s->nb_snapshots; i++) {
2651 sn = s->snapshots + i;
2652 check_refcounts_l1(bs, refcount_table, nb_clusters,
2653 sn->l1_table_offset, sn->l1_size, 0);
2654 }
2655 inc_refcounts(bs, refcount_table, nb_clusters,
2656 s->snapshots_offset, s->snapshots_size);
2657
2658 /* refcount data */
2659 inc_refcounts(bs, refcount_table, nb_clusters,
5fafdf24 2660 s->refcount_table_offset,
585f8587
FB
2661 s->refcount_table_size * sizeof(uint64_t));
2662 for(i = 0; i < s->refcount_table_size; i++) {
2663 int64_t offset;
2664 offset = s->refcount_table[i];
2665 if (offset != 0) {
2666 inc_refcounts(bs, refcount_table, nb_clusters,
2667 offset, s->cluster_size);
2668 }
2669 }
2670
2671 /* compare ref counts */
2672 for(i = 0; i < nb_clusters; i++) {
2673 refcount1 = get_refcount(bs, i);
2674 refcount2 = refcount_table[i];
2675 if (refcount1 != refcount2)
2676 printf("ERROR cluster %d refcount=%d reference=%d\n",
2677 i, refcount1, refcount2);
2678 }
2679
2680 qemu_free(refcount_table);
2681}
2682
2683#if 0
2684static void dump_refcounts(BlockDriverState *bs)
2685{
2686 BDRVQcowState *s = bs->opaque;
2687 int64_t nb_clusters, k, k1, size;
2688 int refcount;
2689
2690 size = bdrv_getlength(s->hd);
6db6c638 2691 nb_clusters = size_to_clusters(s, size);
585f8587
FB
2692 for(k = 0; k < nb_clusters;) {
2693 k1 = k;
2694 refcount = get_refcount(bs, k);
2695 k++;
2696 while (k < nb_clusters && get_refcount(bs, k) == refcount)
2697 k++;
2698 printf("%lld: refcount=%d nb=%lld\n", k, refcount, k - k1);
2699 }
2700}
2701#endif
2702#endif
2703
2704BlockDriver bdrv_qcow2 = {
2705 "qcow2",
2706 sizeof(BDRVQcowState),
2707 qcow_probe,
2708 qcow_open,
2709 NULL,
2710 NULL,
2711 qcow_close,
2712 qcow_create,
2713 qcow_flush,
2714 qcow_is_allocated,
2715 qcow_set_key,
2716 qcow_make_empty,
2717
585f8587
FB
2718 .bdrv_aio_read = qcow_aio_read,
2719 .bdrv_aio_write = qcow_aio_write,
2720 .bdrv_aio_cancel = qcow_aio_cancel,
ce1a14dc 2721 .aiocb_size = sizeof(QCowAIOCB),
585f8587
FB
2722 .bdrv_write_compressed = qcow_write_compressed,
2723
2724 .bdrv_snapshot_create = qcow_snapshot_create,
2725 .bdrv_snapshot_goto = qcow_snapshot_goto,
2726 .bdrv_snapshot_delete = qcow_snapshot_delete,
2727 .bdrv_snapshot_list = qcow_snapshot_list,
2728 .bdrv_get_info = qcow_get_info,
2729};