KVM: Rework VCPU state writeback API
[qemu-kvm/amd-iommu.git] / block / qcow2-refcount.c
blob917fc88d9c888b5265cb8afe4b86fc2607bccadf
1 /*
2 * Block driver for the QCOW version 2 format
4 * Copyright (c) 2004-2006 Fabrice Bellard
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:
13 * The above copyright notice and this permission notice shall be included in
14 * all copies or substantial portions of the Software.
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.
25 #include "qemu-common.h"
26 #include "block_int.h"
27 #include "block/qcow2.h"
29 static int64_t alloc_clusters_noref(BlockDriverState *bs, int64_t size);
30 static int QEMU_WARN_UNUSED_RESULT update_refcount(BlockDriverState *bs,
31 int64_t offset, int64_t length,
32 int addend);
35 static int cache_refcount_updates = 0;
37 static int write_refcount_block(BDRVQcowState *s)
39 size_t size = s->cluster_size;
41 if (s->refcount_block_cache_offset == 0) {
42 return 0;
45 if (bdrv_pwrite(s->hd, s->refcount_block_cache_offset,
46 s->refcount_block_cache, size) != size)
48 return -EIO;
51 return 0;
54 /*********************************************************/
55 /* refcount handling */
57 int qcow2_refcount_init(BlockDriverState *bs)
59 BDRVQcowState *s = bs->opaque;
60 int ret, refcount_table_size2, i;
62 s->refcount_block_cache = qemu_malloc(s->cluster_size);
63 refcount_table_size2 = s->refcount_table_size * sizeof(uint64_t);
64 s->refcount_table = qemu_malloc(refcount_table_size2);
65 if (s->refcount_table_size > 0) {
66 ret = bdrv_pread(s->hd, s->refcount_table_offset,
67 s->refcount_table, refcount_table_size2);
68 if (ret != refcount_table_size2)
69 goto fail;
70 for(i = 0; i < s->refcount_table_size; i++)
71 be64_to_cpus(&s->refcount_table[i]);
73 return 0;
74 fail:
75 return -ENOMEM;
78 void qcow2_refcount_close(BlockDriverState *bs)
80 BDRVQcowState *s = bs->opaque;
81 qemu_free(s->refcount_block_cache);
82 qemu_free(s->refcount_table);
86 static int load_refcount_block(BlockDriverState *bs,
87 int64_t refcount_block_offset)
89 BDRVQcowState *s = bs->opaque;
90 int ret;
92 if (cache_refcount_updates) {
93 write_refcount_block(s);
96 ret = bdrv_pread(s->hd, refcount_block_offset, s->refcount_block_cache,
97 s->cluster_size);
98 if (ret != s->cluster_size)
99 return -EIO;
100 s->refcount_block_cache_offset = refcount_block_offset;
101 return 0;
104 static int get_refcount(BlockDriverState *bs, int64_t cluster_index)
106 BDRVQcowState *s = bs->opaque;
107 int refcount_table_index, block_index;
108 int64_t refcount_block_offset;
110 refcount_table_index = cluster_index >> (s->cluster_bits - REFCOUNT_SHIFT);
111 if (refcount_table_index >= s->refcount_table_size)
112 return 0;
113 refcount_block_offset = s->refcount_table[refcount_table_index];
114 if (!refcount_block_offset)
115 return 0;
116 if (refcount_block_offset != s->refcount_block_cache_offset) {
117 /* better than nothing: return allocated if read error */
118 if (load_refcount_block(bs, refcount_block_offset) < 0)
119 return 1;
121 block_index = cluster_index &
122 ((1 << (s->cluster_bits - REFCOUNT_SHIFT)) - 1);
123 return be16_to_cpu(s->refcount_block_cache[block_index]);
127 * Rounds the refcount table size up to avoid growing the table for each single
128 * refcount block that is allocated.
130 static unsigned int next_refcount_table_size(BDRVQcowState *s,
131 unsigned int min_size)
133 unsigned int min_clusters = (min_size >> (s->cluster_bits - 3)) + 1;
134 unsigned int refcount_table_clusters =
135 MAX(1, s->refcount_table_size >> (s->cluster_bits - 3));
137 while (min_clusters > refcount_table_clusters) {
138 refcount_table_clusters = (refcount_table_clusters * 3 + 1) / 2;
141 return refcount_table_clusters << (s->cluster_bits - 3);
145 /* Checks if two offsets are described by the same refcount block */
146 static int in_same_refcount_block(BDRVQcowState *s, uint64_t offset_a,
147 uint64_t offset_b)
149 uint64_t block_a = offset_a >> (2 * s->cluster_bits - REFCOUNT_SHIFT);
150 uint64_t block_b = offset_b >> (2 * s->cluster_bits - REFCOUNT_SHIFT);
152 return (block_a == block_b);
156 * Loads a refcount block. If it doesn't exist yet, it is allocated first
157 * (including growing the refcount table if needed).
159 * Returns the offset of the refcount block on success or -errno in error case
161 static int64_t alloc_refcount_block(BlockDriverState *bs, int64_t cluster_index)
163 BDRVQcowState *s = bs->opaque;
164 unsigned int refcount_table_index;
165 int ret;
167 /* Find the refcount block for the given cluster */
168 refcount_table_index = cluster_index >> (s->cluster_bits - REFCOUNT_SHIFT);
170 if (refcount_table_index < s->refcount_table_size) {
172 uint64_t refcount_block_offset =
173 s->refcount_table[refcount_table_index];
175 /* If it's already there, we're done */
176 if (refcount_block_offset) {
177 if (refcount_block_offset != s->refcount_block_cache_offset) {
178 ret = load_refcount_block(bs, refcount_block_offset);
179 if (ret < 0) {
180 return ret;
183 return refcount_block_offset;
188 * If we came here, we need to allocate something. Something is at least
189 * a cluster for the new refcount block. It may also include a new refcount
190 * table if the old refcount table is too small.
192 * Note that allocating clusters here needs some special care:
194 * - We can't use the normal qcow2_alloc_clusters(), it would try to
195 * increase the refcount and very likely we would end up with an endless
196 * recursion. Instead we must place the refcount blocks in a way that
197 * they can describe them themselves.
199 * - We need to consider that at this point we are inside update_refcounts
200 * and doing the initial refcount increase. This means that some clusters
201 * have already been allocated by the caller, but their refcount isn't
202 * accurate yet. free_cluster_index tells us where this allocation ends
203 * as long as we don't overwrite it by freeing clusters.
205 * - alloc_clusters_noref and qcow2_free_clusters may load a different
206 * refcount block into the cache
209 if (cache_refcount_updates) {
210 ret = write_refcount_block(s);
211 if (ret < 0) {
212 return ret;
216 /* Allocate the refcount block itself and mark it as used */
217 uint64_t new_block = alloc_clusters_noref(bs, s->cluster_size);
218 memset(s->refcount_block_cache, 0, s->cluster_size);
219 s->refcount_block_cache_offset = new_block;
221 #ifdef DEBUG_ALLOC2
222 fprintf(stderr, "qcow2: Allocate refcount block %d for %" PRIx64
223 " at %" PRIx64 "\n",
224 refcount_table_index, cluster_index << s->cluster_bits, new_block);
225 #endif
227 if (in_same_refcount_block(s, new_block, cluster_index << s->cluster_bits)) {
228 /* The block describes itself, need to update the cache */
229 int block_index = (new_block >> s->cluster_bits) &
230 ((1 << (s->cluster_bits - REFCOUNT_SHIFT)) - 1);
231 s->refcount_block_cache[block_index] = cpu_to_be16(1);
232 } else {
233 /* Described somewhere else. This can recurse at most twice before we
234 * arrive at a block that describes itself. */
235 ret = update_refcount(bs, new_block, s->cluster_size, 1);
236 if (ret < 0) {
237 goto fail_block;
241 /* Now the new refcount block needs to be written to disk */
242 ret = bdrv_pwrite(s->hd, new_block, s->refcount_block_cache,
243 s->cluster_size);
244 if (ret < 0) {
245 goto fail_block;
248 /* If the refcount table is big enough, just hook the block up there */
249 if (refcount_table_index < s->refcount_table_size) {
250 uint64_t data64 = cpu_to_be64(new_block);
251 ret = bdrv_pwrite(s->hd,
252 s->refcount_table_offset + refcount_table_index * sizeof(uint64_t),
253 &data64, sizeof(data64));
254 if (ret < 0) {
255 goto fail_block;
258 s->refcount_table[refcount_table_index] = new_block;
259 return new_block;
263 * If we come here, we need to grow the refcount table. Again, a new
264 * refcount table needs some space and we can't simply allocate to avoid
265 * endless recursion.
267 * Therefore let's grab new refcount blocks at the end of the image, which
268 * will describe themselves and the new refcount table. This way we can
269 * reference them only in the new table and do the switch to the new
270 * refcount table at once without producing an inconsistent state in
271 * between.
273 /* Calculate the number of refcount blocks needed so far */
274 uint64_t refcount_block_clusters = 1 << (s->cluster_bits - REFCOUNT_SHIFT);
275 uint64_t blocks_used = (s->free_cluster_index +
276 refcount_block_clusters - 1) / refcount_block_clusters;
278 /* And now we need at least one block more for the new metadata */
279 uint64_t table_size = next_refcount_table_size(s, blocks_used + 1);
280 uint64_t last_table_size;
281 uint64_t blocks_clusters;
282 do {
283 uint64_t table_clusters = size_to_clusters(s, table_size);
284 blocks_clusters = 1 +
285 ((table_clusters + refcount_block_clusters - 1)
286 / refcount_block_clusters);
287 uint64_t meta_clusters = table_clusters + blocks_clusters;
289 last_table_size = table_size;
290 table_size = next_refcount_table_size(s, blocks_used +
291 ((meta_clusters + refcount_block_clusters - 1)
292 / refcount_block_clusters));
294 } while (last_table_size != table_size);
296 #ifdef DEBUG_ALLOC2
297 fprintf(stderr, "qcow2: Grow refcount table %" PRId32 " => %" PRId64 "\n",
298 s->refcount_table_size, table_size);
299 #endif
301 /* Create the new refcount table and blocks */
302 uint64_t meta_offset = (blocks_used * refcount_block_clusters) *
303 s->cluster_size;
304 uint64_t table_offset = meta_offset + blocks_clusters * s->cluster_size;
305 uint16_t *new_blocks = qemu_mallocz(blocks_clusters * s->cluster_size);
306 uint64_t *new_table = qemu_mallocz(table_size * sizeof(uint64_t));
308 assert(meta_offset >= (s->free_cluster_index * s->cluster_size));
310 /* Fill the new refcount table */
311 memcpy(new_table, s->refcount_table,
312 s->refcount_table_size * sizeof(uint64_t));
313 new_table[refcount_table_index] = new_block;
315 int i;
316 for (i = 0; i < blocks_clusters; i++) {
317 new_table[blocks_used + i] = meta_offset + (i * s->cluster_size);
320 /* Fill the refcount blocks */
321 uint64_t table_clusters = size_to_clusters(s, table_size * sizeof(uint64_t));
322 int block = 0;
323 for (i = 0; i < table_clusters + blocks_clusters; i++) {
324 new_blocks[block++] = cpu_to_be16(1);
327 /* Write refcount blocks to disk */
328 ret = bdrv_pwrite(s->hd, meta_offset, new_blocks,
329 blocks_clusters * s->cluster_size);
330 qemu_free(new_blocks);
331 if (ret < 0) {
332 goto fail_table;
335 /* Write refcount table to disk */
336 for(i = 0; i < table_size; i++) {
337 cpu_to_be64s(&new_table[i]);
340 ret = bdrv_pwrite(s->hd, table_offset, new_table,
341 table_size * sizeof(uint64_t));
342 if (ret < 0) {
343 goto fail_table;
346 for(i = 0; i < table_size; i++) {
347 cpu_to_be64s(&new_table[i]);
350 /* Hook up the new refcount table in the qcow2 header */
351 uint8_t data[12];
352 cpu_to_be64w((uint64_t*)data, table_offset);
353 cpu_to_be32w((uint32_t*)(data + 8), table_clusters);
354 ret = bdrv_pwrite(s->hd, offsetof(QCowHeader, refcount_table_offset),
355 data, sizeof(data));
356 if (ret < 0) {
357 goto fail_table;
360 /* And switch it in memory */
361 uint64_t old_table_offset = s->refcount_table_offset;
362 uint64_t old_table_size = s->refcount_table_size;
364 qemu_free(s->refcount_table);
365 s->refcount_table = new_table;
366 s->refcount_table_size = table_size;
367 s->refcount_table_offset = table_offset;
369 /* Free old table. Remember, we must not change free_cluster_index */
370 uint64_t old_free_cluster_index = s->free_cluster_index;
371 qcow2_free_clusters(bs, old_table_offset, old_table_size * sizeof(uint64_t));
372 s->free_cluster_index = old_free_cluster_index;
374 ret = load_refcount_block(bs, new_block);
375 if (ret < 0) {
376 goto fail_block;
379 return new_block;
381 fail_table:
382 qemu_free(new_table);
383 fail_block:
384 s->refcount_block_cache_offset = 0;
385 return ret;
388 #define REFCOUNTS_PER_SECTOR (512 >> REFCOUNT_SHIFT)
389 static int write_refcount_block_entries(BDRVQcowState *s,
390 int64_t refcount_block_offset, int first_index, int last_index)
392 size_t size;
394 if (cache_refcount_updates) {
395 return 0;
398 first_index &= ~(REFCOUNTS_PER_SECTOR - 1);
399 last_index = (last_index + REFCOUNTS_PER_SECTOR)
400 & ~(REFCOUNTS_PER_SECTOR - 1);
402 size = (last_index - first_index) << REFCOUNT_SHIFT;
403 if (bdrv_pwrite(s->hd,
404 refcount_block_offset + (first_index << REFCOUNT_SHIFT),
405 &s->refcount_block_cache[first_index], size) != size)
407 return -EIO;
410 return 0;
413 /* XXX: cache several refcount block clusters ? */
414 static int QEMU_WARN_UNUSED_RESULT update_refcount(BlockDriverState *bs,
415 int64_t offset, int64_t length, int addend)
417 BDRVQcowState *s = bs->opaque;
418 int64_t start, last, cluster_offset;
419 int64_t refcount_block_offset = 0;
420 int64_t table_index = -1, old_table_index;
421 int first_index = -1, last_index = -1;
422 int ret;
424 #ifdef DEBUG_ALLOC2
425 printf("update_refcount: offset=%" PRId64 " size=%" PRId64 " addend=%d\n",
426 offset, length, addend);
427 #endif
428 if (length < 0) {
429 return -EINVAL;
430 } else if (length == 0) {
431 return 0;
434 start = offset & ~(s->cluster_size - 1);
435 last = (offset + length - 1) & ~(s->cluster_size - 1);
436 for(cluster_offset = start; cluster_offset <= last;
437 cluster_offset += s->cluster_size)
439 int block_index, refcount;
440 int64_t cluster_index = cluster_offset >> s->cluster_bits;
441 int64_t new_block;
443 /* Only write refcount block to disk when we are done with it */
444 old_table_index = table_index;
445 table_index = cluster_index >> (s->cluster_bits - REFCOUNT_SHIFT);
446 if ((old_table_index >= 0) && (table_index != old_table_index)) {
448 if (write_refcount_block_entries(s, refcount_block_offset,
449 first_index, last_index) < 0)
451 return -EIO;
454 first_index = -1;
455 last_index = -1;
458 /* Load the refcount block and allocate it if needed */
459 new_block = alloc_refcount_block(bs, cluster_index);
460 if (new_block < 0) {
461 ret = new_block;
462 goto fail;
464 refcount_block_offset = new_block;
466 /* we can update the count and save it */
467 block_index = cluster_index &
468 ((1 << (s->cluster_bits - REFCOUNT_SHIFT)) - 1);
469 if (first_index == -1 || block_index < first_index) {
470 first_index = block_index;
472 if (block_index > last_index) {
473 last_index = block_index;
476 refcount = be16_to_cpu(s->refcount_block_cache[block_index]);
477 refcount += addend;
478 if (refcount < 0 || refcount > 0xffff) {
479 ret = -EINVAL;
480 goto fail;
482 if (refcount == 0 && cluster_index < s->free_cluster_index) {
483 s->free_cluster_index = cluster_index;
485 s->refcount_block_cache[block_index] = cpu_to_be16(refcount);
488 ret = 0;
489 fail:
491 /* Write last changed block to disk */
492 if (refcount_block_offset != 0) {
493 if (write_refcount_block_entries(s, refcount_block_offset,
494 first_index, last_index) < 0)
496 return ret < 0 ? ret : -EIO;
501 * Try do undo any updates if an error is returned (This may succeed in
502 * some cases like ENOSPC for allocating a new refcount block)
504 if (ret < 0) {
505 int dummy;
506 dummy = update_refcount(bs, offset, cluster_offset - offset, -addend);
509 return ret;
512 /* addend must be 1 or -1 */
513 static int update_cluster_refcount(BlockDriverState *bs,
514 int64_t cluster_index,
515 int addend)
517 BDRVQcowState *s = bs->opaque;
518 int ret;
520 ret = update_refcount(bs, cluster_index << s->cluster_bits, 1, addend);
521 if (ret < 0) {
522 return ret;
525 return get_refcount(bs, cluster_index);
530 /*********************************************************/
531 /* cluster allocation functions */
535 /* return < 0 if error */
536 static int64_t alloc_clusters_noref(BlockDriverState *bs, int64_t size)
538 BDRVQcowState *s = bs->opaque;
539 int i, nb_clusters;
541 nb_clusters = size_to_clusters(s, size);
542 retry:
543 for(i = 0; i < nb_clusters; i++) {
544 int64_t i = s->free_cluster_index++;
545 if (get_refcount(bs, i) != 0)
546 goto retry;
548 #ifdef DEBUG_ALLOC2
549 printf("alloc_clusters: size=%" PRId64 " -> %" PRId64 "\n",
550 size,
551 (s->free_cluster_index - nb_clusters) << s->cluster_bits);
552 #endif
553 return (s->free_cluster_index - nb_clusters) << s->cluster_bits;
556 int64_t qcow2_alloc_clusters(BlockDriverState *bs, int64_t size)
558 int64_t offset;
559 int ret;
561 offset = alloc_clusters_noref(bs, size);
562 ret = update_refcount(bs, offset, size, 1);
563 if (ret < 0) {
564 return ret;
566 return offset;
569 /* only used to allocate compressed sectors. We try to allocate
570 contiguous sectors. size must be <= cluster_size */
571 int64_t qcow2_alloc_bytes(BlockDriverState *bs, int size)
573 BDRVQcowState *s = bs->opaque;
574 int64_t offset, cluster_offset;
575 int free_in_cluster;
577 assert(size > 0 && size <= s->cluster_size);
578 if (s->free_byte_offset == 0) {
579 s->free_byte_offset = qcow2_alloc_clusters(bs, s->cluster_size);
580 if (s->free_byte_offset < 0) {
581 return s->free_byte_offset;
584 redo:
585 free_in_cluster = s->cluster_size -
586 (s->free_byte_offset & (s->cluster_size - 1));
587 if (size <= free_in_cluster) {
588 /* enough space in current cluster */
589 offset = s->free_byte_offset;
590 s->free_byte_offset += size;
591 free_in_cluster -= size;
592 if (free_in_cluster == 0)
593 s->free_byte_offset = 0;
594 if ((offset & (s->cluster_size - 1)) != 0)
595 update_cluster_refcount(bs, offset >> s->cluster_bits, 1);
596 } else {
597 offset = qcow2_alloc_clusters(bs, s->cluster_size);
598 if (offset < 0) {
599 return offset;
601 cluster_offset = s->free_byte_offset & ~(s->cluster_size - 1);
602 if ((cluster_offset + s->cluster_size) == offset) {
603 /* we are lucky: contiguous data */
604 offset = s->free_byte_offset;
605 update_cluster_refcount(bs, offset >> s->cluster_bits, 1);
606 s->free_byte_offset += size;
607 } else {
608 s->free_byte_offset = offset;
609 goto redo;
612 return offset;
615 void qcow2_free_clusters(BlockDriverState *bs,
616 int64_t offset, int64_t size)
618 int ret;
620 ret = update_refcount(bs, offset, size, -1);
621 if (ret < 0) {
622 fprintf(stderr, "qcow2_free_clusters failed: %s\n", strerror(-ret));
623 abort();
628 * free_any_clusters
630 * free clusters according to its type: compressed or not
634 void qcow2_free_any_clusters(BlockDriverState *bs,
635 uint64_t cluster_offset, int nb_clusters)
637 BDRVQcowState *s = bs->opaque;
639 /* free the cluster */
641 if (cluster_offset & QCOW_OFLAG_COMPRESSED) {
642 int nb_csectors;
643 nb_csectors = ((cluster_offset >> s->csize_shift) &
644 s->csize_mask) + 1;
645 qcow2_free_clusters(bs,
646 (cluster_offset & s->cluster_offset_mask) & ~511,
647 nb_csectors * 512);
648 return;
651 qcow2_free_clusters(bs, cluster_offset, nb_clusters << s->cluster_bits);
653 return;
658 /*********************************************************/
659 /* snapshots and image creation */
663 void qcow2_create_refcount_update(QCowCreateState *s, int64_t offset,
664 int64_t size)
666 int refcount;
667 int64_t start, last, cluster_offset;
668 uint16_t *p;
670 start = offset & ~(s->cluster_size - 1);
671 last = (offset + size - 1) & ~(s->cluster_size - 1);
672 for(cluster_offset = start; cluster_offset <= last;
673 cluster_offset += s->cluster_size) {
674 p = &s->refcount_block[cluster_offset >> s->cluster_bits];
675 refcount = be16_to_cpu(*p);
676 refcount++;
677 *p = cpu_to_be16(refcount);
681 /* update the refcounts of snapshots and the copied flag */
682 int qcow2_update_snapshot_refcount(BlockDriverState *bs,
683 int64_t l1_table_offset, int l1_size, int addend)
685 BDRVQcowState *s = bs->opaque;
686 uint64_t *l1_table, *l2_table, l2_offset, offset, l1_size2, l1_allocated;
687 int64_t old_offset, old_l2_offset;
688 int l2_size, i, j, l1_modified, l2_modified, nb_csectors, refcount;
690 qcow2_l2_cache_reset(bs);
691 cache_refcount_updates = 1;
693 l2_table = NULL;
694 l1_table = NULL;
695 l1_size2 = l1_size * sizeof(uint64_t);
696 if (l1_table_offset != s->l1_table_offset) {
697 if (l1_size2 != 0) {
698 l1_table = qemu_mallocz(align_offset(l1_size2, 512));
699 } else {
700 l1_table = NULL;
702 l1_allocated = 1;
703 if (bdrv_pread(s->hd, l1_table_offset,
704 l1_table, l1_size2) != l1_size2)
705 goto fail;
706 for(i = 0;i < l1_size; i++)
707 be64_to_cpus(&l1_table[i]);
708 } else {
709 assert(l1_size == s->l1_size);
710 l1_table = s->l1_table;
711 l1_allocated = 0;
714 l2_size = s->l2_size * sizeof(uint64_t);
715 l2_table = qemu_malloc(l2_size);
716 l1_modified = 0;
717 for(i = 0; i < l1_size; i++) {
718 l2_offset = l1_table[i];
719 if (l2_offset) {
720 old_l2_offset = l2_offset;
721 l2_offset &= ~QCOW_OFLAG_COPIED;
722 l2_modified = 0;
723 if (bdrv_pread(s->hd, l2_offset, l2_table, l2_size) != l2_size)
724 goto fail;
725 for(j = 0; j < s->l2_size; j++) {
726 offset = be64_to_cpu(l2_table[j]);
727 if (offset != 0) {
728 old_offset = offset;
729 offset &= ~QCOW_OFLAG_COPIED;
730 if (offset & QCOW_OFLAG_COMPRESSED) {
731 nb_csectors = ((offset >> s->csize_shift) &
732 s->csize_mask) + 1;
733 if (addend != 0) {
734 int ret;
735 ret = update_refcount(bs,
736 (offset & s->cluster_offset_mask) & ~511,
737 nb_csectors * 512, addend);
738 if (ret < 0) {
739 goto fail;
742 /* compressed clusters are never modified */
743 refcount = 2;
744 } else {
745 if (addend != 0) {
746 refcount = update_cluster_refcount(bs, offset >> s->cluster_bits, addend);
747 } else {
748 refcount = get_refcount(bs, offset >> s->cluster_bits);
752 if (refcount == 1) {
753 offset |= QCOW_OFLAG_COPIED;
755 if (offset != old_offset) {
756 l2_table[j] = cpu_to_be64(offset);
757 l2_modified = 1;
761 if (l2_modified) {
762 if (bdrv_pwrite(s->hd,
763 l2_offset, l2_table, l2_size) != l2_size)
764 goto fail;
767 if (addend != 0) {
768 refcount = update_cluster_refcount(bs, l2_offset >> s->cluster_bits, addend);
769 } else {
770 refcount = get_refcount(bs, l2_offset >> s->cluster_bits);
772 if (refcount == 1) {
773 l2_offset |= QCOW_OFLAG_COPIED;
775 if (l2_offset != old_l2_offset) {
776 l1_table[i] = l2_offset;
777 l1_modified = 1;
781 if (l1_modified) {
782 for(i = 0; i < l1_size; i++)
783 cpu_to_be64s(&l1_table[i]);
784 if (bdrv_pwrite(s->hd, l1_table_offset, l1_table,
785 l1_size2) != l1_size2)
786 goto fail;
787 for(i = 0; i < l1_size; i++)
788 be64_to_cpus(&l1_table[i]);
790 if (l1_allocated)
791 qemu_free(l1_table);
792 qemu_free(l2_table);
793 cache_refcount_updates = 0;
794 write_refcount_block(s);
795 return 0;
796 fail:
797 if (l1_allocated)
798 qemu_free(l1_table);
799 qemu_free(l2_table);
800 cache_refcount_updates = 0;
801 write_refcount_block(s);
802 return -EIO;
808 /*********************************************************/
809 /* refcount checking functions */
814 * Increases the refcount for a range of clusters in a given refcount table.
815 * This is used to construct a temporary refcount table out of L1 and L2 tables
816 * which can be compared the the refcount table saved in the image.
818 * Returns the number of errors in the image that were found
820 static int inc_refcounts(BlockDriverState *bs,
821 uint16_t *refcount_table,
822 int refcount_table_size,
823 int64_t offset, int64_t size)
825 BDRVQcowState *s = bs->opaque;
826 int64_t start, last, cluster_offset;
827 int k;
828 int errors = 0;
830 if (size <= 0)
831 return 0;
833 start = offset & ~(s->cluster_size - 1);
834 last = (offset + size - 1) & ~(s->cluster_size - 1);
835 for(cluster_offset = start; cluster_offset <= last;
836 cluster_offset += s->cluster_size) {
837 k = cluster_offset >> s->cluster_bits;
838 if (k < 0 || k >= refcount_table_size) {
839 fprintf(stderr, "ERROR: invalid cluster offset=0x%" PRIx64 "\n",
840 cluster_offset);
841 errors++;
842 } else {
843 if (++refcount_table[k] == 0) {
844 fprintf(stderr, "ERROR: overflow cluster offset=0x%" PRIx64
845 "\n", cluster_offset);
846 errors++;
851 return errors;
855 * Increases the refcount in the given refcount table for the all clusters
856 * referenced in the L2 table. While doing so, performs some checks on L2
857 * entries.
859 * Returns the number of errors found by the checks or -errno if an internal
860 * error occurred.
862 static int check_refcounts_l2(BlockDriverState *bs,
863 uint16_t *refcount_table, int refcount_table_size, int64_t l2_offset,
864 int check_copied)
866 BDRVQcowState *s = bs->opaque;
867 uint64_t *l2_table, offset;
868 int i, l2_size, nb_csectors, refcount;
869 int errors = 0;
871 /* Read L2 table from disk */
872 l2_size = s->l2_size * sizeof(uint64_t);
873 l2_table = qemu_malloc(l2_size);
875 if (bdrv_pread(s->hd, l2_offset, l2_table, l2_size) != l2_size)
876 goto fail;
878 /* Do the actual checks */
879 for(i = 0; i < s->l2_size; i++) {
880 offset = be64_to_cpu(l2_table[i]);
881 if (offset != 0) {
882 if (offset & QCOW_OFLAG_COMPRESSED) {
883 /* Compressed clusters don't have QCOW_OFLAG_COPIED */
884 if (offset & QCOW_OFLAG_COPIED) {
885 fprintf(stderr, "ERROR: cluster %" PRId64 ": "
886 "copied flag must never be set for compressed "
887 "clusters\n", offset >> s->cluster_bits);
888 offset &= ~QCOW_OFLAG_COPIED;
889 errors++;
892 /* Mark cluster as used */
893 nb_csectors = ((offset >> s->csize_shift) &
894 s->csize_mask) + 1;
895 offset &= s->cluster_offset_mask;
896 errors += inc_refcounts(bs, refcount_table,
897 refcount_table_size,
898 offset & ~511, nb_csectors * 512);
899 } else {
900 /* QCOW_OFLAG_COPIED must be set iff refcount == 1 */
901 if (check_copied) {
902 uint64_t entry = offset;
903 offset &= ~QCOW_OFLAG_COPIED;
904 refcount = get_refcount(bs, offset >> s->cluster_bits);
905 if ((refcount == 1) != ((entry & QCOW_OFLAG_COPIED) != 0)) {
906 fprintf(stderr, "ERROR OFLAG_COPIED: offset=%"
907 PRIx64 " refcount=%d\n", entry, refcount);
908 errors++;
912 /* Mark cluster as used */
913 offset &= ~QCOW_OFLAG_COPIED;
914 errors += inc_refcounts(bs, refcount_table,
915 refcount_table_size,
916 offset, s->cluster_size);
918 /* Correct offsets are cluster aligned */
919 if (offset & (s->cluster_size - 1)) {
920 fprintf(stderr, "ERROR offset=%" PRIx64 ": Cluster is not "
921 "properly aligned; L2 entry corrupted.\n", offset);
922 errors++;
928 qemu_free(l2_table);
929 return errors;
931 fail:
932 fprintf(stderr, "ERROR: I/O error in check_refcounts_l1\n");
933 qemu_free(l2_table);
934 return -EIO;
938 * Increases the refcount for the L1 table, its L2 tables and all referenced
939 * clusters in the given refcount table. While doing so, performs some checks
940 * on L1 and L2 entries.
942 * Returns the number of errors found by the checks or -errno if an internal
943 * error occurred.
945 static int check_refcounts_l1(BlockDriverState *bs,
946 uint16_t *refcount_table,
947 int refcount_table_size,
948 int64_t l1_table_offset, int l1_size,
949 int check_copied)
951 BDRVQcowState *s = bs->opaque;
952 uint64_t *l1_table, l2_offset, l1_size2;
953 int i, refcount, ret;
954 int errors = 0;
956 l1_size2 = l1_size * sizeof(uint64_t);
958 /* Mark L1 table as used */
959 errors += inc_refcounts(bs, refcount_table, refcount_table_size,
960 l1_table_offset, l1_size2);
962 /* Read L1 table entries from disk */
963 if (l1_size2 == 0) {
964 l1_table = NULL;
965 } else {
966 l1_table = qemu_malloc(l1_size2);
967 if (bdrv_pread(s->hd, l1_table_offset,
968 l1_table, l1_size2) != l1_size2)
969 goto fail;
970 for(i = 0;i < l1_size; i++)
971 be64_to_cpus(&l1_table[i]);
974 /* Do the actual checks */
975 for(i = 0; i < l1_size; i++) {
976 l2_offset = l1_table[i];
977 if (l2_offset) {
978 /* QCOW_OFLAG_COPIED must be set iff refcount == 1 */
979 if (check_copied) {
980 refcount = get_refcount(bs, (l2_offset & ~QCOW_OFLAG_COPIED)
981 >> s->cluster_bits);
982 if ((refcount == 1) != ((l2_offset & QCOW_OFLAG_COPIED) != 0)) {
983 fprintf(stderr, "ERROR OFLAG_COPIED: l2_offset=%" PRIx64
984 " refcount=%d\n", l2_offset, refcount);
985 errors++;
989 /* Mark L2 table as used */
990 l2_offset &= ~QCOW_OFLAG_COPIED;
991 errors += inc_refcounts(bs, refcount_table,
992 refcount_table_size,
993 l2_offset,
994 s->cluster_size);
996 /* L2 tables are cluster aligned */
997 if (l2_offset & (s->cluster_size - 1)) {
998 fprintf(stderr, "ERROR l2_offset=%" PRIx64 ": Table is not "
999 "cluster aligned; L1 entry corrupted\n", l2_offset);
1000 errors++;
1003 /* Process and check L2 entries */
1004 ret = check_refcounts_l2(bs, refcount_table, refcount_table_size,
1005 l2_offset, check_copied);
1006 if (ret < 0) {
1007 goto fail;
1009 errors += ret;
1012 qemu_free(l1_table);
1013 return errors;
1015 fail:
1016 fprintf(stderr, "ERROR: I/O error in check_refcounts_l1\n");
1017 qemu_free(l1_table);
1018 return -EIO;
1022 * Checks an image for refcount consistency.
1024 * Returns 0 if no errors are found, the number of errors in case the image is
1025 * detected as corrupted, and -errno when an internal error occured.
1027 int qcow2_check_refcounts(BlockDriverState *bs)
1029 BDRVQcowState *s = bs->opaque;
1030 int64_t size;
1031 int nb_clusters, refcount1, refcount2, i;
1032 QCowSnapshot *sn;
1033 uint16_t *refcount_table;
1034 int ret, errors = 0;
1036 size = bdrv_getlength(s->hd);
1037 nb_clusters = size_to_clusters(s, size);
1038 refcount_table = qemu_mallocz(nb_clusters * sizeof(uint16_t));
1040 /* header */
1041 errors += inc_refcounts(bs, refcount_table, nb_clusters,
1042 0, s->cluster_size);
1044 /* current L1 table */
1045 ret = check_refcounts_l1(bs, refcount_table, nb_clusters,
1046 s->l1_table_offset, s->l1_size, 1);
1047 if (ret < 0) {
1048 return ret;
1050 errors += ret;
1052 /* snapshots */
1053 for(i = 0; i < s->nb_snapshots; i++) {
1054 sn = s->snapshots + i;
1055 check_refcounts_l1(bs, refcount_table, nb_clusters,
1056 sn->l1_table_offset, sn->l1_size, 0);
1058 errors += inc_refcounts(bs, refcount_table, nb_clusters,
1059 s->snapshots_offset, s->snapshots_size);
1061 /* refcount data */
1062 errors += inc_refcounts(bs, refcount_table, nb_clusters,
1063 s->refcount_table_offset,
1064 s->refcount_table_size * sizeof(uint64_t));
1065 for(i = 0; i < s->refcount_table_size; i++) {
1066 int64_t offset;
1067 offset = s->refcount_table[i];
1069 /* Refcount blocks are cluster aligned */
1070 if (offset & (s->cluster_size - 1)) {
1071 fprintf(stderr, "ERROR refcount block %d is not "
1072 "cluster aligned; refcount table entry corrupted\n", i);
1073 errors++;
1076 if (offset != 0) {
1077 errors += inc_refcounts(bs, refcount_table, nb_clusters,
1078 offset, s->cluster_size);
1079 if (refcount_table[offset / s->cluster_size] != 1) {
1080 fprintf(stderr, "ERROR refcount block %d refcount=%d\n",
1081 i, refcount_table[offset / s->cluster_size]);
1086 /* compare ref counts */
1087 for(i = 0; i < nb_clusters; i++) {
1088 refcount1 = get_refcount(bs, i);
1089 refcount2 = refcount_table[i];
1090 if (refcount1 != refcount2) {
1091 fprintf(stderr, "ERROR cluster %d refcount=%d reference=%d\n",
1092 i, refcount1, refcount2);
1093 errors++;
1097 qemu_free(refcount_table);
1099 return errors;