hardfloat: implement float32/64 addition and subtraction
[qemu/ericb.git] / block / qcow2-refcount.c
blob1c63ac244ac3aafd04b02f6f84af0c9ee91e537b
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/osdep.h"
26 #include "qapi/error.h"
27 #include "qemu-common.h"
28 #include "block/block_int.h"
29 #include "qcow2.h"
30 #include "qemu/range.h"
31 #include "qemu/bswap.h"
32 #include "qemu/cutils.h"
34 static int64_t alloc_clusters_noref(BlockDriverState *bs, uint64_t size,
35 uint64_t max);
36 static int QEMU_WARN_UNUSED_RESULT update_refcount(BlockDriverState *bs,
37 int64_t offset, int64_t length, uint64_t addend,
38 bool decrease, enum qcow2_discard_type type);
40 static uint64_t get_refcount_ro0(const void *refcount_array, uint64_t index);
41 static uint64_t get_refcount_ro1(const void *refcount_array, uint64_t index);
42 static uint64_t get_refcount_ro2(const void *refcount_array, uint64_t index);
43 static uint64_t get_refcount_ro3(const void *refcount_array, uint64_t index);
44 static uint64_t get_refcount_ro4(const void *refcount_array, uint64_t index);
45 static uint64_t get_refcount_ro5(const void *refcount_array, uint64_t index);
46 static uint64_t get_refcount_ro6(const void *refcount_array, uint64_t index);
48 static void set_refcount_ro0(void *refcount_array, uint64_t index,
49 uint64_t value);
50 static void set_refcount_ro1(void *refcount_array, uint64_t index,
51 uint64_t value);
52 static void set_refcount_ro2(void *refcount_array, uint64_t index,
53 uint64_t value);
54 static void set_refcount_ro3(void *refcount_array, uint64_t index,
55 uint64_t value);
56 static void set_refcount_ro4(void *refcount_array, uint64_t index,
57 uint64_t value);
58 static void set_refcount_ro5(void *refcount_array, uint64_t index,
59 uint64_t value);
60 static void set_refcount_ro6(void *refcount_array, uint64_t index,
61 uint64_t value);
64 static Qcow2GetRefcountFunc *const get_refcount_funcs[] = {
65 &get_refcount_ro0,
66 &get_refcount_ro1,
67 &get_refcount_ro2,
68 &get_refcount_ro3,
69 &get_refcount_ro4,
70 &get_refcount_ro5,
71 &get_refcount_ro6
74 static Qcow2SetRefcountFunc *const set_refcount_funcs[] = {
75 &set_refcount_ro0,
76 &set_refcount_ro1,
77 &set_refcount_ro2,
78 &set_refcount_ro3,
79 &set_refcount_ro4,
80 &set_refcount_ro5,
81 &set_refcount_ro6
85 /*********************************************************/
86 /* refcount handling */
88 static void update_max_refcount_table_index(BDRVQcow2State *s)
90 unsigned i = s->refcount_table_size - 1;
91 while (i > 0 && (s->refcount_table[i] & REFT_OFFSET_MASK) == 0) {
92 i--;
94 /* Set s->max_refcount_table_index to the index of the last used entry */
95 s->max_refcount_table_index = i;
98 int qcow2_refcount_init(BlockDriverState *bs)
100 BDRVQcow2State *s = bs->opaque;
101 unsigned int refcount_table_size2, i;
102 int ret;
104 assert(s->refcount_order >= 0 && s->refcount_order <= 6);
106 s->get_refcount = get_refcount_funcs[s->refcount_order];
107 s->set_refcount = set_refcount_funcs[s->refcount_order];
109 assert(s->refcount_table_size <= INT_MAX / sizeof(uint64_t));
110 refcount_table_size2 = s->refcount_table_size * sizeof(uint64_t);
111 s->refcount_table = g_try_malloc(refcount_table_size2);
113 if (s->refcount_table_size > 0) {
114 if (s->refcount_table == NULL) {
115 ret = -ENOMEM;
116 goto fail;
118 BLKDBG_EVENT(bs->file, BLKDBG_REFTABLE_LOAD);
119 ret = bdrv_pread(bs->file, s->refcount_table_offset,
120 s->refcount_table, refcount_table_size2);
121 if (ret < 0) {
122 goto fail;
124 for(i = 0; i < s->refcount_table_size; i++)
125 be64_to_cpus(&s->refcount_table[i]);
126 update_max_refcount_table_index(s);
128 return 0;
129 fail:
130 return ret;
133 void qcow2_refcount_close(BlockDriverState *bs)
135 BDRVQcow2State *s = bs->opaque;
136 g_free(s->refcount_table);
140 static uint64_t get_refcount_ro0(const void *refcount_array, uint64_t index)
142 return (((const uint8_t *)refcount_array)[index / 8] >> (index % 8)) & 0x1;
145 static void set_refcount_ro0(void *refcount_array, uint64_t index,
146 uint64_t value)
148 assert(!(value >> 1));
149 ((uint8_t *)refcount_array)[index / 8] &= ~(0x1 << (index % 8));
150 ((uint8_t *)refcount_array)[index / 8] |= value << (index % 8);
153 static uint64_t get_refcount_ro1(const void *refcount_array, uint64_t index)
155 return (((const uint8_t *)refcount_array)[index / 4] >> (2 * (index % 4)))
156 & 0x3;
159 static void set_refcount_ro1(void *refcount_array, uint64_t index,
160 uint64_t value)
162 assert(!(value >> 2));
163 ((uint8_t *)refcount_array)[index / 4] &= ~(0x3 << (2 * (index % 4)));
164 ((uint8_t *)refcount_array)[index / 4] |= value << (2 * (index % 4));
167 static uint64_t get_refcount_ro2(const void *refcount_array, uint64_t index)
169 return (((const uint8_t *)refcount_array)[index / 2] >> (4 * (index % 2)))
170 & 0xf;
173 static void set_refcount_ro2(void *refcount_array, uint64_t index,
174 uint64_t value)
176 assert(!(value >> 4));
177 ((uint8_t *)refcount_array)[index / 2] &= ~(0xf << (4 * (index % 2)));
178 ((uint8_t *)refcount_array)[index / 2] |= value << (4 * (index % 2));
181 static uint64_t get_refcount_ro3(const void *refcount_array, uint64_t index)
183 return ((const uint8_t *)refcount_array)[index];
186 static void set_refcount_ro3(void *refcount_array, uint64_t index,
187 uint64_t value)
189 assert(!(value >> 8));
190 ((uint8_t *)refcount_array)[index] = value;
193 static uint64_t get_refcount_ro4(const void *refcount_array, uint64_t index)
195 return be16_to_cpu(((const uint16_t *)refcount_array)[index]);
198 static void set_refcount_ro4(void *refcount_array, uint64_t index,
199 uint64_t value)
201 assert(!(value >> 16));
202 ((uint16_t *)refcount_array)[index] = cpu_to_be16(value);
205 static uint64_t get_refcount_ro5(const void *refcount_array, uint64_t index)
207 return be32_to_cpu(((const uint32_t *)refcount_array)[index]);
210 static void set_refcount_ro5(void *refcount_array, uint64_t index,
211 uint64_t value)
213 assert(!(value >> 32));
214 ((uint32_t *)refcount_array)[index] = cpu_to_be32(value);
217 static uint64_t get_refcount_ro6(const void *refcount_array, uint64_t index)
219 return be64_to_cpu(((const uint64_t *)refcount_array)[index]);
222 static void set_refcount_ro6(void *refcount_array, uint64_t index,
223 uint64_t value)
225 ((uint64_t *)refcount_array)[index] = cpu_to_be64(value);
229 static int load_refcount_block(BlockDriverState *bs,
230 int64_t refcount_block_offset,
231 void **refcount_block)
233 BDRVQcow2State *s = bs->opaque;
235 BLKDBG_EVENT(bs->file, BLKDBG_REFBLOCK_LOAD);
236 return qcow2_cache_get(bs, s->refcount_block_cache, refcount_block_offset,
237 refcount_block);
241 * Retrieves the refcount of the cluster given by its index and stores it in
242 * *refcount. Returns 0 on success and -errno on failure.
244 int qcow2_get_refcount(BlockDriverState *bs, int64_t cluster_index,
245 uint64_t *refcount)
247 BDRVQcow2State *s = bs->opaque;
248 uint64_t refcount_table_index, block_index;
249 int64_t refcount_block_offset;
250 int ret;
251 void *refcount_block;
253 refcount_table_index = cluster_index >> s->refcount_block_bits;
254 if (refcount_table_index >= s->refcount_table_size) {
255 *refcount = 0;
256 return 0;
258 refcount_block_offset =
259 s->refcount_table[refcount_table_index] & REFT_OFFSET_MASK;
260 if (!refcount_block_offset) {
261 *refcount = 0;
262 return 0;
265 if (offset_into_cluster(s, refcount_block_offset)) {
266 qcow2_signal_corruption(bs, true, -1, -1, "Refblock offset %#" PRIx64
267 " unaligned (reftable index: %#" PRIx64 ")",
268 refcount_block_offset, refcount_table_index);
269 return -EIO;
272 ret = qcow2_cache_get(bs, s->refcount_block_cache, refcount_block_offset,
273 &refcount_block);
274 if (ret < 0) {
275 return ret;
278 block_index = cluster_index & (s->refcount_block_size - 1);
279 *refcount = s->get_refcount(refcount_block, block_index);
281 qcow2_cache_put(s->refcount_block_cache, &refcount_block);
283 return 0;
286 /* Checks if two offsets are described by the same refcount block */
287 static int in_same_refcount_block(BDRVQcow2State *s, uint64_t offset_a,
288 uint64_t offset_b)
290 uint64_t block_a = offset_a >> (s->cluster_bits + s->refcount_block_bits);
291 uint64_t block_b = offset_b >> (s->cluster_bits + s->refcount_block_bits);
293 return (block_a == block_b);
297 * Loads a refcount block. If it doesn't exist yet, it is allocated first
298 * (including growing the refcount table if needed).
300 * Returns 0 on success or -errno in error case
302 static int alloc_refcount_block(BlockDriverState *bs,
303 int64_t cluster_index, void **refcount_block)
305 BDRVQcow2State *s = bs->opaque;
306 unsigned int refcount_table_index;
307 int64_t ret;
309 BLKDBG_EVENT(bs->file, BLKDBG_REFBLOCK_ALLOC);
311 /* Find the refcount block for the given cluster */
312 refcount_table_index = cluster_index >> s->refcount_block_bits;
314 if (refcount_table_index < s->refcount_table_size) {
316 uint64_t refcount_block_offset =
317 s->refcount_table[refcount_table_index] & REFT_OFFSET_MASK;
319 /* If it's already there, we're done */
320 if (refcount_block_offset) {
321 if (offset_into_cluster(s, refcount_block_offset)) {
322 qcow2_signal_corruption(bs, true, -1, -1, "Refblock offset %#"
323 PRIx64 " unaligned (reftable index: "
324 "%#x)", refcount_block_offset,
325 refcount_table_index);
326 return -EIO;
329 return load_refcount_block(bs, refcount_block_offset,
330 refcount_block);
335 * If we came here, we need to allocate something. Something is at least
336 * a cluster for the new refcount block. It may also include a new refcount
337 * table if the old refcount table is too small.
339 * Note that allocating clusters here needs some special care:
341 * - We can't use the normal qcow2_alloc_clusters(), it would try to
342 * increase the refcount and very likely we would end up with an endless
343 * recursion. Instead we must place the refcount blocks in a way that
344 * they can describe them themselves.
346 * - We need to consider that at this point we are inside update_refcounts
347 * and potentially doing an initial refcount increase. This means that
348 * some clusters have already been allocated by the caller, but their
349 * refcount isn't accurate yet. If we allocate clusters for metadata, we
350 * need to return -EAGAIN to signal the caller that it needs to restart
351 * the search for free clusters.
353 * - alloc_clusters_noref and qcow2_free_clusters may load a different
354 * refcount block into the cache
357 *refcount_block = NULL;
359 /* We write to the refcount table, so we might depend on L2 tables */
360 ret = qcow2_cache_flush(bs, s->l2_table_cache);
361 if (ret < 0) {
362 return ret;
365 /* Allocate the refcount block itself and mark it as used */
366 int64_t new_block = alloc_clusters_noref(bs, s->cluster_size, INT64_MAX);
367 if (new_block < 0) {
368 return new_block;
371 /* If we're allocating the block at offset 0 then something is wrong */
372 if (new_block == 0) {
373 qcow2_signal_corruption(bs, true, -1, -1, "Preventing invalid "
374 "allocation of refcount block at offset 0");
375 return -EIO;
378 #ifdef DEBUG_ALLOC2
379 fprintf(stderr, "qcow2: Allocate refcount block %d for %" PRIx64
380 " at %" PRIx64 "\n",
381 refcount_table_index, cluster_index << s->cluster_bits, new_block);
382 #endif
384 if (in_same_refcount_block(s, new_block, cluster_index << s->cluster_bits)) {
385 /* Zero the new refcount block before updating it */
386 ret = qcow2_cache_get_empty(bs, s->refcount_block_cache, new_block,
387 refcount_block);
388 if (ret < 0) {
389 goto fail;
392 memset(*refcount_block, 0, s->cluster_size);
394 /* The block describes itself, need to update the cache */
395 int block_index = (new_block >> s->cluster_bits) &
396 (s->refcount_block_size - 1);
397 s->set_refcount(*refcount_block, block_index, 1);
398 } else {
399 /* Described somewhere else. This can recurse at most twice before we
400 * arrive at a block that describes itself. */
401 ret = update_refcount(bs, new_block, s->cluster_size, 1, false,
402 QCOW2_DISCARD_NEVER);
403 if (ret < 0) {
404 goto fail;
407 ret = qcow2_cache_flush(bs, s->refcount_block_cache);
408 if (ret < 0) {
409 goto fail;
412 /* Initialize the new refcount block only after updating its refcount,
413 * update_refcount uses the refcount cache itself */
414 ret = qcow2_cache_get_empty(bs, s->refcount_block_cache, new_block,
415 refcount_block);
416 if (ret < 0) {
417 goto fail;
420 memset(*refcount_block, 0, s->cluster_size);
423 /* Now the new refcount block needs to be written to disk */
424 BLKDBG_EVENT(bs->file, BLKDBG_REFBLOCK_ALLOC_WRITE);
425 qcow2_cache_entry_mark_dirty(s->refcount_block_cache, *refcount_block);
426 ret = qcow2_cache_flush(bs, s->refcount_block_cache);
427 if (ret < 0) {
428 goto fail;
431 /* If the refcount table is big enough, just hook the block up there */
432 if (refcount_table_index < s->refcount_table_size) {
433 uint64_t data64 = cpu_to_be64(new_block);
434 BLKDBG_EVENT(bs->file, BLKDBG_REFBLOCK_ALLOC_HOOKUP);
435 ret = bdrv_pwrite_sync(bs->file,
436 s->refcount_table_offset + refcount_table_index * sizeof(uint64_t),
437 &data64, sizeof(data64));
438 if (ret < 0) {
439 goto fail;
442 s->refcount_table[refcount_table_index] = new_block;
443 /* If there's a hole in s->refcount_table then it can happen
444 * that refcount_table_index < s->max_refcount_table_index */
445 s->max_refcount_table_index =
446 MAX(s->max_refcount_table_index, refcount_table_index);
448 /* The new refcount block may be where the caller intended to put its
449 * data, so let it restart the search. */
450 return -EAGAIN;
453 qcow2_cache_put(s->refcount_block_cache, refcount_block);
456 * If we come here, we need to grow the refcount table. Again, a new
457 * refcount table needs some space and we can't simply allocate to avoid
458 * endless recursion.
460 * Therefore let's grab new refcount blocks at the end of the image, which
461 * will describe themselves and the new refcount table. This way we can
462 * reference them only in the new table and do the switch to the new
463 * refcount table at once without producing an inconsistent state in
464 * between.
466 BLKDBG_EVENT(bs->file, BLKDBG_REFTABLE_GROW);
468 /* Calculate the number of refcount blocks needed so far; this will be the
469 * basis for calculating the index of the first cluster used for the
470 * self-describing refcount structures which we are about to create.
472 * Because we reached this point, there cannot be any refcount entries for
473 * cluster_index or higher indices yet. However, because new_block has been
474 * allocated to describe that cluster (and it will assume this role later
475 * on), we cannot use that index; also, new_block may actually have a higher
476 * cluster index than cluster_index, so it needs to be taken into account
477 * here (and 1 needs to be added to its value because that cluster is used).
479 uint64_t blocks_used = DIV_ROUND_UP(MAX(cluster_index + 1,
480 (new_block >> s->cluster_bits) + 1),
481 s->refcount_block_size);
483 /* Create the new refcount table and blocks */
484 uint64_t meta_offset = (blocks_used * s->refcount_block_size) *
485 s->cluster_size;
487 ret = qcow2_refcount_area(bs, meta_offset, 0, false,
488 refcount_table_index, new_block);
489 if (ret < 0) {
490 return ret;
493 ret = load_refcount_block(bs, new_block, refcount_block);
494 if (ret < 0) {
495 return ret;
498 /* If we were trying to do the initial refcount update for some cluster
499 * allocation, we might have used the same clusters to store newly
500 * allocated metadata. Make the caller search some new space. */
501 return -EAGAIN;
503 fail:
504 if (*refcount_block != NULL) {
505 qcow2_cache_put(s->refcount_block_cache, refcount_block);
507 return ret;
511 * Starting at @start_offset, this function creates new self-covering refcount
512 * structures: A new refcount table and refcount blocks which cover all of
513 * themselves, and a number of @additional_clusters beyond their end.
514 * @start_offset must be at the end of the image file, that is, there must be
515 * only empty space beyond it.
516 * If @exact_size is false, the refcount table will have 50 % more entries than
517 * necessary so it will not need to grow again soon.
518 * If @new_refblock_offset is not zero, it contains the offset of a refcount
519 * block that should be entered into the new refcount table at index
520 * @new_refblock_index.
522 * Returns: The offset after the new refcount structures (i.e. where the
523 * @additional_clusters may be placed) on success, -errno on error.
525 int64_t qcow2_refcount_area(BlockDriverState *bs, uint64_t start_offset,
526 uint64_t additional_clusters, bool exact_size,
527 int new_refblock_index,
528 uint64_t new_refblock_offset)
530 BDRVQcow2State *s = bs->opaque;
531 uint64_t total_refblock_count_u64, additional_refblock_count;
532 int total_refblock_count, table_size, area_reftable_index, table_clusters;
533 int i;
534 uint64_t table_offset, block_offset, end_offset;
535 int ret;
536 uint64_t *new_table;
538 assert(!(start_offset % s->cluster_size));
540 qcow2_refcount_metadata_size(start_offset / s->cluster_size +
541 additional_clusters,
542 s->cluster_size, s->refcount_order,
543 !exact_size, &total_refblock_count_u64);
544 if (total_refblock_count_u64 > QCOW_MAX_REFTABLE_SIZE) {
545 return -EFBIG;
547 total_refblock_count = total_refblock_count_u64;
549 /* Index in the refcount table of the first refcount block to cover the area
550 * of refcount structures we are about to create; we know that
551 * @total_refblock_count can cover @start_offset, so this will definitely
552 * fit into an int. */
553 area_reftable_index = (start_offset / s->cluster_size) /
554 s->refcount_block_size;
556 if (exact_size) {
557 table_size = total_refblock_count;
558 } else {
559 table_size = total_refblock_count +
560 DIV_ROUND_UP(total_refblock_count, 2);
562 /* The qcow2 file can only store the reftable size in number of clusters */
563 table_size = ROUND_UP(table_size, s->cluster_size / sizeof(uint64_t));
564 table_clusters = (table_size * sizeof(uint64_t)) / s->cluster_size;
566 if (table_size > QCOW_MAX_REFTABLE_SIZE) {
567 return -EFBIG;
570 new_table = g_try_new0(uint64_t, table_size);
572 assert(table_size > 0);
573 if (new_table == NULL) {
574 ret = -ENOMEM;
575 goto fail;
578 /* Fill the new refcount table */
579 if (table_size > s->max_refcount_table_index) {
580 /* We're actually growing the reftable */
581 memcpy(new_table, s->refcount_table,
582 (s->max_refcount_table_index + 1) * sizeof(uint64_t));
583 } else {
584 /* Improbable case: We're shrinking the reftable. However, the caller
585 * has assured us that there is only empty space beyond @start_offset,
586 * so we can simply drop all of the refblocks that won't fit into the
587 * new reftable. */
588 memcpy(new_table, s->refcount_table, table_size * sizeof(uint64_t));
591 if (new_refblock_offset) {
592 assert(new_refblock_index < total_refblock_count);
593 new_table[new_refblock_index] = new_refblock_offset;
596 /* Count how many new refblocks we have to create */
597 additional_refblock_count = 0;
598 for (i = area_reftable_index; i < total_refblock_count; i++) {
599 if (!new_table[i]) {
600 additional_refblock_count++;
604 table_offset = start_offset + additional_refblock_count * s->cluster_size;
605 end_offset = table_offset + table_clusters * s->cluster_size;
607 /* Fill the refcount blocks, and create new ones, if necessary */
608 block_offset = start_offset;
609 for (i = area_reftable_index; i < total_refblock_count; i++) {
610 void *refblock_data;
611 uint64_t first_offset_covered;
613 /* Reuse an existing refblock if possible, create a new one otherwise */
614 if (new_table[i]) {
615 ret = qcow2_cache_get(bs, s->refcount_block_cache, new_table[i],
616 &refblock_data);
617 if (ret < 0) {
618 goto fail;
620 } else {
621 ret = qcow2_cache_get_empty(bs, s->refcount_block_cache,
622 block_offset, &refblock_data);
623 if (ret < 0) {
624 goto fail;
626 memset(refblock_data, 0, s->cluster_size);
627 qcow2_cache_entry_mark_dirty(s->refcount_block_cache,
628 refblock_data);
630 new_table[i] = block_offset;
631 block_offset += s->cluster_size;
634 /* First host offset covered by this refblock */
635 first_offset_covered = (uint64_t)i * s->refcount_block_size *
636 s->cluster_size;
637 if (first_offset_covered < end_offset) {
638 int j, end_index;
640 /* Set the refcount of all of the new refcount structures to 1 */
642 if (first_offset_covered < start_offset) {
643 assert(i == area_reftable_index);
644 j = (start_offset - first_offset_covered) / s->cluster_size;
645 assert(j < s->refcount_block_size);
646 } else {
647 j = 0;
650 end_index = MIN((end_offset - first_offset_covered) /
651 s->cluster_size,
652 s->refcount_block_size);
654 for (; j < end_index; j++) {
655 /* The caller guaranteed us this space would be empty */
656 assert(s->get_refcount(refblock_data, j) == 0);
657 s->set_refcount(refblock_data, j, 1);
660 qcow2_cache_entry_mark_dirty(s->refcount_block_cache,
661 refblock_data);
664 qcow2_cache_put(s->refcount_block_cache, &refblock_data);
667 assert(block_offset == table_offset);
669 /* Write refcount blocks to disk */
670 BLKDBG_EVENT(bs->file, BLKDBG_REFBLOCK_ALLOC_WRITE_BLOCKS);
671 ret = qcow2_cache_flush(bs, s->refcount_block_cache);
672 if (ret < 0) {
673 goto fail;
676 /* Write refcount table to disk */
677 for (i = 0; i < total_refblock_count; i++) {
678 cpu_to_be64s(&new_table[i]);
681 BLKDBG_EVENT(bs->file, BLKDBG_REFBLOCK_ALLOC_WRITE_TABLE);
682 ret = bdrv_pwrite_sync(bs->file, table_offset, new_table,
683 table_size * sizeof(uint64_t));
684 if (ret < 0) {
685 goto fail;
688 for (i = 0; i < total_refblock_count; i++) {
689 be64_to_cpus(&new_table[i]);
692 /* Hook up the new refcount table in the qcow2 header */
693 struct QEMU_PACKED {
694 uint64_t d64;
695 uint32_t d32;
696 } data;
697 data.d64 = cpu_to_be64(table_offset);
698 data.d32 = cpu_to_be32(table_clusters);
699 BLKDBG_EVENT(bs->file, BLKDBG_REFBLOCK_ALLOC_SWITCH_TABLE);
700 ret = bdrv_pwrite_sync(bs->file,
701 offsetof(QCowHeader, refcount_table_offset),
702 &data, sizeof(data));
703 if (ret < 0) {
704 goto fail;
707 /* And switch it in memory */
708 uint64_t old_table_offset = s->refcount_table_offset;
709 uint64_t old_table_size = s->refcount_table_size;
711 g_free(s->refcount_table);
712 s->refcount_table = new_table;
713 s->refcount_table_size = table_size;
714 s->refcount_table_offset = table_offset;
715 update_max_refcount_table_index(s);
717 /* Free old table. */
718 qcow2_free_clusters(bs, old_table_offset, old_table_size * sizeof(uint64_t),
719 QCOW2_DISCARD_OTHER);
721 return end_offset;
723 fail:
724 g_free(new_table);
725 return ret;
728 void qcow2_process_discards(BlockDriverState *bs, int ret)
730 BDRVQcow2State *s = bs->opaque;
731 Qcow2DiscardRegion *d, *next;
733 QTAILQ_FOREACH_SAFE(d, &s->discards, next, next) {
734 QTAILQ_REMOVE(&s->discards, d, next);
736 /* Discard is optional, ignore the return value */
737 if (ret >= 0) {
738 bdrv_pdiscard(bs->file, d->offset, d->bytes);
741 g_free(d);
745 static void update_refcount_discard(BlockDriverState *bs,
746 uint64_t offset, uint64_t length)
748 BDRVQcow2State *s = bs->opaque;
749 Qcow2DiscardRegion *d, *p, *next;
751 QTAILQ_FOREACH(d, &s->discards, next) {
752 uint64_t new_start = MIN(offset, d->offset);
753 uint64_t new_end = MAX(offset + length, d->offset + d->bytes);
755 if (new_end - new_start <= length + d->bytes) {
756 /* There can't be any overlap, areas ending up here have no
757 * references any more and therefore shouldn't get freed another
758 * time. */
759 assert(d->bytes + length == new_end - new_start);
760 d->offset = new_start;
761 d->bytes = new_end - new_start;
762 goto found;
766 d = g_malloc(sizeof(*d));
767 *d = (Qcow2DiscardRegion) {
768 .bs = bs,
769 .offset = offset,
770 .bytes = length,
772 QTAILQ_INSERT_TAIL(&s->discards, d, next);
774 found:
775 /* Merge discard requests if they are adjacent now */
776 QTAILQ_FOREACH_SAFE(p, &s->discards, next, next) {
777 if (p == d
778 || p->offset > d->offset + d->bytes
779 || d->offset > p->offset + p->bytes)
781 continue;
784 /* Still no overlap possible */
785 assert(p->offset == d->offset + d->bytes
786 || d->offset == p->offset + p->bytes);
788 QTAILQ_REMOVE(&s->discards, p, next);
789 d->offset = MIN(d->offset, p->offset);
790 d->bytes += p->bytes;
791 g_free(p);
795 /* XXX: cache several refcount block clusters ? */
796 /* @addend is the absolute value of the addend; if @decrease is set, @addend
797 * will be subtracted from the current refcount, otherwise it will be added */
798 static int QEMU_WARN_UNUSED_RESULT update_refcount(BlockDriverState *bs,
799 int64_t offset,
800 int64_t length,
801 uint64_t addend,
802 bool decrease,
803 enum qcow2_discard_type type)
805 BDRVQcow2State *s = bs->opaque;
806 int64_t start, last, cluster_offset;
807 void *refcount_block = NULL;
808 int64_t old_table_index = -1;
809 int ret;
811 #ifdef DEBUG_ALLOC2
812 fprintf(stderr, "update_refcount: offset=%" PRId64 " size=%" PRId64
813 " addend=%s%" PRIu64 "\n", offset, length, decrease ? "-" : "",
814 addend);
815 #endif
816 if (length < 0) {
817 return -EINVAL;
818 } else if (length == 0) {
819 return 0;
822 if (decrease) {
823 qcow2_cache_set_dependency(bs, s->refcount_block_cache,
824 s->l2_table_cache);
827 start = start_of_cluster(s, offset);
828 last = start_of_cluster(s, offset + length - 1);
829 for(cluster_offset = start; cluster_offset <= last;
830 cluster_offset += s->cluster_size)
832 int block_index;
833 uint64_t refcount;
834 int64_t cluster_index = cluster_offset >> s->cluster_bits;
835 int64_t table_index = cluster_index >> s->refcount_block_bits;
837 /* Load the refcount block and allocate it if needed */
838 if (table_index != old_table_index) {
839 if (refcount_block) {
840 qcow2_cache_put(s->refcount_block_cache, &refcount_block);
842 ret = alloc_refcount_block(bs, cluster_index, &refcount_block);
843 /* If the caller needs to restart the search for free clusters,
844 * try the same ones first to see if they're still free. */
845 if (ret == -EAGAIN) {
846 if (s->free_cluster_index > (start >> s->cluster_bits)) {
847 s->free_cluster_index = (start >> s->cluster_bits);
850 if (ret < 0) {
851 goto fail;
854 old_table_index = table_index;
856 qcow2_cache_entry_mark_dirty(s->refcount_block_cache, refcount_block);
858 /* we can update the count and save it */
859 block_index = cluster_index & (s->refcount_block_size - 1);
861 refcount = s->get_refcount(refcount_block, block_index);
862 if (decrease ? (refcount - addend > refcount)
863 : (refcount + addend < refcount ||
864 refcount + addend > s->refcount_max))
866 ret = -EINVAL;
867 goto fail;
869 if (decrease) {
870 refcount -= addend;
871 } else {
872 refcount += addend;
874 if (refcount == 0 && cluster_index < s->free_cluster_index) {
875 s->free_cluster_index = cluster_index;
877 s->set_refcount(refcount_block, block_index, refcount);
879 if (refcount == 0) {
880 void *table;
882 table = qcow2_cache_is_table_offset(s->refcount_block_cache,
883 offset);
884 if (table != NULL) {
885 qcow2_cache_put(s->refcount_block_cache, &refcount_block);
886 qcow2_cache_discard(s->refcount_block_cache, table);
889 table = qcow2_cache_is_table_offset(s->l2_table_cache, offset);
890 if (table != NULL) {
891 qcow2_cache_discard(s->l2_table_cache, table);
894 if (s->discard_passthrough[type]) {
895 update_refcount_discard(bs, cluster_offset, s->cluster_size);
900 ret = 0;
901 fail:
902 if (!s->cache_discards) {
903 qcow2_process_discards(bs, ret);
906 /* Write last changed block to disk */
907 if (refcount_block) {
908 qcow2_cache_put(s->refcount_block_cache, &refcount_block);
912 * Try do undo any updates if an error is returned (This may succeed in
913 * some cases like ENOSPC for allocating a new refcount block)
915 if (ret < 0) {
916 int dummy;
917 dummy = update_refcount(bs, offset, cluster_offset - offset, addend,
918 !decrease, QCOW2_DISCARD_NEVER);
919 (void)dummy;
922 return ret;
926 * Increases or decreases the refcount of a given cluster.
928 * @addend is the absolute value of the addend; if @decrease is set, @addend
929 * will be subtracted from the current refcount, otherwise it will be added.
931 * On success 0 is returned; on failure -errno is returned.
933 int qcow2_update_cluster_refcount(BlockDriverState *bs,
934 int64_t cluster_index,
935 uint64_t addend, bool decrease,
936 enum qcow2_discard_type type)
938 BDRVQcow2State *s = bs->opaque;
939 int ret;
941 ret = update_refcount(bs, cluster_index << s->cluster_bits, 1, addend,
942 decrease, type);
943 if (ret < 0) {
944 return ret;
947 return 0;
952 /*********************************************************/
953 /* cluster allocation functions */
957 /* return < 0 if error */
958 static int64_t alloc_clusters_noref(BlockDriverState *bs, uint64_t size,
959 uint64_t max)
961 BDRVQcow2State *s = bs->opaque;
962 uint64_t i, nb_clusters, refcount;
963 int ret;
965 /* We can't allocate clusters if they may still be queued for discard. */
966 if (s->cache_discards) {
967 qcow2_process_discards(bs, 0);
970 nb_clusters = size_to_clusters(s, size);
971 retry:
972 for(i = 0; i < nb_clusters; i++) {
973 uint64_t next_cluster_index = s->free_cluster_index++;
974 ret = qcow2_get_refcount(bs, next_cluster_index, &refcount);
976 if (ret < 0) {
977 return ret;
978 } else if (refcount != 0) {
979 goto retry;
983 /* Make sure that all offsets in the "allocated" range are representable
984 * in the requested max */
985 if (s->free_cluster_index > 0 &&
986 s->free_cluster_index - 1 > (max >> s->cluster_bits))
988 return -EFBIG;
991 #ifdef DEBUG_ALLOC2
992 fprintf(stderr, "alloc_clusters: size=%" PRId64 " -> %" PRId64 "\n",
993 size,
994 (s->free_cluster_index - nb_clusters) << s->cluster_bits);
995 #endif
996 return (s->free_cluster_index - nb_clusters) << s->cluster_bits;
999 int64_t qcow2_alloc_clusters(BlockDriverState *bs, uint64_t size)
1001 int64_t offset;
1002 int ret;
1004 BLKDBG_EVENT(bs->file, BLKDBG_CLUSTER_ALLOC);
1005 do {
1006 offset = alloc_clusters_noref(bs, size, QCOW_MAX_CLUSTER_OFFSET);
1007 if (offset < 0) {
1008 return offset;
1011 ret = update_refcount(bs, offset, size, 1, false, QCOW2_DISCARD_NEVER);
1012 } while (ret == -EAGAIN);
1014 if (ret < 0) {
1015 return ret;
1018 return offset;
1021 int64_t qcow2_alloc_clusters_at(BlockDriverState *bs, uint64_t offset,
1022 int64_t nb_clusters)
1024 BDRVQcow2State *s = bs->opaque;
1025 uint64_t cluster_index, refcount;
1026 uint64_t i;
1027 int ret;
1029 assert(nb_clusters >= 0);
1030 if (nb_clusters == 0) {
1031 return 0;
1034 do {
1035 /* Check how many clusters there are free */
1036 cluster_index = offset >> s->cluster_bits;
1037 for(i = 0; i < nb_clusters; i++) {
1038 ret = qcow2_get_refcount(bs, cluster_index++, &refcount);
1039 if (ret < 0) {
1040 return ret;
1041 } else if (refcount != 0) {
1042 break;
1046 /* And then allocate them */
1047 ret = update_refcount(bs, offset, i << s->cluster_bits, 1, false,
1048 QCOW2_DISCARD_NEVER);
1049 } while (ret == -EAGAIN);
1051 if (ret < 0) {
1052 return ret;
1055 return i;
1058 /* only used to allocate compressed sectors. We try to allocate
1059 contiguous sectors. size must be <= cluster_size */
1060 int64_t qcow2_alloc_bytes(BlockDriverState *bs, int size)
1062 BDRVQcow2State *s = bs->opaque;
1063 int64_t offset;
1064 size_t free_in_cluster;
1065 int ret;
1067 BLKDBG_EVENT(bs->file, BLKDBG_CLUSTER_ALLOC_BYTES);
1068 assert(size > 0 && size <= s->cluster_size);
1069 assert(!s->free_byte_offset || offset_into_cluster(s, s->free_byte_offset));
1071 offset = s->free_byte_offset;
1073 if (offset) {
1074 uint64_t refcount;
1075 ret = qcow2_get_refcount(bs, offset >> s->cluster_bits, &refcount);
1076 if (ret < 0) {
1077 return ret;
1080 if (refcount == s->refcount_max) {
1081 offset = 0;
1085 free_in_cluster = s->cluster_size - offset_into_cluster(s, offset);
1086 do {
1087 if (!offset || free_in_cluster < size) {
1088 int64_t new_cluster;
1090 new_cluster = alloc_clusters_noref(bs, s->cluster_size,
1091 MIN(s->cluster_offset_mask,
1092 QCOW_MAX_CLUSTER_OFFSET));
1093 if (new_cluster < 0) {
1094 return new_cluster;
1097 if (new_cluster == 0) {
1098 qcow2_signal_corruption(bs, true, -1, -1, "Preventing invalid "
1099 "allocation of compressed cluster "
1100 "at offset 0");
1101 return -EIO;
1104 if (!offset || ROUND_UP(offset, s->cluster_size) != new_cluster) {
1105 offset = new_cluster;
1106 free_in_cluster = s->cluster_size;
1107 } else {
1108 free_in_cluster += s->cluster_size;
1112 assert(offset);
1113 ret = update_refcount(bs, offset, size, 1, false, QCOW2_DISCARD_NEVER);
1114 if (ret < 0) {
1115 offset = 0;
1117 } while (ret == -EAGAIN);
1118 if (ret < 0) {
1119 return ret;
1122 /* The cluster refcount was incremented; refcount blocks must be flushed
1123 * before the caller's L2 table updates. */
1124 qcow2_cache_set_dependency(bs, s->l2_table_cache, s->refcount_block_cache);
1126 s->free_byte_offset = offset + size;
1127 if (!offset_into_cluster(s, s->free_byte_offset)) {
1128 s->free_byte_offset = 0;
1131 return offset;
1134 void qcow2_free_clusters(BlockDriverState *bs,
1135 int64_t offset, int64_t size,
1136 enum qcow2_discard_type type)
1138 int ret;
1140 BLKDBG_EVENT(bs->file, BLKDBG_CLUSTER_FREE);
1141 ret = update_refcount(bs, offset, size, 1, true, type);
1142 if (ret < 0) {
1143 fprintf(stderr, "qcow2_free_clusters failed: %s\n", strerror(-ret));
1144 /* TODO Remember the clusters to free them later and avoid leaking */
1149 * Free a cluster using its L2 entry (handles clusters of all types, e.g.
1150 * normal cluster, compressed cluster, etc.)
1152 void qcow2_free_any_clusters(BlockDriverState *bs, uint64_t l2_entry,
1153 int nb_clusters, enum qcow2_discard_type type)
1155 BDRVQcow2State *s = bs->opaque;
1157 switch (qcow2_get_cluster_type(l2_entry)) {
1158 case QCOW2_CLUSTER_COMPRESSED:
1160 int nb_csectors;
1161 nb_csectors = ((l2_entry >> s->csize_shift) &
1162 s->csize_mask) + 1;
1163 qcow2_free_clusters(bs,
1164 (l2_entry & s->cluster_offset_mask) & ~511,
1165 nb_csectors * 512, type);
1167 break;
1168 case QCOW2_CLUSTER_NORMAL:
1169 case QCOW2_CLUSTER_ZERO_ALLOC:
1170 if (offset_into_cluster(s, l2_entry & L2E_OFFSET_MASK)) {
1171 qcow2_signal_corruption(bs, false, -1, -1,
1172 "Cannot free unaligned cluster %#llx",
1173 l2_entry & L2E_OFFSET_MASK);
1174 } else {
1175 qcow2_free_clusters(bs, l2_entry & L2E_OFFSET_MASK,
1176 nb_clusters << s->cluster_bits, type);
1178 break;
1179 case QCOW2_CLUSTER_ZERO_PLAIN:
1180 case QCOW2_CLUSTER_UNALLOCATED:
1181 break;
1182 default:
1183 abort();
1187 int coroutine_fn qcow2_write_caches(BlockDriverState *bs)
1189 BDRVQcow2State *s = bs->opaque;
1190 int ret;
1192 ret = qcow2_cache_write(bs, s->l2_table_cache);
1193 if (ret < 0) {
1194 return ret;
1197 if (qcow2_need_accurate_refcounts(s)) {
1198 ret = qcow2_cache_write(bs, s->refcount_block_cache);
1199 if (ret < 0) {
1200 return ret;
1204 return 0;
1207 int coroutine_fn qcow2_flush_caches(BlockDriverState *bs)
1209 int ret = qcow2_write_caches(bs);
1210 if (ret < 0) {
1211 return ret;
1214 return bdrv_flush(bs->file->bs);
1217 /*********************************************************/
1218 /* snapshots and image creation */
1222 /* update the refcounts of snapshots and the copied flag */
1223 int qcow2_update_snapshot_refcount(BlockDriverState *bs,
1224 int64_t l1_table_offset, int l1_size, int addend)
1226 BDRVQcow2State *s = bs->opaque;
1227 uint64_t *l1_table, *l2_slice, l2_offset, entry, l1_size2, refcount;
1228 bool l1_allocated = false;
1229 int64_t old_entry, old_l2_offset;
1230 unsigned slice, slice_size2, n_slices;
1231 int i, j, l1_modified = 0, nb_csectors;
1232 int ret;
1234 assert(addend >= -1 && addend <= 1);
1236 l2_slice = NULL;
1237 l1_table = NULL;
1238 l1_size2 = l1_size * sizeof(uint64_t);
1239 slice_size2 = s->l2_slice_size * sizeof(uint64_t);
1240 n_slices = s->cluster_size / slice_size2;
1242 s->cache_discards = true;
1244 /* WARNING: qcow2_snapshot_goto relies on this function not using the
1245 * l1_table_offset when it is the current s->l1_table_offset! Be careful
1246 * when changing this! */
1247 if (l1_table_offset != s->l1_table_offset) {
1248 l1_table = g_try_malloc0(ROUND_UP(l1_size2, 512));
1249 if (l1_size2 && l1_table == NULL) {
1250 ret = -ENOMEM;
1251 goto fail;
1253 l1_allocated = true;
1255 ret = bdrv_pread(bs->file, l1_table_offset, l1_table, l1_size2);
1256 if (ret < 0) {
1257 goto fail;
1260 for (i = 0; i < l1_size; i++) {
1261 be64_to_cpus(&l1_table[i]);
1263 } else {
1264 assert(l1_size == s->l1_size);
1265 l1_table = s->l1_table;
1266 l1_allocated = false;
1269 for (i = 0; i < l1_size; i++) {
1270 l2_offset = l1_table[i];
1271 if (l2_offset) {
1272 old_l2_offset = l2_offset;
1273 l2_offset &= L1E_OFFSET_MASK;
1275 if (offset_into_cluster(s, l2_offset)) {
1276 qcow2_signal_corruption(bs, true, -1, -1, "L2 table offset %#"
1277 PRIx64 " unaligned (L1 index: %#x)",
1278 l2_offset, i);
1279 ret = -EIO;
1280 goto fail;
1283 for (slice = 0; slice < n_slices; slice++) {
1284 ret = qcow2_cache_get(bs, s->l2_table_cache,
1285 l2_offset + slice * slice_size2,
1286 (void **) &l2_slice);
1287 if (ret < 0) {
1288 goto fail;
1291 for (j = 0; j < s->l2_slice_size; j++) {
1292 uint64_t cluster_index;
1293 uint64_t offset;
1295 entry = be64_to_cpu(l2_slice[j]);
1296 old_entry = entry;
1297 entry &= ~QCOW_OFLAG_COPIED;
1298 offset = entry & L2E_OFFSET_MASK;
1300 switch (qcow2_get_cluster_type(entry)) {
1301 case QCOW2_CLUSTER_COMPRESSED:
1302 nb_csectors = ((entry >> s->csize_shift) &
1303 s->csize_mask) + 1;
1304 if (addend != 0) {
1305 ret = update_refcount(
1306 bs, (entry & s->cluster_offset_mask) & ~511,
1307 nb_csectors * 512, abs(addend), addend < 0,
1308 QCOW2_DISCARD_SNAPSHOT);
1309 if (ret < 0) {
1310 goto fail;
1313 /* compressed clusters are never modified */
1314 refcount = 2;
1315 break;
1317 case QCOW2_CLUSTER_NORMAL:
1318 case QCOW2_CLUSTER_ZERO_ALLOC:
1319 if (offset_into_cluster(s, offset)) {
1320 /* Here l2_index means table (not slice) index */
1321 int l2_index = slice * s->l2_slice_size + j;
1322 qcow2_signal_corruption(
1323 bs, true, -1, -1, "Cluster "
1324 "allocation offset %#" PRIx64
1325 " unaligned (L2 offset: %#"
1326 PRIx64 ", L2 index: %#x)",
1327 offset, l2_offset, l2_index);
1328 ret = -EIO;
1329 goto fail;
1332 cluster_index = offset >> s->cluster_bits;
1333 assert(cluster_index);
1334 if (addend != 0) {
1335 ret = qcow2_update_cluster_refcount(
1336 bs, cluster_index, abs(addend), addend < 0,
1337 QCOW2_DISCARD_SNAPSHOT);
1338 if (ret < 0) {
1339 goto fail;
1343 ret = qcow2_get_refcount(bs, cluster_index, &refcount);
1344 if (ret < 0) {
1345 goto fail;
1347 break;
1349 case QCOW2_CLUSTER_ZERO_PLAIN:
1350 case QCOW2_CLUSTER_UNALLOCATED:
1351 refcount = 0;
1352 break;
1354 default:
1355 abort();
1358 if (refcount == 1) {
1359 entry |= QCOW_OFLAG_COPIED;
1361 if (entry != old_entry) {
1362 if (addend > 0) {
1363 qcow2_cache_set_dependency(bs, s->l2_table_cache,
1364 s->refcount_block_cache);
1366 l2_slice[j] = cpu_to_be64(entry);
1367 qcow2_cache_entry_mark_dirty(s->l2_table_cache,
1368 l2_slice);
1372 qcow2_cache_put(s->l2_table_cache, (void **) &l2_slice);
1375 if (addend != 0) {
1376 ret = qcow2_update_cluster_refcount(bs, l2_offset >>
1377 s->cluster_bits,
1378 abs(addend), addend < 0,
1379 QCOW2_DISCARD_SNAPSHOT);
1380 if (ret < 0) {
1381 goto fail;
1384 ret = qcow2_get_refcount(bs, l2_offset >> s->cluster_bits,
1385 &refcount);
1386 if (ret < 0) {
1387 goto fail;
1388 } else if (refcount == 1) {
1389 l2_offset |= QCOW_OFLAG_COPIED;
1391 if (l2_offset != old_l2_offset) {
1392 l1_table[i] = l2_offset;
1393 l1_modified = 1;
1398 ret = bdrv_flush(bs);
1399 fail:
1400 if (l2_slice) {
1401 qcow2_cache_put(s->l2_table_cache, (void **) &l2_slice);
1404 s->cache_discards = false;
1405 qcow2_process_discards(bs, ret);
1407 /* Update L1 only if it isn't deleted anyway (addend = -1) */
1408 if (ret == 0 && addend >= 0 && l1_modified) {
1409 for (i = 0; i < l1_size; i++) {
1410 cpu_to_be64s(&l1_table[i]);
1413 ret = bdrv_pwrite_sync(bs->file, l1_table_offset,
1414 l1_table, l1_size2);
1416 for (i = 0; i < l1_size; i++) {
1417 be64_to_cpus(&l1_table[i]);
1420 if (l1_allocated)
1421 g_free(l1_table);
1422 return ret;
1428 /*********************************************************/
1429 /* refcount checking functions */
1432 static uint64_t refcount_array_byte_size(BDRVQcow2State *s, uint64_t entries)
1434 /* This assertion holds because there is no way we can address more than
1435 * 2^(64 - 9) clusters at once (with cluster size 512 = 2^9, and because
1436 * offsets have to be representable in bytes); due to every cluster
1437 * corresponding to one refcount entry, we are well below that limit */
1438 assert(entries < (UINT64_C(1) << (64 - 9)));
1440 /* Thanks to the assertion this will not overflow, because
1441 * s->refcount_order < 7.
1442 * (note: x << s->refcount_order == x * s->refcount_bits) */
1443 return DIV_ROUND_UP(entries << s->refcount_order, 8);
1447 * Reallocates *array so that it can hold new_size entries. *size must contain
1448 * the current number of entries in *array. If the reallocation fails, *array
1449 * and *size will not be modified and -errno will be returned. If the
1450 * reallocation is successful, *array will be set to the new buffer, *size
1451 * will be set to new_size and 0 will be returned. The size of the reallocated
1452 * refcount array buffer will be aligned to a cluster boundary, and the newly
1453 * allocated area will be zeroed.
1455 static int realloc_refcount_array(BDRVQcow2State *s, void **array,
1456 int64_t *size, int64_t new_size)
1458 int64_t old_byte_size, new_byte_size;
1459 void *new_ptr;
1461 /* Round to clusters so the array can be directly written to disk */
1462 old_byte_size = size_to_clusters(s, refcount_array_byte_size(s, *size))
1463 * s->cluster_size;
1464 new_byte_size = size_to_clusters(s, refcount_array_byte_size(s, new_size))
1465 * s->cluster_size;
1467 if (new_byte_size == old_byte_size) {
1468 *size = new_size;
1469 return 0;
1472 assert(new_byte_size > 0);
1474 if (new_byte_size > SIZE_MAX) {
1475 return -ENOMEM;
1478 new_ptr = g_try_realloc(*array, new_byte_size);
1479 if (!new_ptr) {
1480 return -ENOMEM;
1483 if (new_byte_size > old_byte_size) {
1484 memset((char *)new_ptr + old_byte_size, 0,
1485 new_byte_size - old_byte_size);
1488 *array = new_ptr;
1489 *size = new_size;
1491 return 0;
1495 * Increases the refcount for a range of clusters in a given refcount table.
1496 * This is used to construct a temporary refcount table out of L1 and L2 tables
1497 * which can be compared to the refcount table saved in the image.
1499 * Modifies the number of errors in res.
1501 int qcow2_inc_refcounts_imrt(BlockDriverState *bs, BdrvCheckResult *res,
1502 void **refcount_table,
1503 int64_t *refcount_table_size,
1504 int64_t offset, int64_t size)
1506 BDRVQcow2State *s = bs->opaque;
1507 uint64_t start, last, cluster_offset, k, refcount;
1508 int ret;
1510 if (size <= 0) {
1511 return 0;
1514 start = start_of_cluster(s, offset);
1515 last = start_of_cluster(s, offset + size - 1);
1516 for(cluster_offset = start; cluster_offset <= last;
1517 cluster_offset += s->cluster_size) {
1518 k = cluster_offset >> s->cluster_bits;
1519 if (k >= *refcount_table_size) {
1520 ret = realloc_refcount_array(s, refcount_table,
1521 refcount_table_size, k + 1);
1522 if (ret < 0) {
1523 res->check_errors++;
1524 return ret;
1528 refcount = s->get_refcount(*refcount_table, k);
1529 if (refcount == s->refcount_max) {
1530 fprintf(stderr, "ERROR: overflow cluster offset=0x%" PRIx64
1531 "\n", cluster_offset);
1532 fprintf(stderr, "Use qemu-img amend to increase the refcount entry "
1533 "width or qemu-img convert to create a clean copy if the "
1534 "image cannot be opened for writing\n");
1535 res->corruptions++;
1536 continue;
1538 s->set_refcount(*refcount_table, k, refcount + 1);
1541 return 0;
1544 /* Flags for check_refcounts_l1() and check_refcounts_l2() */
1545 enum {
1546 CHECK_FRAG_INFO = 0x2, /* update BlockFragInfo counters */
1550 * Increases the refcount in the given refcount table for the all clusters
1551 * referenced in the L2 table. While doing so, performs some checks on L2
1552 * entries.
1554 * Returns the number of errors found by the checks or -errno if an internal
1555 * error occurred.
1557 static int check_refcounts_l2(BlockDriverState *bs, BdrvCheckResult *res,
1558 void **refcount_table,
1559 int64_t *refcount_table_size, int64_t l2_offset,
1560 int flags, BdrvCheckMode fix)
1562 BDRVQcow2State *s = bs->opaque;
1563 uint64_t *l2_table, l2_entry;
1564 uint64_t next_contiguous_offset = 0;
1565 int i, l2_size, nb_csectors, ret;
1567 /* Read L2 table from disk */
1568 l2_size = s->l2_size * sizeof(uint64_t);
1569 l2_table = g_malloc(l2_size);
1571 ret = bdrv_pread(bs->file, l2_offset, l2_table, l2_size);
1572 if (ret < 0) {
1573 fprintf(stderr, "ERROR: I/O error in check_refcounts_l2\n");
1574 res->check_errors++;
1575 goto fail;
1578 /* Do the actual checks */
1579 for(i = 0; i < s->l2_size; i++) {
1580 l2_entry = be64_to_cpu(l2_table[i]);
1582 switch (qcow2_get_cluster_type(l2_entry)) {
1583 case QCOW2_CLUSTER_COMPRESSED:
1584 /* Compressed clusters don't have QCOW_OFLAG_COPIED */
1585 if (l2_entry & QCOW_OFLAG_COPIED) {
1586 fprintf(stderr, "ERROR: coffset=0x%" PRIx64 ": "
1587 "copied flag must never be set for compressed "
1588 "clusters\n", l2_entry & s->cluster_offset_mask);
1589 l2_entry &= ~QCOW_OFLAG_COPIED;
1590 res->corruptions++;
1593 /* Mark cluster as used */
1594 nb_csectors = ((l2_entry >> s->csize_shift) &
1595 s->csize_mask) + 1;
1596 l2_entry &= s->cluster_offset_mask;
1597 ret = qcow2_inc_refcounts_imrt(bs, res,
1598 refcount_table, refcount_table_size,
1599 l2_entry & ~511, nb_csectors * 512);
1600 if (ret < 0) {
1601 goto fail;
1604 if (flags & CHECK_FRAG_INFO) {
1605 res->bfi.allocated_clusters++;
1606 res->bfi.compressed_clusters++;
1608 /* Compressed clusters are fragmented by nature. Since they
1609 * take up sub-sector space but we only have sector granularity
1610 * I/O we need to re-read the same sectors even for adjacent
1611 * compressed clusters.
1613 res->bfi.fragmented_clusters++;
1615 break;
1617 case QCOW2_CLUSTER_ZERO_ALLOC:
1618 case QCOW2_CLUSTER_NORMAL:
1620 uint64_t offset = l2_entry & L2E_OFFSET_MASK;
1622 if (flags & CHECK_FRAG_INFO) {
1623 res->bfi.allocated_clusters++;
1624 if (next_contiguous_offset &&
1625 offset != next_contiguous_offset) {
1626 res->bfi.fragmented_clusters++;
1628 next_contiguous_offset = offset + s->cluster_size;
1631 /* Correct offsets are cluster aligned */
1632 if (offset_into_cluster(s, offset)) {
1633 if (qcow2_get_cluster_type(l2_entry) ==
1634 QCOW2_CLUSTER_ZERO_ALLOC)
1636 fprintf(stderr, "%s offset=%" PRIx64 ": Preallocated zero "
1637 "cluster is not properly aligned; L2 entry "
1638 "corrupted.\n",
1639 fix & BDRV_FIX_ERRORS ? "Repairing" : "ERROR",
1640 offset);
1641 if (fix & BDRV_FIX_ERRORS) {
1642 uint64_t l2e_offset =
1643 l2_offset + (uint64_t)i * sizeof(uint64_t);
1645 l2_entry = QCOW_OFLAG_ZERO;
1646 l2_table[i] = cpu_to_be64(l2_entry);
1647 ret = qcow2_pre_write_overlap_check(bs,
1648 QCOW2_OL_ACTIVE_L2 | QCOW2_OL_INACTIVE_L2,
1649 l2e_offset, sizeof(uint64_t));
1650 if (ret < 0) {
1651 fprintf(stderr, "ERROR: Overlap check failed\n");
1652 res->check_errors++;
1653 /* Something is seriously wrong, so abort checking
1654 * this L2 table */
1655 goto fail;
1658 ret = bdrv_pwrite_sync(bs->file, l2e_offset,
1659 &l2_table[i], sizeof(uint64_t));
1660 if (ret < 0) {
1661 fprintf(stderr, "ERROR: Failed to overwrite L2 "
1662 "table entry: %s\n", strerror(-ret));
1663 res->check_errors++;
1664 /* Do not abort, continue checking the rest of this
1665 * L2 table's entries */
1666 } else {
1667 res->corruptions_fixed++;
1668 /* Skip marking the cluster as used
1669 * (it is unused now) */
1670 continue;
1672 } else {
1673 res->corruptions++;
1675 } else {
1676 fprintf(stderr, "ERROR offset=%" PRIx64 ": Data cluster is "
1677 "not properly aligned; L2 entry corrupted.\n", offset);
1678 res->corruptions++;
1682 /* Mark cluster as used */
1683 ret = qcow2_inc_refcounts_imrt(bs, res,
1684 refcount_table, refcount_table_size,
1685 offset, s->cluster_size);
1686 if (ret < 0) {
1687 goto fail;
1689 break;
1692 case QCOW2_CLUSTER_ZERO_PLAIN:
1693 case QCOW2_CLUSTER_UNALLOCATED:
1694 break;
1696 default:
1697 abort();
1701 g_free(l2_table);
1702 return 0;
1704 fail:
1705 g_free(l2_table);
1706 return ret;
1710 * Increases the refcount for the L1 table, its L2 tables and all referenced
1711 * clusters in the given refcount table. While doing so, performs some checks
1712 * on L1 and L2 entries.
1714 * Returns the number of errors found by the checks or -errno if an internal
1715 * error occurred.
1717 static int check_refcounts_l1(BlockDriverState *bs,
1718 BdrvCheckResult *res,
1719 void **refcount_table,
1720 int64_t *refcount_table_size,
1721 int64_t l1_table_offset, int l1_size,
1722 int flags, BdrvCheckMode fix)
1724 BDRVQcow2State *s = bs->opaque;
1725 uint64_t *l1_table = NULL, l2_offset, l1_size2;
1726 int i, ret;
1728 l1_size2 = l1_size * sizeof(uint64_t);
1730 /* Mark L1 table as used */
1731 ret = qcow2_inc_refcounts_imrt(bs, res, refcount_table, refcount_table_size,
1732 l1_table_offset, l1_size2);
1733 if (ret < 0) {
1734 goto fail;
1737 /* Read L1 table entries from disk */
1738 if (l1_size2 > 0) {
1739 l1_table = g_try_malloc(l1_size2);
1740 if (l1_table == NULL) {
1741 ret = -ENOMEM;
1742 res->check_errors++;
1743 goto fail;
1745 ret = bdrv_pread(bs->file, l1_table_offset, l1_table, l1_size2);
1746 if (ret < 0) {
1747 fprintf(stderr, "ERROR: I/O error in check_refcounts_l1\n");
1748 res->check_errors++;
1749 goto fail;
1751 for(i = 0;i < l1_size; i++)
1752 be64_to_cpus(&l1_table[i]);
1755 /* Do the actual checks */
1756 for(i = 0; i < l1_size; i++) {
1757 l2_offset = l1_table[i];
1758 if (l2_offset) {
1759 /* Mark L2 table as used */
1760 l2_offset &= L1E_OFFSET_MASK;
1761 ret = qcow2_inc_refcounts_imrt(bs, res,
1762 refcount_table, refcount_table_size,
1763 l2_offset, s->cluster_size);
1764 if (ret < 0) {
1765 goto fail;
1768 /* L2 tables are cluster aligned */
1769 if (offset_into_cluster(s, l2_offset)) {
1770 fprintf(stderr, "ERROR l2_offset=%" PRIx64 ": Table is not "
1771 "cluster aligned; L1 entry corrupted\n", l2_offset);
1772 res->corruptions++;
1775 /* Process and check L2 entries */
1776 ret = check_refcounts_l2(bs, res, refcount_table,
1777 refcount_table_size, l2_offset, flags,
1778 fix);
1779 if (ret < 0) {
1780 goto fail;
1784 g_free(l1_table);
1785 return 0;
1787 fail:
1788 g_free(l1_table);
1789 return ret;
1793 * Checks the OFLAG_COPIED flag for all L1 and L2 entries.
1795 * This function does not print an error message nor does it increment
1796 * check_errors if qcow2_get_refcount fails (this is because such an error will
1797 * have been already detected and sufficiently signaled by the calling function
1798 * (qcow2_check_refcounts) by the time this function is called).
1800 static int check_oflag_copied(BlockDriverState *bs, BdrvCheckResult *res,
1801 BdrvCheckMode fix)
1803 BDRVQcow2State *s = bs->opaque;
1804 uint64_t *l2_table = qemu_blockalign(bs, s->cluster_size);
1805 int ret;
1806 uint64_t refcount;
1807 int i, j;
1808 bool repair;
1810 if (fix & BDRV_FIX_ERRORS) {
1811 /* Always repair */
1812 repair = true;
1813 } else if (fix & BDRV_FIX_LEAKS) {
1814 /* Repair only if that seems safe: This function is always
1815 * called after the refcounts have been fixed, so the refcount
1816 * is accurate if that repair was successful */
1817 repair = !res->check_errors && !res->corruptions && !res->leaks;
1818 } else {
1819 repair = false;
1822 for (i = 0; i < s->l1_size; i++) {
1823 uint64_t l1_entry = s->l1_table[i];
1824 uint64_t l2_offset = l1_entry & L1E_OFFSET_MASK;
1825 bool l2_dirty = false;
1827 if (!l2_offset) {
1828 continue;
1831 ret = qcow2_get_refcount(bs, l2_offset >> s->cluster_bits,
1832 &refcount);
1833 if (ret < 0) {
1834 /* don't print message nor increment check_errors */
1835 continue;
1837 if ((refcount == 1) != ((l1_entry & QCOW_OFLAG_COPIED) != 0)) {
1838 fprintf(stderr, "%s OFLAG_COPIED L2 cluster: l1_index=%d "
1839 "l1_entry=%" PRIx64 " refcount=%" PRIu64 "\n",
1840 repair ? "Repairing" : "ERROR", i, l1_entry, refcount);
1841 if (repair) {
1842 s->l1_table[i] = refcount == 1
1843 ? l1_entry | QCOW_OFLAG_COPIED
1844 : l1_entry & ~QCOW_OFLAG_COPIED;
1845 ret = qcow2_write_l1_entry(bs, i);
1846 if (ret < 0) {
1847 res->check_errors++;
1848 goto fail;
1850 res->corruptions_fixed++;
1851 } else {
1852 res->corruptions++;
1856 ret = bdrv_pread(bs->file, l2_offset, l2_table,
1857 s->l2_size * sizeof(uint64_t));
1858 if (ret < 0) {
1859 fprintf(stderr, "ERROR: Could not read L2 table: %s\n",
1860 strerror(-ret));
1861 res->check_errors++;
1862 goto fail;
1865 for (j = 0; j < s->l2_size; j++) {
1866 uint64_t l2_entry = be64_to_cpu(l2_table[j]);
1867 uint64_t data_offset = l2_entry & L2E_OFFSET_MASK;
1868 QCow2ClusterType cluster_type = qcow2_get_cluster_type(l2_entry);
1870 if (cluster_type == QCOW2_CLUSTER_NORMAL ||
1871 cluster_type == QCOW2_CLUSTER_ZERO_ALLOC) {
1872 ret = qcow2_get_refcount(bs,
1873 data_offset >> s->cluster_bits,
1874 &refcount);
1875 if (ret < 0) {
1876 /* don't print message nor increment check_errors */
1877 continue;
1879 if ((refcount == 1) != ((l2_entry & QCOW_OFLAG_COPIED) != 0)) {
1880 fprintf(stderr, "%s OFLAG_COPIED data cluster: "
1881 "l2_entry=%" PRIx64 " refcount=%" PRIu64 "\n",
1882 repair ? "Repairing" : "ERROR", l2_entry, refcount);
1883 if (repair) {
1884 l2_table[j] = cpu_to_be64(refcount == 1
1885 ? l2_entry | QCOW_OFLAG_COPIED
1886 : l2_entry & ~QCOW_OFLAG_COPIED);
1887 l2_dirty = true;
1888 res->corruptions_fixed++;
1889 } else {
1890 res->corruptions++;
1896 if (l2_dirty) {
1897 ret = qcow2_pre_write_overlap_check(bs, QCOW2_OL_ACTIVE_L2,
1898 l2_offset, s->cluster_size);
1899 if (ret < 0) {
1900 fprintf(stderr, "ERROR: Could not write L2 table; metadata "
1901 "overlap check failed: %s\n", strerror(-ret));
1902 res->check_errors++;
1903 goto fail;
1906 ret = bdrv_pwrite(bs->file, l2_offset, l2_table,
1907 s->cluster_size);
1908 if (ret < 0) {
1909 fprintf(stderr, "ERROR: Could not write L2 table: %s\n",
1910 strerror(-ret));
1911 res->check_errors++;
1912 goto fail;
1917 ret = 0;
1919 fail:
1920 qemu_vfree(l2_table);
1921 return ret;
1925 * Checks consistency of refblocks and accounts for each refblock in
1926 * *refcount_table.
1928 static int check_refblocks(BlockDriverState *bs, BdrvCheckResult *res,
1929 BdrvCheckMode fix, bool *rebuild,
1930 void **refcount_table, int64_t *nb_clusters)
1932 BDRVQcow2State *s = bs->opaque;
1933 int64_t i, size;
1934 int ret;
1936 for(i = 0; i < s->refcount_table_size; i++) {
1937 uint64_t offset, cluster;
1938 offset = s->refcount_table[i];
1939 cluster = offset >> s->cluster_bits;
1941 /* Refcount blocks are cluster aligned */
1942 if (offset_into_cluster(s, offset)) {
1943 fprintf(stderr, "ERROR refcount block %" PRId64 " is not "
1944 "cluster aligned; refcount table entry corrupted\n", i);
1945 res->corruptions++;
1946 *rebuild = true;
1947 continue;
1950 if (cluster >= *nb_clusters) {
1951 fprintf(stderr, "%s refcount block %" PRId64 " is outside image\n",
1952 fix & BDRV_FIX_ERRORS ? "Repairing" : "ERROR", i);
1954 if (fix & BDRV_FIX_ERRORS) {
1955 int64_t new_nb_clusters;
1956 Error *local_err = NULL;
1958 if (offset > INT64_MAX - s->cluster_size) {
1959 ret = -EINVAL;
1960 goto resize_fail;
1963 ret = bdrv_truncate(bs->file, offset + s->cluster_size,
1964 PREALLOC_MODE_OFF, &local_err);
1965 if (ret < 0) {
1966 error_report_err(local_err);
1967 goto resize_fail;
1969 size = bdrv_getlength(bs->file->bs);
1970 if (size < 0) {
1971 ret = size;
1972 goto resize_fail;
1975 new_nb_clusters = size_to_clusters(s, size);
1976 assert(new_nb_clusters >= *nb_clusters);
1978 ret = realloc_refcount_array(s, refcount_table,
1979 nb_clusters, new_nb_clusters);
1980 if (ret < 0) {
1981 res->check_errors++;
1982 return ret;
1985 if (cluster >= *nb_clusters) {
1986 ret = -EINVAL;
1987 goto resize_fail;
1990 res->corruptions_fixed++;
1991 ret = qcow2_inc_refcounts_imrt(bs, res,
1992 refcount_table, nb_clusters,
1993 offset, s->cluster_size);
1994 if (ret < 0) {
1995 return ret;
1997 /* No need to check whether the refcount is now greater than 1:
1998 * This area was just allocated and zeroed, so it can only be
1999 * exactly 1 after qcow2_inc_refcounts_imrt() */
2000 continue;
2002 resize_fail:
2003 res->corruptions++;
2004 *rebuild = true;
2005 fprintf(stderr, "ERROR could not resize image: %s\n",
2006 strerror(-ret));
2007 } else {
2008 res->corruptions++;
2010 continue;
2013 if (offset != 0) {
2014 ret = qcow2_inc_refcounts_imrt(bs, res, refcount_table, nb_clusters,
2015 offset, s->cluster_size);
2016 if (ret < 0) {
2017 return ret;
2019 if (s->get_refcount(*refcount_table, cluster) != 1) {
2020 fprintf(stderr, "ERROR refcount block %" PRId64
2021 " refcount=%" PRIu64 "\n", i,
2022 s->get_refcount(*refcount_table, cluster));
2023 res->corruptions++;
2024 *rebuild = true;
2029 return 0;
2033 * Calculates an in-memory refcount table.
2035 static int calculate_refcounts(BlockDriverState *bs, BdrvCheckResult *res,
2036 BdrvCheckMode fix, bool *rebuild,
2037 void **refcount_table, int64_t *nb_clusters)
2039 BDRVQcow2State *s = bs->opaque;
2040 int64_t i;
2041 QCowSnapshot *sn;
2042 int ret;
2044 if (!*refcount_table) {
2045 int64_t old_size = 0;
2046 ret = realloc_refcount_array(s, refcount_table,
2047 &old_size, *nb_clusters);
2048 if (ret < 0) {
2049 res->check_errors++;
2050 return ret;
2054 /* header */
2055 ret = qcow2_inc_refcounts_imrt(bs, res, refcount_table, nb_clusters,
2056 0, s->cluster_size);
2057 if (ret < 0) {
2058 return ret;
2061 /* current L1 table */
2062 ret = check_refcounts_l1(bs, res, refcount_table, nb_clusters,
2063 s->l1_table_offset, s->l1_size, CHECK_FRAG_INFO,
2064 fix);
2065 if (ret < 0) {
2066 return ret;
2069 /* snapshots */
2070 for (i = 0; i < s->nb_snapshots; i++) {
2071 sn = s->snapshots + i;
2072 if (offset_into_cluster(s, sn->l1_table_offset)) {
2073 fprintf(stderr, "ERROR snapshot %s (%s) l1_offset=%#" PRIx64 ": "
2074 "L1 table is not cluster aligned; snapshot table entry "
2075 "corrupted\n", sn->id_str, sn->name, sn->l1_table_offset);
2076 res->corruptions++;
2077 continue;
2079 if (sn->l1_size > QCOW_MAX_L1_SIZE / sizeof(uint64_t)) {
2080 fprintf(stderr, "ERROR snapshot %s (%s) l1_size=%#" PRIx32 ": "
2081 "L1 table is too large; snapshot table entry corrupted\n",
2082 sn->id_str, sn->name, sn->l1_size);
2083 res->corruptions++;
2084 continue;
2086 ret = check_refcounts_l1(bs, res, refcount_table, nb_clusters,
2087 sn->l1_table_offset, sn->l1_size, 0, fix);
2088 if (ret < 0) {
2089 return ret;
2092 ret = qcow2_inc_refcounts_imrt(bs, res, refcount_table, nb_clusters,
2093 s->snapshots_offset, s->snapshots_size);
2094 if (ret < 0) {
2095 return ret;
2098 /* refcount data */
2099 ret = qcow2_inc_refcounts_imrt(bs, res, refcount_table, nb_clusters,
2100 s->refcount_table_offset,
2101 s->refcount_table_size * sizeof(uint64_t));
2102 if (ret < 0) {
2103 return ret;
2106 /* encryption */
2107 if (s->crypto_header.length) {
2108 ret = qcow2_inc_refcounts_imrt(bs, res, refcount_table, nb_clusters,
2109 s->crypto_header.offset,
2110 s->crypto_header.length);
2111 if (ret < 0) {
2112 return ret;
2116 /* bitmaps */
2117 ret = qcow2_check_bitmaps_refcounts(bs, res, refcount_table, nb_clusters);
2118 if (ret < 0) {
2119 return ret;
2122 return check_refblocks(bs, res, fix, rebuild, refcount_table, nb_clusters);
2126 * Compares the actual reference count for each cluster in the image against the
2127 * refcount as reported by the refcount structures on-disk.
2129 static void compare_refcounts(BlockDriverState *bs, BdrvCheckResult *res,
2130 BdrvCheckMode fix, bool *rebuild,
2131 int64_t *highest_cluster,
2132 void *refcount_table, int64_t nb_clusters)
2134 BDRVQcow2State *s = bs->opaque;
2135 int64_t i;
2136 uint64_t refcount1, refcount2;
2137 int ret;
2139 for (i = 0, *highest_cluster = 0; i < nb_clusters; i++) {
2140 ret = qcow2_get_refcount(bs, i, &refcount1);
2141 if (ret < 0) {
2142 fprintf(stderr, "Can't get refcount for cluster %" PRId64 ": %s\n",
2143 i, strerror(-ret));
2144 res->check_errors++;
2145 continue;
2148 refcount2 = s->get_refcount(refcount_table, i);
2150 if (refcount1 > 0 || refcount2 > 0) {
2151 *highest_cluster = i;
2154 if (refcount1 != refcount2) {
2155 /* Check if we're allowed to fix the mismatch */
2156 int *num_fixed = NULL;
2157 if (refcount1 == 0) {
2158 *rebuild = true;
2159 } else if (refcount1 > refcount2 && (fix & BDRV_FIX_LEAKS)) {
2160 num_fixed = &res->leaks_fixed;
2161 } else if (refcount1 < refcount2 && (fix & BDRV_FIX_ERRORS)) {
2162 num_fixed = &res->corruptions_fixed;
2165 fprintf(stderr, "%s cluster %" PRId64 " refcount=%" PRIu64
2166 " reference=%" PRIu64 "\n",
2167 num_fixed != NULL ? "Repairing" :
2168 refcount1 < refcount2 ? "ERROR" :
2169 "Leaked",
2170 i, refcount1, refcount2);
2172 if (num_fixed) {
2173 ret = update_refcount(bs, i << s->cluster_bits, 1,
2174 refcount_diff(refcount1, refcount2),
2175 refcount1 > refcount2,
2176 QCOW2_DISCARD_ALWAYS);
2177 if (ret >= 0) {
2178 (*num_fixed)++;
2179 continue;
2183 /* And if we couldn't, print an error */
2184 if (refcount1 < refcount2) {
2185 res->corruptions++;
2186 } else {
2187 res->leaks++;
2194 * Allocates clusters using an in-memory refcount table (IMRT) in contrast to
2195 * the on-disk refcount structures.
2197 * On input, *first_free_cluster tells where to start looking, and need not
2198 * actually be a free cluster; the returned offset will not be before that
2199 * cluster. On output, *first_free_cluster points to the first gap found, even
2200 * if that gap was too small to be used as the returned offset.
2202 * Note that *first_free_cluster is a cluster index whereas the return value is
2203 * an offset.
2205 static int64_t alloc_clusters_imrt(BlockDriverState *bs,
2206 int cluster_count,
2207 void **refcount_table,
2208 int64_t *imrt_nb_clusters,
2209 int64_t *first_free_cluster)
2211 BDRVQcow2State *s = bs->opaque;
2212 int64_t cluster = *first_free_cluster, i;
2213 bool first_gap = true;
2214 int contiguous_free_clusters;
2215 int ret;
2217 /* Starting at *first_free_cluster, find a range of at least cluster_count
2218 * continuously free clusters */
2219 for (contiguous_free_clusters = 0;
2220 cluster < *imrt_nb_clusters &&
2221 contiguous_free_clusters < cluster_count;
2222 cluster++)
2224 if (!s->get_refcount(*refcount_table, cluster)) {
2225 contiguous_free_clusters++;
2226 if (first_gap) {
2227 /* If this is the first free cluster found, update
2228 * *first_free_cluster accordingly */
2229 *first_free_cluster = cluster;
2230 first_gap = false;
2232 } else if (contiguous_free_clusters) {
2233 contiguous_free_clusters = 0;
2237 /* If contiguous_free_clusters is greater than zero, it contains the number
2238 * of continuously free clusters until the current cluster; the first free
2239 * cluster in the current "gap" is therefore
2240 * cluster - contiguous_free_clusters */
2242 /* If no such range could be found, grow the in-memory refcount table
2243 * accordingly to append free clusters at the end of the image */
2244 if (contiguous_free_clusters < cluster_count) {
2245 /* contiguous_free_clusters clusters are already empty at the image end;
2246 * we need cluster_count clusters; therefore, we have to allocate
2247 * cluster_count - contiguous_free_clusters new clusters at the end of
2248 * the image (which is the current value of cluster; note that cluster
2249 * may exceed old_imrt_nb_clusters if *first_free_cluster pointed beyond
2250 * the image end) */
2251 ret = realloc_refcount_array(s, refcount_table, imrt_nb_clusters,
2252 cluster + cluster_count
2253 - contiguous_free_clusters);
2254 if (ret < 0) {
2255 return ret;
2259 /* Go back to the first free cluster */
2260 cluster -= contiguous_free_clusters;
2261 for (i = 0; i < cluster_count; i++) {
2262 s->set_refcount(*refcount_table, cluster + i, 1);
2265 return cluster << s->cluster_bits;
2269 * Creates a new refcount structure based solely on the in-memory information
2270 * given through *refcount_table. All necessary allocations will be reflected
2271 * in that array.
2273 * On success, the old refcount structure is leaked (it will be covered by the
2274 * new refcount structure).
2276 static int rebuild_refcount_structure(BlockDriverState *bs,
2277 BdrvCheckResult *res,
2278 void **refcount_table,
2279 int64_t *nb_clusters)
2281 BDRVQcow2State *s = bs->opaque;
2282 int64_t first_free_cluster = 0, reftable_offset = -1, cluster = 0;
2283 int64_t refblock_offset, refblock_start, refblock_index;
2284 uint32_t reftable_size = 0;
2285 uint64_t *on_disk_reftable = NULL;
2286 void *on_disk_refblock;
2287 int ret = 0;
2288 struct {
2289 uint64_t reftable_offset;
2290 uint32_t reftable_clusters;
2291 } QEMU_PACKED reftable_offset_and_clusters;
2293 qcow2_cache_empty(bs, s->refcount_block_cache);
2295 write_refblocks:
2296 for (; cluster < *nb_clusters; cluster++) {
2297 if (!s->get_refcount(*refcount_table, cluster)) {
2298 continue;
2301 refblock_index = cluster >> s->refcount_block_bits;
2302 refblock_start = refblock_index << s->refcount_block_bits;
2304 /* Don't allocate a cluster in a refblock already written to disk */
2305 if (first_free_cluster < refblock_start) {
2306 first_free_cluster = refblock_start;
2308 refblock_offset = alloc_clusters_imrt(bs, 1, refcount_table,
2309 nb_clusters, &first_free_cluster);
2310 if (refblock_offset < 0) {
2311 fprintf(stderr, "ERROR allocating refblock: %s\n",
2312 strerror(-refblock_offset));
2313 res->check_errors++;
2314 ret = refblock_offset;
2315 goto fail;
2318 if (reftable_size <= refblock_index) {
2319 uint32_t old_reftable_size = reftable_size;
2320 uint64_t *new_on_disk_reftable;
2322 reftable_size = ROUND_UP((refblock_index + 1) * sizeof(uint64_t),
2323 s->cluster_size) / sizeof(uint64_t);
2324 new_on_disk_reftable = g_try_realloc(on_disk_reftable,
2325 reftable_size *
2326 sizeof(uint64_t));
2327 if (!new_on_disk_reftable) {
2328 res->check_errors++;
2329 ret = -ENOMEM;
2330 goto fail;
2332 on_disk_reftable = new_on_disk_reftable;
2334 memset(on_disk_reftable + old_reftable_size, 0,
2335 (reftable_size - old_reftable_size) * sizeof(uint64_t));
2337 /* The offset we have for the reftable is now no longer valid;
2338 * this will leak that range, but we can easily fix that by running
2339 * a leak-fixing check after this rebuild operation */
2340 reftable_offset = -1;
2341 } else {
2342 assert(on_disk_reftable);
2344 on_disk_reftable[refblock_index] = refblock_offset;
2346 /* If this is apparently the last refblock (for now), try to squeeze the
2347 * reftable in */
2348 if (refblock_index == (*nb_clusters - 1) >> s->refcount_block_bits &&
2349 reftable_offset < 0)
2351 uint64_t reftable_clusters = size_to_clusters(s, reftable_size *
2352 sizeof(uint64_t));
2353 reftable_offset = alloc_clusters_imrt(bs, reftable_clusters,
2354 refcount_table, nb_clusters,
2355 &first_free_cluster);
2356 if (reftable_offset < 0) {
2357 fprintf(stderr, "ERROR allocating reftable: %s\n",
2358 strerror(-reftable_offset));
2359 res->check_errors++;
2360 ret = reftable_offset;
2361 goto fail;
2365 ret = qcow2_pre_write_overlap_check(bs, 0, refblock_offset,
2366 s->cluster_size);
2367 if (ret < 0) {
2368 fprintf(stderr, "ERROR writing refblock: %s\n", strerror(-ret));
2369 goto fail;
2372 /* The size of *refcount_table is always cluster-aligned, therefore the
2373 * write operation will not overflow */
2374 on_disk_refblock = (void *)((char *) *refcount_table +
2375 refblock_index * s->cluster_size);
2377 ret = bdrv_write(bs->file, refblock_offset / BDRV_SECTOR_SIZE,
2378 on_disk_refblock, s->cluster_sectors);
2379 if (ret < 0) {
2380 fprintf(stderr, "ERROR writing refblock: %s\n", strerror(-ret));
2381 goto fail;
2384 /* Go to the end of this refblock */
2385 cluster = refblock_start + s->refcount_block_size - 1;
2388 if (reftable_offset < 0) {
2389 uint64_t post_refblock_start, reftable_clusters;
2391 post_refblock_start = ROUND_UP(*nb_clusters, s->refcount_block_size);
2392 reftable_clusters = size_to_clusters(s,
2393 reftable_size * sizeof(uint64_t));
2394 /* Not pretty but simple */
2395 if (first_free_cluster < post_refblock_start) {
2396 first_free_cluster = post_refblock_start;
2398 reftable_offset = alloc_clusters_imrt(bs, reftable_clusters,
2399 refcount_table, nb_clusters,
2400 &first_free_cluster);
2401 if (reftable_offset < 0) {
2402 fprintf(stderr, "ERROR allocating reftable: %s\n",
2403 strerror(-reftable_offset));
2404 res->check_errors++;
2405 ret = reftable_offset;
2406 goto fail;
2409 goto write_refblocks;
2412 for (refblock_index = 0; refblock_index < reftable_size; refblock_index++) {
2413 cpu_to_be64s(&on_disk_reftable[refblock_index]);
2416 ret = qcow2_pre_write_overlap_check(bs, 0, reftable_offset,
2417 reftable_size * sizeof(uint64_t));
2418 if (ret < 0) {
2419 fprintf(stderr, "ERROR writing reftable: %s\n", strerror(-ret));
2420 goto fail;
2423 assert(reftable_size < INT_MAX / sizeof(uint64_t));
2424 ret = bdrv_pwrite(bs->file, reftable_offset, on_disk_reftable,
2425 reftable_size * sizeof(uint64_t));
2426 if (ret < 0) {
2427 fprintf(stderr, "ERROR writing reftable: %s\n", strerror(-ret));
2428 goto fail;
2431 /* Enter new reftable into the image header */
2432 reftable_offset_and_clusters.reftable_offset = cpu_to_be64(reftable_offset);
2433 reftable_offset_and_clusters.reftable_clusters =
2434 cpu_to_be32(size_to_clusters(s, reftable_size * sizeof(uint64_t)));
2435 ret = bdrv_pwrite_sync(bs->file,
2436 offsetof(QCowHeader, refcount_table_offset),
2437 &reftable_offset_and_clusters,
2438 sizeof(reftable_offset_and_clusters));
2439 if (ret < 0) {
2440 fprintf(stderr, "ERROR setting reftable: %s\n", strerror(-ret));
2441 goto fail;
2444 for (refblock_index = 0; refblock_index < reftable_size; refblock_index++) {
2445 be64_to_cpus(&on_disk_reftable[refblock_index]);
2447 s->refcount_table = on_disk_reftable;
2448 s->refcount_table_offset = reftable_offset;
2449 s->refcount_table_size = reftable_size;
2450 update_max_refcount_table_index(s);
2452 return 0;
2454 fail:
2455 g_free(on_disk_reftable);
2456 return ret;
2460 * Checks an image for refcount consistency.
2462 * Returns 0 if no errors are found, the number of errors in case the image is
2463 * detected as corrupted, and -errno when an internal error occurred.
2465 int qcow2_check_refcounts(BlockDriverState *bs, BdrvCheckResult *res,
2466 BdrvCheckMode fix)
2468 BDRVQcow2State *s = bs->opaque;
2469 BdrvCheckResult pre_compare_res;
2470 int64_t size, highest_cluster, nb_clusters;
2471 void *refcount_table = NULL;
2472 bool rebuild = false;
2473 int ret;
2475 size = bdrv_getlength(bs->file->bs);
2476 if (size < 0) {
2477 res->check_errors++;
2478 return size;
2481 nb_clusters = size_to_clusters(s, size);
2482 if (nb_clusters > INT_MAX) {
2483 res->check_errors++;
2484 return -EFBIG;
2487 res->bfi.total_clusters =
2488 size_to_clusters(s, bs->total_sectors * BDRV_SECTOR_SIZE);
2490 ret = calculate_refcounts(bs, res, fix, &rebuild, &refcount_table,
2491 &nb_clusters);
2492 if (ret < 0) {
2493 goto fail;
2496 /* In case we don't need to rebuild the refcount structure (but want to fix
2497 * something), this function is immediately called again, in which case the
2498 * result should be ignored */
2499 pre_compare_res = *res;
2500 compare_refcounts(bs, res, 0, &rebuild, &highest_cluster, refcount_table,
2501 nb_clusters);
2503 if (rebuild && (fix & BDRV_FIX_ERRORS)) {
2504 BdrvCheckResult old_res = *res;
2505 int fresh_leaks = 0;
2507 fprintf(stderr, "Rebuilding refcount structure\n");
2508 ret = rebuild_refcount_structure(bs, res, &refcount_table,
2509 &nb_clusters);
2510 if (ret < 0) {
2511 goto fail;
2514 res->corruptions = 0;
2515 res->leaks = 0;
2517 /* Because the old reftable has been exchanged for a new one the
2518 * references have to be recalculated */
2519 rebuild = false;
2520 memset(refcount_table, 0, refcount_array_byte_size(s, nb_clusters));
2521 ret = calculate_refcounts(bs, res, 0, &rebuild, &refcount_table,
2522 &nb_clusters);
2523 if (ret < 0) {
2524 goto fail;
2527 if (fix & BDRV_FIX_LEAKS) {
2528 /* The old refcount structures are now leaked, fix it; the result
2529 * can be ignored, aside from leaks which were introduced by
2530 * rebuild_refcount_structure() that could not be fixed */
2531 BdrvCheckResult saved_res = *res;
2532 *res = (BdrvCheckResult){ 0 };
2534 compare_refcounts(bs, res, BDRV_FIX_LEAKS, &rebuild,
2535 &highest_cluster, refcount_table, nb_clusters);
2536 if (rebuild) {
2537 fprintf(stderr, "ERROR rebuilt refcount structure is still "
2538 "broken\n");
2541 /* Any leaks accounted for here were introduced by
2542 * rebuild_refcount_structure() because that function has created a
2543 * new refcount structure from scratch */
2544 fresh_leaks = res->leaks;
2545 *res = saved_res;
2548 if (res->corruptions < old_res.corruptions) {
2549 res->corruptions_fixed += old_res.corruptions - res->corruptions;
2551 if (res->leaks < old_res.leaks) {
2552 res->leaks_fixed += old_res.leaks - res->leaks;
2554 res->leaks += fresh_leaks;
2555 } else if (fix) {
2556 if (rebuild) {
2557 fprintf(stderr, "ERROR need to rebuild refcount structures\n");
2558 res->check_errors++;
2559 ret = -EIO;
2560 goto fail;
2563 if (res->leaks || res->corruptions) {
2564 *res = pre_compare_res;
2565 compare_refcounts(bs, res, fix, &rebuild, &highest_cluster,
2566 refcount_table, nb_clusters);
2570 /* check OFLAG_COPIED */
2571 ret = check_oflag_copied(bs, res, fix);
2572 if (ret < 0) {
2573 goto fail;
2576 res->image_end_offset = (highest_cluster + 1) * s->cluster_size;
2577 ret = 0;
2579 fail:
2580 g_free(refcount_table);
2582 return ret;
2585 #define overlaps_with(ofs, sz) \
2586 ranges_overlap(offset, size, ofs, sz)
2589 * Checks if the given offset into the image file is actually free to use by
2590 * looking for overlaps with important metadata sections (L1/L2 tables etc.),
2591 * i.e. a sanity check without relying on the refcount tables.
2593 * The ign parameter specifies what checks not to perform (being a bitmask of
2594 * QCow2MetadataOverlap values), i.e., what sections to ignore.
2596 * Returns:
2597 * - 0 if writing to this offset will not affect the mentioned metadata
2598 * - a positive QCow2MetadataOverlap value indicating one overlapping section
2599 * - a negative value (-errno) indicating an error while performing a check,
2600 * e.g. when bdrv_read failed on QCOW2_OL_INACTIVE_L2
2602 int qcow2_check_metadata_overlap(BlockDriverState *bs, int ign, int64_t offset,
2603 int64_t size)
2605 BDRVQcow2State *s = bs->opaque;
2606 int chk = s->overlap_check & ~ign;
2607 int i, j;
2609 if (!size) {
2610 return 0;
2613 if (chk & QCOW2_OL_MAIN_HEADER) {
2614 if (offset < s->cluster_size) {
2615 return QCOW2_OL_MAIN_HEADER;
2619 /* align range to test to cluster boundaries */
2620 size = ROUND_UP(offset_into_cluster(s, offset) + size, s->cluster_size);
2621 offset = start_of_cluster(s, offset);
2623 if ((chk & QCOW2_OL_ACTIVE_L1) && s->l1_size) {
2624 if (overlaps_with(s->l1_table_offset, s->l1_size * sizeof(uint64_t))) {
2625 return QCOW2_OL_ACTIVE_L1;
2629 if ((chk & QCOW2_OL_REFCOUNT_TABLE) && s->refcount_table_size) {
2630 if (overlaps_with(s->refcount_table_offset,
2631 s->refcount_table_size * sizeof(uint64_t))) {
2632 return QCOW2_OL_REFCOUNT_TABLE;
2636 if ((chk & QCOW2_OL_SNAPSHOT_TABLE) && s->snapshots_size) {
2637 if (overlaps_with(s->snapshots_offset, s->snapshots_size)) {
2638 return QCOW2_OL_SNAPSHOT_TABLE;
2642 if ((chk & QCOW2_OL_INACTIVE_L1) && s->snapshots) {
2643 for (i = 0; i < s->nb_snapshots; i++) {
2644 if (s->snapshots[i].l1_size &&
2645 overlaps_with(s->snapshots[i].l1_table_offset,
2646 s->snapshots[i].l1_size * sizeof(uint64_t))) {
2647 return QCOW2_OL_INACTIVE_L1;
2652 if ((chk & QCOW2_OL_ACTIVE_L2) && s->l1_table) {
2653 for (i = 0; i < s->l1_size; i++) {
2654 if ((s->l1_table[i] & L1E_OFFSET_MASK) &&
2655 overlaps_with(s->l1_table[i] & L1E_OFFSET_MASK,
2656 s->cluster_size)) {
2657 return QCOW2_OL_ACTIVE_L2;
2662 if ((chk & QCOW2_OL_REFCOUNT_BLOCK) && s->refcount_table) {
2663 unsigned last_entry = s->max_refcount_table_index;
2664 assert(last_entry < s->refcount_table_size);
2665 assert(last_entry + 1 == s->refcount_table_size ||
2666 (s->refcount_table[last_entry + 1] & REFT_OFFSET_MASK) == 0);
2667 for (i = 0; i <= last_entry; i++) {
2668 if ((s->refcount_table[i] & REFT_OFFSET_MASK) &&
2669 overlaps_with(s->refcount_table[i] & REFT_OFFSET_MASK,
2670 s->cluster_size)) {
2671 return QCOW2_OL_REFCOUNT_BLOCK;
2676 if ((chk & QCOW2_OL_INACTIVE_L2) && s->snapshots) {
2677 for (i = 0; i < s->nb_snapshots; i++) {
2678 uint64_t l1_ofs = s->snapshots[i].l1_table_offset;
2679 uint32_t l1_sz = s->snapshots[i].l1_size;
2680 uint64_t l1_sz2 = l1_sz * sizeof(uint64_t);
2681 uint64_t *l1;
2682 int ret;
2684 ret = qcow2_validate_table(bs, l1_ofs, l1_sz, sizeof(uint64_t),
2685 QCOW_MAX_L1_SIZE, "", NULL);
2686 if (ret < 0) {
2687 return ret;
2690 l1 = g_try_malloc(l1_sz2);
2692 if (l1_sz2 && l1 == NULL) {
2693 return -ENOMEM;
2696 ret = bdrv_pread(bs->file, l1_ofs, l1, l1_sz2);
2697 if (ret < 0) {
2698 g_free(l1);
2699 return ret;
2702 for (j = 0; j < l1_sz; j++) {
2703 uint64_t l2_ofs = be64_to_cpu(l1[j]) & L1E_OFFSET_MASK;
2704 if (l2_ofs && overlaps_with(l2_ofs, s->cluster_size)) {
2705 g_free(l1);
2706 return QCOW2_OL_INACTIVE_L2;
2710 g_free(l1);
2714 if ((chk & QCOW2_OL_BITMAP_DIRECTORY) &&
2715 (s->autoclear_features & QCOW2_AUTOCLEAR_BITMAPS))
2717 if (overlaps_with(s->bitmap_directory_offset,
2718 s->bitmap_directory_size))
2720 return QCOW2_OL_BITMAP_DIRECTORY;
2724 return 0;
2727 static const char *metadata_ol_names[] = {
2728 [QCOW2_OL_MAIN_HEADER_BITNR] = "qcow2_header",
2729 [QCOW2_OL_ACTIVE_L1_BITNR] = "active L1 table",
2730 [QCOW2_OL_ACTIVE_L2_BITNR] = "active L2 table",
2731 [QCOW2_OL_REFCOUNT_TABLE_BITNR] = "refcount table",
2732 [QCOW2_OL_REFCOUNT_BLOCK_BITNR] = "refcount block",
2733 [QCOW2_OL_SNAPSHOT_TABLE_BITNR] = "snapshot table",
2734 [QCOW2_OL_INACTIVE_L1_BITNR] = "inactive L1 table",
2735 [QCOW2_OL_INACTIVE_L2_BITNR] = "inactive L2 table",
2736 [QCOW2_OL_BITMAP_DIRECTORY_BITNR] = "bitmap directory",
2738 QEMU_BUILD_BUG_ON(QCOW2_OL_MAX_BITNR != ARRAY_SIZE(metadata_ol_names));
2741 * First performs a check for metadata overlaps (through
2742 * qcow2_check_metadata_overlap); if that fails with a negative value (error
2743 * while performing a check), that value is returned. If an impending overlap
2744 * is detected, the BDS will be made unusable, the qcow2 file marked corrupt
2745 * and -EIO returned.
2747 * Returns 0 if there were neither overlaps nor errors while checking for
2748 * overlaps; or a negative value (-errno) on error.
2750 int qcow2_pre_write_overlap_check(BlockDriverState *bs, int ign, int64_t offset,
2751 int64_t size)
2753 int ret = qcow2_check_metadata_overlap(bs, ign, offset, size);
2755 if (ret < 0) {
2756 return ret;
2757 } else if (ret > 0) {
2758 int metadata_ol_bitnr = ctz32(ret);
2759 assert(metadata_ol_bitnr < QCOW2_OL_MAX_BITNR);
2761 qcow2_signal_corruption(bs, true, offset, size, "Preventing invalid "
2762 "write on metadata (overlaps with %s)",
2763 metadata_ol_names[metadata_ol_bitnr]);
2764 return -EIO;
2767 return 0;
2770 /* A pointer to a function of this type is given to walk_over_reftable(). That
2771 * function will create refblocks and pass them to a RefblockFinishOp once they
2772 * are completed (@refblock). @refblock_empty is set if the refblock is
2773 * completely empty.
2775 * Along with the refblock, a corresponding reftable entry is passed, in the
2776 * reftable @reftable (which may be reallocated) at @reftable_index.
2778 * @allocated should be set to true if a new cluster has been allocated.
2780 typedef int (RefblockFinishOp)(BlockDriverState *bs, uint64_t **reftable,
2781 uint64_t reftable_index, uint64_t *reftable_size,
2782 void *refblock, bool refblock_empty,
2783 bool *allocated, Error **errp);
2786 * This "operation" for walk_over_reftable() allocates the refblock on disk (if
2787 * it is not empty) and inserts its offset into the new reftable. The size of
2788 * this new reftable is increased as required.
2790 static int alloc_refblock(BlockDriverState *bs, uint64_t **reftable,
2791 uint64_t reftable_index, uint64_t *reftable_size,
2792 void *refblock, bool refblock_empty, bool *allocated,
2793 Error **errp)
2795 BDRVQcow2State *s = bs->opaque;
2796 int64_t offset;
2798 if (!refblock_empty && reftable_index >= *reftable_size) {
2799 uint64_t *new_reftable;
2800 uint64_t new_reftable_size;
2802 new_reftable_size = ROUND_UP(reftable_index + 1,
2803 s->cluster_size / sizeof(uint64_t));
2804 if (new_reftable_size > QCOW_MAX_REFTABLE_SIZE / sizeof(uint64_t)) {
2805 error_setg(errp,
2806 "This operation would make the refcount table grow "
2807 "beyond the maximum size supported by QEMU, aborting");
2808 return -ENOTSUP;
2811 new_reftable = g_try_realloc(*reftable, new_reftable_size *
2812 sizeof(uint64_t));
2813 if (!new_reftable) {
2814 error_setg(errp, "Failed to increase reftable buffer size");
2815 return -ENOMEM;
2818 memset(new_reftable + *reftable_size, 0,
2819 (new_reftable_size - *reftable_size) * sizeof(uint64_t));
2821 *reftable = new_reftable;
2822 *reftable_size = new_reftable_size;
2825 if (!refblock_empty && !(*reftable)[reftable_index]) {
2826 offset = qcow2_alloc_clusters(bs, s->cluster_size);
2827 if (offset < 0) {
2828 error_setg_errno(errp, -offset, "Failed to allocate refblock");
2829 return offset;
2831 (*reftable)[reftable_index] = offset;
2832 *allocated = true;
2835 return 0;
2839 * This "operation" for walk_over_reftable() writes the refblock to disk at the
2840 * offset specified by the new reftable's entry. It does not modify the new
2841 * reftable or change any refcounts.
2843 static int flush_refblock(BlockDriverState *bs, uint64_t **reftable,
2844 uint64_t reftable_index, uint64_t *reftable_size,
2845 void *refblock, bool refblock_empty, bool *allocated,
2846 Error **errp)
2848 BDRVQcow2State *s = bs->opaque;
2849 int64_t offset;
2850 int ret;
2852 if (reftable_index < *reftable_size && (*reftable)[reftable_index]) {
2853 offset = (*reftable)[reftable_index];
2855 ret = qcow2_pre_write_overlap_check(bs, 0, offset, s->cluster_size);
2856 if (ret < 0) {
2857 error_setg_errno(errp, -ret, "Overlap check failed");
2858 return ret;
2861 ret = bdrv_pwrite(bs->file, offset, refblock, s->cluster_size);
2862 if (ret < 0) {
2863 error_setg_errno(errp, -ret, "Failed to write refblock");
2864 return ret;
2866 } else {
2867 assert(refblock_empty);
2870 return 0;
2874 * This function walks over the existing reftable and every referenced refblock;
2875 * if @new_set_refcount is non-NULL, it is called for every refcount entry to
2876 * create an equal new entry in the passed @new_refblock. Once that
2877 * @new_refblock is completely filled, @operation will be called.
2879 * @status_cb and @cb_opaque are used for the amend operation's status callback.
2880 * @index is the index of the walk_over_reftable() calls and @total is the total
2881 * number of walk_over_reftable() calls per amend operation. Both are used for
2882 * calculating the parameters for the status callback.
2884 * @allocated is set to true if a new cluster has been allocated.
2886 static int walk_over_reftable(BlockDriverState *bs, uint64_t **new_reftable,
2887 uint64_t *new_reftable_index,
2888 uint64_t *new_reftable_size,
2889 void *new_refblock, int new_refblock_size,
2890 int new_refcount_bits,
2891 RefblockFinishOp *operation, bool *allocated,
2892 Qcow2SetRefcountFunc *new_set_refcount,
2893 BlockDriverAmendStatusCB *status_cb,
2894 void *cb_opaque, int index, int total,
2895 Error **errp)
2897 BDRVQcow2State *s = bs->opaque;
2898 uint64_t reftable_index;
2899 bool new_refblock_empty = true;
2900 int refblock_index;
2901 int new_refblock_index = 0;
2902 int ret;
2904 for (reftable_index = 0; reftable_index < s->refcount_table_size;
2905 reftable_index++)
2907 uint64_t refblock_offset = s->refcount_table[reftable_index]
2908 & REFT_OFFSET_MASK;
2910 status_cb(bs, (uint64_t)index * s->refcount_table_size + reftable_index,
2911 (uint64_t)total * s->refcount_table_size, cb_opaque);
2913 if (refblock_offset) {
2914 void *refblock;
2916 if (offset_into_cluster(s, refblock_offset)) {
2917 qcow2_signal_corruption(bs, true, -1, -1, "Refblock offset %#"
2918 PRIx64 " unaligned (reftable index: %#"
2919 PRIx64 ")", refblock_offset,
2920 reftable_index);
2921 error_setg(errp,
2922 "Image is corrupt (unaligned refblock offset)");
2923 return -EIO;
2926 ret = qcow2_cache_get(bs, s->refcount_block_cache, refblock_offset,
2927 &refblock);
2928 if (ret < 0) {
2929 error_setg_errno(errp, -ret, "Failed to retrieve refblock");
2930 return ret;
2933 for (refblock_index = 0; refblock_index < s->refcount_block_size;
2934 refblock_index++)
2936 uint64_t refcount;
2938 if (new_refblock_index >= new_refblock_size) {
2939 /* new_refblock is now complete */
2940 ret = operation(bs, new_reftable, *new_reftable_index,
2941 new_reftable_size, new_refblock,
2942 new_refblock_empty, allocated, errp);
2943 if (ret < 0) {
2944 qcow2_cache_put(s->refcount_block_cache, &refblock);
2945 return ret;
2948 (*new_reftable_index)++;
2949 new_refblock_index = 0;
2950 new_refblock_empty = true;
2953 refcount = s->get_refcount(refblock, refblock_index);
2954 if (new_refcount_bits < 64 && refcount >> new_refcount_bits) {
2955 uint64_t offset;
2957 qcow2_cache_put(s->refcount_block_cache, &refblock);
2959 offset = ((reftable_index << s->refcount_block_bits)
2960 + refblock_index) << s->cluster_bits;
2962 error_setg(errp, "Cannot decrease refcount entry width to "
2963 "%i bits: Cluster at offset %#" PRIx64 " has a "
2964 "refcount of %" PRIu64, new_refcount_bits,
2965 offset, refcount);
2966 return -EINVAL;
2969 if (new_set_refcount) {
2970 new_set_refcount(new_refblock, new_refblock_index++,
2971 refcount);
2972 } else {
2973 new_refblock_index++;
2975 new_refblock_empty = new_refblock_empty && refcount == 0;
2978 qcow2_cache_put(s->refcount_block_cache, &refblock);
2979 } else {
2980 /* No refblock means every refcount is 0 */
2981 for (refblock_index = 0; refblock_index < s->refcount_block_size;
2982 refblock_index++)
2984 if (new_refblock_index >= new_refblock_size) {
2985 /* new_refblock is now complete */
2986 ret = operation(bs, new_reftable, *new_reftable_index,
2987 new_reftable_size, new_refblock,
2988 new_refblock_empty, allocated, errp);
2989 if (ret < 0) {
2990 return ret;
2993 (*new_reftable_index)++;
2994 new_refblock_index = 0;
2995 new_refblock_empty = true;
2998 if (new_set_refcount) {
2999 new_set_refcount(new_refblock, new_refblock_index++, 0);
3000 } else {
3001 new_refblock_index++;
3007 if (new_refblock_index > 0) {
3008 /* Complete the potentially existing partially filled final refblock */
3009 if (new_set_refcount) {
3010 for (; new_refblock_index < new_refblock_size;
3011 new_refblock_index++)
3013 new_set_refcount(new_refblock, new_refblock_index, 0);
3017 ret = operation(bs, new_reftable, *new_reftable_index,
3018 new_reftable_size, new_refblock, new_refblock_empty,
3019 allocated, errp);
3020 if (ret < 0) {
3021 return ret;
3024 (*new_reftable_index)++;
3027 status_cb(bs, (uint64_t)(index + 1) * s->refcount_table_size,
3028 (uint64_t)total * s->refcount_table_size, cb_opaque);
3030 return 0;
3033 int qcow2_change_refcount_order(BlockDriverState *bs, int refcount_order,
3034 BlockDriverAmendStatusCB *status_cb,
3035 void *cb_opaque, Error **errp)
3037 BDRVQcow2State *s = bs->opaque;
3038 Qcow2GetRefcountFunc *new_get_refcount;
3039 Qcow2SetRefcountFunc *new_set_refcount;
3040 void *new_refblock = qemu_blockalign(bs->file->bs, s->cluster_size);
3041 uint64_t *new_reftable = NULL, new_reftable_size = 0;
3042 uint64_t *old_reftable, old_reftable_size, old_reftable_offset;
3043 uint64_t new_reftable_index = 0;
3044 uint64_t i;
3045 int64_t new_reftable_offset = 0, allocated_reftable_size = 0;
3046 int new_refblock_size, new_refcount_bits = 1 << refcount_order;
3047 int old_refcount_order;
3048 int walk_index = 0;
3049 int ret;
3050 bool new_allocation;
3052 assert(s->qcow_version >= 3);
3053 assert(refcount_order >= 0 && refcount_order <= 6);
3055 /* see qcow2_open() */
3056 new_refblock_size = 1 << (s->cluster_bits - (refcount_order - 3));
3058 new_get_refcount = get_refcount_funcs[refcount_order];
3059 new_set_refcount = set_refcount_funcs[refcount_order];
3062 do {
3063 int total_walks;
3065 new_allocation = false;
3067 /* At least we have to do this walk and the one which writes the
3068 * refblocks; also, at least we have to do this loop here at least
3069 * twice (normally), first to do the allocations, and second to
3070 * determine that everything is correctly allocated, this then makes
3071 * three walks in total */
3072 total_walks = MAX(walk_index + 2, 3);
3074 /* First, allocate the structures so they are present in the refcount
3075 * structures */
3076 ret = walk_over_reftable(bs, &new_reftable, &new_reftable_index,
3077 &new_reftable_size, NULL, new_refblock_size,
3078 new_refcount_bits, &alloc_refblock,
3079 &new_allocation, NULL, status_cb, cb_opaque,
3080 walk_index++, total_walks, errp);
3081 if (ret < 0) {
3082 goto done;
3085 new_reftable_index = 0;
3087 if (new_allocation) {
3088 if (new_reftable_offset) {
3089 qcow2_free_clusters(bs, new_reftable_offset,
3090 allocated_reftable_size * sizeof(uint64_t),
3091 QCOW2_DISCARD_NEVER);
3094 new_reftable_offset = qcow2_alloc_clusters(bs, new_reftable_size *
3095 sizeof(uint64_t));
3096 if (new_reftable_offset < 0) {
3097 error_setg_errno(errp, -new_reftable_offset,
3098 "Failed to allocate the new reftable");
3099 ret = new_reftable_offset;
3100 goto done;
3102 allocated_reftable_size = new_reftable_size;
3104 } while (new_allocation);
3106 /* Second, write the new refblocks */
3107 ret = walk_over_reftable(bs, &new_reftable, &new_reftable_index,
3108 &new_reftable_size, new_refblock,
3109 new_refblock_size, new_refcount_bits,
3110 &flush_refblock, &new_allocation, new_set_refcount,
3111 status_cb, cb_opaque, walk_index, walk_index + 1,
3112 errp);
3113 if (ret < 0) {
3114 goto done;
3116 assert(!new_allocation);
3119 /* Write the new reftable */
3120 ret = qcow2_pre_write_overlap_check(bs, 0, new_reftable_offset,
3121 new_reftable_size * sizeof(uint64_t));
3122 if (ret < 0) {
3123 error_setg_errno(errp, -ret, "Overlap check failed");
3124 goto done;
3127 for (i = 0; i < new_reftable_size; i++) {
3128 cpu_to_be64s(&new_reftable[i]);
3131 ret = bdrv_pwrite(bs->file, new_reftable_offset, new_reftable,
3132 new_reftable_size * sizeof(uint64_t));
3134 for (i = 0; i < new_reftable_size; i++) {
3135 be64_to_cpus(&new_reftable[i]);
3138 if (ret < 0) {
3139 error_setg_errno(errp, -ret, "Failed to write the new reftable");
3140 goto done;
3144 /* Empty the refcount cache */
3145 ret = qcow2_cache_flush(bs, s->refcount_block_cache);
3146 if (ret < 0) {
3147 error_setg_errno(errp, -ret, "Failed to flush the refblock cache");
3148 goto done;
3151 /* Update the image header to point to the new reftable; this only updates
3152 * the fields which are relevant to qcow2_update_header(); other fields
3153 * such as s->refcount_table or s->refcount_bits stay stale for now
3154 * (because we have to restore everything if qcow2_update_header() fails) */
3155 old_refcount_order = s->refcount_order;
3156 old_reftable_size = s->refcount_table_size;
3157 old_reftable_offset = s->refcount_table_offset;
3159 s->refcount_order = refcount_order;
3160 s->refcount_table_size = new_reftable_size;
3161 s->refcount_table_offset = new_reftable_offset;
3163 ret = qcow2_update_header(bs);
3164 if (ret < 0) {
3165 s->refcount_order = old_refcount_order;
3166 s->refcount_table_size = old_reftable_size;
3167 s->refcount_table_offset = old_reftable_offset;
3168 error_setg_errno(errp, -ret, "Failed to update the qcow2 header");
3169 goto done;
3172 /* Now update the rest of the in-memory information */
3173 old_reftable = s->refcount_table;
3174 s->refcount_table = new_reftable;
3175 update_max_refcount_table_index(s);
3177 s->refcount_bits = 1 << refcount_order;
3178 s->refcount_max = UINT64_C(1) << (s->refcount_bits - 1);
3179 s->refcount_max += s->refcount_max - 1;
3181 s->refcount_block_bits = s->cluster_bits - (refcount_order - 3);
3182 s->refcount_block_size = 1 << s->refcount_block_bits;
3184 s->get_refcount = new_get_refcount;
3185 s->set_refcount = new_set_refcount;
3187 /* For cleaning up all old refblocks and the old reftable below the "done"
3188 * label */
3189 new_reftable = old_reftable;
3190 new_reftable_size = old_reftable_size;
3191 new_reftable_offset = old_reftable_offset;
3193 done:
3194 if (new_reftable) {
3195 /* On success, new_reftable actually points to the old reftable (and
3196 * new_reftable_size is the old reftable's size); but that is just
3197 * fine */
3198 for (i = 0; i < new_reftable_size; i++) {
3199 uint64_t offset = new_reftable[i] & REFT_OFFSET_MASK;
3200 if (offset) {
3201 qcow2_free_clusters(bs, offset, s->cluster_size,
3202 QCOW2_DISCARD_OTHER);
3205 g_free(new_reftable);
3207 if (new_reftable_offset > 0) {
3208 qcow2_free_clusters(bs, new_reftable_offset,
3209 new_reftable_size * sizeof(uint64_t),
3210 QCOW2_DISCARD_OTHER);
3214 qemu_vfree(new_refblock);
3215 return ret;
3218 static int64_t get_refblock_offset(BlockDriverState *bs, uint64_t offset)
3220 BDRVQcow2State *s = bs->opaque;
3221 uint32_t index = offset_to_reftable_index(s, offset);
3222 int64_t covering_refblock_offset = 0;
3224 if (index < s->refcount_table_size) {
3225 covering_refblock_offset = s->refcount_table[index] & REFT_OFFSET_MASK;
3227 if (!covering_refblock_offset) {
3228 qcow2_signal_corruption(bs, true, -1, -1, "Refblock at %#" PRIx64 " is "
3229 "not covered by the refcount structures",
3230 offset);
3231 return -EIO;
3234 return covering_refblock_offset;
3237 static int qcow2_discard_refcount_block(BlockDriverState *bs,
3238 uint64_t discard_block_offs)
3240 BDRVQcow2State *s = bs->opaque;
3241 int64_t refblock_offs;
3242 uint64_t cluster_index = discard_block_offs >> s->cluster_bits;
3243 uint32_t block_index = cluster_index & (s->refcount_block_size - 1);
3244 void *refblock;
3245 int ret;
3247 refblock_offs = get_refblock_offset(bs, discard_block_offs);
3248 if (refblock_offs < 0) {
3249 return refblock_offs;
3252 assert(discard_block_offs != 0);
3254 ret = qcow2_cache_get(bs, s->refcount_block_cache, refblock_offs,
3255 &refblock);
3256 if (ret < 0) {
3257 return ret;
3260 if (s->get_refcount(refblock, block_index) != 1) {
3261 qcow2_signal_corruption(bs, true, -1, -1, "Invalid refcount:"
3262 " refblock offset %#" PRIx64
3263 ", reftable index %u"
3264 ", block offset %#" PRIx64
3265 ", refcount %#" PRIx64,
3266 refblock_offs,
3267 offset_to_reftable_index(s, discard_block_offs),
3268 discard_block_offs,
3269 s->get_refcount(refblock, block_index));
3270 qcow2_cache_put(s->refcount_block_cache, &refblock);
3271 return -EINVAL;
3273 s->set_refcount(refblock, block_index, 0);
3275 qcow2_cache_entry_mark_dirty(s->refcount_block_cache, refblock);
3277 qcow2_cache_put(s->refcount_block_cache, &refblock);
3279 if (cluster_index < s->free_cluster_index) {
3280 s->free_cluster_index = cluster_index;
3283 refblock = qcow2_cache_is_table_offset(s->refcount_block_cache,
3284 discard_block_offs);
3285 if (refblock) {
3286 /* discard refblock from the cache if refblock is cached */
3287 qcow2_cache_discard(s->refcount_block_cache, refblock);
3289 update_refcount_discard(bs, discard_block_offs, s->cluster_size);
3291 return 0;
3294 int qcow2_shrink_reftable(BlockDriverState *bs)
3296 BDRVQcow2State *s = bs->opaque;
3297 uint64_t *reftable_tmp =
3298 g_malloc(s->refcount_table_size * sizeof(uint64_t));
3299 int i, ret;
3301 for (i = 0; i < s->refcount_table_size; i++) {
3302 int64_t refblock_offs = s->refcount_table[i] & REFT_OFFSET_MASK;
3303 void *refblock;
3304 bool unused_block;
3306 if (refblock_offs == 0) {
3307 reftable_tmp[i] = 0;
3308 continue;
3310 ret = qcow2_cache_get(bs, s->refcount_block_cache, refblock_offs,
3311 &refblock);
3312 if (ret < 0) {
3313 goto out;
3316 /* the refblock has own reference */
3317 if (i == offset_to_reftable_index(s, refblock_offs)) {
3318 uint64_t block_index = (refblock_offs >> s->cluster_bits) &
3319 (s->refcount_block_size - 1);
3320 uint64_t refcount = s->get_refcount(refblock, block_index);
3322 s->set_refcount(refblock, block_index, 0);
3324 unused_block = buffer_is_zero(refblock, s->cluster_size);
3326 s->set_refcount(refblock, block_index, refcount);
3327 } else {
3328 unused_block = buffer_is_zero(refblock, s->cluster_size);
3330 qcow2_cache_put(s->refcount_block_cache, &refblock);
3332 reftable_tmp[i] = unused_block ? 0 : cpu_to_be64(s->refcount_table[i]);
3335 ret = bdrv_pwrite_sync(bs->file, s->refcount_table_offset, reftable_tmp,
3336 s->refcount_table_size * sizeof(uint64_t));
3338 * If the write in the reftable failed the image may contain a partially
3339 * overwritten reftable. In this case it would be better to clear the
3340 * reftable in memory to avoid possible image corruption.
3342 for (i = 0; i < s->refcount_table_size; i++) {
3343 if (s->refcount_table[i] && !reftable_tmp[i]) {
3344 if (ret == 0) {
3345 ret = qcow2_discard_refcount_block(bs, s->refcount_table[i] &
3346 REFT_OFFSET_MASK);
3348 s->refcount_table[i] = 0;
3352 if (!s->cache_discards) {
3353 qcow2_process_discards(bs, ret);
3356 out:
3357 g_free(reftable_tmp);
3358 return ret;
3361 int64_t qcow2_get_last_cluster(BlockDriverState *bs, int64_t size)
3363 BDRVQcow2State *s = bs->opaque;
3364 int64_t i;
3366 for (i = size_to_clusters(s, size) - 1; i >= 0; i--) {
3367 uint64_t refcount;
3368 int ret = qcow2_get_refcount(bs, i, &refcount);
3369 if (ret < 0) {
3370 fprintf(stderr, "Can't get refcount for cluster %" PRId64 ": %s\n",
3371 i, strerror(-ret));
3372 return ret;
3374 if (refcount > 0) {
3375 return i;
3378 qcow2_signal_corruption(bs, true, -1, -1,
3379 "There are no references in the refcount table.");
3380 return -EIO;