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
25 #include "qemu/osdep.h"
28 #include "qapi/error.h"
29 #include "qemu-common.h"
30 #include "block/block_int.h"
31 #include "block/qcow2.h"
32 #include "qemu/bswap.h"
35 int qcow2_grow_l1_table(BlockDriverState *bs, uint64_t min_size,
38 BDRVQcow2State *s = bs->opaque;
39 int new_l1_size2, ret, i;
40 uint64_t *new_l1_table;
41 int64_t old_l1_table_offset, old_l1_size;
42 int64_t new_l1_table_offset, new_l1_size;
45 if (min_size <= s->l1_size)
48 /* Do a sanity check on min_size before trying to calculate new_l1_size
49 * (this prevents overflows during the while loop for the calculation of
51 if (min_size > INT_MAX / sizeof(uint64_t)) {
56 new_l1_size = min_size;
58 /* Bump size up to reduce the number of times we have to grow */
59 new_l1_size = s->l1_size;
60 if (new_l1_size == 0) {
63 while (min_size > new_l1_size) {
64 new_l1_size = (new_l1_size * 3 + 1) / 2;
68 if (new_l1_size > INT_MAX / sizeof(uint64_t)) {
73 fprintf(stderr, "grow l1_table from %d to %" PRId64 "\n",
74 s->l1_size, new_l1_size);
77 new_l1_size2 = sizeof(uint64_t) * new_l1_size;
78 new_l1_table = qemu_try_blockalign(bs->file->bs,
79 align_offset(new_l1_size2, 512));
80 if (new_l1_table == NULL) {
83 memset(new_l1_table, 0, align_offset(new_l1_size2, 512));
85 memcpy(new_l1_table, s->l1_table, s->l1_size * sizeof(uint64_t));
87 /* write new table (align to cluster) */
88 BLKDBG_EVENT(bs->file, BLKDBG_L1_GROW_ALLOC_TABLE);
89 new_l1_table_offset = qcow2_alloc_clusters(bs, new_l1_size2);
90 if (new_l1_table_offset < 0) {
91 qemu_vfree(new_l1_table);
92 return new_l1_table_offset;
95 ret = qcow2_cache_flush(bs, s->refcount_block_cache);
100 /* the L1 position has not yet been updated, so these clusters must
101 * indeed be completely free */
102 ret = qcow2_pre_write_overlap_check(bs, 0, new_l1_table_offset,
108 BLKDBG_EVENT(bs->file, BLKDBG_L1_GROW_WRITE_TABLE);
109 for(i = 0; i < s->l1_size; i++)
110 new_l1_table[i] = cpu_to_be64(new_l1_table[i]);
111 ret = bdrv_pwrite_sync(bs->file, new_l1_table_offset,
112 new_l1_table, new_l1_size2);
115 for(i = 0; i < s->l1_size; i++)
116 new_l1_table[i] = be64_to_cpu(new_l1_table[i]);
119 BLKDBG_EVENT(bs->file, BLKDBG_L1_GROW_ACTIVATE_TABLE);
120 cpu_to_be32w((uint32_t*)data, new_l1_size);
121 stq_be_p(data + 4, new_l1_table_offset);
122 ret = bdrv_pwrite_sync(bs->file, offsetof(QCowHeader, l1_size),
127 qemu_vfree(s->l1_table);
128 old_l1_table_offset = s->l1_table_offset;
129 s->l1_table_offset = new_l1_table_offset;
130 s->l1_table = new_l1_table;
131 old_l1_size = s->l1_size;
132 s->l1_size = new_l1_size;
133 qcow2_free_clusters(bs, old_l1_table_offset, old_l1_size * sizeof(uint64_t),
134 QCOW2_DISCARD_OTHER);
137 qemu_vfree(new_l1_table);
138 qcow2_free_clusters(bs, new_l1_table_offset, new_l1_size2,
139 QCOW2_DISCARD_OTHER);
146 * Loads a L2 table into memory. If the table is in the cache, the cache
147 * is used; otherwise the L2 table is loaded from the image file.
149 * Returns a pointer to the L2 table on success, or NULL if the read from
150 * the image file failed.
153 static int l2_load(BlockDriverState *bs, uint64_t l2_offset,
156 BDRVQcow2State *s = bs->opaque;
158 return qcow2_cache_get(bs, s->l2_table_cache, l2_offset,
163 * Writes one sector of the L1 table to the disk (can't update single entries
164 * and we really don't want bdrv_pread to perform a read-modify-write)
166 #define L1_ENTRIES_PER_SECTOR (512 / 8)
167 int qcow2_write_l1_entry(BlockDriverState *bs, int l1_index)
169 BDRVQcow2State *s = bs->opaque;
170 uint64_t buf[L1_ENTRIES_PER_SECTOR] = { 0 };
174 l1_start_index = l1_index & ~(L1_ENTRIES_PER_SECTOR - 1);
175 for (i = 0; i < L1_ENTRIES_PER_SECTOR && l1_start_index + i < s->l1_size;
178 buf[i] = cpu_to_be64(s->l1_table[l1_start_index + i]);
181 ret = qcow2_pre_write_overlap_check(bs, QCOW2_OL_ACTIVE_L1,
182 s->l1_table_offset + 8 * l1_start_index, sizeof(buf));
187 BLKDBG_EVENT(bs->file, BLKDBG_L1_UPDATE);
188 ret = bdrv_pwrite_sync(bs->file,
189 s->l1_table_offset + 8 * l1_start_index,
201 * Allocate a new l2 entry in the file. If l1_index points to an already
202 * used entry in the L2 table (i.e. we are doing a copy on write for the L2
203 * table) copy the contents of the old L2 table into the newly allocated one.
204 * Otherwise the new table is initialized with zeros.
208 static int l2_allocate(BlockDriverState *bs, int l1_index, uint64_t **table)
210 BDRVQcow2State *s = bs->opaque;
211 uint64_t old_l2_offset;
212 uint64_t *l2_table = NULL;
216 old_l2_offset = s->l1_table[l1_index];
218 trace_qcow2_l2_allocate(bs, l1_index);
220 /* allocate a new l2 entry */
222 l2_offset = qcow2_alloc_clusters(bs, s->l2_size * sizeof(uint64_t));
228 ret = qcow2_cache_flush(bs, s->refcount_block_cache);
233 /* allocate a new entry in the l2 cache */
235 trace_qcow2_l2_allocate_get_empty(bs, l1_index);
236 ret = qcow2_cache_get_empty(bs, s->l2_table_cache, l2_offset, (void**) table);
243 if ((old_l2_offset & L1E_OFFSET_MASK) == 0) {
244 /* if there was no old l2 table, clear the new table */
245 memset(l2_table, 0, s->l2_size * sizeof(uint64_t));
249 /* if there was an old l2 table, read it from the disk */
250 BLKDBG_EVENT(bs->file, BLKDBG_L2_ALLOC_COW_READ);
251 ret = qcow2_cache_get(bs, s->l2_table_cache,
252 old_l2_offset & L1E_OFFSET_MASK,
253 (void**) &old_table);
258 memcpy(l2_table, old_table, s->cluster_size);
260 qcow2_cache_put(bs, s->l2_table_cache, (void **) &old_table);
263 /* write the l2 table to the file */
264 BLKDBG_EVENT(bs->file, BLKDBG_L2_ALLOC_WRITE);
266 trace_qcow2_l2_allocate_write_l2(bs, l1_index);
267 qcow2_cache_entry_mark_dirty(bs, s->l2_table_cache, l2_table);
268 ret = qcow2_cache_flush(bs, s->l2_table_cache);
273 /* update the L1 entry */
274 trace_qcow2_l2_allocate_write_l1(bs, l1_index);
275 s->l1_table[l1_index] = l2_offset | QCOW_OFLAG_COPIED;
276 ret = qcow2_write_l1_entry(bs, l1_index);
282 trace_qcow2_l2_allocate_done(bs, l1_index, 0);
286 trace_qcow2_l2_allocate_done(bs, l1_index, ret);
287 if (l2_table != NULL) {
288 qcow2_cache_put(bs, s->l2_table_cache, (void**) table);
290 s->l1_table[l1_index] = old_l2_offset;
292 qcow2_free_clusters(bs, l2_offset, s->l2_size * sizeof(uint64_t),
293 QCOW2_DISCARD_ALWAYS);
299 * Checks how many clusters in a given L2 table are contiguous in the image
300 * file. As soon as one of the flags in the bitmask stop_flags changes compared
301 * to the first cluster, the search is stopped and the cluster is not counted
302 * as contiguous. (This allows it, for example, to stop at the first compressed
303 * cluster which may require a different handling)
305 static int count_contiguous_clusters(int nb_clusters, int cluster_size,
306 uint64_t *l2_table, uint64_t stop_flags)
309 uint64_t mask = stop_flags | L2E_OFFSET_MASK | QCOW_OFLAG_COMPRESSED;
310 uint64_t first_entry = be64_to_cpu(l2_table[0]);
311 uint64_t offset = first_entry & mask;
316 assert(qcow2_get_cluster_type(first_entry) == QCOW2_CLUSTER_NORMAL);
318 for (i = 0; i < nb_clusters; i++) {
319 uint64_t l2_entry = be64_to_cpu(l2_table[i]) & mask;
320 if (offset + (uint64_t) i * cluster_size != l2_entry) {
328 static int count_contiguous_clusters_by_type(int nb_clusters,
334 for (i = 0; i < nb_clusters; i++) {
335 int type = qcow2_get_cluster_type(be64_to_cpu(l2_table[i]));
337 if (type != wanted_type) {
345 /* The crypt function is compatible with the linux cryptoloop
346 algorithm for < 4 GB images. NOTE: out_buf == in_buf is
348 int qcow2_encrypt_sectors(BDRVQcow2State *s, int64_t sector_num,
349 uint8_t *out_buf, const uint8_t *in_buf,
350 int nb_sectors, bool enc,
360 for(i = 0; i < nb_sectors; i++) {
361 ivec.ll[0] = cpu_to_le64(sector_num);
363 if (qcrypto_cipher_setiv(s->cipher,
364 ivec.b, G_N_ELEMENTS(ivec.b),
369 ret = qcrypto_cipher_encrypt(s->cipher,
375 ret = qcrypto_cipher_decrypt(s->cipher,
391 static int coroutine_fn do_perform_cow(BlockDriverState *bs,
392 uint64_t src_cluster_offset,
393 uint64_t cluster_offset,
394 int offset_in_cluster,
397 BDRVQcow2State *s = bs->opaque;
403 iov.iov_base = qemu_try_blockalign(bs, iov.iov_len);
404 if (iov.iov_base == NULL) {
408 qemu_iovec_init_external(&qiov, &iov, 1);
410 BLKDBG_EVENT(bs->file, BLKDBG_COW_READ);
417 /* Call .bdrv_co_readv() directly instead of using the public block-layer
418 * interface. This avoids double I/O throttling and request tracking,
419 * which can lead to deadlock when block layer copy-on-read is enabled.
421 ret = bs->drv->bdrv_co_preadv(bs, src_cluster_offset + offset_in_cluster,
429 int64_t sector = (cluster_offset + offset_in_cluster)
432 assert((offset_in_cluster & ~BDRV_SECTOR_MASK) == 0);
433 assert((bytes & ~BDRV_SECTOR_MASK) == 0);
434 if (qcow2_encrypt_sectors(s, sector, iov.iov_base, iov.iov_base,
435 bytes >> BDRV_SECTOR_BITS, true, &err) < 0) {
442 ret = qcow2_pre_write_overlap_check(bs, 0,
443 cluster_offset + offset_in_cluster, bytes);
448 BLKDBG_EVENT(bs->file, BLKDBG_COW_WRITE);
449 ret = bdrv_co_pwritev(bs->file->bs, cluster_offset + offset_in_cluster,
457 qemu_vfree(iov.iov_base);
465 * For a given offset of the virtual disk, find the cluster type and offset in
466 * the qcow2 file. The offset is stored in *cluster_offset.
468 * On entry, *bytes is the maximum number of contiguous bytes starting at
469 * offset that we are interested in.
471 * On exit, *bytes is the number of bytes starting at offset that have the same
472 * cluster type and (if applicable) are stored contiguously in the image file.
473 * Compressed clusters are always returned one by one.
475 * Returns the cluster type (QCOW2_CLUSTER_*) on success, -errno in error
478 int qcow2_get_cluster_offset(BlockDriverState *bs, uint64_t offset,
479 unsigned int *bytes, uint64_t *cluster_offset)
481 BDRVQcow2State *s = bs->opaque;
482 unsigned int l2_index;
483 uint64_t l1_index, l2_offset, *l2_table;
485 unsigned int offset_in_cluster, nb_clusters;
486 uint64_t bytes_available, bytes_needed;
489 offset_in_cluster = offset_into_cluster(s, offset);
490 bytes_needed = (uint64_t) *bytes + offset_in_cluster;
492 l1_bits = s->l2_bits + s->cluster_bits;
494 /* compute how many bytes there are between the start of the cluster
495 * containing offset and the end of the l1 entry */
496 bytes_available = (1ULL << l1_bits) - (offset & ((1ULL << l1_bits) - 1))
499 if (bytes_needed > bytes_available) {
500 bytes_needed = bytes_available;
502 assert(bytes_needed <= INT_MAX);
506 /* seek to the l2 offset in the l1 table */
508 l1_index = offset >> l1_bits;
509 if (l1_index >= s->l1_size) {
510 ret = QCOW2_CLUSTER_UNALLOCATED;
514 l2_offset = s->l1_table[l1_index] & L1E_OFFSET_MASK;
516 ret = QCOW2_CLUSTER_UNALLOCATED;
520 if (offset_into_cluster(s, l2_offset)) {
521 qcow2_signal_corruption(bs, true, -1, -1, "L2 table offset %#" PRIx64
522 " unaligned (L1 index: %#" PRIx64 ")",
523 l2_offset, l1_index);
527 /* load the l2 table in memory */
529 ret = l2_load(bs, l2_offset, &l2_table);
534 /* find the cluster offset for the given disk offset */
536 l2_index = (offset >> s->cluster_bits) & (s->l2_size - 1);
537 *cluster_offset = be64_to_cpu(l2_table[l2_index]);
539 /* nb_needed <= INT_MAX, thus nb_clusters <= INT_MAX, too */
540 nb_clusters = size_to_clusters(s, bytes_needed);
542 ret = qcow2_get_cluster_type(*cluster_offset);
544 case QCOW2_CLUSTER_COMPRESSED:
545 /* Compressed clusters can only be processed one by one */
547 *cluster_offset &= L2E_COMPRESSED_OFFSET_SIZE_MASK;
549 case QCOW2_CLUSTER_ZERO:
550 if (s->qcow_version < 3) {
551 qcow2_signal_corruption(bs, true, -1, -1, "Zero cluster entry found"
552 " in pre-v3 image (L2 offset: %#" PRIx64
553 ", L2 index: %#x)", l2_offset, l2_index);
557 c = count_contiguous_clusters_by_type(nb_clusters, &l2_table[l2_index],
561 case QCOW2_CLUSTER_UNALLOCATED:
562 /* how many empty clusters ? */
563 c = count_contiguous_clusters_by_type(nb_clusters, &l2_table[l2_index],
564 QCOW2_CLUSTER_UNALLOCATED);
567 case QCOW2_CLUSTER_NORMAL:
568 /* how many allocated clusters ? */
569 c = count_contiguous_clusters(nb_clusters, s->cluster_size,
570 &l2_table[l2_index], QCOW_OFLAG_ZERO);
571 *cluster_offset &= L2E_OFFSET_MASK;
572 if (offset_into_cluster(s, *cluster_offset)) {
573 qcow2_signal_corruption(bs, true, -1, -1, "Data cluster offset %#"
574 PRIx64 " unaligned (L2 offset: %#" PRIx64
575 ", L2 index: %#x)", *cluster_offset,
576 l2_offset, l2_index);
585 qcow2_cache_put(bs, s->l2_table_cache, (void**) &l2_table);
587 bytes_available = (c * s->cluster_size);
590 if (bytes_available > bytes_needed) {
591 bytes_available = bytes_needed;
594 *bytes = bytes_available - offset_in_cluster;
599 qcow2_cache_put(bs, s->l2_table_cache, (void **)&l2_table);
606 * for a given disk offset, load (and allocate if needed)
609 * the l2 table offset in the qcow2 file and the cluster index
610 * in the l2 table are given to the caller.
612 * Returns 0 on success, -errno in failure case
614 static int get_cluster_table(BlockDriverState *bs, uint64_t offset,
615 uint64_t **new_l2_table,
618 BDRVQcow2State *s = bs->opaque;
619 unsigned int l2_index;
620 uint64_t l1_index, l2_offset;
621 uint64_t *l2_table = NULL;
624 /* seek to the l2 offset in the l1 table */
626 l1_index = offset >> (s->l2_bits + s->cluster_bits);
627 if (l1_index >= s->l1_size) {
628 ret = qcow2_grow_l1_table(bs, l1_index + 1, false);
634 assert(l1_index < s->l1_size);
635 l2_offset = s->l1_table[l1_index] & L1E_OFFSET_MASK;
636 if (offset_into_cluster(s, l2_offset)) {
637 qcow2_signal_corruption(bs, true, -1, -1, "L2 table offset %#" PRIx64
638 " unaligned (L1 index: %#" PRIx64 ")",
639 l2_offset, l1_index);
643 /* seek the l2 table of the given l2 offset */
645 if (s->l1_table[l1_index] & QCOW_OFLAG_COPIED) {
646 /* load the l2 table in memory */
647 ret = l2_load(bs, l2_offset, &l2_table);
652 /* First allocate a new L2 table (and do COW if needed) */
653 ret = l2_allocate(bs, l1_index, &l2_table);
658 /* Then decrease the refcount of the old table */
660 qcow2_free_clusters(bs, l2_offset, s->l2_size * sizeof(uint64_t),
661 QCOW2_DISCARD_OTHER);
665 /* find the cluster offset for the given disk offset */
667 l2_index = (offset >> s->cluster_bits) & (s->l2_size - 1);
669 *new_l2_table = l2_table;
670 *new_l2_index = l2_index;
676 * alloc_compressed_cluster_offset
678 * For a given offset of the disk image, return cluster offset in
681 * If the offset is not found, allocate a new compressed cluster.
683 * Return the cluster offset if successful,
684 * Return 0, otherwise.
688 uint64_t qcow2_alloc_compressed_cluster_offset(BlockDriverState *bs,
692 BDRVQcow2State *s = bs->opaque;
695 int64_t cluster_offset;
698 ret = get_cluster_table(bs, offset, &l2_table, &l2_index);
703 /* Compression can't overwrite anything. Fail if the cluster was already
705 cluster_offset = be64_to_cpu(l2_table[l2_index]);
706 if (cluster_offset & L2E_OFFSET_MASK) {
707 qcow2_cache_put(bs, s->l2_table_cache, (void**) &l2_table);
711 cluster_offset = qcow2_alloc_bytes(bs, compressed_size);
712 if (cluster_offset < 0) {
713 qcow2_cache_put(bs, s->l2_table_cache, (void**) &l2_table);
717 nb_csectors = ((cluster_offset + compressed_size - 1) >> 9) -
718 (cluster_offset >> 9);
720 cluster_offset |= QCOW_OFLAG_COMPRESSED |
721 ((uint64_t)nb_csectors << s->csize_shift);
723 /* update L2 table */
725 /* compressed clusters never have the copied flag */
727 BLKDBG_EVENT(bs->file, BLKDBG_L2_UPDATE_COMPRESSED);
728 qcow2_cache_entry_mark_dirty(bs, s->l2_table_cache, l2_table);
729 l2_table[l2_index] = cpu_to_be64(cluster_offset);
730 qcow2_cache_put(bs, s->l2_table_cache, (void **) &l2_table);
732 return cluster_offset;
735 static int perform_cow(BlockDriverState *bs, QCowL2Meta *m, Qcow2COWRegion *r)
737 BDRVQcow2State *s = bs->opaque;
740 if (r->nb_bytes == 0) {
744 qemu_co_mutex_unlock(&s->lock);
745 ret = do_perform_cow(bs, m->offset, m->alloc_offset, r->offset, r->nb_bytes);
746 qemu_co_mutex_lock(&s->lock);
753 * Before we update the L2 table to actually point to the new cluster, we
754 * need to be sure that the refcounts have been increased and COW was
757 qcow2_cache_depends_on_flush(s->l2_table_cache);
762 int qcow2_alloc_cluster_link_l2(BlockDriverState *bs, QCowL2Meta *m)
764 BDRVQcow2State *s = bs->opaque;
765 int i, j = 0, l2_index, ret;
766 uint64_t *old_cluster, *l2_table;
767 uint64_t cluster_offset = m->alloc_offset;
769 trace_qcow2_cluster_link_l2(qemu_coroutine_self(), m->nb_clusters);
770 assert(m->nb_clusters > 0);
772 old_cluster = g_try_new(uint64_t, m->nb_clusters);
773 if (old_cluster == NULL) {
778 /* copy content of unmodified sectors */
779 ret = perform_cow(bs, m, &m->cow_start);
784 ret = perform_cow(bs, m, &m->cow_end);
789 /* Update L2 table. */
790 if (s->use_lazy_refcounts) {
791 qcow2_mark_dirty(bs);
793 if (qcow2_need_accurate_refcounts(s)) {
794 qcow2_cache_set_dependency(bs, s->l2_table_cache,
795 s->refcount_block_cache);
798 ret = get_cluster_table(bs, m->offset, &l2_table, &l2_index);
802 qcow2_cache_entry_mark_dirty(bs, s->l2_table_cache, l2_table);
804 assert(l2_index + m->nb_clusters <= s->l2_size);
805 for (i = 0; i < m->nb_clusters; i++) {
806 /* if two concurrent writes happen to the same unallocated cluster
807 * each write allocates separate cluster and writes data concurrently.
808 * The first one to complete updates l2 table with pointer to its
809 * cluster the second one has to do RMW (which is done above by
810 * perform_cow()), update l2 table with its cluster pointer and free
811 * old cluster. This is what this loop does */
812 if (l2_table[l2_index + i] != 0) {
813 old_cluster[j++] = l2_table[l2_index + i];
816 l2_table[l2_index + i] = cpu_to_be64((cluster_offset +
817 (i << s->cluster_bits)) | QCOW_OFLAG_COPIED);
821 qcow2_cache_put(bs, s->l2_table_cache, (void **) &l2_table);
824 * If this was a COW, we need to decrease the refcount of the old cluster.
826 * Don't discard clusters that reach a refcount of 0 (e.g. compressed
827 * clusters), the next write will reuse them anyway.
830 for (i = 0; i < j; i++) {
831 qcow2_free_any_clusters(bs, be64_to_cpu(old_cluster[i]), 1,
832 QCOW2_DISCARD_NEVER);
843 * Returns the number of contiguous clusters that can be used for an allocating
844 * write, but require COW to be performed (this includes yet unallocated space,
845 * which must copy from the backing file)
847 static int count_cow_clusters(BDRVQcow2State *s, int nb_clusters,
848 uint64_t *l2_table, int l2_index)
852 for (i = 0; i < nb_clusters; i++) {
853 uint64_t l2_entry = be64_to_cpu(l2_table[l2_index + i]);
854 int cluster_type = qcow2_get_cluster_type(l2_entry);
856 switch(cluster_type) {
857 case QCOW2_CLUSTER_NORMAL:
858 if (l2_entry & QCOW_OFLAG_COPIED) {
862 case QCOW2_CLUSTER_UNALLOCATED:
863 case QCOW2_CLUSTER_COMPRESSED:
864 case QCOW2_CLUSTER_ZERO:
872 assert(i <= nb_clusters);
877 * Check if there already is an AIO write request in flight which allocates
878 * the same cluster. In this case we need to wait until the previous
879 * request has completed and updated the L2 table accordingly.
882 * 0 if there was no dependency. *cur_bytes indicates the number of
883 * bytes from guest_offset that can be read before the next
884 * dependency must be processed (or the request is complete)
886 * -EAGAIN if we had to wait for another request, previously gathered
887 * information on cluster allocation may be invalid now. The caller
888 * must start over anyway, so consider *cur_bytes undefined.
890 static int handle_dependencies(BlockDriverState *bs, uint64_t guest_offset,
891 uint64_t *cur_bytes, QCowL2Meta **m)
893 BDRVQcow2State *s = bs->opaque;
894 QCowL2Meta *old_alloc;
895 uint64_t bytes = *cur_bytes;
897 QLIST_FOREACH(old_alloc, &s->cluster_allocs, next_in_flight) {
899 uint64_t start = guest_offset;
900 uint64_t end = start + bytes;
901 uint64_t old_start = l2meta_cow_start(old_alloc);
902 uint64_t old_end = l2meta_cow_end(old_alloc);
904 if (end <= old_start || start >= old_end) {
905 /* No intersection */
907 if (start < old_start) {
908 /* Stop at the start of a running allocation */
909 bytes = old_start - start;
914 /* Stop if already an l2meta exists. After yielding, it wouldn't
915 * be valid any more, so we'd have to clean up the old L2Metas
916 * and deal with requests depending on them before starting to
917 * gather new ones. Not worth the trouble. */
918 if (bytes == 0 && *m) {
924 /* Wait for the dependency to complete. We need to recheck
925 * the free/allocated clusters when we continue. */
926 qemu_co_mutex_unlock(&s->lock);
927 qemu_co_queue_wait(&old_alloc->dependent_requests);
928 qemu_co_mutex_lock(&s->lock);
934 /* Make sure that existing clusters and new allocations are only used up to
935 * the next dependency if we shortened the request above */
942 * Checks how many already allocated clusters that don't require a copy on
943 * write there are at the given guest_offset (up to *bytes). If
944 * *host_offset is not zero, only physically contiguous clusters beginning at
945 * this host offset are counted.
947 * Note that guest_offset may not be cluster aligned. In this case, the
948 * returned *host_offset points to exact byte referenced by guest_offset and
949 * therefore isn't cluster aligned as well.
952 * 0: if no allocated clusters are available at the given offset.
953 * *bytes is normally unchanged. It is set to 0 if the cluster
954 * is allocated and doesn't need COW, but doesn't have the right
957 * 1: if allocated clusters that don't require a COW are available at
958 * the requested offset. *bytes may have decreased and describes
959 * the length of the area that can be written to.
961 * -errno: in error cases
963 static int handle_copied(BlockDriverState *bs, uint64_t guest_offset,
964 uint64_t *host_offset, uint64_t *bytes, QCowL2Meta **m)
966 BDRVQcow2State *s = bs->opaque;
968 uint64_t cluster_offset;
970 uint64_t nb_clusters;
971 unsigned int keep_clusters;
974 trace_qcow2_handle_copied(qemu_coroutine_self(), guest_offset, *host_offset,
977 assert(*host_offset == 0 || offset_into_cluster(s, guest_offset)
978 == offset_into_cluster(s, *host_offset));
981 * Calculate the number of clusters to look for. We stop at L2 table
982 * boundaries to keep things simple.
985 size_to_clusters(s, offset_into_cluster(s, guest_offset) + *bytes);
987 l2_index = offset_to_l2_index(s, guest_offset);
988 nb_clusters = MIN(nb_clusters, s->l2_size - l2_index);
989 assert(nb_clusters <= INT_MAX);
991 /* Find L2 entry for the first involved cluster */
992 ret = get_cluster_table(bs, guest_offset, &l2_table, &l2_index);
997 cluster_offset = be64_to_cpu(l2_table[l2_index]);
999 /* Check how many clusters are already allocated and don't need COW */
1000 if (qcow2_get_cluster_type(cluster_offset) == QCOW2_CLUSTER_NORMAL
1001 && (cluster_offset & QCOW_OFLAG_COPIED))
1003 /* If a specific host_offset is required, check it */
1004 bool offset_matches =
1005 (cluster_offset & L2E_OFFSET_MASK) == *host_offset;
1007 if (offset_into_cluster(s, cluster_offset & L2E_OFFSET_MASK)) {
1008 qcow2_signal_corruption(bs, true, -1, -1, "Data cluster offset "
1009 "%#llx unaligned (guest offset: %#" PRIx64
1010 ")", cluster_offset & L2E_OFFSET_MASK,
1016 if (*host_offset != 0 && !offset_matches) {
1022 /* We keep all QCOW_OFLAG_COPIED clusters */
1024 count_contiguous_clusters(nb_clusters, s->cluster_size,
1025 &l2_table[l2_index],
1026 QCOW_OFLAG_COPIED | QCOW_OFLAG_ZERO);
1027 assert(keep_clusters <= nb_clusters);
1029 *bytes = MIN(*bytes,
1030 keep_clusters * s->cluster_size
1031 - offset_into_cluster(s, guest_offset));
1040 qcow2_cache_put(bs, s->l2_table_cache, (void **) &l2_table);
1042 /* Only return a host offset if we actually made progress. Otherwise we
1043 * would make requirements for handle_alloc() that it can't fulfill */
1045 *host_offset = (cluster_offset & L2E_OFFSET_MASK)
1046 + offset_into_cluster(s, guest_offset);
1053 * Allocates new clusters for the given guest_offset.
1055 * At most *nb_clusters are allocated, and on return *nb_clusters is updated to
1056 * contain the number of clusters that have been allocated and are contiguous
1057 * in the image file.
1059 * If *host_offset is non-zero, it specifies the offset in the image file at
1060 * which the new clusters must start. *nb_clusters can be 0 on return in this
1061 * case if the cluster at host_offset is already in use. If *host_offset is
1062 * zero, the clusters can be allocated anywhere in the image file.
1064 * *host_offset is updated to contain the offset into the image file at which
1065 * the first allocated cluster starts.
1067 * Return 0 on success and -errno in error cases. -EAGAIN means that the
1068 * function has been waiting for another request and the allocation must be
1069 * restarted, but the whole request should not be failed.
1071 static int do_alloc_cluster_offset(BlockDriverState *bs, uint64_t guest_offset,
1072 uint64_t *host_offset, uint64_t *nb_clusters)
1074 BDRVQcow2State *s = bs->opaque;
1076 trace_qcow2_do_alloc_clusters_offset(qemu_coroutine_self(), guest_offset,
1077 *host_offset, *nb_clusters);
1079 /* Allocate new clusters */
1080 trace_qcow2_cluster_alloc_phys(qemu_coroutine_self());
1081 if (*host_offset == 0) {
1082 int64_t cluster_offset =
1083 qcow2_alloc_clusters(bs, *nb_clusters * s->cluster_size);
1084 if (cluster_offset < 0) {
1085 return cluster_offset;
1087 *host_offset = cluster_offset;
1090 int64_t ret = qcow2_alloc_clusters_at(bs, *host_offset, *nb_clusters);
1100 * Allocates new clusters for an area that either is yet unallocated or needs a
1101 * copy on write. If *host_offset is non-zero, clusters are only allocated if
1102 * the new allocation can match the specified host offset.
1104 * Note that guest_offset may not be cluster aligned. In this case, the
1105 * returned *host_offset points to exact byte referenced by guest_offset and
1106 * therefore isn't cluster aligned as well.
1109 * 0: if no clusters could be allocated. *bytes is set to 0,
1110 * *host_offset is left unchanged.
1112 * 1: if new clusters were allocated. *bytes may be decreased if the
1113 * new allocation doesn't cover all of the requested area.
1114 * *host_offset is updated to contain the host offset of the first
1115 * newly allocated cluster.
1117 * -errno: in error cases
1119 static int handle_alloc(BlockDriverState *bs, uint64_t guest_offset,
1120 uint64_t *host_offset, uint64_t *bytes, QCowL2Meta **m)
1122 BDRVQcow2State *s = bs->opaque;
1126 uint64_t nb_clusters;
1129 uint64_t alloc_cluster_offset;
1131 trace_qcow2_handle_alloc(qemu_coroutine_self(), guest_offset, *host_offset,
1136 * Calculate the number of clusters to look for. We stop at L2 table
1137 * boundaries to keep things simple.
1140 size_to_clusters(s, offset_into_cluster(s, guest_offset) + *bytes);
1142 l2_index = offset_to_l2_index(s, guest_offset);
1143 nb_clusters = MIN(nb_clusters, s->l2_size - l2_index);
1144 assert(nb_clusters <= INT_MAX);
1146 /* Find L2 entry for the first involved cluster */
1147 ret = get_cluster_table(bs, guest_offset, &l2_table, &l2_index);
1152 entry = be64_to_cpu(l2_table[l2_index]);
1154 /* For the moment, overwrite compressed clusters one by one */
1155 if (entry & QCOW_OFLAG_COMPRESSED) {
1158 nb_clusters = count_cow_clusters(s, nb_clusters, l2_table, l2_index);
1161 /* This function is only called when there were no non-COW clusters, so if
1162 * we can't find any unallocated or COW clusters either, something is
1163 * wrong with our code. */
1164 assert(nb_clusters > 0);
1166 qcow2_cache_put(bs, s->l2_table_cache, (void **) &l2_table);
1168 /* Allocate, if necessary at a given offset in the image file */
1169 alloc_cluster_offset = start_of_cluster(s, *host_offset);
1170 ret = do_alloc_cluster_offset(bs, guest_offset, &alloc_cluster_offset,
1176 /* Can't extend contiguous allocation */
1177 if (nb_clusters == 0) {
1182 /* !*host_offset would overwrite the image header and is reserved for "no
1183 * host offset preferred". If 0 was a valid host offset, it'd trigger the
1184 * following overlap check; do that now to avoid having an invalid value in
1186 if (!alloc_cluster_offset) {
1187 ret = qcow2_pre_write_overlap_check(bs, 0, alloc_cluster_offset,
1188 nb_clusters * s->cluster_size);
1194 * Save info needed for meta data update.
1196 * requested_bytes: Number of bytes from the start of the first
1197 * newly allocated cluster to the end of the (possibly shortened
1198 * before) write request.
1200 * avail_bytes: Number of bytes from the start of the first
1201 * newly allocated to the end of the last newly allocated cluster.
1203 * nb_bytes: The number of bytes from the start of the first
1204 * newly allocated cluster to the end of the area that the write
1205 * request actually writes to (excluding COW at the end)
1207 uint64_t requested_bytes = *bytes + offset_into_cluster(s, guest_offset);
1208 int avail_bytes = MIN(INT_MAX, nb_clusters << s->cluster_bits);
1209 int nb_bytes = MIN(requested_bytes, avail_bytes);
1210 QCowL2Meta *old_m = *m;
1212 *m = g_malloc0(sizeof(**m));
1214 **m = (QCowL2Meta) {
1217 .alloc_offset = alloc_cluster_offset,
1218 .offset = start_of_cluster(s, guest_offset),
1219 .nb_clusters = nb_clusters,
1223 .nb_bytes = offset_into_cluster(s, guest_offset),
1227 .nb_bytes = avail_bytes - nb_bytes,
1230 qemu_co_queue_init(&(*m)->dependent_requests);
1231 QLIST_INSERT_HEAD(&s->cluster_allocs, *m, next_in_flight);
1233 *host_offset = alloc_cluster_offset + offset_into_cluster(s, guest_offset);
1234 *bytes = MIN(*bytes, nb_bytes - offset_into_cluster(s, guest_offset));
1235 assert(*bytes != 0);
1240 if (*m && (*m)->nb_clusters > 0) {
1241 QLIST_REMOVE(*m, next_in_flight);
1247 * alloc_cluster_offset
1249 * For a given offset on the virtual disk, find the cluster offset in qcow2
1250 * file. If the offset is not found, allocate a new cluster.
1252 * If the cluster was already allocated, m->nb_clusters is set to 0 and
1253 * other fields in m are meaningless.
1255 * If the cluster is newly allocated, m->nb_clusters is set to the number of
1256 * contiguous clusters that have been allocated. In this case, the other
1257 * fields of m are valid and contain information about the first allocated
1260 * If the request conflicts with another write request in flight, the coroutine
1261 * is queued and will be reentered when the dependency has completed.
1263 * Return 0 on success and -errno in error cases
1265 int qcow2_alloc_cluster_offset(BlockDriverState *bs, uint64_t offset,
1266 unsigned int *bytes, uint64_t *host_offset,
1269 BDRVQcow2State *s = bs->opaque;
1270 uint64_t start, remaining;
1271 uint64_t cluster_offset;
1275 trace_qcow2_alloc_clusters_offset(qemu_coroutine_self(), offset, *bytes);
1287 if (!*host_offset) {
1288 *host_offset = start_of_cluster(s, cluster_offset);
1291 assert(remaining >= cur_bytes);
1294 remaining -= cur_bytes;
1295 cluster_offset += cur_bytes;
1297 if (remaining == 0) {
1301 cur_bytes = remaining;
1304 * Now start gathering as many contiguous clusters as possible:
1306 * 1. Check for overlaps with in-flight allocations
1308 * a) Overlap not in the first cluster -> shorten this request and
1309 * let the caller handle the rest in its next loop iteration.
1311 * b) Real overlaps of two requests. Yield and restart the search
1312 * for contiguous clusters (the situation could have changed
1313 * while we were sleeping)
1315 * c) TODO: Request starts in the same cluster as the in-flight
1316 * allocation ends. Shorten the COW of the in-fight allocation,
1317 * set cluster_offset to write to the same cluster and set up
1318 * the right synchronisation between the in-flight request and
1321 ret = handle_dependencies(bs, start, &cur_bytes, m);
1322 if (ret == -EAGAIN) {
1323 /* Currently handle_dependencies() doesn't yield if we already had
1324 * an allocation. If it did, we would have to clean up the L2Meta
1325 * structs before starting over. */
1328 } else if (ret < 0) {
1330 } else if (cur_bytes == 0) {
1333 /* handle_dependencies() may have decreased cur_bytes (shortened
1334 * the allocations below) so that the next dependency is processed
1335 * correctly during the next loop iteration. */
1339 * 2. Count contiguous COPIED clusters.
1341 ret = handle_copied(bs, start, &cluster_offset, &cur_bytes, m);
1346 } else if (cur_bytes == 0) {
1351 * 3. If the request still hasn't completed, allocate new clusters,
1352 * considering any cluster_offset of steps 1c or 2.
1354 ret = handle_alloc(bs, start, &cluster_offset, &cur_bytes, m);
1360 assert(cur_bytes == 0);
1365 *bytes -= remaining;
1367 assert(*host_offset != 0);
1372 static int decompress_buffer(uint8_t *out_buf, int out_buf_size,
1373 const uint8_t *buf, int buf_size)
1375 z_stream strm1, *strm = &strm1;
1378 memset(strm, 0, sizeof(*strm));
1380 strm->next_in = (uint8_t *)buf;
1381 strm->avail_in = buf_size;
1382 strm->next_out = out_buf;
1383 strm->avail_out = out_buf_size;
1385 ret = inflateInit2(strm, -12);
1388 ret = inflate(strm, Z_FINISH);
1389 out_len = strm->next_out - out_buf;
1390 if ((ret != Z_STREAM_END && ret != Z_BUF_ERROR) ||
1391 out_len != out_buf_size) {
1399 int qcow2_decompress_cluster(BlockDriverState *bs, uint64_t cluster_offset)
1401 BDRVQcow2State *s = bs->opaque;
1402 int ret, csize, nb_csectors, sector_offset;
1405 coffset = cluster_offset & s->cluster_offset_mask;
1406 if (s->cluster_cache_offset != coffset) {
1407 nb_csectors = ((cluster_offset >> s->csize_shift) & s->csize_mask) + 1;
1408 sector_offset = coffset & 511;
1409 csize = nb_csectors * 512 - sector_offset;
1410 BLKDBG_EVENT(bs->file, BLKDBG_READ_COMPRESSED);
1411 ret = bdrv_read(bs->file, coffset >> 9, s->cluster_data,
1416 if (decompress_buffer(s->cluster_cache, s->cluster_size,
1417 s->cluster_data + sector_offset, csize) < 0) {
1420 s->cluster_cache_offset = coffset;
1426 * This discards as many clusters of nb_clusters as possible at once (i.e.
1427 * all clusters in the same L2 table) and returns the number of discarded
1430 static int discard_single_l2(BlockDriverState *bs, uint64_t offset,
1431 uint64_t nb_clusters, enum qcow2_discard_type type,
1434 BDRVQcow2State *s = bs->opaque;
1440 ret = get_cluster_table(bs, offset, &l2_table, &l2_index);
1445 /* Limit nb_clusters to one L2 table */
1446 nb_clusters = MIN(nb_clusters, s->l2_size - l2_index);
1447 assert(nb_clusters <= INT_MAX);
1449 for (i = 0; i < nb_clusters; i++) {
1450 uint64_t old_l2_entry;
1452 old_l2_entry = be64_to_cpu(l2_table[l2_index + i]);
1455 * If full_discard is false, make sure that a discarded area reads back
1456 * as zeroes for v3 images (we cannot do it for v2 without actually
1457 * writing a zero-filled buffer). We can skip the operation if the
1458 * cluster is already marked as zero, or if it's unallocated and we
1459 * don't have a backing file.
1461 * TODO We might want to use bdrv_get_block_status(bs) here, but we're
1462 * holding s->lock, so that doesn't work today.
1464 * If full_discard is true, the sector should not read back as zeroes,
1465 * but rather fall through to the backing file.
1467 switch (qcow2_get_cluster_type(old_l2_entry)) {
1468 case QCOW2_CLUSTER_UNALLOCATED:
1469 if (full_discard || !bs->backing) {
1474 case QCOW2_CLUSTER_ZERO:
1475 if (!full_discard) {
1480 case QCOW2_CLUSTER_NORMAL:
1481 case QCOW2_CLUSTER_COMPRESSED:
1488 /* First remove L2 entries */
1489 qcow2_cache_entry_mark_dirty(bs, s->l2_table_cache, l2_table);
1490 if (!full_discard && s->qcow_version >= 3) {
1491 l2_table[l2_index + i] = cpu_to_be64(QCOW_OFLAG_ZERO);
1493 l2_table[l2_index + i] = cpu_to_be64(0);
1496 /* Then decrease the refcount */
1497 qcow2_free_any_clusters(bs, old_l2_entry, 1, type);
1500 qcow2_cache_put(bs, s->l2_table_cache, (void **) &l2_table);
1505 int qcow2_discard_clusters(BlockDriverState *bs, uint64_t offset,
1506 int nb_sectors, enum qcow2_discard_type type, bool full_discard)
1508 BDRVQcow2State *s = bs->opaque;
1509 uint64_t end_offset;
1510 uint64_t nb_clusters;
1513 end_offset = offset + (nb_sectors << BDRV_SECTOR_BITS);
1515 /* Round start up and end down */
1516 offset = align_offset(offset, s->cluster_size);
1517 end_offset = start_of_cluster(s, end_offset);
1519 if (offset > end_offset) {
1523 nb_clusters = size_to_clusters(s, end_offset - offset);
1525 s->cache_discards = true;
1527 /* Each L2 table is handled by its own loop iteration */
1528 while (nb_clusters > 0) {
1529 ret = discard_single_l2(bs, offset, nb_clusters, type, full_discard);
1535 offset += (ret * s->cluster_size);
1540 s->cache_discards = false;
1541 qcow2_process_discards(bs, ret);
1547 * This zeroes as many clusters of nb_clusters as possible at once (i.e.
1548 * all clusters in the same L2 table) and returns the number of zeroed
1551 static int zero_single_l2(BlockDriverState *bs, uint64_t offset,
1552 uint64_t nb_clusters)
1554 BDRVQcow2State *s = bs->opaque;
1560 ret = get_cluster_table(bs, offset, &l2_table, &l2_index);
1565 /* Limit nb_clusters to one L2 table */
1566 nb_clusters = MIN(nb_clusters, s->l2_size - l2_index);
1567 assert(nb_clusters <= INT_MAX);
1569 for (i = 0; i < nb_clusters; i++) {
1570 uint64_t old_offset;
1572 old_offset = be64_to_cpu(l2_table[l2_index + i]);
1574 /* Update L2 entries */
1575 qcow2_cache_entry_mark_dirty(bs, s->l2_table_cache, l2_table);
1576 if (old_offset & QCOW_OFLAG_COMPRESSED) {
1577 l2_table[l2_index + i] = cpu_to_be64(QCOW_OFLAG_ZERO);
1578 qcow2_free_any_clusters(bs, old_offset, 1, QCOW2_DISCARD_REQUEST);
1580 l2_table[l2_index + i] |= cpu_to_be64(QCOW_OFLAG_ZERO);
1584 qcow2_cache_put(bs, s->l2_table_cache, (void **) &l2_table);
1589 int qcow2_zero_clusters(BlockDriverState *bs, uint64_t offset, int nb_sectors)
1591 BDRVQcow2State *s = bs->opaque;
1592 uint64_t nb_clusters;
1595 /* The zero flag is only supported by version 3 and newer */
1596 if (s->qcow_version < 3) {
1600 /* Each L2 table is handled by its own loop iteration */
1601 nb_clusters = size_to_clusters(s, nb_sectors << BDRV_SECTOR_BITS);
1603 s->cache_discards = true;
1605 while (nb_clusters > 0) {
1606 ret = zero_single_l2(bs, offset, nb_clusters);
1612 offset += (ret * s->cluster_size);
1617 s->cache_discards = false;
1618 qcow2_process_discards(bs, ret);
1624 * Expands all zero clusters in a specific L1 table (or deallocates them, for
1625 * non-backed non-pre-allocated zero clusters).
1627 * l1_entries and *visited_l1_entries are used to keep track of progress for
1628 * status_cb(). l1_entries contains the total number of L1 entries and
1629 * *visited_l1_entries counts all visited L1 entries.
1631 static int expand_zero_clusters_in_l1(BlockDriverState *bs, uint64_t *l1_table,
1632 int l1_size, int64_t *visited_l1_entries,
1634 BlockDriverAmendStatusCB *status_cb,
1637 BDRVQcow2State *s = bs->opaque;
1638 bool is_active_l1 = (l1_table == s->l1_table);
1639 uint64_t *l2_table = NULL;
1643 if (!is_active_l1) {
1644 /* inactive L2 tables require a buffer to be stored in when loading
1646 l2_table = qemu_try_blockalign(bs->file->bs, s->cluster_size);
1647 if (l2_table == NULL) {
1652 for (i = 0; i < l1_size; i++) {
1653 uint64_t l2_offset = l1_table[i] & L1E_OFFSET_MASK;
1654 bool l2_dirty = false;
1655 uint64_t l2_refcount;
1659 (*visited_l1_entries)++;
1661 status_cb(bs, *visited_l1_entries, l1_entries, cb_opaque);
1666 if (offset_into_cluster(s, l2_offset)) {
1667 qcow2_signal_corruption(bs, true, -1, -1, "L2 table offset %#"
1668 PRIx64 " unaligned (L1 index: %#x)",
1675 /* get active L2 tables from cache */
1676 ret = qcow2_cache_get(bs, s->l2_table_cache, l2_offset,
1677 (void **)&l2_table);
1679 /* load inactive L2 tables from disk */
1680 ret = bdrv_read(bs->file, l2_offset / BDRV_SECTOR_SIZE,
1681 (void *)l2_table, s->cluster_sectors);
1687 ret = qcow2_get_refcount(bs, l2_offset >> s->cluster_bits,
1693 for (j = 0; j < s->l2_size; j++) {
1694 uint64_t l2_entry = be64_to_cpu(l2_table[j]);
1695 int64_t offset = l2_entry & L2E_OFFSET_MASK;
1696 int cluster_type = qcow2_get_cluster_type(l2_entry);
1697 bool preallocated = offset != 0;
1699 if (cluster_type != QCOW2_CLUSTER_ZERO) {
1703 if (!preallocated) {
1705 /* not backed; therefore we can simply deallocate the
1712 offset = qcow2_alloc_clusters(bs, s->cluster_size);
1718 if (l2_refcount > 1) {
1719 /* For shared L2 tables, set the refcount accordingly (it is
1720 * already 1 and needs to be l2_refcount) */
1721 ret = qcow2_update_cluster_refcount(bs,
1722 offset >> s->cluster_bits,
1723 refcount_diff(1, l2_refcount), false,
1724 QCOW2_DISCARD_OTHER);
1726 qcow2_free_clusters(bs, offset, s->cluster_size,
1727 QCOW2_DISCARD_OTHER);
1733 if (offset_into_cluster(s, offset)) {
1734 qcow2_signal_corruption(bs, true, -1, -1, "Data cluster offset "
1735 "%#" PRIx64 " unaligned (L2 offset: %#"
1736 PRIx64 ", L2 index: %#x)", offset,
1738 if (!preallocated) {
1739 qcow2_free_clusters(bs, offset, s->cluster_size,
1740 QCOW2_DISCARD_ALWAYS);
1746 ret = qcow2_pre_write_overlap_check(bs, 0, offset, s->cluster_size);
1748 if (!preallocated) {
1749 qcow2_free_clusters(bs, offset, s->cluster_size,
1750 QCOW2_DISCARD_ALWAYS);
1755 ret = bdrv_pwrite_zeroes(bs->file->bs, offset, s->cluster_size, 0);
1757 if (!preallocated) {
1758 qcow2_free_clusters(bs, offset, s->cluster_size,
1759 QCOW2_DISCARD_ALWAYS);
1764 if (l2_refcount == 1) {
1765 l2_table[j] = cpu_to_be64(offset | QCOW_OFLAG_COPIED);
1767 l2_table[j] = cpu_to_be64(offset);
1774 qcow2_cache_entry_mark_dirty(bs, s->l2_table_cache, l2_table);
1775 qcow2_cache_depends_on_flush(s->l2_table_cache);
1777 qcow2_cache_put(bs, s->l2_table_cache, (void **) &l2_table);
1780 ret = qcow2_pre_write_overlap_check(bs,
1781 QCOW2_OL_INACTIVE_L2 | QCOW2_OL_ACTIVE_L2, l2_offset,
1787 ret = bdrv_write(bs->file, l2_offset / BDRV_SECTOR_SIZE,
1788 (void *)l2_table, s->cluster_sectors);
1795 (*visited_l1_entries)++;
1797 status_cb(bs, *visited_l1_entries, l1_entries, cb_opaque);
1805 if (!is_active_l1) {
1806 qemu_vfree(l2_table);
1808 qcow2_cache_put(bs, s->l2_table_cache, (void **) &l2_table);
1815 * For backed images, expands all zero clusters on the image. For non-backed
1816 * images, deallocates all non-pre-allocated zero clusters (and claims the
1817 * allocation for pre-allocated ones). This is important for downgrading to a
1818 * qcow2 version which doesn't yet support metadata zero clusters.
1820 int qcow2_expand_zero_clusters(BlockDriverState *bs,
1821 BlockDriverAmendStatusCB *status_cb,
1824 BDRVQcow2State *s = bs->opaque;
1825 uint64_t *l1_table = NULL;
1826 int64_t l1_entries = 0, visited_l1_entries = 0;
1831 l1_entries = s->l1_size;
1832 for (i = 0; i < s->nb_snapshots; i++) {
1833 l1_entries += s->snapshots[i].l1_size;
1837 ret = expand_zero_clusters_in_l1(bs, s->l1_table, s->l1_size,
1838 &visited_l1_entries, l1_entries,
1839 status_cb, cb_opaque);
1844 /* Inactive L1 tables may point to active L2 tables - therefore it is
1845 * necessary to flush the L2 table cache before trying to access the L2
1846 * tables pointed to by inactive L1 entries (else we might try to expand
1847 * zero clusters that have already been expanded); furthermore, it is also
1848 * necessary to empty the L2 table cache, since it may contain tables which
1849 * are now going to be modified directly on disk, bypassing the cache.
1850 * qcow2_cache_empty() does both for us. */
1851 ret = qcow2_cache_empty(bs, s->l2_table_cache);
1856 for (i = 0; i < s->nb_snapshots; i++) {
1857 int l1_sectors = DIV_ROUND_UP(s->snapshots[i].l1_size *
1858 sizeof(uint64_t), BDRV_SECTOR_SIZE);
1860 l1_table = g_realloc(l1_table, l1_sectors * BDRV_SECTOR_SIZE);
1862 ret = bdrv_read(bs->file,
1863 s->snapshots[i].l1_table_offset / BDRV_SECTOR_SIZE,
1864 (void *)l1_table, l1_sectors);
1869 for (j = 0; j < s->snapshots[i].l1_size; j++) {
1870 be64_to_cpus(&l1_table[j]);
1873 ret = expand_zero_clusters_in_l1(bs, l1_table, s->snapshots[i].l1_size,
1874 &visited_l1_entries, l1_entries,
1875 status_cb, cb_opaque);