2009-05-28 18:07:04 +04:00
|
|
|
/*
|
|
|
|
* Block driver for the QCOW version 2 format
|
|
|
|
*
|
|
|
|
* Copyright (c) 2004-2006 Fabrice Bellard
|
|
|
|
*
|
|
|
|
* Permission is hereby granted, free of charge, to any person obtaining a copy
|
|
|
|
* of this software and associated documentation files (the "Software"), to deal
|
|
|
|
* in the Software without restriction, including without limitation the rights
|
|
|
|
* to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
|
|
|
|
* copies of the Software, and to permit persons to whom the Software is
|
|
|
|
* furnished to do so, subject to the following conditions:
|
|
|
|
*
|
|
|
|
* The above copyright notice and this permission notice shall be included in
|
|
|
|
* all copies or substantial portions of the Software.
|
|
|
|
*
|
|
|
|
* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
|
|
|
|
* IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
|
|
|
|
* FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
|
|
|
|
* THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
|
|
|
|
* LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
|
|
|
|
* OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
|
|
|
|
* THE SOFTWARE.
|
|
|
|
*/
|
|
|
|
|
|
|
|
#include "qemu-common.h"
|
2012-12-17 21:19:44 +04:00
|
|
|
#include "block/block_int.h"
|
2009-05-28 18:07:04 +04:00
|
|
|
#include "block/qcow2.h"
|
2013-08-30 16:34:25 +04:00
|
|
|
#include "qemu/range.h"
|
|
|
|
#include "qapi/qmp/types.h"
|
2009-05-28 18:07:04 +04:00
|
|
|
|
2014-03-26 16:05:51 +04:00
|
|
|
static int64_t alloc_clusters_noref(BlockDriverState *bs, uint64_t size);
|
2010-02-23 18:40:53 +03:00
|
|
|
static int QEMU_WARN_UNUSED_RESULT update_refcount(BlockDriverState *bs,
|
2009-05-28 18:07:04 +04:00
|
|
|
int64_t offset, int64_t length,
|
2013-06-19 15:44:18 +04:00
|
|
|
int addend, enum qcow2_discard_type type);
|
2009-05-28 18:07:04 +04:00
|
|
|
|
2009-06-26 22:19:38 +04:00
|
|
|
|
2009-05-28 18:07:04 +04:00
|
|
|
/*********************************************************/
|
|
|
|
/* refcount handling */
|
|
|
|
|
2009-05-28 18:07:07 +04:00
|
|
|
int qcow2_refcount_init(BlockDriverState *bs)
|
2009-05-28 18:07:04 +04:00
|
|
|
{
|
|
|
|
BDRVQcowState *s = bs->opaque;
|
2014-03-26 16:05:43 +04:00
|
|
|
unsigned int refcount_table_size2, i;
|
|
|
|
int ret;
|
2009-05-28 18:07:04 +04:00
|
|
|
|
2014-03-26 16:05:43 +04:00
|
|
|
assert(s->refcount_table_size <= INT_MAX / sizeof(uint64_t));
|
2009-05-28 18:07:04 +04:00
|
|
|
refcount_table_size2 = s->refcount_table_size * sizeof(uint64_t);
|
2011-08-21 07:09:37 +04:00
|
|
|
s->refcount_table = g_malloc(refcount_table_size2);
|
2009-05-28 18:07:04 +04:00
|
|
|
if (s->refcount_table_size > 0) {
|
2010-04-14 16:17:38 +04:00
|
|
|
BLKDBG_EVENT(bs->file, BLKDBG_REFTABLE_LOAD);
|
|
|
|
ret = bdrv_pread(bs->file, s->refcount_table_offset,
|
2009-05-28 18:07:04 +04:00
|
|
|
s->refcount_table, refcount_table_size2);
|
|
|
|
if (ret != refcount_table_size2)
|
|
|
|
goto fail;
|
|
|
|
for(i = 0; i < s->refcount_table_size; i++)
|
|
|
|
be64_to_cpus(&s->refcount_table[i]);
|
|
|
|
}
|
|
|
|
return 0;
|
|
|
|
fail:
|
|
|
|
return -ENOMEM;
|
|
|
|
}
|
|
|
|
|
2009-05-28 18:07:07 +04:00
|
|
|
void qcow2_refcount_close(BlockDriverState *bs)
|
2009-05-28 18:07:04 +04:00
|
|
|
{
|
|
|
|
BDRVQcowState *s = bs->opaque;
|
2011-08-21 07:09:37 +04:00
|
|
|
g_free(s->refcount_table);
|
2009-05-28 18:07:04 +04:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
static int load_refcount_block(BlockDriverState *bs,
|
2011-01-10 19:17:28 +03:00
|
|
|
int64_t refcount_block_offset,
|
|
|
|
void **refcount_block)
|
2009-05-28 18:07:04 +04:00
|
|
|
{
|
|
|
|
BDRVQcowState *s = bs->opaque;
|
|
|
|
int ret;
|
2009-06-26 22:19:38 +04:00
|
|
|
|
2010-04-14 16:17:38 +04:00
|
|
|
BLKDBG_EVENT(bs->file, BLKDBG_REFBLOCK_LOAD);
|
2011-01-10 19:17:28 +03:00
|
|
|
ret = qcow2_cache_get(bs, s->refcount_block_cache, refcount_block_offset,
|
|
|
|
refcount_block);
|
2010-06-04 13:27:08 +04:00
|
|
|
|
2011-01-10 19:17:28 +03:00
|
|
|
return ret;
|
2009-05-28 18:07:04 +04:00
|
|
|
}
|
|
|
|
|
2010-06-04 13:16:11 +04:00
|
|
|
/*
|
|
|
|
* Returns the refcount of the cluster given by its index. Any non-negative
|
|
|
|
* return value is the refcount of the cluster, negative values are -errno
|
|
|
|
* and indicate an error.
|
|
|
|
*/
|
2009-05-28 18:07:04 +04:00
|
|
|
static int get_refcount(BlockDriverState *bs, int64_t cluster_index)
|
|
|
|
{
|
|
|
|
BDRVQcowState *s = bs->opaque;
|
2014-03-26 16:05:49 +04:00
|
|
|
uint64_t refcount_table_index, block_index;
|
2009-05-28 18:07:04 +04:00
|
|
|
int64_t refcount_block_offset;
|
2010-06-04 13:16:11 +04:00
|
|
|
int ret;
|
2011-01-10 19:17:28 +03:00
|
|
|
uint16_t *refcount_block;
|
|
|
|
uint16_t refcount;
|
2009-05-28 18:07:04 +04:00
|
|
|
|
|
|
|
refcount_table_index = cluster_index >> (s->cluster_bits - REFCOUNT_SHIFT);
|
|
|
|
if (refcount_table_index >= s->refcount_table_size)
|
|
|
|
return 0;
|
2014-03-08 02:10:12 +04:00
|
|
|
refcount_block_offset =
|
|
|
|
s->refcount_table[refcount_table_index] & REFT_OFFSET_MASK;
|
2009-05-28 18:07:04 +04:00
|
|
|
if (!refcount_block_offset)
|
|
|
|
return 0;
|
2011-01-10 19:17:28 +03:00
|
|
|
|
|
|
|
ret = qcow2_cache_get(bs, s->refcount_block_cache, refcount_block_offset,
|
|
|
|
(void**) &refcount_block);
|
|
|
|
if (ret < 0) {
|
|
|
|
return ret;
|
2009-05-28 18:07:04 +04:00
|
|
|
}
|
2011-01-10 19:17:28 +03:00
|
|
|
|
2009-05-28 18:07:04 +04:00
|
|
|
block_index = cluster_index &
|
|
|
|
((1 << (s->cluster_bits - REFCOUNT_SHIFT)) - 1);
|
2011-01-10 19:17:28 +03:00
|
|
|
refcount = be16_to_cpu(refcount_block[block_index]);
|
|
|
|
|
|
|
|
ret = qcow2_cache_put(bs, s->refcount_block_cache,
|
|
|
|
(void**) &refcount_block);
|
|
|
|
if (ret < 0) {
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
|
|
|
return refcount;
|
2009-05-28 18:07:04 +04:00
|
|
|
}
|
|
|
|
|
2010-02-23 18:40:52 +03:00
|
|
|
/*
|
|
|
|
* Rounds the refcount table size up to avoid growing the table for each single
|
|
|
|
* refcount block that is allocated.
|
|
|
|
*/
|
|
|
|
static unsigned int next_refcount_table_size(BDRVQcowState *s,
|
|
|
|
unsigned int min_size)
|
|
|
|
{
|
|
|
|
unsigned int min_clusters = (min_size >> (s->cluster_bits - 3)) + 1;
|
|
|
|
unsigned int refcount_table_clusters =
|
|
|
|
MAX(1, s->refcount_table_size >> (s->cluster_bits - 3));
|
|
|
|
|
|
|
|
while (min_clusters > refcount_table_clusters) {
|
|
|
|
refcount_table_clusters = (refcount_table_clusters * 3 + 1) / 2;
|
|
|
|
}
|
|
|
|
|
|
|
|
return refcount_table_clusters << (s->cluster_bits - 3);
|
|
|
|
}
|
|
|
|
|
2010-02-23 18:40:53 +03:00
|
|
|
|
|
|
|
/* Checks if two offsets are described by the same refcount block */
|
|
|
|
static int in_same_refcount_block(BDRVQcowState *s, uint64_t offset_a,
|
|
|
|
uint64_t offset_b)
|
|
|
|
{
|
|
|
|
uint64_t block_a = offset_a >> (2 * s->cluster_bits - REFCOUNT_SHIFT);
|
|
|
|
uint64_t block_b = offset_b >> (2 * s->cluster_bits - REFCOUNT_SHIFT);
|
|
|
|
|
|
|
|
return (block_a == block_b);
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Loads a refcount block. If it doesn't exist yet, it is allocated first
|
|
|
|
* (including growing the refcount table if needed).
|
|
|
|
*
|
2011-01-10 19:17:28 +03:00
|
|
|
* Returns 0 on success or -errno in error case
|
2010-02-23 18:40:53 +03:00
|
|
|
*/
|
2011-01-10 19:17:28 +03:00
|
|
|
static int alloc_refcount_block(BlockDriverState *bs,
|
|
|
|
int64_t cluster_index, uint16_t **refcount_block)
|
2009-05-28 18:07:04 +04:00
|
|
|
{
|
|
|
|
BDRVQcowState *s = bs->opaque;
|
2010-02-23 18:40:53 +03:00
|
|
|
unsigned int refcount_table_index;
|
|
|
|
int ret;
|
|
|
|
|
2010-04-14 16:17:38 +04:00
|
|
|
BLKDBG_EVENT(bs->file, BLKDBG_REFBLOCK_ALLOC);
|
2010-03-15 19:38:05 +03:00
|
|
|
|
2010-02-23 18:40:53 +03:00
|
|
|
/* Find the refcount block for the given cluster */
|
|
|
|
refcount_table_index = cluster_index >> (s->cluster_bits - REFCOUNT_SHIFT);
|
|
|
|
|
|
|
|
if (refcount_table_index < s->refcount_table_size) {
|
|
|
|
|
|
|
|
uint64_t refcount_block_offset =
|
2012-03-16 17:09:08 +04:00
|
|
|
s->refcount_table[refcount_table_index] & REFT_OFFSET_MASK;
|
2010-02-23 18:40:53 +03:00
|
|
|
|
|
|
|
/* If it's already there, we're done */
|
|
|
|
if (refcount_block_offset) {
|
2011-01-10 19:17:28 +03:00
|
|
|
return load_refcount_block(bs, refcount_block_offset,
|
|
|
|
(void**) refcount_block);
|
2010-02-23 18:40:53 +03:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* If we came here, we need to allocate something. Something is at least
|
|
|
|
* a cluster for the new refcount block. It may also include a new refcount
|
|
|
|
* table if the old refcount table is too small.
|
|
|
|
*
|
|
|
|
* Note that allocating clusters here needs some special care:
|
|
|
|
*
|
|
|
|
* - We can't use the normal qcow2_alloc_clusters(), it would try to
|
|
|
|
* increase the refcount and very likely we would end up with an endless
|
|
|
|
* recursion. Instead we must place the refcount blocks in a way that
|
|
|
|
* they can describe them themselves.
|
|
|
|
*
|
|
|
|
* - We need to consider that at this point we are inside update_refcounts
|
2014-03-28 21:06:31 +04:00
|
|
|
* and potentially doing an initial refcount increase. This means that
|
|
|
|
* some clusters have already been allocated by the caller, but their
|
|
|
|
* refcount isn't accurate yet. If we allocate clusters for metadata, we
|
|
|
|
* need to return -EAGAIN to signal the caller that it needs to restart
|
|
|
|
* the search for free clusters.
|
2010-02-23 18:40:53 +03:00
|
|
|
*
|
|
|
|
* - alloc_clusters_noref and qcow2_free_clusters may load a different
|
|
|
|
* refcount block into the cache
|
|
|
|
*/
|
|
|
|
|
2011-01-10 19:17:28 +03:00
|
|
|
*refcount_block = NULL;
|
|
|
|
|
|
|
|
/* We write to the refcount table, so we might depend on L2 tables */
|
2013-03-04 18:02:30 +04:00
|
|
|
ret = qcow2_cache_flush(bs, s->l2_table_cache);
|
|
|
|
if (ret < 0) {
|
|
|
|
return ret;
|
|
|
|
}
|
2010-02-23 18:40:53 +03:00
|
|
|
|
|
|
|
/* Allocate the refcount block itself and mark it as used */
|
2010-06-04 13:22:39 +04:00
|
|
|
int64_t new_block = alloc_clusters_noref(bs, s->cluster_size);
|
|
|
|
if (new_block < 0) {
|
|
|
|
return new_block;
|
|
|
|
}
|
2009-05-28 18:07:04 +04:00
|
|
|
|
|
|
|
#ifdef DEBUG_ALLOC2
|
2010-02-23 18:40:53 +03:00
|
|
|
fprintf(stderr, "qcow2: Allocate refcount block %d for %" PRIx64
|
|
|
|
" at %" PRIx64 "\n",
|
|
|
|
refcount_table_index, cluster_index << s->cluster_bits, new_block);
|
2009-05-28 18:07:04 +04:00
|
|
|
#endif
|
2010-02-23 18:40:53 +03:00
|
|
|
|
|
|
|
if (in_same_refcount_block(s, new_block, cluster_index << s->cluster_bits)) {
|
2010-05-28 14:05:45 +04:00
|
|
|
/* Zero the new refcount block before updating it */
|
2011-01-10 19:17:28 +03:00
|
|
|
ret = qcow2_cache_get_empty(bs, s->refcount_block_cache, new_block,
|
|
|
|
(void**) refcount_block);
|
|
|
|
if (ret < 0) {
|
|
|
|
goto fail_block;
|
|
|
|
}
|
|
|
|
|
|
|
|
memset(*refcount_block, 0, s->cluster_size);
|
2010-05-28 14:05:45 +04:00
|
|
|
|
2010-02-23 18:40:53 +03:00
|
|
|
/* The block describes itself, need to update the cache */
|
|
|
|
int block_index = (new_block >> s->cluster_bits) &
|
|
|
|
((1 << (s->cluster_bits - REFCOUNT_SHIFT)) - 1);
|
2011-01-10 19:17:28 +03:00
|
|
|
(*refcount_block)[block_index] = cpu_to_be16(1);
|
2010-02-23 18:40:53 +03:00
|
|
|
} else {
|
|
|
|
/* Described somewhere else. This can recurse at most twice before we
|
|
|
|
* arrive at a block that describes itself. */
|
2013-06-19 15:44:18 +04:00
|
|
|
ret = update_refcount(bs, new_block, s->cluster_size, 1,
|
|
|
|
QCOW2_DISCARD_NEVER);
|
2010-02-23 18:40:53 +03:00
|
|
|
if (ret < 0) {
|
|
|
|
goto fail_block;
|
|
|
|
}
|
2010-05-28 14:05:45 +04:00
|
|
|
|
2013-03-04 18:02:30 +04:00
|
|
|
ret = qcow2_cache_flush(bs, s->refcount_block_cache);
|
|
|
|
if (ret < 0) {
|
|
|
|
goto fail_block;
|
|
|
|
}
|
2010-09-17 18:36:58 +04:00
|
|
|
|
2010-05-28 14:05:45 +04:00
|
|
|
/* Initialize the new refcount block only after updating its refcount,
|
|
|
|
* update_refcount uses the refcount cache itself */
|
2011-01-10 19:17:28 +03:00
|
|
|
ret = qcow2_cache_get_empty(bs, s->refcount_block_cache, new_block,
|
|
|
|
(void**) refcount_block);
|
|
|
|
if (ret < 0) {
|
|
|
|
goto fail_block;
|
|
|
|
}
|
|
|
|
|
|
|
|
memset(*refcount_block, 0, s->cluster_size);
|
2010-02-23 18:40:53 +03:00
|
|
|
}
|
|
|
|
|
|
|
|
/* Now the new refcount block needs to be written to disk */
|
2010-04-14 16:17:38 +04:00
|
|
|
BLKDBG_EVENT(bs->file, BLKDBG_REFBLOCK_ALLOC_WRITE);
|
2011-01-10 19:17:28 +03:00
|
|
|
qcow2_cache_entry_mark_dirty(s->refcount_block_cache, *refcount_block);
|
|
|
|
ret = qcow2_cache_flush(bs, s->refcount_block_cache);
|
2010-02-23 18:40:53 +03:00
|
|
|
if (ret < 0) {
|
|
|
|
goto fail_block;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* If the refcount table is big enough, just hook the block up there */
|
|
|
|
if (refcount_table_index < s->refcount_table_size) {
|
|
|
|
uint64_t data64 = cpu_to_be64(new_block);
|
2010-04-14 16:17:38 +04:00
|
|
|
BLKDBG_EVENT(bs->file, BLKDBG_REFBLOCK_ALLOC_HOOKUP);
|
2010-06-16 19:44:35 +04:00
|
|
|
ret = bdrv_pwrite_sync(bs->file,
|
2010-02-23 18:40:53 +03:00
|
|
|
s->refcount_table_offset + refcount_table_index * sizeof(uint64_t),
|
|
|
|
&data64, sizeof(data64));
|
|
|
|
if (ret < 0) {
|
|
|
|
goto fail_block;
|
|
|
|
}
|
|
|
|
|
|
|
|
s->refcount_table[refcount_table_index] = new_block;
|
2014-03-28 21:06:31 +04:00
|
|
|
|
|
|
|
/* The new refcount block may be where the caller intended to put its
|
|
|
|
* data, so let it restart the search. */
|
|
|
|
return -EAGAIN;
|
2011-01-10 19:17:28 +03:00
|
|
|
}
|
|
|
|
|
|
|
|
ret = qcow2_cache_put(bs, s->refcount_block_cache, (void**) refcount_block);
|
|
|
|
if (ret < 0) {
|
|
|
|
goto fail_block;
|
2010-02-23 18:40:53 +03:00
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* If we come here, we need to grow the refcount table. Again, a new
|
|
|
|
* refcount table needs some space and we can't simply allocate to avoid
|
|
|
|
* endless recursion.
|
|
|
|
*
|
|
|
|
* Therefore let's grab new refcount blocks at the end of the image, which
|
|
|
|
* will describe themselves and the new refcount table. This way we can
|
|
|
|
* reference them only in the new table and do the switch to the new
|
|
|
|
* refcount table at once without producing an inconsistent state in
|
|
|
|
* between.
|
|
|
|
*/
|
2010-04-14 16:17:38 +04:00
|
|
|
BLKDBG_EVENT(bs->file, BLKDBG_REFTABLE_GROW);
|
2010-03-15 19:38:05 +03:00
|
|
|
|
2010-02-23 18:40:53 +03:00
|
|
|
/* Calculate the number of refcount blocks needed so far */
|
|
|
|
uint64_t refcount_block_clusters = 1 << (s->cluster_bits - REFCOUNT_SHIFT);
|
2014-03-28 21:06:31 +04:00
|
|
|
uint64_t blocks_used = DIV_ROUND_UP(cluster_index, refcount_block_clusters);
|
2010-02-23 18:40:53 +03:00
|
|
|
|
2014-03-26 16:05:50 +04:00
|
|
|
if (blocks_used > QCOW_MAX_REFTABLE_SIZE / sizeof(uint64_t)) {
|
|
|
|
return -EFBIG;
|
|
|
|
}
|
|
|
|
|
2010-02-23 18:40:53 +03:00
|
|
|
/* And now we need at least one block more for the new metadata */
|
|
|
|
uint64_t table_size = next_refcount_table_size(s, blocks_used + 1);
|
|
|
|
uint64_t last_table_size;
|
|
|
|
uint64_t blocks_clusters;
|
|
|
|
do {
|
2012-10-26 18:33:32 +04:00
|
|
|
uint64_t table_clusters =
|
|
|
|
size_to_clusters(s, table_size * sizeof(uint64_t));
|
2010-02-23 18:40:53 +03:00
|
|
|
blocks_clusters = 1 +
|
|
|
|
((table_clusters + refcount_block_clusters - 1)
|
|
|
|
/ refcount_block_clusters);
|
|
|
|
uint64_t meta_clusters = table_clusters + blocks_clusters;
|
|
|
|
|
|
|
|
last_table_size = table_size;
|
|
|
|
table_size = next_refcount_table_size(s, blocks_used +
|
|
|
|
((meta_clusters + refcount_block_clusters - 1)
|
|
|
|
/ refcount_block_clusters));
|
|
|
|
|
|
|
|
} while (last_table_size != table_size);
|
|
|
|
|
|
|
|
#ifdef DEBUG_ALLOC2
|
|
|
|
fprintf(stderr, "qcow2: Grow refcount table %" PRId32 " => %" PRId64 "\n",
|
|
|
|
s->refcount_table_size, table_size);
|
|
|
|
#endif
|
|
|
|
|
|
|
|
/* Create the new refcount table and blocks */
|
|
|
|
uint64_t meta_offset = (blocks_used * refcount_block_clusters) *
|
|
|
|
s->cluster_size;
|
|
|
|
uint64_t table_offset = meta_offset + blocks_clusters * s->cluster_size;
|
2011-08-21 07:09:37 +04:00
|
|
|
uint16_t *new_blocks = g_malloc0(blocks_clusters * s->cluster_size);
|
|
|
|
uint64_t *new_table = g_malloc0(table_size * sizeof(uint64_t));
|
2010-02-23 18:40:53 +03:00
|
|
|
|
|
|
|
/* Fill the new refcount table */
|
2009-05-28 18:07:04 +04:00
|
|
|
memcpy(new_table, s->refcount_table,
|
2010-02-23 18:40:53 +03:00
|
|
|
s->refcount_table_size * sizeof(uint64_t));
|
|
|
|
new_table[refcount_table_index] = new_block;
|
|
|
|
|
|
|
|
int i;
|
|
|
|
for (i = 0; i < blocks_clusters; i++) {
|
|
|
|
new_table[blocks_used + i] = meta_offset + (i * s->cluster_size);
|
|
|
|
}
|
|
|
|
|
|
|
|
/* Fill the refcount blocks */
|
|
|
|
uint64_t table_clusters = size_to_clusters(s, table_size * sizeof(uint64_t));
|
|
|
|
int block = 0;
|
|
|
|
for (i = 0; i < table_clusters + blocks_clusters; i++) {
|
|
|
|
new_blocks[block++] = cpu_to_be16(1);
|
|
|
|
}
|
|
|
|
|
|
|
|
/* Write refcount blocks to disk */
|
2010-04-14 16:17:38 +04:00
|
|
|
BLKDBG_EVENT(bs->file, BLKDBG_REFBLOCK_ALLOC_WRITE_BLOCKS);
|
2010-06-16 19:44:35 +04:00
|
|
|
ret = bdrv_pwrite_sync(bs->file, meta_offset, new_blocks,
|
2010-02-23 18:40:53 +03:00
|
|
|
blocks_clusters * s->cluster_size);
|
2011-08-21 07:09:37 +04:00
|
|
|
g_free(new_blocks);
|
2010-02-23 18:40:53 +03:00
|
|
|
if (ret < 0) {
|
|
|
|
goto fail_table;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* Write refcount table to disk */
|
|
|
|
for(i = 0; i < table_size; i++) {
|
|
|
|
cpu_to_be64s(&new_table[i]);
|
|
|
|
}
|
|
|
|
|
2010-04-14 16:17:38 +04:00
|
|
|
BLKDBG_EVENT(bs->file, BLKDBG_REFBLOCK_ALLOC_WRITE_TABLE);
|
2010-06-16 19:44:35 +04:00
|
|
|
ret = bdrv_pwrite_sync(bs->file, table_offset, new_table,
|
2010-02-23 18:40:53 +03:00
|
|
|
table_size * sizeof(uint64_t));
|
|
|
|
if (ret < 0) {
|
|
|
|
goto fail_table;
|
|
|
|
}
|
|
|
|
|
|
|
|
for(i = 0; i < table_size; i++) {
|
2012-04-28 11:38:08 +04:00
|
|
|
be64_to_cpus(&new_table[i]);
|
2010-02-23 18:40:53 +03:00
|
|
|
}
|
2009-05-28 18:07:04 +04:00
|
|
|
|
2010-02-23 18:40:53 +03:00
|
|
|
/* Hook up the new refcount table in the qcow2 header */
|
|
|
|
uint8_t data[12];
|
2009-05-28 18:07:04 +04:00
|
|
|
cpu_to_be64w((uint64_t*)data, table_offset);
|
2010-02-23 18:40:53 +03:00
|
|
|
cpu_to_be32w((uint32_t*)(data + 8), table_clusters);
|
2010-04-14 16:17:38 +04:00
|
|
|
BLKDBG_EVENT(bs->file, BLKDBG_REFBLOCK_ALLOC_SWITCH_TABLE);
|
2010-06-16 19:44:35 +04:00
|
|
|
ret = bdrv_pwrite_sync(bs->file, offsetof(QCowHeader, refcount_table_offset),
|
2010-02-23 18:40:53 +03:00
|
|
|
data, sizeof(data));
|
|
|
|
if (ret < 0) {
|
|
|
|
goto fail_table;
|
2010-01-20 17:03:03 +03:00
|
|
|
}
|
|
|
|
|
2010-02-23 18:40:53 +03:00
|
|
|
/* And switch it in memory */
|
|
|
|
uint64_t old_table_offset = s->refcount_table_offset;
|
|
|
|
uint64_t old_table_size = s->refcount_table_size;
|
|
|
|
|
2011-08-21 07:09:37 +04:00
|
|
|
g_free(s->refcount_table);
|
2009-05-28 18:07:04 +04:00
|
|
|
s->refcount_table = new_table;
|
2010-02-23 18:40:53 +03:00
|
|
|
s->refcount_table_size = table_size;
|
2009-05-28 18:07:04 +04:00
|
|
|
s->refcount_table_offset = table_offset;
|
|
|
|
|
2014-03-28 21:06:31 +04:00
|
|
|
/* Free old table. */
|
2013-06-19 15:44:18 +04:00
|
|
|
qcow2_free_clusters(bs, old_table_offset, old_table_size * sizeof(uint64_t),
|
|
|
|
QCOW2_DISCARD_OTHER);
|
2009-05-28 18:07:04 +04:00
|
|
|
|
2011-01-10 19:17:28 +03:00
|
|
|
ret = load_refcount_block(bs, new_block, (void**) refcount_block);
|
2010-02-23 18:40:53 +03:00
|
|
|
if (ret < 0) {
|
2011-01-10 19:17:28 +03:00
|
|
|
return ret;
|
2009-05-28 18:07:04 +04:00
|
|
|
}
|
|
|
|
|
2014-03-28 21:06:31 +04:00
|
|
|
/* If we were trying to do the initial refcount update for some cluster
|
|
|
|
* allocation, we might have used the same clusters to store newly
|
|
|
|
* allocated metadata. Make the caller search some new space. */
|
|
|
|
return -EAGAIN;
|
2009-05-28 18:07:04 +04:00
|
|
|
|
2010-02-23 18:40:53 +03:00
|
|
|
fail_table:
|
2011-08-21 07:09:37 +04:00
|
|
|
g_free(new_table);
|
2010-02-23 18:40:53 +03:00
|
|
|
fail_block:
|
2011-01-10 19:17:28 +03:00
|
|
|
if (*refcount_block != NULL) {
|
|
|
|
qcow2_cache_put(bs, s->refcount_block_cache, (void**) refcount_block);
|
2009-06-26 22:19:38 +04:00
|
|
|
}
|
2011-01-10 19:17:28 +03:00
|
|
|
return ret;
|
2009-06-16 13:31:30 +04:00
|
|
|
}
|
|
|
|
|
2013-06-19 15:44:20 +04:00
|
|
|
void qcow2_process_discards(BlockDriverState *bs, int ret)
|
|
|
|
{
|
|
|
|
BDRVQcowState *s = bs->opaque;
|
|
|
|
Qcow2DiscardRegion *d, *next;
|
|
|
|
|
|
|
|
QTAILQ_FOREACH_SAFE(d, &s->discards, next, next) {
|
|
|
|
QTAILQ_REMOVE(&s->discards, d, next);
|
|
|
|
|
|
|
|
/* Discard is optional, ignore the return value */
|
|
|
|
if (ret >= 0) {
|
|
|
|
bdrv_discard(bs->file,
|
|
|
|
d->offset >> BDRV_SECTOR_BITS,
|
|
|
|
d->bytes >> BDRV_SECTOR_BITS);
|
|
|
|
}
|
|
|
|
|
|
|
|
g_free(d);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
static void update_refcount_discard(BlockDriverState *bs,
|
|
|
|
uint64_t offset, uint64_t length)
|
|
|
|
{
|
|
|
|
BDRVQcowState *s = bs->opaque;
|
|
|
|
Qcow2DiscardRegion *d, *p, *next;
|
|
|
|
|
|
|
|
QTAILQ_FOREACH(d, &s->discards, next) {
|
|
|
|
uint64_t new_start = MIN(offset, d->offset);
|
|
|
|
uint64_t new_end = MAX(offset + length, d->offset + d->bytes);
|
|
|
|
|
|
|
|
if (new_end - new_start <= length + d->bytes) {
|
|
|
|
/* There can't be any overlap, areas ending up here have no
|
|
|
|
* references any more and therefore shouldn't get freed another
|
|
|
|
* time. */
|
|
|
|
assert(d->bytes + length == new_end - new_start);
|
|
|
|
d->offset = new_start;
|
|
|
|
d->bytes = new_end - new_start;
|
|
|
|
goto found;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
d = g_malloc(sizeof(*d));
|
|
|
|
*d = (Qcow2DiscardRegion) {
|
|
|
|
.bs = bs,
|
|
|
|
.offset = offset,
|
|
|
|
.bytes = length,
|
|
|
|
};
|
|
|
|
QTAILQ_INSERT_TAIL(&s->discards, d, next);
|
|
|
|
|
|
|
|
found:
|
|
|
|
/* Merge discard requests if they are adjacent now */
|
|
|
|
QTAILQ_FOREACH_SAFE(p, &s->discards, next, next) {
|
|
|
|
if (p == d
|
|
|
|
|| p->offset > d->offset + d->bytes
|
|
|
|
|| d->offset > p->offset + p->bytes)
|
|
|
|
{
|
|
|
|
continue;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* Still no overlap possible */
|
|
|
|
assert(p->offset == d->offset + d->bytes
|
|
|
|
|| d->offset == p->offset + p->bytes);
|
|
|
|
|
|
|
|
QTAILQ_REMOVE(&s->discards, p, next);
|
|
|
|
d->offset = MIN(d->offset, p->offset);
|
|
|
|
d->bytes += p->bytes;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2009-05-28 18:07:04 +04:00
|
|
|
/* XXX: cache several refcount block clusters ? */
|
2010-01-20 17:03:06 +03:00
|
|
|
static int QEMU_WARN_UNUSED_RESULT update_refcount(BlockDriverState *bs,
|
2013-06-19 15:44:18 +04:00
|
|
|
int64_t offset, int64_t length, int addend, enum qcow2_discard_type type)
|
2009-05-28 18:07:04 +04:00
|
|
|
{
|
|
|
|
BDRVQcowState *s = bs->opaque;
|
|
|
|
int64_t start, last, cluster_offset;
|
2011-01-10 19:17:28 +03:00
|
|
|
uint16_t *refcount_block = NULL;
|
|
|
|
int64_t old_table_index = -1;
|
2010-01-20 17:03:04 +03:00
|
|
|
int ret;
|
2009-05-28 18:07:04 +04:00
|
|
|
|
|
|
|
#ifdef DEBUG_ALLOC2
|
2011-08-25 11:23:55 +04:00
|
|
|
fprintf(stderr, "update_refcount: offset=%" PRId64 " size=%" PRId64 " addend=%d\n",
|
2009-05-28 18:07:04 +04:00
|
|
|
offset, length, addend);
|
|
|
|
#endif
|
2010-01-20 17:03:05 +03:00
|
|
|
if (length < 0) {
|
2009-05-28 18:07:04 +04:00
|
|
|
return -EINVAL;
|
2010-01-20 17:03:05 +03:00
|
|
|
} else if (length == 0) {
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2011-01-10 19:17:28 +03:00
|
|
|
if (addend < 0) {
|
|
|
|
qcow2_cache_set_dependency(bs, s->refcount_block_cache,
|
|
|
|
s->l2_table_cache);
|
|
|
|
}
|
|
|
|
|
2013-12-05 10:32:34 +04:00
|
|
|
start = start_of_cluster(s, offset);
|
|
|
|
last = start_of_cluster(s, offset + length - 1);
|
2009-05-28 18:07:04 +04:00
|
|
|
for(cluster_offset = start; cluster_offset <= last;
|
|
|
|
cluster_offset += s->cluster_size)
|
|
|
|
{
|
|
|
|
int block_index, refcount;
|
|
|
|
int64_t cluster_index = cluster_offset >> s->cluster_bits;
|
2011-01-10 19:17:28 +03:00
|
|
|
int64_t table_index =
|
|
|
|
cluster_index >> (s->cluster_bits - REFCOUNT_SHIFT);
|
2009-05-28 18:07:04 +04:00
|
|
|
|
2011-01-10 19:17:28 +03:00
|
|
|
/* Load the refcount block and allocate it if needed */
|
|
|
|
if (table_index != old_table_index) {
|
|
|
|
if (refcount_block) {
|
|
|
|
ret = qcow2_cache_put(bs, s->refcount_block_cache,
|
|
|
|
(void**) &refcount_block);
|
|
|
|
if (ret < 0) {
|
|
|
|
goto fail;
|
|
|
|
}
|
|
|
|
}
|
2009-06-16 13:31:30 +04:00
|
|
|
|
2011-01-10 19:17:28 +03:00
|
|
|
ret = alloc_refcount_block(bs, cluster_index, &refcount_block);
|
2010-05-28 11:47:44 +04:00
|
|
|
if (ret < 0) {
|
2011-01-10 19:17:28 +03:00
|
|
|
goto fail;
|
2009-05-28 18:07:04 +04:00
|
|
|
}
|
|
|
|
}
|
2011-01-10 19:17:28 +03:00
|
|
|
old_table_index = table_index;
|
2009-05-28 18:07:04 +04:00
|
|
|
|
2011-01-10 19:17:28 +03:00
|
|
|
qcow2_cache_entry_mark_dirty(s->refcount_block_cache, refcount_block);
|
2009-05-28 18:07:04 +04:00
|
|
|
|
|
|
|
/* we can update the count and save it */
|
|
|
|
block_index = cluster_index &
|
|
|
|
((1 << (s->cluster_bits - REFCOUNT_SHIFT)) - 1);
|
|
|
|
|
2011-01-10 19:17:28 +03:00
|
|
|
refcount = be16_to_cpu(refcount_block[block_index]);
|
2009-05-28 18:07:04 +04:00
|
|
|
refcount += addend;
|
2010-01-20 17:03:04 +03:00
|
|
|
if (refcount < 0 || refcount > 0xffff) {
|
|
|
|
ret = -EINVAL;
|
|
|
|
goto fail;
|
|
|
|
}
|
2009-05-28 18:07:04 +04:00
|
|
|
if (refcount == 0 && cluster_index < s->free_cluster_index) {
|
|
|
|
s->free_cluster_index = cluster_index;
|
|
|
|
}
|
2011-01-10 19:17:28 +03:00
|
|
|
refcount_block[block_index] = cpu_to_be16(refcount);
|
2013-06-19 15:44:20 +04:00
|
|
|
|
2013-06-19 15:44:19 +04:00
|
|
|
if (refcount == 0 && s->discard_passthrough[type]) {
|
2013-06-19 15:44:20 +04:00
|
|
|
update_refcount_discard(bs, cluster_offset, s->cluster_size);
|
2013-06-19 15:44:19 +04:00
|
|
|
}
|
2009-05-28 18:07:04 +04:00
|
|
|
}
|
|
|
|
|
2010-01-20 17:03:04 +03:00
|
|
|
ret = 0;
|
|
|
|
fail:
|
2013-06-19 15:44:20 +04:00
|
|
|
if (!s->cache_discards) {
|
|
|
|
qcow2_process_discards(bs, ret);
|
|
|
|
}
|
|
|
|
|
2009-05-28 18:07:04 +04:00
|
|
|
/* Write last changed block to disk */
|
2011-01-10 19:17:28 +03:00
|
|
|
if (refcount_block) {
|
2010-05-28 11:47:44 +04:00
|
|
|
int wret;
|
2011-01-10 19:17:28 +03:00
|
|
|
wret = qcow2_cache_put(bs, s->refcount_block_cache,
|
|
|
|
(void**) &refcount_block);
|
2010-05-28 11:47:44 +04:00
|
|
|
if (wret < 0) {
|
|
|
|
return ret < 0 ? ret : wret;
|
2009-05-28 18:07:04 +04:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2010-01-20 17:03:04 +03:00
|
|
|
/*
|
|
|
|
* Try do undo any updates if an error is returned (This may succeed in
|
|
|
|
* some cases like ENOSPC for allocating a new refcount block)
|
|
|
|
*/
|
|
|
|
if (ret < 0) {
|
|
|
|
int dummy;
|
2013-06-19 15:44:18 +04:00
|
|
|
dummy = update_refcount(bs, offset, cluster_offset - offset, -addend,
|
|
|
|
QCOW2_DISCARD_NEVER);
|
2010-10-13 22:38:07 +04:00
|
|
|
(void)dummy;
|
2010-01-20 17:03:04 +03:00
|
|
|
}
|
|
|
|
|
|
|
|
return ret;
|
2009-05-28 18:07:04 +04:00
|
|
|
}
|
|
|
|
|
2010-06-04 13:16:11 +04:00
|
|
|
/*
|
|
|
|
* Increases or decreases the refcount of a given cluster by one.
|
|
|
|
* addend must be 1 or -1.
|
|
|
|
*
|
|
|
|
* If the return value is non-negative, it is the new refcount of the cluster.
|
|
|
|
* If it is negative, it is -errno and indicates an error.
|
|
|
|
*/
|
2013-09-03 12:09:52 +04:00
|
|
|
int qcow2_update_cluster_refcount(BlockDriverState *bs,
|
|
|
|
int64_t cluster_index,
|
|
|
|
int addend,
|
|
|
|
enum qcow2_discard_type type)
|
2009-05-28 18:07:04 +04:00
|
|
|
{
|
|
|
|
BDRVQcowState *s = bs->opaque;
|
|
|
|
int ret;
|
|
|
|
|
2013-06-19 15:44:18 +04:00
|
|
|
ret = update_refcount(bs, cluster_index << s->cluster_bits, 1, addend,
|
|
|
|
type);
|
2009-05-28 18:07:04 +04:00
|
|
|
if (ret < 0) {
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
|
|
|
return get_refcount(bs, cluster_index);
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
/*********************************************************/
|
|
|
|
/* cluster allocation functions */
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
/* return < 0 if error */
|
2014-03-26 16:05:51 +04:00
|
|
|
static int64_t alloc_clusters_noref(BlockDriverState *bs, uint64_t size)
|
2009-05-28 18:07:04 +04:00
|
|
|
{
|
|
|
|
BDRVQcowState *s = bs->opaque;
|
2014-03-26 16:05:51 +04:00
|
|
|
uint64_t i, nb_clusters;
|
|
|
|
int refcount;
|
2009-05-28 18:07:04 +04:00
|
|
|
|
|
|
|
nb_clusters = size_to_clusters(s, size);
|
|
|
|
retry:
|
|
|
|
for(i = 0; i < nb_clusters; i++) {
|
2014-03-26 16:05:51 +04:00
|
|
|
uint64_t next_cluster_index = s->free_cluster_index++;
|
2010-06-04 13:22:39 +04:00
|
|
|
refcount = get_refcount(bs, next_cluster_index);
|
|
|
|
|
|
|
|
if (refcount < 0) {
|
|
|
|
return refcount;
|
|
|
|
} else if (refcount != 0) {
|
2009-05-28 18:07:04 +04:00
|
|
|
goto retry;
|
2010-06-04 13:22:39 +04:00
|
|
|
}
|
2009-05-28 18:07:04 +04:00
|
|
|
}
|
|
|
|
#ifdef DEBUG_ALLOC2
|
2011-08-25 11:23:55 +04:00
|
|
|
fprintf(stderr, "alloc_clusters: size=%" PRId64 " -> %" PRId64 "\n",
|
2009-05-28 18:07:04 +04:00
|
|
|
size,
|
|
|
|
(s->free_cluster_index - nb_clusters) << s->cluster_bits);
|
|
|
|
#endif
|
|
|
|
return (s->free_cluster_index - nb_clusters) << s->cluster_bits;
|
|
|
|
}
|
|
|
|
|
2014-03-26 16:05:51 +04:00
|
|
|
int64_t qcow2_alloc_clusters(BlockDriverState *bs, uint64_t size)
|
2009-05-28 18:07:04 +04:00
|
|
|
{
|
|
|
|
int64_t offset;
|
2010-01-20 17:03:06 +03:00
|
|
|
int ret;
|
2009-05-28 18:07:04 +04:00
|
|
|
|
2010-04-14 16:17:38 +04:00
|
|
|
BLKDBG_EVENT(bs->file, BLKDBG_CLUSTER_ALLOC);
|
2014-03-28 21:06:31 +04:00
|
|
|
do {
|
|
|
|
offset = alloc_clusters_noref(bs, size);
|
|
|
|
if (offset < 0) {
|
|
|
|
return offset;
|
|
|
|
}
|
|
|
|
|
|
|
|
ret = update_refcount(bs, offset, size, 1, QCOW2_DISCARD_NEVER);
|
|
|
|
} while (ret == -EAGAIN);
|
2010-06-04 13:22:39 +04:00
|
|
|
|
2010-01-20 17:03:06 +03:00
|
|
|
if (ret < 0) {
|
|
|
|
return ret;
|
|
|
|
}
|
2010-09-17 18:36:58 +04:00
|
|
|
|
2009-05-28 18:07:04 +04:00
|
|
|
return offset;
|
|
|
|
}
|
|
|
|
|
2012-03-02 22:35:58 +04:00
|
|
|
int qcow2_alloc_clusters_at(BlockDriverState *bs, uint64_t offset,
|
|
|
|
int nb_clusters)
|
|
|
|
{
|
|
|
|
BDRVQcowState *s = bs->opaque;
|
|
|
|
uint64_t cluster_index;
|
2014-01-26 07:12:38 +04:00
|
|
|
uint64_t i;
|
|
|
|
int refcount, ret;
|
|
|
|
|
|
|
|
assert(nb_clusters >= 0);
|
|
|
|
if (nb_clusters == 0) {
|
|
|
|
return 0;
|
|
|
|
}
|
2012-03-02 22:35:58 +04:00
|
|
|
|
2014-03-28 21:06:31 +04:00
|
|
|
do {
|
|
|
|
/* Check how many clusters there are free */
|
|
|
|
cluster_index = offset >> s->cluster_bits;
|
|
|
|
for(i = 0; i < nb_clusters; i++) {
|
|
|
|
refcount = get_refcount(bs, cluster_index++);
|
|
|
|
|
|
|
|
if (refcount < 0) {
|
|
|
|
return refcount;
|
|
|
|
} else if (refcount != 0) {
|
|
|
|
break;
|
|
|
|
}
|
2012-03-02 22:35:58 +04:00
|
|
|
}
|
|
|
|
|
2014-03-28 21:06:31 +04:00
|
|
|
/* And then allocate them */
|
|
|
|
ret = update_refcount(bs, offset, i << s->cluster_bits, 1,
|
|
|
|
QCOW2_DISCARD_NEVER);
|
|
|
|
} while (ret == -EAGAIN);
|
2012-04-20 17:50:39 +04:00
|
|
|
|
2012-03-02 22:35:58 +04:00
|
|
|
if (ret < 0) {
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
|
|
|
return i;
|
|
|
|
}
|
|
|
|
|
2009-05-28 18:07:04 +04:00
|
|
|
/* only used to allocate compressed sectors. We try to allocate
|
|
|
|
contiguous sectors. size must be <= cluster_size */
|
2009-05-28 18:07:07 +04:00
|
|
|
int64_t qcow2_alloc_bytes(BlockDriverState *bs, int size)
|
2009-05-28 18:07:04 +04:00
|
|
|
{
|
|
|
|
BDRVQcowState *s = bs->opaque;
|
|
|
|
int64_t offset, cluster_offset;
|
|
|
|
int free_in_cluster;
|
|
|
|
|
2010-04-14 16:17:38 +04:00
|
|
|
BLKDBG_EVENT(bs->file, BLKDBG_CLUSTER_ALLOC_BYTES);
|
2009-05-28 18:07:04 +04:00
|
|
|
assert(size > 0 && size <= s->cluster_size);
|
|
|
|
if (s->free_byte_offset == 0) {
|
2012-06-18 17:00:57 +04:00
|
|
|
offset = qcow2_alloc_clusters(bs, s->cluster_size);
|
|
|
|
if (offset < 0) {
|
|
|
|
return offset;
|
2010-01-20 17:04:01 +03:00
|
|
|
}
|
2012-06-18 17:00:57 +04:00
|
|
|
s->free_byte_offset = offset;
|
2009-05-28 18:07:04 +04:00
|
|
|
}
|
|
|
|
redo:
|
|
|
|
free_in_cluster = s->cluster_size -
|
2013-12-05 10:32:34 +04:00
|
|
|
offset_into_cluster(s, s->free_byte_offset);
|
2009-05-28 18:07:04 +04:00
|
|
|
if (size <= free_in_cluster) {
|
|
|
|
/* enough space in current cluster */
|
|
|
|
offset = s->free_byte_offset;
|
|
|
|
s->free_byte_offset += size;
|
|
|
|
free_in_cluster -= size;
|
|
|
|
if (free_in_cluster == 0)
|
|
|
|
s->free_byte_offset = 0;
|
2013-12-05 10:32:34 +04:00
|
|
|
if (offset_into_cluster(s, offset) != 0)
|
2013-09-03 12:09:52 +04:00
|
|
|
qcow2_update_cluster_refcount(bs, offset >> s->cluster_bits, 1,
|
|
|
|
QCOW2_DISCARD_NEVER);
|
2009-05-28 18:07:04 +04:00
|
|
|
} else {
|
2009-05-28 18:07:07 +04:00
|
|
|
offset = qcow2_alloc_clusters(bs, s->cluster_size);
|
2010-01-20 17:04:01 +03:00
|
|
|
if (offset < 0) {
|
|
|
|
return offset;
|
|
|
|
}
|
2013-12-05 10:32:34 +04:00
|
|
|
cluster_offset = start_of_cluster(s, s->free_byte_offset);
|
2009-05-28 18:07:04 +04:00
|
|
|
if ((cluster_offset + s->cluster_size) == offset) {
|
|
|
|
/* we are lucky: contiguous data */
|
|
|
|
offset = s->free_byte_offset;
|
2013-09-03 12:09:52 +04:00
|
|
|
qcow2_update_cluster_refcount(bs, offset >> s->cluster_bits, 1,
|
|
|
|
QCOW2_DISCARD_NEVER);
|
2009-05-28 18:07:04 +04:00
|
|
|
s->free_byte_offset += size;
|
|
|
|
} else {
|
|
|
|
s->free_byte_offset = offset;
|
|
|
|
goto redo;
|
|
|
|
}
|
|
|
|
}
|
2010-09-17 18:57:48 +04:00
|
|
|
|
2013-03-04 18:02:32 +04:00
|
|
|
/* The cluster refcount was incremented, either by qcow2_alloc_clusters()
|
2013-09-03 12:09:52 +04:00
|
|
|
* or explicitly by qcow2_update_cluster_refcount(). Refcount blocks must
|
|
|
|
* be flushed before the caller's L2 table updates.
|
2013-03-04 18:02:32 +04:00
|
|
|
*/
|
|
|
|
qcow2_cache_set_dependency(bs, s->l2_table_cache, s->refcount_block_cache);
|
2009-05-28 18:07:04 +04:00
|
|
|
return offset;
|
|
|
|
}
|
|
|
|
|
2009-05-28 18:07:07 +04:00
|
|
|
void qcow2_free_clusters(BlockDriverState *bs,
|
2013-06-19 15:44:18 +04:00
|
|
|
int64_t offset, int64_t size,
|
|
|
|
enum qcow2_discard_type type)
|
2009-05-28 18:07:04 +04:00
|
|
|
{
|
2010-01-20 17:03:06 +03:00
|
|
|
int ret;
|
|
|
|
|
2010-04-14 16:17:38 +04:00
|
|
|
BLKDBG_EVENT(bs->file, BLKDBG_CLUSTER_FREE);
|
2013-06-19 15:44:18 +04:00
|
|
|
ret = update_refcount(bs, offset, size, -1, type);
|
2010-01-20 17:03:06 +03:00
|
|
|
if (ret < 0) {
|
|
|
|
fprintf(stderr, "qcow2_free_clusters failed: %s\n", strerror(-ret));
|
2010-04-21 13:37:52 +04:00
|
|
|
/* TODO Remember the clusters to free them later and avoid leaking */
|
2010-01-20 17:03:06 +03:00
|
|
|
}
|
2009-05-28 18:07:04 +04:00
|
|
|
}
|
|
|
|
|
2009-05-28 18:07:05 +04:00
|
|
|
/*
|
2012-03-27 15:09:22 +04:00
|
|
|
* Free a cluster using its L2 entry (handles clusters of all types, e.g.
|
|
|
|
* normal cluster, compressed cluster, etc.)
|
2009-05-28 18:07:05 +04:00
|
|
|
*/
|
2013-06-19 15:44:18 +04:00
|
|
|
void qcow2_free_any_clusters(BlockDriverState *bs, uint64_t l2_entry,
|
|
|
|
int nb_clusters, enum qcow2_discard_type type)
|
2009-05-28 18:07:05 +04:00
|
|
|
{
|
|
|
|
BDRVQcowState *s = bs->opaque;
|
|
|
|
|
2012-03-27 15:09:22 +04:00
|
|
|
switch (qcow2_get_cluster_type(l2_entry)) {
|
|
|
|
case QCOW2_CLUSTER_COMPRESSED:
|
|
|
|
{
|
|
|
|
int nb_csectors;
|
|
|
|
nb_csectors = ((l2_entry >> s->csize_shift) &
|
|
|
|
s->csize_mask) + 1;
|
|
|
|
qcow2_free_clusters(bs,
|
|
|
|
(l2_entry & s->cluster_offset_mask) & ~511,
|
2013-06-19 15:44:18 +04:00
|
|
|
nb_csectors * 512, type);
|
2012-03-27 15:09:22 +04:00
|
|
|
}
|
|
|
|
break;
|
|
|
|
case QCOW2_CLUSTER_NORMAL:
|
2013-10-09 12:44:28 +04:00
|
|
|
case QCOW2_CLUSTER_ZERO:
|
|
|
|
if (l2_entry & L2E_OFFSET_MASK) {
|
|
|
|
qcow2_free_clusters(bs, l2_entry & L2E_OFFSET_MASK,
|
|
|
|
nb_clusters << s->cluster_bits, type);
|
|
|
|
}
|
2012-03-27 15:09:22 +04:00
|
|
|
break;
|
|
|
|
case QCOW2_CLUSTER_UNALLOCATED:
|
|
|
|
break;
|
|
|
|
default:
|
|
|
|
abort();
|
2009-05-28 18:07:05 +04:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2009-05-28 18:07:04 +04:00
|
|
|
|
|
|
|
|
|
|
|
/*********************************************************/
|
|
|
|
/* snapshots and image creation */
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
/* update the refcounts of snapshots and the copied flag */
|
2009-05-28 18:07:07 +04:00
|
|
|
int qcow2_update_snapshot_refcount(BlockDriverState *bs,
|
|
|
|
int64_t l1_table_offset, int l1_size, int addend)
|
2009-05-28 18:07:04 +04:00
|
|
|
{
|
|
|
|
BDRVQcowState *s = bs->opaque;
|
|
|
|
uint64_t *l1_table, *l2_table, l2_offset, offset, l1_size2, l1_allocated;
|
|
|
|
int64_t old_offset, old_l2_offset;
|
2011-07-19 15:01:48 +04:00
|
|
|
int i, j, l1_modified = 0, nb_csectors, refcount;
|
2011-01-10 19:17:28 +03:00
|
|
|
int ret;
|
2009-05-28 18:07:04 +04:00
|
|
|
|
|
|
|
l2_table = NULL;
|
|
|
|
l1_table = NULL;
|
|
|
|
l1_size2 = l1_size * sizeof(uint64_t);
|
2011-11-16 18:20:45 +04:00
|
|
|
|
2013-06-19 15:44:20 +04:00
|
|
|
s->cache_discards = true;
|
|
|
|
|
2011-11-16 18:20:45 +04:00
|
|
|
/* WARNING: qcow2_snapshot_goto relies on this function not using the
|
|
|
|
* l1_table_offset when it is the current s->l1_table_offset! Be careful
|
|
|
|
* when changing this! */
|
2009-05-28 18:07:04 +04:00
|
|
|
if (l1_table_offset != s->l1_table_offset) {
|
2013-01-22 14:07:56 +04:00
|
|
|
l1_table = g_malloc0(align_offset(l1_size2, 512));
|
2009-05-28 18:07:04 +04:00
|
|
|
l1_allocated = 1;
|
2013-04-05 14:51:31 +04:00
|
|
|
|
|
|
|
ret = bdrv_pread(bs->file, l1_table_offset, l1_table, l1_size2);
|
|
|
|
if (ret < 0) {
|
2009-05-28 18:07:04 +04:00
|
|
|
goto fail;
|
2011-07-19 15:01:48 +04:00
|
|
|
}
|
|
|
|
|
2009-05-28 18:07:04 +04:00
|
|
|
for(i = 0;i < l1_size; i++)
|
|
|
|
be64_to_cpus(&l1_table[i]);
|
|
|
|
} else {
|
|
|
|
assert(l1_size == s->l1_size);
|
|
|
|
l1_table = s->l1_table;
|
|
|
|
l1_allocated = 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
for(i = 0; i < l1_size; i++) {
|
|
|
|
l2_offset = l1_table[i];
|
|
|
|
if (l2_offset) {
|
|
|
|
old_l2_offset = l2_offset;
|
2012-02-23 18:40:55 +04:00
|
|
|
l2_offset &= L1E_OFFSET_MASK;
|
2011-01-10 19:17:28 +03:00
|
|
|
|
|
|
|
ret = qcow2_cache_get(bs, s->l2_table_cache, l2_offset,
|
|
|
|
(void**) &l2_table);
|
|
|
|
if (ret < 0) {
|
2009-05-28 18:07:04 +04:00
|
|
|
goto fail;
|
2011-01-10 19:17:28 +03:00
|
|
|
}
|
|
|
|
|
2009-05-28 18:07:04 +04:00
|
|
|
for(j = 0; j < s->l2_size; j++) {
|
2013-08-30 12:40:14 +04:00
|
|
|
uint64_t cluster_index;
|
|
|
|
|
2009-05-28 18:07:04 +04:00
|
|
|
offset = be64_to_cpu(l2_table[j]);
|
2013-08-30 12:40:14 +04:00
|
|
|
old_offset = offset;
|
|
|
|
offset &= ~QCOW_OFLAG_COPIED;
|
|
|
|
|
|
|
|
switch (qcow2_get_cluster_type(offset)) {
|
|
|
|
case QCOW2_CLUSTER_COMPRESSED:
|
2009-05-28 18:07:04 +04:00
|
|
|
nb_csectors = ((offset >> s->csize_shift) &
|
|
|
|
s->csize_mask) + 1;
|
2010-01-20 17:03:06 +03:00
|
|
|
if (addend != 0) {
|
|
|
|
ret = update_refcount(bs,
|
|
|
|
(offset & s->cluster_offset_mask) & ~511,
|
2013-06-19 15:44:18 +04:00
|
|
|
nb_csectors * 512, addend,
|
|
|
|
QCOW2_DISCARD_SNAPSHOT);
|
2010-01-20 17:03:06 +03:00
|
|
|
if (ret < 0) {
|
|
|
|
goto fail;
|
|
|
|
}
|
|
|
|
}
|
2009-05-28 18:07:04 +04:00
|
|
|
/* compressed clusters are never modified */
|
|
|
|
refcount = 2;
|
2013-08-30 12:40:14 +04:00
|
|
|
break;
|
|
|
|
|
|
|
|
case QCOW2_CLUSTER_NORMAL:
|
|
|
|
case QCOW2_CLUSTER_ZERO:
|
|
|
|
cluster_index = (offset & L2E_OFFSET_MASK) >> s->cluster_bits;
|
|
|
|
if (!cluster_index) {
|
|
|
|
/* unallocated */
|
|
|
|
refcount = 0;
|
|
|
|
break;
|
|
|
|
}
|
2009-05-28 18:07:04 +04:00
|
|
|
if (addend != 0) {
|
2013-09-03 12:09:52 +04:00
|
|
|
refcount = qcow2_update_cluster_refcount(bs,
|
|
|
|
cluster_index, addend,
|
|
|
|
QCOW2_DISCARD_SNAPSHOT);
|
2009-05-28 18:07:04 +04:00
|
|
|
} else {
|
2012-02-23 18:40:55 +04:00
|
|
|
refcount = get_refcount(bs, cluster_index);
|
2009-05-28 18:07:04 +04:00
|
|
|
}
|
2010-06-04 13:16:11 +04:00
|
|
|
|
|
|
|
if (refcount < 0) {
|
2013-04-05 14:51:31 +04:00
|
|
|
ret = refcount;
|
2010-06-04 13:16:11 +04:00
|
|
|
goto fail;
|
|
|
|
}
|
2013-08-30 12:40:14 +04:00
|
|
|
break;
|
2009-05-28 18:07:04 +04:00
|
|
|
|
2013-08-30 12:40:14 +04:00
|
|
|
case QCOW2_CLUSTER_UNALLOCATED:
|
|
|
|
refcount = 0;
|
|
|
|
break;
|
|
|
|
|
|
|
|
default:
|
|
|
|
abort();
|
|
|
|
}
|
|
|
|
|
|
|
|
if (refcount == 1) {
|
|
|
|
offset |= QCOW_OFLAG_COPIED;
|
|
|
|
}
|
|
|
|
if (offset != old_offset) {
|
|
|
|
if (addend > 0) {
|
|
|
|
qcow2_cache_set_dependency(bs, s->l2_table_cache,
|
|
|
|
s->refcount_block_cache);
|
2009-05-28 18:07:04 +04:00
|
|
|
}
|
2013-08-30 12:40:14 +04:00
|
|
|
l2_table[j] = cpu_to_be64(offset);
|
|
|
|
qcow2_cache_entry_mark_dirty(s->l2_table_cache, l2_table);
|
2009-05-28 18:07:04 +04:00
|
|
|
}
|
|
|
|
}
|
2011-01-10 19:17:28 +03:00
|
|
|
|
|
|
|
ret = qcow2_cache_put(bs, s->l2_table_cache, (void**) &l2_table);
|
|
|
|
if (ret < 0) {
|
|
|
|
goto fail;
|
2009-05-28 18:07:04 +04:00
|
|
|
}
|
|
|
|
|
2011-01-10 19:17:28 +03:00
|
|
|
|
2009-05-28 18:07:04 +04:00
|
|
|
if (addend != 0) {
|
2013-09-03 12:09:52 +04:00
|
|
|
refcount = qcow2_update_cluster_refcount(bs, l2_offset >>
|
|
|
|
s->cluster_bits, addend, QCOW2_DISCARD_SNAPSHOT);
|
2009-05-28 18:07:04 +04:00
|
|
|
} else {
|
|
|
|
refcount = get_refcount(bs, l2_offset >> s->cluster_bits);
|
|
|
|
}
|
2010-06-04 13:16:11 +04:00
|
|
|
if (refcount < 0) {
|
2013-04-05 14:51:31 +04:00
|
|
|
ret = refcount;
|
2010-06-04 13:16:11 +04:00
|
|
|
goto fail;
|
|
|
|
} else if (refcount == 1) {
|
2009-05-28 18:07:04 +04:00
|
|
|
l2_offset |= QCOW_OFLAG_COPIED;
|
|
|
|
}
|
|
|
|
if (l2_offset != old_l2_offset) {
|
|
|
|
l1_table[i] = l2_offset;
|
|
|
|
l1_modified = 1;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
2011-07-19 15:01:48 +04:00
|
|
|
|
2013-03-04 18:02:33 +04:00
|
|
|
ret = bdrv_flush(bs);
|
2011-07-19 15:01:48 +04:00
|
|
|
fail:
|
|
|
|
if (l2_table) {
|
|
|
|
qcow2_cache_put(bs, s->l2_table_cache, (void**) &l2_table);
|
|
|
|
}
|
|
|
|
|
2013-06-19 15:44:20 +04:00
|
|
|
s->cache_discards = false;
|
|
|
|
qcow2_process_discards(bs, ret);
|
|
|
|
|
2011-11-16 18:20:45 +04:00
|
|
|
/* Update L1 only if it isn't deleted anyway (addend = -1) */
|
2013-04-05 14:57:10 +04:00
|
|
|
if (ret == 0 && addend >= 0 && l1_modified) {
|
|
|
|
for (i = 0; i < l1_size; i++) {
|
2009-05-28 18:07:04 +04:00
|
|
|
cpu_to_be64s(&l1_table[i]);
|
2013-04-05 14:57:10 +04:00
|
|
|
}
|
|
|
|
|
|
|
|
ret = bdrv_pwrite_sync(bs->file, l1_table_offset, l1_table, l1_size2);
|
|
|
|
|
|
|
|
for (i = 0; i < l1_size; i++) {
|
2009-05-28 18:07:04 +04:00
|
|
|
be64_to_cpus(&l1_table[i]);
|
2013-04-05 14:57:10 +04:00
|
|
|
}
|
2009-05-28 18:07:04 +04:00
|
|
|
}
|
|
|
|
if (l1_allocated)
|
2011-08-21 07:09:37 +04:00
|
|
|
g_free(l1_table);
|
2011-07-19 15:01:48 +04:00
|
|
|
return ret;
|
2009-05-28 18:07:04 +04:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
/*********************************************************/
|
|
|
|
/* refcount checking functions */
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Increases the refcount for a range of clusters in a given refcount table.
|
|
|
|
* This is used to construct a temporary refcount table out of L1 and L2 tables
|
|
|
|
* which can be compared the the refcount table saved in the image.
|
|
|
|
*
|
2010-06-29 14:37:54 +04:00
|
|
|
* Modifies the number of errors in res.
|
2009-05-28 18:07:04 +04:00
|
|
|
*/
|
2010-06-29 14:37:54 +04:00
|
|
|
static void inc_refcounts(BlockDriverState *bs,
|
|
|
|
BdrvCheckResult *res,
|
2009-05-28 18:07:04 +04:00
|
|
|
uint16_t *refcount_table,
|
|
|
|
int refcount_table_size,
|
|
|
|
int64_t offset, int64_t size)
|
|
|
|
{
|
|
|
|
BDRVQcowState *s = bs->opaque;
|
2014-03-26 16:05:52 +04:00
|
|
|
uint64_t start, last, cluster_offset, k;
|
2009-05-28 18:07:04 +04:00
|
|
|
|
|
|
|
if (size <= 0)
|
2010-06-29 14:37:54 +04:00
|
|
|
return;
|
2009-05-28 18:07:04 +04:00
|
|
|
|
2013-12-05 10:32:34 +04:00
|
|
|
start = start_of_cluster(s, offset);
|
|
|
|
last = start_of_cluster(s, offset + size - 1);
|
2009-05-28 18:07:04 +04:00
|
|
|
for(cluster_offset = start; cluster_offset <= last;
|
|
|
|
cluster_offset += s->cluster_size) {
|
|
|
|
k = cluster_offset >> s->cluster_bits;
|
2014-03-26 16:05:52 +04:00
|
|
|
if (k >= refcount_table_size) {
|
2010-06-29 14:37:54 +04:00
|
|
|
fprintf(stderr, "Warning: cluster offset=0x%" PRIx64 " is after "
|
|
|
|
"the end of the image file, can't properly check refcounts.\n",
|
|
|
|
cluster_offset);
|
|
|
|
res->check_errors++;
|
2009-05-28 18:07:04 +04:00
|
|
|
} else {
|
|
|
|
if (++refcount_table[k] == 0) {
|
|
|
|
fprintf(stderr, "ERROR: overflow cluster offset=0x%" PRIx64
|
|
|
|
"\n", cluster_offset);
|
2010-06-29 14:37:54 +04:00
|
|
|
res->corruptions++;
|
2009-05-28 18:07:04 +04:00
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2013-02-07 20:15:01 +04:00
|
|
|
/* Flags for check_refcounts_l1() and check_refcounts_l2() */
|
|
|
|
enum {
|
2013-02-07 20:15:02 +04:00
|
|
|
CHECK_FRAG_INFO = 0x2, /* update BlockFragInfo counters */
|
2013-02-07 20:15:01 +04:00
|
|
|
};
|
|
|
|
|
2009-05-28 18:07:04 +04:00
|
|
|
/*
|
|
|
|
* Increases the refcount in the given refcount table for the all clusters
|
|
|
|
* referenced in the L2 table. While doing so, performs some checks on L2
|
|
|
|
* entries.
|
|
|
|
*
|
|
|
|
* Returns the number of errors found by the checks or -errno if an internal
|
|
|
|
* error occurred.
|
|
|
|
*/
|
2010-06-29 14:37:54 +04:00
|
|
|
static int check_refcounts_l2(BlockDriverState *bs, BdrvCheckResult *res,
|
2009-05-28 18:07:04 +04:00
|
|
|
uint16_t *refcount_table, int refcount_table_size, int64_t l2_offset,
|
2013-02-07 20:15:01 +04:00
|
|
|
int flags)
|
2009-05-28 18:07:04 +04:00
|
|
|
{
|
|
|
|
BDRVQcowState *s = bs->opaque;
|
2012-03-27 15:44:56 +04:00
|
|
|
uint64_t *l2_table, l2_entry;
|
2013-02-07 20:15:02 +04:00
|
|
|
uint64_t next_contiguous_offset = 0;
|
2013-08-30 16:34:27 +04:00
|
|
|
int i, l2_size, nb_csectors;
|
2009-05-28 18:07:04 +04:00
|
|
|
|
|
|
|
/* Read L2 table from disk */
|
|
|
|
l2_size = s->l2_size * sizeof(uint64_t);
|
2011-08-21 07:09:37 +04:00
|
|
|
l2_table = g_malloc(l2_size);
|
2009-05-28 18:07:04 +04:00
|
|
|
|
2010-04-14 16:17:38 +04:00
|
|
|
if (bdrv_pread(bs->file, l2_offset, l2_table, l2_size) != l2_size)
|
2009-05-28 18:07:04 +04:00
|
|
|
goto fail;
|
|
|
|
|
|
|
|
/* Do the actual checks */
|
|
|
|
for(i = 0; i < s->l2_size; i++) {
|
2012-03-27 15:44:56 +04:00
|
|
|
l2_entry = be64_to_cpu(l2_table[i]);
|
|
|
|
|
|
|
|
switch (qcow2_get_cluster_type(l2_entry)) {
|
|
|
|
case QCOW2_CLUSTER_COMPRESSED:
|
|
|
|
/* Compressed clusters don't have QCOW_OFLAG_COPIED */
|
|
|
|
if (l2_entry & QCOW_OFLAG_COPIED) {
|
|
|
|
fprintf(stderr, "ERROR: cluster %" PRId64 ": "
|
|
|
|
"copied flag must never be set for compressed "
|
|
|
|
"clusters\n", l2_entry >> s->cluster_bits);
|
|
|
|
l2_entry &= ~QCOW_OFLAG_COPIED;
|
|
|
|
res->corruptions++;
|
|
|
|
}
|
2009-05-28 18:07:04 +04:00
|
|
|
|
2012-03-27 15:44:56 +04:00
|
|
|
/* Mark cluster as used */
|
|
|
|
nb_csectors = ((l2_entry >> s->csize_shift) &
|
|
|
|
s->csize_mask) + 1;
|
|
|
|
l2_entry &= s->cluster_offset_mask;
|
|
|
|
inc_refcounts(bs, res, refcount_table, refcount_table_size,
|
|
|
|
l2_entry & ~511, nb_csectors * 512);
|
2013-02-07 20:15:02 +04:00
|
|
|
|
|
|
|
if (flags & CHECK_FRAG_INFO) {
|
|
|
|
res->bfi.allocated_clusters++;
|
2013-02-07 20:15:05 +04:00
|
|
|
res->bfi.compressed_clusters++;
|
2013-02-07 20:15:02 +04:00
|
|
|
|
|
|
|
/* Compressed clusters are fragmented by nature. Since they
|
|
|
|
* take up sub-sector space but we only have sector granularity
|
|
|
|
* I/O we need to re-read the same sectors even for adjacent
|
|
|
|
* compressed clusters.
|
|
|
|
*/
|
|
|
|
res->bfi.fragmented_clusters++;
|
|
|
|
}
|
2012-03-27 15:44:56 +04:00
|
|
|
break;
|
2009-05-28 18:07:04 +04:00
|
|
|
|
2012-03-16 18:02:38 +04:00
|
|
|
case QCOW2_CLUSTER_ZERO:
|
|
|
|
if ((l2_entry & L2E_OFFSET_MASK) == 0) {
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
/* fall through */
|
|
|
|
|
2012-03-27 15:44:56 +04:00
|
|
|
case QCOW2_CLUSTER_NORMAL:
|
|
|
|
{
|
|
|
|
uint64_t offset = l2_entry & L2E_OFFSET_MASK;
|
2009-05-28 18:07:04 +04:00
|
|
|
|
2013-02-07 20:15:02 +04:00
|
|
|
if (flags & CHECK_FRAG_INFO) {
|
|
|
|
res->bfi.allocated_clusters++;
|
|
|
|
if (next_contiguous_offset &&
|
|
|
|
offset != next_contiguous_offset) {
|
|
|
|
res->bfi.fragmented_clusters++;
|
|
|
|
}
|
|
|
|
next_contiguous_offset = offset + s->cluster_size;
|
|
|
|
}
|
|
|
|
|
2012-03-27 15:44:56 +04:00
|
|
|
/* Mark cluster as used */
|
|
|
|
inc_refcounts(bs, res, refcount_table,refcount_table_size,
|
|
|
|
offset, s->cluster_size);
|
|
|
|
|
|
|
|
/* Correct offsets are cluster aligned */
|
2013-12-05 10:32:34 +04:00
|
|
|
if (offset_into_cluster(s, offset)) {
|
2012-03-27 15:44:56 +04:00
|
|
|
fprintf(stderr, "ERROR offset=%" PRIx64 ": Cluster is not "
|
|
|
|
"properly aligned; L2 entry corrupted.\n", offset);
|
|
|
|
res->corruptions++;
|
|
|
|
}
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
|
|
|
|
case QCOW2_CLUSTER_UNALLOCATED:
|
|
|
|
break;
|
|
|
|
|
|
|
|
default:
|
|
|
|
abort();
|
2009-05-28 18:07:04 +04:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2011-08-21 07:09:37 +04:00
|
|
|
g_free(l2_table);
|
2010-06-29 14:37:54 +04:00
|
|
|
return 0;
|
2009-05-28 18:07:04 +04:00
|
|
|
|
|
|
|
fail:
|
2010-06-29 14:37:54 +04:00
|
|
|
fprintf(stderr, "ERROR: I/O error in check_refcounts_l2\n");
|
2011-08-21 07:09:37 +04:00
|
|
|
g_free(l2_table);
|
2009-05-28 18:07:04 +04:00
|
|
|
return -EIO;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Increases the refcount for the L1 table, its L2 tables and all referenced
|
|
|
|
* clusters in the given refcount table. While doing so, performs some checks
|
|
|
|
* on L1 and L2 entries.
|
|
|
|
*
|
|
|
|
* Returns the number of errors found by the checks or -errno if an internal
|
|
|
|
* error occurred.
|
|
|
|
*/
|
|
|
|
static int check_refcounts_l1(BlockDriverState *bs,
|
2010-06-29 14:37:54 +04:00
|
|
|
BdrvCheckResult *res,
|
2009-05-28 18:07:04 +04:00
|
|
|
uint16_t *refcount_table,
|
|
|
|
int refcount_table_size,
|
|
|
|
int64_t l1_table_offset, int l1_size,
|
2013-02-07 20:15:01 +04:00
|
|
|
int flags)
|
2009-05-28 18:07:04 +04:00
|
|
|
{
|
|
|
|
BDRVQcowState *s = bs->opaque;
|
|
|
|
uint64_t *l1_table, l2_offset, l1_size2;
|
2013-08-30 16:34:27 +04:00
|
|
|
int i, ret;
|
2009-05-28 18:07:04 +04:00
|
|
|
|
|
|
|
l1_size2 = l1_size * sizeof(uint64_t);
|
|
|
|
|
|
|
|
/* Mark L1 table as used */
|
2010-06-29 14:37:54 +04:00
|
|
|
inc_refcounts(bs, res, refcount_table, refcount_table_size,
|
|
|
|
l1_table_offset, l1_size2);
|
2009-05-28 18:07:04 +04:00
|
|
|
|
|
|
|
/* Read L1 table entries from disk */
|
2009-11-27 19:35:55 +03:00
|
|
|
if (l1_size2 == 0) {
|
|
|
|
l1_table = NULL;
|
|
|
|
} else {
|
2011-08-21 07:09:37 +04:00
|
|
|
l1_table = g_malloc(l1_size2);
|
2010-04-14 16:17:38 +04:00
|
|
|
if (bdrv_pread(bs->file, l1_table_offset,
|
2009-11-27 19:35:55 +03:00
|
|
|
l1_table, l1_size2) != l1_size2)
|
|
|
|
goto fail;
|
|
|
|
for(i = 0;i < l1_size; i++)
|
|
|
|
be64_to_cpus(&l1_table[i]);
|
|
|
|
}
|
2009-05-28 18:07:04 +04:00
|
|
|
|
|
|
|
/* Do the actual checks */
|
|
|
|
for(i = 0; i < l1_size; i++) {
|
|
|
|
l2_offset = l1_table[i];
|
|
|
|
if (l2_offset) {
|
|
|
|
/* Mark L2 table as used */
|
2012-03-27 15:44:56 +04:00
|
|
|
l2_offset &= L1E_OFFSET_MASK;
|
2010-06-29 14:37:54 +04:00
|
|
|
inc_refcounts(bs, res, refcount_table, refcount_table_size,
|
|
|
|
l2_offset, s->cluster_size);
|
2009-05-28 18:07:04 +04:00
|
|
|
|
|
|
|
/* L2 tables are cluster aligned */
|
2013-12-05 10:32:34 +04:00
|
|
|
if (offset_into_cluster(s, l2_offset)) {
|
2009-05-28 18:07:04 +04:00
|
|
|
fprintf(stderr, "ERROR l2_offset=%" PRIx64 ": Table is not "
|
|
|
|
"cluster aligned; L1 entry corrupted\n", l2_offset);
|
2010-06-29 14:37:54 +04:00
|
|
|
res->corruptions++;
|
2009-05-28 18:07:04 +04:00
|
|
|
}
|
|
|
|
|
|
|
|
/* Process and check L2 entries */
|
2010-06-29 14:37:54 +04:00
|
|
|
ret = check_refcounts_l2(bs, res, refcount_table,
|
2013-02-07 20:15:01 +04:00
|
|
|
refcount_table_size, l2_offset, flags);
|
2009-05-28 18:07:04 +04:00
|
|
|
if (ret < 0) {
|
|
|
|
goto fail;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
2011-08-21 07:09:37 +04:00
|
|
|
g_free(l1_table);
|
2010-06-29 14:37:54 +04:00
|
|
|
return 0;
|
2009-05-28 18:07:04 +04:00
|
|
|
|
|
|
|
fail:
|
|
|
|
fprintf(stderr, "ERROR: I/O error in check_refcounts_l1\n");
|
2010-06-29 14:37:54 +04:00
|
|
|
res->check_errors++;
|
2011-08-21 07:09:37 +04:00
|
|
|
g_free(l1_table);
|
2009-05-28 18:07:04 +04:00
|
|
|
return -EIO;
|
|
|
|
}
|
|
|
|
|
2013-08-30 16:34:27 +04:00
|
|
|
/*
|
|
|
|
* Checks the OFLAG_COPIED flag for all L1 and L2 entries.
|
|
|
|
*
|
|
|
|
* This function does not print an error message nor does it increment
|
|
|
|
* check_errors if get_refcount fails (this is because such an error will have
|
|
|
|
* been already detected and sufficiently signaled by the calling function
|
|
|
|
* (qcow2_check_refcounts) by the time this function is called).
|
|
|
|
*/
|
2013-08-30 16:34:28 +04:00
|
|
|
static int check_oflag_copied(BlockDriverState *bs, BdrvCheckResult *res,
|
|
|
|
BdrvCheckMode fix)
|
2013-08-30 16:34:27 +04:00
|
|
|
{
|
|
|
|
BDRVQcowState *s = bs->opaque;
|
|
|
|
uint64_t *l2_table = qemu_blockalign(bs, s->cluster_size);
|
|
|
|
int ret;
|
|
|
|
int refcount;
|
|
|
|
int i, j;
|
|
|
|
|
|
|
|
for (i = 0; i < s->l1_size; i++) {
|
|
|
|
uint64_t l1_entry = s->l1_table[i];
|
|
|
|
uint64_t l2_offset = l1_entry & L1E_OFFSET_MASK;
|
2013-08-30 16:34:28 +04:00
|
|
|
bool l2_dirty = false;
|
2013-08-30 16:34:27 +04:00
|
|
|
|
|
|
|
if (!l2_offset) {
|
|
|
|
continue;
|
|
|
|
}
|
|
|
|
|
|
|
|
refcount = get_refcount(bs, l2_offset >> s->cluster_bits);
|
|
|
|
if (refcount < 0) {
|
|
|
|
/* don't print message nor increment check_errors */
|
|
|
|
continue;
|
|
|
|
}
|
|
|
|
if ((refcount == 1) != ((l1_entry & QCOW_OFLAG_COPIED) != 0)) {
|
2013-08-30 16:34:28 +04:00
|
|
|
fprintf(stderr, "%s OFLAG_COPIED L2 cluster: l1_index=%d "
|
2013-08-30 16:34:27 +04:00
|
|
|
"l1_entry=%" PRIx64 " refcount=%d\n",
|
2013-08-30 16:34:28 +04:00
|
|
|
fix & BDRV_FIX_ERRORS ? "Repairing" :
|
|
|
|
"ERROR",
|
2013-08-30 16:34:27 +04:00
|
|
|
i, l1_entry, refcount);
|
2013-08-30 16:34:28 +04:00
|
|
|
if (fix & BDRV_FIX_ERRORS) {
|
|
|
|
s->l1_table[i] = refcount == 1
|
|
|
|
? l1_entry | QCOW_OFLAG_COPIED
|
|
|
|
: l1_entry & ~QCOW_OFLAG_COPIED;
|
|
|
|
ret = qcow2_write_l1_entry(bs, i);
|
|
|
|
if (ret < 0) {
|
|
|
|
res->check_errors++;
|
|
|
|
goto fail;
|
|
|
|
}
|
|
|
|
res->corruptions_fixed++;
|
|
|
|
} else {
|
|
|
|
res->corruptions++;
|
|
|
|
}
|
2013-08-30 16:34:27 +04:00
|
|
|
}
|
|
|
|
|
|
|
|
ret = bdrv_pread(bs->file, l2_offset, l2_table,
|
|
|
|
s->l2_size * sizeof(uint64_t));
|
|
|
|
if (ret < 0) {
|
|
|
|
fprintf(stderr, "ERROR: Could not read L2 table: %s\n",
|
|
|
|
strerror(-ret));
|
|
|
|
res->check_errors++;
|
|
|
|
goto fail;
|
|
|
|
}
|
|
|
|
|
|
|
|
for (j = 0; j < s->l2_size; j++) {
|
|
|
|
uint64_t l2_entry = be64_to_cpu(l2_table[j]);
|
|
|
|
uint64_t data_offset = l2_entry & L2E_OFFSET_MASK;
|
|
|
|
int cluster_type = qcow2_get_cluster_type(l2_entry);
|
|
|
|
|
|
|
|
if ((cluster_type == QCOW2_CLUSTER_NORMAL) ||
|
|
|
|
((cluster_type == QCOW2_CLUSTER_ZERO) && (data_offset != 0))) {
|
|
|
|
refcount = get_refcount(bs, data_offset >> s->cluster_bits);
|
|
|
|
if (refcount < 0) {
|
|
|
|
/* don't print message nor increment check_errors */
|
|
|
|
continue;
|
|
|
|
}
|
|
|
|
if ((refcount == 1) != ((l2_entry & QCOW_OFLAG_COPIED) != 0)) {
|
2013-08-30 16:34:28 +04:00
|
|
|
fprintf(stderr, "%s OFLAG_COPIED data cluster: "
|
2013-08-30 16:34:27 +04:00
|
|
|
"l2_entry=%" PRIx64 " refcount=%d\n",
|
2013-08-30 16:34:28 +04:00
|
|
|
fix & BDRV_FIX_ERRORS ? "Repairing" :
|
|
|
|
"ERROR",
|
2013-08-30 16:34:27 +04:00
|
|
|
l2_entry, refcount);
|
2013-08-30 16:34:28 +04:00
|
|
|
if (fix & BDRV_FIX_ERRORS) {
|
|
|
|
l2_table[j] = cpu_to_be64(refcount == 1
|
|
|
|
? l2_entry | QCOW_OFLAG_COPIED
|
|
|
|
: l2_entry & ~QCOW_OFLAG_COPIED);
|
|
|
|
l2_dirty = true;
|
|
|
|
res->corruptions_fixed++;
|
|
|
|
} else {
|
|
|
|
res->corruptions++;
|
|
|
|
}
|
2013-08-30 16:34:27 +04:00
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
2013-08-30 16:34:28 +04:00
|
|
|
|
|
|
|
if (l2_dirty) {
|
2013-10-10 13:09:23 +04:00
|
|
|
ret = qcow2_pre_write_overlap_check(bs, QCOW2_OL_ACTIVE_L2,
|
|
|
|
l2_offset, s->cluster_size);
|
2013-08-30 16:34:28 +04:00
|
|
|
if (ret < 0) {
|
|
|
|
fprintf(stderr, "ERROR: Could not write L2 table; metadata "
|
|
|
|
"overlap check failed: %s\n", strerror(-ret));
|
|
|
|
res->check_errors++;
|
|
|
|
goto fail;
|
|
|
|
}
|
|
|
|
|
|
|
|
ret = bdrv_pwrite(bs->file, l2_offset, l2_table, s->cluster_size);
|
|
|
|
if (ret < 0) {
|
|
|
|
fprintf(stderr, "ERROR: Could not write L2 table: %s\n",
|
|
|
|
strerror(-ret));
|
|
|
|
res->check_errors++;
|
|
|
|
goto fail;
|
|
|
|
}
|
|
|
|
}
|
2013-08-30 16:34:27 +04:00
|
|
|
}
|
|
|
|
|
|
|
|
ret = 0;
|
|
|
|
|
|
|
|
fail:
|
|
|
|
qemu_vfree(l2_table);
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
2013-09-02 11:25:10 +04:00
|
|
|
/*
|
|
|
|
* Writes one sector of the refcount table to the disk
|
|
|
|
*/
|
|
|
|
#define RT_ENTRIES_PER_SECTOR (512 / sizeof(uint64_t))
|
|
|
|
static int write_reftable_entry(BlockDriverState *bs, int rt_index)
|
|
|
|
{
|
|
|
|
BDRVQcowState *s = bs->opaque;
|
|
|
|
uint64_t buf[RT_ENTRIES_PER_SECTOR];
|
|
|
|
int rt_start_index;
|
|
|
|
int i, ret;
|
|
|
|
|
|
|
|
rt_start_index = rt_index & ~(RT_ENTRIES_PER_SECTOR - 1);
|
|
|
|
for (i = 0; i < RT_ENTRIES_PER_SECTOR; i++) {
|
|
|
|
buf[i] = cpu_to_be64(s->refcount_table[rt_start_index + i]);
|
|
|
|
}
|
|
|
|
|
2013-10-10 13:09:23 +04:00
|
|
|
ret = qcow2_pre_write_overlap_check(bs, QCOW2_OL_REFCOUNT_TABLE,
|
2013-09-02 11:25:10 +04:00
|
|
|
s->refcount_table_offset + rt_start_index * sizeof(uint64_t),
|
|
|
|
sizeof(buf));
|
|
|
|
if (ret < 0) {
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
|
|
|
BLKDBG_EVENT(bs->file, BLKDBG_REFTABLE_UPDATE);
|
|
|
|
ret = bdrv_pwrite_sync(bs->file, s->refcount_table_offset +
|
|
|
|
rt_start_index * sizeof(uint64_t), buf, sizeof(buf));
|
|
|
|
if (ret < 0) {
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Allocates a new cluster for the given refcount block (represented by its
|
|
|
|
* offset in the image file) and copies the current content there. This function
|
|
|
|
* does _not_ decrement the reference count for the currently occupied cluster.
|
|
|
|
*
|
|
|
|
* This function prints an informative message to stderr on error (and returns
|
2014-03-18 02:04:51 +04:00
|
|
|
* -errno); on success, the offset of the newly allocated cluster is returned.
|
2013-09-02 11:25:10 +04:00
|
|
|
*/
|
|
|
|
static int64_t realloc_refcount_block(BlockDriverState *bs, int reftable_index,
|
|
|
|
uint64_t offset)
|
|
|
|
{
|
|
|
|
BDRVQcowState *s = bs->opaque;
|
|
|
|
int64_t new_offset = 0;
|
|
|
|
void *refcount_block = NULL;
|
|
|
|
int ret;
|
|
|
|
|
|
|
|
/* allocate new refcount block */
|
|
|
|
new_offset = qcow2_alloc_clusters(bs, s->cluster_size);
|
|
|
|
if (new_offset < 0) {
|
|
|
|
fprintf(stderr, "Could not allocate new cluster: %s\n",
|
|
|
|
strerror(-new_offset));
|
|
|
|
ret = new_offset;
|
2014-03-18 02:04:52 +04:00
|
|
|
goto done;
|
2013-09-02 11:25:10 +04:00
|
|
|
}
|
|
|
|
|
|
|
|
/* fetch current refcount block content */
|
|
|
|
ret = qcow2_cache_get(bs, s->refcount_block_cache, offset, &refcount_block);
|
|
|
|
if (ret < 0) {
|
|
|
|
fprintf(stderr, "Could not fetch refcount block: %s\n", strerror(-ret));
|
2014-03-18 02:04:52 +04:00
|
|
|
goto fail_free_cluster;
|
2013-09-02 11:25:10 +04:00
|
|
|
}
|
|
|
|
|
|
|
|
/* new block has not yet been entered into refcount table, therefore it is
|
|
|
|
* no refcount block yet (regarding this check) */
|
2013-10-10 13:09:23 +04:00
|
|
|
ret = qcow2_pre_write_overlap_check(bs, 0, new_offset, s->cluster_size);
|
2013-09-02 11:25:10 +04:00
|
|
|
if (ret < 0) {
|
|
|
|
fprintf(stderr, "Could not write refcount block; metadata overlap "
|
|
|
|
"check failed: %s\n", strerror(-ret));
|
|
|
|
/* the image will be marked corrupt, so don't even attempt on freeing
|
|
|
|
* the cluster */
|
2014-03-18 02:04:52 +04:00
|
|
|
goto done;
|
2013-09-02 11:25:10 +04:00
|
|
|
}
|
|
|
|
|
|
|
|
/* write to new block */
|
|
|
|
ret = bdrv_write(bs->file, new_offset / BDRV_SECTOR_SIZE, refcount_block,
|
|
|
|
s->cluster_sectors);
|
|
|
|
if (ret < 0) {
|
|
|
|
fprintf(stderr, "Could not write refcount block: %s\n", strerror(-ret));
|
2014-03-18 02:04:52 +04:00
|
|
|
goto fail_free_cluster;
|
2013-09-02 11:25:10 +04:00
|
|
|
}
|
|
|
|
|
|
|
|
/* update refcount table */
|
2013-12-05 10:32:34 +04:00
|
|
|
assert(!offset_into_cluster(s, new_offset));
|
2013-09-02 11:25:10 +04:00
|
|
|
s->refcount_table[reftable_index] = new_offset;
|
|
|
|
ret = write_reftable_entry(bs, reftable_index);
|
|
|
|
if (ret < 0) {
|
|
|
|
fprintf(stderr, "Could not update refcount table: %s\n",
|
|
|
|
strerror(-ret));
|
2014-03-18 02:04:52 +04:00
|
|
|
goto fail_free_cluster;
|
2013-09-02 11:25:10 +04:00
|
|
|
}
|
|
|
|
|
2014-03-18 02:04:52 +04:00
|
|
|
goto done;
|
|
|
|
|
|
|
|
fail_free_cluster:
|
|
|
|
qcow2_free_clusters(bs, new_offset, s->cluster_size, QCOW2_DISCARD_OTHER);
|
|
|
|
|
|
|
|
done:
|
2013-09-02 11:25:10 +04:00
|
|
|
if (refcount_block) {
|
2014-03-18 02:04:52 +04:00
|
|
|
/* This should never fail, as it would only do so if the given refcount
|
|
|
|
* block cannot be found in the cache. As this is impossible as long as
|
|
|
|
* there are no bugs, assert the success. */
|
|
|
|
int tmp = qcow2_cache_put(bs, s->refcount_block_cache, &refcount_block);
|
|
|
|
assert(tmp == 0);
|
2013-09-02 11:25:10 +04:00
|
|
|
}
|
2014-03-18 02:04:52 +04:00
|
|
|
|
2013-09-02 11:25:10 +04:00
|
|
|
if (ret < 0) {
|
|
|
|
return ret;
|
|
|
|
}
|
2014-03-18 02:04:52 +04:00
|
|
|
|
2013-09-02 11:25:10 +04:00
|
|
|
return new_offset;
|
|
|
|
}
|
|
|
|
|
2009-05-28 18:07:04 +04:00
|
|
|
/*
|
|
|
|
* Checks an image for refcount consistency.
|
|
|
|
*
|
|
|
|
* Returns 0 if no errors are found, the number of errors in case the image is
|
2011-04-28 19:20:38 +04:00
|
|
|
* detected as corrupted, and -errno when an internal error occurred.
|
2009-05-28 18:07:04 +04:00
|
|
|
*/
|
2012-05-11 20:18:36 +04:00
|
|
|
int qcow2_check_refcounts(BlockDriverState *bs, BdrvCheckResult *res,
|
|
|
|
BdrvCheckMode fix)
|
2009-05-28 18:07:04 +04:00
|
|
|
{
|
|
|
|
BDRVQcowState *s = bs->opaque;
|
2014-03-26 16:05:52 +04:00
|
|
|
int64_t size, i, highest_cluster, nb_clusters;
|
|
|
|
int refcount1, refcount2;
|
2009-05-28 18:07:04 +04:00
|
|
|
QCowSnapshot *sn;
|
|
|
|
uint16_t *refcount_table;
|
2010-06-29 14:37:54 +04:00
|
|
|
int ret;
|
2009-05-28 18:07:04 +04:00
|
|
|
|
2010-04-14 16:17:38 +04:00
|
|
|
size = bdrv_getlength(bs->file);
|
2009-05-28 18:07:04 +04:00
|
|
|
nb_clusters = size_to_clusters(s, size);
|
2014-03-26 16:05:52 +04:00
|
|
|
if (nb_clusters > INT_MAX) {
|
|
|
|
res->check_errors++;
|
|
|
|
return -EFBIG;
|
|
|
|
}
|
|
|
|
|
2011-08-21 07:09:37 +04:00
|
|
|
refcount_table = g_malloc0(nb_clusters * sizeof(uint16_t));
|
2009-05-28 18:07:04 +04:00
|
|
|
|
2013-03-26 20:49:54 +04:00
|
|
|
res->bfi.total_clusters =
|
|
|
|
size_to_clusters(s, bs->total_sectors * BDRV_SECTOR_SIZE);
|
|
|
|
|
2009-05-28 18:07:04 +04:00
|
|
|
/* header */
|
2010-06-29 14:37:54 +04:00
|
|
|
inc_refcounts(bs, res, refcount_table, nb_clusters,
|
|
|
|
0, s->cluster_size);
|
2009-05-28 18:07:04 +04:00
|
|
|
|
|
|
|
/* current L1 table */
|
2010-06-29 14:37:54 +04:00
|
|
|
ret = check_refcounts_l1(bs, res, refcount_table, nb_clusters,
|
2013-09-30 11:21:07 +04:00
|
|
|
s->l1_table_offset, s->l1_size, CHECK_FRAG_INFO);
|
2009-05-28 18:07:04 +04:00
|
|
|
if (ret < 0) {
|
2011-06-01 12:50:00 +04:00
|
|
|
goto fail;
|
2009-05-28 18:07:04 +04:00
|
|
|
}
|
|
|
|
|
|
|
|
/* snapshots */
|
|
|
|
for(i = 0; i < s->nb_snapshots; i++) {
|
|
|
|
sn = s->snapshots + i;
|
2010-06-29 14:37:54 +04:00
|
|
|
ret = check_refcounts_l1(bs, res, refcount_table, nb_clusters,
|
|
|
|
sn->l1_table_offset, sn->l1_size, 0);
|
|
|
|
if (ret < 0) {
|
2011-06-01 12:50:00 +04:00
|
|
|
goto fail;
|
2010-06-29 14:37:54 +04:00
|
|
|
}
|
2009-05-28 18:07:04 +04:00
|
|
|
}
|
2010-06-29 14:37:54 +04:00
|
|
|
inc_refcounts(bs, res, refcount_table, nb_clusters,
|
|
|
|
s->snapshots_offset, s->snapshots_size);
|
2009-05-28 18:07:04 +04:00
|
|
|
|
|
|
|
/* refcount data */
|
2010-06-29 14:37:54 +04:00
|
|
|
inc_refcounts(bs, res, refcount_table, nb_clusters,
|
|
|
|
s->refcount_table_offset,
|
|
|
|
s->refcount_table_size * sizeof(uint64_t));
|
|
|
|
|
2009-05-28 18:07:04 +04:00
|
|
|
for(i = 0; i < s->refcount_table_size; i++) {
|
2010-06-22 14:31:45 +04:00
|
|
|
uint64_t offset, cluster;
|
2009-05-28 18:07:04 +04:00
|
|
|
offset = s->refcount_table[i];
|
2010-06-22 14:31:45 +04:00
|
|
|
cluster = offset >> s->cluster_bits;
|
2010-02-23 18:40:54 +03:00
|
|
|
|
|
|
|
/* Refcount blocks are cluster aligned */
|
2013-12-05 10:32:34 +04:00
|
|
|
if (offset_into_cluster(s, offset)) {
|
2012-05-11 20:18:36 +04:00
|
|
|
fprintf(stderr, "ERROR refcount block %" PRId64 " is not "
|
2010-02-23 18:40:54 +03:00
|
|
|
"cluster aligned; refcount table entry corrupted\n", i);
|
2010-06-29 14:37:54 +04:00
|
|
|
res->corruptions++;
|
2010-06-22 14:31:45 +04:00
|
|
|
continue;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (cluster >= nb_clusters) {
|
2012-05-11 20:18:36 +04:00
|
|
|
fprintf(stderr, "ERROR refcount block %" PRId64
|
|
|
|
" is outside image\n", i);
|
2010-06-29 14:37:54 +04:00
|
|
|
res->corruptions++;
|
2010-06-22 14:31:45 +04:00
|
|
|
continue;
|
2010-02-23 18:40:54 +03:00
|
|
|
}
|
|
|
|
|
2009-05-28 18:07:04 +04:00
|
|
|
if (offset != 0) {
|
2010-06-29 14:37:54 +04:00
|
|
|
inc_refcounts(bs, res, refcount_table, nb_clusters,
|
|
|
|
offset, s->cluster_size);
|
2010-06-22 14:31:45 +04:00
|
|
|
if (refcount_table[cluster] != 1) {
|
2013-09-02 11:25:10 +04:00
|
|
|
fprintf(stderr, "%s refcount block %" PRId64
|
2012-05-11 20:18:36 +04:00
|
|
|
" refcount=%d\n",
|
2013-09-02 11:25:10 +04:00
|
|
|
fix & BDRV_FIX_ERRORS ? "Repairing" :
|
|
|
|
"ERROR",
|
2010-06-22 14:31:45 +04:00
|
|
|
i, refcount_table[cluster]);
|
2013-09-02 11:25:10 +04:00
|
|
|
|
|
|
|
if (fix & BDRV_FIX_ERRORS) {
|
|
|
|
int64_t new_offset;
|
|
|
|
|
|
|
|
new_offset = realloc_refcount_block(bs, i, offset);
|
|
|
|
if (new_offset < 0) {
|
|
|
|
res->corruptions++;
|
|
|
|
continue;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* update refcounts */
|
|
|
|
if ((new_offset >> s->cluster_bits) >= nb_clusters) {
|
|
|
|
/* increase refcount_table size if necessary */
|
|
|
|
int old_nb_clusters = nb_clusters;
|
|
|
|
nb_clusters = (new_offset >> s->cluster_bits) + 1;
|
|
|
|
refcount_table = g_realloc(refcount_table,
|
|
|
|
nb_clusters * sizeof(uint16_t));
|
|
|
|
memset(&refcount_table[old_nb_clusters], 0, (nb_clusters
|
|
|
|
- old_nb_clusters) * sizeof(uint16_t));
|
|
|
|
}
|
|
|
|
refcount_table[cluster]--;
|
|
|
|
inc_refcounts(bs, res, refcount_table, nb_clusters,
|
|
|
|
new_offset, s->cluster_size);
|
|
|
|
|
|
|
|
res->corruptions_fixed++;
|
|
|
|
} else {
|
|
|
|
res->corruptions++;
|
|
|
|
}
|
2010-02-23 18:40:54 +03:00
|
|
|
}
|
2009-05-28 18:07:04 +04:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
/* compare ref counts */
|
2013-01-28 15:59:46 +04:00
|
|
|
for (i = 0, highest_cluster = 0; i < nb_clusters; i++) {
|
2009-05-28 18:07:04 +04:00
|
|
|
refcount1 = get_refcount(bs, i);
|
2010-06-04 13:16:11 +04:00
|
|
|
if (refcount1 < 0) {
|
2012-05-11 20:18:36 +04:00
|
|
|
fprintf(stderr, "Can't get refcount for cluster %" PRId64 ": %s\n",
|
2010-06-04 13:16:11 +04:00
|
|
|
i, strerror(-refcount1));
|
2010-06-29 14:37:54 +04:00
|
|
|
res->check_errors++;
|
2010-06-22 14:35:07 +04:00
|
|
|
continue;
|
2010-06-04 13:16:11 +04:00
|
|
|
}
|
|
|
|
|
2009-05-28 18:07:04 +04:00
|
|
|
refcount2 = refcount_table[i];
|
2013-01-28 15:59:46 +04:00
|
|
|
|
|
|
|
if (refcount1 > 0 || refcount2 > 0) {
|
|
|
|
highest_cluster = i;
|
|
|
|
}
|
|
|
|
|
2009-05-28 18:07:04 +04:00
|
|
|
if (refcount1 != refcount2) {
|
2012-05-11 20:18:36 +04:00
|
|
|
|
|
|
|
/* Check if we're allowed to fix the mismatch */
|
|
|
|
int *num_fixed = NULL;
|
|
|
|
if (refcount1 > refcount2 && (fix & BDRV_FIX_LEAKS)) {
|
|
|
|
num_fixed = &res->leaks_fixed;
|
|
|
|
} else if (refcount1 < refcount2 && (fix & BDRV_FIX_ERRORS)) {
|
|
|
|
num_fixed = &res->corruptions_fixed;
|
|
|
|
}
|
|
|
|
|
|
|
|
fprintf(stderr, "%s cluster %" PRId64 " refcount=%d reference=%d\n",
|
|
|
|
num_fixed != NULL ? "Repairing" :
|
|
|
|
refcount1 < refcount2 ? "ERROR" :
|
|
|
|
"Leaked",
|
2009-05-28 18:07:04 +04:00
|
|
|
i, refcount1, refcount2);
|
2012-05-11 20:18:36 +04:00
|
|
|
|
|
|
|
if (num_fixed) {
|
|
|
|
ret = update_refcount(bs, i << s->cluster_bits, 1,
|
2013-06-19 15:44:18 +04:00
|
|
|
refcount2 - refcount1,
|
|
|
|
QCOW2_DISCARD_ALWAYS);
|
2012-05-11 20:18:36 +04:00
|
|
|
if (ret >= 0) {
|
|
|
|
(*num_fixed)++;
|
|
|
|
continue;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
/* And if we couldn't, print an error */
|
2010-06-29 14:37:54 +04:00
|
|
|
if (refcount1 < refcount2) {
|
|
|
|
res->corruptions++;
|
|
|
|
} else {
|
|
|
|
res->leaks++;
|
|
|
|
}
|
2009-05-28 18:07:04 +04:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2013-08-30 16:34:27 +04:00
|
|
|
/* check OFLAG_COPIED */
|
2013-08-30 16:34:28 +04:00
|
|
|
ret = check_oflag_copied(bs, res, fix);
|
2013-08-30 16:34:27 +04:00
|
|
|
if (ret < 0) {
|
|
|
|
goto fail;
|
|
|
|
}
|
|
|
|
|
2013-01-28 15:59:46 +04:00
|
|
|
res->image_end_offset = (highest_cluster + 1) * s->cluster_size;
|
2011-06-01 12:50:00 +04:00
|
|
|
ret = 0;
|
|
|
|
|
|
|
|
fail:
|
2011-08-21 07:09:37 +04:00
|
|
|
g_free(refcount_table);
|
2009-05-28 18:07:04 +04:00
|
|
|
|
2011-06-01 12:50:00 +04:00
|
|
|
return ret;
|
2009-05-28 18:07:04 +04:00
|
|
|
}
|
|
|
|
|
2013-08-30 16:34:25 +04:00
|
|
|
#define overlaps_with(ofs, sz) \
|
|
|
|
ranges_overlap(offset, size, ofs, sz)
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Checks if the given offset into the image file is actually free to use by
|
|
|
|
* looking for overlaps with important metadata sections (L1/L2 tables etc.),
|
|
|
|
* i.e. a sanity check without relying on the refcount tables.
|
|
|
|
*
|
2013-10-10 13:09:23 +04:00
|
|
|
* The ign parameter specifies what checks not to perform (being a bitmask of
|
|
|
|
* QCow2MetadataOverlap values), i.e., what sections to ignore.
|
2013-08-30 16:34:25 +04:00
|
|
|
*
|
|
|
|
* Returns:
|
|
|
|
* - 0 if writing to this offset will not affect the mentioned metadata
|
|
|
|
* - a positive QCow2MetadataOverlap value indicating one overlapping section
|
|
|
|
* - a negative value (-errno) indicating an error while performing a check,
|
|
|
|
* e.g. when bdrv_read failed on QCOW2_OL_INACTIVE_L2
|
|
|
|
*/
|
2013-10-10 13:09:23 +04:00
|
|
|
int qcow2_check_metadata_overlap(BlockDriverState *bs, int ign, int64_t offset,
|
2013-08-30 16:34:25 +04:00
|
|
|
int64_t size)
|
|
|
|
{
|
|
|
|
BDRVQcowState *s = bs->opaque;
|
2013-10-10 13:09:24 +04:00
|
|
|
int chk = s->overlap_check & ~ign;
|
2013-08-30 16:34:25 +04:00
|
|
|
int i, j;
|
|
|
|
|
|
|
|
if (!size) {
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (chk & QCOW2_OL_MAIN_HEADER) {
|
|
|
|
if (offset < s->cluster_size) {
|
|
|
|
return QCOW2_OL_MAIN_HEADER;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
/* align range to test to cluster boundaries */
|
|
|
|
size = align_offset(offset_into_cluster(s, offset) + size, s->cluster_size);
|
|
|
|
offset = start_of_cluster(s, offset);
|
|
|
|
|
|
|
|
if ((chk & QCOW2_OL_ACTIVE_L1) && s->l1_size) {
|
|
|
|
if (overlaps_with(s->l1_table_offset, s->l1_size * sizeof(uint64_t))) {
|
|
|
|
return QCOW2_OL_ACTIVE_L1;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
if ((chk & QCOW2_OL_REFCOUNT_TABLE) && s->refcount_table_size) {
|
|
|
|
if (overlaps_with(s->refcount_table_offset,
|
|
|
|
s->refcount_table_size * sizeof(uint64_t))) {
|
|
|
|
return QCOW2_OL_REFCOUNT_TABLE;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
if ((chk & QCOW2_OL_SNAPSHOT_TABLE) && s->snapshots_size) {
|
|
|
|
if (overlaps_with(s->snapshots_offset, s->snapshots_size)) {
|
|
|
|
return QCOW2_OL_SNAPSHOT_TABLE;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
if ((chk & QCOW2_OL_INACTIVE_L1) && s->snapshots) {
|
|
|
|
for (i = 0; i < s->nb_snapshots; i++) {
|
|
|
|
if (s->snapshots[i].l1_size &&
|
|
|
|
overlaps_with(s->snapshots[i].l1_table_offset,
|
|
|
|
s->snapshots[i].l1_size * sizeof(uint64_t))) {
|
|
|
|
return QCOW2_OL_INACTIVE_L1;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
if ((chk & QCOW2_OL_ACTIVE_L2) && s->l1_table) {
|
|
|
|
for (i = 0; i < s->l1_size; i++) {
|
|
|
|
if ((s->l1_table[i] & L1E_OFFSET_MASK) &&
|
|
|
|
overlaps_with(s->l1_table[i] & L1E_OFFSET_MASK,
|
|
|
|
s->cluster_size)) {
|
|
|
|
return QCOW2_OL_ACTIVE_L2;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
if ((chk & QCOW2_OL_REFCOUNT_BLOCK) && s->refcount_table) {
|
|
|
|
for (i = 0; i < s->refcount_table_size; i++) {
|
|
|
|
if ((s->refcount_table[i] & REFT_OFFSET_MASK) &&
|
|
|
|
overlaps_with(s->refcount_table[i] & REFT_OFFSET_MASK,
|
|
|
|
s->cluster_size)) {
|
|
|
|
return QCOW2_OL_REFCOUNT_BLOCK;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
if ((chk & QCOW2_OL_INACTIVE_L2) && s->snapshots) {
|
|
|
|
for (i = 0; i < s->nb_snapshots; i++) {
|
|
|
|
uint64_t l1_ofs = s->snapshots[i].l1_table_offset;
|
|
|
|
uint32_t l1_sz = s->snapshots[i].l1_size;
|
2013-10-09 12:42:56 +04:00
|
|
|
uint64_t l1_sz2 = l1_sz * sizeof(uint64_t);
|
|
|
|
uint64_t *l1 = g_malloc(l1_sz2);
|
2013-08-30 16:34:25 +04:00
|
|
|
int ret;
|
|
|
|
|
2013-10-09 12:42:56 +04:00
|
|
|
ret = bdrv_pread(bs->file, l1_ofs, l1, l1_sz2);
|
2013-08-30 16:34:25 +04:00
|
|
|
if (ret < 0) {
|
|
|
|
g_free(l1);
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
|
|
|
for (j = 0; j < l1_sz; j++) {
|
2013-09-30 10:59:28 +04:00
|
|
|
uint64_t l2_ofs = be64_to_cpu(l1[j]) & L1E_OFFSET_MASK;
|
|
|
|
if (l2_ofs && overlaps_with(l2_ofs, s->cluster_size)) {
|
2013-08-30 16:34:25 +04:00
|
|
|
g_free(l1);
|
|
|
|
return QCOW2_OL_INACTIVE_L2;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
g_free(l1);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
static const char *metadata_ol_names[] = {
|
|
|
|
[QCOW2_OL_MAIN_HEADER_BITNR] = "qcow2_header",
|
|
|
|
[QCOW2_OL_ACTIVE_L1_BITNR] = "active L1 table",
|
|
|
|
[QCOW2_OL_ACTIVE_L2_BITNR] = "active L2 table",
|
|
|
|
[QCOW2_OL_REFCOUNT_TABLE_BITNR] = "refcount table",
|
|
|
|
[QCOW2_OL_REFCOUNT_BLOCK_BITNR] = "refcount block",
|
|
|
|
[QCOW2_OL_SNAPSHOT_TABLE_BITNR] = "snapshot table",
|
|
|
|
[QCOW2_OL_INACTIVE_L1_BITNR] = "inactive L1 table",
|
|
|
|
[QCOW2_OL_INACTIVE_L2_BITNR] = "inactive L2 table",
|
|
|
|
};
|
|
|
|
|
|
|
|
/*
|
|
|
|
* First performs a check for metadata overlaps (through
|
|
|
|
* qcow2_check_metadata_overlap); if that fails with a negative value (error
|
|
|
|
* while performing a check), that value is returned. If an impending overlap
|
|
|
|
* is detected, the BDS will be made unusable, the qcow2 file marked corrupt
|
|
|
|
* and -EIO returned.
|
|
|
|
*
|
|
|
|
* Returns 0 if there were neither overlaps nor errors while checking for
|
|
|
|
* overlaps; or a negative value (-errno) on error.
|
|
|
|
*/
|
2013-10-10 13:09:23 +04:00
|
|
|
int qcow2_pre_write_overlap_check(BlockDriverState *bs, int ign, int64_t offset,
|
2013-08-30 16:34:25 +04:00
|
|
|
int64_t size)
|
|
|
|
{
|
2013-10-10 13:09:23 +04:00
|
|
|
int ret = qcow2_check_metadata_overlap(bs, ign, offset, size);
|
2013-08-30 16:34:25 +04:00
|
|
|
|
|
|
|
if (ret < 0) {
|
|
|
|
return ret;
|
|
|
|
} else if (ret > 0) {
|
|
|
|
int metadata_ol_bitnr = ffs(ret) - 1;
|
|
|
|
char *message;
|
|
|
|
QObject *data;
|
|
|
|
|
|
|
|
assert(metadata_ol_bitnr < QCOW2_OL_MAX_BITNR);
|
|
|
|
|
|
|
|
fprintf(stderr, "qcow2: Preventing invalid write on metadata (overlaps "
|
|
|
|
"with %s); image marked as corrupt.\n",
|
|
|
|
metadata_ol_names[metadata_ol_bitnr]);
|
|
|
|
message = g_strdup_printf("Prevented %s overwrite",
|
|
|
|
metadata_ol_names[metadata_ol_bitnr]);
|
|
|
|
data = qobject_from_jsonf("{ 'device': %s, 'msg': %s, 'offset': %"
|
|
|
|
PRId64 ", 'size': %" PRId64 " }", bs->device_name, message,
|
|
|
|
offset, size);
|
|
|
|
monitor_protocol_event(QEVENT_BLOCK_IMAGE_CORRUPTED, data);
|
|
|
|
g_free(message);
|
|
|
|
qobject_decref(data);
|
|
|
|
|
|
|
|
qcow2_mark_corrupt(bs);
|
|
|
|
bs->drv = NULL; /* make BDS unusable */
|
|
|
|
return -EIO;
|
|
|
|
}
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|