*/
#include "qemu-common.h"
-#include "block_int.h"
+#include "block/block_int.h"
#include "block/qcow2.h"
+#include "qemu/range.h"
+#include "qapi/qmp/types.h"
static int64_t alloc_clusters_noref(BlockDriverState *bs, int64_t size);
static int QEMU_WARN_UNUSED_RESULT update_refcount(BlockDriverState *bs,
int64_t offset, int64_t length,
- int addend);
+ int addend, enum qcow2_discard_type type);
-static int cache_refcount_updates = 0;
-
-static int write_refcount_block(BDRVQcowState *s)
-{
- size_t size = s->cluster_size;
-
- if (s->refcount_block_cache_offset == 0) {
- return 0;
- }
-
- if (bdrv_pwrite(s->hd, s->refcount_block_cache_offset,
- s->refcount_block_cache, size) != size)
- {
- return -EIO;
- }
-
- return 0;
-}
-
/*********************************************************/
/* refcount handling */
BDRVQcowState *s = bs->opaque;
int ret, refcount_table_size2, i;
- s->refcount_block_cache = qemu_malloc(s->cluster_size);
refcount_table_size2 = s->refcount_table_size * sizeof(uint64_t);
- s->refcount_table = qemu_malloc(refcount_table_size2);
+ s->refcount_table = g_malloc(refcount_table_size2);
if (s->refcount_table_size > 0) {
- ret = bdrv_pread(s->hd, s->refcount_table_offset,
+ BLKDBG_EVENT(bs->file, BLKDBG_REFTABLE_LOAD);
+ ret = bdrv_pread(bs->file, s->refcount_table_offset,
s->refcount_table, refcount_table_size2);
if (ret != refcount_table_size2)
goto fail;
void qcow2_refcount_close(BlockDriverState *bs)
{
BDRVQcowState *s = bs->opaque;
- qemu_free(s->refcount_block_cache);
- qemu_free(s->refcount_table);
+ g_free(s->refcount_table);
}
static int load_refcount_block(BlockDriverState *bs,
- int64_t refcount_block_offset)
+ int64_t refcount_block_offset,
+ void **refcount_block)
{
BDRVQcowState *s = bs->opaque;
int ret;
- if (cache_refcount_updates) {
- write_refcount_block(s);
- }
+ BLKDBG_EVENT(bs->file, BLKDBG_REFBLOCK_LOAD);
+ ret = qcow2_cache_get(bs, s->refcount_block_cache, refcount_block_offset,
+ refcount_block);
- ret = bdrv_pread(s->hd, refcount_block_offset, s->refcount_block_cache,
- s->cluster_size);
- if (ret != s->cluster_size)
- return -EIO;
- s->refcount_block_cache_offset = refcount_block_offset;
- return 0;
+ return ret;
}
+/*
+ * 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.
+ */
static int get_refcount(BlockDriverState *bs, int64_t cluster_index)
{
BDRVQcowState *s = bs->opaque;
int refcount_table_index, block_index;
int64_t refcount_block_offset;
+ int ret;
+ uint16_t *refcount_block;
+ uint16_t refcount;
refcount_table_index = cluster_index >> (s->cluster_bits - REFCOUNT_SHIFT);
if (refcount_table_index >= s->refcount_table_size)
refcount_block_offset = s->refcount_table[refcount_table_index];
if (!refcount_block_offset)
return 0;
- if (refcount_block_offset != s->refcount_block_cache_offset) {
- /* better than nothing: return allocated if read error */
- if (load_refcount_block(bs, refcount_block_offset) < 0)
- return 1;
+
+ ret = qcow2_cache_get(bs, s->refcount_block_cache, refcount_block_offset,
+ (void**) &refcount_block);
+ if (ret < 0) {
+ return ret;
}
+
block_index = cluster_index &
((1 << (s->cluster_bits - REFCOUNT_SHIFT)) - 1);
- return be16_to_cpu(s->refcount_block_cache[block_index]);
+ 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;
}
/*
* Loads a refcount block. If it doesn't exist yet, it is allocated first
* (including growing the refcount table if needed).
*
- * Returns the offset of the refcount block on success or -errno in error case
+ * Returns 0 on success or -errno in error case
*/
-static int64_t alloc_refcount_block(BlockDriverState *bs, int64_t cluster_index)
+static int alloc_refcount_block(BlockDriverState *bs,
+ int64_t cluster_index, uint16_t **refcount_block)
{
BDRVQcowState *s = bs->opaque;
unsigned int refcount_table_index;
int ret;
+ BLKDBG_EVENT(bs->file, BLKDBG_REFBLOCK_ALLOC);
+
/* 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 =
- s->refcount_table[refcount_table_index];
+ s->refcount_table[refcount_table_index] & REFT_OFFSET_MASK;
/* If it's already there, we're done */
if (refcount_block_offset) {
- if (refcount_block_offset != s->refcount_block_cache_offset) {
- ret = load_refcount_block(bs, refcount_block_offset);
- if (ret < 0) {
- return ret;
- }
- }
- return refcount_block_offset;
+ return load_refcount_block(bs, refcount_block_offset,
+ (void**) refcount_block);
}
}
* refcount block into the cache
*/
- if (cache_refcount_updates) {
- ret = write_refcount_block(s);
- if (ret < 0) {
- return ret;
- }
+ *refcount_block = NULL;
+
+ /* We write to the refcount table, so we might depend on L2 tables */
+ ret = qcow2_cache_flush(bs, s->l2_table_cache);
+ if (ret < 0) {
+ return ret;
}
/* Allocate the refcount block itself and mark it as used */
- uint64_t new_block = alloc_clusters_noref(bs, s->cluster_size);
- memset(s->refcount_block_cache, 0, s->cluster_size);
- s->refcount_block_cache_offset = new_block;
+ int64_t new_block = alloc_clusters_noref(bs, s->cluster_size);
+ if (new_block < 0) {
+ return new_block;
+ }
#ifdef DEBUG_ALLOC2
fprintf(stderr, "qcow2: Allocate refcount block %d for %" PRIx64
#endif
if (in_same_refcount_block(s, new_block, cluster_index << s->cluster_bits)) {
+ /* Zero the new refcount block before updating it */
+ 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);
+
/* The block describes itself, need to update the cache */
int block_index = (new_block >> s->cluster_bits) &
((1 << (s->cluster_bits - REFCOUNT_SHIFT)) - 1);
- s->refcount_block_cache[block_index] = cpu_to_be16(1);
+ (*refcount_block)[block_index] = cpu_to_be16(1);
} else {
/* Described somewhere else. This can recurse at most twice before we
* arrive at a block that describes itself. */
- ret = update_refcount(bs, new_block, s->cluster_size, 1);
+ ret = update_refcount(bs, new_block, s->cluster_size, 1,
+ QCOW2_DISCARD_NEVER);
+ if (ret < 0) {
+ goto fail_block;
+ }
+
+ ret = qcow2_cache_flush(bs, s->refcount_block_cache);
if (ret < 0) {
goto fail_block;
}
+
+ /* Initialize the new refcount block only after updating its refcount,
+ * update_refcount uses the refcount cache itself */
+ 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);
}
/* Now the new refcount block needs to be written to disk */
- ret = bdrv_pwrite(s->hd, new_block, s->refcount_block_cache,
- s->cluster_size);
+ BLKDBG_EVENT(bs->file, BLKDBG_REFBLOCK_ALLOC_WRITE);
+ qcow2_cache_entry_mark_dirty(s->refcount_block_cache, *refcount_block);
+ ret = qcow2_cache_flush(bs, s->refcount_block_cache);
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);
- ret = bdrv_pwrite(s->hd,
+ BLKDBG_EVENT(bs->file, BLKDBG_REFBLOCK_ALLOC_HOOKUP);
+ ret = bdrv_pwrite_sync(bs->file,
s->refcount_table_offset + refcount_table_index * sizeof(uint64_t),
&data64, sizeof(data64));
if (ret < 0) {
}
s->refcount_table[refcount_table_index] = new_block;
- return new_block;
+ return 0;
+ }
+
+ ret = qcow2_cache_put(bs, s->refcount_block_cache, (void**) refcount_block);
+ if (ret < 0) {
+ goto fail_block;
}
/*
* refcount table at once without producing an inconsistent state in
* between.
*/
+ BLKDBG_EVENT(bs->file, BLKDBG_REFTABLE_GROW);
+
/* Calculate the number of refcount blocks needed so far */
uint64_t refcount_block_clusters = 1 << (s->cluster_bits - REFCOUNT_SHIFT);
uint64_t blocks_used = (s->free_cluster_index +
uint64_t last_table_size;
uint64_t blocks_clusters;
do {
- uint64_t table_clusters = size_to_clusters(s, table_size);
+ uint64_t table_clusters =
+ size_to_clusters(s, table_size * sizeof(uint64_t));
blocks_clusters = 1 +
((table_clusters + refcount_block_clusters - 1)
/ refcount_block_clusters);
uint64_t meta_offset = (blocks_used * refcount_block_clusters) *
s->cluster_size;
uint64_t table_offset = meta_offset + blocks_clusters * s->cluster_size;
- uint16_t *new_blocks = qemu_mallocz(blocks_clusters * s->cluster_size);
- uint64_t *new_table = qemu_mallocz(table_size * sizeof(uint64_t));
+ uint16_t *new_blocks = g_malloc0(blocks_clusters * s->cluster_size);
+ uint64_t *new_table = g_malloc0(table_size * sizeof(uint64_t));
assert(meta_offset >= (s->free_cluster_index * s->cluster_size));
}
/* Write refcount blocks to disk */
- ret = bdrv_pwrite(s->hd, meta_offset, new_blocks,
+ BLKDBG_EVENT(bs->file, BLKDBG_REFBLOCK_ALLOC_WRITE_BLOCKS);
+ ret = bdrv_pwrite_sync(bs->file, meta_offset, new_blocks,
blocks_clusters * s->cluster_size);
- qemu_free(new_blocks);
+ g_free(new_blocks);
if (ret < 0) {
goto fail_table;
}
cpu_to_be64s(&new_table[i]);
}
- ret = bdrv_pwrite(s->hd, table_offset, new_table,
+ BLKDBG_EVENT(bs->file, BLKDBG_REFBLOCK_ALLOC_WRITE_TABLE);
+ ret = bdrv_pwrite_sync(bs->file, table_offset, new_table,
table_size * sizeof(uint64_t));
if (ret < 0) {
goto fail_table;
}
for(i = 0; i < table_size; i++) {
- cpu_to_be64s(&new_table[i]);
+ be64_to_cpus(&new_table[i]);
}
/* Hook up the new refcount table in the qcow2 header */
uint8_t data[12];
cpu_to_be64w((uint64_t*)data, table_offset);
cpu_to_be32w((uint32_t*)(data + 8), table_clusters);
- ret = bdrv_pwrite(s->hd, offsetof(QCowHeader, refcount_table_offset),
+ BLKDBG_EVENT(bs->file, BLKDBG_REFBLOCK_ALLOC_SWITCH_TABLE);
+ ret = bdrv_pwrite_sync(bs->file, offsetof(QCowHeader, refcount_table_offset),
data, sizeof(data));
if (ret < 0) {
goto fail_table;
uint64_t old_table_offset = s->refcount_table_offset;
uint64_t old_table_size = s->refcount_table_size;
- qemu_free(s->refcount_table);
+ g_free(s->refcount_table);
s->refcount_table = new_table;
s->refcount_table_size = table_size;
s->refcount_table_offset = table_offset;
/* Free old table. Remember, we must not change free_cluster_index */
uint64_t old_free_cluster_index = s->free_cluster_index;
- qcow2_free_clusters(bs, old_table_offset, old_table_size * sizeof(uint64_t));
+ qcow2_free_clusters(bs, old_table_offset, old_table_size * sizeof(uint64_t),
+ QCOW2_DISCARD_OTHER);
s->free_cluster_index = old_free_cluster_index;
- ret = load_refcount_block(bs, new_block);
+ ret = load_refcount_block(bs, new_block, (void**) refcount_block);
if (ret < 0) {
- goto fail_block;
+ return ret;
}
- return new_block;
+ return 0;
fail_table:
- qemu_free(new_table);
+ g_free(new_table);
fail_block:
- s->refcount_block_cache_offset = 0;
+ if (*refcount_block != NULL) {
+ qcow2_cache_put(bs, s->refcount_block_cache, (void**) refcount_block);
+ }
return ret;
}
-#define REFCOUNTS_PER_SECTOR (512 >> REFCOUNT_SHIFT)
-static int write_refcount_block_entries(BDRVQcowState *s,
- int64_t refcount_block_offset, int first_index, int last_index)
+void qcow2_process_discards(BlockDriverState *bs, int ret)
{
- size_t size;
+ BDRVQcowState *s = bs->opaque;
+ Qcow2DiscardRegion *d, *next;
- if (cache_refcount_updates) {
- return 0;
- }
+ 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);
+ }
- first_index &= ~(REFCOUNTS_PER_SECTOR - 1);
- last_index = (last_index + REFCOUNTS_PER_SECTOR)
- & ~(REFCOUNTS_PER_SECTOR - 1);
+ g_free(d);
+ }
+}
- size = (last_index - first_index) << REFCOUNT_SHIFT;
- if (bdrv_pwrite(s->hd,
- refcount_block_offset + (first_index << REFCOUNT_SHIFT),
- &s->refcount_block_cache[first_index], size) != size)
- {
- return -EIO;
+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;
+ }
}
- return 0;
+ 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;
+ }
}
/* XXX: cache several refcount block clusters ? */
static int QEMU_WARN_UNUSED_RESULT update_refcount(BlockDriverState *bs,
- int64_t offset, int64_t length, int addend)
+ int64_t offset, int64_t length, int addend, enum qcow2_discard_type type)
{
BDRVQcowState *s = bs->opaque;
int64_t start, last, cluster_offset;
- int64_t refcount_block_offset = 0;
- int64_t table_index = -1, old_table_index;
- int first_index = -1, last_index = -1;
+ uint16_t *refcount_block = NULL;
+ int64_t old_table_index = -1;
int ret;
#ifdef DEBUG_ALLOC2
- printf("update_refcount: offset=%" PRId64 " size=%" PRId64 " addend=%d\n",
+ fprintf(stderr, "update_refcount: offset=%" PRId64 " size=%" PRId64 " addend=%d\n",
offset, length, addend);
#endif
if (length < 0) {
return 0;
}
+ if (addend < 0) {
+ qcow2_cache_set_dependency(bs, s->refcount_block_cache,
+ s->l2_table_cache);
+ }
+
start = offset & ~(s->cluster_size - 1);
last = (offset + length - 1) & ~(s->cluster_size - 1);
for(cluster_offset = start; cluster_offset <= last;
{
int block_index, refcount;
int64_t cluster_index = cluster_offset >> s->cluster_bits;
- int64_t new_block;
-
- /* Only write refcount block to disk when we are done with it */
- old_table_index = table_index;
- table_index = cluster_index >> (s->cluster_bits - REFCOUNT_SHIFT);
- if ((old_table_index >= 0) && (table_index != old_table_index)) {
+ int64_t table_index =
+ cluster_index >> (s->cluster_bits - REFCOUNT_SHIFT);
- if (write_refcount_block_entries(s, refcount_block_offset,
- first_index, last_index) < 0)
- {
- return -EIO;
+ /* 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;
+ }
}
- first_index = -1;
- last_index = -1;
+ ret = alloc_refcount_block(bs, cluster_index, &refcount_block);
+ if (ret < 0) {
+ goto fail;
+ }
}
+ old_table_index = table_index;
- /* Load the refcount block and allocate it if needed */
- new_block = alloc_refcount_block(bs, cluster_index);
- if (new_block < 0) {
- ret = new_block;
- goto fail;
- }
- refcount_block_offset = new_block;
+ qcow2_cache_entry_mark_dirty(s->refcount_block_cache, refcount_block);
/* we can update the count and save it */
block_index = cluster_index &
((1 << (s->cluster_bits - REFCOUNT_SHIFT)) - 1);
- if (first_index == -1 || block_index < first_index) {
- first_index = block_index;
- }
- if (block_index > last_index) {
- last_index = block_index;
- }
- refcount = be16_to_cpu(s->refcount_block_cache[block_index]);
+ refcount = be16_to_cpu(refcount_block[block_index]);
refcount += addend;
if (refcount < 0 || refcount > 0xffff) {
ret = -EINVAL;
if (refcount == 0 && cluster_index < s->free_cluster_index) {
s->free_cluster_index = cluster_index;
}
- s->refcount_block_cache[block_index] = cpu_to_be16(refcount);
+ refcount_block[block_index] = cpu_to_be16(refcount);
+
+ if (refcount == 0 && s->discard_passthrough[type]) {
+ update_refcount_discard(bs, cluster_offset, s->cluster_size);
+ }
}
ret = 0;
fail:
+ if (!s->cache_discards) {
+ qcow2_process_discards(bs, ret);
+ }
/* Write last changed block to disk */
- if (refcount_block_offset != 0) {
- if (write_refcount_block_entries(s, refcount_block_offset,
- first_index, last_index) < 0)
- {
- return ret < 0 ? ret : -EIO;
+ if (refcount_block) {
+ int wret;
+ wret = qcow2_cache_put(bs, s->refcount_block_cache,
+ (void**) &refcount_block);
+ if (wret < 0) {
+ return ret < 0 ? ret : wret;
}
}
*/
if (ret < 0) {
int dummy;
- dummy = update_refcount(bs, offset, cluster_offset - offset, -addend);
+ dummy = update_refcount(bs, offset, cluster_offset - offset, -addend,
+ QCOW2_DISCARD_NEVER);
+ (void)dummy;
}
return ret;
}
-/* addend must be 1 or -1 */
-static int update_cluster_refcount(BlockDriverState *bs,
- int64_t cluster_index,
- int addend)
+/*
+ * 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.
+ */
+int qcow2_update_cluster_refcount(BlockDriverState *bs,
+ int64_t cluster_index,
+ int addend,
+ enum qcow2_discard_type type)
{
BDRVQcowState *s = bs->opaque;
int ret;
- ret = update_refcount(bs, cluster_index << s->cluster_bits, 1, addend);
+ ret = update_refcount(bs, cluster_index << s->cluster_bits, 1, addend,
+ type);
if (ret < 0) {
return ret;
}
static int64_t alloc_clusters_noref(BlockDriverState *bs, int64_t size)
{
BDRVQcowState *s = bs->opaque;
- int i, nb_clusters;
+ int i, nb_clusters, refcount;
nb_clusters = size_to_clusters(s, size);
retry:
for(i = 0; i < nb_clusters; i++) {
- int64_t i = s->free_cluster_index++;
- if (get_refcount(bs, i) != 0)
+ int64_t next_cluster_index = s->free_cluster_index++;
+ refcount = get_refcount(bs, next_cluster_index);
+
+ if (refcount < 0) {
+ return refcount;
+ } else if (refcount != 0) {
goto retry;
+ }
}
#ifdef DEBUG_ALLOC2
- printf("alloc_clusters: size=%" PRId64 " -> %" PRId64 "\n",
+ fprintf(stderr, "alloc_clusters: size=%" PRId64 " -> %" PRId64 "\n",
size,
(s->free_cluster_index - nb_clusters) << s->cluster_bits);
#endif
int64_t offset;
int ret;
+ BLKDBG_EVENT(bs->file, BLKDBG_CLUSTER_ALLOC);
offset = alloc_clusters_noref(bs, size);
- ret = update_refcount(bs, offset, size, 1);
+ if (offset < 0) {
+ return offset;
+ }
+
+ ret = update_refcount(bs, offset, size, 1, QCOW2_DISCARD_NEVER);
if (ret < 0) {
return ret;
}
+
return offset;
}
+int qcow2_alloc_clusters_at(BlockDriverState *bs, uint64_t offset,
+ int nb_clusters)
+{
+ BDRVQcowState *s = bs->opaque;
+ uint64_t cluster_index;
+ uint64_t old_free_cluster_index;
+ int i, refcount, ret;
+
+ /* 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;
+ }
+ }
+
+ /* And then allocate them */
+ old_free_cluster_index = s->free_cluster_index;
+ s->free_cluster_index = cluster_index + i;
+
+ ret = update_refcount(bs, offset, i << s->cluster_bits, 1,
+ QCOW2_DISCARD_NEVER);
+ if (ret < 0) {
+ return ret;
+ }
+
+ s->free_cluster_index = old_free_cluster_index;
+
+ return i;
+}
+
/* only used to allocate compressed sectors. We try to allocate
contiguous sectors. size must be <= cluster_size */
int64_t qcow2_alloc_bytes(BlockDriverState *bs, int size)
int64_t offset, cluster_offset;
int free_in_cluster;
+ BLKDBG_EVENT(bs->file, BLKDBG_CLUSTER_ALLOC_BYTES);
assert(size > 0 && size <= s->cluster_size);
if (s->free_byte_offset == 0) {
- s->free_byte_offset = qcow2_alloc_clusters(bs, s->cluster_size);
- if (s->free_byte_offset < 0) {
- return s->free_byte_offset;
+ offset = qcow2_alloc_clusters(bs, s->cluster_size);
+ if (offset < 0) {
+ return offset;
}
+ s->free_byte_offset = offset;
}
redo:
free_in_cluster = s->cluster_size -
if (free_in_cluster == 0)
s->free_byte_offset = 0;
if ((offset & (s->cluster_size - 1)) != 0)
- update_cluster_refcount(bs, offset >> s->cluster_bits, 1);
+ qcow2_update_cluster_refcount(bs, offset >> s->cluster_bits, 1,
+ QCOW2_DISCARD_NEVER);
} else {
offset = qcow2_alloc_clusters(bs, s->cluster_size);
if (offset < 0) {
if ((cluster_offset + s->cluster_size) == offset) {
/* we are lucky: contiguous data */
offset = s->free_byte_offset;
- update_cluster_refcount(bs, offset >> s->cluster_bits, 1);
+ qcow2_update_cluster_refcount(bs, offset >> s->cluster_bits, 1,
+ QCOW2_DISCARD_NEVER);
s->free_byte_offset += size;
} else {
s->free_byte_offset = offset;
goto redo;
}
}
+
+ /* The cluster refcount was incremented, either by qcow2_alloc_clusters()
+ * or explicitly by qcow2_update_cluster_refcount(). Refcount blocks must
+ * be flushed before the caller's L2 table updates.
+ */
+ qcow2_cache_set_dependency(bs, s->l2_table_cache, s->refcount_block_cache);
return offset;
}
void qcow2_free_clusters(BlockDriverState *bs,
- int64_t offset, int64_t size)
+ int64_t offset, int64_t size,
+ enum qcow2_discard_type type)
{
int ret;
- ret = update_refcount(bs, offset, size, -1);
+ BLKDBG_EVENT(bs->file, BLKDBG_CLUSTER_FREE);
+ ret = update_refcount(bs, offset, size, -1, type);
if (ret < 0) {
fprintf(stderr, "qcow2_free_clusters failed: %s\n", strerror(-ret));
- abort();
+ /* TODO Remember the clusters to free them later and avoid leaking */
}
}
/*
- * free_any_clusters
- *
- * free clusters according to its type: compressed or not
- *
+ * Free a cluster using its L2 entry (handles clusters of all types, e.g.
+ * normal cluster, compressed cluster, etc.)
*/
-
-void qcow2_free_any_clusters(BlockDriverState *bs,
- uint64_t cluster_offset, int nb_clusters)
+void qcow2_free_any_clusters(BlockDriverState *bs, uint64_t l2_entry,
+ int nb_clusters, enum qcow2_discard_type type)
{
BDRVQcowState *s = bs->opaque;
- /* free the cluster */
-
- if (cluster_offset & QCOW_OFLAG_COMPRESSED) {
- int nb_csectors;
- nb_csectors = ((cluster_offset >> s->csize_shift) &
- s->csize_mask) + 1;
- qcow2_free_clusters(bs,
- (cluster_offset & s->cluster_offset_mask) & ~511,
- nb_csectors * 512);
- return;
+ 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,
+ nb_csectors * 512, type);
+ }
+ break;
+ case QCOW2_CLUSTER_NORMAL:
+ 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);
+ }
+ break;
+ case QCOW2_CLUSTER_UNALLOCATED:
+ break;
+ default:
+ abort();
}
-
- qcow2_free_clusters(bs, cluster_offset, nb_clusters << s->cluster_bits);
-
- return;
}
-void qcow2_create_refcount_update(QCowCreateState *s, int64_t offset,
- int64_t size)
-{
- int refcount;
- int64_t start, last, cluster_offset;
- uint16_t *p;
-
- start = offset & ~(s->cluster_size - 1);
- last = (offset + size - 1) & ~(s->cluster_size - 1);
- for(cluster_offset = start; cluster_offset <= last;
- cluster_offset += s->cluster_size) {
- p = &s->refcount_block[cluster_offset >> s->cluster_bits];
- refcount = be16_to_cpu(*p);
- refcount++;
- *p = cpu_to_be16(refcount);
- }
-}
-
/* update the refcounts of snapshots and the copied flag */
int qcow2_update_snapshot_refcount(BlockDriverState *bs,
int64_t l1_table_offset, int l1_size, int addend)
BDRVQcowState *s = bs->opaque;
uint64_t *l1_table, *l2_table, l2_offset, offset, l1_size2, l1_allocated;
int64_t old_offset, old_l2_offset;
- int l2_size, i, j, l1_modified, l2_modified, nb_csectors, refcount;
-
- qcow2_l2_cache_reset(bs);
- cache_refcount_updates = 1;
+ int i, j, l1_modified = 0, nb_csectors, refcount;
+ int ret;
l2_table = NULL;
l1_table = NULL;
l1_size2 = l1_size * sizeof(uint64_t);
+
+ s->cache_discards = true;
+
+ /* 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! */
if (l1_table_offset != s->l1_table_offset) {
- if (l1_size2 != 0) {
- l1_table = qemu_mallocz(align_offset(l1_size2, 512));
- } else {
- l1_table = NULL;
- }
+ l1_table = g_malloc0(align_offset(l1_size2, 512));
l1_allocated = 1;
- if (bdrv_pread(s->hd, l1_table_offset,
- l1_table, l1_size2) != l1_size2)
+
+ ret = bdrv_pread(bs->file, l1_table_offset, l1_table, l1_size2);
+ if (ret < 0) {
goto fail;
+ }
+
for(i = 0;i < l1_size; i++)
be64_to_cpus(&l1_table[i]);
} else {
l1_allocated = 0;
}
- l2_size = s->l2_size * sizeof(uint64_t);
- l2_table = qemu_malloc(l2_size);
- l1_modified = 0;
for(i = 0; i < l1_size; i++) {
l2_offset = l1_table[i];
if (l2_offset) {
old_l2_offset = l2_offset;
- l2_offset &= ~QCOW_OFLAG_COPIED;
- l2_modified = 0;
- if (bdrv_pread(s->hd, l2_offset, l2_table, l2_size) != l2_size)
+ l2_offset &= L1E_OFFSET_MASK;
+
+ ret = qcow2_cache_get(bs, s->l2_table_cache, l2_offset,
+ (void**) &l2_table);
+ if (ret < 0) {
goto fail;
+ }
+
for(j = 0; j < s->l2_size; j++) {
+ uint64_t cluster_index;
+
offset = be64_to_cpu(l2_table[j]);
- if (offset != 0) {
- old_offset = offset;
- offset &= ~QCOW_OFLAG_COPIED;
- if (offset & QCOW_OFLAG_COMPRESSED) {
+ old_offset = offset;
+ offset &= ~QCOW_OFLAG_COPIED;
+
+ switch (qcow2_get_cluster_type(offset)) {
+ case QCOW2_CLUSTER_COMPRESSED:
nb_csectors = ((offset >> s->csize_shift) &
s->csize_mask) + 1;
if (addend != 0) {
- int ret;
ret = update_refcount(bs,
(offset & s->cluster_offset_mask) & ~511,
- nb_csectors * 512, addend);
+ nb_csectors * 512, addend,
+ QCOW2_DISCARD_SNAPSHOT);
if (ret < 0) {
goto fail;
}
}
/* compressed clusters are never modified */
refcount = 2;
- } else {
+ 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;
+ }
if (addend != 0) {
- refcount = update_cluster_refcount(bs, offset >> s->cluster_bits, addend);
+ refcount = qcow2_update_cluster_refcount(bs,
+ cluster_index, addend,
+ QCOW2_DISCARD_SNAPSHOT);
} else {
- refcount = get_refcount(bs, offset >> s->cluster_bits);
+ refcount = get_refcount(bs, cluster_index);
}
- }
- if (refcount == 1) {
- offset |= QCOW_OFLAG_COPIED;
- }
- if (offset != old_offset) {
- l2_table[j] = cpu_to_be64(offset);
- l2_modified = 1;
+ if (refcount < 0) {
+ ret = refcount;
+ goto fail;
+ }
+ break;
+
+ 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);
}
+ l2_table[j] = cpu_to_be64(offset);
+ qcow2_cache_entry_mark_dirty(s->l2_table_cache, l2_table);
}
}
- if (l2_modified) {
- if (bdrv_pwrite(s->hd,
- l2_offset, l2_table, l2_size) != l2_size)
- goto fail;
+
+ ret = qcow2_cache_put(bs, s->l2_table_cache, (void**) &l2_table);
+ if (ret < 0) {
+ goto fail;
}
+
if (addend != 0) {
- refcount = update_cluster_refcount(bs, l2_offset >> s->cluster_bits, addend);
+ refcount = qcow2_update_cluster_refcount(bs, l2_offset >>
+ s->cluster_bits, addend, QCOW2_DISCARD_SNAPSHOT);
} else {
refcount = get_refcount(bs, l2_offset >> s->cluster_bits);
}
- if (refcount == 1) {
+ if (refcount < 0) {
+ ret = refcount;
+ goto fail;
+ } else if (refcount == 1) {
l2_offset |= QCOW_OFLAG_COPIED;
}
if (l2_offset != old_l2_offset) {
}
}
}
- if (l1_modified) {
- for(i = 0; i < l1_size; i++)
+
+ ret = bdrv_flush(bs);
+fail:
+ if (l2_table) {
+ qcow2_cache_put(bs, s->l2_table_cache, (void**) &l2_table);
+ }
+
+ s->cache_discards = false;
+ qcow2_process_discards(bs, ret);
+
+ /* Update L1 only if it isn't deleted anyway (addend = -1) */
+ if (ret == 0 && addend >= 0 && l1_modified) {
+ for (i = 0; i < l1_size; i++) {
cpu_to_be64s(&l1_table[i]);
- if (bdrv_pwrite(s->hd, l1_table_offset, l1_table,
- l1_size2) != l1_size2)
- goto fail;
- for(i = 0; i < l1_size; i++)
+ }
+
+ ret = bdrv_pwrite_sync(bs->file, l1_table_offset, l1_table, l1_size2);
+
+ for (i = 0; i < l1_size; i++) {
be64_to_cpus(&l1_table[i]);
+ }
}
if (l1_allocated)
- qemu_free(l1_table);
- qemu_free(l2_table);
- cache_refcount_updates = 0;
- write_refcount_block(s);
- return 0;
- fail:
- if (l1_allocated)
- qemu_free(l1_table);
- qemu_free(l2_table);
- cache_refcount_updates = 0;
- write_refcount_block(s);
- return -EIO;
+ g_free(l1_table);
+ return ret;
}
* 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.
*
- * Returns the number of errors in the image that were found
+ * Modifies the number of errors in res.
*/
-static int inc_refcounts(BlockDriverState *bs,
+static void inc_refcounts(BlockDriverState *bs,
+ BdrvCheckResult *res,
uint16_t *refcount_table,
int refcount_table_size,
int64_t offset, int64_t size)
BDRVQcowState *s = bs->opaque;
int64_t start, last, cluster_offset;
int k;
- int errors = 0;
if (size <= 0)
- return 0;
+ return;
start = offset & ~(s->cluster_size - 1);
last = (offset + size - 1) & ~(s->cluster_size - 1);
for(cluster_offset = start; cluster_offset <= last;
cluster_offset += s->cluster_size) {
k = cluster_offset >> s->cluster_bits;
- if (k < 0 || k >= refcount_table_size) {
+ if (k < 0) {
fprintf(stderr, "ERROR: invalid cluster offset=0x%" PRIx64 "\n",
cluster_offset);
- errors++;
+ res->corruptions++;
+ } else if (k >= refcount_table_size) {
+ 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++;
} else {
if (++refcount_table[k] == 0) {
fprintf(stderr, "ERROR: overflow cluster offset=0x%" PRIx64
"\n", cluster_offset);
- errors++;
+ res->corruptions++;
}
}
}
-
- return errors;
}
+/* Flags for check_refcounts_l1() and check_refcounts_l2() */
+enum {
+ CHECK_FRAG_INFO = 0x2, /* update BlockFragInfo counters */
+};
+
/*
* 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
* Returns the number of errors found by the checks or -errno if an internal
* error occurred.
*/
-static int check_refcounts_l2(BlockDriverState *bs,
+static int check_refcounts_l2(BlockDriverState *bs, BdrvCheckResult *res,
uint16_t *refcount_table, int refcount_table_size, int64_t l2_offset,
- int check_copied)
+ int flags)
{
BDRVQcowState *s = bs->opaque;
- uint64_t *l2_table, offset;
- int i, l2_size, nb_csectors, refcount;
- int errors = 0;
+ uint64_t *l2_table, l2_entry;
+ uint64_t next_contiguous_offset = 0;
+ int i, l2_size, nb_csectors;
/* Read L2 table from disk */
l2_size = s->l2_size * sizeof(uint64_t);
- l2_table = qemu_malloc(l2_size);
+ l2_table = g_malloc(l2_size);
- if (bdrv_pread(s->hd, l2_offset, l2_table, l2_size) != l2_size)
+ if (bdrv_pread(bs->file, l2_offset, l2_table, l2_size) != l2_size)
goto fail;
/* Do the actual checks */
for(i = 0; i < s->l2_size; i++) {
- offset = be64_to_cpu(l2_table[i]);
- if (offset != 0) {
- if (offset & QCOW_OFLAG_COMPRESSED) {
- /* Compressed clusters don't have QCOW_OFLAG_COPIED */
- if (offset & QCOW_OFLAG_COPIED) {
- fprintf(stderr, "ERROR: cluster %" PRId64 ": "
- "copied flag must never be set for compressed "
- "clusters\n", offset >> s->cluster_bits);
- offset &= ~QCOW_OFLAG_COPIED;
- errors++;
- }
+ 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++;
+ }
- /* Mark cluster as used */
- nb_csectors = ((offset >> s->csize_shift) &
- s->csize_mask) + 1;
- offset &= s->cluster_offset_mask;
- errors += inc_refcounts(bs, refcount_table,
- refcount_table_size,
- offset & ~511, nb_csectors * 512);
- } else {
- /* QCOW_OFLAG_COPIED must be set iff refcount == 1 */
- if (check_copied) {
- uint64_t entry = offset;
- offset &= ~QCOW_OFLAG_COPIED;
- refcount = get_refcount(bs, offset >> s->cluster_bits);
- if ((refcount == 1) != ((entry & QCOW_OFLAG_COPIED) != 0)) {
- fprintf(stderr, "ERROR OFLAG_COPIED: offset=%"
- PRIx64 " refcount=%d\n", entry, refcount);
- errors++;
- }
- }
+ /* 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);
+
+ if (flags & CHECK_FRAG_INFO) {
+ res->bfi.allocated_clusters++;
+ res->bfi.compressed_clusters++;
+
+ /* 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++;
+ }
+ break;
- /* Mark cluster as used */
- offset &= ~QCOW_OFLAG_COPIED;
- errors += inc_refcounts(bs, refcount_table,
- refcount_table_size,
- offset, s->cluster_size);
-
- /* Correct offsets are cluster aligned */
- if (offset & (s->cluster_size - 1)) {
- fprintf(stderr, "ERROR offset=%" PRIx64 ": Cluster is not "
- "properly aligned; L2 entry corrupted.\n", offset);
- errors++;
+ case QCOW2_CLUSTER_ZERO:
+ if ((l2_entry & L2E_OFFSET_MASK) == 0) {
+ break;
+ }
+ /* fall through */
+
+ case QCOW2_CLUSTER_NORMAL:
+ {
+ uint64_t offset = l2_entry & L2E_OFFSET_MASK;
+
+ 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;
+ }
+
+ /* Mark cluster as used */
+ inc_refcounts(bs, res, refcount_table,refcount_table_size,
+ offset, s->cluster_size);
+
+ /* Correct offsets are cluster aligned */
+ if (offset & (s->cluster_size - 1)) {
+ 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();
}
}
- qemu_free(l2_table);
- return errors;
+ g_free(l2_table);
+ return 0;
fail:
- fprintf(stderr, "ERROR: I/O error in check_refcounts_l1\n");
- qemu_free(l2_table);
+ fprintf(stderr, "ERROR: I/O error in check_refcounts_l2\n");
+ g_free(l2_table);
return -EIO;
}
* error occurred.
*/
static int check_refcounts_l1(BlockDriverState *bs,
+ BdrvCheckResult *res,
uint16_t *refcount_table,
int refcount_table_size,
int64_t l1_table_offset, int l1_size,
- int check_copied)
+ int flags)
{
BDRVQcowState *s = bs->opaque;
uint64_t *l1_table, l2_offset, l1_size2;
- int i, refcount, ret;
- int errors = 0;
+ int i, ret;
l1_size2 = l1_size * sizeof(uint64_t);
/* Mark L1 table as used */
- errors += inc_refcounts(bs, refcount_table, refcount_table_size,
- l1_table_offset, l1_size2);
+ inc_refcounts(bs, res, refcount_table, refcount_table_size,
+ l1_table_offset, l1_size2);
/* Read L1 table entries from disk */
if (l1_size2 == 0) {
l1_table = NULL;
} else {
- l1_table = qemu_malloc(l1_size2);
- if (bdrv_pread(s->hd, l1_table_offset,
+ l1_table = g_malloc(l1_size2);
+ if (bdrv_pread(bs->file, l1_table_offset,
l1_table, l1_size2) != l1_size2)
goto fail;
for(i = 0;i < l1_size; i++)
for(i = 0; i < l1_size; i++) {
l2_offset = l1_table[i];
if (l2_offset) {
- /* QCOW_OFLAG_COPIED must be set iff refcount == 1 */
- if (check_copied) {
- refcount = get_refcount(bs, (l2_offset & ~QCOW_OFLAG_COPIED)
- >> s->cluster_bits);
- if ((refcount == 1) != ((l2_offset & QCOW_OFLAG_COPIED) != 0)) {
- fprintf(stderr, "ERROR OFLAG_COPIED: l2_offset=%" PRIx64
- " refcount=%d\n", l2_offset, refcount);
- errors++;
- }
- }
-
/* Mark L2 table as used */
- l2_offset &= ~QCOW_OFLAG_COPIED;
- errors += inc_refcounts(bs, refcount_table,
- refcount_table_size,
- l2_offset,
- s->cluster_size);
+ l2_offset &= L1E_OFFSET_MASK;
+ inc_refcounts(bs, res, refcount_table, refcount_table_size,
+ l2_offset, s->cluster_size);
/* L2 tables are cluster aligned */
if (l2_offset & (s->cluster_size - 1)) {
fprintf(stderr, "ERROR l2_offset=%" PRIx64 ": Table is not "
"cluster aligned; L1 entry corrupted\n", l2_offset);
- errors++;
+ res->corruptions++;
}
/* Process and check L2 entries */
- ret = check_refcounts_l2(bs, refcount_table, refcount_table_size,
- l2_offset, check_copied);
+ ret = check_refcounts_l2(bs, res, refcount_table,
+ refcount_table_size, l2_offset, flags);
if (ret < 0) {
goto fail;
}
- errors += ret;
}
}
- qemu_free(l1_table);
- return errors;
+ g_free(l1_table);
+ return 0;
fail:
fprintf(stderr, "ERROR: I/O error in check_refcounts_l1\n");
- qemu_free(l1_table);
+ res->check_errors++;
+ g_free(l1_table);
return -EIO;
}
+/*
+ * 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).
+ */
+static int check_oflag_copied(BlockDriverState *bs, BdrvCheckResult *res,
+ BdrvCheckMode fix)
+{
+ 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;
+ bool l2_dirty = false;
+
+ 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)) {
+ fprintf(stderr, "%s OFLAG_COPIED L2 cluster: l1_index=%d "
+ "l1_entry=%" PRIx64 " refcount=%d\n",
+ fix & BDRV_FIX_ERRORS ? "Repairing" :
+ "ERROR",
+ i, l1_entry, refcount);
+ 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++;
+ }
+ }
+
+ 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)) {
+ fprintf(stderr, "%s OFLAG_COPIED data cluster: "
+ "l2_entry=%" PRIx64 " refcount=%d\n",
+ fix & BDRV_FIX_ERRORS ? "Repairing" :
+ "ERROR",
+ l2_entry, refcount);
+ 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++;
+ }
+ }
+ }
+ }
+
+ if (l2_dirty) {
+ ret = qcow2_pre_write_overlap_check(bs, QCOW2_OL_ACTIVE_L2,
+ l2_offset, s->cluster_size);
+ 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;
+ }
+ }
+ }
+
+ ret = 0;
+
+fail:
+ qemu_vfree(l2_table);
+ return ret;
+}
+
+/*
+ * 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]);
+ }
+
+ ret = qcow2_pre_write_overlap_check(bs, QCOW2_OL_REFCOUNT_TABLE,
+ 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
+ * -errno); on success, 0 is returned.
+ */
+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;
+ goto fail;
+ }
+
+ /* 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));
+ goto fail;
+ }
+
+ /* new block has not yet been entered into refcount table, therefore it is
+ * no refcount block yet (regarding this check) */
+ ret = qcow2_pre_write_overlap_check(bs, 0, new_offset, s->cluster_size);
+ 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 */
+ new_offset = 0;
+ goto fail;
+ }
+
+ /* 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));
+ goto fail;
+ }
+
+ /* update refcount table */
+ assert(!(new_offset & (s->cluster_size - 1)));
+ 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));
+ goto fail;
+ }
+
+fail:
+ if (new_offset && (ret < 0)) {
+ qcow2_free_clusters(bs, new_offset, s->cluster_size,
+ QCOW2_DISCARD_ALWAYS);
+ }
+ if (refcount_block) {
+ if (ret < 0) {
+ qcow2_cache_put(bs, s->refcount_block_cache, &refcount_block);
+ } else {
+ ret = qcow2_cache_put(bs, s->refcount_block_cache, &refcount_block);
+ }
+ }
+ if (ret < 0) {
+ return ret;
+ }
+ return new_offset;
+}
+
/*
* Checks an image for refcount consistency.
*
* Returns 0 if no errors are found, the number of errors in case the image is
- * detected as corrupted, and -errno when an internal error occured.
+ * detected as corrupted, and -errno when an internal error occurred.
*/
-int qcow2_check_refcounts(BlockDriverState *bs)
+int qcow2_check_refcounts(BlockDriverState *bs, BdrvCheckResult *res,
+ BdrvCheckMode fix)
{
BDRVQcowState *s = bs->opaque;
- int64_t size;
- int nb_clusters, refcount1, refcount2, i;
+ int64_t size, i, highest_cluster;
+ int nb_clusters, refcount1, refcount2;
QCowSnapshot *sn;
uint16_t *refcount_table;
- int ret, errors = 0;
+ int ret;
- size = bdrv_getlength(s->hd);
+ size = bdrv_getlength(bs->file);
nb_clusters = size_to_clusters(s, size);
- refcount_table = qemu_mallocz(nb_clusters * sizeof(uint16_t));
+ refcount_table = g_malloc0(nb_clusters * sizeof(uint16_t));
+
+ res->bfi.total_clusters =
+ size_to_clusters(s, bs->total_sectors * BDRV_SECTOR_SIZE);
/* header */
- errors += inc_refcounts(bs, refcount_table, nb_clusters,
- 0, s->cluster_size);
+ inc_refcounts(bs, res, refcount_table, nb_clusters,
+ 0, s->cluster_size);
/* current L1 table */
- ret = check_refcounts_l1(bs, refcount_table, nb_clusters,
- s->l1_table_offset, s->l1_size, 1);
+ ret = check_refcounts_l1(bs, res, refcount_table, nb_clusters,
+ s->l1_table_offset, s->l1_size, CHECK_FRAG_INFO);
if (ret < 0) {
- return ret;
+ goto fail;
}
- errors += ret;
/* snapshots */
for(i = 0; i < s->nb_snapshots; i++) {
sn = s->snapshots + i;
- check_refcounts_l1(bs, refcount_table, nb_clusters,
- sn->l1_table_offset, sn->l1_size, 0);
+ ret = check_refcounts_l1(bs, res, refcount_table, nb_clusters,
+ sn->l1_table_offset, sn->l1_size, 0);
+ if (ret < 0) {
+ goto fail;
+ }
}
- errors += inc_refcounts(bs, refcount_table, nb_clusters,
- s->snapshots_offset, s->snapshots_size);
+ inc_refcounts(bs, res, refcount_table, nb_clusters,
+ s->snapshots_offset, s->snapshots_size);
/* refcount data */
- errors += inc_refcounts(bs, refcount_table, nb_clusters,
- s->refcount_table_offset,
- s->refcount_table_size * sizeof(uint64_t));
+ inc_refcounts(bs, res, refcount_table, nb_clusters,
+ s->refcount_table_offset,
+ s->refcount_table_size * sizeof(uint64_t));
+
for(i = 0; i < s->refcount_table_size; i++) {
- int64_t offset;
+ uint64_t offset, cluster;
offset = s->refcount_table[i];
+ cluster = offset >> s->cluster_bits;
/* Refcount blocks are cluster aligned */
if (offset & (s->cluster_size - 1)) {
- fprintf(stderr, "ERROR refcount block %d is not "
+ fprintf(stderr, "ERROR refcount block %" PRId64 " is not "
"cluster aligned; refcount table entry corrupted\n", i);
- errors++;
+ res->corruptions++;
+ continue;
+ }
+
+ if (cluster >= nb_clusters) {
+ fprintf(stderr, "ERROR refcount block %" PRId64
+ " is outside image\n", i);
+ res->corruptions++;
+ continue;
}
if (offset != 0) {
- errors += inc_refcounts(bs, refcount_table, nb_clusters,
- offset, s->cluster_size);
- if (refcount_table[offset / s->cluster_size] != 1) {
- fprintf(stderr, "ERROR refcount block %d refcount=%d\n",
- i, refcount_table[offset / s->cluster_size]);
+ inc_refcounts(bs, res, refcount_table, nb_clusters,
+ offset, s->cluster_size);
+ if (refcount_table[cluster] != 1) {
+ fprintf(stderr, "%s refcount block %" PRId64
+ " refcount=%d\n",
+ fix & BDRV_FIX_ERRORS ? "Repairing" :
+ "ERROR",
+ i, refcount_table[cluster]);
+
+ 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++;
+ }
}
}
}
/* compare ref counts */
- for(i = 0; i < nb_clusters; i++) {
+ for (i = 0, highest_cluster = 0; i < nb_clusters; i++) {
refcount1 = get_refcount(bs, i);
+ if (refcount1 < 0) {
+ fprintf(stderr, "Can't get refcount for cluster %" PRId64 ": %s\n",
+ i, strerror(-refcount1));
+ res->check_errors++;
+ continue;
+ }
+
refcount2 = refcount_table[i];
+
+ if (refcount1 > 0 || refcount2 > 0) {
+ highest_cluster = i;
+ }
+
if (refcount1 != refcount2) {
- fprintf(stderr, "ERROR cluster %d refcount=%d reference=%d\n",
+
+ /* 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",
i, refcount1, refcount2);
- errors++;
+
+ if (num_fixed) {
+ ret = update_refcount(bs, i << s->cluster_bits, 1,
+ refcount2 - refcount1,
+ QCOW2_DISCARD_ALWAYS);
+ if (ret >= 0) {
+ (*num_fixed)++;
+ continue;
+ }
+ }
+
+ /* And if we couldn't, print an error */
+ if (refcount1 < refcount2) {
+ res->corruptions++;
+ } else {
+ res->leaks++;
+ }
+ }
+ }
+
+ /* check OFLAG_COPIED */
+ ret = check_oflag_copied(bs, res, fix);
+ if (ret < 0) {
+ goto fail;
+ }
+
+ res->image_end_offset = (highest_cluster + 1) * s->cluster_size;
+ ret = 0;
+
+fail:
+ g_free(refcount_table);
+
+ return ret;
+}
+
+#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.
+ *
+ * The ign parameter specifies what checks not to perform (being a bitmask of
+ * QCow2MetadataOverlap values), i.e., what sections to ignore.
+ *
+ * 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
+ */
+int qcow2_check_metadata_overlap(BlockDriverState *bs, int ign, int64_t offset,
+ int64_t size)
+{
+ BDRVQcowState *s = bs->opaque;
+ int chk = s->overlap_check & ~ign;
+ 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;
}
}
- qemu_free(refcount_table);
+ 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;
+ }
+ }
+ }
- return errors;
+ 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;
+ uint64_t l1_sz2 = l1_sz * sizeof(uint64_t);
+ uint64_t *l1 = g_malloc(l1_sz2);
+ int ret;
+
+ ret = bdrv_pread(bs->file, l1_ofs, l1, l1_sz2);
+ if (ret < 0) {
+ g_free(l1);
+ return ret;
+ }
+
+ for (j = 0; j < l1_sz; j++) {
+ uint64_t l2_ofs = be64_to_cpu(l1[j]) & L1E_OFFSET_MASK;
+ if (l2_ofs && overlaps_with(l2_ofs, s->cluster_size)) {
+ 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.
+ */
+int qcow2_pre_write_overlap_check(BlockDriverState *bs, int ign, int64_t offset,
+ int64_t size)
+{
+ int ret = qcow2_check_metadata_overlap(bs, ign, offset, size);
+
+ 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;
+}
projects / qemu.git / blobdiff