#include <zlib.h>
#include "qemu-common.h"
-#include "block_int.h"
+#include "block/block_int.h"
#include "block/qcow2.h"
+#include "trace.h"
int qcow2_grow_l1_table(BlockDriverState *bs, int min_size, bool exact_size)
{
old_l2_offset = s->l1_table[l1_index];
+ trace_qcow2_l2_allocate(bs, l1_index);
+
/* allocate a new l2 entry */
l2_offset = qcow2_alloc_clusters(bs, s->l2_size * sizeof(uint64_t));
/* allocate a new entry in the l2 cache */
+ trace_qcow2_l2_allocate_get_empty(bs, l1_index);
ret = qcow2_cache_get_empty(bs, s->l2_table_cache, l2_offset, (void**) table);
if (ret < 0) {
return ret;
l2_table = *table;
- if (old_l2_offset == 0) {
+ if ((old_l2_offset & L1E_OFFSET_MASK) == 0) {
/* if there was no old l2 table, clear the new table */
memset(l2_table, 0, s->l2_size * sizeof(uint64_t));
} else {
/* if there was an old l2 table, read it from the disk */
BLKDBG_EVENT(bs->file, BLKDBG_L2_ALLOC_COW_READ);
- ret = qcow2_cache_get(bs, s->l2_table_cache, old_l2_offset,
+ ret = qcow2_cache_get(bs, s->l2_table_cache,
+ old_l2_offset & L1E_OFFSET_MASK,
(void**) &old_table);
if (ret < 0) {
goto fail;
/* write the l2 table to the file */
BLKDBG_EVENT(bs->file, BLKDBG_L2_ALLOC_WRITE);
+ trace_qcow2_l2_allocate_write_l2(bs, l1_index);
qcow2_cache_entry_mark_dirty(s->l2_table_cache, l2_table);
ret = qcow2_cache_flush(bs, s->l2_table_cache);
if (ret < 0) {
}
/* update the L1 entry */
+ trace_qcow2_l2_allocate_write_l1(bs, l1_index);
s->l1_table[l1_index] = l2_offset | QCOW_OFLAG_COPIED;
ret = write_l1_entry(bs, l1_index);
if (ret < 0) {
}
*table = l2_table;
+ trace_qcow2_l2_allocate_done(bs, l1_index, 0);
return 0;
fail:
+ trace_qcow2_l2_allocate_done(bs, l1_index, ret);
qcow2_cache_put(bs, s->l2_table_cache, (void**) table);
s->l1_table[l1_index] = old_l2_offset;
return ret;
}
+/*
+ * Checks how many clusters in a given L2 table are contiguous in the image
+ * file. As soon as one of the flags in the bitmask stop_flags changes compared
+ * to the first cluster, the search is stopped and the cluster is not counted
+ * as contiguous. (This allows it, for example, to stop at the first compressed
+ * cluster which may require a different handling)
+ */
static int count_contiguous_clusters(uint64_t nb_clusters, int cluster_size,
- uint64_t *l2_table, uint64_t start, uint64_t mask)
+ uint64_t *l2_table, uint64_t start, uint64_t stop_flags)
{
int i;
- uint64_t offset = be64_to_cpu(l2_table[0]) & ~mask;
+ uint64_t mask = stop_flags | L2E_OFFSET_MASK;
+ uint64_t offset = be64_to_cpu(l2_table[0]) & mask;
if (!offset)
return 0;
- for (i = start; i < start + nb_clusters; i++)
- if (offset + (uint64_t) i * cluster_size != (be64_to_cpu(l2_table[i]) & ~mask))
+ for (i = start; i < start + nb_clusters; i++) {
+ uint64_t l2_entry = be64_to_cpu(l2_table[i]) & mask;
+ if (offset + (uint64_t) i * cluster_size != l2_entry) {
break;
+ }
+ }
return (i - start);
}
static int count_contiguous_free_clusters(uint64_t nb_clusters, uint64_t *l2_table)
{
- int i = 0;
+ int i;
- while(nb_clusters-- && l2_table[i] == 0)
- i++;
+ for (i = 0; i < nb_clusters; i++) {
+ int type = qcow2_get_cluster_type(be64_to_cpu(l2_table[i]));
+
+ if (type != QCOW2_CLUSTER_UNALLOCATED) {
+ break;
+ }
+ }
return i;
}
*
* on exit, *num is the number of contiguous sectors we can read.
*
- * Return 0, if the offset is found
- * Return -errno, otherwise.
- *
+ * Returns the cluster type (QCOW2_CLUSTER_*) on success, -errno in error
+ * cases.
*/
-
int qcow2_get_cluster_offset(BlockDriverState *bs, uint64_t offset,
int *num, uint64_t *cluster_offset)
{
/* seek the the l2 offset in the l1 table */
l1_index = offset >> l1_bits;
- if (l1_index >= s->l1_size)
+ if (l1_index >= s->l1_size) {
+ ret = QCOW2_CLUSTER_UNALLOCATED;
goto out;
+ }
- l2_offset = s->l1_table[l1_index];
-
- /* seek the l2 table of the given l2 offset */
-
- if (!l2_offset)
+ l2_offset = s->l1_table[l1_index] & L1E_OFFSET_MASK;
+ if (!l2_offset) {
+ ret = QCOW2_CLUSTER_UNALLOCATED;
goto out;
+ }
/* load the l2 table in memory */
- l2_offset &= ~QCOW_OFLAG_COPIED;
ret = l2_load(bs, l2_offset, &l2_table);
if (ret < 0) {
return ret;
*cluster_offset = be64_to_cpu(l2_table[l2_index]);
nb_clusters = size_to_clusters(s, nb_needed << 9);
- if (!*cluster_offset) {
+ ret = qcow2_get_cluster_type(*cluster_offset);
+ switch (ret) {
+ case QCOW2_CLUSTER_COMPRESSED:
+ /* Compressed clusters can only be processed one by one */
+ c = 1;
+ *cluster_offset &= L2E_COMPRESSED_OFFSET_SIZE_MASK;
+ break;
+ case QCOW2_CLUSTER_ZERO:
+ if (s->qcow_version < 3) {
+ return -EIO;
+ }
+ c = count_contiguous_clusters(nb_clusters, s->cluster_size,
+ &l2_table[l2_index], 0,
+ QCOW_OFLAG_COMPRESSED | QCOW_OFLAG_ZERO);
+ *cluster_offset = 0;
+ break;
+ case QCOW2_CLUSTER_UNALLOCATED:
/* how many empty clusters ? */
c = count_contiguous_free_clusters(nb_clusters, &l2_table[l2_index]);
- } else {
+ *cluster_offset = 0;
+ break;
+ case QCOW2_CLUSTER_NORMAL:
/* how many allocated clusters ? */
c = count_contiguous_clusters(nb_clusters, s->cluster_size,
- &l2_table[l2_index], 0, QCOW_OFLAG_COPIED);
+ &l2_table[l2_index], 0,
+ QCOW_OFLAG_COMPRESSED | QCOW_OFLAG_ZERO);
+ *cluster_offset &= L2E_OFFSET_MASK;
+ break;
+ default:
+ abort();
}
qcow2_cache_put(bs, s->l2_table_cache, (void**) &l2_table);
- nb_available = (c * s->cluster_sectors);
+ nb_available = (c * s->cluster_sectors);
+
out:
if (nb_available > nb_needed)
nb_available = nb_needed;
*num = nb_available - index_in_cluster;
- *cluster_offset &=~QCOW_OFLAG_COPIED;
- return 0;
+ return ret;
}
/*
*/
static int get_cluster_table(BlockDriverState *bs, uint64_t offset,
uint64_t **new_l2_table,
- uint64_t *new_l2_offset,
int *new_l2_index)
{
BDRVQcowState *s = bs->opaque;
return ret;
}
}
- l2_offset = s->l1_table[l1_index];
+
+ l2_offset = s->l1_table[l1_index] & L1E_OFFSET_MASK;
/* seek the l2 table of the given l2 offset */
- if (l2_offset & QCOW_OFLAG_COPIED) {
+ if (s->l1_table[l1_index] & QCOW_OFLAG_COPIED) {
/* load the l2 table in memory */
- l2_offset &= ~QCOW_OFLAG_COPIED;
ret = l2_load(bs, l2_offset, &l2_table);
if (ret < 0) {
return ret;
if (l2_offset) {
qcow2_free_clusters(bs, l2_offset, s->l2_size * sizeof(uint64_t));
}
- l2_offset = s->l1_table[l1_index] & ~QCOW_OFLAG_COPIED;
}
/* find the cluster offset for the given disk offset */
l2_index = (offset >> s->cluster_bits) & (s->l2_size - 1);
*new_l2_table = l2_table;
- *new_l2_offset = l2_offset;
*new_l2_index = l2_index;
return 0;
{
BDRVQcowState *s = bs->opaque;
int l2_index, ret;
- uint64_t l2_offset, *l2_table;
+ uint64_t *l2_table;
int64_t cluster_offset;
int nb_csectors;
- ret = get_cluster_table(bs, offset, &l2_table, &l2_offset, &l2_index);
+ ret = get_cluster_table(bs, offset, &l2_table, &l2_index);
if (ret < 0) {
return 0;
}
+ /* Compression can't overwrite anything. Fail if the cluster was already
+ * allocated. */
cluster_offset = be64_to_cpu(l2_table[l2_index]);
- if (cluster_offset & QCOW_OFLAG_COPIED) {
+ if (cluster_offset & L2E_OFFSET_MASK) {
qcow2_cache_put(bs, s->l2_table_cache, (void**) &l2_table);
return 0;
}
- if (cluster_offset)
- qcow2_free_any_clusters(bs, cluster_offset, 1);
-
cluster_offset = qcow2_alloc_bytes(bs, compressed_size);
if (cluster_offset < 0) {
qcow2_cache_put(bs, s->l2_table_cache, (void**) &l2_table);
return cluster_offset;
}
-int qcow2_alloc_cluster_link_l2(BlockDriverState *bs, QCowL2Meta *m)
+static int perform_cow(BlockDriverState *bs, QCowL2Meta *m, Qcow2COWRegion *r)
{
BDRVQcowState *s = bs->opaque;
- int i, j = 0, l2_index, ret;
- uint64_t *old_cluster, start_sect, l2_offset, *l2_table;
- uint64_t cluster_offset = m->cluster_offset;
- bool cow = false;
+ int ret;
- if (m->nb_clusters == 0)
+ if (r->nb_sectors == 0) {
return 0;
+ }
- old_cluster = g_malloc(m->nb_clusters * sizeof(uint64_t));
+ qemu_co_mutex_unlock(&s->lock);
+ ret = copy_sectors(bs, m->offset / BDRV_SECTOR_SIZE, m->alloc_offset,
+ r->offset / BDRV_SECTOR_SIZE,
+ r->offset / BDRV_SECTOR_SIZE + r->nb_sectors);
+ qemu_co_mutex_lock(&s->lock);
- /* copy content of unmodified sectors */
- start_sect = (m->offset & ~(s->cluster_size - 1)) >> 9;
- if (m->n_start) {
- cow = true;
- qemu_co_mutex_unlock(&s->lock);
- ret = copy_sectors(bs, start_sect, cluster_offset, 0, m->n_start);
- qemu_co_mutex_lock(&s->lock);
- if (ret < 0)
- goto err;
- }
-
- if (m->nb_available & (s->cluster_sectors - 1)) {
- uint64_t end = m->nb_available & ~(uint64_t)(s->cluster_sectors - 1);
- cow = true;
- qemu_co_mutex_unlock(&s->lock);
- ret = copy_sectors(bs, start_sect + end, cluster_offset + (end << 9),
- m->nb_available - end, s->cluster_sectors);
- qemu_co_mutex_lock(&s->lock);
- if (ret < 0)
- goto err;
+ if (ret < 0) {
+ return ret;
}
/*
- * Update L2 table.
- *
* Before we update the L2 table to actually point to the new cluster, we
* need to be sure that the refcounts have been increased and COW was
* handled.
*/
- if (cow) {
- qcow2_cache_depends_on_flush(s->l2_table_cache);
+ qcow2_cache_depends_on_flush(s->l2_table_cache);
+
+ return 0;
+}
+
+int qcow2_alloc_cluster_link_l2(BlockDriverState *bs, QCowL2Meta *m)
+{
+ BDRVQcowState *s = bs->opaque;
+ int i, j = 0, l2_index, ret;
+ uint64_t *old_cluster, *l2_table;
+ uint64_t cluster_offset = m->alloc_offset;
+
+ trace_qcow2_cluster_link_l2(qemu_coroutine_self(), m->nb_clusters);
+ assert(m->nb_clusters > 0);
+
+ old_cluster = g_malloc(m->nb_clusters * sizeof(uint64_t));
+
+ /* copy content of unmodified sectors */
+ ret = perform_cow(bs, m, &m->cow_start);
+ if (ret < 0) {
+ goto err;
}
- qcow2_cache_set_dependency(bs, s->l2_table_cache, s->refcount_block_cache);
- ret = get_cluster_table(bs, m->offset, &l2_table, &l2_offset, &l2_index);
+ ret = perform_cow(bs, m, &m->cow_end);
+ if (ret < 0) {
+ goto err;
+ }
+
+ /* Update L2 table. */
+ if (s->use_lazy_refcounts) {
+ qcow2_mark_dirty(bs);
+ }
+ if (qcow2_need_accurate_refcounts(s)) {
+ qcow2_cache_set_dependency(bs, s->l2_table_cache,
+ s->refcount_block_cache);
+ }
+
+ ret = get_cluster_table(bs, m->offset, &l2_table, &l2_index);
if (ret < 0) {
goto err;
}
*/
if (j != 0) {
for (i = 0; i < j; i++) {
- qcow2_free_any_clusters(bs,
- be64_to_cpu(old_cluster[i]) & ~QCOW_OFLAG_COPIED, 1);
+ qcow2_free_any_clusters(bs, be64_to_cpu(old_cluster[i]), 1);
}
}
}
/*
- * alloc_cluster_offset
+ * Returns the number of contiguous clusters that can be used for an allocating
+ * write, but require COW to be performed (this includes yet unallocated space,
+ * which must copy from the backing file)
+ */
+static int count_cow_clusters(BDRVQcowState *s, int nb_clusters,
+ uint64_t *l2_table, int l2_index)
+{
+ int i;
+
+ for (i = 0; i < nb_clusters; i++) {
+ uint64_t l2_entry = be64_to_cpu(l2_table[l2_index + i]);
+ int cluster_type = qcow2_get_cluster_type(l2_entry);
+
+ switch(cluster_type) {
+ case QCOW2_CLUSTER_NORMAL:
+ if (l2_entry & QCOW_OFLAG_COPIED) {
+ goto out;
+ }
+ break;
+ case QCOW2_CLUSTER_UNALLOCATED:
+ case QCOW2_CLUSTER_COMPRESSED:
+ case QCOW2_CLUSTER_ZERO:
+ break;
+ default:
+ abort();
+ }
+ }
+
+out:
+ assert(i <= nb_clusters);
+ return i;
+}
+
+/*
+ * Check if there already is an AIO write request in flight which allocates
+ * the same cluster. In this case we need to wait until the previous
+ * request has completed and updated the L2 table accordingly.
*
- * For a given offset of the disk image, return cluster offset in qcow2 file.
- * If the offset is not found, allocate a new cluster.
+ * Returns:
+ * 0 if there was no dependency. *cur_bytes indicates the number of
+ * bytes from guest_offset that can be read before the next
+ * dependency must be processed (or the request is complete)
*
- * If the cluster was already allocated, m->nb_clusters is set to 0,
- * other fields in m are meaningless.
+ * -EAGAIN if we had to wait for another request, previously gathered
+ * information on cluster allocation may be invalid now. The caller
+ * must start over anyway, so consider *cur_bytes undefined.
+ */
+static int handle_dependencies(BlockDriverState *bs, uint64_t guest_offset,
+ uint64_t *cur_bytes, QCowL2Meta **m)
+{
+ BDRVQcowState *s = bs->opaque;
+ QCowL2Meta *old_alloc;
+ uint64_t bytes = *cur_bytes;
+
+ QLIST_FOREACH(old_alloc, &s->cluster_allocs, next_in_flight) {
+
+ uint64_t start = guest_offset;
+ uint64_t end = start + bytes;
+ uint64_t old_start = l2meta_cow_start(old_alloc);
+ uint64_t old_end = l2meta_cow_end(old_alloc);
+
+ if (end <= old_start || start >= old_end) {
+ /* No intersection */
+ } else {
+ if (start < old_start) {
+ /* Stop at the start of a running allocation */
+ bytes = old_start - start;
+ } else {
+ bytes = 0;
+ }
+
+ /* Stop if already an l2meta exists. After yielding, it wouldn't
+ * be valid any more, so we'd have to clean up the old L2Metas
+ * and deal with requests depending on them before starting to
+ * gather new ones. Not worth the trouble. */
+ if (bytes == 0 && *m) {
+ *cur_bytes = 0;
+ return 0;
+ }
+
+ if (bytes == 0) {
+ /* Wait for the dependency to complete. We need to recheck
+ * the free/allocated clusters when we continue. */
+ qemu_co_mutex_unlock(&s->lock);
+ qemu_co_queue_wait(&old_alloc->dependent_requests);
+ qemu_co_mutex_lock(&s->lock);
+ return -EAGAIN;
+ }
+ }
+ }
+
+ /* Make sure that existing clusters and new allocations are only used up to
+ * the next dependency if we shortened the request above */
+ *cur_bytes = bytes;
+
+ return 0;
+}
+
+/*
+ * Checks how many already allocated clusters that don't require a copy on
+ * write there are at the given guest_offset (up to *bytes). If
+ * *host_offset is not zero, only physically contiguous clusters beginning at
+ * this host offset are counted.
*
- * If the cluster is newly allocated, m->nb_clusters is set to the number of
- * contiguous clusters that have been allocated. In this case, the other
- * fields of m are valid and contain information about the first allocated
- * cluster.
+ * Note that guest_offset may not be cluster aligned. In this case, the
+ * returned *host_offset points to exact byte referenced by guest_offset and
+ * therefore isn't cluster aligned as well.
*
- * If the request conflicts with another write request in flight, the coroutine
- * is queued and will be reentered when the dependency has completed.
+ * Returns:
+ * 0: if no allocated clusters are available at the given offset.
+ * *bytes is normally unchanged. It is set to 0 if the cluster
+ * is allocated and doesn't need COW, but doesn't have the right
+ * physical offset.
*
- * Return 0 on success and -errno in error cases
+ * 1: if allocated clusters that don't require a COW are available at
+ * the requested offset. *bytes may have decreased and describes
+ * the length of the area that can be written to.
+ *
+ * -errno: in error cases
*/
-int qcow2_alloc_cluster_offset(BlockDriverState *bs, uint64_t offset,
- int n_start, int n_end, int *num, QCowL2Meta *m)
+static int handle_copied(BlockDriverState *bs, uint64_t guest_offset,
+ uint64_t *host_offset, uint64_t *bytes, QCowL2Meta **m)
{
BDRVQcowState *s = bs->opaque;
- int l2_index, ret;
- uint64_t l2_offset, *l2_table;
- int64_t cluster_offset;
- unsigned int nb_clusters, i = 0;
- QCowL2Meta *old_alloc;
+ int l2_index;
+ uint64_t cluster_offset;
+ uint64_t *l2_table;
+ unsigned int nb_clusters;
+ unsigned int keep_clusters;
+ int ret, pret;
- ret = get_cluster_table(bs, offset, &l2_table, &l2_offset, &l2_index);
+ trace_qcow2_handle_copied(qemu_coroutine_self(), guest_offset, *host_offset,
+ *bytes);
+
+ assert(*host_offset == 0 || offset_into_cluster(s, guest_offset)
+ == offset_into_cluster(s, *host_offset));
+
+ /*
+ * Calculate the number of clusters to look for. We stop at L2 table
+ * boundaries to keep things simple.
+ */
+ nb_clusters =
+ size_to_clusters(s, offset_into_cluster(s, guest_offset) + *bytes);
+
+ l2_index = offset_to_l2_index(s, guest_offset);
+ nb_clusters = MIN(nb_clusters, s->l2_size - l2_index);
+
+ /* Find L2 entry for the first involved cluster */
+ ret = get_cluster_table(bs, guest_offset, &l2_table, &l2_index);
if (ret < 0) {
return ret;
}
-again:
- nb_clusters = size_to_clusters(s, n_end << 9);
+ cluster_offset = be64_to_cpu(l2_table[l2_index]);
- nb_clusters = MIN(nb_clusters, s->l2_size - l2_index);
+ /* Check how many clusters are already allocated and don't need COW */
+ if (qcow2_get_cluster_type(cluster_offset) == QCOW2_CLUSTER_NORMAL
+ && (cluster_offset & QCOW_OFLAG_COPIED))
+ {
+ /* If a specific host_offset is required, check it */
+ bool offset_matches =
+ (cluster_offset & L2E_OFFSET_MASK) == *host_offset;
+
+ if (*host_offset != 0 && !offset_matches) {
+ *bytes = 0;
+ ret = 0;
+ goto out;
+ }
- cluster_offset = be64_to_cpu(l2_table[l2_index]);
+ /* We keep all QCOW_OFLAG_COPIED clusters */
+ keep_clusters =
+ count_contiguous_clusters(nb_clusters, s->cluster_size,
+ &l2_table[l2_index], 0,
+ QCOW_OFLAG_COPIED | QCOW_OFLAG_ZERO);
+ assert(keep_clusters <= nb_clusters);
- /* We keep all QCOW_OFLAG_COPIED clusters */
+ *bytes = MIN(*bytes,
+ keep_clusters * s->cluster_size
+ - offset_into_cluster(s, guest_offset));
- if (cluster_offset & QCOW_OFLAG_COPIED) {
- nb_clusters = count_contiguous_clusters(nb_clusters, s->cluster_size,
- &l2_table[l2_index], 0, 0);
+ ret = 1;
+ } else {
+ ret = 0;
+ }
- cluster_offset &= ~QCOW_OFLAG_COPIED;
- m->nb_clusters = 0;
+ /* Cleanup */
+out:
+ pret = qcow2_cache_put(bs, s->l2_table_cache, (void**) &l2_table);
+ if (pret < 0) {
+ return pret;
+ }
- goto out;
+ /* Only return a host offset if we actually made progress. Otherwise we
+ * would make requirements for handle_alloc() that it can't fulfill */
+ if (ret) {
+ *host_offset = (cluster_offset & L2E_OFFSET_MASK)
+ + offset_into_cluster(s, guest_offset);
}
- /* for the moment, multiple compressed clusters are not managed */
+ return ret;
+}
- if (cluster_offset & QCOW_OFLAG_COMPRESSED)
- nb_clusters = 1;
+/*
+ * Allocates new clusters for the given guest_offset.
+ *
+ * At most *nb_clusters are allocated, and on return *nb_clusters is updated to
+ * contain the number of clusters that have been allocated and are contiguous
+ * in the image file.
+ *
+ * If *host_offset is non-zero, it specifies the offset in the image file at
+ * which the new clusters must start. *nb_clusters can be 0 on return in this
+ * case if the cluster at host_offset is already in use. If *host_offset is
+ * zero, the clusters can be allocated anywhere in the image file.
+ *
+ * *host_offset is updated to contain the offset into the image file at which
+ * the first allocated cluster starts.
+ *
+ * Return 0 on success and -errno in error cases. -EAGAIN means that the
+ * function has been waiting for another request and the allocation must be
+ * restarted, but the whole request should not be failed.
+ */
+static int do_alloc_cluster_offset(BlockDriverState *bs, uint64_t guest_offset,
+ uint64_t *host_offset, unsigned int *nb_clusters)
+{
+ BDRVQcowState *s = bs->opaque;
- /* how many available clusters ? */
+ trace_qcow2_do_alloc_clusters_offset(qemu_coroutine_self(), guest_offset,
+ *host_offset, *nb_clusters);
- while (i < nb_clusters) {
- i += count_contiguous_clusters(nb_clusters - i, s->cluster_size,
- &l2_table[l2_index], i, 0);
- if ((i >= nb_clusters) || be64_to_cpu(l2_table[l2_index + i])) {
- break;
+ /* Allocate new clusters */
+ trace_qcow2_cluster_alloc_phys(qemu_coroutine_self());
+ if (*host_offset == 0) {
+ int64_t cluster_offset =
+ qcow2_alloc_clusters(bs, *nb_clusters * s->cluster_size);
+ if (cluster_offset < 0) {
+ return cluster_offset;
}
-
- i += count_contiguous_free_clusters(nb_clusters - i,
- &l2_table[l2_index + i]);
- if (i >= nb_clusters) {
- break;
+ *host_offset = cluster_offset;
+ return 0;
+ } else {
+ int ret = qcow2_alloc_clusters_at(bs, *host_offset, *nb_clusters);
+ if (ret < 0) {
+ return ret;
}
+ *nb_clusters = ret;
+ return 0;
+ }
+}
+
+/*
+ * Allocates new clusters for an area that either is yet unallocated or needs a
+ * copy on write. If *host_offset is non-zero, clusters are only allocated if
+ * the new allocation can match the specified host offset.
+ *
+ * Note that guest_offset may not be cluster aligned. In this case, the
+ * returned *host_offset points to exact byte referenced by guest_offset and
+ * therefore isn't cluster aligned as well.
+ *
+ * Returns:
+ * 0: if no clusters could be allocated. *bytes is set to 0,
+ * *host_offset is left unchanged.
+ *
+ * 1: if new clusters were allocated. *bytes may be decreased if the
+ * new allocation doesn't cover all of the requested area.
+ * *host_offset is updated to contain the host offset of the first
+ * newly allocated cluster.
+ *
+ * -errno: in error cases
+ */
+static int handle_alloc(BlockDriverState *bs, uint64_t guest_offset,
+ uint64_t *host_offset, uint64_t *bytes, QCowL2Meta **m)
+{
+ BDRVQcowState *s = bs->opaque;
+ int l2_index;
+ uint64_t *l2_table;
+ uint64_t entry;
+ unsigned int nb_clusters;
+ int ret;
- cluster_offset = be64_to_cpu(l2_table[l2_index + i]);
+ uint64_t alloc_cluster_offset;
- if ((cluster_offset & QCOW_OFLAG_COPIED) ||
- (cluster_offset & QCOW_OFLAG_COMPRESSED))
- break;
- }
- assert(i <= nb_clusters);
- nb_clusters = i;
+ trace_qcow2_handle_alloc(qemu_coroutine_self(), guest_offset, *host_offset,
+ *bytes);
+ assert(*bytes > 0);
/*
- * Check if there already is an AIO write request in flight which allocates
- * the same cluster. In this case we need to wait until the previous
- * request has completed and updated the L2 table accordingly.
+ * Calculate the number of clusters to look for. We stop at L2 table
+ * boundaries to keep things simple.
*/
- QLIST_FOREACH(old_alloc, &s->cluster_allocs, next_in_flight) {
-
- uint64_t start = offset >> s->cluster_bits;
- uint64_t end = start + nb_clusters;
- uint64_t old_start = old_alloc->offset >> s->cluster_bits;
- uint64_t old_end = old_start + old_alloc->nb_clusters;
+ nb_clusters =
+ size_to_clusters(s, offset_into_cluster(s, guest_offset) + *bytes);
- if (end < old_start || start > old_end) {
- /* No intersection */
- } else {
- if (start < old_start) {
- /* Stop at the start of a running allocation */
- nb_clusters = old_start - start;
- } else {
- nb_clusters = 0;
- }
-
- if (nb_clusters == 0) {
- /* Wait for the dependency to complete. We need to recheck
- * the free/allocated clusters when we continue. */
- qemu_co_mutex_unlock(&s->lock);
- qemu_co_queue_wait(&old_alloc->dependent_requests);
- qemu_co_mutex_lock(&s->lock);
- goto again;
- }
- }
- }
+ l2_index = offset_to_l2_index(s, guest_offset);
+ nb_clusters = MIN(nb_clusters, s->l2_size - l2_index);
- if (!nb_clusters) {
- abort();
+ /* Find L2 entry for the first involved cluster */
+ ret = get_cluster_table(bs, guest_offset, &l2_table, &l2_index);
+ if (ret < 0) {
+ return ret;
}
- /* save info needed for meta data update */
- m->offset = offset;
- m->n_start = n_start;
- m->nb_clusters = nb_clusters;
+ entry = be64_to_cpu(l2_table[l2_index]);
- QLIST_INSERT_HEAD(&s->cluster_allocs, m, next_in_flight);
-
- /* allocate a new cluster */
-
- cluster_offset = qcow2_alloc_clusters(bs, nb_clusters * s->cluster_size);
- if (cluster_offset < 0) {
- ret = cluster_offset;
- goto fail;
+ /* For the moment, overwrite compressed clusters one by one */
+ if (entry & QCOW_OFLAG_COMPRESSED) {
+ nb_clusters = 1;
+ } else {
+ nb_clusters = count_cow_clusters(s, nb_clusters, l2_table, l2_index);
}
-out:
+ /* This function is only called when there were no non-COW clusters, so if
+ * we can't find any unallocated or COW clusters either, something is
+ * wrong with our code. */
+ assert(nb_clusters > 0);
+
ret = qcow2_cache_put(bs, s->l2_table_cache, (void**) &l2_table);
if (ret < 0) {
- goto fail_put;
+ return ret;
}
- m->nb_available = MIN(nb_clusters << (s->cluster_bits - 9), n_end);
- m->cluster_offset = cluster_offset;
+ /* Allocate, if necessary at a given offset in the image file */
+ alloc_cluster_offset = start_of_cluster(s, *host_offset);
+ ret = do_alloc_cluster_offset(bs, guest_offset, &alloc_cluster_offset,
+ &nb_clusters);
+ if (ret < 0) {
+ goto fail;
+ }
- *num = m->nb_available - n_start;
+ /* Can't extend contiguous allocation */
+ if (nb_clusters == 0) {
+ *bytes = 0;
+ return 0;
+ }
- return 0;
+ /*
+ * Save info needed for meta data update.
+ *
+ * requested_sectors: Number of sectors from the start of the first
+ * newly allocated cluster to the end of the (possibly shortened
+ * before) write request.
+ *
+ * avail_sectors: Number of sectors from the start of the first
+ * newly allocated to the end of the last newly allocated cluster.
+ *
+ * nb_sectors: The number of sectors from the start of the first
+ * newly allocated cluster to the end of the area that the write
+ * request actually writes to (excluding COW at the end)
+ */
+ int requested_sectors =
+ (*bytes + offset_into_cluster(s, guest_offset))
+ >> BDRV_SECTOR_BITS;
+ int avail_sectors = nb_clusters
+ << (s->cluster_bits - BDRV_SECTOR_BITS);
+ int alloc_n_start = offset_into_cluster(s, guest_offset)
+ >> BDRV_SECTOR_BITS;
+ int nb_sectors = MIN(requested_sectors, avail_sectors);
+ QCowL2Meta *old_m = *m;
+
+ *m = g_malloc0(sizeof(**m));
+
+ **m = (QCowL2Meta) {
+ .next = old_m,
+
+ .alloc_offset = alloc_cluster_offset,
+ .offset = start_of_cluster(s, guest_offset),
+ .nb_clusters = nb_clusters,
+ .nb_available = nb_sectors,
+
+ .cow_start = {
+ .offset = 0,
+ .nb_sectors = alloc_n_start,
+ },
+ .cow_end = {
+ .offset = nb_sectors * BDRV_SECTOR_SIZE,
+ .nb_sectors = avail_sectors - nb_sectors,
+ },
+ };
+ qemu_co_queue_init(&(*m)->dependent_requests);
+ QLIST_INSERT_HEAD(&s->cluster_allocs, *m, next_in_flight);
+
+ *host_offset = alloc_cluster_offset + offset_into_cluster(s, guest_offset);
+ *bytes = MIN(*bytes, (nb_sectors * BDRV_SECTOR_SIZE)
+ - offset_into_cluster(s, guest_offset));
+ assert(*bytes != 0);
+
+ return 1;
fail:
- qcow2_cache_put(bs, s->l2_table_cache, (void**) &l2_table);
-fail_put:
- QLIST_REMOVE(m, next_in_flight);
+ if (*m && (*m)->nb_clusters > 0) {
+ QLIST_REMOVE(*m, next_in_flight);
+ }
return ret;
}
+/*
+ * alloc_cluster_offset
+ *
+ * For a given offset on the virtual disk, find the cluster offset in qcow2
+ * file. If the offset is not found, allocate a new cluster.
+ *
+ * If the cluster was already allocated, m->nb_clusters is set to 0 and
+ * other fields in m are meaningless.
+ *
+ * If the cluster is newly allocated, m->nb_clusters is set to the number of
+ * contiguous clusters that have been allocated. In this case, the other
+ * fields of m are valid and contain information about the first allocated
+ * cluster.
+ *
+ * If the request conflicts with another write request in flight, the coroutine
+ * is queued and will be reentered when the dependency has completed.
+ *
+ * Return 0 on success and -errno in error cases
+ */
+int qcow2_alloc_cluster_offset(BlockDriverState *bs, uint64_t offset,
+ int n_start, int n_end, int *num, uint64_t *host_offset, QCowL2Meta **m)
+{
+ BDRVQcowState *s = bs->opaque;
+ uint64_t start, remaining;
+ uint64_t cluster_offset;
+ uint64_t cur_bytes;
+ int ret;
+
+ trace_qcow2_alloc_clusters_offset(qemu_coroutine_self(), offset,
+ n_start, n_end);
+
+ assert(n_start * BDRV_SECTOR_SIZE == offset_into_cluster(s, offset));
+ offset = start_of_cluster(s, offset);
+
+again:
+ start = offset + (n_start << BDRV_SECTOR_BITS);
+ remaining = (n_end - n_start) << BDRV_SECTOR_BITS;
+ cluster_offset = 0;
+ *host_offset = 0;
+ cur_bytes = 0;
+ *m = NULL;
+
+ while (true) {
+
+ if (!*host_offset) {
+ *host_offset = start_of_cluster(s, cluster_offset);
+ }
+
+ assert(remaining >= cur_bytes);
+
+ start += cur_bytes;
+ remaining -= cur_bytes;
+ cluster_offset += cur_bytes;
+
+ if (remaining == 0) {
+ break;
+ }
+
+ cur_bytes = remaining;
+
+ /*
+ * Now start gathering as many contiguous clusters as possible:
+ *
+ * 1. Check for overlaps with in-flight allocations
+ *
+ * a) Overlap not in the first cluster -> shorten this request and
+ * let the caller handle the rest in its next loop iteration.
+ *
+ * b) Real overlaps of two requests. Yield and restart the search
+ * for contiguous clusters (the situation could have changed
+ * while we were sleeping)
+ *
+ * c) TODO: Request starts in the same cluster as the in-flight
+ * allocation ends. Shorten the COW of the in-fight allocation,
+ * set cluster_offset to write to the same cluster and set up
+ * the right synchronisation between the in-flight request and
+ * the new one.
+ */
+ ret = handle_dependencies(bs, start, &cur_bytes, m);
+ if (ret == -EAGAIN) {
+ /* Currently handle_dependencies() doesn't yield if we already had
+ * an allocation. If it did, we would have to clean up the L2Meta
+ * structs before starting over. */
+ assert(*m == NULL);
+ goto again;
+ } else if (ret < 0) {
+ return ret;
+ } else if (cur_bytes == 0) {
+ break;
+ } else {
+ /* handle_dependencies() may have decreased cur_bytes (shortened
+ * the allocations below) so that the next dependency is processed
+ * correctly during the next loop iteration. */
+ }
+
+ /*
+ * 2. Count contiguous COPIED clusters.
+ */
+ ret = handle_copied(bs, start, &cluster_offset, &cur_bytes, m);
+ if (ret < 0) {
+ return ret;
+ } else if (ret) {
+ continue;
+ } else if (cur_bytes == 0) {
+ break;
+ }
+
+ /*
+ * 3. If the request still hasn't completed, allocate new clusters,
+ * considering any cluster_offset of steps 1c or 2.
+ */
+ ret = handle_alloc(bs, start, &cluster_offset, &cur_bytes, m);
+ if (ret < 0) {
+ return ret;
+ } else if (ret) {
+ continue;
+ } else {
+ assert(cur_bytes == 0);
+ break;
+ }
+ }
+
+ *num = (n_end - n_start) - (remaining >> BDRV_SECTOR_BITS);
+ assert(*num > 0);
+ assert(*host_offset != 0);
+
+ return 0;
+}
+
static int decompress_buffer(uint8_t *out_buf, int out_buf_size,
const uint8_t *buf, int buf_size)
{
unsigned int nb_clusters)
{
BDRVQcowState *s = bs->opaque;
- uint64_t l2_offset, *l2_table;
+ uint64_t *l2_table;
int l2_index;
int ret;
int i;
- ret = get_cluster_table(bs, offset, &l2_table, &l2_offset, &l2_index);
+ ret = get_cluster_table(bs, offset, &l2_table, &l2_index);
if (ret < 0) {
return ret;
}
uint64_t old_offset;
old_offset = be64_to_cpu(l2_table[l2_index + i]);
- old_offset &= ~QCOW_OFLAG_COPIED;
-
- if (old_offset == 0) {
+ if ((old_offset & L2E_OFFSET_MASK) == 0) {
continue;
}
return 0;
}
+
+/*
+ * This zeroes as many clusters of nb_clusters as possible at once (i.e.
+ * all clusters in the same L2 table) and returns the number of zeroed
+ * clusters.
+ */
+static int zero_single_l2(BlockDriverState *bs, uint64_t offset,
+ unsigned int nb_clusters)
+{
+ BDRVQcowState *s = bs->opaque;
+ uint64_t *l2_table;
+ int l2_index;
+ int ret;
+ int i;
+
+ ret = get_cluster_table(bs, offset, &l2_table, &l2_index);
+ if (ret < 0) {
+ return ret;
+ }
+
+ /* Limit nb_clusters to one L2 table */
+ nb_clusters = MIN(nb_clusters, s->l2_size - l2_index);
+
+ for (i = 0; i < nb_clusters; i++) {
+ uint64_t old_offset;
+
+ old_offset = be64_to_cpu(l2_table[l2_index + i]);
+
+ /* Update L2 entries */
+ qcow2_cache_entry_mark_dirty(s->l2_table_cache, l2_table);
+ if (old_offset & QCOW_OFLAG_COMPRESSED) {
+ l2_table[l2_index + i] = cpu_to_be64(QCOW_OFLAG_ZERO);
+ qcow2_free_any_clusters(bs, old_offset, 1);
+ } else {
+ l2_table[l2_index + i] |= cpu_to_be64(QCOW_OFLAG_ZERO);
+ }
+ }
+
+ ret = qcow2_cache_put(bs, s->l2_table_cache, (void**) &l2_table);
+ if (ret < 0) {
+ return ret;
+ }
+
+ return nb_clusters;
+}
+
+int qcow2_zero_clusters(BlockDriverState *bs, uint64_t offset, int nb_sectors)
+{
+ BDRVQcowState *s = bs->opaque;
+ unsigned int nb_clusters;
+ int ret;
+
+ /* The zero flag is only supported by version 3 and newer */
+ if (s->qcow_version < 3) {
+ return -ENOTSUP;
+ }
+
+ /* Each L2 table is handled by its own loop iteration */
+ nb_clusters = size_to_clusters(s, nb_sectors << BDRV_SECTOR_BITS);
+
+ while (nb_clusters > 0) {
+ ret = zero_single_l2(bs, offset, nb_clusters);
+ if (ret < 0) {
+ return ret;
+ }
+
+ nb_clusters -= ret;
+ offset += (ret * s->cluster_size);
+ }
+
+ return 0;
+}