[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
[Qemu-devel] [RFC V6 19/33] block: Add qcow2_dedup format and image crea
From: |
Benoît Canet |
Subject: |
[Qemu-devel] [RFC V6 19/33] block: Add qcow2_dedup format and image creation code. |
Date: |
Wed, 6 Feb 2013 13:31:52 +0100 |
Signed-off-by: Benoit Canet <address@hidden>
---
block/qcow2.c | 188 +++++++++++++++++++++++++++++++++++++++++----
block/qcow2.h | 2 +
include/block/block_int.h | 1 +
3 files changed, 178 insertions(+), 13 deletions(-)
diff --git a/block/qcow2.c b/block/qcow2.c
index ad202fa..9cbb2f0 100644
--- a/block/qcow2.c
+++ b/block/qcow2.c
@@ -277,6 +277,11 @@ int qcow2_mark_dirty(BlockDriverState *bs)
return qcow2_add_feature(bs, QCOW2_INCOMPAT_DIRTY);
}
+static int qcow2_activate_dedup(BlockDriverState *bs)
+{
+ return qcow2_add_feature(bs, QCOW2_INCOMPAT_DEDUP);
+}
+
/*
* Clears an incompatible feature bit and flushes before if necessary.
* Only call this function when there are no pending requests, it does not
@@ -1267,7 +1272,8 @@ static int preallocate(BlockDriverState *bs)
static int qcow2_create2(const char *filename, int64_t total_size,
const char *backing_file, const char *backing_format,
int flags, size_t cluster_size, int prealloc,
- QEMUOptionParameter *options, int version)
+ QEMUOptionParameter *options, int version,
+ bool dedup, uint8_t hash_algo)
{
/* Calculate cluster_bits */
int cluster_bits;
@@ -1294,8 +1300,10 @@ static int qcow2_create2(const char *filename, int64_t
total_size,
* size for any qcow2 image.
*/
BlockDriverState* bs;
+ BDRVQcowState *s;
QCowHeader header;
- uint8_t* refcount_table;
+ uint8_t *tables;
+ int size;
int ret;
ret = bdrv_create_file(filename, options);
@@ -1337,10 +1345,11 @@ static int qcow2_create2(const char *filename, int64_t
total_size,
goto out;
}
- /* Write an empty refcount table */
- refcount_table = g_malloc0(cluster_size);
- ret = bdrv_pwrite(bs, cluster_size, refcount_table, cluster_size);
- g_free(refcount_table);
+ /* Write an empty refcount table + extra space for dedup table if needed */
+ size = dedup ? 2 : 1;
+ tables = g_malloc0(size * cluster_size);
+ ret = bdrv_pwrite(bs, cluster_size, tables, size * cluster_size);
+ g_free(tables);
if (ret < 0) {
goto out;
@@ -1351,7 +1360,7 @@ static int qcow2_create2(const char *filename, int64_t
total_size,
/*
* And now open the image and make it consistent first (i.e. increase the
* refcount of the cluster that is occupied by the header and the refcount
- * table)
+ * table and the eventual dedup table)
*/
BlockDriver* drv = bdrv_find_format("qcow2");
assert(drv != NULL);
@@ -1361,7 +1370,8 @@ static int qcow2_create2(const char *filename, int64_t
total_size,
goto out;
}
- ret = qcow2_alloc_clusters(bs, 2 * cluster_size);
+ size++; /* Add a cluster for the header */
+ ret = qcow2_alloc_clusters(bs, size * cluster_size);
if (ret < 0) {
goto out;
@@ -1371,11 +1381,29 @@ static int qcow2_create2(const char *filename, int64_t
total_size,
}
/* Okay, now that we have a valid image, let's give it the right size */
+ s = bs->opaque;
ret = bdrv_truncate(bs, total_size * BDRV_SECTOR_SIZE);
if (ret < 0) {
goto out;
}
+ if (dedup) {
+ s->has_dedup = true;
+ s->dedup_table_offset = cluster_size * 2;
+ s->dedup_table_size = cluster_size / sizeof(uint64_t);
+ s->dedup_hash_algo = hash_algo;
+
+ ret = qcow2_activate_dedup(bs);
+ if (ret < 0) {
+ goto out;
+ }
+
+ ret = qcow2_update_header(bs);
+ if (ret < 0) {
+ goto out;
+ }
+ }
+
/* Want a backing file? There you go.*/
if (backing_file) {
ret = bdrv_change_backing_file(bs, backing_file, backing_format);
@@ -1401,15 +1429,41 @@ out:
return ret;
}
+static int qcow2_warn_if_version_3_is_needed(int version,
+ bool has_feature,
+ const char *feature)
+{
+ if (version < 3 && has_feature) {
+ fprintf(stderr, "%s only supported with compatibility "
+ "level 1.1 and above (use compat=1.1 or greater)\n",
+ feature);
+ return -EINVAL;
+ }
+ return 0;
+}
+
+static int8_t qcow2_get_dedup_hash_algo(char *value)
+{
+ if (!value || !strcmp(value, "sha256")) {
+ return QCOW_HASH_SHA256;
+ }
+
+ error_printf("Unsupported deduplication hash algorithm.\n");
+ return -EINVAL;
+}
+
static int qcow2_create(const char *filename, QEMUOptionParameter *options)
{
const char *backing_file = NULL;
const char *backing_fmt = NULL;
uint64_t sectors = 0;
int flags = 0;
+ int ret;
size_t cluster_size = DEFAULT_CLUSTER_SIZE;
int prealloc = 0;
int version = 2;
+ bool dedup = false;
+ int8_t hash_algo = 0;
/* Read out options */
while (options && options->name) {
@@ -1447,24 +1501,42 @@ static int qcow2_create(const char *filename,
QEMUOptionParameter *options)
}
} else if (!strcmp(options->name, BLOCK_OPT_LAZY_REFCOUNTS)) {
flags |= options->value.n ? BLOCK_FLAG_LAZY_REFCOUNTS : 0;
+ } else if (!strcmp(options->name, BLOCK_OPT_DEDUP)) {
+ hash_algo = qcow2_get_dedup_hash_algo(options->value.s);
+ if (hash_algo < 0) {
+ return hash_algo;
+ }
+ dedup = true;
}
options++;
}
+ if (dedup) {
+ version = 3;
+ }
+
if (backing_file && prealloc) {
fprintf(stderr, "Backing file and preallocation cannot be used at "
"the same time\n");
return -EINVAL;
}
- if (version < 3 && (flags & BLOCK_FLAG_LAZY_REFCOUNTS)) {
- fprintf(stderr, "Lazy refcounts only supported with compatibility "
- "level 1.1 and above (use compat=1.1 or greater)\n");
- return -EINVAL;
+ ret = qcow2_warn_if_version_3_is_needed(version,
+ flags & BLOCK_FLAG_LAZY_REFCOUNTS,
+ "Lazy refcounts");
+ if (ret < 0) {
+ return ret;
+ }
+ ret = qcow2_warn_if_version_3_is_needed(version,
+ dedup,
+ "Deduplication");
+ if (ret < 0) {
+ return ret;
}
return qcow2_create2(filename, sectors, backing_file, backing_fmt, flags,
- cluster_size, prealloc, options, version);
+ cluster_size, prealloc, options, version,
+ dedup, hash_algo);
}
static int qcow2_make_empty(BlockDriverState *bs)
@@ -1770,6 +1842,56 @@ static QEMUOptionParameter qcow2_create_options[] = {
{ NULL }
};
+static QEMUOptionParameter qcow2_dedup_create_options[] = {
+ {
+ .name = BLOCK_OPT_SIZE,
+ .type = OPT_SIZE,
+ .help = "Virtual disk size"
+ },
+ {
+ .name = BLOCK_OPT_COMPAT_LEVEL,
+ .type = OPT_STRING,
+ .help = "Compatibility level (0.10 or 1.1)",
+ },
+ {
+ .name = BLOCK_OPT_BACKING_FILE,
+ .type = OPT_STRING,
+ .help = "File name of a base image"
+ },
+ {
+ .name = BLOCK_OPT_BACKING_FMT,
+ .type = OPT_STRING,
+ .help = "Image format of the base image"
+ },
+ {
+ .name = BLOCK_OPT_ENCRYPT,
+ .type = OPT_FLAG,
+ .help = "Encrypt the image"
+ },
+ {
+ .name = BLOCK_OPT_CLUSTER_SIZE,
+ .type = OPT_SIZE,
+ .help = "qcow2 cluster size",
+ .value = { .n = DEFAULT_DEDUP_CLUSTER_SIZE },
+ },
+ {
+ .name = BLOCK_OPT_PREALLOC,
+ .type = OPT_STRING,
+ .help = "Preallocation mode (allowed values: off, metadata)"
+ },
+ {
+ .name = BLOCK_OPT_LAZY_REFCOUNTS,
+ .type = OPT_FLAG,
+ .help = "Postpone refcount updates",
+ },
+ {
+ .name = BLOCK_OPT_DEDUP,
+ .type = OPT_STRING,
+ .help = "Deduplication",
+ },
+ { NULL }
+};
+
static BlockDriver bdrv_qcow2 = {
.format_name = "qcow2",
.instance_size = sizeof(BDRVQcowState),
@@ -1809,9 +1931,49 @@ static BlockDriver bdrv_qcow2 = {
.bdrv_check = qcow2_check,
};
+static BlockDriver bdrv_qcow2_dedup = {
+ .format_name = "qcow2_dedup",
+ .instance_size = sizeof(BDRVQcowState),
+ .bdrv_probe = qcow2_probe,
+ .bdrv_open = qcow2_open,
+ .bdrv_close = qcow2_close,
+ .bdrv_reopen_prepare = qcow2_reopen_prepare,
+ .bdrv_create = qcow2_create,
+ .bdrv_co_is_allocated = qcow2_co_is_allocated,
+ .bdrv_set_key = qcow2_set_key,
+ .bdrv_make_empty = qcow2_make_empty,
+
+ .bdrv_co_readv = qcow2_co_readv,
+ .bdrv_co_writev = qcow2_co_writev,
+ .bdrv_co_flush_to_os = qcow2_co_flush_to_os,
+
+ .bdrv_co_write_zeroes = qcow2_co_write_zeroes,
+ .bdrv_co_discard = qcow2_co_discard,
+ .bdrv_truncate = qcow2_truncate,
+ .bdrv_write_compressed = qcow2_write_compressed,
+
+ .bdrv_snapshot_create = qcow2_snapshot_create,
+ .bdrv_snapshot_goto = qcow2_snapshot_goto,
+ .bdrv_snapshot_delete = qcow2_snapshot_delete,
+ .bdrv_snapshot_list = qcow2_snapshot_list,
+ .bdrv_snapshot_load_tmp = qcow2_snapshot_load_tmp,
+ .bdrv_get_info = qcow2_get_info,
+
+ .bdrv_save_vmstate = qcow2_save_vmstate,
+ .bdrv_load_vmstate = qcow2_load_vmstate,
+
+ .bdrv_change_backing_file = qcow2_change_backing_file,
+
+ .bdrv_invalidate_cache = qcow2_invalidate_cache,
+
+ .create_options = qcow2_dedup_create_options,
+ .bdrv_check = qcow2_check,
+};
+
static void bdrv_qcow2_init(void)
{
bdrv_register(&bdrv_qcow2);
+ bdrv_register(&bdrv_qcow2_dedup);
}
block_init(bdrv_qcow2_init);
diff --git a/block/qcow2.h b/block/qcow2.h
index 77ffa0b..5b51005 100644
--- a/block/qcow2.h
+++ b/block/qcow2.h
@@ -60,7 +60,9 @@
/* Must be at least 4 to cover all cases of refcount table growth */
#define REFCOUNT_CACHE_SIZE 4
+#define DEDUP_CACHE_SIZE 4
#define DEFAULT_CLUSTER_SIZE 65536
+#define DEFAULT_DEDUP_CLUSTER_SIZE 4096
#define HASH_LENGTH 32
/* indicate that the hash structure is empty and miss offset */
diff --git a/include/block/block_int.h b/include/block/block_int.h
index eaad53e..62c72fc 100644
--- a/include/block/block_int.h
+++ b/include/block/block_int.h
@@ -57,6 +57,7 @@
#define BLOCK_OPT_COMPAT_LEVEL "compat"
#define BLOCK_OPT_LAZY_REFCOUNTS "lazy_refcounts"
#define BLOCK_OPT_ADAPTER_TYPE "adapter_type"
+#define BLOCK_OPT_DEDUP "dedup"
typedef struct BdrvTrackedRequest BdrvTrackedRequest;
--
1.7.10.4
- [Qemu-devel] [RFC V6 18/33] qcow2: Extract qcow2_add_feature and qcow2_remove_feature., (continued)
- [Qemu-devel] [RFC V6 18/33] qcow2: Extract qcow2_add_feature and qcow2_remove_feature., Benoît Canet, 2013/02/06
- [Qemu-devel] [RFC V6 16/33] qcow2: Extract qcow2_do_table_init., Benoît Canet, 2013/02/06
- [Qemu-devel] [RFC V6 20/33] qcow2: Add a deduplication boolean to update_refcount., Benoît Canet, 2013/02/06
- [Qemu-devel] [RFC V6 21/33] qcow2: Drop hash for a given cluster when dedup makes refcount > 2^16/2., Benoît Canet, 2013/02/06
- [Qemu-devel] [RFC V6 22/33] qcow2: Remove hash when cluster is deleted., Benoît Canet, 2013/02/06
- [Qemu-devel] [RFC V6 19/33] block: Add qcow2_dedup format and image creation code.,
Benoît Canet <=
- [Qemu-devel] [RFC V6 25/33] qcow2: Serialize write requests when deduplication is activated., Benoît Canet, 2013/02/06
- [Qemu-devel] [RFC V6 30/33] qcow2: Integrate SKEIN hash algorithm in deduplication., Benoît Canet, 2013/02/06
- [Qemu-devel] [RFC V6 32/33] qemu-iotests: Filter dedup=on/off so existing tests don't break., Benoît Canet, 2013/02/06
- [Qemu-devel] [RFC V6 31/33] qcow: Set large dedup hash block size., Benoît Canet, 2013/02/06