From: Max Reitz <mreitz@redhat.com>
To: qemu-devel@nongnu.org
Cc: Kevin Wolf <kwolf@redhat.com>, Peter Lieven <pl@kamp.de>,
Stefan Hajnoczi <stefanha@redhat.com>,
Max Reitz <mreitz@redhat.com>
Subject: [Qemu-devel] [PATCH 18/21] qcow2: Add function for refcount order amendment
Date: Mon, 10 Nov 2014 14:45:56 +0100 [thread overview]
Message-ID: <1415627159-15941-19-git-send-email-mreitz@redhat.com> (raw)
In-Reply-To: <1415627159-15941-1-git-send-email-mreitz@redhat.com>
Add a function qcow2_change_refcount_order() which allows changing the
refcount order of a qcow2 image.
Signed-off-by: Max Reitz <mreitz@redhat.com>
---
block/qcow2-refcount.c | 424 +++++++++++++++++++++++++++++++++++++++++++++++++
block/qcow2.h | 4 +
2 files changed, 428 insertions(+)
diff --git a/block/qcow2-refcount.c b/block/qcow2-refcount.c
index 08b2ddb..59b6437 100644
--- a/block/qcow2-refcount.c
+++ b/block/qcow2-refcount.c
@@ -2473,3 +2473,427 @@ int qcow2_pre_write_overlap_check(BlockDriverState *bs, int ign, int64_t offset,
return 0;
}
+
+/**
+ * This "operation" for walk_over_reftable() allocates the refblock on disk (if
+ * it is not empty) and inserts its offset into the new reftable. The size of
+ * this new reftable is increased as required.
+ */
+static int alloc_refblock(BlockDriverState *bs, uint64_t **reftable,
+ uint64_t reftable_index, uint64_t *reftable_size,
+ void *refblock, bool refblock_empty, Error **errp)
+{
+ BDRVQcowState *s = bs->opaque;
+ int64_t offset;
+
+ if (!refblock_empty && reftable_index >= *reftable_size) {
+ uint64_t *new_reftable;
+ uint64_t new_reftable_size;
+
+ new_reftable_size = ROUND_UP(reftable_index + 1,
+ s->cluster_size / sizeof(uint64_t));
+ if (new_reftable_size > QCOW_MAX_REFTABLE_SIZE / sizeof(uint64_t)) {
+ error_setg(errp,
+ "This operation would make the refcount table grow "
+ "beyond the maximum size supported by QEMU, aborting");
+ return -ENOTSUP;
+ }
+
+ new_reftable = g_try_realloc(*reftable, new_reftable_size *
+ sizeof(uint64_t));
+ if (!new_reftable) {
+ error_setg(errp, "Failed to increase reftable buffer size");
+ return -ENOMEM;
+ }
+
+ memset(new_reftable + *reftable_size, 0,
+ (new_reftable_size - *reftable_size) * sizeof(uint64_t));
+
+ *reftable = new_reftable;
+ *reftable_size = new_reftable_size;
+ }
+
+ if (refblock_empty) {
+ if (reftable_index < *reftable_size) {
+ (*reftable)[reftable_index] = 0;
+ }
+ } else {
+ offset = qcow2_alloc_clusters(bs, s->cluster_size);
+ if (offset < 0) {
+ error_setg_errno(errp, -offset, "Failed to allocate refblock");
+ return offset;
+ }
+ (*reftable)[reftable_index++] = offset;
+ }
+
+ return 0;
+}
+
+/**
+ * This "operation" for walk_over_reftable() writes the refblock to disk at the
+ * offset specified by the new reftable's entry. It does not modify the new
+ * reftable or change any refcounts.
+ */
+static int flush_refblock(BlockDriverState *bs, uint64_t **reftable,
+ uint64_t reftable_index, uint64_t *reftable_size,
+ void *refblock, bool refblock_empty, Error **errp)
+{
+ BDRVQcowState *s = bs->opaque;
+ int64_t offset;
+ int ret;
+
+ if (refblock_empty) {
+ if (reftable_index < *reftable_size) {
+ assert((*reftable)[reftable_index] == 0);
+ }
+ } else {
+ /* The first pass with alloc_refblock() made the reftable large enough
+ */
+ assert(reftable_index < *reftable_size);
+ offset = (*reftable)[reftable_index];
+ assert(offset != 0);
+
+ ret = qcow2_pre_write_overlap_check(bs, 0, offset, s->cluster_size);
+ if (ret < 0) {
+ error_setg_errno(errp, -ret, "Overlap check failed");
+ return ret;
+ }
+
+ ret = bdrv_pwrite(bs->file, offset, refblock, s->cluster_size);
+ if (ret < 0) {
+ error_setg_errno(errp, -ret, "Failed to write refblock");
+ return ret;
+ }
+ }
+
+ return 0;
+}
+
+/**
+ * This function walks over the existing reftable and every referenced refblock;
+ * if @new_set_refcount is non-NULL, it is called for every refcount entry to
+ * create an equal new entry in the passed @new_refblock. Once that
+ * @new_refblock is completely filled, @operation will be called.
+ *
+ * @operation is expected to combine the @new_refblock and its entry in the new
+ * reftable (which is described by the parameters starting with "reftable").
+ * @refblock_empty is set if all entries in the refblock are zero.
+ *
+ * @status_cb and @cb_opaque are used for the amend operation's status callback.
+ * @index is the index of the walk_over_reftable() calls and @total is the total
+ * number of walk_over_reftable() calls per amend operation. Both are used for
+ * calculating the parameters for the status callback.
+ */
+static int walk_over_reftable(BlockDriverState *bs, uint64_t **new_reftable,
+ uint64_t *new_reftable_index,
+ uint64_t *new_reftable_size,
+ void *new_refblock, int new_refblock_size,
+ int new_refcount_bits,
+ int (*operation)(BlockDriverState *bs,
+ uint64_t **reftable,
+ uint64_t reftable_index,
+ uint64_t *reftable_size,
+ void *refblock,
+ bool refblock_empty,
+ Error **errp),
+ Qcow2SetRefcountFunc *new_set_refcount,
+ BlockDriverAmendStatusCB *status_cb,
+ void *cb_opaque, int index, int total,
+ Error **errp)
+{
+ BDRVQcowState *s = bs->opaque;
+ uint64_t reftable_index;
+ bool new_refblock_empty = true;
+ int refblock_index;
+ int new_refblock_index = 0;
+ int ret;
+
+ for (reftable_index = 0; reftable_index < s->refcount_table_size;
+ reftable_index++)
+ {
+ uint64_t refblock_offset = s->refcount_table[reftable_index]
+ & REFT_OFFSET_MASK;
+
+ status_cb(bs, (uint64_t)index * s->refcount_table_size + reftable_index,
+ (uint64_t)total * s->refcount_table_size, cb_opaque);
+
+ if (refblock_offset) {
+ void *refblock;
+
+ if (offset_into_cluster(s, refblock_offset)) {
+ qcow2_signal_corruption(bs, true, -1, -1, "Refblock offset %#"
+ PRIx64 " unaligned (reftable index: %#"
+ PRIx64 ")", refblock_offset,
+ reftable_index);
+ error_setg(errp,
+ "Image is corrupt (unaligned refblock offset)");
+ return -EIO;
+ }
+
+ ret = qcow2_cache_get(bs, s->refcount_block_cache, refblock_offset,
+ &refblock);
+ if (ret < 0) {
+ error_setg_errno(errp, -ret, "Failed to retrieve refblock");
+ return ret;
+ }
+
+ for (refblock_index = 0; refblock_index < s->refcount_block_size;
+ refblock_index++)
+ {
+ uint64_t refcount;
+
+ if (new_refblock_index >= new_refblock_size) {
+ /* new_refblock is now complete */
+ ret = operation(bs, new_reftable, *new_reftable_index,
+ new_reftable_size, new_refblock,
+ new_refblock_empty, errp);
+ if (ret < 0) {
+ qcow2_cache_put(bs, s->refcount_block_cache, &refblock);
+ return ret;
+ }
+
+ (*new_reftable_index)++;
+ new_refblock_index = 0;
+ new_refblock_empty = true;
+ }
+
+ refcount = s->get_refcount(refblock, refblock_index);
+ if (new_refcount_bits < 64 && refcount >> new_refcount_bits) {
+ uint64_t offset;
+
+ qcow2_cache_put(bs, s->refcount_block_cache, &refblock);
+
+ offset = ((reftable_index << s->refcount_block_bits)
+ + refblock_index) << s->cluster_bits;
+
+ error_setg(errp, "Cannot decrease refcount entry width to "
+ "%i bits: Cluster at offset %#" PRIx64 " has a "
+ "refcount of %" PRIu64, new_refcount_bits,
+ offset, refcount);
+ return -EINVAL;
+ }
+
+ if (new_set_refcount) {
+ new_set_refcount(new_refblock, new_refblock_index++, refcount);
+ } else {
+ new_refblock_index++;
+ }
+ new_refblock_empty = new_refblock_empty && refcount == 0;
+ }
+
+ ret = qcow2_cache_put(bs, s->refcount_block_cache, &refblock);
+ if (ret < 0) {
+ error_setg_errno(errp, -ret, "Failed to put refblock back into "
+ "the cache");
+ return ret;
+ }
+ } else {
+ /* No refblock means every refcount is 0 */
+ for (refblock_index = 0; refblock_index < s->refcount_block_size;
+ refblock_index++)
+ {
+ if (new_refblock_index >= new_refblock_size) {
+ /* new_refblock is now complete */
+ ret = operation(bs, new_reftable, *new_reftable_index,
+ new_reftable_size, new_refblock,
+ new_refblock_empty, errp);
+ if (ret < 0) {
+ return ret;
+ }
+
+ (*new_reftable_index)++;
+ new_refblock_index = 0;
+ new_refblock_empty = true;
+ }
+
+ if (new_set_refcount) {
+ new_set_refcount(new_refblock, new_refblock_index++, 0);
+ } else {
+ new_refblock_index++;
+ }
+ }
+ }
+ }
+
+ if (new_refblock_index > 0) {
+ /* Complete the potentially existing partially filled final refblock */
+ if (new_set_refcount) {
+ for (; new_refblock_index < new_refblock_size;
+ new_refblock_index++)
+ {
+ new_set_refcount(new_refblock, new_refblock_index, 0);
+ }
+ }
+
+ ret = operation(bs, new_reftable, *new_reftable_index,
+ new_reftable_size, new_refblock, new_refblock_empty,
+ errp);
+ if (ret < 0) {
+ return ret;
+ }
+
+ (*new_reftable_index)++;
+ }
+
+ return 0;
+}
+
+int qcow2_change_refcount_order(BlockDriverState *bs, int refcount_order,
+ BlockDriverAmendStatusCB *status_cb,
+ void *cb_opaque, Error **errp)
+{
+ BDRVQcowState *s = bs->opaque;
+ Qcow2GetRefcountFunc *new_get_refcount;
+ Qcow2SetRefcountFunc *new_set_refcount;
+ void *new_refblock = qemu_blockalign(bs->file, s->cluster_size);
+ uint64_t *new_reftable = NULL, new_reftable_size = 0;
+ uint64_t *old_reftable, old_reftable_size, old_reftable_offset;
+ uint64_t new_reftable_index = 0;
+ uint64_t i;
+ int64_t new_reftable_offset;
+ int new_refblock_size, new_refcount_bits = 1 << refcount_order;
+ int old_refcount_order;
+ int ret;
+
+ assert(s->qcow_version >= 3);
+ assert(refcount_order >= 0 && refcount_order <= 6);
+
+ /* see qcow2_open() */
+ new_refblock_size = 1 << (s->cluster_bits - (refcount_order - 3));
+
+ get_refcount_functions(refcount_order,
+ &new_get_refcount, &new_set_refcount);
+
+
+ /* First, allocate the structures so they are present in the refcount
+ * structures */
+ ret = walk_over_reftable(bs, &new_reftable, &new_reftable_index,
+ &new_reftable_size, NULL, new_refblock_size,
+ new_refcount_bits, &alloc_refblock, NULL,
+ status_cb, cb_opaque, 0, 2, errp);
+ if (ret < 0) {
+ goto done;
+ }
+
+ /* The new_reftable_size is now valid and will not be changed anymore,
+ * so we can now allocate the reftable */
+ new_reftable_offset = qcow2_alloc_clusters(bs, new_reftable_size *
+ sizeof(uint64_t));
+ if (new_reftable_offset < 0) {
+ error_setg_errno(errp, -new_reftable_offset,
+ "Failed to allocate the new reftable");
+ ret = new_reftable_offset;
+ goto done;
+ }
+
+ new_reftable_index = 0;
+
+ /* Second, write the new refblocks */
+ ret = walk_over_reftable(bs, &new_reftable, &new_reftable_index,
+ &new_reftable_size, new_refblock,
+ new_refblock_size, new_refcount_bits,
+ &flush_refblock, new_set_refcount,
+ status_cb, cb_opaque, 1, 2, errp);
+ if (ret < 0) {
+ goto done;
+ }
+
+
+ /* Write the new reftable */
+ ret = qcow2_pre_write_overlap_check(bs, 0, new_reftable_offset,
+ new_reftable_size * sizeof(uint64_t));
+ if (ret < 0) {
+ error_setg_errno(errp, -ret, "Overlap check failed");
+ goto done;
+ }
+
+ for (i = 0; i < new_reftable_size; i++) {
+ cpu_to_be64s(&new_reftable[i]);
+ }
+
+ ret = bdrv_pwrite(bs->file, new_reftable_offset, new_reftable,
+ new_reftable_size * sizeof(uint64_t));
+
+ for (i = 0; i < new_reftable_size; i++) {
+ be64_to_cpus(&new_reftable[i]);
+ }
+
+ if (ret < 0) {
+ error_setg_errno(errp, -ret, "Failed to write the new reftable");
+ goto done;
+ }
+
+
+ /* Empty the refcount cache */
+ ret = qcow2_cache_flush(bs, s->refcount_block_cache);
+ if (ret < 0) {
+ error_setg_errno(errp, -ret, "Failed to flush the refblock cache");
+ goto done;
+ }
+
+ /* Update the image header to point to the new reftable; this only updates
+ * the fields which are relevant to qcow2_update_header(); other fields
+ * such as s->refcount_table or s->refcount_bits stay stale for now
+ * (because we have to restore everything if qcow2_update_header() fails) */
+ old_refcount_order = s->refcount_order;
+ old_reftable_size = s->refcount_table_size;
+ old_reftable_offset = s->refcount_table_offset;
+
+ s->refcount_order = refcount_order;
+ s->refcount_table_size = new_reftable_size;
+ s->refcount_table_offset = new_reftable_offset;
+
+ ret = qcow2_update_header(bs);
+ if (ret < 0) {
+ s->refcount_order = old_refcount_order;
+ s->refcount_table_size = old_reftable_size;
+ s->refcount_table_offset = old_reftable_offset;
+ error_setg_errno(errp, -ret, "Failed to update the qcow2 header");
+ goto done;
+ }
+
+ /* Now update the rest of the in-memory information */
+ old_reftable = s->refcount_table;
+ s->refcount_table = new_reftable;
+
+ /* For cleaning up all old refblocks */
+ new_reftable = old_reftable;
+ new_reftable_size = old_reftable_size;
+
+ s->refcount_bits = 1 << refcount_order;
+ if (refcount_order < 6) {
+ s->refcount_max = (UINT64_C(1) << s->refcount_bits) - 1;
+ } else {
+ s->refcount_max = INT64_MAX;
+ }
+
+ s->refcount_block_bits = s->cluster_bits - (refcount_order - 3);
+ s->refcount_block_size = 1 << s->refcount_block_bits;
+
+ s->get_refcount = new_get_refcount;
+ s->set_refcount = new_set_refcount;
+
+ /* And free the old reftable (the old refblocks are freed below the "done"
+ * label) */
+ qcow2_free_clusters(bs, old_reftable_offset,
+ old_reftable_size * sizeof(uint64_t),
+ QCOW2_DISCARD_NEVER);
+
+done:
+ if (new_reftable) {
+ /* On success, new_reftable actually points to the old reftable (and
+ * new_reftable_size is the old reftable's size); but that is just
+ * fine */
+ for (i = 0; i < new_reftable_size; i++) {
+ uint64_t offset = new_reftable[i] & REFT_OFFSET_MASK;
+ if (offset) {
+ qcow2_free_clusters(bs, offset, s->cluster_size,
+ QCOW2_DISCARD_NEVER);
+ }
+ }
+ g_free(new_reftable);
+ }
+
+ qemu_vfree(new_refblock);
+ return ret;
+}
diff --git a/block/qcow2.h b/block/qcow2.h
index fe12c54..5b96519 100644
--- a/block/qcow2.h
+++ b/block/qcow2.h
@@ -526,6 +526,10 @@ int qcow2_check_metadata_overlap(BlockDriverState *bs, int ign, int64_t offset,
int qcow2_pre_write_overlap_check(BlockDriverState *bs, int ign, int64_t offset,
int64_t size);
+int qcow2_change_refcount_order(BlockDriverState *bs, int refcount_order,
+ BlockDriverAmendStatusCB *status_cb,
+ void *cb_opaque, Error **errp);
+
/* qcow2-cluster.c functions */
int qcow2_grow_l1_table(BlockDriverState *bs, uint64_t min_size,
bool exact_size);
--
1.9.3
next prev parent reply other threads:[~2014-11-10 13:46 UTC|newest]
Thread overview: 75+ messages / expand[flat|nested] mbox.gz Atom feed top
2014-11-10 13:45 [Qemu-devel] [PATCH 00/21] qcow2: Support refcount orders != 4 Max Reitz
2014-11-10 13:45 ` [Qemu-devel] [PATCH 01/21] qcow2: Add two new fields to BDRVQcowState Max Reitz
2014-11-10 19:00 ` Eric Blake
2014-11-10 13:45 ` [Qemu-devel] [PATCH 02/21] qcow2: Add refcount_width to format-specific info Max Reitz
2014-11-10 19:06 ` Eric Blake
2014-11-11 8:11 ` Max Reitz
2014-11-11 15:49 ` Eric Blake
2014-11-10 13:45 ` [Qemu-devel] [PATCH 03/21] qcow2: Use 64 bits for refcount values Max Reitz
2014-11-10 20:59 ` Eric Blake
2014-11-11 8:12 ` Max Reitz
2014-11-11 9:22 ` Kevin Wolf
2014-11-11 9:25 ` Max Reitz
2014-11-11 9:26 ` Max Reitz
2014-11-11 9:36 ` Kevin Wolf
2014-11-10 13:45 ` [Qemu-devel] [PATCH 04/21] qcow2: Respect error in qcow2_alloc_bytes() Max Reitz
2014-11-10 21:05 ` Eric Blake
2014-11-10 13:45 ` [Qemu-devel] [PATCH 05/21] qcow2: Refcount overflow and qcow2_alloc_bytes() Max Reitz
2014-11-10 21:12 ` Eric Blake
2014-11-11 8:22 ` Max Reitz
2014-11-11 16:13 ` Eric Blake
2014-11-11 16:18 ` Max Reitz
2014-11-11 19:49 ` Eric Blake
2014-11-12 8:52 ` Max Reitz
2014-11-10 13:45 ` [Qemu-devel] [PATCH 06/21] qcow2: Helper function for refcount modification Max Reitz
2014-11-10 22:30 ` Eric Blake
2014-11-11 8:35 ` Max Reitz
2014-11-11 9:43 ` Max Reitz
2014-11-11 10:56 ` Max Reitz
2014-11-10 13:45 ` [Qemu-devel] [PATCH 07/21] qcow2: Helper for refcount array size calculation Max Reitz
2014-11-10 22:49 ` Eric Blake
2014-11-11 8:37 ` Max Reitz
2014-11-11 10:08 ` Max Reitz
2014-11-10 13:45 ` [Qemu-devel] [PATCH 08/21] qcow2: More helpers for refcount modification Max Reitz
2014-11-11 0:29 ` Eric Blake
2014-11-11 8:42 ` Max Reitz
2014-11-10 13:45 ` [Qemu-devel] [PATCH 09/21] qcow2: Open images with refcount order != 4 Max Reitz
2014-11-10 17:03 ` Eric Blake
2014-11-10 17:06 ` Max Reitz
2014-11-10 13:45 ` [Qemu-devel] [PATCH 10/21] qcow2: refcount_order parameter for qcow2_create2 Max Reitz
2014-11-11 5:40 ` Eric Blake
2014-11-11 8:48 ` Max Reitz
2014-11-10 13:45 ` [Qemu-devel] [PATCH 11/21] iotests: Prepare for refcount_width option Max Reitz
2014-11-11 17:57 ` Eric Blake
2014-11-12 8:41 ` Max Reitz
2014-11-10 13:45 ` [Qemu-devel] [PATCH 12/21] qcow2: Allow creation with refcount order != 4 Max Reitz
2014-11-11 18:05 ` Eric Blake
2014-11-12 8:47 ` Max Reitz
2014-11-10 13:45 ` [Qemu-devel] [PATCH 13/21] block: Add opaque value to the amend CB Max Reitz
2014-11-11 18:08 ` Eric Blake
2014-11-10 13:45 ` [Qemu-devel] [PATCH 14/21] qcow2: Use error_report() in qcow2_amend_options() Max Reitz
2014-11-11 18:11 ` Eric Blake
2014-11-12 8:47 ` Max Reitz
2014-11-10 13:45 ` [Qemu-devel] [PATCH 15/21] qcow2: Use abort() instead of assert(false) Max Reitz
2014-11-11 18:12 ` Eric Blake
2014-11-12 8:48 ` Max Reitz
2014-11-10 13:45 ` [Qemu-devel] [PATCH 16/21] qcow2: Split upgrade/downgrade paths for amend Max Reitz
2014-11-11 18:14 ` Eric Blake
2014-11-10 13:45 ` [Qemu-devel] [PATCH 17/21] qcow2: Use intermediate helper CB " Max Reitz
2014-11-11 21:05 ` Eric Blake
2014-11-12 9:10 ` Max Reitz
2014-11-10 13:45 ` Max Reitz [this message]
2014-11-12 4:15 ` [Qemu-devel] [PATCH 18/21] qcow2: Add function for refcount order amendment Eric Blake
2014-11-12 9:55 ` Max Reitz
2014-11-12 13:50 ` Eric Blake
2014-11-12 14:19 ` Eric Blake
2014-11-12 14:21 ` Max Reitz
2014-11-12 17:45 ` Max Reitz
2014-11-12 20:21 ` Eric Blake
2014-11-10 13:45 ` [Qemu-devel] [PATCH 19/21] qcow2: Invoke refcount order amendment function Max Reitz
2014-11-12 4:36 ` Eric Blake
2014-11-10 13:45 ` [Qemu-devel] [PATCH 20/21] qcow2: Point to amend function in check Max Reitz
2014-11-12 4:38 ` Eric Blake
2014-11-10 13:45 ` [Qemu-devel] [PATCH 21/21] iotests: Add test for different refcount widths Max Reitz
2014-11-11 19:53 ` Eric Blake
2014-11-12 8:58 ` Max Reitz
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=1415627159-15941-19-git-send-email-mreitz@redhat.com \
--to=mreitz@redhat.com \
--cc=kwolf@redhat.com \
--cc=pl@kamp.de \
--cc=qemu-devel@nongnu.org \
--cc=stefanha@redhat.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).