From: Alberto Garcia <berto@igalia.com>
To: qemu-devel@nongnu.org
Cc: Kevin Wolf <kwolf@redhat.com>,
Anton Nefedov <anton.nefedov@virtuozzo.com>,
Alberto Garcia <berto@igalia.com>,
qemu-block@nongnu.org, Max Reitz <mreitz@redhat.com>,
Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>,
"Denis V . Lunev" <den@openvz.org>
Subject: [RFC PATCH v2 03/26] qcow2: Process QCOW2_CLUSTER_ZERO_ALLOC clusters in handle_copied()
Date: Sun, 27 Oct 2019 00:25:05 +0300 [thread overview]
Message-ID: <fe73f28b44ecaea8a0104e11078f38f563da5925.1572125022.git.berto@igalia.com> (raw)
In-Reply-To: <cover.1572125022.git.berto@igalia.com>
When writing to a qcow2 file there are two functions that take a
virtual offset and return a host offset, possibly allocating new
clusters if necessary:
- handle_copied() looks for normal data clusters that are already
allocated and have a reference count of 1. In those clusters we
can simply write the data and there is no need to perform any
copy-on-write.
- handle_alloc() looks for clusters that do need copy-on-write,
either because they haven't been allocated yet, because their
reference count is != 1 or because they are ZERO_ALLOC clusters.
The ZERO_ALLOC case is a bit special because those are clusters that
are already allocated and they could perfectly be dealt with in
handle_copied() (as long as copy-on-write is performed when required).
In fact, there is extra code specifically for them in handle_alloc()
that tries to reuse the existing allocation if possible and frees them
otherwise.
This patch changes the handling of ZERO_ALLOC clusters so the
semantics of these two functions are now like this:
- handle_copied() looks for clusters that are already allocated and
which we can overwrite (NORMAL and ZERO_ALLOC clusters with a
reference count of 1).
- handle_alloc() looks for clusters for which we need a new
allocation (all other cases).
One importante difference after this change is that clusters found in
handle_copied() may now require copy-on-write, but this will be anyway
necessary once we add support for subclusters.
Signed-off-by: Alberto Garcia <berto@igalia.com>
---
block/qcow2-cluster.c | 177 +++++++++++++++++++++++-------------------
1 file changed, 96 insertions(+), 81 deletions(-)
diff --git a/block/qcow2-cluster.c b/block/qcow2-cluster.c
index aa1010a515..ee6b46f917 100644
--- a/block/qcow2-cluster.c
+++ b/block/qcow2-cluster.c
@@ -1021,7 +1021,8 @@ void qcow2_alloc_cluster_abort(BlockDriverState *bs, QCowL2Meta *m)
/*
* For a given write request, create a new QCowL2Meta structure and
- * add it to @m.
+ * add it to @m. If the write request does not need copy-on-write or
+ * changes to the L2 metadata then this function does nothing.
*
* @host_offset points to the beginning of the first cluster.
*
@@ -1034,15 +1035,51 @@ void qcow2_alloc_cluster_abort(BlockDriverState *bs, QCowL2Meta *m)
*/
static void calculate_l2_meta(BlockDriverState *bs, uint64_t host_offset,
uint64_t guest_offset, uint64_t bytes,
- QCowL2Meta **m, bool keep_old)
+ uint64_t *l2_slice, QCowL2Meta **m, bool keep_old)
{
BDRVQcow2State *s = bs->opaque;
- unsigned cow_start_from = 0;
+ int l2_index = offset_to_l2_slice_index(s, guest_offset);
+ uint64_t l2_entry;
+ unsigned cow_start_from, cow_end_to;
unsigned cow_start_to = offset_into_cluster(s, guest_offset);
unsigned cow_end_from = cow_start_to + bytes;
- unsigned cow_end_to = ROUND_UP(cow_end_from, s->cluster_size);
unsigned nb_clusters = size_to_clusters(s, cow_end_from);
QCowL2Meta *old_m = *m;
+ QCow2ClusterType type;
+
+ /* Return if there's no COW (all clusters are normal and we keep them) */
+ if (keep_old) {
+ int i;
+ for (i = 0; i < nb_clusters; i++) {
+ l2_entry = be64_to_cpu(l2_slice[l2_index + i]);
+ if (qcow2_get_cluster_type(bs, l2_entry) != QCOW2_CLUSTER_NORMAL) {
+ break;
+ }
+ }
+ if (i == nb_clusters) {
+ return;
+ }
+ }
+
+ /* Get the L2 entry from the first cluster */
+ l2_entry = be64_to_cpu(l2_slice[l2_index]);
+ type = qcow2_get_cluster_type(bs, l2_entry);
+
+ if (type == QCOW2_CLUSTER_NORMAL && keep_old) {
+ cow_start_from = cow_start_to;
+ } else {
+ cow_start_from = 0;
+ }
+
+ /* Get the L2 entry from the last cluster */
+ l2_entry = be64_to_cpu(l2_slice[l2_index + nb_clusters - 1]);
+ type = qcow2_get_cluster_type(bs, l2_entry);
+
+ if (type == QCOW2_CLUSTER_NORMAL && keep_old) {
+ cow_end_to = cow_end_from;
+ } else {
+ cow_end_to = ROUND_UP(cow_end_from, s->cluster_size);
+ }
*m = g_malloc0(sizeof(**m));
**m = (QCowL2Meta) {
@@ -1068,18 +1105,18 @@ static void calculate_l2_meta(BlockDriverState *bs, uint64_t host_offset,
QLIST_INSERT_HEAD(&s->cluster_allocs, *m, next_in_flight);
}
-/* Returns true if writing to a cluster requires COW */
+/* Returns true if the cluster is unallocated or has refcount > 1 */
static bool cluster_needs_cow(BlockDriverState *bs, uint64_t l2_entry)
{
switch (qcow2_get_cluster_type(bs, l2_entry)) {
case QCOW2_CLUSTER_NORMAL:
+ case QCOW2_CLUSTER_ZERO_ALLOC:
if (l2_entry & QCOW_OFLAG_COPIED) {
return false;
}
case QCOW2_CLUSTER_UNALLOCATED:
case QCOW2_CLUSTER_COMPRESSED:
case QCOW2_CLUSTER_ZERO_PLAIN:
- case QCOW2_CLUSTER_ZERO_ALLOC:
return true;
default:
abort();
@@ -1087,20 +1124,34 @@ static bool cluster_needs_cow(BlockDriverState *bs, uint64_t l2_entry)
}
/*
- * 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)
+ * Returns the number of contiguous clusters that can be written to
+ * using one single write request, starting from @l2_index.
+ * At most @nb_clusters are checked.
+ *
+ * If @want_cow is true this counts clusters that are either
+ * unallocated, or allocated but with refcount > 1.
+ *
+ * If @want_cow is false this counts clusters that are already
+ * allocated and can be written to using their current locations
+ * (including QCOW2_CLUSTER_ZERO_ALLOC).
*/
static int count_cow_clusters(BlockDriverState *bs, int nb_clusters,
- uint64_t *l2_slice, int l2_index)
+ uint64_t *l2_slice, int l2_index, bool want_cow)
{
+ BDRVQcow2State *s = bs->opaque;
+ uint64_t l2_entry = be64_to_cpu(l2_slice[l2_index]);
+ uint64_t expected_offset = l2_entry & L2E_OFFSET_MASK;
int i;
for (i = 0; i < nb_clusters; i++) {
- uint64_t l2_entry = be64_to_cpu(l2_slice[l2_index + i]);
- if (!cluster_needs_cow(bs, l2_entry)) {
+ l2_entry = be64_to_cpu(l2_slice[l2_index + i]);
+ if (cluster_needs_cow(bs, l2_entry) != want_cow) {
break;
}
+ if (!want_cow && expected_offset != (l2_entry & L2E_OFFSET_MASK)) {
+ break;
+ }
+ expected_offset += s->cluster_size;
}
assert(i <= nb_clusters);
@@ -1228,18 +1279,17 @@ static int handle_copied(BlockDriverState *bs, uint64_t guest_offset,
cluster_offset = be64_to_cpu(l2_slice[l2_index]);
- /* Check how many clusters are already allocated and don't need COW */
- if (qcow2_get_cluster_type(bs, cluster_offset) == QCOW2_CLUSTER_NORMAL
- && (cluster_offset & QCOW_OFLAG_COPIED))
- {
+ if (!cluster_needs_cow(bs, cluster_offset)) {
/* If a specific host_offset is required, check it */
bool offset_matches =
(cluster_offset & L2E_OFFSET_MASK) == *host_offset;
if (offset_into_cluster(s, cluster_offset & L2E_OFFSET_MASK)) {
- qcow2_signal_corruption(bs, true, -1, -1, "Data cluster offset "
+ qcow2_signal_corruption(bs, true, -1, -1, "%s cluster offset "
"%#llx unaligned (guest offset: %#" PRIx64
- ")", cluster_offset & L2E_OFFSET_MASK,
+ ")", cluster_offset & QCOW_OFLAG_ZERO ?
+ "Preallocated zero" : "Data",
+ cluster_offset & L2E_OFFSET_MASK,
guest_offset);
ret = -EIO;
goto out;
@@ -1252,15 +1302,17 @@ static int handle_copied(BlockDriverState *bs, uint64_t guest_offset,
}
/* We keep all QCOW_OFLAG_COPIED clusters */
- keep_clusters =
- count_contiguous_clusters(bs, nb_clusters, s->cluster_size,
- &l2_slice[l2_index],
- QCOW_OFLAG_COPIED | QCOW_OFLAG_ZERO);
+ keep_clusters = count_cow_clusters(bs, nb_clusters, l2_slice,
+ l2_index, false);
assert(keep_clusters <= nb_clusters);
*bytes = MIN(*bytes,
keep_clusters * s->cluster_size
- offset_into_cluster(s, guest_offset));
+ assert(*bytes != 0);
+
+ calculate_l2_meta(bs, cluster_offset & L2E_OFFSET_MASK, guest_offset,
+ *bytes, l2_slice, m, true);
ret = 1;
} else {
@@ -1361,12 +1413,10 @@ static int handle_alloc(BlockDriverState *bs, uint64_t guest_offset,
BDRVQcow2State *s = bs->opaque;
int l2_index;
uint64_t *l2_slice;
- uint64_t entry;
uint64_t nb_clusters;
int ret;
- bool keep_old_clusters = false;
- uint64_t alloc_cluster_offset = INV_OFFSET;
+ uint64_t alloc_cluster_offset;
trace_qcow2_handle_alloc(qemu_coroutine_self(), guest_offset, *host_offset,
*bytes);
@@ -1392,67 +1442,31 @@ static int handle_alloc(BlockDriverState *bs, uint64_t guest_offset,
return ret;
}
- entry = be64_to_cpu(l2_slice[l2_index]);
- nb_clusters = count_cow_clusters(bs, nb_clusters, l2_slice, l2_index);
+ nb_clusters = count_cow_clusters(bs, nb_clusters, l2_slice, l2_index, true);
/* 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);
- if (qcow2_get_cluster_type(bs, entry) == QCOW2_CLUSTER_ZERO_ALLOC &&
- (entry & QCOW_OFLAG_COPIED) &&
- (*host_offset == INV_OFFSET ||
- start_of_cluster(s, *host_offset) == (entry & L2E_OFFSET_MASK)))
- {
- int preallocated_nb_clusters;
-
- if (offset_into_cluster(s, entry & L2E_OFFSET_MASK)) {
- qcow2_signal_corruption(bs, true, -1, -1, "Preallocated zero "
- "cluster offset %#llx unaligned (guest "
- "offset: %#" PRIx64 ")",
- entry & L2E_OFFSET_MASK, guest_offset);
- ret = -EIO;
- goto fail;
- }
-
- /* Try to reuse preallocated zero clusters; contiguous normal clusters
- * would be fine, too, but count_cow_clusters() above has limited
- * nb_clusters already to a range of COW clusters */
- preallocated_nb_clusters =
- count_contiguous_clusters(bs, nb_clusters, s->cluster_size,
- &l2_slice[l2_index], QCOW_OFLAG_COPIED);
- assert(preallocated_nb_clusters > 0);
-
- nb_clusters = preallocated_nb_clusters;
- alloc_cluster_offset = entry & L2E_OFFSET_MASK;
-
- /* We want to reuse these clusters, so qcow2_alloc_cluster_link_l2()
- * should not free them. */
- keep_old_clusters = true;
+ /* Allocate, if necessary at a given offset in the image file */
+ alloc_cluster_offset = *host_offset == INV_OFFSET ? INV_OFFSET :
+ start_of_cluster(s, *host_offset);
+ ret = do_alloc_cluster_offset(bs, guest_offset, &alloc_cluster_offset,
+ &nb_clusters);
+ if (ret < 0) {
+ goto out;
}
- qcow2_cache_put(s->l2_table_cache, (void **) &l2_slice);
-
- if (alloc_cluster_offset == INV_OFFSET) {
- /* Allocate, if necessary at a given offset in the image file */
- alloc_cluster_offset = *host_offset == INV_OFFSET ? INV_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;
- }
-
- /* Can't extend contiguous allocation */
- if (nb_clusters == 0) {
- *bytes = 0;
- return 0;
- }
-
- assert(alloc_cluster_offset != INV_OFFSET);
+ /* Can't extend contiguous allocation */
+ if (nb_clusters == 0) {
+ *bytes = 0;
+ ret = 0;
+ goto out;
}
+ assert(alloc_cluster_offset != INV_OFFSET);
+
/*
* Save info needed for meta data update.
*
@@ -1475,13 +1489,14 @@ static int handle_alloc(BlockDriverState *bs, uint64_t guest_offset,
*bytes = MIN(*bytes, nb_bytes - offset_into_cluster(s, guest_offset));
assert(*bytes != 0);
- calculate_l2_meta(bs, alloc_cluster_offset, guest_offset, *bytes,
- m, keep_old_clusters);
+ calculate_l2_meta(bs, alloc_cluster_offset, guest_offset, *bytes, l2_slice,
+ m, false);
- return 1;
+ ret = 1;
-fail:
- if (*m && (*m)->nb_clusters > 0) {
+out:
+ qcow2_cache_put(s->l2_table_cache, (void **) &l2_slice);
+ if (ret < 0 && *m && (*m)->nb_clusters > 0) {
QLIST_REMOVE(*m, next_in_flight);
}
return ret;
--
2.20.1
next prev parent reply other threads:[~2019-10-26 21:45 UTC|newest]
Thread overview: 65+ messages / expand[flat|nested] mbox.gz Atom feed top
2019-10-26 21:25 [RFC PATCH v2 00/26] Add subcluster allocation to qcow2 Alberto Garcia
2019-10-26 21:25 ` [RFC PATCH v2 01/26] qcow2: Add calculate_l2_meta() Alberto Garcia
2019-10-28 12:50 ` Vladimir Sementsov-Ogievskiy
2019-10-30 15:56 ` Alberto Garcia
2019-10-30 12:04 ` Max Reitz
2019-10-30 16:02 ` Alberto Garcia
2019-10-26 21:25 ` [RFC PATCH v2 02/26] qcow2: Split cluster_needs_cow() out of count_cow_clusters() Alberto Garcia
2019-10-28 13:55 ` Vladimir Sementsov-Ogievskiy
2019-10-26 21:25 ` Alberto Garcia [this message]
2019-10-30 14:24 ` [RFC PATCH v2 03/26] qcow2: Process QCOW2_CLUSTER_ZERO_ALLOC clusters in handle_copied() Max Reitz
2019-11-13 14:25 ` Alberto Garcia
2019-10-26 21:25 ` [RFC PATCH v2 04/26] qcow2: Add get_l2_entry() and set_l2_entry() Alberto Garcia
2019-10-26 21:25 ` [RFC PATCH v2 05/26] qcow2: Document the Extended L2 Entries feature Alberto Garcia
2019-10-30 16:23 ` Max Reitz
2019-10-30 22:38 ` Alberto Garcia
2019-10-26 21:25 ` [RFC PATCH v2 06/26] qcow2: Add dummy has_subclusters() function Alberto Garcia
2019-10-26 21:25 ` [RFC PATCH v2 07/26] qcow2: Add subcluster-related fields to BDRVQcow2State Alberto Garcia
2019-10-26 21:25 ` [RFC PATCH v2 08/26] qcow2: Add offset_to_sc_index() Alberto Garcia
2019-10-26 21:25 ` [RFC PATCH v2 09/26] qcow2: Add l2_entry_size() Alberto Garcia
2019-10-30 16:47 ` Max Reitz
2019-10-26 21:25 ` [RFC PATCH v2 10/26] qcow2: Update get/set_l2_entry() and add get/set_l2_bitmap() Alberto Garcia
2019-10-30 16:55 ` Max Reitz
2019-11-14 13:57 ` Alberto Garcia
2019-10-26 21:25 ` [RFC PATCH v2 11/26] qcow2: Add qcow2_get_subcluster_type() Alberto Garcia
2019-11-04 12:35 ` Max Reitz
2019-11-04 15:01 ` Max Reitz
2019-10-26 21:25 ` [RFC PATCH v2 12/26] qcow2: Handle QCOW2_CLUSTER_UNALLOCATED_SUBCLUSTER Alberto Garcia
2019-11-04 12:57 ` Max Reitz
2019-11-04 13:03 ` Alberto Garcia
2019-11-04 13:10 ` Max Reitz
2019-11-07 14:44 ` Alberto Garcia
2019-10-26 21:25 ` [RFC PATCH v2 13/26] qcow2: Add subcluster support to calculate_l2_meta() Alberto Garcia
2019-11-04 14:21 ` Max Reitz
2019-11-08 15:18 ` Alberto Garcia
2019-10-26 21:25 ` [RFC PATCH v2 14/26] qcow2: Add subcluster support to qcow2_get_cluster_offset() Alberto Garcia
2019-11-04 14:58 ` Max Reitz
2019-11-08 15:42 ` Alberto Garcia
2019-11-11 8:42 ` Max Reitz
2019-10-26 21:25 ` [RFC PATCH v2 15/26] qcow2: Add subcluster support to zero_in_l2_slice() Alberto Garcia
2019-11-04 15:04 ` Max Reitz
2019-11-04 15:10 ` Max Reitz
2019-11-14 15:31 ` Alberto Garcia
2019-10-26 21:25 ` [RFC PATCH v2 16/26] qcow2: Add subcluster support to discard_in_l2_slice() Alberto Garcia
2019-11-04 15:07 ` Max Reitz
2019-11-14 15:33 ` Alberto Garcia
2019-11-14 16:22 ` Max Reitz
2019-10-26 21:25 ` [RFC PATCH v2 17/26] qcow2: Add subcluster support to check_refcounts_l2() Alberto Garcia
2019-10-26 21:25 ` [RFC PATCH v2 18/26] qcow2: Add subcluster support to expand_zero_clusters_in_l1() Alberto Garcia
2019-11-05 11:05 ` Max Reitz
2019-11-14 15:43 ` Alberto Garcia
2019-10-26 21:25 ` [RFC PATCH v2 19/26] qcow2: Fix offset calculation in handle_dependencies() Alberto Garcia
2019-10-26 21:25 ` [RFC PATCH v2 20/26] qcow2: Update L2 bitmap in qcow2_alloc_cluster_link_l2() Alberto Garcia
2019-11-05 11:43 ` Max Reitz
2019-11-14 16:30 ` Alberto Garcia
2019-10-26 21:25 ` [RFC PATCH v2 21/26] qcow2: Clear the L2 bitmap when allocating a compressed cluster Alberto Garcia
2019-10-26 21:25 ` [RFC PATCH v2 22/26] qcow2: Add subcluster support to handle_alloc_space() Alberto Garcia
2019-11-05 12:05 ` Max Reitz
2019-10-26 21:25 ` [RFC PATCH v2 23/26] qcow2: Restrict qcow2_co_pwrite_zeroes() to full clusters only Alberto Garcia
2019-10-26 21:25 ` [RFC PATCH v2 24/26] qcow2: Add the 'extended_l2' option and the QCOW2_INCOMPAT_EXTL2 bit Alberto Garcia
2019-11-05 12:47 ` Max Reitz
2019-11-15 13:35 ` Alberto Garcia
2019-10-26 21:25 ` [RFC PATCH v2 25/26] qcow2: Allow preallocation and backing files if extended_l2 is set Alberto Garcia
2019-11-05 13:11 ` Max Reitz
2019-10-26 21:25 ` [RFC PATCH v2 26/26] iotests: Add tests for qcow2 images with extended L2 entries Alberto Garcia
2019-11-05 13:32 ` [RFC PATCH v2 00/26] Add subcluster allocation to qcow2 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=fe73f28b44ecaea8a0104e11078f38f563da5925.1572125022.git.berto@igalia.com \
--to=berto@igalia.com \
--cc=anton.nefedov@virtuozzo.com \
--cc=den@openvz.org \
--cc=kwolf@redhat.com \
--cc=mreitz@redhat.com \
--cc=qemu-block@nongnu.org \
--cc=qemu-devel@nongnu.org \
--cc=vsementsov@virtuozzo.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).