From: Anton Nefedov <anton.nefedov@virtuozzo.com>
To: qemu-devel@nongnu.org
Cc: qemu-block@nongnu.org, kwolf@redhat.com, mreitz@redhat.com,
eblake@redhat.com, den@virtuozzo.com, berto@igalia.com,
Anton Nefedov <anton.nefedov@virtuozzo.com>
Subject: [Qemu-devel] [PATCH v5 13/15] qcow2: do not zero out clusters if already preallocated
Date: Wed, 1 Nov 2017 18:44:06 +0300 [thread overview]
Message-ID: <1509551048-129830-14-git-send-email-anton.nefedov@virtuozzo.com> (raw)
In-Reply-To: <1509551048-129830-1-git-send-email-anton.nefedov@virtuozzo.com>
if space preallocation feature is used, it can be detected that the space
is already zeroes, so COW can be skipped without additional zeroing of
the clusters.
Signed-off-by: Anton Nefedov <anton.nefedov@virtuozzo.com>
---
block/qcow2.h | 6 ++++++
block/qcow2-cluster.c | 2 ++
block/qcow2.c | 45 +++++++++++++++++++++++++++++++--------------
3 files changed, 39 insertions(+), 14 deletions(-)
diff --git a/block/qcow2.h b/block/qcow2.h
index 579838d..5d96748 100644
--- a/block/qcow2.h
+++ b/block/qcow2.h
@@ -363,6 +363,12 @@ typedef struct QCowL2Meta
bool keep_old_clusters;
/**
+ * True if the area is allocated at the end of data area
+ * (i.e. >= BDRVQcow2State::data_end)
+ */
+ bool clusters_are_trailing;
+
+ /**
* Requests that overlap with this allocation and wait to be restarted
* when the allocating request has completed.
*/
diff --git a/block/qcow2-cluster.c b/block/qcow2-cluster.c
index 9b24cab..5698919 100644
--- a/block/qcow2-cluster.c
+++ b/block/qcow2-cluster.c
@@ -1251,6 +1251,7 @@ static int handle_alloc(BlockDriverState *bs, uint64_t guest_offset,
uint64_t *host_offset, uint64_t *bytes, QCowL2Meta **m)
{
BDRVQcow2State *s = bs->opaque;
+ const uint64_t old_data_end = s->data_end;
int l2_index;
uint64_t *l2_table;
uint64_t entry;
@@ -1372,6 +1373,7 @@ static int handle_alloc(BlockDriverState *bs, uint64_t guest_offset,
.alloc_offset = alloc_cluster_offset,
.offset = start_of_cluster(s, guest_offset),
.nb_clusters = nb_clusters,
+ .clusters_are_trailing = alloc_cluster_offset >= old_data_end,
.keep_old_clusters = keep_old_clusters,
diff --git a/block/qcow2.c b/block/qcow2.c
index ef65b5f..6019cf4 100644
--- a/block/qcow2.c
+++ b/block/qcow2.c
@@ -1975,26 +1975,40 @@ static bool is_zero_cow(BlockDriverState *bs, QCowL2Meta *m)
/*
* If the specified area is beyond EOF, allocates it + prealloc_size
* bytes ahead.
+ *
+ * Returns
+ * true if the space is allocated and contains zeroes
*/
-static void coroutine_fn handle_prealloc(BlockDriverState *bs,
+static bool coroutine_fn handle_prealloc(BlockDriverState *bs,
const QCowL2Meta *m)
{
BDRVQcow2State *s = bs->opaque;
uint64_t start = m->alloc_offset;
uint64_t end = start + m->nb_clusters * s->cluster_size;
+ int ret;
int64_t flen = bdrv_getlength(bs->file->bs);
if (flen < 0) {
- return;
+ return false;
}
if (end > flen) {
/* try to alloc host space in one chunk for better locality */
- bdrv_co_pwrite_zeroes(bs->file, flen,
- QEMU_ALIGN_UP(end + s->prealloc_size - flen,
- s->cluster_size),
- BDRV_REQ_ALLOCATE);
+ ret = bdrv_co_pwrite_zeroes(bs->file, flen,
+ QEMU_ALIGN_UP(end + s->prealloc_size - flen,
+ s->cluster_size),
+ BDRV_REQ_ALLOCATE);
+ if (ret < 0) {
+ return false;
+ }
}
+
+ /* We're safe to assume that the area is zeroes if the area
+ * was allocated at the end of data (s->data_end).
+ * In this case, the only way for file length to be bigger is that
+ * the area was preallocated by this or another request.
+ */
+ return m->clusters_are_trailing;
}
static void handle_alloc_space(BlockDriverState *bs, QCowL2Meta *l2meta)
@@ -2004,9 +2018,10 @@ static void handle_alloc_space(BlockDriverState *bs, QCowL2Meta *l2meta)
for (m = l2meta; m != NULL; m = m->next) {
int ret;
+ bool preallocated_zeroes = false;
if (s->prealloc_size) {
- handle_prealloc(bs, m);
+ preallocated_zeroes = handle_prealloc(bs, m);
}
if (!m->cow_start.nb_bytes && !m->cow_end.nb_bytes) {
@@ -2017,13 +2032,15 @@ static void handle_alloc_space(BlockDriverState *bs, QCowL2Meta *l2meta)
continue;
}
- /* instead of writing zero COW buffers,
- efficiently zero out the whole clusters */
- ret = bdrv_co_pwrite_zeroes(bs->file, m->alloc_offset,
- m->nb_clusters * s->cluster_size,
- BDRV_REQ_ALLOCATE);
- if (ret < 0) {
- continue;
+ if (!preallocated_zeroes) {
+ /* instead of writing zero COW buffers,
+ efficiently zero out the whole clusters */
+ ret = bdrv_co_pwrite_zeroes(bs->file, m->alloc_offset,
+ m->nb_clusters * s->cluster_size,
+ BDRV_REQ_ALLOCATE);
+ if (ret < 0) {
+ continue;
+ }
}
trace_qcow2_skip_cow(qemu_coroutine_self(), m->offset, m->nb_clusters);
--
2.7.4
next prev parent reply other threads:[~2017-11-01 15:44 UTC|newest]
Thread overview: 24+ messages / expand[flat|nested] mbox.gz Atom feed top
2017-11-01 15:43 [Qemu-devel] [PATCH v5 00/15] qcow2: space preallocation and COW improvements Anton Nefedov
2017-11-01 15:43 ` [Qemu-devel] [PATCH v5 01/15] mirror: inherit supported write/zero flags Anton Nefedov
2018-01-11 14:14 ` Alberto Garcia
2017-11-01 15:43 ` [Qemu-devel] [PATCH v5 02/15] blkverify: set " Anton Nefedov
2018-01-11 14:17 ` Alberto Garcia
2017-11-01 15:43 ` [Qemu-devel] [PATCH v5 03/15] block: introduce BDRV_REQ_ALLOCATE flag Anton Nefedov
2017-11-01 15:43 ` [Qemu-devel] [PATCH v5 04/15] block: treat BDRV_REQ_ALLOCATE as serialising Anton Nefedov
2017-11-01 15:43 ` [Qemu-devel] [PATCH v5 05/15] file-posix: support BDRV_REQ_ALLOCATE Anton Nefedov
2017-11-01 15:43 ` [Qemu-devel] [PATCH v5 06/15] block: support BDRV_REQ_ALLOCATE in passthrough drivers Anton Nefedov
2017-11-01 15:44 ` [Qemu-devel] [PATCH v5 07/15] qcow2: move is_zero() up Anton Nefedov
2018-01-15 15:14 ` Alberto Garcia
2017-11-01 15:44 ` [Qemu-devel] [PATCH v5 08/15] qcow2: skip writing zero buffers to empty COW areas Anton Nefedov
2018-01-15 15:31 ` Alberto Garcia
2018-01-15 18:21 ` Anton Nefedov
2018-01-15 20:12 ` Alberto Garcia
2017-11-01 15:44 ` [Qemu-devel] [PATCH v5 09/15] qcow2: preallocation at image expand Anton Nefedov
2017-11-01 15:44 ` [Qemu-devel] [PATCH v5 10/15] qcow2: set inactive flag Anton Nefedov
2017-11-01 15:44 ` [Qemu-devel] [PATCH v5 11/15] qcow2: truncate preallocated space Anton Nefedov
2017-11-01 15:44 ` [Qemu-devel] [PATCH v5 12/15] qcow2: check space leak at the end of the image Anton Nefedov
2017-11-01 15:44 ` Anton Nefedov [this message]
2017-11-01 15:44 ` [Qemu-devel] [PATCH v5 14/15] iotest 198: test BDRV_REQ_ALLOCATE Anton Nefedov
2017-11-01 15:44 ` [Qemu-devel] [PATCH v5 15/15] iotest 134: test cluster-misaligned encrypted write Anton Nefedov
2017-12-04 22:32 ` [Qemu-devel] [Qemu-block] [PATCH v5 00/15] qcow2: space preallocation and COW improvements John Snow
2017-12-05 17:28 ` Anton Nefedov
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=1509551048-129830-14-git-send-email-anton.nefedov@virtuozzo.com \
--to=anton.nefedov@virtuozzo.com \
--cc=berto@igalia.com \
--cc=den@virtuozzo.com \
--cc=eblake@redhat.com \
--cc=kwolf@redhat.com \
--cc=mreitz@redhat.com \
--cc=qemu-block@nongnu.org \
--cc=qemu-devel@nongnu.org \
/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).