From: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
To: qemu-block@nongnu.org
Cc: fam@euphon.net, kwolf@redhat.com, vsementsov@virtuozzo.com,
armbru@redhat.com, qemu-devel@nongnu.org, mreitz@redhat.com,
stefanha@redhat.com, den@openvz.org, jsnow@redhat.com
Subject: [Qemu-devel] [PATCH v3 6/7] block/backup: teach backup_cow_with_bounce_buffer to copy more at once
Date: Sat, 10 Aug 2019 22:31:54 +0300 [thread overview]
Message-ID: <20190810193155.58637-7-vsementsov@virtuozzo.com> (raw)
In-Reply-To: <20190810193155.58637-1-vsementsov@virtuozzo.com>
backup_cow_with_offload can transfer more than one cluster. Let
backup_cow_with_bounce_buffer behave similarly. It reduces the number
of IO requests, since there is no need to copy cluster by cluster.
Logic around bounce_buffer allocation changed: we can't just allocate
one-cluster-sized buffer to share for all iterations. We can't also
allocate buffer of full-request length it may be too large, so
BACKUP_MAX_BOUNCE_BUFFER is introduced. And finally, allocation logic
is to allocate a buffer sufficient to handle all remaining iterations
at the point where we need the buffer for the first time.
Bonus: get rid of pointer-to-pointer.
Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
---
block/backup.c | 65 +++++++++++++++++++++++++++++++-------------------
1 file changed, 41 insertions(+), 24 deletions(-)
diff --git a/block/backup.c b/block/backup.c
index d482d93458..65f7212c85 100644
--- a/block/backup.c
+++ b/block/backup.c
@@ -27,6 +27,7 @@
#include "qemu/error-report.h"
#define BACKUP_CLUSTER_SIZE_DEFAULT (1 << 16)
+#define BACKUP_MAX_BOUNCE_BUFFER (64 * 1024 * 1024)
typedef struct CowRequest {
int64_t start_byte;
@@ -98,44 +99,55 @@ static void cow_request_end(CowRequest *req)
qemu_co_queue_restart_all(&req->wait_queue);
}
-/* Copy range to target with a bounce buffer and return the bytes copied. If
- * error occurred, return a negative error number */
+/*
+ * Copy range to target with a bounce buffer and return the bytes copied. If
+ * error occurred, return a negative error number
+ *
+ * @bounce_buffer is assumed to enough to store
+ * MIN(BACKUP_MAX_BOUNCE_BUFFER, @end - @start) bytes
+ */
static int coroutine_fn backup_cow_with_bounce_buffer(BackupBlockJob *job,
int64_t start,
int64_t end,
bool is_write_notifier,
bool *error_is_read,
- void **bounce_buffer)
+ void *bounce_buffer)
{
int ret;
BlockBackend *blk = job->common.blk;
- int nbytes;
+ int nbytes, remaining_bytes;
int read_flags = is_write_notifier ? BDRV_REQ_NO_SERIALISING : 0;
assert(QEMU_IS_ALIGNED(start, job->cluster_size));
- bdrv_reset_dirty_bitmap(job->copy_bitmap, start, job->cluster_size);
- nbytes = MIN(job->cluster_size, job->len - start);
- if (!*bounce_buffer) {
- *bounce_buffer = blk_blockalign(blk, job->cluster_size);
- }
+ bdrv_reset_dirty_bitmap(job->copy_bitmap, start, end - start);
+ nbytes = MIN(end - start, job->len - start);
- ret = blk_co_pread(blk, start, nbytes, *bounce_buffer, read_flags);
- if (ret < 0) {
- trace_backup_do_cow_read_fail(job, start, ret);
- if (error_is_read) {
- *error_is_read = true;
+
+ remaining_bytes = nbytes;
+ while (remaining_bytes) {
+ int chunk = MIN(BACKUP_MAX_BOUNCE_BUFFER, remaining_bytes);
+
+ ret = blk_co_pread(blk, start, chunk, bounce_buffer, read_flags);
+ if (ret < 0) {
+ trace_backup_do_cow_read_fail(job, start, ret);
+ if (error_is_read) {
+ *error_is_read = true;
+ }
+ goto fail;
}
- goto fail;
- }
- ret = blk_co_pwrite(job->target, start, nbytes, *bounce_buffer,
- job->write_flags);
- if (ret < 0) {
- trace_backup_do_cow_write_fail(job, start, ret);
- if (error_is_read) {
- *error_is_read = false;
+ ret = blk_co_pwrite(job->target, start, chunk, bounce_buffer,
+ job->write_flags);
+ if (ret < 0) {
+ trace_backup_do_cow_write_fail(job, start, ret);
+ if (error_is_read) {
+ *error_is_read = false;
+ }
+ goto fail;
}
- goto fail;
+
+ start += chunk;
+ remaining_bytes -= chunk;
}
return nbytes;
@@ -301,9 +313,14 @@ static int coroutine_fn backup_do_cow(BackupBlockJob *job,
}
}
if (!job->use_copy_range) {
+ if (!bounce_buffer) {
+ size_t len = MIN(BACKUP_MAX_BOUNCE_BUFFER,
+ MAX(dirty_end - start, end - dirty_end));
+ bounce_buffer = blk_try_blockalign(job->common.blk, len);
+ }
ret = backup_cow_with_bounce_buffer(job, start, dirty_end,
is_write_notifier,
- error_is_read, &bounce_buffer);
+ error_is_read, bounce_buffer);
}
if (ret < 0) {
break;
--
2.18.0
next prev parent reply other threads:[~2019-08-10 19:34 UTC|newest]
Thread overview: 24+ messages / expand[flat|nested] mbox.gz Atom feed top
2019-08-10 19:31 [Qemu-devel] [PATCH v3 0/7] backup improvements Vladimir Sementsov-Ogievskiy
2019-08-10 19:31 ` [Qemu-devel] [PATCH v3 1/7] block/backup: deal with zero detection Vladimir Sementsov-Ogievskiy
2019-08-10 19:31 ` [Qemu-devel] [PATCH v3 2/7] block/backup: refactor write_flags Vladimir Sementsov-Ogievskiy
2019-08-10 19:31 ` [Qemu-devel] [PATCH v3 3/7] block/io: handle alignment and max_transfer for copy_range Vladimir Sementsov-Ogievskiy
2019-08-12 14:48 ` Max Reitz
2019-08-20 15:18 ` Vladimir Sementsov-Ogievskiy
2019-08-10 19:31 ` [Qemu-devel] [PATCH v3 4/7] block/backup: drop handling of " Vladimir Sementsov-Ogievskiy
2019-08-10 19:31 ` [Qemu-devel] [PATCH v3 5/7] block/backup: fix backup_cow_with_offload for last cluster Vladimir Sementsov-Ogievskiy
2019-08-10 19:31 ` Vladimir Sementsov-Ogievskiy [this message]
2019-08-12 15:10 ` [Qemu-devel] [PATCH v3 6/7] block/backup: teach backup_cow_with_bounce_buffer to copy more at once Max Reitz
2019-08-12 15:47 ` Vladimir Sementsov-Ogievskiy
2019-08-12 16:11 ` Max Reitz
2019-08-12 16:37 ` Vladimir Sementsov-Ogievskiy
2019-08-13 14:14 ` Vladimir Sementsov-Ogievskiy
2019-08-13 14:23 ` Max Reitz
2019-08-13 14:39 ` Vladimir Sementsov-Ogievskiy
2019-08-13 14:57 ` Max Reitz
2019-08-13 15:00 ` Vladimir Sementsov-Ogievskiy
2019-08-13 15:02 ` Max Reitz
2019-08-13 15:32 ` Vladimir Sementsov-Ogievskiy
2019-08-13 16:30 ` Max Reitz
2019-08-13 16:45 ` Vladimir Sementsov-Ogievskiy
2019-08-13 16:51 ` Max Reitz
2019-08-10 19:31 ` [Qemu-devel] [PATCH v3 7/7] block/backup: merge duplicated logic into backup_do_cow Vladimir Sementsov-Ogievskiy
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=20190810193155.58637-7-vsementsov@virtuozzo.com \
--to=vsementsov@virtuozzo.com \
--cc=armbru@redhat.com \
--cc=den@openvz.org \
--cc=fam@euphon.net \
--cc=jsnow@redhat.com \
--cc=kwolf@redhat.com \
--cc=mreitz@redhat.com \
--cc=qemu-block@nongnu.org \
--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).