From: "Denis V. Lunev" <den@openvz.org>
Cc: Kevin Wolf <kwolf@redhat.com>, "Denis V. Lunev" <den@openvz.org>,
qemu-devel@nongnu.org, Stefan Hajnoczi <stefanha@redhat.com>,
Roman Karan <rkagan@parallels.com>
Subject: [Qemu-devel] [PATCH 26/27] block/parallels: optimize linear image expansion
Date: Tue, 10 Mar 2015 11:51:20 +0300 [thread overview]
Message-ID: <1425977481-13317-27-git-send-email-den@openvz.org> (raw)
In-Reply-To: <1425977481-13317-1-git-send-email-den@openvz.org>
Plain image expansion spends a lot of time to update image file size.
This seriously affects the performance. The following simple test
qemu_img create -f parallels -o cluster_size=64k ./1.hds 64G
qemu_io -n -c "write -P 0x11 0 1024M" ./1.hds
could be improved if the format driver will pre-allocate some space
in the image file with a reasonable chunk.
This patch preallocates 128 Mb using bdrv_write_zeroes, which should
normally use fallocate() call inside. Fallback to older truncate()
could be used as a fallback using image open options thanks to the
previous patch.
The benefit is around 15%.
Signed-off-by: Denis V. Lunev <den@openvz.org>
CC: Roman Karan <rkagan@parallels.com>
CC: Kevin Wolf <kwolf@redhat.com>
CC: Stefan Hajnoczi <stefanha@redhat.com>
---
block/parallels.c | 35 ++++++++++++++++++++++++++++++++---
1 file changed, 32 insertions(+), 3 deletions(-)
diff --git a/block/parallels.c b/block/parallels.c
index 989efe4..c6343c5 100644
--- a/block/parallels.c
+++ b/block/parallels.c
@@ -84,6 +84,7 @@ typedef struct BDRVParallelsState {
uint32_t *bat_bitmap;
unsigned int bat_size;
+ int64_t data_end;
uint64_t prealloc_size;
ParallelsPreallocMode prealloc_mode;
@@ -194,7 +195,20 @@ static int64_t allocate_cluster(BlockDriverState *bs, int64_t sector_num)
}
pos = bdrv_getlength(bs->file) >> BDRV_SECTOR_BITS;
- bdrv_truncate(bs->file, (pos + s->tracks) << BDRV_SECTOR_BITS);
+ if (s->data_end + s->tracks > pos) {
+ int ret;
+ if (s->prealloc_mode == PRL_PREALLOC_MODE_FALLOCATE)
+ ret = bdrv_write_zeroes(bs->file, s->data_end,
+ s->prealloc_size, 0);
+ else
+ ret = bdrv_truncate(bs->file,
+ (s->data_end + s->prealloc_size) << BDRV_SECTOR_BITS);
+ if (ret < 0) {
+ return ret;
+ }
+ }
+ pos = s->data_end;
+ s->data_end += s->tracks;
s->bat_bitmap[idx] = cpu_to_le32(pos / s->off_multiplier);
@@ -533,7 +547,7 @@ static int parallels_open(BlockDriverState *bs, QDict *options, int flags,
{
BDRVParallelsState *s = bs->opaque;
ParallelsHeader ph;
- int ret, size;
+ int ret, size, i;
QemuOpts *opts = NULL;
Error *local_err = NULL;
char *buf;
@@ -583,7 +597,11 @@ static int parallels_open(BlockDriverState *bs, QDict *options, int flags,
ret = -ENOMEM;
goto fail;
}
- if (le32_to_cpu(ph.data_off) < s->header_size) {
+ s->data_end = le32_to_cpu(ph.data_off);
+ if (s->data_end == 0) {
+ s->data_end = ROUND_UP(bat_entry_off(s->bat_size), BDRV_SECTOR_SIZE);
+ }
+ if (s->data_end < s->header_size) {
/* there is not enough unused space to fit to block align between BAT
and actual data. We can't avoid read-modify-write... */
s->header_size = size;
@@ -595,6 +613,13 @@ static int parallels_open(BlockDriverState *bs, QDict *options, int flags,
}
s->bat_bitmap = (uint32_t *)(s->header + 1);
+ for (i = 0; i < s->bat_size; i++) {
+ int64_t off = bat2sect(s, i);
+ if (off >= s->data_end) {
+ s->data_end = off + s->tracks;
+ }
+ }
+
if (le32_to_cpu(ph.inuse) == HEADER_INUSE_MAGIC) {
/* Image was not closed correctly. The check is mandatory */
s->header_unclean = true;
@@ -665,6 +690,10 @@ static void parallels_close(BlockDriverState *bs)
parallels_update_header(bs);
}
+ if (bs->open_flags & BDRV_O_RDWR) {
+ bdrv_truncate(bs->file, s->data_end << BDRV_SECTOR_BITS);
+ }
+
g_free(s->bat_dirty_bmap);
qemu_vfree(s->header);
}
--
1.9.1
next prev parent reply other threads:[~2015-03-10 8:51 UTC|newest]
Thread overview: 53+ messages / expand[flat|nested] mbox.gz Atom feed top
2015-03-10 8:50 [Qemu-devel] [PATCH v3 0/27] write/create for Parallels images with reasonable performance Denis V. Lunev
2015-03-10 8:50 ` [Qemu-devel] [PATCH 01/27] iotests, parallels: quote TEST_IMG in 076 test to be path-safe Denis V. Lunev
2015-03-10 8:50 ` [Qemu-devel] [PATCH 02/27] block/parallels: rename parallels_header to ParallelsHeader Denis V. Lunev
2015-03-10 8:50 ` [Qemu-devel] [PATCH 03/27] block/parallels: switch to bdrv_read Denis V. Lunev
2015-03-10 8:50 ` [Qemu-devel] [PATCH 04/27] block/parallels: read up to cluster end in one go Denis V. Lunev
2015-03-10 8:50 ` [Qemu-devel] [PATCH 05/27] block/parallels: add get_block_status Denis V. Lunev
2015-03-10 8:51 ` [Qemu-devel] [PATCH 06/27] block/parallels: provide _co_readv routine for parallels format driver Denis V. Lunev
2015-03-10 8:51 ` [Qemu-devel] [PATCH 07/27] block/parallels: replace magic constants 4, 64 with proper sizeofs Denis V. Lunev
2015-03-10 8:51 ` [Qemu-devel] [PATCH 08/27] block/parallels: _co_writev callback for Parallels format Denis V. Lunev
2015-03-10 8:51 ` [Qemu-devel] [PATCH 09/27] iotests, parallels: test for write into Parallels image Denis V. Lunev
2015-03-10 8:51 ` [Qemu-devel] [PATCH 10/27] block/parallels: support parallels image creation Denis V. Lunev
2015-03-10 8:51 ` [Qemu-devel] [PATCH 11/27] iotests, parallels: test for newly created parallels image via qemu-img Denis V. Lunev
2015-03-10 8:51 ` [Qemu-devel] [PATCH 12/27] parallels: change copyright information in the image header Denis V. Lunev
2015-03-10 8:51 ` [Qemu-devel] [PATCH 13/27] block/parallels: rename catalog_ names to bat_ Denis V. Lunev
2015-03-10 8:51 ` [Qemu-devel] [PATCH 14/27] block/parallels: create bat2sect helper Denis V. Lunev
2015-03-10 12:14 ` Roman Kagan
2015-03-10 8:51 ` [Qemu-devel] [PATCH 15/27] block/parallels: keep BAT bitmap data in little endian in memory Denis V. Lunev
2015-03-10 12:16 ` Roman Kagan
2015-03-10 8:51 ` [Qemu-devel] [PATCH 16/27] block/parallels: read parallels image header and BAT into single buffer Denis V. Lunev
2015-03-10 14:07 ` Roman Kagan
2015-03-10 8:51 ` [Qemu-devel] [PATCH 17/27] block/parallels: move parallels_open/probe to the very end of the file Denis V. Lunev
2015-03-10 14:09 ` Roman Kagan
2015-03-10 8:51 ` [Qemu-devel] [PATCH 18/27] block/parallels: implement parallels_check method of block driver Denis V. Lunev
2015-03-10 14:24 ` Roman Kagan
2015-03-10 14:29 ` Denis V. Lunev
2015-03-10 8:51 ` [Qemu-devel] [PATCH 19/27] block/parallels: implement incorrect close detection Denis V. Lunev
2015-03-10 14:38 ` Roman Kagan
2015-03-10 14:44 ` Denis V. Lunev
2015-03-10 14:51 ` Roman Kagan
2015-03-10 8:51 ` [Qemu-devel] [PATCH 20/27] iotests, parallels: check for incorrectly closed image in tests Denis V. Lunev
2015-03-10 14:39 ` Roman Kagan
2015-03-10 8:51 ` [Qemu-devel] [PATCH 21/27] block/parallels: no need to flush on each block allocation table update Denis V. Lunev
2015-03-10 14:41 ` Roman Kagan
2015-03-10 8:51 ` [Qemu-devel] [PATCH 22/27] block/parallels: improve image reading performance Denis V. Lunev
2015-03-10 14:50 ` Roman Kagan
2015-03-10 8:51 ` [Qemu-devel] [PATCH 23/27] block/parallels: create bat_entry_off helper Denis V. Lunev
2015-03-10 14:52 ` Roman Kagan
2015-03-10 8:51 ` [Qemu-devel] [PATCH 24/27] block/parallels: delay writing to BAT till bdrv_co_flush_to_os Denis V. Lunev
2015-03-10 15:04 ` Roman Kagan
2015-03-10 8:51 ` [Qemu-devel] [PATCH 25/27] block/parallels: add prealloc-mode and prealloc-size open paramemets Denis V. Lunev
2015-03-10 15:05 ` Roman Kagan
2015-03-10 8:51 ` Denis V. Lunev [this message]
2015-03-10 15:10 ` [Qemu-devel] [PATCH 26/27] block/parallels: optimize linear image expansion Roman Karan
2015-03-10 8:51 ` [Qemu-devel] [PATCH 27/27] block/parallels: improve image writing performance further Denis V. Lunev
2015-03-11 9:25 ` Roman Kagan
-- strict thread matches above, loose matches on Subject: below --
2015-03-11 10:27 [Qemu-devel] [PATCH v3 0/27] write/create for Parallels images with reasonable performance Denis V. Lunev
2015-03-11 10:28 ` [Qemu-devel] [PATCH 26/27] block/parallels: optimize linear image expansion Denis V. Lunev
2015-04-22 14:18 ` Stefan Hajnoczi
2015-04-22 14:25 ` Denis V. Lunev
2015-04-22 15:41 ` Denis V. Lunev
2015-04-23 9:26 ` Stefan Hajnoczi
2015-04-28 7:46 [Qemu-devel] [PATCH v4 0/27] write/create for Parallels images with reasonable performance Denis V. Lunev
2015-04-28 7:46 ` [Qemu-devel] [PATCH 26/27] block/parallels: optimize linear image expansion Denis V. Lunev
2015-04-28 11:00 ` Roman Kagan
2015-05-18 16:32 ` Stefan Hajnoczi
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=1425977481-13317-27-git-send-email-den@openvz.org \
--to=den@openvz.org \
--cc=kwolf@redhat.com \
--cc=qemu-devel@nongnu.org \
--cc=rkagan@parallels.com \
--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).