From: Paolo Bonzini <pbonzini@redhat.com>
To: qemu-devel@nongnu.org
Cc: kwolf@redhat.com, jcody@redhat.com
Subject: [Qemu-devel] [PATCH v2 36/45] block: implement dirty bitmap using HBitmap
Date: Wed, 26 Sep 2012 17:56:42 +0200 [thread overview]
Message-ID: <1348675011-8794-37-git-send-email-pbonzini@redhat.com> (raw)
In-Reply-To: <1348675011-8794-1-git-send-email-pbonzini@redhat.com>
This actually uses the dirty bitmap in the block layer, and converts
mirroring to use an HBitmapIter.
Reviewed-by: Laszlo Ersek <lersek@redhat.com> (except block/mirror.c parts)
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
---
Makefile.objs | 2 +-
block.c | 94 ++++++++++------------------------------------------------
block.h | 6 ++--
block/mirror.c | 12 ++++++--
block_int.h | 4 +--
trace-events | 1 +
6 file modificati, 33 inserzioni(+), 86 rimozioni(-)
diff --git a/Makefile.objs b/Makefile.objs
index 9401516..b7bf9f1 100644
--- a/Makefile.objs
+++ b/Makefile.objs
@@ -43,7 +43,7 @@ coroutine-obj-$(CONFIG_WIN32) += coroutine-win32.o
block-obj-y = cutils.o iov.o cache-utils.o qemu-option.o module.o async.o
block-obj-y += nbd.o block.o blockjob.o aio.o aes.o qemu-config.o
-block-obj-y += qemu-progress.o qemu-sockets.o
+block-obj-y += qemu-progress.o qemu-sockets.o hbitmap.o
block-obj-y += $(coroutine-obj-y) $(qobject-obj-y) $(version-obj-y)
block-obj-$(CONFIG_POSIX) += posix-aio-compat.o
block-obj-$(CONFIG_LINUX_AIO) += linux-aio.o
diff --git a/block.c b/block.c
index 0f57ec1..af4237b 100644
--- a/block.c
+++ b/block.c
@@ -1267,7 +1267,6 @@ static void bdrv_move_feature_fields(BlockDriverState *bs_dest,
bs_dest->iostatus = bs_src->iostatus;
/* dirty bitmap */
- bs_dest->dirty_count = bs_src->dirty_count;
bs_dest->dirty_bitmap = bs_src->dirty_bitmap;
/* job */
@@ -1872,36 +1871,6 @@ int bdrv_read_unthrottled(BlockDriverState *bs, int64_t sector_num,
return ret;
}
-#define BITS_PER_LONG (sizeof(unsigned long) * 8)
-
-static void set_dirty_bitmap(BlockDriverState *bs, int64_t sector_num,
- int nb_sectors, int dirty)
-{
- int64_t start, end;
- unsigned long val, idx, bit;
-
- start = sector_num / BDRV_SECTORS_PER_DIRTY_CHUNK;
- end = (sector_num + nb_sectors - 1) / BDRV_SECTORS_PER_DIRTY_CHUNK;
-
- for (; start <= end; start++) {
- idx = start / BITS_PER_LONG;
- bit = start % BITS_PER_LONG;
- val = bs->dirty_bitmap[idx];
- if (dirty) {
- if (!(val & (1UL << bit))) {
- bs->dirty_count++;
- val |= 1UL << bit;
- }
- } else {
- if (val & (1UL << bit)) {
- bs->dirty_count--;
- val &= ~(1UL << bit);
- }
- }
- bs->dirty_bitmap[idx] = val;
- }
-}
-
/* Return < 0 if error. Important errors are:
-EIO generic I/O error (may happen for all errors)
-ENOMEDIUM No media inserted.
@@ -4044,18 +4013,15 @@ void bdrv_set_dirty_tracking(BlockDriverState *bs, int enable)
{
int64_t bitmap_size;
- bs->dirty_count = 0;
if (enable) {
if (!bs->dirty_bitmap) {
- bitmap_size = (bdrv_getlength(bs) >> BDRV_SECTOR_BITS) +
- BDRV_SECTORS_PER_DIRTY_CHUNK * BITS_PER_LONG - 1;
- bitmap_size /= BDRV_SECTORS_PER_DIRTY_CHUNK * BITS_PER_LONG;
-
- bs->dirty_bitmap = g_new0(unsigned long, bitmap_size);
+ bitmap_size = (bdrv_getlength(bs) >> BDRV_SECTOR_BITS);
+ bs->dirty_bitmap = hbitmap_alloc(bitmap_size,
+ BDRV_LOG_SECTORS_PER_DIRTY_CHUNK);
}
} else {
if (bs->dirty_bitmap) {
- g_free(bs->dirty_bitmap);
+ hbitmap_free(bs->dirty_bitmap);
bs->dirty_bitmap = NULL;
}
}
@@ -4063,67 +4029,37 @@ void bdrv_set_dirty_tracking(BlockDriverState *bs, int enable)
int bdrv_get_dirty(BlockDriverState *bs, int64_t sector)
{
- int64_t chunk = sector / (int64_t)BDRV_SECTORS_PER_DIRTY_CHUNK;
-
- if (bs->dirty_bitmap &&
- (sector << BDRV_SECTOR_BITS) < bdrv_getlength(bs)) {
- return !!(bs->dirty_bitmap[chunk / BITS_PER_LONG] &
- (1UL << (chunk % BITS_PER_LONG)));
+ if (bs->dirty_bitmap) {
+ return hbitmap_get(bs->dirty_bitmap, sector);
} else {
return 0;
}
}
-int64_t bdrv_get_next_dirty(BlockDriverState *bs, int64_t sector)
+void bdrv_dirty_iter_init(BlockDriverState *bs, HBitmapIter *hbi)
{
- int64_t chunk;
- int bit, elem;
-
- /* Avoid an infinite loop. */
- assert(bs->dirty_count > 0);
-
- sector = (sector | (BDRV_SECTORS_PER_DIRTY_CHUNK - 1)) + 1;
- chunk = sector / (int64_t)BDRV_SECTORS_PER_DIRTY_CHUNK;
-
- QEMU_BUILD_BUG_ON(sizeof(bs->dirty_bitmap[0]) * 8 != BITS_PER_LONG);
- elem = chunk / BITS_PER_LONG;
- bit = chunk % BITS_PER_LONG;
- for (;;) {
- if (sector >= bs->total_sectors) {
- sector = 0;
- bit = elem = 0;
- }
- if (bit == 0 && bs->dirty_bitmap[elem] == 0) {
- sector += BDRV_SECTORS_PER_DIRTY_CHUNK * BITS_PER_LONG;
- elem++;
- } else {
- if (bs->dirty_bitmap[elem] & (1UL << bit)) {
- return sector;
- }
- sector += BDRV_SECTORS_PER_DIRTY_CHUNK;
- if (++bit == BITS_PER_LONG) {
- bit = 0;
- elem++;
- }
- }
- }
+ hbitmap_iter_init(hbi, bs->dirty_bitmap, 0);
}
void bdrv_set_dirty(BlockDriverState *bs, int64_t cur_sector,
int nr_sectors)
{
- set_dirty_bitmap(bs, cur_sector, nr_sectors, 1);
+ hbitmap_set(bs->dirty_bitmap, cur_sector, nr_sectors);
}
void bdrv_reset_dirty(BlockDriverState *bs, int64_t cur_sector,
int nr_sectors)
{
- set_dirty_bitmap(bs, cur_sector, nr_sectors, 0);
+ hbitmap_reset(bs->dirty_bitmap, cur_sector, nr_sectors);
}
int64_t bdrv_get_dirty_count(BlockDriverState *bs)
{
- return bs->dirty_count;
+ if (bs->dirty_bitmap) {
+ return hbitmap_count(bs->dirty_bitmap) >> BDRV_LOG_SECTORS_PER_DIRTY_CHUNK;
+ } else {
+ return 0;
+ }
}
void bdrv_set_in_use(BlockDriverState *bs, int in_use)
diff --git a/block.h b/block.h
index 9097edd..676135d 100644
--- a/block.h
+++ b/block.h
@@ -344,13 +344,15 @@ int bdrv_img_create(const char *filename, const char *fmt,
void bdrv_set_buffer_alignment(BlockDriverState *bs, int align);
void *qemu_blockalign(BlockDriverState *bs, size_t size);
-#define BDRV_SECTORS_PER_DIRTY_CHUNK 2048
+#define BDRV_SECTORS_PER_DIRTY_CHUNK (1 << BDRV_LOG_SECTORS_PER_DIRTY_CHUNK)
+#define BDRV_LOG_SECTORS_PER_DIRTY_CHUNK 11
+struct HBitmapIter;
void bdrv_set_dirty_tracking(BlockDriverState *bs, int enable);
int bdrv_get_dirty(BlockDriverState *bs, int64_t sector);
void bdrv_set_dirty(BlockDriverState *bs, int64_t cur_sector, int nr_sectors);
void bdrv_reset_dirty(BlockDriverState *bs, int64_t cur_sector, int nr_sectors);
-int64_t bdrv_get_next_dirty(BlockDriverState *bs, int64_t sector);
+void bdrv_dirty_iter_init(BlockDriverState *bs, struct HBitmapIter *hbi);
int64_t bdrv_get_dirty_count(BlockDriverState *bs);
void bdrv_enable_copy_on_read(BlockDriverState *bs);
diff --git a/block/mirror.c b/block/mirror.c
index caec272..72e0986 100644
--- a/block/mirror.c
+++ b/block/mirror.c
@@ -36,6 +36,7 @@ typedef struct MirrorBlockJob {
bool synced;
bool complete;
int64_t sector_num;
+ HBitmapIter hbi;
uint8_t *buf;
} MirrorBlockJob;
@@ -62,8 +63,15 @@ static int coroutine_fn mirror_iteration(MirrorBlockJob *s,
int64_t end;
struct iovec iov;
+ s->sector_num = hbitmap_iter_next(&s->hbi);
+ if (s->sector_num < 0) {
+ bdrv_dirty_iter_init(source, &s->hbi);
+ s->sector_num = hbitmap_iter_next(&s->hbi);
+ trace_mirror_restart_iter(s, bdrv_get_dirty_count(source));
+ assert(s->sector_num >= 0);
+ }
+
end = s->common.len >> BDRV_SECTOR_BITS;
- s->sector_num = bdrv_get_next_dirty(source, s->sector_num);
nb_sectors = MIN(BDRV_SECTORS_PER_DIRTY_CHUNK, end - s->sector_num);
bdrv_reset_dirty(source, s->sector_num, nb_sectors);
@@ -136,7 +144,7 @@ static void coroutine_fn mirror_run(void *opaque)
}
}
- s->sector_num = -1;
+ bdrv_dirty_iter_init(bs, &s->hbi);
for (;;) {
uint64_t delay_ns;
int64_t cnt;
diff --git a/block_int.h b/block_int.h
index a533c7b..c09a679 100644
--- a/block_int.h
+++ b/block_int.h
@@ -32,6 +32,7 @@
#include "qapi-types.h"
#include "qerror.h"
#include "monitor.h"
+#include "hbitmap.h"
#define BLOCK_FLAG_ENCRYPT 1
#define BLOCK_FLAG_COMPAT6 4
@@ -267,8 +268,7 @@ struct BlockDriverState {
bool iostatus_enabled;
BlockDeviceIoStatus iostatus;
char device_name[32];
- unsigned long *dirty_bitmap;
- int64_t dirty_count;
+ HBitmap *dirty_bitmap;
int in_use; /* users other than guest access, eg. block migration */
QTAILQ_ENTRY(BlockDriverState) list;
diff --git a/trace-events b/trace-events
index e2aa164..99818d5 100644
--- a/trace-events
+++ b/trace-events
@@ -77,6 +77,7 @@ stream_start(void *bs, void *base, void *s, void *co, void *opaque) "bs %p base
# block/mirror.c
mirror_start(void *bs, void *s, void *co, void *opaque) "bs %p s %p co %p opaque %p"
+mirror_restart_iter(void *s, int64_t cnt) "s %p dirty count %"PRId64
mirror_before_flush(void *s) "s %p"
mirror_before_drain(void *s, int64_t cnt) "s %p dirty count %"PRId64
mirror_before_sleep(void *s, int64_t cnt, int synced) "s %p dirty count %"PRId64" synced %d"
--
1.7.12
next prev parent reply other threads:[~2012-09-26 15:59 UTC|newest]
Thread overview: 102+ messages / expand[flat|nested] mbox.gz Atom feed top
2012-09-26 15:56 [Qemu-devel] [PATCH v2 00/45] Block job improvements for 1.3 Paolo Bonzini
2012-09-26 15:56 ` [Qemu-devel] [PATCH v2 01/45] qerror/block: introduce QERR_BLOCK_JOB_NOT_ACTIVE Paolo Bonzini
2012-09-26 15:56 ` [Qemu-devel] [PATCH v2 02/45] blockdev: rename block_stream_cb to a generic block_job_cb Paolo Bonzini
2012-09-27 11:56 ` Kevin Wolf
2012-09-26 15:56 ` [Qemu-devel] [PATCH v2 03/45] block: fix documentation of block_job_cancel_sync Paolo Bonzini
2012-09-27 12:03 ` Kevin Wolf
2012-09-27 12:08 ` Paolo Bonzini
2012-09-27 12:13 ` Kevin Wolf
2012-09-26 15:56 ` [Qemu-devel] [PATCH v2 04/45] block: move job APIs to separate files Paolo Bonzini
2012-09-26 15:56 ` [Qemu-devel] [PATCH v2 05/45] block: add block_job_query Paolo Bonzini
2012-09-26 15:56 ` [Qemu-devel] [PATCH v2 06/45] block: add support for job pause/resume Paolo Bonzini
2012-09-26 17:31 ` Eric Blake
2012-09-27 12:18 ` Kevin Wolf
2012-09-27 12:27 ` Paolo Bonzini
2012-09-27 12:45 ` Kevin Wolf
2012-09-27 12:57 ` Paolo Bonzini
2012-09-27 13:51 ` Kevin Wolf
2012-09-26 15:56 ` [Qemu-devel] [PATCH v2 07/45] qmp: add block-job-pause and block-job-resume Paolo Bonzini
2012-09-26 17:45 ` Eric Blake
2012-09-27 9:23 ` Paolo Bonzini
2012-09-26 15:56 ` [Qemu-devel] [PATCH v2 08/45] qemu-iotests: add test for pausing a streaming operation Paolo Bonzini
2012-09-26 15:56 ` [Qemu-devel] [PATCH v2 09/45] block: rename block_job_complete to block_job_completed Paolo Bonzini
2012-09-27 12:30 ` Kevin Wolf
2012-09-27 20:31 ` Jeff Cody
2012-09-28 11:00 ` Paolo Bonzini
2012-09-26 15:56 ` [Qemu-devel] [PATCH v2 10/45] iostatus: rename BlockErrorAction, BlockQMPEventAction Paolo Bonzini
2012-09-26 15:56 ` [Qemu-devel] [PATCH v2 11/45] iostatus: move BlockdevOnError declaration to QAPI Paolo Bonzini
2012-09-26 17:54 ` Eric Blake
2012-09-27 9:23 ` Paolo Bonzini
2012-09-26 15:56 ` [Qemu-devel] [PATCH v2 12/45] iostatus: change is_read to a bool Paolo Bonzini
2012-09-26 15:56 ` [Qemu-devel] [PATCH v2 13/45] iostatus: reorganize io error code Paolo Bonzini
2012-09-26 15:56 ` [Qemu-devel] [PATCH v2 14/45] block: introduce block job error Paolo Bonzini
2012-09-26 19:10 ` Eric Blake
2012-09-26 19:27 ` Eric Blake
2012-09-27 9:24 ` Paolo Bonzini
2012-09-27 13:41 ` Kevin Wolf
2012-09-27 14:50 ` Paolo Bonzini
2012-09-26 15:56 ` [Qemu-devel] [PATCH v2 15/45] stream: add on-error argument Paolo Bonzini
2012-09-26 20:53 ` Eric Blake
2012-09-26 15:56 ` [Qemu-devel] [PATCH v2 16/45] blkdebug: process all set_state rules in the old state Paolo Bonzini
2012-09-26 15:56 ` [Qemu-devel] [PATCH v2 17/45] qemu-iotests: map underscore to dash in QMP argument names Paolo Bonzini
2012-09-26 15:56 ` [Qemu-devel] [PATCH v2 18/45] qemu-iotests: add tests for streaming error handling Paolo Bonzini
2012-09-26 15:56 ` [Qemu-devel] [PATCH v2 19/45] block: add bdrv_query_info Paolo Bonzini
2012-10-15 15:42 ` Kevin Wolf
2012-09-26 15:56 ` [Qemu-devel] [PATCH v2 20/45] block: add bdrv_query_stats Paolo Bonzini
2012-09-26 15:56 ` [Qemu-devel] [PATCH v2 21/45] block: add bdrv_open_backing_file Paolo Bonzini
2012-09-27 18:14 ` Jeff Cody
2012-09-26 15:56 ` [Qemu-devel] [PATCH v2 22/45] block: introduce new dirty bitmap functionality Paolo Bonzini
2012-09-26 15:56 ` [Qemu-devel] [PATCH v2 23/45] block: export dirty bitmap information in query-block Paolo Bonzini
2012-10-15 16:08 ` Kevin Wolf
2012-09-26 15:56 ` [Qemu-devel] [PATCH v2 24/45] block: add block-job-complete Paolo Bonzini
2012-09-26 15:56 ` [Qemu-devel] [PATCH v2 25/45] block: introduce BLOCK_JOB_READY event Paolo Bonzini
2012-09-27 0:01 ` Eric Blake
2012-09-27 9:25 ` Paolo Bonzini
2012-09-26 15:56 ` [Qemu-devel] [PATCH v2 26/45] mirror: introduce mirror job Paolo Bonzini
2012-10-15 16:57 ` Kevin Wolf
2012-10-16 6:36 ` Paolo Bonzini
2012-10-16 8:24 ` Kevin Wolf
2012-10-16 8:35 ` Paolo Bonzini
2012-09-26 15:56 ` [Qemu-devel] [PATCH v2 27/45] qmp: add drive-mirror command Paolo Bonzini
2012-09-27 0:14 ` Eric Blake
2012-09-27 19:49 ` Jeff Cody
2012-10-15 17:33 ` Kevin Wolf
2012-10-16 6:39 ` Paolo Bonzini
2012-10-18 13:13 ` Paolo Bonzini
2012-09-26 15:56 ` [Qemu-devel] [PATCH v2 28/45] mirror: implement completion Paolo Bonzini
2012-10-15 17:49 ` Kevin Wolf
2012-09-26 15:56 ` [Qemu-devel] [PATCH v2 29/45] qemu-iotests: add mirroring test case Paolo Bonzini
2012-09-27 0:26 ` Eric Blake
2012-10-18 12:43 ` Kevin Wolf
2012-10-18 12:50 ` Paolo Bonzini
2012-10-18 13:08 ` Kevin Wolf
2012-09-26 15:56 ` [Qemu-devel] [PATCH v2 30/45] iostatus: forward block_job_iostatus_reset to block job Paolo Bonzini
2012-09-26 15:56 ` [Qemu-devel] [PATCH v2 31/45] mirror: add support for on-source-error/on-target-error Paolo Bonzini
2012-10-18 13:07 ` Kevin Wolf
2012-10-18 13:10 ` Paolo Bonzini
2012-10-18 13:56 ` Kevin Wolf
2012-10-18 14:52 ` Paolo Bonzini
2012-10-19 8:04 ` Kevin Wolf
2012-10-19 9:30 ` Paolo Bonzini
2012-09-26 15:56 ` [Qemu-devel] [PATCH v2 32/45] qmp: add pull_event function Paolo Bonzini
2012-09-26 17:17 ` Luiz Capitulino
2012-09-26 15:56 ` [Qemu-devel] [PATCH v2 33/45] qemu-iotests: add testcases for mirroring on-source-error/on-target-error Paolo Bonzini
2012-09-26 15:56 ` [Qemu-devel] [PATCH v2 34/45] host-utils: add ffsl Paolo Bonzini
2012-09-27 1:14 ` Eric Blake
2012-09-26 15:56 ` [Qemu-devel] [PATCH v2 35/45] add hierarchical bitmap data type and test cases Paolo Bonzini
2012-09-27 2:53 ` Eric Blake
2012-09-27 9:27 ` Paolo Bonzini
2012-10-24 14:41 ` Kevin Wolf
2012-10-24 14:50 ` Paolo Bonzini
2012-09-26 15:56 ` Paolo Bonzini [this message]
2012-09-26 15:56 ` [Qemu-devel] [PATCH v2 37/45] block: make round_to_clusters public Paolo Bonzini
2012-09-26 15:56 ` [Qemu-devel] [PATCH v2 38/45] mirror: perform COW if the cluster size is bigger than the granularity Paolo Bonzini
2012-09-26 15:56 ` [Qemu-devel] [PATCH v2 39/45] block: return count of dirty sectors, not chunks Paolo Bonzini
2012-09-26 15:56 ` [Qemu-devel] [PATCH v2 40/45] block: allow customizing the granularity of the dirty bitmap Paolo Bonzini
2012-09-26 15:56 ` [Qemu-devel] [PATCH v2 41/45] mirror: allow customizing the granularity Paolo Bonzini
2012-09-26 15:56 ` [Qemu-devel] [PATCH v2 42/45] mirror: switch mirror_iteration to AIO Paolo Bonzini
2012-09-26 15:56 ` [Qemu-devel] [PATCH v2 43/45] mirror: add buf-size argument to drive-mirror Paolo Bonzini
2012-09-26 15:56 ` [Qemu-devel] [PATCH v2 44/45] mirror: support more than one in-flight AIO operation Paolo Bonzini
2012-09-26 15:56 ` [Qemu-devel] [PATCH v2 45/45] mirror: support arbitrarily-sized iterations Paolo Bonzini
2012-09-27 14:05 ` [Qemu-devel] [PATCH v2 00/45] Block job improvements for 1.3 Kevin Wolf
2012-09-27 14:57 ` Paolo Bonzini
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=1348675011-8794-37-git-send-email-pbonzini@redhat.com \
--to=pbonzini@redhat.com \
--cc=jcody@redhat.com \
--cc=kwolf@redhat.com \
--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).