diff --git a/block/mirror.c b/block/mirror.c
index 20cb1e777f..307bcf101e 100644
--- a/block/mirror.c
+++ b/block/mirror.c
@@ -15,6 +15,7 @@
#include "block/blockjob.h"
#include "block/block_int.h"
#include "qemu/ratelimit.h"
+#include "qemu/bitmap.h"
enum {
/*
@@ -36,6 +37,8 @@ typedef struct MirrorBlockJob {
bool synced;
bool should_complete;
int64_t sector_num;
+ size_t buf_size;
+ unsigned long *cow_bitmap;
HBitmapIter hbi;
uint8_t *buf;
} MirrorBlockJob;
@@ -60,7 +63,7 @@ static int coroutine_fn mirror_iteration(MirrorBlockJob *s,
BlockDriverState *target = s->target;
QEMUIOVector qiov;
int ret, nb_sectors;
- int64_t end;
+ int64_t end, sector_num, chunk_num;
struct iovec iov;
s->sector_num = hbitmap_iter_next(&s->hbi);
@@ -71,32 +74,53 @@ static int coroutine_fn mirror_iteration(MirrorBlockJob *s,
assert(s->sector_num >= 0);
}
+ /* If we have no backing file yet in the destination, and the cluster size
+ * is very large, we need to do COW ourselves. The first time a cluster is
+ * copied, copy it entirely.
+ *
+ * Because both BDRV_SECTORS_PER_DIRTY_CHUNK and the cluster size are
+ * powers of two, the number of sectors to copy cannot exceed one cluster.
+ */
+ sector_num = s->sector_num;
+ nb_sectors = BDRV_SECTORS_PER_DIRTY_CHUNK;
+ chunk_num = sector_num / BDRV_SECTORS_PER_DIRTY_CHUNK;
+ if (s->cow_bitmap && !test_bit(chunk_num, s->cow_bitmap)) {
+ trace_mirror_cow(s, sector_num);
+ bdrv_round_to_clusters(s->target,
+ sector_num, BDRV_SECTORS_PER_DIRTY_CHUNK,
+ §or_num, &nb_sectors);
+ }
+
end = s->common.len >> BDRV_SECTOR_BITS;
- nb_sectors = MIN(BDRV_SECTORS_PER_DIRTY_CHUNK, end - s->sector_num);
- bdrv_reset_dirty(source, s->sector_num, nb_sectors);
+ nb_sectors = MIN(nb_sectors, end - sector_num);
+ bdrv_reset_dirty(source, sector_num, nb_sectors);
/* Copy the dirty cluster. */
iov.iov_base = s->buf;
iov.iov_len = nb_sectors * 512;
qemu_iovec_init_external(&qiov, &iov, 1);
- trace_mirror_one_iteration(s, s->sector_num, nb_sectors);
- ret = bdrv_co_readv(source, s->sector_num, nb_sectors, &qiov);
+ trace_mirror_one_iteration(s, sector_num, nb_sectors);
+ ret = bdrv_co_readv(source, sector_num, nb_sectors, &qiov);
if (ret < 0) {
*p_action = mirror_error_action(s, true, -ret);
goto fail;
}
- ret = bdrv_co_writev(target, s->sector_num, nb_sectors, &qiov);
+ ret = bdrv_co_writev(target, sector_num, nb_sectors, &qiov);
if (ret < 0) {
*p_action = mirror_error_action(s, false, -ret);
s->synced = false;
goto fail;
}
+ if (s->cow_bitmap) {
+ bitmap_set(s->cow_bitmap, sector_num / BDRV_SECTORS_PER_DIRTY_CHUNK,
+ nb_sectors / BDRV_SECTORS_PER_DIRTY_CHUNK);
+ }
return 0;
fail:
/* Try again later. */
- bdrv_set_dirty(source, s->sector_num, nb_sectors);
+ bdrv_set_dirty(source, sector_num, nb_sectors);
return ret;
}
@@ -104,7 +128,9 @@ static void coroutine_fn mirror_run(void *opaque)
{
MirrorBlockJob *s = opaque;
BlockDriverState *bs = s->common.bs;
- int64_t sector_num, end;
+ int64_t sector_num, end, length;
+ BlockDriverInfo bdi;
+ char backing_filename[1024];
int ret = 0;
int n;
@@ -118,8 +144,23 @@ static void coroutine_fn mirror_run(void *opaque)
return;
}
+ /* If we have no backing file yet in the destination, we cannot let
+ * the destination do COW. Instead, we copy sectors around the
+ * dirty data if needed. We need a bitmap to do that.
+ */
+ bdrv_get_backing_filename(s->target, backing_filename,
+ sizeof(backing_filename));
+ if (backing_filename[0] && !s->target->backing_hd) {
+ bdrv_get_info(s->target, &bdi);
+ if (s->buf_size < bdi.cluster_size) {
+ s->buf_size = bdi.cluster_size;
+ length = (bdrv_getlength(bs) + BLOCK_SIZE - 1) / BLOCK_SIZE;
+ s->cow_bitmap = bitmap_new(length);
+ }
+ }
+
end = s->common.len >> BDRV_SECTOR_BITS;
- s->buf = qemu_blockalign(bs, BLOCK_SIZE);
+ s->buf = qemu_blockalign(bs, s->buf_size);
if (s->mode != MIRROR_SYNC_MODE_NONE) {
/* First part, loop on the sectors and initialize the dirty bitmap. */
@@ -234,6 +275,7 @@ static void coroutine_fn mirror_run(void *opaque)
immediate_exit:
qemu_vfree(s->buf);
+ g_free(s->cow_bitmap);
bdrv_set_dirty_tracking(bs, false);
bdrv_iostatus_disable(s->target);
if (s->should_complete && ret == 0) {
@@ -320,6 +362,8 @@ void mirror_start(BlockDriverState *bs, BlockDriverState *target,
s->on_target_error = on_target_error;
s->target = target;
s->mode = mode;
+ s->buf_size = BLOCK_SIZE;
+
bdrv_set_dirty_tracking(bs, true);
bdrv_set_enable_write_cache(s->target, true);
bdrv_set_on_error(s->target, on_target_error, on_target_error);
diff --git a/blockdev.c b/blockdev.c
index 9126587c45..1eb62b637c 100644
--- a/blockdev.c
+++ b/blockdev.c
@@ -1197,7 +1197,6 @@ void qmp_drive_mirror(const char *device, const char *target,
bool has_on_target_error, BlockdevOnError on_target_error,
Error **errp)
{
- BlockDriverInfo bdi;
BlockDriverState *bs;
BlockDriverState *source, *target_bs;
BlockDriver *proto_drv;
@@ -1288,6 +1287,9 @@ void qmp_drive_mirror(const char *device, const char *target,
return;
}
+ /* Mirroring takes care of copy-on-write using the source's backing
+ * file.
+ */
target_bs = bdrv_new("");
ret = bdrv_open(target_bs, target, flags | BDRV_O_NO_BACKING, drv);
@@ -1297,17 +1299,6 @@ void qmp_drive_mirror(const char *device, const char *target,
return;
}
- /* We need a backing file if we will copy parts of a cluster. */
- if (bdrv_get_info(target_bs, &bdi) >= 0 && bdi.cluster_size != 0 &&
- bdi.cluster_size >= BDRV_SECTORS_PER_DIRTY_CHUNK * 512) {
- ret = bdrv_open_backing_file(target_bs);
- if (ret < 0) {
- bdrv_delete(target_bs);
- error_set(errp, QERR_OPEN_FILE_FAILED, target);
- return;
- }
- }
-
mirror_start(bs, target_bs, speed, sync, on_source_error, on_target_error,
block_job_cb, bs, &local_err);
if (local_err != NULL) {
diff --git a/tests/qemu-iotests/041 b/tests/qemu-iotests/041
index c6eb851871..a1299b348e 100755
--- a/tests/qemu-iotests/041
+++ b/tests/qemu-iotests/041
@@ -292,6 +292,27 @@ class TestMirrorNoBacking(ImageMirroringTestCase):
self.assertTrue(self.compare_images(test_img, target_img),
'target image does not match source after mirroring')
+ def test_large_cluster(self):
+ self.assert_no_active_mirrors()
+
+ # qemu-img create fails if the image is not there
+ qemu_img('create', '-f', iotests.imgfmt, '-o', 'size=%d'
+ %(TestMirrorNoBacking.image_len), target_backing_img)
+ qemu_img('create', '-f', iotests.imgfmt, '-o', 'cluster_size=%d,backing_file=%s'
+ % (TestMirrorNoBacking.image_len, target_backing_img), target_img)
+ os.remove(target_backing_img)
+
+ result = self.vm.qmp('drive-mirror', device='drive0', sync='full',
+ mode='existing', target=target_img)
+ self.assert_qmp(result, 'return', {})
+
+ self.complete_and_wait()
+ result = self.vm.qmp('query-block')
+ self.assert_qmp(result, 'return[0]/inserted/file', target_img)
+ self.vm.shutdown()
+ self.assertTrue(self.compare_images(test_img, target_img),
+ 'target image does not match source after mirroring')
+
class TestReadErrors(ImageMirroringTestCase):
image_len = 2 * 1024 * 1024 # MB
@@ -330,6 +351,9 @@ new_state = "1"
'-o', 'backing_file=blkdebug:%s:%s,backing_fmt=raw'
% (self.blkdebug_file, backing_img),
test_img)
+ # Write something for tests that use sync='top'
+ qemu_io('-c', 'write %d 512' % (self.MIRROR_GRANULARITY + 65536),
+ test_img)
self.vm = iotests.VM().add_drive(test_img)
self.vm.launch()
@@ -383,6 +407,32 @@ new_state = "1"
self.complete_and_wait()
self.vm.shutdown()
+ def test_large_cluster(self):
+ self.assert_no_active_mirrors()
+
+ # Test COW into the target image. The first half of the
+ # cluster at MIRROR_GRANULARITY has to be copied from
+ # backing_img, even though sync='top'.
+ qemu_img('create', '-f', iotests.imgfmt, '-ocluster_size=131072,backing_file=%s' %(backing_img), target_img)
+ result = self.vm.qmp('drive-mirror', device='drive0', sync='top',
+ on_source_error='ignore',
+ mode='existing', target=target_img)
+ self.assert_qmp(result, 'return', {})
+
+ event = self.vm.get_qmp_event(wait=True)
+ self.assertEquals(event['event'], 'BLOCK_JOB_ERROR')
+ self.assert_qmp(event, 'data/device', 'drive0')
+ self.assert_qmp(event, 'data/operation', 'read')
+ result = self.vm.qmp('query-block-jobs')
+ self.assert_qmp(result, 'return[0]/paused', False)
+ self.complete_and_wait()
+ self.vm.shutdown()
+
+ # Detach blkdebug to compare images successfully
+ qemu_img('rebase', '-f', iotests.imgfmt, '-u', '-b', backing_img, test_img)
+ self.assertTrue(self.compare_images(test_img, target_img),
+ 'target image does not match source after mirroring')
+
def test_stop_read(self):
self.assert_no_active_mirrors()
diff --git a/tests/qemu-iotests/041.out b/tests/qemu-iotests/041.out
index 71009c239f..3a89159833 100644
--- a/tests/qemu-iotests/041.out
+++ b/tests/qemu-iotests/041.out
@@ -1,5 +1,5 @@
-..................
+....................
----------------------------------------------------------------------
-Ran 18 tests
+Ran 20 tests
OK
diff --git a/trace-events b/trace-events
index 61ed349765..ffa27568ee 100644
--- a/trace-events
+++ b/trace-events
@@ -84,6 +84,7 @@ 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"
mirror_one_iteration(void *s, int64_t sector_num, int nb_sectors) "s %p sector_num %"PRId64" nb_sectors %d"
+mirror_cow(void *s, int64_t sector_num) "s %p sector_num %"PRId64
# blockdev.c
qmp_block_job_cancel(void *job) "job %p"