mirror: support arbitrarily-sized iterations

Yet another optimization is to extend the mirroring iteration to include more
adjacent dirty blocks.  This limits the number of I/O operations and makes
mirroring efficient even with a small granularity.  Most of the infrastructure
is already in place; we only need to put a loop around the computation of
the origin and sector count of the iteration.

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
This commit is contained in:
Paolo Bonzini 2013-01-22 09:03:15 +01:00 committed by Kevin Wolf
parent 402a47411b
commit 884fea4e87
2 changed files with 69 additions and 31 deletions

View File

@ -140,7 +140,7 @@ static void coroutine_fn mirror_iteration(MirrorBlockJob *s)
{ {
BlockDriverState *source = s->common.bs; BlockDriverState *source = s->common.bs;
int nb_sectors, sectors_per_chunk, nb_chunks; int nb_sectors, sectors_per_chunk, nb_chunks;
int64_t end, sector_num, chunk_num, next_sector, hbitmap_next_sector; int64_t end, sector_num, next_chunk, next_sector, hbitmap_next_sector;
MirrorOp *op; MirrorOp *op;
s->sector_num = hbitmap_iter_next(&s->hbi); s->sector_num = hbitmap_iter_next(&s->hbi);
@ -152,45 +152,82 @@ static void coroutine_fn mirror_iteration(MirrorBlockJob *s)
} }
hbitmap_next_sector = s->sector_num; hbitmap_next_sector = s->sector_num;
/* 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 the granularity and the cluster size are powers of two, the
* number of sectors to copy cannot exceed one cluster.
*/
sector_num = s->sector_num; sector_num = s->sector_num;
sectors_per_chunk = nb_sectors = s->granularity >> BDRV_SECTOR_BITS; sectors_per_chunk = s->granularity >> BDRV_SECTOR_BITS;
chunk_num = sector_num / sectors_per_chunk; end = s->common.len >> BDRV_SECTOR_BITS;
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, sectors_per_chunk,
&sector_num, &nb_sectors);
/* The rounding may make us copy sectors before the /* Extend the QEMUIOVector to include all adjacent blocks that will
* first dirty one. * be copied in this operation.
*
* We have to do this if we have no backing file yet in the destination,
* and the cluster size is very large. Then we need to do COW ourselves.
* The first time a cluster is copied, copy it entirely. Note that,
* because both the granularity and the cluster size are powers of two,
* the number of sectors to copy cannot exceed one cluster.
*
* We also want to extend the QEMUIOVector to include more adjacent
* dirty blocks if possible, to limit the number of I/O operations and
* run efficiently even with a small granularity.
*/ */
chunk_num = sector_num / sectors_per_chunk; nb_chunks = 0;
} nb_sectors = 0;
next_sector = sector_num;
next_chunk = sector_num / sectors_per_chunk;
/* Wait for I/O to this cluster (from a previous iteration) to be done. */ /* Wait for I/O to this cluster (from a previous iteration) to be done. */
while (test_bit(chunk_num, s->in_flight_bitmap)) { while (test_bit(next_chunk, s->in_flight_bitmap)) {
trace_mirror_yield_in_flight(s, sector_num, s->in_flight); trace_mirror_yield_in_flight(s, sector_num, s->in_flight);
qemu_coroutine_yield(); qemu_coroutine_yield();
} }
end = s->common.len >> BDRV_SECTOR_BITS; do {
nb_sectors = MIN(nb_sectors, end - sector_num); int added_sectors, added_chunks;
nb_chunks = (nb_sectors + sectors_per_chunk - 1) / sectors_per_chunk;
while (s->buf_free_count < nb_chunks) { if (!bdrv_get_dirty(source, next_sector) ||
test_bit(next_chunk, s->in_flight_bitmap)) {
assert(nb_sectors > 0);
break;
}
added_sectors = sectors_per_chunk;
if (s->cow_bitmap && !test_bit(next_chunk, s->cow_bitmap)) {
bdrv_round_to_clusters(s->target,
next_sector, added_sectors,
&next_sector, &added_sectors);
/* On the first iteration, the rounding may make us copy
* sectors before the first dirty one.
*/
if (next_sector < sector_num) {
assert(nb_sectors == 0);
sector_num = next_sector;
next_chunk = next_sector / sectors_per_chunk;
}
}
added_sectors = MIN(added_sectors, end - (sector_num + nb_sectors));
added_chunks = (added_sectors + sectors_per_chunk - 1) / sectors_per_chunk;
/* When doing COW, it may happen that there is not enough space for
* a full cluster. Wait if that is the case.
*/
while (nb_chunks == 0 && s->buf_free_count < added_chunks) {
trace_mirror_yield_buf_busy(s, nb_chunks, s->in_flight); trace_mirror_yield_buf_busy(s, nb_chunks, s->in_flight);
qemu_coroutine_yield(); qemu_coroutine_yield();
} }
if (s->buf_free_count < nb_chunks + added_chunks) {
trace_mirror_break_buf_busy(s, nb_chunks, s->in_flight);
break;
}
/* We have enough free space to copy these sectors. */ /* We have enough free space to copy these sectors. */
bitmap_set(s->in_flight_bitmap, chunk_num, nb_chunks); bitmap_set(s->in_flight_bitmap, next_chunk, added_chunks);
nb_sectors += added_sectors;
nb_chunks += added_chunks;
next_sector += added_sectors;
next_chunk += added_chunks;
} while (next_sector < end);
/* Allocate a MirrorOp that is used as an AIO callback. */ /* Allocate a MirrorOp that is used as an AIO callback. */
op = g_slice_new(MirrorOp); op = g_slice_new(MirrorOp);

View File

@ -89,6 +89,7 @@ mirror_iteration_done(void *s, int64_t sector_num, int nb_sectors, int ret) "s %
mirror_yield(void *s, int64_t cnt, int buf_free_count, int in_flight) "s %p dirty count %"PRId64" free buffers %d in_flight %d" mirror_yield(void *s, int64_t cnt, int buf_free_count, int in_flight) "s %p dirty count %"PRId64" free buffers %d in_flight %d"
mirror_yield_in_flight(void *s, int64_t sector_num, int in_flight) "s %p sector_num %"PRId64" in_flight %d" mirror_yield_in_flight(void *s, int64_t sector_num, int in_flight) "s %p sector_num %"PRId64" in_flight %d"
mirror_yield_buf_busy(void *s, int nb_chunks, int in_flight) "s %p requested chunks %d in_flight %d" mirror_yield_buf_busy(void *s, int nb_chunks, int in_flight) "s %p requested chunks %d in_flight %d"
mirror_break_buf_busy(void *s, int nb_chunks, int in_flight) "s %p requested chunks %d in_flight %d"
# blockdev.c # blockdev.c
qmp_block_job_cancel(void *job) "job %p" qmp_block_job_cancel(void *job) "job %p"