If COW areas of the newly allocated clusters are zeroes on the backing image: (even if preallocation feature is not used or it cannot detect if the image already reads as zeroes, e.g. writing to a hole / preallocated zero cluster) efficient bdrv_write_zeroes(flags=BDRV_REQ_ALLOCATE) can be used on the whole cluster instead of writing explicit zero buffers later in perform_cow().
iotest 060: write to the discarded cluster does not trigger COW anymore. so, break on write_aio event instead, will work for the test (but write won't fail anymore, so update reference output) iotest 066: cluster-alignment areas that were not really COWed are now detected as zeroes, hence the initial write has to be exactly the same size for the maps to match Signed-off-by: Anton Nefedov <anton.nefe...@virtuozzo.com> --- block/qcow2.c | 22 +++++++++++++++++----- tests/qemu-iotests/060 | 2 +- tests/qemu-iotests/060.out | 3 ++- tests/qemu-iotests/066 | 2 +- tests/qemu-iotests/066.out | 4 ++-- 5 files changed, 23 insertions(+), 10 deletions(-) diff --git a/block/qcow2.c b/block/qcow2.c index e49ad50..9c49d40 100644 --- a/block/qcow2.c +++ b/block/qcow2.c @@ -2039,13 +2039,25 @@ static void handle_alloc_space(BlockDriverState *bs, QCowL2Meta *l2meta) QCowL2Meta *m; for (m = l2meta; m != NULL; m = m->next) { - if (s->prealloc_size && handle_prealloc(bs, m)) { - if (check_zero_cow(bs, m)) { - trace_qcow2_skip_cow(qemu_coroutine_self(), m->offset, - m->nb_clusters); - m->zero_cow = true; + bool preallocated_zeroes = s->prealloc_size && handle_prealloc(bs, m); + + if (!check_zero_cow(bs, m)) { + continue; + } + + if (!preallocated_zeroes && + (m->cow_start.nb_bytes != 0 || m->cow_end.nb_bytes != 0)) + { + if (bdrv_co_pwrite_zeroes(bs->file, m->alloc_offset, + m->nb_clusters * s->cluster_size, + BDRV_REQ_ALLOCATE) != 0) + { + continue; } } + + trace_qcow2_skip_cow(qemu_coroutine_self(), m->offset, m->nb_clusters); + m->zero_cow = true; } } diff --git a/tests/qemu-iotests/060 b/tests/qemu-iotests/060 index 8e95c45..3a0f096 100755 --- a/tests/qemu-iotests/060 +++ b/tests/qemu-iotests/060 @@ -160,7 +160,7 @@ poke_file "$TEST_IMG" '131084' "\x00\x00" # 0x2000c # any unallocated cluster, leading to an attempt to overwrite the second L2 # table. Finally, resume the COW write and see it fail (but not crash). echo "open -o file.driver=blkdebug $TEST_IMG -break cow_read 0 +break write_aio 0 aio_write 0k 1k wait_break 0 write 64k 64k diff --git a/tests/qemu-iotests/060.out b/tests/qemu-iotests/060.out index a20e267..290ccec 100644 --- a/tests/qemu-iotests/060.out +++ b/tests/qemu-iotests/060.out @@ -114,7 +114,8 @@ qcow2: Marking image as corrupt: Preventing invalid write on metadata (overlaps blkdebug: Suspended request '0' write failed: Input/output error blkdebug: Resuming request '0' -aio_write failed: No medium found +wrote 1024/1024 bytes at offset 0 +1 KiB, X ops; XX:XX:XX.X (XXX YYY/sec and XXX ops/sec) === Testing unallocated image header === diff --git a/tests/qemu-iotests/066 b/tests/qemu-iotests/066 index 8638217..3c216a1 100755 --- a/tests/qemu-iotests/066 +++ b/tests/qemu-iotests/066 @@ -71,7 +71,7 @@ echo _make_test_img $IMG_SIZE # Create data clusters (not aligned to an L2 table) -$QEMU_IO -c 'write -P 42 1M 256k' "$TEST_IMG" | _filter_qemu_io +$QEMU_IO -c "write -P 42 $(((1024 + 32) * 1024)) 192k" "$TEST_IMG" | _filter_qemu_io orig_map=$($QEMU_IMG map --output=json "$TEST_IMG") # Convert the data clusters to preallocated zero clusters diff --git a/tests/qemu-iotests/066.out b/tests/qemu-iotests/066.out index f94aa5c..81ef795 100644 --- a/tests/qemu-iotests/066.out +++ b/tests/qemu-iotests/066.out @@ -22,8 +22,8 @@ Offset Length Mapped to File === Writing to preallocated zero clusters === Formatting 'TEST_DIR/t.IMGFMT', fmt=IMGFMT size=67109376 -wrote 262144/262144 bytes at offset 1048576 -256 KiB, X ops; XX:XX:XX.X (XXX YYY/sec and XXX ops/sec) +wrote 196608/196608 bytes at offset 1081344 +192 KiB, X ops; XX:XX:XX.X (XXX YYY/sec and XXX ops/sec) wrote 262144/262144 bytes at offset 1048576 256 KiB, X ops; XX:XX:XX.X (XXX YYY/sec and XXX ops/sec) wrote 196608/196608 bytes at offset 1081344 -- 2.7.4