| Message ID | 20190408162617.258535-3-vsementsov@virtuozzo.com |
|---|---|
| State | New |
| Headers | show |
| Series | avoid lseek on block_status | expand |
Am 08.04.2019 um 18:26 hat Vladimir Sementsov-Ogievskiy geschrieben: > drv_co_block_status digs bs->file for additional, more accurate search > for hole inside region, reported as DATA by bs since 5daa74a6ebc. > > This accuracy is not free: assume we have qcow2 disk. Actually, qcow2 > knows, where are holes and where is data. But every block_status > request calls lseek additionally. Assume a big disk, full of > data, in any iterative copying block job (or img convert) we'll call > lseek(HOLE) on every iteration, and each of these lseeks will have to > iterate through all metadata up to the end of file. It's obviously > ineffective behavior. And for many scenarios we don't need this lseek > at all. > > However, lseek is needed when we have metadata-preallocated image. > > So, let's detect metadata-preallocation case and don't dig qcow2's > protocol file in other cases. > > The idea is to compare allocation size in POV of filesystem with > allocations size in POV of Qcow2 (by refcounts). If allocation in fs is > significantly lower, consider it as metadata-preallocation case. > > 102 iotest changed, as our detector can't detect shrinked file as > metadata-preallocation, which don't seem to be wrong, as with metadata > preallocation we always have valid file length. > > Other two iotests tiny changed QMP output sequence, which should be > exactly because skipped lseek at mirror beginning. No, these hunks don't show an improvement. An earlier {'return':{}} means that the block job coroutine has yielded earlier, so it's doing a blocking operation where it didn't do one before the patch. What happens is that qcow2_detect_metadata_preallocation() causes additional I/O by reading in the refcount block. I'll modify the commit message accordingly while applying. Kevin
On 08.04.19 18:26, Vladimir Sementsov-Ogievskiy wrote: > drv_co_block_status digs bs->file for additional, more accurate search > for hole inside region, reported as DATA by bs since 5daa74a6ebc. > > This accuracy is not free: assume we have qcow2 disk. Actually, qcow2 > knows, where are holes and where is data. But every block_status > request calls lseek additionally. Assume a big disk, full of > data, in any iterative copying block job (or img convert) we'll call > lseek(HOLE) on every iteration, and each of these lseeks will have to > iterate through all metadata up to the end of file. It's obviously > ineffective behavior. And for many scenarios we don't need this lseek > at all. > > However, lseek is needed when we have metadata-preallocated image. > > So, let's detect metadata-preallocation case and don't dig qcow2's > protocol file in other cases. > > The idea is to compare allocation size in POV of filesystem with > allocations size in POV of Qcow2 (by refcounts). If allocation in fs is > significantly lower, consider it as metadata-preallocation case. > > 102 iotest changed, as our detector can't detect shrinked file as > metadata-preallocation, which don't seem to be wrong, as with metadata > preallocation we always have valid file length. > > Other two iotests tiny changed QMP output sequence, which should be > exactly because skipped lseek at mirror beginning. > > Suggested-by: Denis V. Lunev <den@openvz.org> > Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com> > --- > block/qcow2.h | 4 ++++ > include/block/block.h | 8 +++++++- > block/io.c | 9 ++++++++- > block/qcow2-refcount.c | 32 ++++++++++++++++++++++++++++++++ > block/qcow2.c | 11 +++++++++++ > tests/qemu-iotests/102 | 2 +- > tests/qemu-iotests/102.out | 3 ++- > tests/qemu-iotests/141.out | 2 +- > tests/qemu-iotests/144.out | 2 +- > 9 files changed, 67 insertions(+), 6 deletions(-) For me, this patch breaks iotests 141 (for qed) and 211 (for vdi): > 141 1s ... [17:11:53] [17:11:53] - output mismatch (see 141.out.bad) > --- tests/qemu-iotests/141.out 2019-05-27 17:11:43.327664282 +0200 > +++ build/tests/qemu-iotests/141.out.bad 2019-05-27 17:11:53.949439880 +0200 > @@ -42,9 +42,9 @@ > {"return": {}} > {"timestamp": {"seconds": TIMESTAMP, "microseconds": TIMESTAMP}, "event": "JOB_STATUS_CHANGE", "data": {"status": "created", "id": "job0"}} > {"timestamp": {"seconds": TIMESTAMP, "microseconds": TIMESTAMP}, "event": "JOB_STATUS_CHANGE", "data": {"status": "running", "id": "job0"}} > -{"return": {}} > {"timestamp": {"seconds": TIMESTAMP, "microseconds": TIMESTAMP}, "event": "JOB_STATUS_CHANGE", "data": {"status": "ready", "id": "job0"}} > {"timestamp": {"seconds": TIMESTAMP, "microseconds": TIMESTAMP}, "event": "BLOCK_JOB_READY", "data": {"device": "job0", "len": 0, "offset": 0, "speed": 0, "type": "commit"}} > +{"return": {}} > {"error": {"class": "GenericError", "desc": "Node 'drv0' is busy: block device is in use by block job: commit"}} > {"return": {}} > {"timestamp": {"seconds": TIMESTAMP, "microseconds": TIMESTAMP}, "event": "JOB_STATUS_CHANGE", "data": {"status": "waiting", "id": "job0"}} and > 211 5s ... [17:11:54] [17:11:58] - output mismatch (see 211.out.bad) > --- tests/qemu-iotests/211.out 2019-05-22 19:58:34.870273427 +0200 > +++ build/tests/qemu-iotests/211.out.bad 2019-05-27 17:11:58.259348827 +0200 > @@ -55,8 +55,7 @@ > virtual size: 32 MiB (33554432 bytes) > cluster_size: 1048576 > > -[{ "start": 0, "length": 3072, "depth": 0, "zero": false, "data": true, "offset": 1024}, > -{ "start": 3072, "length": 33551360, "depth": 0, "zero": true, "data": true, "offset": 4096}] > +[{ "start": 0, "length": 33554432, "depth": 0, "zero": false, "data": true, "offset": 1024}] > > === Invalid BlockdevRef === Doesn’t look too bad, but still, broken iotests are broken iotests. :/ Max
27.05.2019 18:13, Max Reitz wrote: > On 08.04.19 18:26, Vladimir Sementsov-Ogievskiy wrote: >> drv_co_block_status digs bs->file for additional, more accurate search >> for hole inside region, reported as DATA by bs since 5daa74a6ebc. >> >> This accuracy is not free: assume we have qcow2 disk. Actually, qcow2 >> knows, where are holes and where is data. But every block_status >> request calls lseek additionally. Assume a big disk, full of >> data, in any iterative copying block job (or img convert) we'll call >> lseek(HOLE) on every iteration, and each of these lseeks will have to >> iterate through all metadata up to the end of file. It's obviously >> ineffective behavior. And for many scenarios we don't need this lseek >> at all. >> >> However, lseek is needed when we have metadata-preallocated image. >> >> So, let's detect metadata-preallocation case and don't dig qcow2's >> protocol file in other cases. >> >> The idea is to compare allocation size in POV of filesystem with >> allocations size in POV of Qcow2 (by refcounts). If allocation in fs is >> significantly lower, consider it as metadata-preallocation case. >> >> 102 iotest changed, as our detector can't detect shrinked file as >> metadata-preallocation, which don't seem to be wrong, as with metadata >> preallocation we always have valid file length. >> >> Other two iotests tiny changed QMP output sequence, which should be >> exactly because skipped lseek at mirror beginning. >> >> Suggested-by: Denis V. Lunev <den@openvz.org> >> Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com> >> --- >> block/qcow2.h | 4 ++++ >> include/block/block.h | 8 +++++++- >> block/io.c | 9 ++++++++- >> block/qcow2-refcount.c | 32 ++++++++++++++++++++++++++++++++ >> block/qcow2.c | 11 +++++++++++ >> tests/qemu-iotests/102 | 2 +- >> tests/qemu-iotests/102.out | 3 ++- >> tests/qemu-iotests/141.out | 2 +- >> tests/qemu-iotests/144.out | 2 +- >> 9 files changed, 67 insertions(+), 6 deletions(-) > > For me, this patch breaks iotests 141 (for qed) and 211 (for vdi): > >> 141 1s ... [17:11:53] [17:11:53] - output mismatch (see 141.out.bad) >> --- tests/qemu-iotests/141.out 2019-05-27 17:11:43.327664282 +0200 >> +++ build/tests/qemu-iotests/141.out.bad 2019-05-27 17:11:53.949439880 +0200 >> @@ -42,9 +42,9 @@ >> {"return": {}} >> {"timestamp": {"seconds": TIMESTAMP, "microseconds": TIMESTAMP}, "event": "JOB_STATUS_CHANGE", "data": {"status": "created", "id": "job0"}} >> {"timestamp": {"seconds": TIMESTAMP, "microseconds": TIMESTAMP}, "event": "JOB_STATUS_CHANGE", "data": {"status": "running", "id": "job0"}} >> -{"return": {}} >> {"timestamp": {"seconds": TIMESTAMP, "microseconds": TIMESTAMP}, "event": "JOB_STATUS_CHANGE", "data": {"status": "ready", "id": "job0"}} >> {"timestamp": {"seconds": TIMESTAMP, "microseconds": TIMESTAMP}, "event": "BLOCK_JOB_READY", "data": {"device": "job0", "len": 0, "offset": 0, "speed": 0, "type": "commit"}} >> +{"return": {}} >> {"error": {"class": "GenericError", "desc": "Node 'drv0' is busy: block device is in use by block job: commit"}} >> {"return": {}} >> {"timestamp": {"seconds": TIMESTAMP, "microseconds": TIMESTAMP}, "event": "JOB_STATUS_CHANGE", "data": {"status": "waiting", "id": "job0"}} > > and > >> 211 5s ... [17:11:54] [17:11:58] - output mismatch (see 211.out.bad) >> --- tests/qemu-iotests/211.out 2019-05-22 19:58:34.870273427 +0200 >> +++ build/tests/qemu-iotests/211.out.bad 2019-05-27 17:11:58.259348827 +0200 >> @@ -55,8 +55,7 @@ >> virtual size: 32 MiB (33554432 bytes) >> cluster_size: 1048576 >> >> -[{ "start": 0, "length": 3072, "depth": 0, "zero": false, "data": true, "offset": 1024}, >> -{ "start": 3072, "length": 33551360, "depth": 0, "zero": true, "data": true, "offset": 4096}] >> +[{ "start": 0, "length": 33554432, "depth": 0, "zero": false, "data": true, "offset": 1024}] >> >> === Invalid BlockdevRef === > > Doesn’t look too bad, but still, broken iotests are broken iotests. :/ > You are right, thanks for pointing to this! So, here is my investigation: about 211: aha, looks like I just didn't implement metadata preallocation detection for vdi, and default behavior is changed. I don't know vdi code, but after fast look, it seems that it may be as simple as (and it fixes the test): diff --git a/block/vdi.c b/block/vdi.c index d7ef6628e7..a72333dcf4 100644 --- a/block/vdi.c +++ b/block/vdi.c @@ -542,7 +542,9 @@ static int coroutine_fn vdi_co_block_status(BlockDriverState *bs, *map = s->header.offset_data + (uint64_t)bmap_entry * s->block_size + index_in_block; *file = bs->file->bs; - return BDRV_BLOCK_DATA | BDRV_BLOCK_OFFSET_VALID; + + return BDRV_BLOCK_DATA | BDRV_BLOCK_OFFSET_VALID | + (s->header.image_type == VDI_TYPE_STATIC ? BDRV_BLOCK_RECURSE : 0); } Or, we can rollback vdi behavior to pre-patch state: diff --git a/block/vdi.c b/block/vdi.c index d7ef6628e7..074256d845 100644 --- a/block/vdi.c +++ b/block/vdi.c @@ -542,7 +542,7 @@ static int coroutine_fn vdi_co_block_status(BlockDriverState *bs, *map = s->header.offset_data + (uint64_t)bmap_entry * s->block_size + index_in_block; *file = bs->file->bs; - return BDRV_BLOCK_DATA | BDRV_BLOCK_OFFSET_VALID; + return BDRV_BLOCK_DATA | BDRV_BLOCK_OFFSET_VALID | BDRV_BLOCK_RECURSE; } static int coroutine_fn about 141: aha, it's a difference between qcow2 and qed, because of changed io sequence for qcow2, as Kevin described "What happens is that qcow2_detect_metadata_preallocation() causes additional I/O by reading in the refcount block.". I think correct way to fix it is to filter {"return": {}} elements, as we can't be sure where they occur among events: diff --git a/tests/qemu-iotests/141 b/tests/qemu-iotests/141 index 2197a82d45..2704695921 100755 --- a/tests/qemu-iotests/141 +++ b/tests/qemu-iotests/141 @@ -58,16 +58,20 @@ test_blockjob() }}}" \ 'return' + # We may get {"return": {}} result of the following command among events in + # the following output or after all events in the next output (if $2 is an + # event, not 'return'). So, filter it here and in the following output to + # avoid race in test output. _send_qemu_cmd $QEMU_HANDLE \ "$1" \ "$2" \ - | _filter_img_create + | _filter_img_create | _filter_qmp_empty_return # We want this to return an error because the block job is still running _send_qemu_cmd $QEMU_HANDLE \ "{'execute': 'blockdev-del', 'arguments': {'node-name': 'drv0'}}" \ - 'error' | _filter_generated_node_ids + 'error' | _filter_generated_node_ids | _filter_qmp_empty_return _send_qemu_cmd $QEMU_HANDLE \ "{'execute': 'block-job-cancel', diff --git a/tests/qemu-iotests/141.out b/tests/qemu-iotests/141.out index 4d71d9dcae..dbd3bdef6c 100644 --- a/tests/qemu-iotests/141.out +++ b/tests/qemu-iotests/141.out @@ -10,7 +10,6 @@ Formatting 'TEST_DIR/t.IMGFMT', fmt=IMGFMT size=1048576 backing_file=TEST_DIR/m. Formatting 'TEST_DIR/o.IMGFMT', fmt=IMGFMT size=1048576 backing_file=TEST_DIR/t.IMGFMT backing_fmt=IMGFMT {"timestamp": {"seconds": TIMESTAMP, "microseconds": TIMESTAMP}, "event": "JOB_STATUS_CHANGE", "data": {"status": "created", "id": "job0"}} {"timestamp": {"seconds": TIMESTAMP, "microseconds": TIMESTAMP}, "event": "JOB_STATUS_CHANGE", "data": {"status": "running", "id": "job0"}} -{"return": {}} {"error": {"class": "GenericError", "desc": "Node drv0 is in use"}} {"return": {}} {"timestamp": {"seconds": TIMESTAMP, "microseconds": TIMESTAMP}, "event": "JOB_STATUS_CHANGE", "data": {"status": "aborting", "id": "job0"}} @@ -27,7 +26,6 @@ Formatting 'TEST_DIR/o.IMGFMT', fmt=IMGFMT size=1048576 backing_file=TEST_DIR/t. {"timestamp": {"seconds": TIMESTAMP, "microseconds": TIMESTAMP}, "event": "JOB_STATUS_CHANGE", "data": {"status": "running", "id": "job0"}} {"timestamp": {"seconds": TIMESTAMP, "microseconds": TIMESTAMP}, "event": "JOB_STATUS_CHANGE", "data": {"status": "ready", "id": "job0"}} {"timestamp": {"seconds": TIMESTAMP, "microseconds": TIMESTAMP}, "event": "BLOCK_JOB_READY", "data": {"device": "job0", "len": 0, "offset": 0, "speed": 0, "type": "mirror"}} -{"return": {}} {"error": {"class": "GenericError", "desc": "Node 'drv0' is busy: block device is in use by block job: mirror"}} {"return": {}} {"timestamp": {"seconds": TIMESTAMP, "microseconds": TIMESTAMP}, "event": "JOB_STATUS_CHANGE", "data": {"status": "waiting", "id": "job0"}} @@ -42,7 +40,6 @@ Formatting 'TEST_DIR/o.IMGFMT', fmt=IMGFMT size=1048576 backing_file=TEST_DIR/t. {"return": {}} {"timestamp": {"seconds": TIMESTAMP, "microseconds": TIMESTAMP}, "event": "JOB_STATUS_CHANGE", "data": {"status": "created", "id": "job0"}} {"timestamp": {"seconds": TIMESTAMP, "microseconds": TIMESTAMP}, "event": "JOB_STATUS_CHANGE", "data": {"status": "running", "id": "job0"}} -{"return": {}} {"timestamp": {"seconds": TIMESTAMP, "microseconds": TIMESTAMP}, "event": "JOB_STATUS_CHANGE", "data": {"status": "ready", "id": "job0"}} {"timestamp": {"seconds": TIMESTAMP, "microseconds": TIMESTAMP}, "event": "BLOCK_JOB_READY", "data": {"device": "job0", "len": 0, "offset": 0, "speed": 0, "type": "commit"}} {"error": {"class": "GenericError", "desc": "Node 'drv0' is busy: block device is in use by block job: commit"}} @@ -61,7 +58,6 @@ wrote 1048576/1048576 bytes at offset 0 {"return": {}} {"timestamp": {"seconds": TIMESTAMP, "microseconds": TIMESTAMP}, "event": "JOB_STATUS_CHANGE", "data": {"status": "created", "id": "job0"}} {"timestamp": {"seconds": TIMESTAMP, "microseconds": TIMESTAMP}, "event": "JOB_STATUS_CHANGE", "data": {"status": "running", "id": "job0"}} -{"return": {}} {"error": {"class": "GenericError", "desc": "Node drv0 is in use"}} {"return": {}} {"timestamp": {"seconds": TIMESTAMP, "microseconds": TIMESTAMP}, "event": "JOB_STATUS_CHANGE", "data": {"status": "aborting", "id": "job0"}} @@ -77,7 +73,6 @@ wrote 1048576/1048576 bytes at offset 0 {"return": {}} {"timestamp": {"seconds": TIMESTAMP, "microseconds": TIMESTAMP}, "event": "JOB_STATUS_CHANGE", "data": {"status": "created", "id": "job0"}} {"timestamp": {"seconds": TIMESTAMP, "microseconds": TIMESTAMP}, "event": "JOB_STATUS_CHANGE", "data": {"status": "running", "id": "job0"}} -{"return": {}} {"error": {"class": "GenericError", "desc": "Node drv0 is in use"}} {"return": {}} {"timestamp": {"seconds": TIMESTAMP, "microseconds": TIMESTAMP}, "event": "JOB_STATUS_CHANGE", "data": {"status": "aborting", "id": "job0"}} diff --git a/tests/qemu-iotests/common.filter b/tests/qemu-iotests/common.filter index 35fddc746f..8e9235d6fe 100644 --- a/tests/qemu-iotests/common.filter +++ b/tests/qemu-iotests/common.filter @@ -219,5 +219,10 @@ _filter_nbd() -e 's#\(foo\|PORT/\?\|.sock\): Failed to .*$#\1#' } +_filter_qmp_empty_return() +{ + grep -v '{"return": {}}' +} + # make sure this script returns success true
On 28.05.19 08:39, Vladimir Sementsov-Ogievskiy wrote: > 27.05.2019 18:13, Max Reitz wrote: >> On 08.04.19 18:26, Vladimir Sementsov-Ogievskiy wrote: >>> drv_co_block_status digs bs->file for additional, more accurate search >>> for hole inside region, reported as DATA by bs since 5daa74a6ebc. >>> >>> This accuracy is not free: assume we have qcow2 disk. Actually, qcow2 >>> knows, where are holes and where is data. But every block_status >>> request calls lseek additionally. Assume a big disk, full of >>> data, in any iterative copying block job (or img convert) we'll call >>> lseek(HOLE) on every iteration, and each of these lseeks will have to >>> iterate through all metadata up to the end of file. It's obviously >>> ineffective behavior. And for many scenarios we don't need this lseek >>> at all. >>> >>> However, lseek is needed when we have metadata-preallocated image. >>> >>> So, let's detect metadata-preallocation case and don't dig qcow2's >>> protocol file in other cases. >>> >>> The idea is to compare allocation size in POV of filesystem with >>> allocations size in POV of Qcow2 (by refcounts). If allocation in fs is >>> significantly lower, consider it as metadata-preallocation case. >>> >>> 102 iotest changed, as our detector can't detect shrinked file as >>> metadata-preallocation, which don't seem to be wrong, as with metadata >>> preallocation we always have valid file length. >>> >>> Other two iotests tiny changed QMP output sequence, which should be >>> exactly because skipped lseek at mirror beginning. >>> >>> Suggested-by: Denis V. Lunev <den@openvz.org> >>> Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com> >>> --- >>> block/qcow2.h | 4 ++++ >>> include/block/block.h | 8 +++++++- >>> block/io.c | 9 ++++++++- >>> block/qcow2-refcount.c | 32 ++++++++++++++++++++++++++++++++ >>> block/qcow2.c | 11 +++++++++++ >>> tests/qemu-iotests/102 | 2 +- >>> tests/qemu-iotests/102.out | 3 ++- >>> tests/qemu-iotests/141.out | 2 +- >>> tests/qemu-iotests/144.out | 2 +- >>> 9 files changed, 67 insertions(+), 6 deletions(-) >> >> For me, this patch breaks iotests 141 (for qed) and 211 (for vdi): >> >>> 141 1s ... [17:11:53] [17:11:53] - output mismatch (see 141.out.bad) >>> --- tests/qemu-iotests/141.out 2019-05-27 17:11:43.327664282 +0200 >>> +++ build/tests/qemu-iotests/141.out.bad 2019-05-27 17:11:53.949439880 +0200 >>> @@ -42,9 +42,9 @@ >>> {"return": {}} >>> {"timestamp": {"seconds": TIMESTAMP, "microseconds": TIMESTAMP}, "event": "JOB_STATUS_CHANGE", "data": {"status": "created", "id": "job0"}} >>> {"timestamp": {"seconds": TIMESTAMP, "microseconds": TIMESTAMP}, "event": "JOB_STATUS_CHANGE", "data": {"status": "running", "id": "job0"}} >>> -{"return": {}} >>> {"timestamp": {"seconds": TIMESTAMP, "microseconds": TIMESTAMP}, "event": "JOB_STATUS_CHANGE", "data": {"status": "ready", "id": "job0"}} >>> {"timestamp": {"seconds": TIMESTAMP, "microseconds": TIMESTAMP}, "event": "BLOCK_JOB_READY", "data": {"device": "job0", "len": 0, "offset": 0, "speed": 0, "type": "commit"}} >>> +{"return": {}} >>> {"error": {"class": "GenericError", "desc": "Node 'drv0' is busy: block device is in use by block job: commit"}} >>> {"return": {}} >>> {"timestamp": {"seconds": TIMESTAMP, "microseconds": TIMESTAMP}, "event": "JOB_STATUS_CHANGE", "data": {"status": "waiting", "id": "job0"}} >> >> and >> >>> 211 5s ... [17:11:54] [17:11:58] - output mismatch (see 211.out.bad) >>> --- tests/qemu-iotests/211.out 2019-05-22 19:58:34.870273427 +0200 >>> +++ build/tests/qemu-iotests/211.out.bad 2019-05-27 17:11:58.259348827 +0200 >>> @@ -55,8 +55,7 @@ >>> virtual size: 32 MiB (33554432 bytes) >>> cluster_size: 1048576 >>> >>> -[{ "start": 0, "length": 3072, "depth": 0, "zero": false, "data": true, "offset": 1024}, >>> -{ "start": 3072, "length": 33551360, "depth": 0, "zero": true, "data": true, "offset": 4096}] >>> +[{ "start": 0, "length": 33554432, "depth": 0, "zero": false, "data": true, "offset": 1024}] >>> >>> === Invalid BlockdevRef === >> >> Doesn’t look too bad, but still, broken iotests are broken iotests. :/ >> > > > You are right, thanks for pointing to this! So, here is my investigation: > > about 211: aha, looks like I just didn't implement metadata preallocation detection for vdi, > and default behavior is changed. I don't know vdi code, but after fast look, it seems that > it may be as simple as (and it fixes the test): > > > diff --git a/block/vdi.c b/block/vdi.c > index d7ef6628e7..a72333dcf4 100644 > --- a/block/vdi.c > +++ b/block/vdi.c > @@ -542,7 +542,9 @@ static int coroutine_fn vdi_co_block_status(BlockDriverState *bs, > *map = s->header.offset_data + (uint64_t)bmap_entry * s->block_size + > index_in_block; > *file = bs->file->bs; > - return BDRV_BLOCK_DATA | BDRV_BLOCK_OFFSET_VALID; > + > + return BDRV_BLOCK_DATA | BDRV_BLOCK_OFFSET_VALID | > + (s->header.image_type == VDI_TYPE_STATIC ? BDRV_BLOCK_RECURSE : 0); > } That would be the same as what this patch implements for qcow2, yes. > Or, we can rollback vdi behavior to pre-patch state: > > > > diff --git a/block/vdi.c b/block/vdi.c > index d7ef6628e7..074256d845 100644 > --- a/block/vdi.c > +++ b/block/vdi.c > @@ -542,7 +542,7 @@ static int coroutine_fn vdi_co_block_status(BlockDriverState *bs, > *map = s->header.offset_data + (uint64_t)bmap_entry * s->block_size + > index_in_block; > *file = bs->file->bs; > - return BDRV_BLOCK_DATA | BDRV_BLOCK_OFFSET_VALID; > + return BDRV_BLOCK_DATA | BDRV_BLOCK_OFFSET_VALID | BDRV_BLOCK_RECURSE; > } > > static int coroutine_fn > > > > > about 141: aha, it's a difference between qcow2 and qed, because of changed io sequence for > qcow2, as Kevin described "What happens is that qcow2_detect_metadata_preallocation() causes > additional I/O by reading in the refcount block.". I think correct way to fix it is to filter > {"return": {}} elements, as we can't be sure where they occur among events: > > > > > diff --git a/tests/qemu-iotests/141 b/tests/qemu-iotests/141 > index 2197a82d45..2704695921 100755 > --- a/tests/qemu-iotests/141 > +++ b/tests/qemu-iotests/141 > @@ -58,16 +58,20 @@ test_blockjob() > }}}" \ > 'return' > > + # We may get {"return": {}} result of the following command among events in > + # the following output or after all events in the next output (if $2 is an > + # event, not 'return'). So, filter it here and in the following output to > + # avoid race in test output. > _send_qemu_cmd $QEMU_HANDLE \ > "$1" \ > "$2" \ > - | _filter_img_create > + | _filter_img_create | _filter_qmp_empty_return > > # We want this to return an error because the block job is still running > _send_qemu_cmd $QEMU_HANDLE \ > "{'execute': 'blockdev-del', > 'arguments': {'node-name': 'drv0'}}" \ > - 'error' | _filter_generated_node_ids > + 'error' | _filter_generated_node_ids | _filter_qmp_empty_return > > _send_qemu_cmd $QEMU_HANDLE \ > "{'execute': 'block-job-cancel', > diff --git a/tests/qemu-iotests/141.out b/tests/qemu-iotests/141.out > index 4d71d9dcae..dbd3bdef6c 100644 > --- a/tests/qemu-iotests/141.out > +++ b/tests/qemu-iotests/141.out > @@ -10,7 +10,6 @@ Formatting 'TEST_DIR/t.IMGFMT', fmt=IMGFMT size=1048576 backing_file=TEST_DIR/m. > Formatting 'TEST_DIR/o.IMGFMT', fmt=IMGFMT size=1048576 backing_file=TEST_DIR/t.IMGFMT backing_fmt=IMGFMT > {"timestamp": {"seconds": TIMESTAMP, "microseconds": TIMESTAMP}, "event": "JOB_STATUS_CHANGE", "data": {"status": "created", "id": "job0"}} > {"timestamp": {"seconds": TIMESTAMP, "microseconds": TIMESTAMP}, "event": "JOB_STATUS_CHANGE", "data": {"status": "running", "id": "job0"}} > -{"return": {}} > {"error": {"class": "GenericError", "desc": "Node drv0 is in use"}} > {"return": {}} > {"timestamp": {"seconds": TIMESTAMP, "microseconds": TIMESTAMP}, "event": "JOB_STATUS_CHANGE", "data": {"status": "aborting", "id": "job0"}} > @@ -27,7 +26,6 @@ Formatting 'TEST_DIR/o.IMGFMT', fmt=IMGFMT size=1048576 backing_file=TEST_DIR/t. > {"timestamp": {"seconds": TIMESTAMP, "microseconds": TIMESTAMP}, "event": "JOB_STATUS_CHANGE", "data": {"status": "running", "id": "job0"}} > {"timestamp": {"seconds": TIMESTAMP, "microseconds": TIMESTAMP}, "event": "JOB_STATUS_CHANGE", "data": {"status": "ready", "id": "job0"}} > {"timestamp": {"seconds": TIMESTAMP, "microseconds": TIMESTAMP}, "event": "BLOCK_JOB_READY", "data": {"device": "job0", "len": 0, "offset": 0, "speed": 0, "type": "mirror"}} > -{"return": {}} > {"error": {"class": "GenericError", "desc": "Node 'drv0' is busy: block device is in use by block job: mirror"}} > {"return": {}} > {"timestamp": {"seconds": TIMESTAMP, "microseconds": TIMESTAMP}, "event": "JOB_STATUS_CHANGE", "data": {"status": "waiting", "id": "job0"}} > @@ -42,7 +40,6 @@ Formatting 'TEST_DIR/o.IMGFMT', fmt=IMGFMT size=1048576 backing_file=TEST_DIR/t. > {"return": {}} > {"timestamp": {"seconds": TIMESTAMP, "microseconds": TIMESTAMP}, "event": "JOB_STATUS_CHANGE", "data": {"status": "created", "id": "job0"}} > {"timestamp": {"seconds": TIMESTAMP, "microseconds": TIMESTAMP}, "event": "JOB_STATUS_CHANGE", "data": {"status": "running", "id": "job0"}} > -{"return": {}} > {"timestamp": {"seconds": TIMESTAMP, "microseconds": TIMESTAMP}, "event": "JOB_STATUS_CHANGE", "data": {"status": "ready", "id": "job0"}} > {"timestamp": {"seconds": TIMESTAMP, "microseconds": TIMESTAMP}, "event": "BLOCK_JOB_READY", "data": {"device": "job0", "len": 0, "offset": 0, "speed": 0, "type": "commit"}} > {"error": {"class": "GenericError", "desc": "Node 'drv0' is busy: block device is in use by block job: commit"}} > @@ -61,7 +58,6 @@ wrote 1048576/1048576 bytes at offset 0 > {"return": {}} > {"timestamp": {"seconds": TIMESTAMP, "microseconds": TIMESTAMP}, "event": "JOB_STATUS_CHANGE", "data": {"status": "created", "id": "job0"}} > {"timestamp": {"seconds": TIMESTAMP, "microseconds": TIMESTAMP}, "event": "JOB_STATUS_CHANGE", "data": {"status": "running", "id": "job0"}} > -{"return": {}} > {"error": {"class": "GenericError", "desc": "Node drv0 is in use"}} > {"return": {}} > {"timestamp": {"seconds": TIMESTAMP, "microseconds": TIMESTAMP}, "event": "JOB_STATUS_CHANGE", "data": {"status": "aborting", "id": "job0"}} > @@ -77,7 +73,6 @@ wrote 1048576/1048576 bytes at offset 0 > {"return": {}} > {"timestamp": {"seconds": TIMESTAMP, "microseconds": TIMESTAMP}, "event": "JOB_STATUS_CHANGE", "data": {"status": "created", "id": "job0"}} > {"timestamp": {"seconds": TIMESTAMP, "microseconds": TIMESTAMP}, "event": "JOB_STATUS_CHANGE", "data": {"status": "running", "id": "job0"}} > -{"return": {}} > {"error": {"class": "GenericError", "desc": "Node drv0 is in use"}} > {"return": {}} > {"timestamp": {"seconds": TIMESTAMP, "microseconds": TIMESTAMP}, "event": "JOB_STATUS_CHANGE", "data": {"status": "aborting", "id": "job0"}} > diff --git a/tests/qemu-iotests/common.filter b/tests/qemu-iotests/common.filter > index 35fddc746f..8e9235d6fe 100644 > --- a/tests/qemu-iotests/common.filter > +++ b/tests/qemu-iotests/common.filter > @@ -219,5 +219,10 @@ _filter_nbd() > -e 's#\(foo\|PORT/\?\|.sock\): Failed to .*$#\1#' > } > > +_filter_qmp_empty_return() > +{ > + grep -v '{"return": {}}' > +} > + > # make sure this script returns success > true I’m not really a fan of this, but I don’t think there is an alternative for a bash test. So that does look like the best solution for me. Max
diff --git a/block/qcow2.h b/block/qcow2.h index fdee297f33..b6135d8271 100644 --- a/block/qcow2.h +++ b/block/qcow2.h @@ -350,6 +350,9 @@ typedef struct BDRVQcow2State { int nb_compress_threads; BdrvChild *data_file; + + bool metadata_preallocation_checked; + bool metadata_preallocation; } BDRVQcow2State; typedef struct Qcow2COWRegion { @@ -643,6 +646,7 @@ int qcow2_change_refcount_order(BlockDriverState *bs, int refcount_order, void *cb_opaque, Error **errp); int qcow2_shrink_reftable(BlockDriverState *bs); int64_t qcow2_get_last_cluster(BlockDriverState *bs, int64_t size); +int qcow2_detect_metadata_preallocation(BlockDriverState *bs); /* qcow2-cluster.c functions */ int qcow2_grow_l1_table(BlockDriverState *bs, uint64_t min_size, diff --git a/include/block/block.h b/include/block/block.h index c7a26199aa..1f2f08e4ee 100644 --- a/include/block/block.h +++ b/include/block/block.h @@ -156,10 +156,15 @@ typedef struct HDGeometry { * BDRV_BLOCK_EOF: the returned pnum covers through end of file for this * layer, set by block layer * - * Internal flag: + * Internal flags: * BDRV_BLOCK_RAW: for use by passthrough drivers, such as raw, to request * that the block layer recompute the answer from the returned * BDS; must be accompanied by just BDRV_BLOCK_OFFSET_VALID. + * BDRV_BLOCK_RECURSE: request that the block layer will recursively search for + * zeroes in file child of current block node inside + * returned region. Only valid together with both + * BDRV_BLOCK_DATA and BDRV_BLOCK_OFFSET_VALID. Should not + * appear with BDRV_BLOCK_ZERO. * * If BDRV_BLOCK_OFFSET_VALID is set, the map parameter represents the * host offset within the returned BDS that is allocated for the @@ -184,6 +189,7 @@ typedef struct HDGeometry { #define BDRV_BLOCK_RAW 0x08 #define BDRV_BLOCK_ALLOCATED 0x10 #define BDRV_BLOCK_EOF 0x20 +#define BDRV_BLOCK_RECURSE 0x40 #define BDRV_BLOCK_OFFSET_MASK BDRV_SECTOR_MASK typedef QSIMPLEQ_HEAD(BlockReopenQueue, BlockReopenQueueEntry) BlockReopenQueue; diff --git a/block/io.c b/block/io.c index dfc153b8d8..8595d4b504 100644 --- a/block/io.c +++ b/block/io.c @@ -2121,6 +2121,12 @@ static int coroutine_fn bdrv_co_block_status(BlockDriverState *bs, */ assert(*pnum && QEMU_IS_ALIGNED(*pnum, align) && align > offset - aligned_offset); + if (ret & BDRV_BLOCK_RECURSE) { + assert(ret & BDRV_BLOCK_DATA); + assert(ret & BDRV_BLOCK_OFFSET_VALID); + assert(!(ret & BDRV_BLOCK_ZERO)); + } + *pnum -= offset - aligned_offset; if (*pnum > bytes) { *pnum = bytes; @@ -2151,7 +2157,8 @@ static int coroutine_fn bdrv_co_block_status(BlockDriverState *bs, } } - if (want_zero && local_file && local_file != bs && + if (want_zero && ret & BDRV_BLOCK_RECURSE && + local_file && local_file != bs && (ret & BDRV_BLOCK_DATA) && !(ret & BDRV_BLOCK_ZERO) && (ret & BDRV_BLOCK_OFFSET_VALID)) { int64_t file_pnum; diff --git a/block/qcow2-refcount.c b/block/qcow2-refcount.c index e0fe322500..5786ba93e0 100644 --- a/block/qcow2-refcount.c +++ b/block/qcow2-refcount.c @@ -3422,3 +3422,35 @@ int64_t qcow2_get_last_cluster(BlockDriverState *bs, int64_t size) "There are no references in the refcount table."); return -EIO; } + +int qcow2_detect_metadata_preallocation(BlockDriverState *bs) +{ + BDRVQcow2State *s = bs->opaque; + int64_t i, end_cluster, cluster_count = 0, threshold; + int64_t file_length, real_allocation, real_clusters; + + file_length = bdrv_getlength(bs->file->bs); + if (file_length < 0) { + return file_length; + } + + real_allocation = bdrv_get_allocated_file_size(bs->file->bs); + if (real_allocation < 0) { + return real_allocation; + } + + real_clusters = real_allocation / s->cluster_size; + threshold = MAX(real_clusters * 10 / 9, real_clusters + 2); + + end_cluster = size_to_clusters(s, file_length); + for (i = 0; i < end_cluster && cluster_count < threshold; i++) { + uint64_t refcount; + int ret = qcow2_get_refcount(bs, i, &refcount); + if (ret < 0) { + return ret; + } + cluster_count += !!refcount; + } + + return cluster_count >= threshold; +} diff --git a/block/qcow2.c b/block/qcow2.c index d507ee0686..e89e1c4fb1 100644 --- a/block/qcow2.c +++ b/block/qcow2.c @@ -1900,6 +1900,12 @@ static int coroutine_fn qcow2_co_block_status(BlockDriverState *bs, unsigned int bytes; int status = 0; + if (!s->metadata_preallocation_checked) { + ret = qcow2_detect_metadata_preallocation(bs); + s->metadata_preallocation = (ret == 1); + s->metadata_preallocation_checked = true; + } + bytes = MIN(INT_MAX, count); qemu_co_mutex_lock(&s->lock); ret = qcow2_get_cluster_offset(bs, offset, &bytes, &cluster_offset); @@ -1922,6 +1928,11 @@ static int coroutine_fn qcow2_co_block_status(BlockDriverState *bs, } else if (ret != QCOW2_CLUSTER_UNALLOCATED) { status |= BDRV_BLOCK_DATA; } + if (s->metadata_preallocation && (status & BDRV_BLOCK_DATA) && + (status & BDRV_BLOCK_OFFSET_VALID)) + { + status |= BDRV_BLOCK_RECURSE; + } return status; } diff --git a/tests/qemu-iotests/102 b/tests/qemu-iotests/102 index cedd2b25dc..125402db2b 100755 --- a/tests/qemu-iotests/102 +++ b/tests/qemu-iotests/102 @@ -56,7 +56,7 @@ $QEMU_IO -c 'write 0 64k' "$TEST_IMG" | _filter_qemu_io $QEMU_IMG resize -f raw --shrink "$TEST_IMG" $((5 * 64 * 1024)) $QEMU_IO -c map "$TEST_IMG" -$QEMU_IMG map "$TEST_IMG" +$QEMU_IMG map "$TEST_IMG" | _filter_qemu_img_map echo echo '=== Testing map on an image file truncated outside of qemu ===' diff --git a/tests/qemu-iotests/102.out b/tests/qemu-iotests/102.out index 4401b08fee..cd2fdc7f96 100644 --- a/tests/qemu-iotests/102.out +++ b/tests/qemu-iotests/102.out @@ -7,7 +7,8 @@ wrote 65536/65536 bytes at offset 0 64 KiB, X ops; XX:XX:XX.X (XXX YYY/sec and XXX ops/sec) Image resized. 64 KiB (0x10000) bytes allocated at offset 0 bytes (0x0) -Offset Length Mapped to File +Offset Length File +0 0x10000 TEST_DIR/t.IMGFMT === Testing map on an image file truncated outside of qemu === diff --git a/tests/qemu-iotests/141.out b/tests/qemu-iotests/141.out index 41c7291258..4d71d9dcae 100644 --- a/tests/qemu-iotests/141.out +++ b/tests/qemu-iotests/141.out @@ -42,9 +42,9 @@ Formatting 'TEST_DIR/o.IMGFMT', fmt=IMGFMT size=1048576 backing_file=TEST_DIR/t. {"return": {}} {"timestamp": {"seconds": TIMESTAMP, "microseconds": TIMESTAMP}, "event": "JOB_STATUS_CHANGE", "data": {"status": "created", "id": "job0"}} {"timestamp": {"seconds": TIMESTAMP, "microseconds": TIMESTAMP}, "event": "JOB_STATUS_CHANGE", "data": {"status": "running", "id": "job0"}} +{"return": {}} {"timestamp": {"seconds": TIMESTAMP, "microseconds": TIMESTAMP}, "event": "JOB_STATUS_CHANGE", "data": {"status": "ready", "id": "job0"}} {"timestamp": {"seconds": TIMESTAMP, "microseconds": TIMESTAMP}, "event": "BLOCK_JOB_READY", "data": {"device": "job0", "len": 0, "offset": 0, "speed": 0, "type": "commit"}} -{"return": {}} {"error": {"class": "GenericError", "desc": "Node 'drv0' is busy: block device is in use by block job: commit"}} {"return": {}} {"timestamp": {"seconds": TIMESTAMP, "microseconds": TIMESTAMP}, "event": "JOB_STATUS_CHANGE", "data": {"status": "waiting", "id": "job0"}} diff --git a/tests/qemu-iotests/144.out b/tests/qemu-iotests/144.out index 55299201e4..a9a8216bea 100644 --- a/tests/qemu-iotests/144.out +++ b/tests/qemu-iotests/144.out @@ -14,10 +14,10 @@ Formatting 'TEST_DIR/tmp.qcow2', fmt=qcow2 size=536870912 backing_file=TEST_DIR/ {"timestamp": {"seconds": TIMESTAMP, "microseconds": TIMESTAMP}, "event": "JOB_STATUS_CHANGE", "data": {"status": "created", "id": "virtio0"}} {"timestamp": {"seconds": TIMESTAMP, "microseconds": TIMESTAMP}, "event": "JOB_STATUS_CHANGE", "data": {"status": "running", "id": "virtio0"}} +{"return": {}} {"timestamp": {"seconds": TIMESTAMP, "microseconds": TIMESTAMP}, "event": "JOB_STATUS_CHANGE", "data": {"status": "ready", "id": "virtio0"}} {"timestamp": {"seconds": TIMESTAMP, "microseconds": TIMESTAMP}, "event": "BLOCK_JOB_READY", "data": {"device": "virtio0", "len": 0, "offset": 0, "speed": 0, "type": "commit"}} {"return": {}} -{"return": {}} {"timestamp": {"seconds": TIMESTAMP, "microseconds": TIMESTAMP}, "event": "JOB_STATUS_CHANGE", "data": {"status": "waiting", "id": "virtio0"}} {"timestamp": {"seconds": TIMESTAMP, "microseconds": TIMESTAMP}, "event": "JOB_STATUS_CHANGE", "data": {"status": "pending", "id": "virtio0"}} {"timestamp": {"seconds": TIMESTAMP, "microseconds": TIMESTAMP}, "event": "BLOCK_JOB_COMPLETED", "data": {"device": "virtio0", "len": 0, "offset": 0, "speed": 0, "type": "commit"}}
drv_co_block_status digs bs->file for additional, more accurate search for hole inside region, reported as DATA by bs since 5daa74a6ebc. This accuracy is not free: assume we have qcow2 disk. Actually, qcow2 knows, where are holes and where is data. But every block_status request calls lseek additionally. Assume a big disk, full of data, in any iterative copying block job (or img convert) we'll call lseek(HOLE) on every iteration, and each of these lseeks will have to iterate through all metadata up to the end of file. It's obviously ineffective behavior. And for many scenarios we don't need this lseek at all. However, lseek is needed when we have metadata-preallocated image. So, let's detect metadata-preallocation case and don't dig qcow2's protocol file in other cases. The idea is to compare allocation size in POV of filesystem with allocations size in POV of Qcow2 (by refcounts). If allocation in fs is significantly lower, consider it as metadata-preallocation case. 102 iotest changed, as our detector can't detect shrinked file as metadata-preallocation, which don't seem to be wrong, as with metadata preallocation we always have valid file length. Other two iotests tiny changed QMP output sequence, which should be exactly because skipped lseek at mirror beginning. Suggested-by: Denis V. Lunev <den@openvz.org> Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com> --- block/qcow2.h | 4 ++++ include/block/block.h | 8 +++++++- block/io.c | 9 ++++++++- block/qcow2-refcount.c | 32 ++++++++++++++++++++++++++++++++ block/qcow2.c | 11 +++++++++++ tests/qemu-iotests/102 | 2 +- tests/qemu-iotests/102.out | 3 ++- tests/qemu-iotests/141.out | 2 +- tests/qemu-iotests/144.out | 2 +- 9 files changed, 67 insertions(+), 6 deletions(-)