Re: [Qemu-devel] [PATCH v2 2/2] block: avoid recursive block_status call if possible

[Max Reitz <mreitz@redhat.com>]

[-- Attachment #1: Type: text/plain, Size: 4055 bytes --]

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


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 488 bytes --]