On 07/17/2013 08:03 AM, Wenchao Xia wrote:
> This series allow user to read internal snapshot's contents without qemu-img
> convert. Another purpose is that, when qemu is online and have taken an
> internal snapshot, let user invoke qemu-nbd to do any thing on it except write.
> 
> This brings two interesting issues:
> 1 is it safe to let qemu-nbd and qemu access that file at same time?

Probably not, for the same reason we tell people to not use qemu-img
while qemu is active on a file.

> I think it is safe, since qemu-nbd is read only. The data will be correct from
> qemu-nbd, if qemu does not delete that snapshot when qemu-nbd is running, and
> data is flushed to storage after qemu take that snapshot so that qemu-nbd
> would see the correct data.

You're making assumptions that qemu won't be touching any metadata in a
manner in which the read-only qemu-nbd could get confused; I'm not sure
we are ready to make that guarantee.  I think the export has to be from
the running qemu process itself, rather than from a second process.

> 
> 2 should an nbd-server exporting internal snapshot be added in qemu?
> I think no. Compared with driver-backup, the snapshot, or COW happens
> in storage level, so it allows another program to read it itself. Actually
> it should be OK to let another server other than qemu's host, do the
> export I/O job, if data is flushed.

Unfortunately, I disagree, and think the answer to this question is yes,
we need to do the export from within the single qemu process, if we want
to guarantee safety.

> 
> Next step:
> As demonstrated before, an explict API should be added, which make sure
> all I/O request is flushed and sent to underlining storage, and cache
> is sync if it is writeback type. So at qemu level, we can make sure
> no request is left behind, before qemu-nbd start. That API should
> also benefit 3rd party block snapshot solution, such as LVM2.
> 
> More:
> With this patch and previous qcow2 snapshot at block device level, I think
> export/import/backup solution around qcow2, is nearly complete at qemu's
> level. It can do similar things as backing chain but with better performance,
> Some small optimization place are left:
> 
> 1 compare two snapshot's data to get the diff with help of qcow2's L1/L2 table.
> 2 convertion between external snapshot and internal snapshot.
> 
> This series need following series applied first:
> [PATCH V5 0/8] add internal snapshot support at block device level
> http://lists.nongnu.org/archive/html/qemu-devel/2013-07/msg01831.html
> 
> Wenchao Xia (4):
>   1 snapshot: distinguish id and name in load_tmp
>   2 qemu-nbd: support internal snapshot export
>   3 qemu-nbd: add doc for internal snapshot export
>   4 qemu-iotests: add 057 internal snapshot export with qemu-nbd case
> 
>  block/qcow2-snapshot.c     |   16 +++++++-
>  block/qcow2.h              |    5 ++-
>  block/snapshot.c           |   37 ++++++++++++++++++-
>  include/block/block_int.h  |    4 ++-
>  include/block/snapshot.h   |    4 ++-
>  qemu-img.c                 |    7 +++-
>  qemu-nbd.c                 |   56 ++++++++++++++++++++++++++++-
>  qemu-nbd.texi              |    3 ++
>  tests/qemu-iotests/057     |   87 ++++++++++++++++++++++++++++++++++++++++++++
>  tests/qemu-iotests/057.out |   26 +++++++++++++
>  tests/qemu-iotests/group   |    1 +
>  11 files changed, 237 insertions(+), 9 deletions(-)
>  create mode 100755 tests/qemu-iotests/057
>  create mode 100644 tests/qemu-iotests/057.out
> 
> 
> 
> 
> 

-- 
Eric Blake   eblake redhat com    +1-919-301-3266
Libvirt virtualization library http://libvirt.org