Re: [Qemu-devel] [RFC][PATCH 00/14 v9] introducing a new, dedicated memory dump mechanism

[Eric Blake <eblake@redhat.com>]

[-- Attachment #1: Type: text/plain, Size: 1628 bytes --]

On 03/14/2012 11:26 AM, Luiz Capitulino wrote:
> On Wed, 14 Mar 2012 10:03:15 +0800
> Wen Congyang <wency@cn.fujitsu.com> wrote:
> 
>> Changes from v8 to v9:
>> 1. remove async support(it will be reimplemented after QAPI async commands support
>>    is finished)
> 
> I gave my review on this one (concentrating on the QMP part only), and one
> important aspect of this command is that it's a long synchronous operation.
> 
> As it runs with vCPUs stopped, the only drawback is that libvirt won't be able
> to run other commands in parallel and won't be able to cancel it either. Also
> note that this command is most likely to be executed when the guest crashes.
> 
> Given all this, it's fine with me to have this as a synchronous command, but
> I'd like to get an ACK from libvirt and another one from Jan and/or Anthony.

The inability to cancel is annoying (which is why I first suggested that
we add async handling and cancel back in v2 or so); but your arguments that:

1. async support will be coming in qemu 1.2, and polluting the waters
with ad-hoc async stuff in 1.1 will just make it harder
2. dump typically only gets run when a guest will no longer be running,
therefore having it be blocking is not the end of the world, as long as
the user doesn't mind the inability to cancel

are sufficient that I'm okay with a synchronous-only version for qemu
1.1.  It won't be the first time libvirt has had a long-running command
it couldn't cancel (think 'savevm', for instance).

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


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 620 bytes --]