Re: [Qemu-devel] [libvirt] [PATCH RFC 0/4] Allow hibernation on guests

[Michael Roth <mdroth@linux.vnet.ibm.com>]

On 01/30/2012 06:57 AM, Luiz Capitulino wrote:
> On Thu, 26 Jan 2012 16:57:01 -0600
> Anthony Liguori<anthony@codemonkey.ws>  wrote:
>
>> On 01/26/2012 01:35 PM, Luiz Capitulino wrote:
>>> On Thu, 26 Jan 2012 08:18:03 -0700
>>> Eric Blake<eblake@redhat.com>   wrote:
>>>
>>>> [adding qemu-devel]
>>>>
>>>> On 01/26/2012 07:46 AM, Daniel P. Berrange wrote:
>>>>>> One thing, that you'll probably notice is this
>>>>>> 'set-support-level' command. Basically, it tells GA what qemu version
>>>>>> is it running on. Ideally, this should be done as soon as
>>>>>> GA starts up. However, that cannot be determined from outside
>>>>>> world as GA doesn't emit any events yet.
>>>>>> Ideally^2 this command should be left out as it should be qemu
>>>>>> who tells its own agent this kind of information.
>>>>>> Anyway, I was going to call this command in qemuProcess{Startup,
>>>>>> Reconnect,Attach}, but it won't work. We need to un-pause guest CPUs
>>>>>> so guest can boot and start GA, but that implies returning from qemuProcess*.
>>>>>>
>>>>>> So I am setting this just before 'guest-suspend' command, as
>>>>>> there is one more thing about GA. It is unable to remember anything
>>>>>> upon its restart (GA process). Which has BTW show flaw
>>>>>> in our current code with FS freeze&   thaw. If we freeze guest
>>>>>> FS, and somebody restart GA, the simple FS Thaw will not succeed as
>>>>>> GA thinks FS are not frozen. But that's a different cup of tea.
>>>>>>
>>>>>> Because of what written above, we need to call set-level
>>>>>> on every suspend.
>>>>>
>>>>>
>>>>> IMHO all this says that the 'set-level' command is a conceptually
>>>>> unfixably broken design&   should be killed in QEMU before it turns
>>>>> into an even bigger mess.
>>>
>>> Can you elaborate on this? Michal and I talked on irc about making the
>>> compatibility level persistent, would that help?
>>>
>>>>> Once we're in a situation where we need to call 'set-level' prior
>>>>> to every single invocation, you might as well just allow the QEMU
>>>>> version number to be passed in directly as an arg to the command
>>>>> you are running directly thus avoiding this horrificness.
>>>>
>>>> Qemu folks, would you care to chime in on this?
>>>>
>>>> Exactly how is the set-level command supposed to work?  As I understand
>>>> it, the goal is that if the guest has qemu-ga 1.1 installed, but is
>>>> being run by qemu 1.0, then we want to ensure that any guest agent
>>>> command supported by qemu-ga 1.1 but requiring features of qemu not
>>>> present in qemu 1.0 will be properly rejected.
>>>
>>> Not exactly, the default support of qemu-ga is qemu 1.0. This means that by
>>> default qemu-ga will only support qemu 1.0 even when running on qemu 2.0. This
>>> way the set-support-level command allows you to specify that qemu 2.0 features
>>> are supported.
>>
>> Version numbers are meaningless.  What happens when a bunch of features get
>> backported by RHEL such that qemu-ga 1.0 ends up being a frankenstein version of
>> 2.0?
>>
>> The feature negotiation mechanism we have in QMP is the existence of a command.
>>    If we're in a position where we're trying to disable part of a command, it
>> simply means that we should have multiple commands such that we can just remove
>> the disabled part entirely.
>
> You may have a point that we shouldn't be using the version number for that,
> but just switching to multiple commands doesn't solve the fundamental problem.

Agreed, but the multiple commands isn't really the fix here, it's 
libvirt querying for the "wakeup" command that Gerd's patches add.

We implemented set-version-level as a way to let management tools 
obliviously report something simple, like the version information it 
parses from QEMU already, to let the guest agent Do The Right Thing 
without any deep insight into what it's host requirements are (which is 
also why we opted for versions over specific capabilities flags).

But we can't rely on version levels applying in all cases, a distro 
might backport s3 support for 1.0, modify the guest agent version level 
dependencies accordingly, and thus render the that agent incompatible 
with, say, RHEL.

So it was a naive approach on my part, and the issues the libvirt folks 
noted with not knowing if a reset occurred since the last 
set-support-level make it even less desirable (we could persist it as 
you suggested, and I am working on patches to persist fsfreeze state, 
but it's not worth trying to fix set-support-level).

And at least in this case we have an easy out: libvirt knows it can't 
resume a guest without the presence of a QMP command to do so, and that 
doesn't require any intimate knowledge of how the agent works, it's just 
common sense.

Breaking the suspend/hibernate stuff into multiple commands avoids the 
need for libvirt to special case based on specific parameters to the 
command. We can also imagine distro-specific implementations of the 
agent disabling s4 due to things like not having s4 patches for virtio 
in place, so it makes it easier (possible) to discover those situations 
as well. It may be that we opt for a single command for libvirt, but at 
least on the agent side it should be multiple discoverable ones.

Hopefully this doesn't set anyone back too much :( IMHO it's the right 
way to go though.

>
> The fundamental problem is that, S3 in current (and old) qemu has two known bugs:
>
>   1. The screen is left black after S3 (it's a bug in seabios)
>   2. QEMU resumes the guest immediately (Gerd posted patches to address this)
>
> We're going to address both issues in 1.1. However, if qemu-ga is installed in
> an old qemu and S3 is used, the bugs will be triggered.
>
> We need a way for qemu-ga to query qemu about the existence of a working S3
> support. The set-support-level solves that.
>
> Another option would be to disable (or enable) S3 by default in qemu-ga, and let
> the admin enable (or disable it) according to S3 support being fixed in qemu.
>