Re: [Qemu-devel] Re: [libvirt] Re: [PATCH 2/3] Introduce monitor 'wait' command

[Anthony Liguori]

Jamie Lokier wrote:
> Anthony Liguori wrote:
>   
> > It doesn't.  When an app enables events, we would start queuing them, 
> > but if it didn't consume them in a timely manner (or at all), we would 
> > start leaking memory badly.
> >
> > We want to be robust even in the face of poorly written management 
> > apps/scripts so we need some expiration function too.
> >     
>
> What happens when an app stops reading the monitor channel for a
> little while, and there's enough monitor output to fill TCP buffers or
> terminal buffers?  Does it block QEMU?  Does QEMU drop arbitrary bytes
> from the stream, corrupting the output syntax?
>   

Depends on the type of character device.  They all have different 
properties in this regard.  Basically, you're stuck in a losing 
proposition.  Either you drop output, buffer memory indefinitely, or put 
the application to sleep.  Different character devices make different 
trade offs.

> If you send events only to the monitor which requests them, then you
> could say that they are sent immediately to that monitor, and if the
> app stops reading the monitor, whatever normally happens when it stops
> reading happens to these events.
>
> In other words, no need for arbitrary expiration time.  Makes it
> determinstic at least.
>   

You're basically saying that if something isn't connected, drop them.  
If it is connected, do a monitor_printf() such that you're never queuing 
events.  Entirely reasonable and I've considered it.

However, I do like the idea though of QEMU queuing events for a certain 
period of time.  Not everyone always has something connected to a 
monitor.  I may notice that my NFS server (which runs in a VM) is not 
responding, VNC to the system, switch to the monitor, and take a look at 
the event log.  If I can get the past 10 minutes of events, I may see 
something useful like a host IO failure.

> >  Monitor "sessions" are ill-defined 
> > though b/c of things like tcp:// reconnects so I wouldn't want to do that.
> >     
>
> Oh dear.  Is defining it insurmountable?
>
> Why can't each TCP (re)connection be a new monitor?
>   

You get a notification on reconnect but not on disconnect.  Basically 
CharDriverState is not designed around a connect model.  The fact that 
it has any notion of reconnect today is really a big hack.

CharDriverState could definitely use a rewrite.  It hasn't aged well at all.

Regards,

Anthony Liguori