Re: DOM0_GETDOMAININFO intended behavior

[Christian Limpach <christian.limpach@gmail.com>]

On 6/4/05, Daniel Stekloff <dsteklof@us.ibm.com> wrote:
> On Saturday 04 June 2005 01:20, Ian Pratt wrote:
> > > :-)
> > >
> > > I asked the same question a week or two ago and yes it is
> > > intended. It provides for easy iteration. I think it is kind
> > > of a hackish interface as it isn't uniform with *all* the
> > > other DOM0 calls. However, it works and once you know to
> > > check the domid return value your code will work as expected.
> >
> > I suppose we could add a flag to indicate whether you were iterating or
> > wanted information on a specific domain. I think currently the only case
> > where we want to iterate is 'xm list'.
> 
> 
> If you're ok with changing the interface, wouldn't it be better for the
> hypercall to get and return just the requested domain info while you move the
> interation to libxc? The library could have two calls, one which calls the
> other: xc_domain_getinfo() and xc_domain_getinfo_list()

This doesn't scale well if your domain id space is sparse, which it
can easily become on a server which has many domains restarting
frequently.  This is the reason for the current behaviour.

I don't see what's wrong with the current behaviour anyway, how hard
is it to check if the call succeeded and then additionally check if
information for the requested domain was returned?  We could avoid the
additional check by having a different return code for when the call
succeeded but returned information for a different domain than the
requested one.

    christian