Xen devel - Jul 2006 - [PATCH][XEND] Fix HVM Domain Creation When No Vifs Configured

Hi,

Currently - with the new qemu code, HVM domain creation will fail if a
vif isn''t configured. Xm-test will not pass sanity tests to run. The
new
qemu expects "-net none", otherwise it defaults to "-net nic -net
user". 

This patch simply adds "-net none" to the qemu-dm cmdline if no vifs
are
configured. 

Please apply this patch or something similar, xm-test is broken for HVM
testing without it.

Thanks,

Dan


_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel

Dan,
> Currently - with the new qemu code, HVM domain creation will fail if a
> vif isn''t configured. Xm-test will not pass sanity tests to run.
The new
> qemu expects "-net none", otherwise it defaults to "-net nic
-net
> user". 
> 
> This patch simply adds "-net none" to the qemu-dm cmdline if no
vifs are
> configured. 
> 
> Please apply this patch or something similar, xm-test is broken for HVM
> testing without it.
I''ve been meaning to look into this.  I think I''d like to take
a slightly
different approach:  change the default in qemu to be -net none.
I''ll get to it eventually, but if you want to look into providing a
patch,
that would be great!

    christian


_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel

On Thu, 2006-07-13 at 22:32 +0100, Christian Limpach
wrote:
> Dan,
> 
> > Currently - with the new qemu code, HVM domain creation will fail if a
> > vif isn''t configured. Xm-test will not pass sanity tests to
run. The new
> > qemu expects "-net none", otherwise it defaults to
"-net nic -net
> > user". 
> > 
> > This patch simply adds "-net none" to the qemu-dm cmdline if
no vifs are
> > configured. 
> > 
> > Please apply this patch or something similar, xm-test is broken for
HVM
> > testing without it.
> 
> I''ve been meaning to look into this.  I think I''d like to
take a slightly
> different approach:  change the default in qemu to be -net none.
> I''ll get to it eventually, but if you want to look into providing
a patch,
> that would be great!

I thought about that... but shouldn''t we keep the number of changes to
qemu to a minimum? Why support more and more specific xen changes to
qemu than necessary? 

Thanks,

Dan


_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel

On 13/7/06 10:46, "Daniel Stekloff" <dsteklof@us.ibm.com>
wrote:
> On Thu, 2006-07-13 at 22:32 +0100, Christian Limpach wrote:
>> Dan,
>> 
>>> Currently - with the new qemu code, HVM domain creation will fail
if a
>>> vif isn''t configured. Xm-test will not pass sanity tests
to run. The new
>>> qemu expects "-net none", otherwise it defaults to
"-net nic -net
>>> user". 
>>> 
>>> This patch simply adds "-net none" to the qemu-dm cmdline
if no vifs are
>>> configured. 
>>> 
>>> Please apply this patch or something similar, xm-test is broken for
HVM
>>> testing without it.
>> 
>> I''ve been meaning to look into this.  I think I''d
like to take a slightly
>> different approach:  change the default in qemu to be -net none.
>> I''ll get to it eventually, but if you want to look into
providing a patch,
>> that would be great!
> 
> 
> I thought about that... but shouldn''t we keep the number of
changes to
> qemu to a minimum? Why support more and more specific xen changes to
> qemu than necessary?
Sure, but this is a very simple and not very intrusive change.
I think it''s preferable to fix things at the source, in this case
it''s
clearly qemu having an unreasonable default for our usage model.

    christian


_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel

On Thu, 2006-07-13 at 23:04 +0100, Christian Limpach
wrote:
> On 13/7/06 10:46, "Daniel Stekloff" <dsteklof@us.ibm.com>
wrote:
> > On Thu, 2006-07-13 at 22:32 +0100, Christian Limpach wrote:
> >> Dan,
> >> 
> >>> Currently - with the new qemu code, HVM domain creation will
fail if a
> >>> vif isn''t configured. Xm-test will not pass sanity
tests to run. The new
> >>> qemu expects "-net none", otherwise it defaults to
"-net nic -net
> >>> user". 
> >>> 
> >>> This patch simply adds "-net none" to the qemu-dm
cmdline if no vifs are
> >>> configured. 
> >>> 
> >>> Please apply this patch or something similar, xm-test is
broken for HVM
> >>> testing without it.
> >> 
> >> I''ve been meaning to look into this.  I think
I''d like to take a slightly
> >> different approach:  change the default in qemu to be -net none.
> >> I''ll get to it eventually, but if you want to look into
providing a patch,
> >> that would be great!
> > 
> > 
> > I thought about that... but shouldn''t we keep the number of
changes to
> > qemu to a minimum? Why support more and more specific xen changes to
> > qemu than necessary?
> 
> Sure, but this is a very simple and not very intrusive change.
> I think it''s preferable to fix things at the source, in this case
it''s
> clearly qemu having an unreasonable default for our usage model.

Ok, how''s this.

Thanks,

Dan



_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel

On 7/13/06, Daniel Stekloff <dsteklof@us.ibm.com>
wrote:
> On Thu, 2006-07-13 at 23:04 +0100, Christian Limpach wrote:
> > On 13/7/06 10:46, "Daniel Stekloff"
<dsteklof@us.ibm.com> wrote:
> > > On Thu, 2006-07-13 at 22:32 +0100, Christian Limpach wrote:
> > >> Dan,
> > >>
> > >>> Currently - with the new qemu code, HVM domain creation
will fail if a
> > >>> vif isn''t configured. Xm-test will not pass
sanity tests to run. The new
> > >>> qemu expects "-net none", otherwise it defaults
to "-net nic -net
> > >>> user".
> > >>>
> > >>> This patch simply adds "-net none" to the
qemu-dm cmdline if no vifs are
> > >>> configured.
> > >>>
> > >>> Please apply this patch or something similar, xm-test is
broken for HVM
> > >>> testing without it.
> > >>
> > >> I''ve been meaning to look into this.  I think
I''d like to take a slightly
> > >> different approach:  change the default in qemu to be -net
none.
> > >> I''ll get to it eventually, but if you want to look
into providing a patch,
> > >> that would be great!
> > >
> > >
> > > I thought about that... but shouldn''t we keep the number
of changes to
> > > qemu to a minimum? Why support more and more specific xen changes
to
> > > qemu than necessary?
> >
> > Sure, but this is a very simple and not very intrusive change.
> > I think it''s preferable to fix things at the source, in this
case it''s
> > clearly qemu having an unreasonable default for our usage model.
>
>
> Ok, how''s this.
I''ve checked this in but #ifdef''ed the code away instead of
deleting it.

     christian

_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel