Re: [Qemu-devel] [RESEND][PATCH v3] Generalize -machine command line option

[Alexander Graf]

On 25.07.2011, at 14:49, Richard W.M. Jones wrote:

> On Mon, Jul 25, 2011 at 07:47:51AM -0500, Anthony Liguori wrote:
>> On 07/25/2011 07:44 AM, Alexander Graf wrote:
>>> 
>>> On 25.07.2011, at 14:05, Peter Maydell wrote:
>>> 
>>>> On 25 July 2011 12:48, Peter Maydell<address@hidden>  wrote:
>>>>> For ARM you absolutely should not be relying on the default
>>>>> machine type (not least because it's an incredibly ancient
>>>>> dev board which nobody uses any more). An ARM kernel is
>>>>> generally fairly specific to the hardware platform being
>>>>> emulated, so you should know which machine you're intending
>>>>> to run on and specify it explicitly.
>>>> 
>>>> In fact having thought about it a bit I'm going to go further
>>>> and say that the whole idea of a "default machine" is a rather
>>>> x86-centric idea -- most architectures don't really have a
>>>> single machine type that's used by just about everybody,
>>>> always has been, and isn't likely to become obsolete in the
>>>> future. So if we're reworking the command line API to
>>>> supersede "-M" then we shouldn't have a default at all.
>>> 
>>> That's not exactly true. For PPC, everyone so far expects a Mac to pop up.
>> 
>> Except if you're running on an IBM Power box, then you definitely
>> expect a pseries guest to pop up.
>> 
>> We really need to enable the default config file (yes, we have a
>> default config file) can express the default machine.
> 
> +1 to this.
> 
> I was going to say there's missing information here, ie. if I had a
> Debian/arm kernel know, what machine should I use, but it looks like a
> config file would provide this missing information.

Well, what we really want is an image file format that also pulls along machine 
config files. Or just have 2 files for now - an image and a machine description 
config. VMs simply are more than just their hard disks ;).


Alex