Re: [Qemu-devel] [PATCH 1/1] s390: IPL device for s390

[Anthony Liguori <aliguori@us.ibm.com>]

On 05/08/2012 07:32 AM, Christian Borntraeger wrote:
> On 04/05/12 20:12, Alexander Graf wrote:
>>
>> On 04.05.2012, at 16:02, Christian Borntraeger wrote:
>>
>>> On 04/05/12 16:00, Christian Borntraeger wrote:
>>>>>> An IPL (booting) on s390 of SCSI disks is done by a firmware component.
>>>>>> Lets implement this scheme as an qemu device that also allows to
>>>>>> configure the IPL like the HMC. We have a parameter iplid that
>>>>>> refers to a disk device and a load parm that specifies the entry
>>>>>> on the disk to be ipled. We also provide a default device
>>>>>> if no -device s390-ipl statement is given.
>>>>>
>>>>> Any reason we can't do this in guest firmware code?
>>>>
>>>> Conceptually guest firmware does not exist in the guest address space
>>>> on s390. It is separate in a storage area called HSA.
>>>> (you could say the existing hardware is semi-hosted, you cant buy it bare
>>>> metal.
>>>> Doing the boot code in guest address space will fail if the guest firmware
>>>> address  collides with the addresses specified by a bootmap.
>>>
>>> Or in other words, this code is closer to the real s390 boxes.
>>
>> Yeah, I see the point. I'd really like to get Anthony's comments on this one first though.
>
> Right.
>
> Anthony, this is the prototype of the IPL device that we have talked about some weeks
> ago. Is an external device to do the IPL process for s390 still ok with you?

I don't really understand the point about collision.  If you chain load 
carefully, you ought to be able to relocate or something like that I would imagine.

But at any rate, for these semi-hosted platforms, I don't see a lot better 
choices than doing these things as devices since there's no "real hardware" to 
model.

So doing this as a device and limiting it to s390 seems like the right strategy 
to me.

Regards,

Anthony Liguori

>
> Christian
>