Re: [PATCH] EFI: Enable booting unified hypervisor/kernel/initrd images

[Jan Beulich <jbeulich@xxxxxxxx>]

On 07.09.2020 20:07, Julien Grall wrote:
> 
> 
> On 07/09/2020 08:11, Jan Beulich wrote:
>> On 04.09.2020 19:35, Julien Grall wrote:
>>> On 04/09/2020 11:06, Trammell Hudson wrote:
>>>> On Friday, September 4, 2020 5:29 AM, Julien Grall <julien@xxxxxxx> wrote:
>>>>
>>>>> On 28/08/2020 12:51, Trammell Hudson wrote:
>>>>>
>>>>>> -   /* PE32+ Subsystem type */
>>>>>>       +#if defined(ARM)
>>>>>>
>>>>>
>>>>> Shouldn't this be defined(aarch64) ?
>>>>
>>>> To be honest I'm not sure and don't have a way to check if
>>>> this code works on ARM.
>>>
>>> The code looks generic enough. I will give a try to build it.
>>>
>>>> Does an Xen EFI build on ARM even
>>>> use the PE32+ format?
>>>
>>> So far, we only support EFI and AArch64. This is built using PE32+ format.
>>
>> But this gets done by some custom means iirc, not by linking to that
>> format? (I'm not certain whether this matters here, so I'm sorry for
>> the noise if it doesn't.)
> 
> The header is written manually in .S but the final binary is looking 
> like a PE32+:
> 
> 42sh> file xen/xen 
> 
> xen/xen: MS-DOS executable, MZ for MS-DOS
> 
> As you asked the question, you must be concerned with this approach. Can 
> you outline what could be the issue?

My thinking merely is: If it's not linked "normally", maybe there are
further odd things? I have no specific concern.

Jan