On Wed Jul 9, 2025 at 4:24 PM CEST, Jan Beulich wrote:
> On 08.07.2025 20:07, Alejandro Vallejo wrote:
>> --- a/xen/include/xen/bootfdt.h
>> +++ b/xen/include/xen/bootfdt.h
>> @@ -107,6 +107,10 @@ struct boot_domain {
>> struct boot_module *initrd;
>>
>> const char *cmdline;
>> +
>> +#if __has_include(<asm/bootfdt.h>)
>> + struct arch_boot_domain arch;
>> +#endif
>> };
>
> I fear I still don't follow this. In patch 1 you made domU a common-but-not-
> x86 field. Why can't domid, which is entirely generic, not similarly be a
> common-but-only-x86-for-now field? What is put in arch-specific structures
> should, in the common case at least, be truly (i.e. conceptually) be arch-
> specific imo.
>
> Jan
I tried to do the same with domU, but it's used by every other architecture so
it got silly adding the same code to 3 different arch-specific headers. I can
definitely move domid to the root structure under a similar ifdef.
Particularly when, as you very well point out, it's bound to be truly common
eventually.
Cheers,
Alejandro
