On 10/12/2025 8:08 am, Jan Beulich wrote: > On 09.12.2025 22:47, Jason Andryuk wrote: >> --- a/xen/include/xen/xen.lds.h >> +++ b/xen/include/xen/xen.lds.h >> @@ -173,6 +173,12 @@ >> _edevice = .; \ >> } :text >> >> +#define SCHEDULER_ARRAY \ >> + . = ALIGN(8); \ > While indeed it was 8 in all original locations, I question that for Arm32 > (and a possible future RV32, for example); imo it wants to be ... > >> + __start_schedulers_array = .; \ >> + *(.data.schedulers) \ >> + __end_schedulers_array = .; >> + >> #ifdef CONFIG_HYPFS >> #define HYPFS_PARAM \ >> . = ALIGN(POINTER_ALIGN); \ > ... exactly like this. Preferably with that change (happy to carry out while > committing, alongside a respective addition to the description, so long as > there's agreement): > Reviewed-by: Jan Beulich <[email protected]>
I thought the same. struct scheduler is entirely pointers, and one unsigned int. I'm pretty sure that this "array" predates the introduction of POINTER_ALIGN. So yes, with it converted to POINTER_ALIGN, Reviewed-by: Andrew Cooper <[email protected]>
