Hi Jan, > On 25 Nov 2024, at 16:21, Jan Beulich <jbeul...@suse.com> wrote: > > On 19.11.2024 09:58, Luca Fancellu wrote: >> --- a/xen/include/xen/vmap.h >> +++ b/xen/include/xen/vmap.h >> @@ -5,12 +5,19 @@ >> * purpose area (VMAP_DEFAULT) and a livepatch-specific area (VMAP_XEN). The >> * latter is used when loading livepatches and the former for everything >> else. >> */ >> -#if !defined(__XEN_VMAP_H__) && defined(VMAP_VIRT_START) >> +#ifndef __XEN_VMAP_H__ >> #define __XEN_VMAP_H__ >> >> #include <xen/mm-frame.h> >> #include <xen/page-size.h> >> >> +/* >> + * MPU systems won't have HAS_VMAP enabled, but will provide implementation >> + * only for some of the functions of this module. So hide the definition for >> + * some of these function to systems where !HAS_VMAP >> + */ >> +#ifdef CONFIG_HAS_VMAP > > What you're hiding are declarations, not definitions.
yes, I realised the mistake after sending > While this may feel like > splitting hair, the question really is: Do the declarations actually need > hiding? IOW won't it suffice to have the definitions unavailable? While this > would mean that wrong uses are flagged only when linking, we do such all the > time when we expect e.g. DCE to remove actual uses of respective identifiers. I misunderstood your comment on the previous version and I thought your preference was to hide the declarations. I’ll try without hiding them and I’ll send the change soon. Cheers, Luca
