On 26.09.2024 11:15, oleksii.kuroc...@gmail.com wrote: > On Thu, 2024-09-26 at 08:23 +0200, Jan Beulich wrote: >> On 25.09.2024 18:08, oleksii.kuroc...@gmail.com wrote: >>> On Wed, 2024-09-25 at 10:36 +0200, Jan Beulich wrote: >>>> PPC's desire to use DECL_SECTION() can certainly be covered by >>>> providing >>>> a (trivial) DECL_SECTION() also for Arm and RISC-V. Seeing that >>>> even >>>> x86 >>>> overrides the default to the trivial form for building xen.efi, >>>> I'm >>>> inclined to suggest we should actually have a way for an arch to >>>> indicate >>>> to xen.lds.h that it wants just the trivial form (avoiding a >>>> later >>>> #undef). >>> If to go with what I suggested before then x86 will look like: >>> >>> diff --git a/xen/arch/x86/xen.lds.S b/xen/arch/x86/xen.lds.S >>> index d48de67cfd..911585541e 100644 >>> --- a/xen/arch/x86/xen.lds.S >>> +++ b/xen/arch/x86/xen.lds.S >>> @@ -3,6 +3,10 @@ >>> >>> #include <xen/cache.h> >>> #include <xen/lib.h> >>> + >>> +#ifdef EFI >>> +#define SIMPLE_DECL_SECTION >>> +#endif >>> #include <xen/xen.lds.h> >>> #include <asm/page.h> >>> #undef ENTRY >>> @@ -12,9 +16,7 @@ >>> >>> #define FORMAT "pei-x86-64" >>> #undef __XEN_VIRT_START >>> -#undef DECL_SECTION >>> #define __XEN_VIRT_START __image_base__ >>> -#define DECL_SECTION(x) x : >>> >>> ENTRY(efi_start) >>> >>> diff --git a/xen/include/xen/xen.lds.h b/xen/include/xen/xen.lds.h >>> index a17810bb28..fb11ba7357 100644 >>> --- a/xen/include/xen/xen.lds.h >>> +++ b/xen/include/xen/xen.lds.h >>> @@ -5,6 +5,8 @@ >>> * Common macros to be used in architecture specific linker >>> scripts. >>> */ >>> >>> +#ifdef SIMPLE_DECL_SECTION >> >> #ifndef I guess? >> >>> @@ -15,6 +17,10 @@ >>> # define DECL_SECTION(x) x : AT(ADDR(x) - __XEN_VIRT_START) >>> #endif >>> >>> +#else /* SIMPLE_DECL_SECION */ >>> +# define DECL_SECTION(x) x : >>> +#endif >>> + >>> /* >>> * To avoid any confusion, please note that the EFI macro does not >>> correspond >>> * to EFI support and is used when linking a native EFI (i.e. >>> PE/COFF) >>> binary, >>> >>> Does it make sense? Or it would be better to follow way for each >>> architecture: >>> #undef DECL_SECTION >>> #define DECL_SECTION(x) x : >> >> Hard to tell which one's better; I was asking myself that same >> question >> when writing an earlier reply. I'm slightly in favor of the form you >> have >> now. > Do you mean moving only a content of section without secname and laddr > ( in case of x86 and PPC ), and alignment to xen.lds.h ?
No. I was solely referring to the DECL_SECTION() aspect. Jan
