Re: [PATCH] symbols: explicitly specify source file name for symtab

[Andrew Cooper <andrew.cooper3@xxxxxxxxxx>]

On 11/05/2026 11:00 am, Jan Beulich wrote:
> If there are any local symbols in an object file, GNU ld will create an
> STT_FILE symbol derived from the object file name if there is none in the
> incoming symbol table. The object file name, however, varies between
> linking passes. As a result, symbol name compression can yield different
> results if any of those local symbols need retaining (Arm [and RISC-V]
> mapping symbols are omitted, for example). If that difference in
> compression would yield a difference in the sizes of symbol_names[] or
> symbols_token_table[], the compare-symbol-tables sanity check will fail.
>
> Fixes: d37d63d4b548 ("symbols: prefix static symbols with their source file 
> names")
> Reported-by: Oleksii Kurochko <oleksii.kurochko@xxxxxxxxx>
> Signed-off-by: Jan Beulich <jbeulich@xxxxxxxx>
> ---
> The observed problem was with a stub generated as Arm64 erratum 843419
> workaround. Such stubs' symbols (imo wrongly) are associated with the last
> input object, rather than the input object they belong to. Also for other
> kinds of stubs, afaict. See
> https://sourceware.org/bugzilla/show_bug.cgi?id=34140.
>
> As per the above, having a Fixes: tag here is questionable.
>
> --- a/xen/arch/x86/Makefile
> +++ b/xen/arch/x86/Makefile
> @@ -200,7 +200,8 @@ $(TARGET).efi: $(objtree)/prelink.o $(no
>  ifeq ($(CONFIG_DEBUG_INFO),y)
>       $(if $(filter --strip-debug,$(EFI_LDFLAGS)),echo,:) "Will strip debug 
> info from $(@F)"
>  endif
> -     $(objtree)/tools/symbols $(all_symbols) --empty > $(dot-target).0s.S
> +     $(objtree)/tools/symbols $(all_symbols) --source-name=$(@F).S --empty \
> +             > $(dot-target).0s.S
>       $(MAKE) $(build)=$(@D) .$(@F).0s.o
>       $(foreach base, $(VIRT_BASE) $(ALT_BASE), \
>                 $(LD) $(call EFI_LDFLAGS,$(base)) -T $(obj)/efi.lds $< 
> $(relocs-dummy) \
> @@ -210,6 +211,7 @@ endif
>               > $(dot-target).1r.S
>       $(NM) -pa --format=sysv $(dot-target).$(VIRT_BASE).0 \
>               | $(objtree)/tools/symbols $(all_symbols) --sysv --sort \
> +                  --source-name=$(@F).S \
>               > $(dot-target).1s.S
>       $(MAKE) $(build)=$(@D) .$(@F).1r.o .$(@F).1s.o
>       $(foreach base, $(VIRT_BASE) $(ALT_BASE), \
> @@ -220,6 +222,7 @@ endif
>               > $(dot-target).2r.S
>       $(NM) -pa --format=sysv $(dot-target).$(VIRT_BASE).1 \
>               | $(objtree)/tools/symbols $(all_symbols) --sysv --sort \
> +                  --source-name=$(@F).S \
>               > $(dot-target).2s.S
>       $(MAKE) $(build)=$(@D) .$(@F).2r.o .$(@F).2s.o
>       $(call compare-symbol-tables, $(dot-target).1r.o, $(dot-target).2r.o)
> --- a/xen/tools/symbols.c
> +++ b/xen/tools/symbols.c
> @@ -66,6 +66,7 @@ int token_profit[0x10000];
>  unsigned char best_table[256][2];
>  unsigned char best_table_len[256];
>  
> +static const char *srcname = "xen-syms.S";
>  
>  static void usage(void)
>  {
> @@ -356,6 +357,7 @@ static void write_src(void)
>       printf("#define ALGN 4\n");
>       printf("#endif\n");
>  
> +     printf("\t.file \"%s\"\n", srcname);
>       printf("\t.section .rodata, \"a\"\n");
>  
>       printf("#ifndef SYMBOLS_ORIGIN\n");
> @@ -679,6 +681,8 @@ int main(int argc, char **argv)
>                               unsorted = true;
>                       else if (strcmp(argv[i], "--sort-by-name") == 0)
>                               sort_by_name = 1;
> +                     else if (strncmp(argv[i], "--source-name=", 14) == 0)
> +                             srcname = argv[i] + 14;
>                       else if (strcmp(argv[i], "--warn-dup") == 0)
>                               warn_dup = true;
>                       else if (strcmp(argv[i], "--error-dup") == 0)

Why does x86 need to plumb the source name in, but the other
architectures don't?

xen-syms.S suffices for both x86 builds AFAICT, so can't it just be
unconditional?

~Andrew