Re: [PATCH 3/4] build: also check for empty .bss.* in .o -> .init.o conversion

[Jan Beulich <jbeulich@xxxxxxxx>]

On 06.08.2020 18:16, Andrew Cooper wrote:
> On 06/08/2020 10:05, Jan Beulich wrote:
>> We're gaining such sections, and like .text.* and .data.* they shouldn't
>> be present in objects subject to automatic to-init conversion. Oddly
>> enough for quite some time we did have an instance breaking this rule,
>> which gets fixed at this occasion, by breaking out the EFI boot
>> allocator functions into its own translation unit.
>>
>> Fixes: c5b9805bc1f7 ("efi: create new early memory allocator")
>> Signed-off-by: Jan Beulich <jbeulich@xxxxxxxx>
> 
> Acked-by: Andrew Cooper <andrew.cooper3@xxxxxxxxxx>

Thanks.

>> ---
>> This likely has a (weak) dependency on "x86/EFI: sanitize build logic"
>> sent several weeks ago, due to the new source file added, as explicit
>> dependencies upon the individual objects in x86/Makefile go away there.
>>
>> --- a/xen/Makefile
>> +++ b/xen/Makefile
>> @@ -355,7 +355,7 @@ $(TARGET): delete-unfresh-files
>>      $(MAKE) -C tools
>>      $(MAKE) -f $(BASEDIR)/Rules.mk include/xen/compile.h
>>      [ -e include/asm ] || ln -sf asm-$(TARGET_ARCH) include/asm
>> -    [ -e arch/$(TARGET_ARCH)/efi ] && for f in boot.c runtime.c compat.c 
>> efi.h;\
>> +    [ -e arch/$(TARGET_ARCH)/efi ] && for f in $$(cd common/efi; echo 
>> *.[ch]); \
>>              do test -r arch/$(TARGET_ARCH)/efi/$$f || \
>>                 ln -nsf ../../../common/efi/$$f arch/$(TARGET_ARCH)/efi/; \
>>              done; \
> 
> Maybe not for this patch, but we need to start removing this (and other)
> symlinking in the tree for proper out-of-tree builds to work.

Yes, but indeed not right here.

> AFAICT, this logic predates both Kconfig and x86's blur into having EFI
> support in xen.gz.

Yes, it was a result of making parts of that code also usable by Arm64.

> Can't we remove all of this by having CONFIG_XEN_PE expressed/selectable
> properly in Kconfig, and gathering all the objects normally, rather than
> bodging all of common/efi/ through arch/efi/ ?

_If_ we settle on Kconfig to be allowed to check compiler (and linker)
features, then yes. This continues to be a pending topic though, so
the switch can't be made like this at this point in time. (It could be
made a Kconfig item now - which, when enabled, implies the assertion
that a capable tool chain is in use.)

Jan