Re: [PATCH v2 2/2] efi/libstub: refactor Makefile to not use lib-y syntax

[Ard Biesheuvel <ardb@kernel.org>]

On Thu, 4 Jun 2020 at 04:20, Masahiro Yamada <masahiroy@kernel.org> wrote:
>
> Documentation/kbuild/makefiles.rst says:
>
>   Use of lib-y is normally restricted to `lib/` and `arch/*/lib`.
>
> This is because lib-y is inteded to be hooked to KBUILD_VMLINUX_LIBS,
> which is passed down to scripts/link-vmlinux.sh.
>
> Besides, lib-y is not so interesting because objects from lib-y are
> mostly linked in normal usecases. For example, lib-y only saves 364
> bytes for x86_64_defconfig. You can see the details in commit
> 7273ad2b08f8 ("kbuild: link lib-y objects to vmlinux forcibly when
> CONFIG_MODULES=y").
>
> I think we should consider to deprecate lib-y syntax at some point
> because we should aim for better solution like dead code elimination
> or LTO.
>
> Other than lib/ and arch/*/lib, this Makefile is the only user of
> lib-y. Replace lib-y with a custom rule.
>
> Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>

Series queued in efi/urgent.

Thanks

> ---
>
> Changes in v2:
>   - Add more description
>
>  drivers/firmware/efi/libstub/Makefile | 49 +++++++++++++++------------
>  1 file changed, 27 insertions(+), 22 deletions(-)
>
> diff --git a/drivers/firmware/efi/libstub/Makefile b/drivers/firmware/efi/libstub/Makefile
> index cce4a7436052..7d81dc45cadf 100644
> --- a/drivers/firmware/efi/libstub/Makefile
> +++ b/drivers/firmware/efi/libstub/Makefile
> @@ -44,7 +44,7 @@ OBJECT_FILES_NON_STANDARD     := y
>  # Prevents link failures: __sanitizer_cov_trace_pc() is not linked in.
>  KCOV_INSTRUMENT                        := n
>
> -lib-y                          := efi-stub-helper.o gop.o secureboot.o tpm.o \
> +stub-obj-y                     := efi-stub-helper.o gop.o secureboot.o tpm.o \
>                                    file.o mem.o random.o randomalloc.o pci.o \
>                                    skip_spaces.o lib-cmdline.o lib-ctype.o \
>                                    alignedmem.o relocate.o vsprintf.o
> @@ -55,15 +55,19 @@ efi-deps-y := fdt_rw.c fdt_ro.c fdt_wip.c fdt.c fdt_empty_tree.c fdt_sw.c
>  $(obj)/lib-%.o: $(srctree)/lib/%.c FORCE
>         $(call if_changed_rule,cc_o_c)
>
> -lib-$(CONFIG_EFI_GENERIC_STUB) += efi-stub.o fdt.o string.o \
> +stub-obj-$(CONFIG_EFI_GENERIC_STUB)    += efi-stub.o fdt.o string.o \
>                                    $(patsubst %.c,lib-%.o,$(efi-deps-y))
>
> -lib-$(CONFIG_ARM)              += arm32-stub.o
> -lib-$(CONFIG_ARM64)            += arm64-stub.o
> -lib-$(CONFIG_X86)              += x86-stub.o
> +stub-obj-$(CONFIG_ARM)         += arm32-stub.o
> +stub-obj-$(CONFIG_ARM64)       += arm64-stub.o
> +stub-obj-$(CONFIG_X86)         += x86-stub.o
>  CFLAGS_arm32-stub.o            := -DTEXT_OFFSET=$(TEXT_OFFSET)
>  CFLAGS_arm64-stub.o            := -DTEXT_OFFSET=$(TEXT_OFFSET)
>
> +targets                                += $(stub-obj-y)
> +stub-obj-y                     := $(patsubst %.o,%.stub.o, $(stub-obj-y))
> +targets                                += $(stub-obj-y)
> +
>  #
>  # For x86, bootloaders like systemd-boot or grub-efi do not zero-initialize the
>  # .bss section, so the .bss section of the EFI stub needs to be included in the
> @@ -83,23 +87,6 @@ STUBCOPY_FLAGS-$(CONFIG_ARM) += --rename-section .data=.data.efistub \
>                                    --rename-section .bss=.bss.efistub,load,alloc
>  STUBCOPY_RELOC-$(CONFIG_ARM)   := R_ARM_ABS
>
> -#
> -# arm64 puts the stub in the kernel proper, which will unnecessarily retain all
> -# code indefinitely unless it is annotated as __init/__initdata/__initconst etc.
> -# So let's apply the __init annotations at the section level, by prefixing
> -# the section names directly. This will ensure that even all the inline string
> -# literals are covered.
> -# The fact that the stub and the kernel proper are essentially the same binary
> -# also means that we need to be extra careful to make sure that the stub does
> -# not rely on any absolute symbol references, considering that the virtual
> -# kernel mapping that the linker uses is not active yet when the stub is
> -# executing. So build all C dependencies of the EFI stub into libstub, and do
> -# a verification pass to see if any absolute relocations exist in any of the
> -# object files.
> -#
> -extra-y                                := $(lib-y)
> -lib-y                          := $(patsubst %.o,%.stub.o,$(lib-y))
> -
>  STUBCOPY_FLAGS-$(CONFIG_ARM64) += --prefix-alloc-sections=.init \
>                                    --prefix-symbols=__efistub_
>  STUBCOPY_RELOC-$(CONFIG_ARM64) := R_AARCH64_ABS
> @@ -121,3 +108,21 @@ quiet_cmd_stubcopy = STUBCPY $@
>                 /bin/false;                                             \
>         fi;                                                             \
>         $(OBJCOPY) $(STUBCOPY_FLAGS-y) $< $@
> +
> +# arm64 puts the stub in the kernel proper, which will unnecessarily retain all
> +# code indefinitely unless it is annotated as __init/__initdata/__initconst etc.
> +# So let's apply the __init annotations at the section level, by prefixing
> +# the section names directly. This will ensure that even all the inline string
> +# literals are covered.
> +# The fact that the stub and the kernel proper are essentially the same binary
> +# also means that we need to be extra careful to make sure that the stub does
> +# not rely on any absolute symbol references, considering that the virtual
> +# kernel mapping that the linker uses is not active yet when the stub is
> +# executing. So build all C dependencies of the EFI stub into libstub, and do
> +# a verification pass to see if any absolute relocations exist in any of the
> +# object files.
> +#
> +extra-y += lib.a
> +
> +$(obj)/lib.a: $(addprefix $(obj)/, $(stub-obj-y)) FORCE
> +       $(call if_changed,ar)
> --
> 2.25.1
>