On Thu, 13 Mar 2025 at 03:09, Masahiro Yamada <masahiroy@xxxxxxxxxx> wrote:
>
> On Tue, Mar 11, 2025 at 8:06 PM Ard Biesheuvel <ardb+git@xxxxxxxxxx> wrote:
> >
> > From: Ard Biesheuvel <ardb@xxxxxxxxxx>
> >
> > Instead of generating the vmlinux.relocs file (needed by the
> > decompressor build to construct the KASLR relocation tables) as a
> > vmlinux postlink step, which is dubious because it depends on data that
> > is stripped from vmlinux before the build completes, generate it from
> > vmlinux.unstripped, which has been introduced specifically for this
> > purpose.
> >
> > This ensures that each artifact is rebuilt as needed, rather than as a
> > side effect of another build rule.
> >
> > This effectively reverts commit
> >
> > 9d9173e9ceb6 ("x86/build: Avoid relocation information in final vmlinux")
> >
> > Signed-off-by: Ard Biesheuvel <ardb@xxxxxxxxxx>
> > ---
> > arch/x86/Makefile.postlink | 38 --------------------
> > arch/x86/boot/compressed/Makefile | 9 +++--
> > 2 files changed, 6 insertions(+), 41 deletions(-)
> >
> > diff --git a/arch/x86/Makefile.postlink b/arch/x86/Makefile.postlink
> > deleted file mode 100644
> > index 445fce66630f..000000000000
> > --- a/arch/x86/Makefile.postlink
> > +++ /dev/null
> > @@ -1,38 +0,0 @@
> > -# SPDX-License-Identifier: GPL-2.0
> > -# ===========================================================================
> > -# Post-link x86 pass
> > -# ===========================================================================
> > -#
> > -# 1. Separate relocations from vmlinux into vmlinux.relocs.
> > -# 2. Strip relocations from vmlinux.
> > -
> > -PHONY := __archpost
> > -__archpost:
> > -
> > --include include/config/auto.conf
> > -include $(srctree)/scripts/Kbuild.include
> > -
> > -CMD_RELOCS = arch/x86/tools/relocs
> > -OUT_RELOCS = arch/x86/boot/compressed
> > -quiet_cmd_relocs = RELOCS $(OUT_RELOCS)/vmlinux.relocs
> > - cmd_relocs = \
> > - mkdir -p $(OUT_RELOCS); \
> > - $(CMD_RELOCS) $@ > $(OUT_RELOCS)/vmlinux.relocs; \
> > - $(CMD_RELOCS) --abs-relocs $@
> > -
> > -# `@true` prevents complaint when there is nothing to be done
> > -
> > -vmlinux vmlinux.unstripped: FORCE
> > - @true
> > -ifeq ($(CONFIG_X86_NEED_RELOCS),y)
> > - $(call cmd,relocs)
> > -endif
> > -
> > -clean:
> > - @rm -f $(OUT_RELOCS)/vmlinux.relocs
> > -
> > -PHONY += FORCE clean
> > -
> > -FORCE:
> > -
> > -.PHONY: $(PHONY)
> > diff --git a/arch/x86/boot/compressed/Makefile b/arch/x86/boot/compressed/Makefile
> > index 606c74f27459..5edee7a9786c 100644
> > --- a/arch/x86/boot/compressed/Makefile
> > +++ b/arch/x86/boot/compressed/Makefile
> > @@ -117,9 +117,12 @@ $(obj)/vmlinux.bin: vmlinux FORCE
> >
> > targets += $(patsubst $(obj)/%,%,$(vmlinux-objs-y)) vmlinux.bin.all vmlinux.relocs
> >
> > -# vmlinux.relocs is created by the vmlinux postlink step.
> > -$(obj)/vmlinux.relocs: vmlinux
> > - @true
> > +CMD_RELOCS = arch/x86/tools/relocs
> > +quiet_cmd_relocs = RELOCS $@
> > + cmd_relocs = $(CMD_RELOCS) $< > $@;$(CMD_RELOCS) --abs-relocs $<
> > +
> > +$(obj)/vmlinux.relocs: vmlinux.unstripped FORCE
> > + $(call if_changed,relocs)
>
> Perhaps, it may make sense to rebuild vmlinux.relocs
> when arch/x86/tools/relocs is changed, but
> I do not see such dependency in the other
> arch/x86/realmode/rm/Makefile.
>
> https://github.com/torvalds/linux/blob/v6.14-rc5/arch/x86/realmode/rm/Makefile#L61
>
>
> So, I decided it is ok.
> If you mind, you can send v3.
>
No I think that's fine: this patch reverts the changes to
arch/x86/boot/compressed/Makefile; if we need to improve it, it should
be a separate change in any case.
