rebuilding object files from asm files
Here's a small kbuild quirk I noticed on 2.6.25rc2. I tried to rebuild the obj file of a assembler file (.S) with make path/to/file.o.
Normally when I do this for .o file only its dependencies and and the object file are built as I expect.
But with an assembler file (like arch/x86/kernel/head_64.o) kbuild starts to rebuild everything.
-Andi
Hi Andi.
Quoted 9 lines · Andi Kleen
Here's a small kbuild quirk I noticed on 2.6.25rc2. I tried to rebuild the obj file of a assembler file (.S) with make
path/to/file.o.Normally when I do this for .o file only its dependencies and and the object file are built as I expect.
But with an assembler file (like
arch/x86/kernel/head_64.o) kbuild starts to rebuild everything.
Took a look at this.
You will notice that there is a difference in the behaviour betweet arch/x86/kernel/entry_64.o and head_64.o from same dir. For entry_64.o it works as expected.
The 'problem' is that kbuild when deciding to use the rule:
%.o: %.S prepare scripts FORCE
$(Q)$(MAKE) $(build)=$(build-dir) $(target-dir)$(notdir $@)
Already have done most of the work and mainly is has defined the dependencies for vmlinux.
As head_64.o is listed in $(head-y) kbuild do as told in following rule:
Note: vmlinux-init is equal to $(head-y) + $(init-y)
# The actual objects are generated when descending,
# make sure no implicit rule kicks in $(sort $(vmlinux-init) $(vmlinux-main)) $(vmlinux-lds): $(vmlinux-dirs) ;
So building head_64.o requires a visit of all directories listed in vmlinux-dirs before it can build head_64.o.
I do not see any clean way to fix this :-(
Sam
Quoted 5 lines
So building head_64.o requires a visit of all directories listed in vmlinux-dirs before it can build head_64.o.
I do not see any clean way to fix this :-(
I see. Thanks for the investigation. If it only affects head_64.S I guess it is not so bad.
-Andi
Quoted 8 lines · Andi Kleen
Quoted 5 lines
So building head_64.o requires a visit of all directories listed in vmlinux-dirs before it can build head_64.o.
I do not see any clean way to fix this :-(
I see. Thanks for the investigation. If it only affects
head_64.SI guess it is not so bad.
All of:
head-y := arch/x86/kernel/head_$(BITS).o
head-$(CONFIG_X86_64) += arch/x86/kernel/head64.o
head-y += arch/x86/kernel/init_task.o
But thats limited anyway.
Sam