[PATCH] x86/build: fix compiler support check for CONFIG_RETPOLINE

[PATCH] x86/build: fix compiler support check for CONFIG_RETPOLINE

[Masahiro Yamada]

It is wrong to add CONFIG option diagnostic to the Makefile parse
stage.

Once you are hit by the error about non-retpoline compiler, the
compilation still breaks even after disabling CONFIG_RETPOLINE.

The easiest fix is to move this check to the "archprepare" like commit
829fe4aa9ac1 ("x86: Allow generating user-space headers without a
compiler") did.

Link: https://lkml.org/lkml/2018/12/4/206
Reported-by: Meelis Roos <mroos@linux.ee>
Fixes: 4cd24de3a098 ("x86/retpoline: Make CONFIG_RETPOLINE depend on compiler support")
Signed-off-by: Masahiro Yamada <yamada.masahiro@socionext.com>
---

 arch/x86/Makefile | 14 ++++++++------
 1 file changed, 8 insertions(+), 6 deletions(-)

diff --git a/arch/x86/Makefile b/arch/x86/Makefile
index f5d7f41..ae0148b 100644
--- a/arch/x86/Makefile
+++ b/arch/x86/Makefile
@@ -219,12 +219,7 @@ KBUILD_CFLAGS += -Wno-sign-compare
 KBUILD_CFLAGS += -fno-asynchronous-unwind-tables
 
 # Avoid indirect branches in kernel to deal with Spectre
-ifdef CONFIG_RETPOLINE
-ifeq ($(RETPOLINE_CFLAGS),)
-  $(error You are building kernel with non-retpoline compiler, please update your compiler.)
-endif
-  KBUILD_CFLAGS += $(RETPOLINE_CFLAGS)
-endif
+KBUILD_CFLAGS += $(RETPOLINE_CFLAGS)
 
 archscripts: scripts_basic
 	$(Q)$(MAKE) $(build)=arch/x86/tools relocs
@@ -307,6 +302,13 @@ ifndef CC_HAVE_ASM_GOTO
 	@echo Compiler lacks asm-goto support.
 	@exit 1
 endif
+ifdef CONFIG_RETPOLINE
+ifeq ($(RETPOLINE_CFLAGS),)
+	@echo "You are building kernel with non-retpoline compiler." >&2
+	@echo "Please update your compiler." >&2
+	@false
+endif
+endif
 
 archclean:
 	$(Q)rm -rf $(objtree)/arch/i386
-- 
2.7.4

Re: [PATCH] x86/build: fix compiler support check for CONFIG_RETPOLINE

[Zhenzhong Duan]

Hi Masahiro,


Thanks for fixing.
There is a question still confusing me. There are many CONFIG_* used in 
arch/x86/Makefile.
Will they also be impacted or not?

# grep CONFIG_ arch/x86/Makefile
ifeq ($(CONFIG_X86_32),y)
         cflags-$(CONFIG_MK8) += $(call cc-option,-march=k8)
         cflags-$(CONFIG_MPSC) += $(call cc-option,-march=nocona)
  ......
drivers-$(CONFIG_MATH_EMULATION) += arch/x86/math-emu/
drivers-$(CONFIG_PCI)      += arch/x86/pci/
drivers-$(CONFIG_OPROFILE) += arch/x86/oprofile/
drivers-$(CONFIG_PM) += arch/x86/power/
drivers-$(CONFIG_FB) += arch/x86/video/
ifeq ($(CONFIG_X86_DECODER_SELFTEST),y)

Thanks
Zhenzhong
On 2018/12/5 11:00, Masahiro Yamada wrote:
> It is wrong to add CONFIG option diagnostic to the Makefile parse
> stage.
>
> Once you are hit by the error about non-retpoline compiler, the
> compilation still breaks even after disabling CONFIG_RETPOLINE.
>
> The easiest fix is to move this check to the "archprepare" like commit
> 829fe4aa9ac1 ("x86: Allow generating user-space headers without a
> compiler") did.
>
> Link: https://lkml.org/lkml/2018/12/4/206
> Reported-by: Meelis Roos <mroos@linux.ee>
> Fixes: 4cd24de3a098 ("x86/retpoline: Make CONFIG_RETPOLINE depend on compiler support")
> Signed-off-by: Masahiro Yamada <yamada.masahiro@socionext.com>
> ---
>
>   arch/x86/Makefile | 14 ++++++++------
>   1 file changed, 8 insertions(+), 6 deletions(-)
>
> diff --git a/arch/x86/Makefile b/arch/x86/Makefile
> index f5d7f41..ae0148b 100644
> --- a/arch/x86/Makefile
> +++ b/arch/x86/Makefile
> @@ -219,12 +219,7 @@ KBUILD_CFLAGS += -Wno-sign-compare
>   KBUILD_CFLAGS += -fno-asynchronous-unwind-tables
>   
>   # Avoid indirect branches in kernel to deal with Spectre
> -ifdef CONFIG_RETPOLINE
> -ifeq ($(RETPOLINE_CFLAGS),)
> -  $(error You are building kernel with non-retpoline compiler, please update your compiler.)
> -endif
> -  KBUILD_CFLAGS += $(RETPOLINE_CFLAGS)
> -endif
> +KBUILD_CFLAGS += $(RETPOLINE_CFLAGS)
>   
>   archscripts: scripts_basic
>   	$(Q)$(MAKE) $(build)=arch/x86/tools relocs
> @@ -307,6 +302,13 @@ ifndef CC_HAVE_ASM_GOTO
>   	@echo Compiler lacks asm-goto support.
>   	@exit 1
>   endif
> +ifdef CONFIG_RETPOLINE
> +ifeq ($(RETPOLINE_CFLAGS),)
> +	@echo "You are building kernel with non-retpoline compiler." >&2
> +	@echo "Please update your compiler." >&2
> +	@false
> +endif
> +endif
>   
>   archclean:
>   	$(Q)rm -rf $(objtree)/arch/i386

Re: [PATCH] x86/build: fix compiler support check for CONFIG_RETPOLINE

[Masahiro Yamada]

On Wed, Dec 5, 2018 at 12:23 PM Zhenzhong Duan
<zhenzhong.duan@oracle.com> wrote:
>
> Hi Masahiro,
>
>
> Thanks for fixing.
> There is a question still confusing me. There are many CONFIG_* used in
> arch/x86/Makefile.
> Will they also be impacted or not?


The top level Makefile (and arch/*/Makefile) is getting huge.
Parsing the whole Makefile everytime means,
it computes variables that may or may not be used.

$(call cc-option,...) is a bit costly,
but other parts are unnoticeable
in terms of performance.


I am addressing the use of $(error ...) / $(warning ...)
in the non-recipe context since they could break the build
for false positive reasons.




> # grep CONFIG_ arch/x86/Makefile
> ifeq ($(CONFIG_X86_32),y)
>          cflags-$(CONFIG_MK8) += $(call cc-option,-march=k8)
>          cflags-$(CONFIG_MPSC) += $(call cc-option,-march=nocona)
>   ......
> drivers-$(CONFIG_MATH_EMULATION) += arch/x86/math-emu/
> drivers-$(CONFIG_PCI)      += arch/x86/pci/
> drivers-$(CONFIG_OPROFILE) += arch/x86/oprofile/
> drivers-$(CONFIG_PM) += arch/x86/power/
> drivers-$(CONFIG_FB) += arch/x86/video/
> ifeq ($(CONFIG_X86_DECODER_SELFTEST),y)
>
> Thanks
> Zhenzhong
> On 2018/12/5 11:00, Masahiro Yamada wrote:
> > It is wrong to add CONFIG option diagnostic to the Makefile parse
> > stage.
> >
> > Once you are hit by the error about non-retpoline compiler, the
> > compilation still breaks even after disabling CONFIG_RETPOLINE.
> >
> > The easiest fix is to move this check to the "archprepare" like commit
> > 829fe4aa9ac1 ("x86: Allow generating user-space headers without a
> > compiler") did.
> >
> > Link: https://lkml.org/lkml/2018/12/4/206
> > Reported-by: Meelis Roos <mroos@linux.ee>
> > Fixes: 4cd24de3a098 ("x86/retpoline: Make CONFIG_RETPOLINE depend on compiler support")
> > Signed-off-by: Masahiro Yamada <yamada.masahiro@socionext.com>
> > ---
> >
> >   arch/x86/Makefile | 14 ++++++++------
> >   1 file changed, 8 insertions(+), 6 deletions(-)
> >
> > diff --git a/arch/x86/Makefile b/arch/x86/Makefile
> > index f5d7f41..ae0148b 100644
> > --- a/arch/x86/Makefile
> > +++ b/arch/x86/Makefile
> > @@ -219,12 +219,7 @@ KBUILD_CFLAGS += -Wno-sign-compare
> >   KBUILD_CFLAGS += -fno-asynchronous-unwind-tables
> >
> >   # Avoid indirect branches in kernel to deal with Spectre
> > -ifdef CONFIG_RETPOLINE
> > -ifeq ($(RETPOLINE_CFLAGS),)
> > -  $(error You are building kernel with non-retpoline compiler, please update your compiler.)
> > -endif
> > -  KBUILD_CFLAGS += $(RETPOLINE_CFLAGS)
> > -endif
> > +KBUILD_CFLAGS += $(RETPOLINE_CFLAGS)
> >
> >   archscripts: scripts_basic
> >       $(Q)$(MAKE) $(build)=arch/x86/tools relocs
> > @@ -307,6 +302,13 @@ ifndef CC_HAVE_ASM_GOTO
> >       @echo Compiler lacks asm-goto support.
> >       @exit 1
> >   endif
> > +ifdef CONFIG_RETPOLINE
> > +ifeq ($(RETPOLINE_CFLAGS),)
> > +     @echo "You are building kernel with non-retpoline compiler." >&2
> > +     @echo "Please update your compiler." >&2
> > +     @false
> > +endif
> > +endif
> >
> >   archclean:
> >       $(Q)rm -rf $(objtree)/arch/i386



--
Best Regards
Masahiro Yamada

Re: [PATCH] x86/build: fix compiler support check for CONFIG_RETPOLINE

[Zhenzhong Duan]

On 2018/12/5 11:00, Masahiro Yamada wrote:
> It is wrong to add CONFIG option diagnostic to the Makefile parse
> stage.
>
> Once you are hit by the error about non-retpoline compiler, the
> compilation still breaks even after disabling CONFIG_RETPOLINE.
>
> The easiest fix is to move this check to the "archprepare" like commit
> 829fe4aa9ac1 ("x86: Allow generating user-space headers without a
> compiler") did.
>
> Link: https://lkml.org/lkml/2018/12/4/206
> Reported-by: Meelis Roos <mroos@linux.ee>
> Fixes: 4cd24de3a098 ("x86/retpoline: Make CONFIG_RETPOLINE depend on compiler support")
> Signed-off-by: Masahiro Yamada <yamada.masahiro@socionext.com>
> ---
>
>   arch/x86/Makefile | 14 ++++++++------
>   1 file changed, 8 insertions(+), 6 deletions(-)
>
> diff --git a/arch/x86/Makefile b/arch/x86/Makefile
> index f5d7f41..ae0148b 100644
> --- a/arch/x86/Makefile
> +++ b/arch/x86/Makefile
> @@ -219,12 +219,7 @@ KBUILD_CFLAGS += -Wno-sign-compare
>   KBUILD_CFLAGS += -fno-asynchronous-unwind-tables
>   
>   # Avoid indirect branches in kernel to deal with Spectre
> -ifdef CONFIG_RETPOLINE
> -ifeq ($(RETPOLINE_CFLAGS),)
> -  $(error You are building kernel with non-retpoline compiler, please update your compiler.)
> -endif
> -  KBUILD_CFLAGS += $(RETPOLINE_CFLAGS)
> -endif
> +KBUILD_CFLAGS += $(RETPOLINE_CFLAGS)

Is it better to also move "# Avoid indirect branches in kernel to deal 
with Spectre"

and "KBUILD_CFLAGS += $(RETPOLINE_CFLAGS)"?

Seems unconditionaly using "KBUILD_CFLAGS += $(RETPOLINE_CFLAGS)" will 
have compiler

using "__x86_indirect_thunk_\reg" call even ifCONFIG_RETPOLINE is 
disabled. I guess

link process will fail?

Thanks

Zhenzhong

>   
>   archscripts: scripts_basic
>   	$(Q)$(MAKE) $(build)=arch/x86/tools relocs
> @@ -307,6 +302,13 @@ ifndef CC_HAVE_ASM_GOTO
>   	@echo Compiler lacks asm-goto support.
>   	@exit 1
>   endif
> +ifdef CONFIG_RETPOLINE
> +ifeq ($(RETPOLINE_CFLAGS),)
> +	@echo "You are building kernel with non-retpoline compiler." >&2
> +	@echo "Please update your compiler." >&2
> +	@false
> +endif
> +endif
>   
>   archclean:
>   	$(Q)rm -rf $(objtree)/arch/i386

Re: [PATCH] x86/build: fix compiler support check for CONFIG_RETPOLINE

[Masahiro Yamada]

On Wed, Dec 5, 2018 at 3:08 PM Zhenzhong Duan <zhenzhong.duan@oracle.com> wrote:
>
>
> On 2018/12/5 11:00, Masahiro Yamada wrote:
> > It is wrong to add CONFIG option diagnostic to the Makefile parse
> > stage.
> >
> > Once you are hit by the error about non-retpoline compiler, the
> > compilation still breaks even after disabling CONFIG_RETPOLINE.
> >
> > The easiest fix is to move this check to the "archprepare" like commit
> > 829fe4aa9ac1 ("x86: Allow generating user-space headers without a
> > compiler") did.
> >
> > Link: https://lkml.org/lkml/2018/12/4/206
> > Reported-by: Meelis Roos <mroos@linux.ee>
> > Fixes: 4cd24de3a098 ("x86/retpoline: Make CONFIG_RETPOLINE depend on compiler support")
> > Signed-off-by: Masahiro Yamada <yamada.masahiro@socionext.com>
> > ---
> >
> >   arch/x86/Makefile | 14 ++++++++------
> >   1 file changed, 8 insertions(+), 6 deletions(-)
> >
> > diff --git a/arch/x86/Makefile b/arch/x86/Makefile
> > index f5d7f41..ae0148b 100644
> > --- a/arch/x86/Makefile
> > +++ b/arch/x86/Makefile
> > @@ -219,12 +219,7 @@ KBUILD_CFLAGS += -Wno-sign-compare
> >   KBUILD_CFLAGS += -fno-asynchronous-unwind-tables
> >
> >   # Avoid indirect branches in kernel to deal with Spectre
> > -ifdef CONFIG_RETPOLINE
> > -ifeq ($(RETPOLINE_CFLAGS),)
> > -  $(error You are building kernel with non-retpoline compiler, please update your compiler.)
> > -endif
> > -  KBUILD_CFLAGS += $(RETPOLINE_CFLAGS)
> > -endif
> > +KBUILD_CFLAGS += $(RETPOLINE_CFLAGS)
>
> Is it better to also move "# Avoid indirect branches in kernel to deal
> with Spectre"
>
> and "KBUILD_CFLAGS += $(RETPOLINE_CFLAGS)"?
>
> Seems unconditionaly using "KBUILD_CFLAGS += $(RETPOLINE_CFLAGS)" will
> have compiler
>
> using "__x86_indirect_thunk_\reg" call even ifCONFIG_RETPOLINE is
> disabled. I guess
>
> link process will fail?


Good catch.

I will send v2.

Thanks.



> Thanks
>
> Zhenzhong
>
> >
> >   archscripts: scripts_basic
> >       $(Q)$(MAKE) $(build)=arch/x86/tools relocs
> > @@ -307,6 +302,13 @@ ifndef CC_HAVE_ASM_GOTO
> >       @echo Compiler lacks asm-goto support.
> >       @exit 1
> >   endif
> > +ifdef CONFIG_RETPOLINE
> > +ifeq ($(RETPOLINE_CFLAGS),)
> > +     @echo "You are building kernel with non-retpoline compiler." >&2
> > +     @echo "Please update your compiler." >&2
> > +     @false
> > +endif
> > +endif
> >
> >   archclean:
> >       $(Q)rm -rf $(objtree)/arch/i386



-- 
Best Regards
Masahiro Yamada