[PATCH v2 1/2] retpolines: Only enable retpoline support when compiler support it

[PATCH v2 1/2] retpolines: Only enable retpoline support when compiler support it

[Zhenzhong Duan]

Since retpoline capable compilers are widely available, make
CONFIG_RETPOLINE hard depend on it.

The check of RETPOLINE is changed to CONFIG_RETPOLINE.

This change is based on suggestion in https://lkml.org/lkml/2018/9/18/1016

Signed-off-by: Zhenzhong Duan <zhenzhong.duan@oracle.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Borislav Petkov <bp@suse.de>
Cc: Daniel Borkmann <daniel@iogearbox.net>
Cc: David Woodhouse <dwmw@amazon.co.uk>
Cc: H. Peter Anvin <hpa@zytor.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
Cc: Andy Lutomirski <luto@kernel.org>
Cc: Masahiro Yamada <yamada.masahiro@socionext.com>
Cc: Michal Marek <michal.lkml@markovi.net>
---
 arch/x86/Kconfig                     |  6 ++----
 arch/x86/Makefile                    |  4 +---
 arch/x86/include/asm/nospec-branch.h | 10 ++++++----
 arch/x86/kernel/cpu/bugs.c           |  2 +-
 scripts/Makefile.build               |  2 --
 5 files changed, 10 insertions(+), 14 deletions(-)

diff --git a/arch/x86/Kconfig b/arch/x86/Kconfig
index ffebfc3f..277e0df 100644
--- a/arch/x86/Kconfig
+++ b/arch/x86/Kconfig
@@ -436,6 +436,8 @@ config GOLDFISH
 
 config RETPOLINE
 	bool "Avoid speculative indirect branches in kernel"
+	depends on $(cc-option,-mindirect-branch=thunk-extern -mindirect-branch-register) || \
+                    $(cc-option,-mretpoline-external-thunk)
 	default y
 	select STACK_VALIDATION if HAVE_STACK_VALIDATION
 	help
@@ -444,10 +446,6 @@ config RETPOLINE
 	  branches. Requires a compiler with -mindirect-branch=thunk-extern
 	  support for full protection. The kernel may run slower.
 
-	  Without compiler support, at least indirect branches in assembler
-	  code are eliminated. Since this includes the syscall entry path,
-	  it is not entirely pointless.
-
 config INTEL_RDT
 	bool "Intel Resource Director Technology support"
 	depends on X86 && CPU_SUP_INTEL
diff --git a/arch/x86/Makefile b/arch/x86/Makefile
index 5b562e4..30573dd 100644
--- a/arch/x86/Makefile
+++ b/arch/x86/Makefile
@@ -222,9 +222,7 @@ KBUILD_CFLAGS += -fno-asynchronous-unwind-tables
 
 # Avoid indirect branches in kernel to deal with Spectre
 ifdef CONFIG_RETPOLINE
-ifneq ($(RETPOLINE_CFLAGS),)
-  KBUILD_CFLAGS += $(RETPOLINE_CFLAGS) -DRETPOLINE
-endif
+  KBUILD_CFLAGS += $(RETPOLINE_CFLAGS)
 endif
 
 archscripts: scripts_basic
diff --git a/arch/x86/include/asm/nospec-branch.h b/arch/x86/include/asm/nospec-branch.h
index 80dc144..8b09cbb 100644
--- a/arch/x86/include/asm/nospec-branch.h
+++ b/arch/x86/include/asm/nospec-branch.h
@@ -162,11 +162,12 @@
 	_ASM_PTR " 999b\n\t"					\
 	".popsection\n\t"
 
-#if defined(CONFIG_X86_64) && defined(RETPOLINE)
+#ifdef CONFIG_RETPOLINE
+#ifdef CONFIG_X86_64
 
 /*
- * Since the inline asm uses the %V modifier which is only in newer GCC,
- * the 64-bit one is dependent on RETPOLINE not CONFIG_RETPOLINE.
+ * Inline asm uses the %V modifier which is only in newer GCC
+ * which is ensured when CONFIG_RETPOLINE is defined.
  */
 # define CALL_NOSPEC						\
 	ANNOTATE_NOSPEC_ALTERNATIVE				\
@@ -181,7 +182,7 @@
 	X86_FEATURE_RETPOLINE_AMD)
 # define THUNK_TARGET(addr) [thunk_target] "r" (addr)
 
-#elif defined(CONFIG_X86_32) && defined(CONFIG_RETPOLINE)
+#else /* CONFIG_X86_32 */
 /*
  * For i386 we use the original ret-equivalent retpoline, because
  * otherwise we'll run out of registers. We don't care about CET
@@ -211,6 +212,7 @@
 	X86_FEATURE_RETPOLINE_AMD)
 
 # define THUNK_TARGET(addr) [thunk_target] "rm" (addr)
+#endif
 #else /* No retpoline for C / inline asm */
 # define CALL_NOSPEC "call *%[thunk_target]\n"
 # define THUNK_TARGET(addr) [thunk_target] "rm" (addr)
diff --git a/arch/x86/kernel/cpu/bugs.c b/arch/x86/kernel/cpu/bugs.c
index c37e66e..d0108fb 100644
--- a/arch/x86/kernel/cpu/bugs.c
+++ b/arch/x86/kernel/cpu/bugs.c
@@ -252,7 +252,7 @@ static void __init spec2_print_if_secure(const char *reason)
 
 static inline bool retp_compiler(void)
 {
-	return __is_defined(RETPOLINE);
+	return __is_defined(CONFIG_RETPOLINE);
 }
 
 static inline bool match_option(const char *arg, int arglen, const char *opt)
diff --git a/scripts/Makefile.build b/scripts/Makefile.build
index a8e7ba9..6a6be9f 100644
--- a/scripts/Makefile.build
+++ b/scripts/Makefile.build
@@ -236,10 +236,8 @@ ifdef CONFIG_GCOV_KERNEL
 objtool_args += --no-unreachable
 endif
 ifdef CONFIG_RETPOLINE
-ifneq ($(RETPOLINE_CFLAGS),)
   objtool_args += --retpoline
 endif
-endif
 
 
 ifdef CONFIG_MODVERSIONS
-- 
1.8.3.1

Re: [PATCH v2 1/2] retpolines: Only enable retpoline support when compiler support it

[Ingo Molnar]

* Zhenzhong Duan <zhenzhong.duan@oracle.com> wrote:

> Since retpoline capable compilers are widely available, make
> CONFIG_RETPOLINE hard depend on it.
> 
> The check of RETPOLINE is changed to CONFIG_RETPOLINE.
> 
> This change is based on suggestion in https://lkml.org/lkml/2018/9/18/1016
>
> Signed-off-by: Zhenzhong Duan <zhenzhong.duan@oracle.com>
> Cc: Thomas Gleixner <tglx@linutronix.de>
> Cc: Peter Zijlstra <peterz@infradead.org>

Please turn such 'based on suggestions' into proper tags as well, i.e. 
something like:

  Suggested-by: David Woodhouse <dwmw2@infradead.org>

>  config RETPOLINE
>  	bool "Avoid speculative indirect branches in kernel"
> +	depends on $(cc-option,-mindirect-branch=thunk-extern -mindirect-branch-register) || \
> +                    $(cc-option,-mretpoline-external-thunk)

At least a comment should be added that this is the retpoline feature 
check for GCC and Clang.

Also, whitespace damage plus the two options should align vertically.

Thanks,

	Ingo

Re: [PATCH v2 1/2] retpolines: Only enable retpoline support when compiler support it

[Thomas Gleixner]

On Thu, 1 Nov 2018, Woodhouse, David wrote:
> On Thu, 2018-11-01 at 10:50 +0100, Ingo Molnar wrote:
> > * Zhenzhong Duan <zhenzhong.duan@oracle.com> wrote:
> > 
> > > Since retpoline capable compilers are widely available, make
> > > CONFIG_RETPOLINE hard depend on it.
> > > 
> > > The check of RETPOLINE is changed to CONFIG_RETPOLINE.
> > > 
> > > This change is based on suggestion in https://lkml.org/lkml/2018/9/18/1016
> > > 
> > > Signed-off-by: Zhenzhong Duan <zhenzhong.duan@oracle.com>
> > > Cc: Thomas Gleixner <tglx@linutronix.de>
> > > Cc: Peter Zijlstra <peterz@infradead.org>
> > 
> > Please turn such 'based on suggestions' into proper tags as well, i.e. 
> > something like:
> > 
> >   Suggested-by: David Woodhouse <dwmw2@infradead.org>
> 
> I think the suggestion came from PeterZ; I just acked it.
> 
> Although on furthe reflection, I think I'd prefer a build break if
> retpoline is enabled in the kernel config and the compiler doesn't
> support it. This patch would make it silently fail to be secure.

Agreed.

Thanks,

	tglx

Re: [PATCH v2 1/2] retpolines: Only enable retpoline support when compiler support it

[Ingo Molnar]

* Thomas Gleixner <tglx@linutronix.de> wrote:

> On Thu, 1 Nov 2018, Woodhouse, David wrote:
> > On Thu, 2018-11-01 at 10:50 +0100, Ingo Molnar wrote:
> > > * Zhenzhong Duan <zhenzhong.duan@oracle.com> wrote:
> > > 
> > > > Since retpoline capable compilers are widely available, make
> > > > CONFIG_RETPOLINE hard depend on it.
> > > > 
> > > > The check of RETPOLINE is changed to CONFIG_RETPOLINE.
> > > > 
> > > > This change is based on suggestion in https://lkml.org/lkml/2018/9/18/1016
> > > > 
> > > > Signed-off-by: Zhenzhong Duan <zhenzhong.duan@oracle.com>
> > > > Cc: Thomas Gleixner <tglx@linutronix.de>
> > > > Cc: Peter Zijlstra <peterz@infradead.org>
> > > 
> > > Please turn such 'based on suggestions' into proper tags as well, i.e. 
> > > something like:
> > > 
> > >   Suggested-by: David Woodhouse <dwmw2@infradead.org>
> > 
> > I think the suggestion came from PeterZ; I just acked it.
> > 
> > Although on furthe reflection, I think I'd prefer a build break if
> > retpoline is enabled in the kernel config and the compiler doesn't
> > support it. This patch would make it silently fail to be secure.
> 
> Agreed.

Yeah, I agree that that's the best policy: if someone wants retpoline 
support it shouldn't silently turn off just because the wrong toolchain 
was used to build the kernel ...

Thanks,

	Ingo