Re: [PATCH v3 04/39] x86/ibt: Base IBT bits

[Peter Zijlstra]

On Thu, Mar 03, 2022 at 12:23:25PM +0100, Peter Zijlstra wrote:

> --- a/arch/x86/Makefile
> +++ b/arch/x86/Makefile
> @@ -62,8 +62,11 @@ export BITS
>  #
>  KBUILD_CFLAGS += -mno-sse -mno-mmx -mno-sse2 -mno-3dnow -mno-avx
>  
> -# Intel CET isn't enabled in the kernel
> +ifeq ($(CONFIG_X86_KERNEL_IBT),y)
> +KBUILD_CFLAGS += $(call cc-option,-fcf-protection=branch)
> +else
>  KBUILD_CFLAGS += $(call cc-option,-fcf-protection=none)
> +endif
>  
>  ifeq ($(CONFIG_X86_32),y)
>          BITS := 32

Joao reported that RETPOLINE=n builds explode; turns out the compilers
default to using NOTRACK prefixes for jump-tables and we explicitly do
not enable that security compromise for the kernel.

Since the compilers don't have explicit control over NOTRACK generation,
blanket disable jump-tables when using IBT without RETPOLINE.

Joao will be submitting GCC and Clang bugreports on this shortly.

diff --git a/arch/x86/Makefile b/arch/x86/Makefile
index d38c18f4bd53..f80a425e7d29 100644
--- a/arch/x86/Makefile
+++ b/arch/x86/Makefile
@@ -63,7 +63,9 @@ export BITS
 KBUILD_CFLAGS += -mno-sse -mno-mmx -mno-sse2 -mno-3dnow -mno-avx
 
 ifeq ($(CONFIG_X86_KERNEL_IBT),y)
-KBUILD_CFLAGS += $(call cc-option,-fcf-protection=branch)
+# Explicitly disable jump-tables, also implied by RETPOLINE=y, for kernel IBT
+# to avoid NOTRACK prefixes.
+KBUILD_CFLAGS += $(call cc-option,-fcf-protection=branch -fno-jump-tables)
 else
 KBUILD_CFLAGS += $(call cc-option,-fcf-protection=none)
 endif