Re: [tip:x86/pti] Revert "x86/speculation: Simplify indirect_branch_prediction_barrier()"

[Peter Zijlstra]

On Tue, Feb 13, 2018 at 12:58:21AM -0800, tip-bot for David Woodhouse wrote:

> --- a/arch/x86/include/asm/nospec-branch.h
> +++ b/arch/x86/include/asm/nospec-branch.h
> @@ -164,10 +164,15 @@ static inline void vmexit_fill_RSB(void)
>  
>  static inline void indirect_branch_prediction_barrier(void)
>  {
> +	asm volatile(ALTERNATIVE("",
> +				 "movl %[msr], %%ecx\n\t"
> +				 "movl %[val], %%eax\n\t"
> +				 "movl $0, %%edx\n\t"
> +				 "wrmsr",
> +				 X86_FEATURE_USE_IBPB)
> +		     : : [msr] "i" (MSR_IA32_PRED_CMD),
> +			 [val] "i" (PRED_CMD_IBPB)
> +		     : "eax", "ecx", "edx", "memory");
>  }

Joe Konno pointed out that we now need the below line too, because we're
using MSR_IA32_PRED_CMD in this header.

With the existing code that's not a problem per-se, but my objtool
retpoline annotation things did do stumble over this.

Do we want to fold it into the objtool annotation patch or have it
separate?

---
 arch/x86/include/asm/nospec-branch.h | 1 +
 1 file changed, 1 insertion(+)

diff --git a/arch/x86/include/asm/nospec-branch.h b/arch/x86/include/asm/nospec-branch.h
index 300cc159b4a0..76b058533e47 100644
--- a/arch/x86/include/asm/nospec-branch.h
+++ b/arch/x86/include/asm/nospec-branch.h
@@ -6,6 +6,7 @@
 #include <asm/alternative.h>
 #include <asm/alternative-asm.h>
 #include <asm/cpufeatures.h>
+#include <asm/msr-index.h>
 
 #ifdef __ASSEMBLY__