[PATCH 7/6] x86: __always_inline __{rd,wr}msr()

[Peter Zijlstra]

On Thu, Jan 07, 2021 at 10:36:43AM +0100, Peter Zijlstra wrote:
> On Wed, Jan 06, 2021 at 09:59:17AM -0800, Randy Dunlap wrote:
> > On 1/6/21 6:36 AM, Peter Zijlstra wrote:
> > > When the compiler fails to inline; we violate nonisntr:
> > > 
> > >   vmlinux.o: warning: objtool: __sev_es_nmi_complete()+0xc7: call to sev_es_wr_ghcb_msr() leaves .noinstr.text section
> > 
> > I am still seeing (a variant of) this one:
> > 
> > vmlinux.o: warning: objtool: __sev_es_nmi_complete()+0xce: call to __wrmsr.constprop.14() leaves .noinstr.text section
> 
> Gah, sorry, I managed to mess up my .config :/ /me goes try again.

OK, restored your original .config and that did indeed reproduce, the
below cures it.

---

Subject: x86: __always_inline __{rd,wr}msr()
From: Peter Zijlstra <peterz@xxxxxxxxxxxxx>
Date: Thu Jan  7 11:14:25 CET 2021

When the compiler choses to not inline the trivial MSR helpers:

  vmlinux.o: warning: objtool: __sev_es_nmi_complete()+0xce: call to __wrmsr.constprop.14() leaves .noinstr.text section

Reported-by: Randy Dunlap <rdunlap@xxxxxxxxxxxxx>
Signed-off-by: Peter Zijlstra (Intel) <peterz@xxxxxxxxxxxxx>
---
 arch/x86/include/asm/msr.h |    4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

--- a/arch/x86/include/asm/msr.h
+++ b/arch/x86/include/asm/msr.h
@@ -86,7 +86,7 @@ static inline void do_trace_rdpmc(unsign
  * think of extending them - you will be slapped with a stinking trout or a frozen
  * shark will reach you, wherever you are! You've been warned.
  */
-static inline unsigned long long notrace __rdmsr(unsigned int msr)
+static __always_inline unsigned long long __rdmsr(unsigned int msr)
 {
 	DECLARE_ARGS(val, low, high);
 
@@ -98,7 +98,7 @@ static inline unsigned long long notrace
 	return EAX_EDX_VAL(val, low, high);
 }
 
-static inline void notrace __wrmsr(unsigned int msr, u32 low, u32 high)
+static __always_inline void __wrmsr(unsigned int msr, u32 low, u32 high)
 {
 	asm volatile("1: wrmsr\n"
 		     "2:\n"