On Mon, Jun 10, 2019 at 01:33:57PM -0500, Josh Poimboeuf wrote: > On Wed, Jun 05, 2019 at 03:08:06PM +0200, Peter Zijlstra wrote: > > --- a/arch/x86/include/asm/static_call.h > > +++ b/arch/x86/include/asm/static_call.h > > @@ -2,6 +2,20 @@ > > #ifndef _ASM_STATIC_CALL_H > > #define _ASM_STATIC_CALL_H > > > > +#include <asm/asm-offsets.h> > > + > > +#ifdef CONFIG_HAVE_STATIC_CALL_INLINE > > + > > +/* > > + * This trampoline is only used during boot / module init, so it's safe to > > use > > + * the indirect branch without a retpoline. > > + */ > > +#define __ARCH_STATIC_CALL_TRAMP_JMP(key, func) > > \ > > + ANNOTATE_RETPOLINE_SAFE \ > > + "jmpq *" __stringify(key) "+" __stringify(SC_KEY_func) "(%rip) \n" > > + > > +#else /* !CONFIG_HAVE_STATIC_CALL_INLINE */ > > I wonder if we can simplify this (and drop the indirect branch) by > getting rid of the above cruft, and instead just use the out-of-line > trampoline as the default for inline as well. > > Then the inline case could fall back to the out-of-line implementation > (by patching the trampoline's jmp dest) before static_call_initialized > is set.
I think I've got that covered. I changed arch_static_call_transform() to (always) first rewrite the trampoline and then the in-line site. That way, when early/module crud comes in that still points at the trampoline, it will jump to the right place. I've not even compiled yet, but it ought to work ;-)
