Re: [PATCH 2/2] x86/paravirt: add extra clobbers with ZERO_CALL_USED_REGS enabled

[Nick Desaulniers]

On Sun, Sep 4, 2022 at 11:02 PM Bill Wendling <morbo@xxxxxxxxxx> wrote:
>
> On Sat, Sep 3, 2022 at 12:18 AM Kees Cook <keescook@xxxxxxxxxxxx> wrote:
> >
> > On Fri, Sep 02, 2022 at 09:37:50PM +0000, Bill Wendling wrote:
> > > [...]
> > >         callq   *pv_ops+536(%rip)
> >
> > Do you know which pv_ops function is this? I can't figure out where
> > pte_offset_kernel() gets converted into a pv_ops call....
> >
> This one is _paravirt_ident_64, I believe. I think that the original
> issue Nathan was seeing was with another seemingly innocuous function.

_paravirt_ident_64 is marked noinstr, which makes me suspect that it
really needs to not be touched at all by the compiler for
these...special features.

Maybe the definition of noinstr in include/linux/compiler_types.h
should be adding __attribute__((zero_call_used_regs(skip)))?

Untested:

```
diff --git a/include/linux/compiler_types.h b/include/linux/compiler_types.h
index 4f2a819fd60a..a51ab77e2da8 100644
--- a/include/linux/compiler_types.h
+++ b/include/linux/compiler_types.h
@@ -226,10 +226,17 @@ struct ftrace_likely_data {
 #define __no_sanitize_or_inline __always_inline
 #endif

+#ifdef CONFIG_ZERO_CALL_USED_REGS
+#define __no_zero_call_used_regs __attribute__((__zero_call_used_reg__(skip)))
+#else
+#define __no_zero_call_used_regs
+#endif
+
 /* Section for code which can't be instrumented at all */
 #define noinstr
         \
        noinline notrace __attribute((__section__(".noinstr.text")))    \
-       __no_kcsan __no_sanitize_address __no_profile __no_sanitize_coverage
+       __no_kcsan __no_sanitize_address __no_profile                   \
+       __no_sanitize_coverage __no_zero_call_used_regs

 #endif /* __KERNEL__ */
```
Or use __has_attribute in include/linux/compiler_attributes.h.
-- 
Thanks,
~Nick Desaulniers