On 15.08.2025 22:41, Andrew Cooper wrote:
> Right now they're inline in {read,write}_gs_shadow(), but we're going to need
> to use these elsewhere to support FRED.
But why "kern"? We're not dealing with GS in kernel / user terms, but in
real / shadow ones. I'm also not quite happy with the double leading
underscores, fwiw.
As to using them elsewhere to support FRED: How can we when they use SWAPGS,
which isn't available under FRED?
> --- a/xen/arch/x86/include/asm/fsgsbase.h
> +++ b/xen/arch/x86/include/asm/fsgsbase.h
> @@ -32,6 +32,17 @@ static inline unsigned long __rdgsbase(void)
> return base;
> }
>
> +static inline unsigned long __rdgskern(void)
> +{
> + unsigned long base;
> +
> + asm_inline volatile ( "swapgs\n\t"
> + "rdgsbase %0\n\t"
> + "swapgs" : "=r" (base) );
Again strictly speaking "=&r", if already you open-code rdgsbase() now.
Jan
