Re: [PATCH v3 09/12] x86: Derive XEN_MSR_PAT from its individual entries

[Jan Beulich <jbeulich@xxxxxxxx>]

On 15.12.2022 00:11, Demi Marie Obenour wrote:
> --- a/xen/arch/x86/include/asm/processor.h
> +++ b/xen/arch/x86/include/asm/processor.h
> @@ -96,7 +96,15 @@
>   * Host IA32_CR_PAT value to cover all memory types.  This is not the default
>   * MSR_PAT value, and is an ABI with PV guests.
>   */
> -#define XEN_MSR_PAT _AC(0x050100070406, ULL)
> +#define XEN_MSR_PAT (X86_MT_WB  << 0x00 | \
> +                     X86_MT_WT  << 0x08 | \
> +                     X86_MT_UCM << 0x10 | \
> +                     X86_MT_UC  << 0x18 | \
> +                     X86_MT_WC  << 0x20 | \
> +                     X86_MT_WP  << 0x28 | \
> +                     X86_MT_UC  << 0x30 | \
> +                     X86_MT_UC  << 0x38 | \

This wants properly parenthesizing (the shifts against the ors), and
at least the last four constants also look to need casting to uint64_t;
in fact I'm surprised that ...

> +                     0)

(interposed: I don't think this really does any good.)

> --- a/xen/arch/x86/mm.c
> +++ b/xen/arch/x86/mm.c
> @@ -6355,6 +6355,12 @@ unsigned long get_upper_mfn_bound(void)
>      return min(max_mfn, 1UL << (paddr_bits - PAGE_SHIFT)) - 1;
>  }
>  
> +static void __init __maybe_unused build_assertions(void)
> +{
> +    BUILD_BUG_ON(XEN_MSR_PAT != 0x050100070406ULL &&
> +                 "wrong XEN_MSR_PAT breaks PV guests");

... this didn't trigger for you. (We also don't normally add such
constructs with a string literal to BUILD_BUG_ON() expressions, not
the least because compilers may take issue with such. I'd like to
suggest to convert this to a comment instead.)

Jan