Re: [PATCH RFC 01/14] x86/asm: add iosubmit_cmds512() based on movdir64b CPU instruction
From: Borislav Petkov
Date: Thu Nov 21 2019 - 05:37:42 EST
On Wed, Nov 20, 2019 at 05:32:51PM -0800, Dan Williams wrote:
> True, but that would be a driver coding mistake flagged by the
> WARN_ON_ONCE, and the failure is static. The driver must check for
> static_cpu_has(X86_FEATURE_MOVDIR64B) once at init,
So if you do that at driver init time, you don't need the static variant
- simply use boot_cpu_has(). But if this function is going to be used on
other platforms, then you need the check in the function and it must be
static_cpu_has() for speed.
The static_cpu_has() thing is a soft-of once check anyway because it
gets patched by alternatives and after that it is 0 overhead.