Re: [PATCH 11/11] x86/fineibt: Add FineIBT+BHI mitigation

[Peter Zijlstra]

On Fri, Feb 14, 2025 at 10:02:12AM -0800, Kees Cook wrote:
> On Fri, Feb 07, 2025 at 01:15:40PM +0100, Peter Zijlstra wrote:
> > Due to FineIBT weakness, add an additional mitigation for BHI.
> > [...]
> > @@ -1020,6 +1036,8 @@ static __init int cfi_parse_cmdline(char
> >  			cfi_mode = CFI_KCFI;
> >  		} else if (!strcmp(str, "fineibt")) {
> >  			cfi_mode = CFI_FINEIBT;
> > +		} else if (IS_ENABLED(CONFIG_FINEIBT_BHI) && !strcmp(str, "fineibt+bhi")) {
> > +			cfi_mode = CFI_FINEIBT_BHI;
> >  		} else if (!strcmp(str, "norand")) {
> >  			cfi_rand = false;
> >  		} else {
> 
> While looking at FineIBT vs entry, I noticed that FineIBT+BHI must be
> explicitly selected at boot. Did you want it to be enabled automatically
> when the compiler supports it and FineIBT is enabled? Does there need to
> be a check for BHI added?

Yes, it needs to be tied in with the whole speculation mitigation crap.
Didn't want to bother with all that just yet though.