Re: [PATCH] x86/speculation: Fix user-mode spectre-v2 protection with KERNEL_IBRS

[Josh Poimboeuf <jpoimboe@kernel.org>]

Getting confused with all the threads, will repost this review to the
lkml version...

On Mon, Feb 20, 2023 at 04:07:30AM -0800, Josh Poimboeuf wrote:
> On Mon, Feb 20, 2023 at 11:39:30AM +0100, KP Singh wrote:
> > +++ b/arch/x86/kernel/cpu/bugs.c
> > @@ -1132,6 +1132,19 @@ static inline bool spectre_v2_in_ibrs_mode(enum spectre_v2_mitigation mode)
> >  	       mode == SPECTRE_V2_EIBRS_LFENCE;
> >  }
> >  
> > +static inline bool spectre_v2_user_no_stibp(enum spectre_v2_mitigation mode)
> > +{
> > +	/* When IBRS or enhanced IBRS is enabled, STIBP is not needed.
> > +	 *
> > +	 * However, With KERNEL_IBRS, the IBRS bit is cleared on return
> > +	 * to user and the user-mode code needs to be able to enable protection
> > +	 * from cross-thread training, either by always enabling STIBP or
> > +	 * by enabling it via prctl.
> > +	 */
> > +	return (spectre_v2_in_ibrs_mode(mode) &&
> > +		!cpu_feature_enabled(X86_FEATURE_KERNEL_IBRS));
> > +}
> 
> The comments and code confused me, they both seem to imply some
> distinction between IBRS and KERNEL_IBRS, but in the kernel those are
> functionally the same thing.  e.g., the kernel doesn't have a user IBRS
> mode.
> 
> And, unless I'm missing some subtlety here, it seems to be a convoluted
> way of saying that eIBRS doesn't need STIBP in user space.
> 
> It would be simpler to just call it spectre_v2_in_eibrs_mode().
> 
> static inline bool spectre_v2_in_eibrs_mode(enum spectre_v2_mitigation mode)
> {
> 	return mode == SPECTRE_V2_EIBRS ||
> 	       mode == SPECTRE_V2_EIBRS_RETPOLINE ||
> 	       mode == SPECTRE_V2_EIBRS_LFENCE;
> }
> 
> And then spectre_v2_in_ibrs_mode() could be changed to call that:
> 
> static inline bool spectre_v2_in_eibrs_mode(enum spectre_v2_mitigation mode)
> {
> 	return spectre_v2_in_eibrs_mode(mode) || mode == SPECTRE_V2_IBRS;
> }
> 
> > @@ -1496,6 +1504,7 @@ static void __init spectre_v2_select_mitigation(void)
> >  		break;
> >  
> >  	case SPECTRE_V2_IBRS:
> > +		pr_err("enabling KERNEL_IBRS");
> 
> Why?
> 
> >  		setup_force_cpu_cap(X86_FEATURE_KERNEL_IBRS);
> >  		if (boot_cpu_has(X86_FEATURE_IBRS_ENHANCED))
> >  			pr_warn(SPECTRE_V2_IBRS_PERF_MSG);
> > @@ -2327,7 +2336,7 @@ static ssize_t mmio_stale_data_show_state(char *buf)
> >  
> >  static char *stibp_state(void)
> >  {
> > -	if (spectre_v2_in_ibrs_mode(spectre_v2_enabled))
> > +	if (spectre_v2_user_no_stibp(spectre_v2_enabled))
> >  		return "";
> 
> This seems like old cruft, can we just remove this check altogether?  In
> the eIBRS case, spectre_v2_user_stibp will already have its default of
> SPECTRE_V2_USER_NONE.
> 
> -- 
> Josh

-- 
Josh