[MODERATED] Re: [PATCH 2/3] v4 more sampling fun 2

On Fri, Mar 27, 2020 at 09:20:41AM -0700, speck for mark gross wrote:
> > > +
> > > +			off:	disable mitigation and remove performance
> > > +				impact to rdrand and rdseed
> > > +
> > >  	srcutree.counter_wrap_check [KNL]
> > >  			Specifies how frequently to check for
> > >  			grace-period sequence counter wrap for the
> > > @@ -397,6 +399,103 @@ static int __init tsx_async_abort_parse_cmdline(char *str)
> > >  }
> > >  early_param("tsx_async_abort", tsx_async_abort_parse_cmdline);
> > >  
> > > +#undef pr_fmt
> > > +#define pr_fmt(fmt)	"SRBDS: " fmt
> > > +
> > > +enum srbds_mitigations srbds_mitigation __ro_after_init = SRBDS_MITIGATION_FULL;
> > > +static const char * const srbds_strings[] = {
> > > +	[SRBDS_NOT_AFFECTED]		= "Not affected",
> > 
> > I don't think the SRBDS_NOT_AFFECTED enum is needed.
> > 
> > The "Not affected" string is already printed by cpu_show_common().
> > 
> > And srbds_configure_mitigation() can just check for
> > boot_cpu_has_bug(X86_BUG_SRBDS) instead of this enum value.
> 
> True.  I prefer having the complete state machine represented by the
> enumeration but, if you feel strongly I'll try to remove it.

If the state isn't actually used by the sysfs printing code then it just
serves to add confusion for a reader of the code.

And notice none of the other mitigations have this.  It's a subtle
inconsistency, which is a recipe for future bugs.

> > > +	[SRBDS_MITIGATION_OFF]		= "Vulnerable",
> > > +	[SRBDS_MITIGATION_UCODE_NEEDED]	= "Vulnerable: no microcode",
> > > +	[SRBDS_MITIGATION_FULL]		= "Mitigated",
> > > +	[SRBDS_NOT_AFFECTED_TSX_OFF]	= "Not affected (TSX disabled)",
> > > +	[SRBDS_HYPERVISOR]		= "Unknown",
> > > +};
> > > +
> > > +static bool srbds_off;
> > > +
> > > +void srbds_configure_mitigation(void)
> > 
> > To me, "configure" sounds like it could be a software thing.
> We are configuring the meditation by honoring the kernel command line in this
> function.
> 
> > 
> > How about update_srbds_msr(), which is also more consistent with some of
> > the other naming in the file.
> 
> If you like but, I was mimicking the TAA implementation when choosing the name
> of this function.

Hm, which TAA function name were you mimicking?

> > > +static void __init srbds_select_mitigation(void)
> > > +{
> > > +	u64 ia32_cap;
> > > +
> > > +	if (boot_cpu_has(X86_FEATURE_HYPERVISOR)) {
> > > +		srbds_mitigation = SRBDS_HYPERVISOR;
> > > +		return;
> > > +	}
> > 
> > It's confusing that this comes before the X86_BUG_SRBDS check.  Is that
> > on purpose?
> VM vendors on keybase wanted it so.
> 
> > 
> > If so, I presume it doesn't have the intended effect, because if booting
> > in a guest on a non-vulnerable CPU, srbds_mitigation will get set to
> > SRBDS_HYPERVISOR, but the shown state will still be "Not affected"
> > because of how cpu_show_common() checks for the bug at the beginning.
> 
> The intended effect is to report "Unknown" for the vulnerability status of
> SRBDS for all guest VM's running this code and to never touch the MSR
> controlling the mitigation.
> 
> I think this is what happens.  I'll double check and confirm with testing.

Maybe I'm blind, but given how cpu_show_common() works, I don't see how
that could possibly happen when boot_cpu_has_bug(X86_BUG_SRBDS) is
false.  If I'm wrong please enlighten me.

> > > +
> > > +	if (!boot_cpu_has_bug(X86_BUG_SRBDS)) {
> > > +		srbds_mitigation = SRBDS_NOT_AFFECTED;
> > > +		return;
> > > +	}
> > > +
> > > +	if (!boot_cpu_has(X86_FEATURE_SRBDS_CTRL)) {
> > > +		srbds_mitigation = SRBDS_MITIGATION_UCODE_NEEDED;
> > > +		return;
> > > +	}
> > > +
> > > +	if (boot_cpu_has_bug(X86_BUG_SRBDS)) {
> > > +		srbds_mitigation = SRBDS_MITIGATION_FULL;
> > > +		/*
> > > +		 * Check to see if this is one of the MDS_NO systems supporting
> > > +		 * TSX that are only exposed to SRBDS when TSX is enabled.
> > > +		 */
> > > +		ia32_cap = x86_read_arch_cap_msr();
> > > +		if (ia32_cap & ARCH_CAP_MDS_NO) {
> > > +			if (!boot_cpu_has(X86_FEATURE_RTM))
> > > +				srbds_mitigation = SRBDS_NOT_AFFECTED_TSX_OFF;
> > > +		}
> > 
> > I wonder if this would be equivalent?
> > 
> > 		if (!boot_cpu_has_bug(X86_BUG_MDS) && boot_cpu_has(X86_BUG_TAA) &&
> > 		    !boot_cpu_has(X86_FEATURE_RTM))
> > 		    	srbds_mitigation = SRBDS_NOT_AFFECTED_TSX_OFF;
> I have no clue if its equivalent but, I find your version much harder to parse
> and my version matches more closely to what the white paper states.

Ok.

> > > +	}
> > > +
> > > +	if (cpu_mitigations_off() || srbds_off) {
> > > +		if (srbds_mitigation != SRBDS_NOT_AFFECTED_TSX_OFF)
> > > +			srbds_mitigation = SRBDS_MITIGATION_OFF;
> > > +	}
> > > +
> > > +	srbds_configure_mitigation();
> > > +}
> > 
> > The other mitigations do a printk with the mitigation status here.
> > Should we not do the same here?
> Perhaps but, I was attempting to limit the extra chatter in the kernel logs.
> If you feel it adds value to the code I'll put it in.  FWIW I think the chatter
> from the other places should be removed.

It should at least be consistent with the others.

> > > +ssize_t cpu_show_special_register_data_sampling(struct device *dev, struct device_attribute *attr, char *buf)
> > > +{
> > 
> > This function name can be renamed to match whatever the filename turns
> > out to be:
> > 
> >   cpu_show_srbds()
> > 
> > or
> > 
> >   cpu_show_special_register_data_buffer_sampling()
> 
> I got some early review feedback to not use the shorter name "srbds" for the
> sysfs inodes so I was trying to avoid the full spelled out the full names.
> 
> "srbds" and "special_register_data_buffer_sampling" do not imply HWRNG security
> risk so IMO I'd like to just use srbds.  What do others think?

+1 for "srbds" for both file name and function name.

> > > +void srbds_configure_mitigation(void);
> > > +
> > >  #ifdef CONFIG_CPU_SUP_INTEL
> > > +
> > >  enum tsx_ctrl_states {
> > >  	TSX_CTRL_ENABLE,
> > >  	TSX_CTRL_DISABLE,
> > > diff --git a/arch/x86/kernel/cpu/intel.c b/arch/x86/kernel/cpu/intel.c
> > > index be82cd5841c3..1b083a2a415b 100644
> > > --- a/arch/x86/kernel/cpu/intel.c
> > > +++ b/arch/x86/kernel/cpu/intel.c
> > > @@ -684,6 +684,8 @@ static void init_intel(struct cpuinfo_x86 *c)
> > >  		tsx_enable();
> > >  	if (tsx_ctrl_state == TSX_CTRL_DISABLE)
> > >  		tsx_disable();
> > > +
> > > +	srbds_configure_mitigation();
> > >  }
> > 
> > I'm not sure this is the right place to call this.  On the boot CPU,
> > this gets called *before* the mitigation is selected, and then again
> > from srbds_select_mitigation() itself.
> > 
> > Maybe it should instead be called from identify_secondary_cpu(), while
> > keeping the existing call from srbds_select_mitigation() so the boot CPU
> > also gets updated.
> The SRBDS MSR needs to be poked from each CPU thread if you want to disable the
> mitigation.  Does identify_secondary_cpu get called per HW thread?

identify_secondary_cpu() gets called for every HW thread, other than the
boot one.

The boot one already does it in srbds_select_mitigation().

> I felt it would be nice to keep the call close to other mitigation setup
> logic, in this case the TAA logic.

That's not the TAA logic, but rather the TSX logic (though yes I realize
they're related).  And anyway SRBDS and TSX have different requirements.

> What do you think it will be more readable if I move it to
> identify_secondary_cpu?

As I said - it gets called *twice* on the boot CPU.  The first time is
before srbds_mitigation has even been configured.  That's confusing.

-- 
Josh