Re: [PATCH v6 09/10] arm64: add sysfs vulnerability show for speculative store bypass

[Andre Przywara <andre.przywara@arm.com>]

On Fri, 5 Apr 2019 15:43:10 +0100
Will Deacon <will.deacon@arm.com> wrote:

> On Fri, Apr 05, 2019 at 11:10:22AM +0100, Andre Przywara wrote:
> > On Wed, 3 Apr 2019 17:50:05 +0100
> > Will Deacon <will.deacon@arm.com> wrote:
> >   
> > > Hi Jeremy,
> > > 
> > > On Thu, Mar 21, 2019 at 06:05:56PM -0500, Jeremy Linton wrote:  
> > > > Return status based on ssbd_state and the arm64 SSBS feature. If
> > > > the mitigation is disabled, or the firmware isn't responding then
> > > > return the expected machine state based on a new blacklist of known
> > > > vulnerable cores.
> > > > 
> > > > Signed-off-by: Jeremy Linton <jeremy.linton@arm.com>
> > > > Reviewed-by: Andre Przywara <andre.przywara@arm.com>
> > > > Tested-by: Stefan Wahren <stefan.wahren@i2se.com>
> > > > ---
> > > >  arch/arm64/kernel/cpu_errata.c | 44 ++++++++++++++++++++++++++++++++++
> > > >  1 file changed, 44 insertions(+)
> > > > 
> > > > diff --git a/arch/arm64/kernel/cpu_errata.c b/arch/arm64/kernel/cpu_errata.c
> > > > index 6958dcdabf7d..172ffbabd597 100644
> > > > --- a/arch/arm64/kernel/cpu_errata.c
> > > > +++ b/arch/arm64/kernel/cpu_errata.c
> > > > @@ -278,6 +278,7 @@ static int detect_harden_bp_fw(void)
> > > >  DEFINE_PER_CPU_READ_MOSTLY(u64, arm64_ssbd_callback_required);
> > > >  
> > > >  int ssbd_state __read_mostly = ARM64_SSBD_KERNEL;
> > > > +static bool __ssb_safe = true;
> > > >  
> > > >  static const struct ssbd_options {
> > > >  	const char	*str;
> > > > @@ -386,6 +387,9 @@ static bool has_ssbd_mitigation(const struct arm64_cpu_capabilities *entry,
> > > >  
> > > >  	WARN_ON(scope != SCOPE_LOCAL_CPU || preemptible());
> > > >  
> > > > +	if (is_midr_in_range_list(read_cpuid_id(), entry->midr_range_list))
> > > > +		__ssb_safe = false;
> > > > +    
> > > 
> > > Does this mean that we assume that CPUs not present in our table are not
> > > affected by speculative store bypass?  
> > 
> > No, not affected are only those where we either have SSBS or the firmware
> > explicitly returns SMCCC_RET_NOT_REQUIRED. This is governed by ssbd_state.
> > So this doesn't affect correctness.  
> 
> I don't think that's true. My TX2, for example, says "Not affected" for
> spec_store_bypass, but we don't actually know whether it's affected or
> not and so it should report "Vulnerable" instead, like we do for spectre_v2
> on the same machine.

Yeah, what I actually meant was that this list doesn't affect whether the workaround gets applied or not. But indeed the reporting is wrong.

> > __ssb_safe is an additional state just used for the sysfs output. But
> > indeed it looks like this is wrong if the CPU is both not listed and the
> > system doesn't provide the firmware interface: AFAICS we would report "Not
> > affected" in this case.  
> 
> Yes, that's what I was getting at.
> 
> > > I don't think that's a good
> > > assumption, because we don't necessary have knowledge about partner or
> > > future CPU implementations, so I think any CPU lists really have to be
> > > whitelists like they are for the other vulnerabilities.  
> > 
> > I think the idea was to cover all those "legacy" systems which have
> > older cores (no SSBS), but didn't get an firmware update. So your old Seattle
> > would truthfully report "Vulnerable", but any random A53 device would
> > report "Not affected", even with ancient firmware.  
> 
> The only manageable way to deal with this is to use a whitelist, just like
> we do for the other vulnerabilities. We shouldn't have to update it for
> long because newer cores should have SSBS.

Agreed. We should start with __ssb_safe = false, and work from there. Seems much safer.

Cheers,
Andre.