Re: [PATCH v3 6/7] arm64: add sysfs vulnerability show for speculative store bypass

[Marc Zyngier]

On 14/01/2019 16:37, Jeremy Linton wrote:
> Hi,
> 
> On 01/14/2019 04:15 AM, Marc Zyngier wrote:
>> On 09/01/2019 23:55, 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@xxxxxxx>
>>> ---
>>>   arch/arm64/kernel/cpu_errata.c | 48 ++++++++++++++++++++++++++++++++++
>>>   1 file changed, 48 insertions(+)
>>>
>>> diff --git a/arch/arm64/kernel/cpu_errata.c b/arch/arm64/kernel/cpu_errata.c
>>> index ee286d606d9b..c8ff96158b94 100644
>>> --- a/arch/arm64/kernel/cpu_errata.c
>>> +++ b/arch/arm64/kernel/cpu_errata.c
>>> @@ -288,6 +288,7 @@ enable_smccc_arch_workaround_1(const struct arm64_cpu_capabilities *entry)
>>>   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;
>>> @@ -385,10 +386,18 @@ static bool has_ssbd_mitigation(const struct arm64_cpu_capabilities *entry,
>>>   {
>>>   	struct arm_smccc_res res;
>>>   	bool required = true;
>>> +	bool is_vul;
>>>   	s32 val;
>>>   
>>>   	WARN_ON(scope != SCOPE_LOCAL_CPU || preemptible());
>>>   
>>> +	is_vul = is_midr_in_range_list(read_cpuid_id(), entry->midr_range_list);
>>> +
>>> +	if (is_vul)
>>> +		__ssb_safe = false;
>>> +
>>> +	arm64_requested_vuln_attrs |= VULN_SSB;
>>> +
>>>   	if (this_cpu_has_cap(ARM64_SSBS)) {
>>>   		required = false;
>>>   		goto out_printmsg;
>>> @@ -422,6 +431,7 @@ static bool has_ssbd_mitigation(const struct arm64_cpu_capabilities *entry,
>>>   		ssbd_state = ARM64_SSBD_UNKNOWN;
>>>   		return false;
>>>   
>>> +	/* machines with mixed mitigation requirements must not return this */
>>>   	case SMCCC_RET_NOT_REQUIRED:
>>>   		pr_info_once("%s mitigation not required\n", entry->desc);
>>>   		ssbd_state = ARM64_SSBD_MITIGATED;
>>> @@ -476,6 +486,17 @@ static bool has_ssbd_mitigation(const struct arm64_cpu_capabilities *entry,
>>>   
>>>   	return required;
>>>   }
>>> +
>>> +/* known vulnerable cores */
>>> +static const struct midr_range arm64_ssb_cpus[] = {
>>> +	MIDR_ALL_VERSIONS(MIDR_CORTEX_A57),
>>> +	MIDR_ALL_VERSIONS(MIDR_CORTEX_A72),
>>> +	MIDR_ALL_VERSIONS(MIDR_CORTEX_A73),
>>> +	MIDR_ALL_VERSIONS(MIDR_CORTEX_A75),
>>> +	MIDR_ALL_VERSIONS(MIDR_CORTEX_A76),
>>> +	{},
>>> +};
>>> +
>>>   #endif	/* CONFIG_ARM64_SSBD */
>>>   
>>>   static void __maybe_unused
>>> @@ -762,6 +783,7 @@ const struct arm64_cpu_capabilities arm64_errata[] = {
>>>   		.capability = ARM64_SSBD,
>>>   		.type = ARM64_CPUCAP_LOCAL_CPU_ERRATUM,
>>>   		.matches = has_ssbd_mitigation,
>>> +		.midr_range_list = arm64_ssb_cpus,
>>>   	},
>>>   #endif
>>>   #ifdef CONFIG_ARM64_ERRATUM_1188873
>>> @@ -809,4 +831,30 @@ ssize_t cpu_show_spectre_v2(struct device *dev, struct device_attribute *attr,
>>>   	return sprintf(buf, "Vulnerable\n");
>>>   }
>>>   
>>> +ssize_t cpu_show_spec_store_bypass(struct device *dev,
>>> +		struct device_attribute *attr, char *buf)
>>> +{
>>> +	/*
>>> +	 *  Two assumptions: First, get_ssbd_state() reflects the worse case
>>> +	 *  for hetrogenous machines, and that if SSBS is supported its
>>> +	 *  supported by all cores.
>>> +	 */
>>> +	switch (arm64_get_ssbd_state()) {
>>> +	case ARM64_SSBD_MITIGATED:
>>> +		return sprintf(buf, "Not affected\n");
>>> +
>>> +	case ARM64_SSBD_KERNEL:
>>> +	case ARM64_SSBD_FORCE_ENABLE:
>>> +		if (cpus_have_cap(ARM64_SSBS))
>>> +			return sprintf(buf, "Not affected\n");
>>> +		return sprintf(buf,
>>> +			"Mitigation: Speculative Store Bypass disabled\n");
>>> +	}
>>> +
>>> +	if (__ssb_safe)
>>> +		return sprintf(buf, "Not affected\n");
>>
>> The kbuild robot reports that this fails if CONFIG_ARM64_SSBD is not
>> selected. What should we print in this case? "Vulnerable"? Or "Unknown"?
> 
> The immediate fix is that the __ssb_safe variable should be in its own 
> conditional block which is  CONFIG_GENERIC_CPU_VULNERABILITIES || 
> CONFIG_ARM64_SSBD. If the mitigation isn't built in then this code won't 
> be run anyway because the sysfs entry won't be populated.

But in that case, we should probably assume that the system is
vulnerable, and we get a different default value for __ssb_safe.

> But, these CONFIG_ conditionals are less than ideal (and would be even 
> uglier if they were made more efficient). My own opinion at this point 
> is that we should really remove the compile time configs and leave the 
> mitigation built all the time. The raw code is fairly small, and we 
> could add in the nospectre_v2 command line options so that users can 
> choose to runtime disable them. That would also remove the need to 
> modify the core cpu vulnerabilities sysfs code.

That'd work for me. The whole thing is now an intractable mess, and I'd
welcome some level of simplification.

Thanks,

	M.
-- 
Jazz is not dead. It just smells funny...