Hi Andre,
On 3/21/19 6:35 PM, Andre Przywara wrote:
> On Thu, 21 Mar 2019 13:54:07 +0100
> Auger Eric <eric.auger@xxxxxxxxxx> wrote:
>
> Hi Eric,
>
> many thanks for looking at this!
> I was about to prepare a new revision.
>
>> Hi Andre,
>>
>> On 3/1/19 12:43 AM, Andre Przywara wrote:
>>> KVM implements the firmware interface for mitigating cache speculation
>>> vulnerabilities. Guests may use this interface to ensure mitigation is
>>> active.
>>> If we want to migrate such a guest to a host with a different support
>>> level for those workarounds, migration might need to fail, to ensure that
>>> critical guests don't loose their protection.
>>>
>>> Introduce a way for userland to save and restore the workarounds state.
>>> On restoring we do checks that make sure we don't downgrade our
>>> mitigation level.
>>>
>>> Signed-off-by: Andre Przywara <andre.przywara@xxxxxxx>
>>> ---
>>> arch/arm/include/asm/kvm_emulate.h | 10 +++
>>> arch/arm/include/uapi/asm/kvm.h | 10 +++
>>> arch/arm64/include/asm/kvm_emulate.h | 14 +++
>>> arch/arm64/include/uapi/asm/kvm.h | 9 ++
>>> virt/kvm/arm/psci.c | 128 +++++++++++++++++++++++----
>>> 5 files changed, 155 insertions(+), 16 deletions(-)
>>>
>>> diff --git a/arch/arm/include/asm/kvm_emulate.h b/arch/arm/include/asm/kvm_emulate.h
>>> index 8927cae7c966..663a02d7e6f4 100644
>>> --- a/arch/arm/include/asm/kvm_emulate.h
>>> +++ b/arch/arm/include/asm/kvm_emulate.h
>>> @@ -283,6 +283,16 @@ static inline unsigned long kvm_vcpu_get_mpidr_aff(struct kvm_vcpu *vcpu)
>>> return vcpu_cp15(vcpu, c0_MPIDR) & MPIDR_HWID_BITMASK;
>>> }
>>>
>>> +static inline bool kvm_arm_get_vcpu_workaround_2_flag(struct kvm_vcpu *vcpu)
>>> +{
>>> + return false;
>>> +}
>>> +
>>> +static inline void kvm_arm_set_vcpu_workaround_2_flag(struct kvm_vcpu *vcpu,
>>> + bool flag)
>>> +{
>>> +}
>>> +
>>> static inline void kvm_vcpu_set_be(struct kvm_vcpu *vcpu)
>>> {
>>> *vcpu_cpsr(vcpu) |= PSR_E_BIT;
>>> diff --git a/arch/arm/include/uapi/asm/kvm.h b/arch/arm/include/uapi/asm/kvm.h
>>> index 4602464ebdfb..ba4d2afe65e3 100644
>>> --- a/arch/arm/include/uapi/asm/kvm.h
>>> +++ b/arch/arm/include/uapi/asm/kvm.h
>>> @@ -214,6 +214,16 @@ struct kvm_vcpu_events {
>>> #define KVM_REG_ARM_FW_REG(r) (KVM_REG_ARM | KVM_REG_SIZE_U64 | \
>>> KVM_REG_ARM_FW | ((r) & 0xffff))
>>> #define KVM_REG_ARM_PSCI_VERSION KVM_REG_ARM_FW_REG(0)
>>> +#define KVM_REG_ARM_SMCCC_ARCH_WORKAROUND_1 KVM_REG_ARM_FW_REG(1)
>>> +#define KVM_REG_ARM_SMCCC_ARCH_WORKAROUND_1_NOT_AVAIL 0
>>> +#define KVM_REG_ARM_SMCCC_ARCH_WORKAROUND_1_AVAIL 1
>>> +#define KVM_REG_ARM_SMCCC_ARCH_WORKAROUND_1_UNAFFECTED 2
>>> +#define KVM_REG_ARM_SMCCC_ARCH_WORKAROUND_2 KVM_REG_ARM_FW_REG(2)
>>> +#define KVM_REG_ARM_SMCCC_ARCH_WORKAROUND_2_NOT_AVAIL 0
>>> +#define KVM_REG_ARM_SMCCC_ARCH_WORKAROUND_2_UNKNOWN 1
>> Would be worth adding a comment saying that values are chosen so that
>> higher values mean better protection. Otherwise it looks strange
>> NOT_AVAIL/AVAIL/UNAFFECTED values are not the same for both workarounds.
>>> +#define KVM_REG_ARM_SMCCC_ARCH_WORKAROUND_2_AVAIL 2
>>> +#define KVM_REG_ARM_SMCCC_ARCH_WORKAROUND_2_UNAFFECTED 3
>>> +#define KVM_REG_ARM_SMCCC_ARCH_WORKAROUND_2_ENABLED (1U << 4)
>>
>>>
>>> /* Device Control API: ARM VGIC */
>>> #define KVM_DEV_ARM_VGIC_GRP_ADDR 0
>>> diff --git a/arch/arm64/include/asm/kvm_emulate.h b/arch/arm64/include/asm/kvm_emulate.h
>>> index d3842791e1c4..c00c17c9adb6 100644
>>> --- a/arch/arm64/include/asm/kvm_emulate.h
>>> +++ b/arch/arm64/include/asm/kvm_emulate.h
>>> @@ -348,6 +348,20 @@ static inline unsigned long kvm_vcpu_get_mpidr_aff(struct kvm_vcpu *vcpu)
>>> return vcpu_read_sys_reg(vcpu, MPIDR_EL1) & MPIDR_HWID_BITMASK;
>>> }
>>>
>>> +static inline bool kvm_arm_get_vcpu_workaround_2_flag(struct kvm_vcpu *vcpu)
>>> +{
>>> + return vcpu->arch.workaround_flags & VCPU_WORKAROUND_2_FLAG;
>>> +}
>>> +
>>> +static inline void kvm_arm_set_vcpu_workaround_2_flag(struct kvm_vcpu *vcpu,
>>> + bool flag)
>>> +{
>>> + if (flag)
>>> + vcpu->arch.workaround_flags |= VCPU_WORKAROUND_2_FLAG;
>>> + else
>>> + vcpu->arch.workaround_flags &= ~VCPU_WORKAROUND_2_FLAG;
>>> +}
>>> +
>>> static inline void kvm_vcpu_set_be(struct kvm_vcpu *vcpu)
>>> {
>>> if (vcpu_mode_is_32bit(vcpu)) {
>>> diff --git a/arch/arm64/include/uapi/asm/kvm.h b/arch/arm64/include/uapi/asm/kvm.h
>>> index 97c3478ee6e7..367e96fe654e 100644
>>> --- a/arch/arm64/include/uapi/asm/kvm.h
>>> +++ b/arch/arm64/include/uapi/asm/kvm.h
>>> @@ -225,6 +225,15 @@ struct kvm_vcpu_events {
>>> #define KVM_REG_ARM_FW_REG(r) (KVM_REG_ARM64 | KVM_REG_SIZE_U64 | \
>>> KVM_REG_ARM_FW | ((r) & 0xffff))
>>> #define KVM_REG_ARM_PSCI_VERSION KVM_REG_ARM_FW_REG(0)
>>> +#define KVM_REG_ARM_SMCCC_ARCH_WORKAROUND_1 KVM_REG_ARM_FW_REG(1)
>>> +#define KVM_REG_ARM_SMCCC_ARCH_WORKAROUND_1_NOT_AVAIL 0
>>> +#define KVM_REG_ARM_SMCCC_ARCH_WORKAROUND_1_AVAIL 1
>>> +#define KVM_REG_ARM_SMCCC_ARCH_WORKAROUND_2 KVM_REG_ARM_FW_REG(2)
>>> +#define KVM_REG_ARM_SMCCC_ARCH_WORKAROUND_2_NOT_AVAIL 0
>>> +#define KVM_REG_ARM_SMCCC_ARCH_WORKAROUND_2_UNKNOWN 1
>>> +#define KVM_REG_ARM_SMCCC_ARCH_WORKAROUND_2_AVAIL 2
>>> +#define KVM_REG_ARM_SMCCC_ARCH_WORKAROUND_2_UNAFFECTED 3
>>> +#define KVM_REG_ARM_SMCCC_ARCH_WORKAROUND_2_ENABLED (1U << 4)
>>>
>>> /* Device Control API: ARM VGIC */
>>> #define KVM_DEV_ARM_VGIC_GRP_ADDR 0
>>> diff --git a/virt/kvm/arm/psci.c b/virt/kvm/arm/psci.c
>>> index 9b73d3ad918a..e65664c09b12 100644
>>> --- a/virt/kvm/arm/psci.c
>>> +++ b/virt/kvm/arm/psci.c
>>> @@ -445,42 +445,97 @@ int kvm_hvc_call_handler(struct kvm_vcpu *vcpu)
>>>
>>> int kvm_arm_get_fw_num_regs(struct kvm_vcpu *vcpu)
>>> {
>>> - return 1; /* PSCI version */
>>> + return 3; /* PSCI version and two workaround registers */
>>> }
>>>
>>> int kvm_arm_copy_fw_reg_indices(struct kvm_vcpu *vcpu, u64 __user *uindices)
>>> {
>>> - if (put_user(KVM_REG_ARM_PSCI_VERSION, uindices))
>>> + if (put_user(KVM_REG_ARM_PSCI_VERSION, uindices++))
>>> + return -EFAULT;
>>> +
>>> + if (put_user(KVM_REG_ARM_SMCCC_ARCH_WORKAROUND_1, uindices++))
>>> + return -EFAULT;
>>> +
>>> + if (put_user(KVM_REG_ARM_SMCCC_ARCH_WORKAROUND_2, uindices++))
>>> return -EFAULT;
>> Wouldn't it make sense to have a const array somewhere listing the FW
>> regs and putting KVM_REG_ARM_FW_REG[i]? Also kvm_arm_get_fw_num_regs
>> could return the ARRAY_SIZE.
>
> Yeah, makes sense. I was cowardly pushing this off to the one adding the next firmware register ;-)
>
>> vcpu arg is never used actually (not related to this patch).
>>>
>>> return 0;
>>> }
>>>
>>> +#define KVM_REG_FEATURE_LEVEL_WIDTH 4
>>> +#define KVM_REG_FEATURE_LEVEL_MASK (BIT(KVM_REG_FEATURE_LEVEL_WIDTH) - 1)
>>> +
>>> +/*
>>> + * Convert the workaround level into an easy-to-compare number, where higher
>>> + * values mean better protection.
>>> + */
>>> +static int get_kernel_wa_level(u64 regid)
>>> +{
>>> + switch (regid) {
>>> + case KVM_REG_ARM_SMCCC_ARCH_WORKAROUND_1:
>>> + if (kvm_arm_harden_branch_predictor())
>>> + return KVM_REG_ARM_SMCCC_ARCH_WORKAROUND_1_AVAIL;
>>> + else
>>> + return KVM_REG_ARM_SMCCC_ARCH_WORKAROUND_1_NOT_AVAIL;
>>> + case KVM_REG_ARM_SMCCC_ARCH_WORKAROUND_2:
>>> + switch (kvm_arm_have_ssbd()) {
>>> + case KVM_SSBD_FORCE_DISABLE:
>>> + return KVM_REG_ARM_SMCCC_ARCH_WORKAROUND_2_NOT_AVAIL;
>>> + case KVM_SSBD_KERNEL:
>>> + return KVM_REG_ARM_SMCCC_ARCH_WORKAROUND_2_AVAIL;
>>> + case KVM_SSBD_FORCE_ENABLE:
>>> + case KVM_SSBD_MITIGATED:
>>> + return KVM_REG_ARM_SMCCC_ARCH_WORKAROUND_2_UNAFFECTED;
>>> + case KVM_SSBD_UNKNOWN:
>>> + default:
>>> + return KVM_REG_ARM_SMCCC_ARCH_WORKAROUND_2_UNKNOWN;
>>> + }
>>> + }
>>> +
>>> + return 0;
>> I would rather return -EINVAL although the function is not called for
>> any invalid reg.
>>> +}
>>> +
>>> int kvm_arm_get_fw_reg(struct kvm_vcpu *vcpu, const struct kvm_one_reg *reg)
>>> {
>>> - if (reg->id == KVM_REG_ARM_PSCI_VERSION) {
>>> - void __user *uaddr = (void __user *)(long)reg->addr;
>>> - u64 val;
>>> + void __user *uaddr = (void __user *)(long)reg->addr;
>>> + u64 val;
>>>
>>> + switch (reg->id) {
>>> + case KVM_REG_ARM_PSCI_VERSION:
>>> val = kvm_psci_version(vcpu, vcpu->kvm);
>>> - if (copy_to_user(uaddr, &val, KVM_REG_SIZE(reg->id)))
>>> - return -EFAULT;
>>> -
>>> - return 0;
>>> + break;
>>> + case KVM_REG_ARM_SMCCC_ARCH_WORKAROUND_1:
>>> + val = get_kernel_wa_level(reg->id) & KVM_REG_FEATURE_LEVEL_MASK;
>> Can get_kernel_wa_level return something outside of
>> KVM_REG_FEATURE_LEVEL_MASK?
>>> + break;
>>> + case KVM_REG_ARM_SMCCC_ARCH_WORKAROUND_2:
>>> + val = get_kernel_wa_level(reg->id) & KVM_REG_FEATURE_LEVEL_MASK;
>> same here
>>> + if (kvm_arm_have_ssbd() == KVM_SSBD_KERNEL &&> + kvm_arm_get_vcpu_workaround_2_flag(vcpu))
>> nit: if (val == KVM_REG_ARM_SMCCC_ARCH_WORKAROUND_2_AVAIL &&
>> kvm_arm_get_vcpu_workaround_2_flag(vcpu)).
>
> Yeah, might be better, since we operate in the KVM_REG_ARM_ namespace here.
>
>>> + val |= KVM_REG_ARM_SMCCC_ARCH_WORKAROUND_2_ENABLED;
>>> + break;
>>> + default:
>>> + return -ENOENT;
>>> }
>>>
>>> - return -EINVAL;
>>> + if (copy_to_user(uaddr, &val, KVM_REG_SIZE(reg->id)))
>>> + return -EFAULT;
>>> +
>>> + return 0;
>>> }
>>>
>>> int kvm_arm_set_fw_reg(struct kvm_vcpu *vcpu, const struct kvm_one_reg *reg)
>>> {
>>> - if (reg->id == KVM_REG_ARM_PSCI_VERSION) {
>>> - void __user *uaddr = (void __user *)(long)reg->addr;
>>> - bool wants_02;
>>> - u64 val;
>>> + void __user *uaddr = (void __user *)(long)reg->addr;
>>> + u64 val;
>>> + int wa_level;
>>> +
>>> + if (copy_from_user(&val, uaddr, KVM_REG_SIZE(reg->id)))
>>> + return -EFAULT;
>>>
>>> - if (copy_from_user(&val, uaddr, KVM_REG_SIZE(reg->id)))
>>> - return -EFAULT;
>>> + switch (reg->id) {
>>> + case KVM_REG_ARM_PSCI_VERSION:
>>> + {
>>> + bool wants_02;
>>>
>>> wants_02 = test_bit(KVM_ARM_VCPU_PSCI_0_2, vcpu->arch.features);
>>>
>>> @@ -497,6 +552,47 @@ int kvm_arm_set_fw_reg(struct kvm_vcpu *vcpu, const struct kvm_one_reg *reg)
>>> vcpu->kvm->arch.psci_version = val;
>>> return 0;
>>> }
>>> + break;
>>> + }
>>> +
>>> + case KVM_REG_ARM_SMCCC_ARCH_WORKAROUND_1:
>>> + if (val & ~KVM_REG_FEATURE_LEVEL_MASK)
>>> + return -EINVAL;
>>> +
>>> + wa_level = val & KVM_REG_FEATURE_LEVEL_MASK;
>> not needed
>
> Yeah, I kept it in for symmetry reasons. I was also afraid that someone
> changes the check above in the future.
>
>>> +
>>> + /* For now we only accept the very same workaround level. */
>>> + if (get_kernel_wa_level(reg->id) != wa_level)
>>> + return -EINVAL;
>>> +
>>> + return 0;
>>> +
>>> + case KVM_REG_ARM_SMCCC_ARCH_WORKAROUND_2:
>>> + if (val & ~(KVM_REG_FEATURE_LEVEL_MASK |
>>> + KVM_REG_ARM_SMCCC_ARCH_WORKAROUND_2_ENABLED))> + return -EINVAL;
>>> +
>>> + wa_level = val & KVM_REG_FEATURE_LEVEL_MASK;
>>> +
>>> + if (get_kernel_wa_level(reg->id) < wa_level)
>>> + return -EINVAL;
>>> +
>> worth a comment?
>
> True.
>
>>> + if (kvm_arm_have_ssbd() != KVM_SSBD_KERNEL)
>>> + return 0;
>>> +
>>> + switch (wa_level) {
>>> + case KVM_REG_ARM_SMCCC_ARCH_WORKAROUND_2_AVAIL:
>>> + kvm_arm_set_vcpu_workaround_2_flag(vcpu,
>>> + val & KVM_REG_ARM_SMCCC_ARCH_WORKAROUND_2_ENABLED);
>>> + break;
>>> + case KVM_REG_ARM_SMCCC_ARCH_WORKAROUND_2_UNAFFECTED:
>>> + kvm_arm_set_vcpu_workaround_2_flag(vcpu, true);
>> Looks strange to me we enable the flag when unaffected.
>
> UNAFFECTED is the *requested* level here, so the original kernel had it.
> It means "always active or not needed". We get here if the local kernel
> has a *switchable* workaround, so we have to make sure it's enabled, to
> match the original workaround level.
> Will add a comment before the switch.
OK I get it now. Adding such kind of comments at main decision levels
would definitively help (me).
Thanks
Eric
>
> Cheers,
> Andre.
>
>
>>
>>> + break;
>>> + }
>>> +
>>> + return 0;
>>> + default:
>>> + return -ENOENT;
>>> }
>>>
>>> return -EINVAL;
>>>
>>
>> Thanks
>>
>> Eric
>
