Re: [RFC PATCH 3/8] KVM: arm64: Add standard secure service calls firmware register

[Raghavendra Rao Ananta <rananta@google.com>]

On Wed, Nov 3, 2021 at 5:15 PM Oliver Upton <oupton@google.com> wrote:
>
> On Tue, Nov 02, 2021 at 12:21:58AM +0000, Raghavendra Rao Ananta wrote:
> > Introduce a firmware register that encapsulates standard secure
> > service calls (owner value 4) as a bitmap. Depending on how the
> > user-space configures the register, the features will be enabled
> > or disabled for the guest. Currently, this includes support only
> > for ARM True Random Number Generator (TRNG) service, with bit-0
> > of the register representing mandatory features of v1.0.
> >
> > Signed-off-by: Raghavendra Rao Ananta <rananta@google.com>
> > ---
> >  Documentation/virt/kvm/arm/hypercalls.rst | 17 +++++
> >  arch/arm64/include/asm/kvm_host.h         |  2 +
> >  arch/arm64/include/uapi/asm/kvm.h         |  6 ++
> >  arch/arm64/kvm/arm.c                      |  8 +++
> >  arch/arm64/kvm/hypercalls.c               | 75 ++++++++++++++++++++++-
> >  arch/arm64/kvm/trng.c                     |  9 +--
> >  include/kvm/arm_hypercalls.h              |  5 ++
> >  7 files changed, 113 insertions(+), 9 deletions(-)
> >
> > diff --git a/Documentation/virt/kvm/arm/hypercalls.rst b/Documentation/virt/kvm/arm/hypercalls.rst
> > index 85dfd682d811..1601919f256d 100644
> > --- a/Documentation/virt/kvm/arm/hypercalls.rst
> > +++ b/Documentation/virt/kvm/arm/hypercalls.rst
> > @@ -20,6 +20,14 @@ pseudo-registers" that can be manipulated using the GET/SET_ONE_REG
> >  interface. These registers can be saved/restored by userspace, and set
> >  to a convenient value if required.
> >
> > +The firmware register KVM_REG_ARM_STD exposes the hypercall services
>
> nit: try to cram BITMAP in the name. IMO, this will help disambiguate
> with version-based FW regs, like KVM_REG_ARM_PSCI_VERSION.
>
> > +in the form of a feature bitmap. Upon VM creation, by default, KVM exposes
> > +all the features to the guest, which can be learnt using GET_ONE_REG
> > +interface. Conversely, the features can be enabled or disabled via the
> > +SET_ONE_REG interface. These registers allow the user-space modification
> > +only until the VM has started running, after which they turn to read-only
> > +registers. SET_ONE_REG in this scenario will return -EBUSY.
> > +
> >  The following register is defined:
> >
> >  * KVM_REG_ARM_PSCI_VERSION:
> > @@ -74,4 +82,13 @@ The following register is defined:
> >      KVM_REG_ARM_SMCCC_ARCH_WORKAROUND_2_NOT_REQUIRED:
> >        The workaround is always active on this vCPU or it is not needed.
> >
> > +* KVM_REG_ARM_STD:
> > +    Controls the bitmap of the ARM Standard Secure Service Calls.
> > +
> > +    The following bits are accepted:
> > +
> > +      KVM_REG_ARM_STD_TRNG_V1_0:
>
> state the bit position as well
>
I was afraid of the name getting too long. But let me see.
> > +        The bit represents the services offered under v1.0 of ARM True Random Number Generator
> > +        (TRNG) specification (ARM DEN 0098).
> > +
> >  .. [1] https://developer.arm.com/-/media/developer/pdf/ARM_DEN_0070A_Firmware_interfaces_for_mitigating_CVE-2017-5715.pdf
> > diff --git a/arch/arm64/include/asm/kvm_host.h b/arch/arm64/include/asm/kvm_host.h
> > index 0b2502494a17..176d6be7b4da 100644
> > --- a/arch/arm64/include/asm/kvm_host.h
> > +++ b/arch/arm64/include/asm/kvm_host.h
> > @@ -105,6 +105,8 @@ struct kvm_arch_memory_slot {
> >  struct hvc_reg_desc {
> >       bool write_disabled;
> >       bool write_attempted;
> > +
> > +     u64 kvm_std_bmap;
> >  };
> >
> >  struct kvm_arch {
> > diff --git a/arch/arm64/include/uapi/asm/kvm.h b/arch/arm64/include/uapi/asm/kvm.h
> > index b3edde68bc3e..6387dea5396d 100644
> > --- a/arch/arm64/include/uapi/asm/kvm.h
> > +++ b/arch/arm64/include/uapi/asm/kvm.h
> > @@ -281,6 +281,12 @@ struct kvm_arm_copy_mte_tags {
> >  #define KVM_REG_ARM_SMCCC_ARCH_WORKAROUND_2_NOT_REQUIRED     3
> >  #define KVM_REG_ARM_SMCCC_ARCH_WORKAROUND_2_ENABLED          (1U << 4)
> >
> > +#define KVM_REG_ARM_STD                      KVM_REG_ARM_FW_REG(3)
> > +enum kvm_reg_arm_std_bmap {
> > +     KVM_REG_ARM_STD_TRNG_V1_0,
> > +     KVM_REG_ARM_STD_BMAP_MAX,
> > +};
> > +
>
> I would recommend just defining the bit values explicitly rather than
> using an enumeration:
>
>   #define KVM_REG_ARM_STD_TRNG_V1_0     (1ULL << 0)
>
> You do lose the convenience of having KVM_REG_ARM_STD_BMAP_MAX.
>
Just curious, any particular reason for this? IMO, going an enum route
could avoid human errors. Anything I'm missing?
> >  /* SVE registers */
> >  #define KVM_REG_ARM64_SVE            (0x15 << KVM_REG_ARM_COPROC_SHIFT)
> >
> > diff --git a/arch/arm64/kvm/arm.c b/arch/arm64/kvm/arm.c
> > index f9a25e439e99..1cf58aa49222 100644
> > --- a/arch/arm64/kvm/arm.c
> > +++ b/arch/arm64/kvm/arm.c
> > @@ -130,6 +130,13 @@ static void set_default_spectre(struct kvm *kvm)
> >               kvm->arch.pfr0_csv3 = 1;
> >  }
> >
> > +static void set_default_hypercalls(struct kvm *kvm)
> > +{
> > +     struct hvc_reg_desc *hvc_desc = &kvm->arch.hvc_desc;
> > +
> > +     hvc_desc->kvm_std_bmap = ARM_SMCCC_STD_FEATURES;
> > +}
> > +
> >  /**
> >   * kvm_arch_init_vm - initializes a VM data structure
> >   * @kvm:     pointer to the KVM struct
> > @@ -156,6 +163,7 @@ int kvm_arch_init_vm(struct kvm *kvm, unsigned long type)
> >       kvm->arch.max_vcpus = kvm_arm_default_max_vcpus();
> >
> >       set_default_spectre(kvm);
> > +     set_default_hypercalls(kvm);
> >
> >       return ret;
> >  out_free_stage2_pgd:
> > diff --git a/arch/arm64/kvm/hypercalls.c b/arch/arm64/kvm/hypercalls.c
> > index 7e873206a05b..0b3006353bf6 100644
> > --- a/arch/arm64/kvm/hypercalls.c
> > +++ b/arch/arm64/kvm/hypercalls.c
> > @@ -60,8 +60,64 @@ static void kvm_ptp_get_time(struct kvm_vcpu *vcpu, u64 *val)
> >
> >  static u64 *kvm_fw_reg_to_bmap(struct kvm *kvm, u64 fw_reg)
> >  {
> > -     /* No firmware registers supporting hvc bitmaps exits yet */
> > -     return NULL;
> > +     struct hvc_reg_desc *hvc_desc = &kvm->arch.hvc_desc;
> > +
> > +     switch (fw_reg) {
> > +     case KVM_REG_ARM_STD:
> > +             return &hvc_desc->kvm_std_bmap;
> > +     default:
> > +             return NULL;
> > +     }
> > +}
> > +
> > +struct kvm_hvc_func_map {
> > +     u32 func_id;
> > +     u64 bmap_bit;
> > +};
> > +
> > +#define HVC_FUNC_MAP_DESC(func, bit) \
> > +     {                                       \
> > +             .func_id = func,                \
> > +             .bmap_bit = bit,                \
> > +     }
> > +
> > +static const struct kvm_hvc_func_map hvc_std_map[] = {
> > +     HVC_FUNC_MAP_DESC(ARM_SMCCC_TRNG_GET_UUID, KVM_REG_ARM_STD_TRNG_V1_0),
> > +     HVC_FUNC_MAP_DESC(ARM_SMCCC_TRNG_RND32, KVM_REG_ARM_STD_TRNG_V1_0),
> > +     HVC_FUNC_MAP_DESC(ARM_SMCCC_TRNG_RND64, KVM_REG_ARM_STD_TRNG_V1_0),
> > +};
> > +
> > +bool kvm_hvc_call_supported(struct kvm_vcpu *vcpu, u32 func_id)
> > +{
> > +     struct kvm *kvm = vcpu->kvm;
> > +     u8 hvc_owner = ARM_SMCCC_OWNER_NUM(func_id);
> > +     const struct kvm_hvc_func_map *hvc_func_map = NULL;
> > +
> > +     u64 fw_reg, *hc_bmap;
> > +     unsigned int map_sz, i;
> > +
> > +     switch (hvc_owner) {
> > +     case ARM_SMCCC_OWNER_STANDARD:
> > +             fw_reg = KVM_REG_ARM_STD;
> > +             hvc_func_map = hvc_std_map;
> > +             map_sz = ARRAY_SIZE(hvc_std_map);
> > +             break;
> > +     default:
> > +             /* Allow all the owners that aren't mapped */
> > +             return true;
> > +     }
> > +
> > +     hc_bmap = kvm_fw_reg_to_bmap(kvm, fw_reg);
> > +     if (!hc_bmap)
> > +             return true;
> > +
> > +     for (i = 0; i < map_sz; i++) {
> > +             if (func_id == hvc_func_map[i].func_id)
> > +                     return *hc_bmap & BIT(hvc_func_map[i].bmap_bit);
> > +     }
>
> Hrm...
>
> Could you instead define a helper function for each service and use a
> switch statement to ensure each function tests the correct bit? This
> would avoid the need to loop over a map.
>
I think so.. I guess I was trying to avoid making too many changes if
we want to support a new func_id.
> > +
> > +     /* Allow all the functions of an owner that aren't mapped */
> > +     return true;
> >  }
> >
> >  int kvm_hvc_call_handler(struct kvm_vcpu *vcpu)
> > @@ -71,6 +127,9 @@ int kvm_hvc_call_handler(struct kvm_vcpu *vcpu)
> >       u32 feature;
> >       gpa_t gpa;
> >
> > +     if (!kvm_hvc_call_supported(vcpu, func_id))
> > +             goto out;
> > +
> >       switch (func_id) {
> >       case ARM_SMCCC_VERSION_FUNC_ID:
> >               val[0] = ARM_SMCCC_VERSION_1_1;
> > @@ -149,6 +208,7 @@ int kvm_hvc_call_handler(struct kvm_vcpu *vcpu)
> >               return kvm_psci_call(vcpu);
> >       }
> >
> > +out:
> >       smccc_set_retval(vcpu, val[0], val[1], val[2], val[3]);
> >       return 1;
> >  }
> > @@ -157,6 +217,7 @@ static const u64 fw_reg_ids[] = {
> >       KVM_REG_ARM_PSCI_VERSION,
> >       KVM_REG_ARM_SMCCC_ARCH_WORKAROUND_1,
> >       KVM_REG_ARM_SMCCC_ARCH_WORKAROUND_2,
> > +     KVM_REG_ARM_STD,
>
> This (and all other FW regs you add) need to be added to the
> get-reg-list selftest. Marc/Andrew have reminded me enough times to do
> this myself, so I'll share suggestion :-P
>
Yes, of course. It's on my todo list. I'll try to include that in the
next patchset.
> >  };
> >
> >  int kvm_arm_get_fw_num_regs(struct kvm_vcpu *vcpu)
> > @@ -219,6 +280,7 @@ static int get_kernel_wa_level(u64 regid)
> >
> >  int kvm_arm_get_fw_reg(struct kvm_vcpu *vcpu, const struct kvm_one_reg *reg)
> >  {
> > +     struct hvc_reg_desc *hvc_desc = &vcpu->kvm->arch.hvc_desc;
> >       void __user *uaddr = (void __user *)(long)reg->addr;
> >       u64 val;
> >
> > @@ -230,6 +292,9 @@ int kvm_arm_get_fw_reg(struct kvm_vcpu *vcpu, const struct kvm_one_reg *reg)
> >       case KVM_REG_ARM_SMCCC_ARCH_WORKAROUND_2:
> >               val = get_kernel_wa_level(reg->id) & KVM_REG_FEATURE_LEVEL_MASK;
> >               break;
> > +     case KVM_REG_ARM_STD:
> > +             val = hvc_desc->kvm_std_bmap;
> > +             break;
> >       default:
> >               return -ENOENT;
> >       }
> > @@ -352,6 +417,12 @@ int kvm_arm_set_fw_reg(struct kvm_vcpu *vcpu, const struct kvm_one_reg *reg)
> >               if (get_kernel_wa_level(reg->id) < wa_level)
> >                       return -EINVAL;
> >
> > +             return 0;
> > +     case KVM_REG_ARM_STD:
> > +             if (val & ~ARM_SMCCC_STD_FEATURES)
> > +                     return -EINVAL;
> > +
> > +             hvc_desc->kvm_std_bmap = val;
> >               return 0;
> >       default:
> >               return -ENOENT;
> > diff --git a/arch/arm64/kvm/trng.c b/arch/arm64/kvm/trng.c
> > index 99bdd7103c9c..6dff765f5b9b 100644
> > --- a/arch/arm64/kvm/trng.c
> > +++ b/arch/arm64/kvm/trng.c
> > @@ -60,14 +60,9 @@ int kvm_trng_call(struct kvm_vcpu *vcpu)
> >               val = ARM_SMCCC_TRNG_VERSION_1_0;
> >               break;
> >       case ARM_SMCCC_TRNG_FEATURES:
> > -             switch (smccc_get_arg1(vcpu)) {
> > -             case ARM_SMCCC_TRNG_VERSION:
> > -             case ARM_SMCCC_TRNG_FEATURES:
> > -             case ARM_SMCCC_TRNG_GET_UUID:
> > -             case ARM_SMCCC_TRNG_RND32:
> > -             case ARM_SMCCC_TRNG_RND64:
> > +             if (kvm_hvc_call_supported(vcpu, smccc_get_arg1(vcpu)))
> >                       val = TRNG_SUCCESS;
> > -             }
> > +
> >               break;
> >       case ARM_SMCCC_TRNG_GET_UUID:
> >               smccc_set_retval(vcpu, le32_to_cpu(u[0]), le32_to_cpu(u[1]),
> > diff --git a/include/kvm/arm_hypercalls.h b/include/kvm/arm_hypercalls.h
> > index 5d38628a8d04..5f01bb139312 100644
> > --- a/include/kvm/arm_hypercalls.h
> > +++ b/include/kvm/arm_hypercalls.h
> > @@ -6,6 +6,9 @@
> >
> >  #include <asm/kvm_emulate.h>
> >
> > +#define ARM_SMCCC_STD_FEATURES \
> > +     GENMASK_ULL(KVM_REG_ARM_STD_BMAP_MAX - 1, 0)
> > +
> >  int kvm_hvc_call_handler(struct kvm_vcpu *vcpu);
> >
> >  static inline u32 smccc_get_function(struct kvm_vcpu *vcpu)
> > @@ -47,4 +50,6 @@ int kvm_arm_copy_fw_reg_indices(struct kvm_vcpu *vcpu, u64 __user *uindices);
> >  int kvm_arm_get_fw_reg(struct kvm_vcpu *vcpu, const struct kvm_one_reg *reg);
> >  int kvm_arm_set_fw_reg(struct kvm_vcpu *vcpu, const struct kvm_one_reg *reg);
> >
> > +bool kvm_hvc_call_supported(struct kvm_vcpu *vcpu, u32 func_id);
> > +
> >  #endif
> > --
> > 2.33.1.1089.g2158813163f-goog
> >
_______________________________________________
kvmarm mailing list
kvmarm@lists.cs.columbia.edu
https://lists.cs.columbia.edu/mailman/listinfo/kvmarm