Re: [PATCH v13 30/48] KVM: arm64: Handle Realm PSCI requests

[Suzuki K Poulose <suzuki.poulose@arm.com>]

On 18/03/2026 15:53, Steven Price wrote:
> The RMM needs to be informed of the target REC when a PSCI call is made
> with an MPIDR argument. Expose an ioctl to the userspace in case the PSCI
> is handled by it.
> 
> [NOTE: A future version of the RMM specification is likely to remove the
> need for this ioctl.]

This will need to stay for the PSCI_CPU_ON case, where the host has to
acknowledge the onlining of a vCPU.

> 
> Co-developed-by: Suzuki K Poulose <suzuki.poulose@arm.com>
> Signed-off-by: Suzuki K Poulose <suzuki.poulose@arm.com>
> Signed-off-by: Steven Price <steven.price@arm.com>

For the record, we can drop the UAPI following our discussions and 
implicitly do the PSCI complete before REC_ENTER, similar to what
we do for the SET_RIPAS request. The VMM/KVM can treat the case
as a normal PSCI_CPU_ON request and return the result as in normal
VMs.

The target REC may ENTER before we complete the reporting, but we can
handle this error case (RMI_ERROR_REC) and return -EAGAIN to the
userspace.

Suzuki


> ---
> Changes since v12:
>   * Chance return code for non-realms to -ENXIO to better represent that
>     the ioctl is invalid for non-realms (checkpatch is insistent that
>     "ENOSYS means 'invalid syscall nr' and nothing else").
> Changes since v11:
>   * RMM->RMI renaming.
> Changes since v6:
>   * Use vcpu_is_rec() rather than kvm_is_realm(vcpu->kvm).
>   * Minor renaming/formatting fixes.
> ---
>   arch/arm64/include/asm/kvm_rmi.h |  3 +++
>   arch/arm64/kvm/arm.c             | 25 +++++++++++++++++++++++++
>   arch/arm64/kvm/psci.c            | 30 ++++++++++++++++++++++++++++++
>   arch/arm64/kvm/rmi.c             | 14 ++++++++++++++
>   4 files changed, 72 insertions(+)
> 
> diff --git a/arch/arm64/include/asm/kvm_rmi.h b/arch/arm64/include/asm/kvm_rmi.h
> index 38208be3c602..1ee5ed0f5ab2 100644
> --- a/arch/arm64/include/asm/kvm_rmi.h
> +++ b/arch/arm64/include/asm/kvm_rmi.h
> @@ -117,6 +117,9 @@ int realm_map_non_secure(struct realm *realm,
>   			 unsigned long size,
>   			 enum kvm_pgtable_prot prot,
>   			 struct kvm_mmu_memory_cache *memcache);
> +int realm_psci_complete(struct kvm_vcpu *source,
> +			struct kvm_vcpu *target,
> +			unsigned long status);
>   
>   static inline bool kvm_realm_is_private_address(struct realm *realm,
>   						unsigned long addr)
> diff --git a/arch/arm64/kvm/arm.c b/arch/arm64/kvm/arm.c
> index 304fb1f2b3ff..61182eb0cf70 100644
> --- a/arch/arm64/kvm/arm.c
> +++ b/arch/arm64/kvm/arm.c
> @@ -1846,6 +1846,22 @@ static int kvm_arm_vcpu_set_events(struct kvm_vcpu *vcpu,
>   	return __kvm_arm_vcpu_set_events(vcpu, events);
>   }
>   
> +static int kvm_arm_vcpu_rmi_psci_complete(struct kvm_vcpu *vcpu,
> +					  struct kvm_arm_rmi_psci_complete *arg)
> +{
> +	struct kvm_vcpu *target = kvm_mpidr_to_vcpu(vcpu->kvm, arg->target_mpidr);
> +
> +	if (!target)
> +		return -EINVAL;
> +
> +	/*
> +	 * RMM v1.0 only supports PSCI_RET_SUCCESS or PSCI_RET_DENIED
> +	 * for the status. But, let us leave it to the RMM to filter
> +	 * for making this future proof.
> +	 */
> +	return realm_psci_complete(vcpu, target, arg->psci_status);
> +}
> +
>   long kvm_arch_vcpu_ioctl(struct file *filp,
>   			 unsigned int ioctl, unsigned long arg)
>   {
> @@ -1974,6 +1990,15 @@ long kvm_arch_vcpu_ioctl(struct file *filp,
>   
>   		return kvm_arm_vcpu_finalize(vcpu, what);
>   	}
> +	case KVM_ARM_VCPU_RMI_PSCI_COMPLETE: {
> +		struct kvm_arm_rmi_psci_complete req;
> +
> +		if (!vcpu_is_rec(vcpu))
> +			return -ENXIO;
> +		if (copy_from_user(&req, argp, sizeof(req)))
> +			return -EFAULT;
> +		return kvm_arm_vcpu_rmi_psci_complete(vcpu, &req);
> +	}
>   	default:
>   		r = -EINVAL;
>   	}
> diff --git a/arch/arm64/kvm/psci.c b/arch/arm64/kvm/psci.c
> index 3b5dbe9a0a0e..a68f3c1878a5 100644
> --- a/arch/arm64/kvm/psci.c
> +++ b/arch/arm64/kvm/psci.c
> @@ -103,6 +103,12 @@ static unsigned long kvm_psci_vcpu_on(struct kvm_vcpu *source_vcpu)
>   
>   	reset_state->reset = true;
>   	kvm_make_request(KVM_REQ_VCPU_RESET, vcpu);
> +	/*
> +	 * Make sure we issue PSCI_COMPLETE before the VCPU can be
> +	 * scheduled.
> +	 */
> +	if (vcpu_is_rec(vcpu))
> +		realm_psci_complete(source_vcpu, vcpu, PSCI_RET_SUCCESS);
>   
>   	/*
>   	 * Make sure the reset request is observed if the RUNNABLE mp_state is
> @@ -115,6 +121,11 @@ static unsigned long kvm_psci_vcpu_on(struct kvm_vcpu *source_vcpu)
>   
>   out_unlock:
>   	spin_unlock(&vcpu->arch.mp_state_lock);
> +	if (vcpu_is_rec(vcpu) && ret != PSCI_RET_SUCCESS) {
> +		realm_psci_complete(source_vcpu, vcpu,
> +				    ret == PSCI_RET_ALREADY_ON ?
> +				    PSCI_RET_SUCCESS : PSCI_RET_DENIED);
> +	}
>   	return ret;
>   }
>   
> @@ -142,6 +153,25 @@ static unsigned long kvm_psci_vcpu_affinity_info(struct kvm_vcpu *vcpu)
>   	/* Ignore other bits of target affinity */
>   	target_affinity &= target_affinity_mask;
>   
> +	if (vcpu_is_rec(vcpu)) {
> +		struct kvm_vcpu *target_vcpu;
> +
> +		/* RMM supports only zero affinity level */
> +		if (lowest_affinity_level != 0)
> +			return PSCI_RET_INVALID_PARAMS;
> +
> +		target_vcpu = kvm_mpidr_to_vcpu(kvm, target_affinity);
> +		if (!target_vcpu)
> +			return PSCI_RET_INVALID_PARAMS;
> +
> +		/*
> +		 * Provide the references of the source and target RECs to the
> +		 * RMM so that the RMM can complete the PSCI request.
> +		 */
> +		realm_psci_complete(vcpu, target_vcpu, PSCI_RET_SUCCESS);
> +		return PSCI_RET_SUCCESS;
> +	}
> +
>   	/*
>   	 * If one or more VCPU matching target affinity are running
>   	 * then ON else OFF
> diff --git a/arch/arm64/kvm/rmi.c b/arch/arm64/kvm/rmi.c
> index 30292814b1ec..e56c8af2ad61 100644
> --- a/arch/arm64/kvm/rmi.c
> +++ b/arch/arm64/kvm/rmi.c
> @@ -353,6 +353,20 @@ static void free_rtt(phys_addr_t phys)
>   	kvm_account_pgtable_pages(phys_to_virt(phys), -1);
>   }
>   
> +int realm_psci_complete(struct kvm_vcpu *source, struct kvm_vcpu *target,
> +			unsigned long status)
> +{
> +	int ret;
> +
> +	ret = rmi_psci_complete(virt_to_phys(source->arch.rec.rec_page),
> +				virt_to_phys(target->arch.rec.rec_page),
> +				status);
> +	if (ret)
> +		return -EINVAL;
> +
> +	return 0;
> +}
> +
>   static int realm_rtt_create(struct realm *realm,
>   			    unsigned long addr,
>   			    int level,