[PATCH v2 20/28] KVM: arm64: Reject non-coherent DMA when using forced writeback

[Oliver Upton <oupton@kernel.org>]

Assigning non-coherent devices to a KVM guest subject to forced
writeback is broken. DMA through cacheable aliases and CMOs work, but
using non-cacheable aliases for buffers on the CPU does not due to the
implicit promotion to Normal writeback.

Non-coherent DMA is fundamentally incompatible with the direction of
travel on KVM/arm64. Draw the line in the sand and bug the VM if FWB is
in use. Users can use the newly-added kernel commandline to fall back to
a non-FWB configuration that *actually works*.

Signed-off-by: Oliver Upton <oupton@kernel.org>
---
 arch/arm64/include/asm/kvm_host.h |  2 ++
 arch/arm64/kvm/arm.c              | 14 ++++++++++++++
 2 files changed, 16 insertions(+)

diff --git a/arch/arm64/include/asm/kvm_host.h b/arch/arm64/include/asm/kvm_host.h
index 1c6632216e32..5672721416a4 100644
--- a/arch/arm64/include/asm/kvm_host.h
+++ b/arch/arm64/include/asm/kvm_host.h
@@ -1565,6 +1565,8 @@ struct kvm *kvm_arch_alloc_vm(void);
 
 #define __KVM_HAVE_ARCH_FLUSH_REMOTE_TLBS_RANGE
 
+#define __KVM_HAVE_ARCH_NONCOHERENT_DMA
+
 #ifdef __KVM_NVHE_HYPERVISOR__
 
 #define kvm_vm_is_protected(kvm)			\
diff --git a/arch/arm64/kvm/arm.c b/arch/arm64/kvm/arm.c
index d593ba928093..f08a450184d9 100644
--- a/arch/arm64/kvm/arm.c
+++ b/arch/arm64/kvm/arm.c
@@ -133,6 +133,20 @@ int kvm_arch_vcpu_should_kick(struct kvm_vcpu *vcpu)
 	return kvm_vcpu_exiting_guest_mode(vcpu) == IN_GUEST_MODE;
 }
 
+void kvm_arch_register_noncoherent_dma(struct kvm *kvm)
+{
+	if (!cpus_have_final_cap(ARM64_HAS_STAGE2_FWB))
+		return;
+
+	pr_err_once("kvm: noncoherent DMA unsupported with FWB, see kvm_arm.nofwb\n");
+	kvm_vm_bugged(kvm);
+}
+
+void kvm_arch_unregister_noncoherent_dma(struct kvm *kvm)
+{
+
+}
+
 int kvm_vm_ioctl_enable_cap(struct kvm *kvm,
 			    struct kvm_enable_cap *cap)
 {
-- 
2.47.3