[v3,8/9] KVM: TDX: Handle TDX PV port I/O hypercall

Commit Message

Binbin Wu Feb. 22, 2025, 1:42 a.m. UTC

From: Isaku Yamahata <[e-mail]>

Emulate port I/O requested by TDX guest via TDVMCALL with leaf
Instruction.IO (same value as EXIT_REASON_IO_INSTRUCTION) according to
TDX Guest Host Communication Interface (GHCI).

All port I/O instructions inside the TDX guest trigger the #VE exception.
On #VE triggered by I/O instructions, TDX guest can call TDVMCALL with
leaf Instruction.IO to request VMM to emulate I/O instructions.

Similar to normal port I/O emulation, try to handle the port I/O in kernel
first, if kernel can't support it, forward the request to userspace.

Note string I/O operations are not supported in TDX.  Guest should unroll
them before calling the TDVMCALL.

Suggested-by: Sean Christopherson <[e-mail]>
Signed-off-by: Isaku Yamahata <[e-mail]>
Co-developed-by: Binbin Wu <[e-mail]>
Signed-off-by: Binbin Wu <[e-mail]>
Reviewed-by: Paolo Bonzini <[e-mail]>
---
Hypercalls exit to userspace v3:
- Rebased to use tdcall_to_vmx_exit_reason().

Hypercalls exit to userspace v2:
- Morph PV port I/O hypercall to EXIT_REASON_IO_INSTRUCTION. (Sean)
- Use vp_enter_args instead of x86 registers.
- Check write is either 0 or 1. (Chao)
- Skip setting return code as TDVMCALL_STATUS_SUCCESS. (Sean)

Hypercalls exit to userspace v1:
- Renamed from "KVM: TDX: Handle TDX PV port io hypercall" to
  "KVM: TDX: Handle TDX PV port I/O hypercall".
- Update changelog.
- Add missing curly brackets.
- Move reset of pio.count to tdx_complete_pio_out() and remove the stale
  comment. (binbin)
- Use TDVMCALL_STATUS prefix for TDX call status codes (Binbin)
- Set status code to TDVMCALL_STATUS_SUCCESS when PIO is handled in kernel.
- Don't write to R11 when it is a write operation for output.

v18:
- Fix out case to set R10 and R11 correctly when user space handled port
  out.
---
 arch/x86/kvm/vmx/tdx.c | 62 ++++++++++++++++++++++++++++++++++++++++++
 1 file changed, 62 insertions(+)

Patch

diff --git a/arch/x86/kvm/vmx/tdx.c b/arch/x86/kvm/vmx/tdx.c
index 7c8356299a25..3a4437520868 100644
--- a/arch/x86/kvm/vmx/tdx.c
+++ b/arch/x86/kvm/vmx/tdx.c
@@ -824,6 +824,8 @@  static bool tdx_guest_state_is_invalid(struct kvm_vcpu *vcpu)
 static __always_inline u32 tdcall_to_vmx_exit_reason(struct kvm_vcpu *vcpu)
 {
 	switch (tdvmcall_leaf(vcpu)) {
+	case EXIT_REASON_IO_INSTRUCTION:
+		return tdvmcall_leaf(vcpu);
 	default:
 		break;
 	}
@@ -1119,6 +1121,64 @@  static int tdx_report_fatal_error(struct kvm_vcpu *vcpu)
 	return 0;
 }
 
+static int tdx_complete_pio_out(struct kvm_vcpu *vcpu)
+{
+	vcpu->arch.pio.count = 0;
+	return 1;
+}
+
+static int tdx_complete_pio_in(struct kvm_vcpu *vcpu)
+{
+	struct x86_emulate_ctxt *ctxt = vcpu->arch.emulate_ctxt;
+	unsigned long val = 0;
+	int ret;
+
+	ret = ctxt->ops->pio_in_emulated(ctxt, vcpu->arch.pio.size,
+					 vcpu->arch.pio.port, &val, 1);
+
+	WARN_ON_ONCE(!ret);
+
+	tdvmcall_set_return_val(vcpu, val);
+
+	return 1;
+}
+
+static int tdx_emulate_io(struct kvm_vcpu *vcpu)
+{
+	struct vcpu_tdx *tdx = to_tdx(vcpu);
+	struct x86_emulate_ctxt *ctxt = vcpu->arch.emulate_ctxt;
+	unsigned long val = 0;
+	unsigned int port;
+	u64 size, write;
+	int ret;
+
+	++vcpu->stat.io_exits;
+
+	size = tdx->vp_enter_args.r12;
+	write = tdx->vp_enter_args.r13;
+	port = tdx->vp_enter_args.r14;
+
+	if ((write != 0 && write != 1) || (size != 1 && size != 2 && size != 4)) {
+		tdvmcall_set_return_code(vcpu, TDVMCALL_STATUS_INVALID_OPERAND);
+		return 1;
+	}
+
+	if (write) {
+		val = tdx->vp_enter_args.r15;
+		ret = ctxt->ops->pio_out_emulated(ctxt, size, port, &val, 1);
+	} else {
+		ret = ctxt->ops->pio_in_emulated(ctxt, size, port, &val, 1);
+	}
+
+	if (!ret)
+		vcpu->arch.complete_userspace_io = write ? tdx_complete_pio_out :
+							   tdx_complete_pio_in;
+	else if (!write)
+		tdvmcall_set_return_val(vcpu, val);
+
+	return ret;
+}
+
 static int handle_tdvmcall(struct kvm_vcpu *vcpu)
 {
 	switch (tdvmcall_leaf(vcpu)) {
@@ -1496,6 +1556,8 @@  int tdx_handle_exit(struct kvm_vcpu *vcpu, fastpath_t fastpath)
 		return handle_tdvmcall(vcpu);
 	case EXIT_REASON_VMCALL:
 		return tdx_emulate_vmcall(vcpu);
+	case EXIT_REASON_IO_INSTRUCTION:
+		return tdx_emulate_io(vcpu);
 	default:
 		break;
 	}