[PATCH v4 1/6] KVM: x86: introduce ISA specific SMM entry/exit callbacks

[Ladi Prosek <lprosek@redhat.com>]

Entering and exiting SMM may require ISA specific handling under certain
circumstances. This commit adds two new callbacks with empty implementations.
Actual functionality will be added in following commits.

* prep_enter_smm() is to be called when injecting an SMM, before any
  SMM related vcpu state has been changed
* post_leave_smm() is to be called when emulating the RSM instruction,
  after all SMM related vcpu state has been restored; the function may
  ask the caller to restore the vcpu state again via the reload_state
  parameter

Signed-off-by: Ladi Prosek <lprosek@redhat.com>
---
 arch/x86/include/asm/kvm_emulate.h |  3 +++
 arch/x86/include/asm/kvm_host.h    |  4 ++++
 arch/x86/kvm/emulate.c             | 30 ++++++++++++++++++++++++++----
 arch/x86/kvm/svm.c                 | 16 ++++++++++++++++
 arch/x86/kvm/vmx.c                 | 16 ++++++++++++++++
 arch/x86/kvm/x86.c                 | 17 +++++++++++++++++
 6 files changed, 82 insertions(+), 4 deletions(-)

diff --git a/arch/x86/include/asm/kvm_emulate.h b/arch/x86/include/asm/kvm_emulate.h
index fa2558e12024..0ba3837173fb 100644
--- a/arch/x86/include/asm/kvm_emulate.h
+++ b/arch/x86/include/asm/kvm_emulate.h
@@ -225,6 +225,9 @@ struct x86_emulate_ops {
 
 	unsigned (*get_hflags)(struct x86_emulate_ctxt *ctxt);
 	void (*set_hflags)(struct x86_emulate_ctxt *ctxt, unsigned hflags);
+	int (*post_leave_smm)(struct x86_emulate_ctxt *ctxt, u64 smbase,
+			      bool *reload_state);
+
 };
 
 typedef u32 __attribute__((vector_size(16))) sse128_t;
diff --git a/arch/x86/include/asm/kvm_host.h b/arch/x86/include/asm/kvm_host.h
index c73e493adf07..769de6d2e684 100644
--- a/arch/x86/include/asm/kvm_host.h
+++ b/arch/x86/include/asm/kvm_host.h
@@ -1061,6 +1061,10 @@ struct kvm_x86_ops {
 	void (*cancel_hv_timer)(struct kvm_vcpu *vcpu);
 
 	void (*setup_mce)(struct kvm_vcpu *vcpu);
+
+	int (*prep_enter_smm)(struct kvm_vcpu *vcpu, char *smstate);
+	int (*post_leave_smm)(struct kvm_vcpu *vcpu, u64 smbase,
+			      bool *reload_state);
 };
 
 struct kvm_arch_async_pf {
diff --git a/arch/x86/kvm/emulate.c b/arch/x86/kvm/emulate.c
index d90cdc77e077..1e6a8a824b8b 100644
--- a/arch/x86/kvm/emulate.c
+++ b/arch/x86/kvm/emulate.c
@@ -2546,9 +2546,18 @@ static int rsm_load_state_64(struct x86_emulate_ctxt *ctxt, u64 smbase)
 	return X86EMUL_CONTINUE;
 }
 
+static int rsm_load_state(struct x86_emulate_ctxt *ctxt, u64 smbase)
+{
+	if (emulator_has_longmode(ctxt))
+		return rsm_load_state_64(ctxt, smbase + 0x8000);
+	else
+		return rsm_load_state_32(ctxt, smbase + 0x8000);
+}
+
 static int em_rsm(struct x86_emulate_ctxt *ctxt)
 {
 	unsigned long cr0, cr4, efer;
+	bool reload_state = false;
 	u64 smbase;
 	int ret;
 
@@ -2591,16 +2600,29 @@ static int em_rsm(struct x86_emulate_ctxt *ctxt)
 	ctxt->ops->set_msr(ctxt, MSR_EFER, efer);
 
 	smbase = ctxt->ops->get_smbase(ctxt);
-	if (emulator_has_longmode(ctxt))
-		ret = rsm_load_state_64(ctxt, smbase + 0x8000);
-	else
-		ret = rsm_load_state_32(ctxt, smbase + 0x8000);
 
+	ret = rsm_load_state(ctxt, smbase);
 	if (ret != X86EMUL_CONTINUE) {
 		/* FIXME: should triple fault */
 		return X86EMUL_UNHANDLEABLE;
 	}
 
+	if (ctxt->ops->post_leave_smm(ctxt, smbase, &reload_state))
+		return X86EMUL_UNHANDLEABLE;
+
+	if (reload_state) {
+		/*
+		 * post_leave_smm() made changes to the vCPU (e.g. entered
+		 * guest mode) and is asking us to load the SMM state-save
+		 * area again.
+		 */
+		ret = rsm_load_state(ctxt, smbase);
+		if (ret != X86EMUL_CONTINUE) {
+			/* FIXME: should triple fault */
+			return X86EMUL_UNHANDLEABLE;
+		}
+	}
+
 	if ((ctxt->ops->get_hflags(ctxt) & X86EMUL_SMM_INSIDE_NMI_MASK) == 0)
 		ctxt->ops->set_nmi_mask(ctxt, false);
 
diff --git a/arch/x86/kvm/svm.c b/arch/x86/kvm/svm.c
index 0e68f0b3cbf7..d9b3e1bea644 100644
--- a/arch/x86/kvm/svm.c
+++ b/arch/x86/kvm/svm.c
@@ -5393,6 +5393,19 @@ static void svm_setup_mce(struct kvm_vcpu *vcpu)
 	vcpu->arch.mcg_cap &= 0x1ff;
 }
 
+static int svm_prep_enter_smm(struct kvm_vcpu *vcpu, char *smstate)
+{
+	/* TODO: Implement */
+	return 0;
+}
+
+static int svm_post_leave_smm(struct kvm_vcpu *vcpu, u64 smbase,
+			      bool *reload_state)
+{
+	/* TODO: Implement */
+	return 0;
+}
+
 static struct kvm_x86_ops svm_x86_ops __ro_after_init = {
 	.cpu_has_kvm_support = has_svm,
 	.disabled_by_bios = is_disabled,
@@ -5503,6 +5516,9 @@ static struct kvm_x86_ops svm_x86_ops __ro_after_init = {
 	.deliver_posted_interrupt = svm_deliver_avic_intr,
 	.update_pi_irte = svm_update_pi_irte,
 	.setup_mce = svm_setup_mce,
+
+	.prep_enter_smm = svm_prep_enter_smm,
+	.post_leave_smm = svm_post_leave_smm,
 };
 
 static int __init svm_init(void)
diff --git a/arch/x86/kvm/vmx.c b/arch/x86/kvm/vmx.c
index a2b804e10c95..15478f413392 100644
--- a/arch/x86/kvm/vmx.c
+++ b/arch/x86/kvm/vmx.c
@@ -11941,6 +11941,19 @@ static void vmx_setup_mce(struct kvm_vcpu *vcpu)
 			~FEATURE_CONTROL_LMCE;
 }
 
+static int vmx_prep_enter_smm(struct kvm_vcpu *vcpu, char *smstate)
+{
+	/* TODO: Implement */
+	return 0;
+}
+
+static int vmx_post_leave_smm(struct kvm_vcpu *vcpu, u64 smbase,
+			      bool *reload_state)
+{
+	/* TODO: Implement */
+	return 0;
+}
+
 static struct kvm_x86_ops vmx_x86_ops __ro_after_init = {
 	.cpu_has_kvm_support = cpu_has_kvm_support,
 	.disabled_by_bios = vmx_disabled_by_bios,
@@ -12066,6 +12079,9 @@ static struct kvm_x86_ops vmx_x86_ops __ro_after_init = {
 #endif
 
 	.setup_mce = vmx_setup_mce,
+
+	.prep_enter_smm = vmx_prep_enter_smm,
+	.post_leave_smm = vmx_post_leave_smm,
 };
 
 static int __init vmx_init(void)
diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c
index 03869eb7fcd6..e9aef1d858a8 100644
--- a/arch/x86/kvm/x86.c
+++ b/arch/x86/kvm/x86.c
@@ -5275,6 +5275,12 @@ static void emulator_set_hflags(struct x86_emulate_ctxt *ctxt, unsigned emul_fla
 	kvm_set_hflags(emul_to_vcpu(ctxt), emul_flags);
 }
 
+static int emulator_post_leave_smm(struct x86_emulate_ctxt *ctxt, u64 smbase,
+				   bool *reload_state)
+{
+	return kvm_x86_ops->post_leave_smm(emul_to_vcpu(ctxt), smbase, reload_state);
+}
+
 static const struct x86_emulate_ops emulate_ops = {
 	.read_gpr            = emulator_read_gpr,
 	.write_gpr           = emulator_write_gpr,
@@ -5316,6 +5322,7 @@ static const struct x86_emulate_ops emulate_ops = {
 	.set_nmi_mask        = emulator_set_nmi_mask,
 	.get_hflags          = emulator_get_hflags,
 	.set_hflags          = emulator_set_hflags,
+	.post_leave_smm      = emulator_post_leave_smm,
 };
 
 static void toggle_interruptibility(struct kvm_vcpu *vcpu, u32 mask)
@@ -6643,6 +6650,7 @@ static void enter_smm(struct kvm_vcpu *vcpu)
 	trace_kvm_enter_smm(vcpu->vcpu_id, vcpu->arch.smbase, true);
 	vcpu->arch.hflags |= HF_SMM_MASK;
 	memset(buf, 0, 512);
+
 	if (guest_cpuid_has(vcpu, X86_FEATURE_LM))
 		enter_smm_save_state_64(vcpu, buf);
 	else
@@ -6650,6 +6658,15 @@ static void enter_smm(struct kvm_vcpu *vcpu)
 
 	kvm_vcpu_write_guest(vcpu, vcpu->arch.smbase + 0xfe00, buf, sizeof(buf));
 
+	/*
+	 * Give prep_enter_smm() a chance to make ISA-specific changes to the
+	 * vCPU state (e.g. leave guest mode) after we've saved the state into
+	 * the SMM state-save area. Clear HF_SMM_MASK temporarily.
+	 */
+	vcpu->arch.hflags &= ~HF_SMM_MASK;
+	kvm_x86_ops->prep_enter_smm(vcpu, buf);
+	vcpu->arch.hflags |= HF_SMM_MASK;
+
 	if (kvm_x86_ops->get_nmi_mask(vcpu))
 		vcpu->arch.hflags |= HF_SMM_INSIDE_NMI_MASK;
 	else
-- 
2.13.5