KVM: SVM: include CR3 in initial VMSA state for SEV-ES guests

Normally guests will set up CR3 themselves, but some guests, such as
kselftests, and potentially CONFIG_PVH guests, rely on being booted
with paging enabled and CR3 initialized to a pre-allocated page table.

Currently CR3 updates via KVM_SET_SREGS* are not loaded into the guest
VMCB until just prior to entering the guest. For SEV-ES/SEV-SNP, this
is too late, since it will have switched over to using the VMSA page
prior to that point, with the VMSA CR3 copied from the VMCB initial
CR3 value: 0.

Address this by sync'ing the CR3 value into the VMCB save area
immediately when KVM_SET_SREGS* is issued so it will find it's way into
the initial VMSA.

Suggested-by: Tom Lendacky <thomas.lendacky@amd.com>
Signed-off-by: Michael Roth <michael.roth@amd.com>
Message-Id: <20211216171358.61140-10-michael.roth@amd.com>
[Remove vmx_post_set_cr3; add a remark about kvm_set_cr3 not calling the
 new hook. - Paolo]
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
diff --git a/arch/x86/include/asm/kvm-x86-ops.h b/arch/x86/include/asm/kvm-x86-ops.h
index 9e50da3..f658bb4 100644
--- a/arch/x86/include/asm/kvm-x86-ops.h
+++ b/arch/x86/include/asm/kvm-x86-ops.h
@@ -35,6 +35,7 @@ KVM_X86_OP(get_cpl)
 KVM_X86_OP(set_segment)
 KVM_X86_OP_NULL(get_cs_db_l_bits)
 KVM_X86_OP(set_cr0)
+KVM_X86_OP_NULL(post_set_cr3)
 KVM_X86_OP(is_valid_cr4)
 KVM_X86_OP(set_cr4)
 KVM_X86_OP(set_efer)
diff --git a/arch/x86/include/asm/kvm_host.h b/arch/x86/include/asm/kvm_host.h
index 623fb7c..6cbf97a 100644
--- a/arch/x86/include/asm/kvm_host.h
+++ b/arch/x86/include/asm/kvm_host.h
@@ -1344,6 +1344,7 @@ struct kvm_x86_ops {
 			    struct kvm_segment *var, int seg);
 	void (*get_cs_db_l_bits)(struct kvm_vcpu *vcpu, int *db, int *l);
 	void (*set_cr0)(struct kvm_vcpu *vcpu, unsigned long cr0);
+	void (*post_set_cr3)(struct kvm_vcpu *vcpu, unsigned long cr3);
 	bool (*is_valid_cr4)(struct kvm_vcpu *vcpu, unsigned long cr0);
 	void (*set_cr4)(struct kvm_vcpu *vcpu, unsigned long cr4);
 	int (*set_efer)(struct kvm_vcpu *vcpu, u64 efer);
diff --git a/arch/x86/kvm/svm/svm.c b/arch/x86/kvm/svm/svm.c
index 5557867..c3d9006 100644
--- a/arch/x86/kvm/svm/svm.c
+++ b/arch/x86/kvm/svm/svm.c
@@ -1801,6 +1801,24 @@ static void svm_set_gdt(struct kvm_vcpu *vcpu, struct desc_ptr *dt)
 	vmcb_mark_dirty(svm->vmcb, VMCB_DT);
 }
 
+static void svm_post_set_cr3(struct kvm_vcpu *vcpu, unsigned long cr3)
+{
+	struct vcpu_svm *svm = to_svm(vcpu);
+
+	/*
+	 * For guests that don't set guest_state_protected, the cr3 update is
+	 * handled via kvm_mmu_load() while entering the guest. For guests
+	 * that do (SEV-ES/SEV-SNP), the cr3 update needs to be written to
+	 * VMCB save area now, since the save area will become the initial
+	 * contents of the VMSA, and future VMCB save area updates won't be
+	 * seen.
+	 */
+	if (sev_es_guest(vcpu->kvm)) {
+		svm->vmcb->save.cr3 = cr3;
+		vmcb_mark_dirty(svm->vmcb, VMCB_CR);
+	}
+}
+
 void svm_set_cr0(struct kvm_vcpu *vcpu, unsigned long cr0)
 {
 	struct vcpu_svm *svm = to_svm(vcpu);
@@ -4624,6 +4642,7 @@ static struct kvm_x86_ops svm_x86_ops __initdata = {
 	.get_cpl = svm_get_cpl,
 	.get_cs_db_l_bits = kvm_get_cs_db_l_bits,
 	.set_cr0 = svm_set_cr0,
+	.post_set_cr3 = svm_post_set_cr3,
 	.is_valid_cr4 = svm_is_valid_cr4,
 	.set_cr4 = svm_set_cr4,
 	.set_efer = svm_set_efer,
diff --git a/arch/x86/kvm/vmx/vmx.c b/arch/x86/kvm/vmx/vmx.c
index 1d53b814..7b5abe2 100644
--- a/arch/x86/kvm/vmx/vmx.c
+++ b/arch/x86/kvm/vmx/vmx.c
@@ -3134,6 +3134,7 @@ static void vmx_load_mmu_pgd(struct kvm_vcpu *vcpu, hpa_t root_hpa,
 		vmcs_writel(GUEST_CR3, guest_cr3);
 }
 
+
 static bool vmx_is_valid_cr4(struct kvm_vcpu *vcpu, unsigned long cr4)
 {
 	/*
diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c
index 6492329f..a9f1044 100644
--- a/arch/x86/kvm/x86.c
+++ b/arch/x86/kvm/x86.c
@@ -1179,6 +1179,7 @@ int kvm_set_cr3(struct kvm_vcpu *vcpu, unsigned long cr3)
 
 	vcpu->arch.cr3 = cr3;
 	kvm_register_mark_dirty(vcpu, VCPU_EXREG_CR3);
+	/* Do not call post_set_cr3, we do not get here for confidential guests.  */
 
 handle_tlb_flush:
 	/*
@@ -10618,6 +10619,7 @@ static int __set_sregs_common(struct kvm_vcpu *vcpu, struct kvm_sregs *sregs,
 	*mmu_reset_needed |= kvm_read_cr3(vcpu) != sregs->cr3;
 	vcpu->arch.cr3 = sregs->cr3;
 	kvm_register_mark_dirty(vcpu, VCPU_EXREG_CR3);
+	static_call_cond(kvm_x86_post_set_cr3)(vcpu, sregs->cr3);
 
 	kvm_set_cr8(vcpu, sregs->cr8);