Merge remote-tracking branch 'kvm/master' into HEAD

Pick commit fdba608f15e2 ("KVM: VMX: Wake vCPU when delivering posted
IRQ even if vCPU == this vCPU").  In addition to fixing a bug, it
also aligns the non-nested and nested usage of triggering posted
interrupts, allowing for additional cleanups.

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

20 files changed, 412 insertions, 73 deletions

diff --git a/Documentation/admin-guide/kernel-parameters.txt b/Documentation/admin-guide/kernel-parameters.txt
index 2711ddb4835a..4eaefb0983a8 100644
--- a/Documentation/admin-guide/kernel-parameters.txt
+++ b/Documentation/admin-guide/kernel-parameters.txt
@@ -2413,8 +2413,12 @@
 			Default is 1 (enabled)
 
 	kvm-intel.emulate_invalid_guest_state=
-			[KVM,Intel] Enable emulation of invalid guest states
-			Default is 0 (disabled)
+			[KVM,Intel] Disable emulation of invalid guest state.
+			Ignored if kvm-intel.enable_unrestricted_guest=1, as
+			guest state is never invalid for unrestricted guests.
+			This param doesn't apply to nested guests (L2), as KVM
+			never emulates invalid L2 guest state.
+			Default is 1 (enabled)
 
 	kvm-intel.flexpriority=
 			[KVM,Intel] Disable FlexPriority feature (TPR shadow).
diff --git a/arch/x86/include/asm/kvm-x86-ops.h b/arch/x86/include/asm/kvm-x86-ops.h
index cefe1d81e2e8..9e50da3ed01a 100644
--- a/arch/x86/include/asm/kvm-x86-ops.h
+++ b/arch/x86/include/asm/kvm-x86-ops.h
@@ -47,6 +47,7 @@ KVM_X86_OP(set_dr7)
 KVM_X86_OP(cache_reg)
 KVM_X86_OP(get_rflags)
 KVM_X86_OP(set_rflags)
+KVM_X86_OP(get_if_flag)
 KVM_X86_OP(tlb_flush_all)
 KVM_X86_OP(tlb_flush_current)
 KVM_X86_OP_NULL(tlb_remote_flush)
diff --git a/arch/x86/include/asm/kvm_host.h b/arch/x86/include/asm/kvm_host.h
index d5fede05eb5f..344ced0b6b68 100644
--- a/arch/x86/include/asm/kvm_host.h
+++ b/arch/x86/include/asm/kvm_host.h
@@ -97,7 +97,7 @@
 	KVM_ARCH_REQ_FLAGS(25, KVM_REQUEST_WAIT | KVM_REQUEST_NO_WAKEUP)
 #define KVM_REQ_TLB_FLUSH_CURRENT	KVM_ARCH_REQ(26)
 #define KVM_REQ_TLB_FLUSH_GUEST \
-	KVM_ARCH_REQ_FLAGS(27, KVM_REQUEST_NO_WAKEUP)
+	KVM_ARCH_REQ_FLAGS(27, KVM_REQUEST_WAIT | KVM_REQUEST_NO_WAKEUP)
 #define KVM_REQ_APF_READY		KVM_ARCH_REQ(28)
 #define KVM_REQ_MSR_FILTER_CHANGED	KVM_ARCH_REQ(29)
 #define KVM_REQ_UPDATE_CPU_DIRTY_LOGGING \
@@ -1354,6 +1354,7 @@ struct kvm_x86_ops {
 	void (*cache_reg)(struct kvm_vcpu *vcpu, enum kvm_reg reg);
 	unsigned long (*get_rflags)(struct kvm_vcpu *vcpu);
 	void (*set_rflags)(struct kvm_vcpu *vcpu, unsigned long rflags);
+	bool (*get_if_flag)(struct kvm_vcpu *vcpu);
 
 	void (*tlb_flush_all)(struct kvm_vcpu *vcpu);
 	void (*tlb_flush_current)(struct kvm_vcpu *vcpu);
diff --git a/arch/x86/kvm/hyperv.c b/arch/x86/kvm/hyperv.c
index a91424ed436d..6e38a7d22e97 100644
--- a/arch/x86/kvm/hyperv.c
+++ b/arch/x86/kvm/hyperv.c
@@ -1923,11 +1923,13 @@ static u64 kvm_hv_send_ipi(struct kvm_vcpu *vcpu, struct kvm_hv_hcall *hc, bool
 
 		all_cpus = send_ipi_ex.vp_set.format == HV_GENERIC_SET_ALL;
 
+		if (all_cpus)
+			goto check_and_send_ipi;
+
 		if (!sparse_banks_len)
 			goto ret_success;
 
-		if (!all_cpus &&
-		    kvm_read_guest(kvm,
+		if (kvm_read_guest(kvm,
 				   hc->ingpa + offsetof(struct hv_send_ipi_ex,
 							vp_set.bank_contents),
 				   sparse_banks,
@@ -1935,6 +1937,7 @@ static u64 kvm_hv_send_ipi(struct kvm_vcpu *vcpu, struct kvm_hv_hcall *hc, bool
 			return HV_STATUS_INVALID_HYPERCALL_INPUT;
 	}
 
+check_and_send_ipi:
 	if ((vector < HV_IPI_LOW_VECTOR) || (vector > HV_IPI_HIGH_VECTOR))
 		return HV_STATUS_INVALID_HYPERCALL_INPUT;
 
diff --git a/arch/x86/kvm/mmu/mmu.c b/arch/x86/kvm/mmu/mmu.c
index 1ccee4d17481..1d275e9d76b5 100644
--- a/arch/x86/kvm/mmu/mmu.c
+++ b/arch/x86/kvm/mmu/mmu.c
@@ -3971,7 +3971,21 @@ out_retry:
 static bool is_page_fault_stale(struct kvm_vcpu *vcpu,
 				struct kvm_page_fault *fault, int mmu_seq)
 {
-	if (is_obsolete_sp(vcpu->kvm, to_shadow_page(vcpu->arch.mmu->root_hpa)))
+	struct kvm_mmu_page *sp = to_shadow_page(vcpu->arch.mmu->root_hpa);
+
+	/* Special roots, e.g. pae_root, are not backed by shadow pages. */
+	if (sp && is_obsolete_sp(vcpu->kvm, sp))
+		return true;
+
+	/*
+	 * Roots without an associated shadow page are considered invalid if
+	 * there is a pending request to free obsolete roots.  The request is
+	 * only a hint that the current root _may_ be obsolete and needs to be
+	 * reloaded, e.g. if the guest frees a PGD that KVM is tracking as a
+	 * previous root, then __kvm_mmu_prepare_zap_page() signals all vCPUs
+	 * to reload even if no vCPU is actively using the root.
+	 */
+	if (!sp && kvm_test_request(KVM_REQ_MMU_RELOAD, vcpu))
 		return true;
 
 	return fault->slot &&
diff --git a/arch/x86/kvm/mmu/tdp_iter.c b/arch/x86/kvm/mmu/tdp_iter.c
index b3ed302c1a35..caa96c270b95 100644
--- a/arch/x86/kvm/mmu/tdp_iter.c
+++ b/arch/x86/kvm/mmu/tdp_iter.c
@@ -26,6 +26,7 @@ static gfn_t round_gfn_for_level(gfn_t gfn, int level)
  */
 void tdp_iter_restart(struct tdp_iter *iter)
 {
+	iter->yielded = false;
 	iter->yielded_gfn = iter->next_last_level_gfn;
 	iter->level = iter->root_level;
 
@@ -160,6 +161,11 @@ static bool try_step_up(struct tdp_iter *iter)
  */
 void tdp_iter_next(struct tdp_iter *iter)
 {
+	if (iter->yielded) {
+		tdp_iter_restart(iter);
+		return;
+	}
+
 	if (try_step_down(iter))
 		return;
 
diff --git a/arch/x86/kvm/mmu/tdp_iter.h b/arch/x86/kvm/mmu/tdp_iter.h
index b1748b988d3a..e19cabbcb65c 100644
--- a/arch/x86/kvm/mmu/tdp_iter.h
+++ b/arch/x86/kvm/mmu/tdp_iter.h
@@ -45,6 +45,12 @@ struct tdp_iter {
 	 * iterator walks off the end of the paging structure.
 	 */
 	bool valid;
+	/*
+	 * True if KVM dropped mmu_lock and yielded in the middle of a walk, in
+	 * which case tdp_iter_next() needs to restart the walk at the root
+	 * level instead of advancing to the next entry.
+	 */
+	bool yielded;
 };
 
 /*
diff --git a/arch/x86/kvm/mmu/tdp_mmu.c b/arch/x86/kvm/mmu/tdp_mmu.c
index b69e47e68307..7b1bc816b7c3 100644
--- a/arch/x86/kvm/mmu/tdp_mmu.c
+++ b/arch/x86/kvm/mmu/tdp_mmu.c
@@ -502,6 +502,8 @@ static inline bool tdp_mmu_set_spte_atomic(struct kvm *kvm,
 					   struct tdp_iter *iter,
 					   u64 new_spte)
 {
+	WARN_ON_ONCE(iter->yielded);
+
 	lockdep_assert_held_read(&kvm->mmu_lock);
 
 	/*
@@ -575,6 +577,8 @@ static inline void __tdp_mmu_set_spte(struct kvm *kvm, struct tdp_iter *iter,
 				      u64 new_spte, bool record_acc_track,
 				      bool record_dirty_log)
 {
+	WARN_ON_ONCE(iter->yielded);
+
 	lockdep_assert_held_write(&kvm->mmu_lock);
 
 	/*
@@ -640,18 +644,19 @@ static inline void tdp_mmu_set_spte_no_dirty_log(struct kvm *kvm,
  * If this function should yield and flush is set, it will perform a remote
  * TLB flush before yielding.
  *
- * If this function yields, it will also reset the tdp_iter's walk over the
- * paging structure and the calling function should skip to the next
- * iteration to allow the iterator to continue its traversal from the
- * paging structure root.
+ * If this function yields, iter->yielded is set and the caller must skip to
+ * the next iteration, where tdp_iter_next() will reset the tdp_iter's walk
+ * over the paging structures to allow the iterator to continue its traversal
+ * from the paging structure root.
  *
- * Return true if this function yielded and the iterator's traversal was reset.
- * Return false if a yield was not needed.
+ * Returns true if this function yielded.
  */
-static inline bool tdp_mmu_iter_cond_resched(struct kvm *kvm,
-					     struct tdp_iter *iter, bool flush,
-					     bool shared)
+static inline bool __must_check tdp_mmu_iter_cond_resched(struct kvm *kvm,
+							  struct tdp_iter *iter,
+							  bool flush, bool shared)
 {
+	WARN_ON(iter->yielded);
+
 	/* Ensure forward progress has been made before yielding. */
 	if (iter->next_last_level_gfn == iter->yielded_gfn)
 		return false;
@@ -671,12 +676,10 @@ static inline bool tdp_mmu_iter_cond_resched(struct kvm *kvm,
 
 		WARN_ON(iter->gfn > iter->next_last_level_gfn);
 
-		tdp_iter_restart(iter);
-
-		return true;
+		iter->yielded = true;
 	}
 
-	return false;
+	return iter->yielded;
 }
 
 /*
diff --git a/arch/x86/kvm/svm/svm.c b/arch/x86/kvm/svm/svm.c
index 208566f63bce..5557867dcb69 100644
--- a/arch/x86/kvm/svm/svm.c
+++ b/arch/x86/kvm/svm/svm.c
@@ -1594,6 +1594,15 @@ static void svm_set_rflags(struct kvm_vcpu *vcpu, unsigned long rflags)
 	to_svm(vcpu)->vmcb->save.rflags = rflags;
 }
 
+static bool svm_get_if_flag(struct kvm_vcpu *vcpu)
+{
+	struct vmcb *vmcb = to_svm(vcpu)->vmcb;
+
+	return sev_es_guest(vcpu->kvm)
+		? vmcb->control.int_state & SVM_GUEST_INTERRUPT_MASK
+		: kvm_get_rflags(vcpu) & X86_EFLAGS_IF;
+}
+
 static void svm_cache_reg(struct kvm_vcpu *vcpu, enum kvm_reg reg)
 {
 	kvm_register_mark_available(vcpu, reg);
@@ -3583,14 +3592,7 @@ bool svm_interrupt_blocked(struct kvm_vcpu *vcpu)
 	if (!gif_set(svm))
 		return true;
 
-	if (sev_es_guest(vcpu->kvm)) {
-		/*
-		 * SEV-ES guests to not expose RFLAGS. Use the VMCB interrupt mask
-		 * bit to determine the state of the IF flag.
-		 */
-		if (!(vmcb->control.int_state & SVM_GUEST_INTERRUPT_MASK))
-			return true;
-	} else if (is_guest_mode(vcpu)) {
+	if (is_guest_mode(vcpu)) {
 		/* As long as interrupts are being delivered...  */
 		if ((svm->nested.ctl.int_ctl & V_INTR_MASKING_MASK)
 		    ? !(svm->vmcb01.ptr->save.rflags & X86_EFLAGS_IF)
@@ -3601,7 +3603,7 @@ bool svm_interrupt_blocked(struct kvm_vcpu *vcpu)
 		if (nested_exit_on_intr(svm))
 			return false;
 	} else {
-		if (!(kvm_get_rflags(vcpu) & X86_EFLAGS_IF))
+		if (!svm_get_if_flag(vcpu))
 			return true;
 	}
 
@@ -4634,6 +4636,7 @@ static struct kvm_x86_ops svm_x86_ops __initdata = {
 	.cache_reg = svm_cache_reg,
 	.get_rflags = svm_get_rflags,
 	.set_rflags = svm_set_rflags,
+	.get_if_flag = svm_get_if_flag,
 
 	.tlb_flush_all = svm_flush_tlb,
 	.tlb_flush_current = svm_flush_tlb,
diff --git a/arch/x86/kvm/vmx/vmx.c b/arch/x86/kvm/vmx/vmx.c
index 63615d242bdf..9bf65e5e4840 100644
--- a/arch/x86/kvm/vmx/vmx.c
+++ b/arch/x86/kvm/vmx/vmx.c
@@ -1372,6 +1372,11 @@ void vmx_set_rflags(struct kvm_vcpu *vcpu, unsigned long rflags)
 		vmx->emulation_required = vmx_emulation_required(vcpu);
 }
 
+static bool vmx_get_if_flag(struct kvm_vcpu *vcpu)
+{
+	return vmx_get_rflags(vcpu) & X86_EFLAGS_IF;
+}
+
 u32 vmx_get_interrupt_shadow(struct kvm_vcpu *vcpu)
 {
 	u32 interruptibility = vmcs_read32(GUEST_INTERRUPTIBILITY_INFO);
@@ -3995,8 +4000,7 @@ static int vmx_deliver_posted_interrupt(struct kvm_vcpu *vcpu, int vector)
 	 * guaranteed to see PID.ON=1 and sync the PIR to IRR if triggering a
 	 * posted interrupt "fails" because vcpu->mode != IN_GUEST_MODE.
 	 */
-	if (vcpu != kvm_get_running_vcpu() &&
-	    !kvm_vcpu_trigger_posted_interrupt(vcpu, false))
+	if (!kvm_vcpu_trigger_posted_interrupt(vcpu, false))
 		kvm_vcpu_kick(vcpu);
 
 	return 0;
@@ -5921,18 +5925,14 @@ static int __vmx_handle_exit(struct kvm_vcpu *vcpu, fastpath_t exit_fastpath)
 		vmx_flush_pml_buffer(vcpu);
 
 	/*
-	 * We should never reach this point with a pending nested VM-Enter, and
-	 * more specifically emulation of L2 due to invalid guest state (see
-	 * below) should never happen as that means we incorrectly allowed a
-	 * nested VM-Enter with an invalid vmcs12.
+	 * KVM should never reach this point with a pending nested VM-Enter.
+	 * More specifically, short-circuiting VM-Entry to emulate L2 due to
+	 * invalid guest state should never happen as that means KVM knowingly
+	 * allowed a nested VM-Enter with an invalid vmcs12.  More below.
 	 */
 	if (KVM_BUG_ON(vmx->nested.nested_run_pending, vcpu->kvm))
 		return -EIO;
 
-	/* If guest state is invalid, start emulating */
-	if (vmx->emulation_required)
-		return handle_invalid_guest_state(vcpu);
-
 	if (is_guest_mode(vcpu)) {
 		/*
 		 * PML is never enabled when running L2, bail immediately if a
@@ -5954,10 +5954,30 @@ static int __vmx_handle_exit(struct kvm_vcpu *vcpu, fastpath_t exit_fastpath)
 		 */
 		nested_mark_vmcs12_pages_dirty(vcpu);
 
+		/*
+		 * Synthesize a triple fault if L2 state is invalid.  In normal
+		 * operation, nested VM-Enter rejects any attempt to enter L2
+		 * with invalid state.  However, those checks are skipped if
+		 * state is being stuffed via RSM or KVM_SET_NESTED_STATE.  If
+		 * L2 state is invalid, it means either L1 modified SMRAM state
+		 * or userspace provided bad state.  Synthesize TRIPLE_FAULT as
+		 * doing so is architecturally allowed in the RSM case, and is
+		 * the least awful solution for the userspace case without
+		 * risking false positives.
+		 */
+		if (vmx->emulation_required) {
+			nested_vmx_vmexit(vcpu, EXIT_REASON_TRIPLE_FAULT, 0, 0);
+			return 1;
+		}
+
 		if (nested_vmx_reflect_vmexit(vcpu))
 			return 1;
 	}
 
+	/* If guest state is invalid, start emulating.  L2 is handled above. */
+	if (vmx->emulation_required)
+		return handle_invalid_guest_state(vcpu);
+
 	if (exit_reason.failed_vmentry) {
 		dump_vmcs(vcpu);
 		vcpu->run->exit_reason = KVM_EXIT_FAIL_ENTRY;
@@ -6652,9 +6672,7 @@ static fastpath_t vmx_vcpu_run(struct kvm_vcpu *vcpu)
 	 * consistency check VM-Exit due to invalid guest state and bail.
 	 */
 	if (unlikely(vmx->emulation_required)) {
-
-		/* We don't emulate invalid state of a nested guest */
-		vmx->fail = is_guest_mode(vcpu);
+		vmx->fail = 0;
 
 		vmx->exit_reason.full = EXIT_REASON_INVALID_STATE;
 		vmx->exit_reason.failed_vmentry = 1;
@@ -7609,6 +7627,7 @@ static struct kvm_x86_ops vmx_x86_ops __initdata = {
 	.cache_reg = vmx_cache_reg,
 	.get_rflags = vmx_get_rflags,
 	.set_rflags = vmx_set_rflags,
+	.get_if_flag = vmx_get_if_flag,
 
 	.tlb_flush_all = vmx_flush_tlb_all,
 	.tlb_flush_current = vmx_flush_tlb_current,
diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c
index 26cb3a4cd0e9..8912f44c640d 100644
--- a/arch/x86/kvm/x86.c
+++ b/arch/x86/kvm/x86.c
@@ -906,7 +906,8 @@ int kvm_set_cr0(struct kvm_vcpu *vcpu, unsigned long cr0)
 	    !load_pdptrs(vcpu, kvm_read_cr3(vcpu)))
 		return 1;
 
-	if (!(cr0 & X86_CR0_PG) && kvm_read_cr4_bits(vcpu, X86_CR4_PCIDE))
+	if (!(cr0 & X86_CR0_PG) &&
+	    (is_64_bit_mode(vcpu) || kvm_read_cr4_bits(vcpu, X86_CR4_PCIDE)))
 		return 1;
 
 	static_call(kvm_x86_set_cr0)(vcpu, cr0);
@@ -1343,7 +1344,7 @@ static const u32 msrs_to_save_all[] = {
 	MSR_IA32_UMWAIT_CONTROL,
 
 	MSR_ARCH_PERFMON_FIXED_CTR0, MSR_ARCH_PERFMON_FIXED_CTR1,
-	MSR_ARCH_PERFMON_FIXED_CTR0 + 2, MSR_ARCH_PERFMON_FIXED_CTR0 + 3,
+	MSR_ARCH_PERFMON_FIXED_CTR0 + 2,
 	MSR_CORE_PERF_FIXED_CTR_CTRL, MSR_CORE_PERF_GLOBAL_STATUS,
 	MSR_CORE_PERF_GLOBAL_CTRL, MSR_CORE_PERF_GLOBAL_OVF_CTRL,
 	MSR_ARCH_PERFMON_PERFCTR0, MSR_ARCH_PERFMON_PERFCTR1,
@@ -3424,7 +3425,7 @@ int kvm_set_msr_common(struct kvm_vcpu *vcpu, struct msr_data *msr_info)
 
 		if (!msr_info->host_initiated)
 			return 1;
-		if (guest_cpuid_has(vcpu, X86_FEATURE_PDCM) && kvm_get_msr_feature(&msr_ent))
+		if (kvm_get_msr_feature(&msr_ent))
 			return 1;
 		if (data & ~msr_ent.data)
 			return 1;
@@ -7144,7 +7145,13 @@ static int emulator_pio_in(struct kvm_vcpu *vcpu, int size,
 			   unsigned short port, void *val, unsigned int count)
 {
 	if (vcpu->arch.pio.count) {
-		/* Complete previous iteration.  */
+		/*
+		 * Complete a previous iteration that required userspace I/O.
+		 * Note, @count isn't guaranteed to match pio.count as userspace
+		 * can modify ECX before rerunning the vCPU.  Ignore any such
+		 * shenanigans as KVM doesn't support modifying the rep count,
+		 * and the emulator ensures @count doesn't overflow the buffer.
+		 */
 	} else {
 		int r = __emulator_pio_in(vcpu, size, port, count);
 		if (!r)
@@ -7153,7 +7160,6 @@ static int emulator_pio_in(struct kvm_vcpu *vcpu, int size,
 		/* Results already available, fall through.  */
 	}
 
-	WARN_ON(count != vcpu->arch.pio.count);
 	complete_emulator_pio_in(vcpu, val);
 	return 1;
 }
@@ -9043,14 +9049,7 @@ static void post_kvm_run_save(struct kvm_vcpu *vcpu)
 {
 	struct kvm_run *kvm_run = vcpu->run;
 
-	/*
-	 * if_flag is obsolete and useless, so do not bother
-	 * setting it for SEV-ES guests.  Userspace can just
-	 * use kvm_run->ready_for_interrupt_injection.
-	 */
-	kvm_run->if_flag = !vcpu->arch.guest_state_protected
-		&& (kvm_get_rflags(vcpu) & X86_EFLAGS_IF) != 0;
-
+	kvm_run->if_flag = static_call(kvm_x86_get_if_flag)(vcpu);
 	kvm_run->cr8 = kvm_get_cr8(vcpu);
 	kvm_run->apic_base = kvm_get_apic_base(vcpu);
 
diff --git a/tools/testing/selftests/kvm/.gitignore b/tools/testing/selftests/kvm/.gitignore
index 3763105029fb..3cb5ac5da087 100644
--- a/tools/testing/selftests/kvm/.gitignore
+++ b/tools/testing/selftests/kvm/.gitignore
@@ -30,10 +30,12 @@
 /x86_64/svm_int_ctl_test
 /x86_64/sync_regs_test
 /x86_64/tsc_msrs_test
+/x86_64/userspace_io_test
 /x86_64/userspace_msr_exit_test
 /x86_64/vmx_apic_access_test
 /x86_64/vmx_close_while_nested_test
 /x86_64/vmx_dirty_log_test
+/x86_64/vmx_invalid_nested_guest_state
 /x86_64/vmx_preemption_timer_test
 /x86_64/vmx_set_nested_state_test
 /x86_64/vmx_tsc_adjust_test
diff --git a/tools/testing/selftests/kvm/Makefile b/tools/testing/selftests/kvm/Makefile
index c4e34717826a..17342b575e85 100644
--- a/tools/testing/selftests/kvm/Makefile
+++ b/tools/testing/selftests/kvm/Makefile
@@ -59,10 +59,12 @@ TEST_GEN_PROGS_x86_64 += x86_64/vmx_preemption_timer_test
 TEST_GEN_PROGS_x86_64 += x86_64/svm_vmcall_test
 TEST_GEN_PROGS_x86_64 += x86_64/svm_int_ctl_test
 TEST_GEN_PROGS_x86_64 += x86_64/sync_regs_test
+TEST_GEN_PROGS_x86_64 += x86_64/userspace_io_test
 TEST_GEN_PROGS_x86_64 += x86_64/userspace_msr_exit_test
 TEST_GEN_PROGS_x86_64 += x86_64/vmx_apic_access_test
 TEST_GEN_PROGS_x86_64 += x86_64/vmx_close_while_nested_test
 TEST_GEN_PROGS_x86_64 += x86_64/vmx_dirty_log_test
+TEST_GEN_PROGS_x86_64 += x86_64/vmx_invalid_nested_guest_state
 TEST_GEN_PROGS_x86_64 += x86_64/vmx_set_nested_state_test
 TEST_GEN_PROGS_x86_64 += x86_64/vmx_tsc_adjust_test
 TEST_GEN_PROGS_x86_64 += x86_64/vmx_nested_tsc_scaling_test
diff --git a/tools/testing/selftests/kvm/include/kvm_util.h b/tools/testing/selftests/kvm/include/kvm_util.h
index 6a1a37f30494..2d62edc49d67 100644
--- a/tools/testing/selftests/kvm/include/kvm_util.h
+++ b/tools/testing/selftests/kvm/include/kvm_util.h
@@ -321,6 +321,7 @@ bool vm_is_unrestricted_guest(struct kvm_vm *vm);
 
 unsigned int vm_get_page_size(struct kvm_vm *vm);
 unsigned int vm_get_page_shift(struct kvm_vm *vm);
+unsigned long vm_compute_max_gfn(struct kvm_vm *vm);
 uint64_t vm_get_max_gfn(struct kvm_vm *vm);
 int vm_get_fd(struct kvm_vm *vm);
 
diff --git a/tools/testing/selftests/kvm/lib/kvm_util.c b/tools/testing/selftests/kvm/lib/kvm_util.c
index 8f2e0bb1ef96..53d2b5d04b82 100644
--- a/tools/testing/selftests/kvm/lib/kvm_util.c
+++ b/tools/testing/selftests/kvm/lib/kvm_util.c
@@ -302,7 +302,7 @@ struct kvm_vm *vm_create(enum vm_guest_mode mode, uint64_t phy_pages, int perm)
 		(1ULL << (vm->va_bits - 1)) >> vm->page_shift);
 
 	/* Limit physical addresses to PA-bits. */
-	vm->max_gfn = ((1ULL << vm->pa_bits) >> vm->page_shift) - 1;
+	vm->max_gfn = vm_compute_max_gfn(vm);
 
 	/* Allocate and setup memory for guest. */
 	vm->vpages_mapped = sparsebit_alloc();
@@ -2328,6 +2328,11 @@ unsigned int vm_get_page_shift(struct kvm_vm *vm)
 	return vm->page_shift;
 }
 
+unsigned long __attribute__((weak)) vm_compute_max_gfn(struct kvm_vm *vm)
+{
+	return ((1ULL << vm->pa_bits) >> vm->page_shift) - 1;
+}
+
 uint64_t vm_get_max_gfn(struct kvm_vm *vm)
 {
 	return vm->max_gfn;
diff --git a/tools/testing/selftests/kvm/lib/x86_64/processor.c b/tools/testing/selftests/kvm/lib/x86_64/processor.c
index 82c39db91369..eef7b34756d5 100644
--- a/tools/testing/selftests/kvm/lib/x86_64/processor.c
+++ b/tools/testing/selftests/kvm/lib/x86_64/processor.c
@@ -1431,3 +1431,71 @@ struct kvm_cpuid2 *vcpu_get_supported_hv_cpuid(struct kvm_vm *vm, uint32_t vcpui
 
 	return cpuid;
 }
+
+#define X86EMUL_CPUID_VENDOR_AuthenticAMD_ebx 0x68747541
+#define X86EMUL_CPUID_VENDOR_AuthenticAMD_ecx 0x444d4163
+#define X86EMUL_CPUID_VENDOR_AuthenticAMD_edx 0x69746e65
+
+static inline unsigned x86_family(unsigned int eax)
+{
+        unsigned int x86;
+
+        x86 = (eax >> 8) & 0xf;
+
+        if (x86 == 0xf)
+                x86 += (eax >> 20) & 0xff;
+
+        return x86;
+}
+
+unsigned long vm_compute_max_gfn(struct kvm_vm *vm)
+{
+	const unsigned long num_ht_pages = 12 << (30 - vm->page_shift); /* 12 GiB */
+	unsigned long ht_gfn, max_gfn, max_pfn;
+	uint32_t eax, ebx, ecx, edx, max_ext_leaf;
+
+	max_gfn = (1ULL << (vm->pa_bits - vm->page_shift)) - 1;
+
+	/* Avoid reserved HyperTransport region on AMD processors.  */
+	eax = ecx = 0;
+	cpuid(&eax, &ebx, &ecx, &edx);
+	if (ebx != X86EMUL_CPUID_VENDOR_AuthenticAMD_ebx ||
+	    ecx != X86EMUL_CPUID_VENDOR_AuthenticAMD_ecx ||
+	    edx != X86EMUL_CPUID_VENDOR_AuthenticAMD_edx)
+		return max_gfn;
+
+	/* On parts with <40 physical address bits, the area is fully hidden */
+	if (vm->pa_bits < 40)
+		return max_gfn;
+
+	/* Before family 17h, the HyperTransport area is just below 1T.  */
+	ht_gfn = (1 << 28) - num_ht_pages;
+	eax = 1;
+	cpuid(&eax, &ebx, &ecx, &edx);
+	if (x86_family(eax) < 0x17)
+		goto done;
+
+	/*
+	 * Otherwise it's at the top of the physical address space, possibly
+	 * reduced due to SME by bits 11:6 of CPUID[0x8000001f].EBX.  Use
+	 * the old conservative value if MAXPHYADDR is not enumerated.
+	 */
+	eax = 0x80000000;
+	cpuid(&eax, &ebx, &ecx, &edx);
+	max_ext_leaf = eax;
+	if (max_ext_leaf < 0x80000008)
+		goto done;
+
+	eax = 0x80000008;
+	cpuid(&eax, &ebx, &ecx, &edx);
+	max_pfn = (1ULL << ((eax & 0xff) - vm->page_shift)) - 1;
+	if (max_ext_leaf >= 0x8000001f) {
+		eax = 0x8000001f;
+		cpuid(&eax, &ebx, &ecx, &edx);
+		max_pfn >>= (ebx >> 6) & 0x3f;
+	}
+
+	ht_gfn = max_pfn - num_ht_pages;
+done:
+	return min(max_gfn, ht_gfn - 1);
+}
diff --git a/tools/testing/selftests/kvm/x86_64/svm_int_ctl_test.c b/tools/testing/selftests/kvm/x86_64/svm_int_ctl_test.c
index df04f56ce859..30a81038df46 100644
--- a/tools/testing/selftests/kvm/x86_64/svm_int_ctl_test.c
+++ b/tools/testing/selftests/kvm/x86_64/svm_int_ctl_test.c
@@ -75,7 +75,7 @@ static void l1_guest_code(struct svm_test_data *svm)
 	vmcb->control.int_ctl &= ~V_INTR_MASKING_MASK;
 
 	/* No intercepts for real and virtual interrupts */
-	vmcb->control.intercept &= ~(1ULL << INTERCEPT_INTR | INTERCEPT_VINTR);
+	vmcb->control.intercept &= ~(BIT(INTERCEPT_INTR) | BIT(INTERCEPT_VINTR));
 
 	/* Make a virtual interrupt VINTR_IRQ_NUMBER pending */
 	vmcb->control.int_ctl |= V_IRQ_MASK | (0x1 << V_INTR_PRIO_SHIFT);
diff --git a/tools/testing/selftests/kvm/x86_64/userspace_io_test.c b/tools/testing/selftests/kvm/x86_64/userspace_io_test.c
new file mode 100644
index 000000000000..e4bef2e05686
--- /dev/null
+++ b/tools/testing/selftests/kvm/x86_64/userspace_io_test.c
@@ -0,0 +1,114 @@
+// SPDX-License-Identifier: GPL-2.0
+#include <fcntl.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+#include <sys/ioctl.h>
+
+#include "test_util.h"
+
+#include "kvm_util.h"
+#include "processor.h"
+
+#define VCPU_ID			1
+
+static void guest_ins_port80(uint8_t *buffer, unsigned int count)
+{
+	unsigned long end;
+
+	if (count == 2)
+		end = (unsigned long)buffer + 1;
+	else
+		end = (unsigned long)buffer + 8192;
+
+	asm volatile("cld; rep; insb" : "+D"(buffer), "+c"(count) : "d"(0x80) : "memory");
+	GUEST_ASSERT_1(count == 0, count);
+	GUEST_ASSERT_2((unsigned long)buffer == end, buffer, end);
+}
+
+static void guest_code(void)
+{
+	uint8_t buffer[8192];
+	int i;
+
+	/*
+	 * Special case tests.  main() will adjust RCX 2 => 1 and 3 => 8192 to
+	 * test that KVM doesn't explode when userspace modifies the "count" on
+	 * a userspace I/O exit.  KVM isn't required to play nice with the I/O
+	 * itself as KVM doesn't support manipulating the count, it just needs
+	 * to not explode or overflow a buffer.
+	 */
+	guest_ins_port80(buffer, 2);
+	guest_ins_port80(buffer, 3);
+
+	/* Verify KVM fills the buffer correctly when not stuffing RCX. */
+	memset(buffer, 0, sizeof(buffer));
+	guest_ins_port80(buffer, 8192);
+	for (i = 0; i < 8192; i++)
+		GUEST_ASSERT_2(buffer[i] == 0xaa, i, buffer[i]);
+
+	GUEST_DONE();
+}
+
+int main(int argc, char *argv[])
+{
+	struct kvm_regs regs;
+	struct kvm_run *run;
+	struct kvm_vm *vm;
+	struct ucall uc;
+	int rc;
+
+	/* Tell stdout not to buffer its content */
+	setbuf(stdout, NULL);
+
+	/* Create VM */
+	vm = vm_create_default(VCPU_ID, 0, guest_code);
+	run = vcpu_state(vm, VCPU_ID);
+
+	memset(&regs, 0, sizeof(regs));
+
+	while (1) {
+		rc = _vcpu_run(vm, VCPU_ID);
+
+		TEST_ASSERT(rc == 0, "vcpu_run failed: %d\n", rc);
+		TEST_ASSERT(run->exit_reason == KVM_EXIT_IO,
+			    "Unexpected exit reason: %u (%s),\n",
+			    run->exit_reason,
+			    exit_reason_str(run->exit_reason));
+
+		if (get_ucall(vm, VCPU_ID, &uc))
+			break;
+
+		TEST_ASSERT(run->io.port == 0x80,
+			    "Expected I/O at port 0x80, got port 0x%x\n", run->io.port);
+
+		/*
+		 * Modify the rep string count in RCX: 2 => 1 and 3 => 8192.
+		 * Note, this abuses KVM's batching of rep string I/O to avoid
+		 * getting stuck in an infinite loop.  That behavior isn't in
+		 * scope from a testing perspective as it's not ABI in any way,
+		 * i.e. it really is abusing internal KVM knowledge.
+		 */
+		vcpu_regs_get(vm, VCPU_ID, &regs);
+		if (regs.rcx == 2)
+			regs.rcx = 1;
+		if (regs.rcx == 3)
+			regs.rcx = 8192;
+		memset((void *)run + run->io.data_offset, 0xaa, 4096);
+		vcpu_regs_set(vm, VCPU_ID, &regs);
+	}
+
+	switch (uc.cmd) {
+	case UCALL_DONE:
+		break;
+	case UCALL_ABORT:
+		TEST_FAIL("%s at %s:%ld : argN+1 = 0x%lx, argN+2 = 0x%lx",
+			  (const char *)uc.args[0], __FILE__, uc.args[1],
+			  uc.args[2], uc.args[3]);
+	default:
+		TEST_FAIL("Unknown ucall %lu", uc.cmd);
+	}
+
+	kvm_vm_free(vm);
+	return 0;
+}
diff --git a/tools/testing/selftests/kvm/x86_64/vmx_invalid_nested_guest_state.c b/tools/testing/selftests/kvm/x86_64/vmx_invalid_nested_guest_state.c
new file mode 100644
index 000000000000..489fbed4ca6f
--- /dev/null
+++ b/tools/testing/selftests/kvm/x86_64/vmx_invalid_nested_guest_state.c
@@ -0,0 +1,105 @@
+// SPDX-License-Identifier: GPL-2.0-only
+#include "test_util.h"
+#include "kvm_util.h"
+#include "processor.h"
+#include "vmx.h"
+
+#include <string.h>
+#include <sys/ioctl.h>
+
+#include "kselftest.h"
+
+#define VCPU_ID	0
+#define ARBITRARY_IO_PORT 0x2000
+
+static struct kvm_vm *vm;
+
+static void l2_guest_code(void)
+{
+	/*
+	 * Generate an exit to L0 userspace, i.e. main(), via I/O to an
+	 * arbitrary port.
+	 */
+	asm volatile("inb %%dx, %%al"
+		     : : [port] "d" (ARBITRARY_IO_PORT) : "rax");
+}
+
+static void l1_guest_code(struct vmx_pages *vmx_pages)
+{
+#define L2_GUEST_STACK_SIZE 64
+	unsigned long l2_guest_stack[L2_GUEST_STACK_SIZE];
+
+	GUEST_ASSERT(prepare_for_vmx_operation(vmx_pages));
+	GUEST_ASSERT(load_vmcs(vmx_pages));
+
+	/* Prepare the VMCS for L2 execution. */
+	prepare_vmcs(vmx_pages, l2_guest_code,
+		     &l2_guest_stack[L2_GUEST_STACK_SIZE]);
+
+	/*
+	 * L2 must be run without unrestricted guest, verify that the selftests
+	 * library hasn't enabled it.  Because KVM selftests jump directly to
+	 * 64-bit mode, unrestricted guest support isn't required.
+	 */
+	GUEST_ASSERT(!(vmreadz(CPU_BASED_VM_EXEC_CONTROL) & CPU_BASED_ACTIVATE_SECONDARY_CONTROLS) ||
+		     !(vmreadz(SECONDARY_VM_EXEC_CONTROL) & SECONDARY_EXEC_UNRESTRICTED_GUEST));
+
+	GUEST_ASSERT(!vmlaunch());
+
+	/* L2 should triple fault after main() stuffs invalid guest state. */
+	GUEST_ASSERT(vmreadz(VM_EXIT_REASON) == EXIT_REASON_TRIPLE_FAULT);
+	GUEST_DONE();
+}
+
+int main(int argc, char *argv[])
+{
+	vm_vaddr_t vmx_pages_gva;
+	struct kvm_sregs sregs;
+	struct kvm_run *run;
+	struct ucall uc;
+
+	nested_vmx_check_supported();
+
+	vm = vm_create_default(VCPU_ID, 0, (void *) l1_guest_code);
+
+	/* Allocate VMX pages and shared descriptors (vmx_pages). */
+	vcpu_alloc_vmx(vm, &vmx_pages_gva);
+	vcpu_args_set(vm, VCPU_ID, 1, vmx_pages_gva);
+
+	vcpu_run(vm, VCPU_ID);
+
+	run = vcpu_state(vm, VCPU_ID);
+
+	/*
+	 * The first exit to L0 userspace should be an I/O access from L2.
+	 * Running L1 should launch L2 without triggering an exit to userspace.
+	 */
+	TEST_ASSERT(run->exit_reason == KVM_EXIT_IO,
+		    "Expected KVM_EXIT_IO, got: %u (%s)\n",
+		    run->exit_reason, exit_reason_str(run->exit_reason));
+
+	TEST_ASSERT(run->io.port == ARBITRARY_IO_PORT,
+		    "Expected IN from port %d from L2, got port %d",
+		    ARBITRARY_IO_PORT, run->io.port);
+
+	/*
+	 * Stuff invalid guest state for L2 by making TR unusuable.  The next
+	 * KVM_RUN should induce a TRIPLE_FAULT in L2 as KVM doesn't support
+	 * emulating invalid guest state for L2.
+	 */
+	memset(&sregs, 0, sizeof(sregs));
+	vcpu_sregs_get(vm, VCPU_ID, &sregs);
+	sregs.tr.unusable = 1;
+	vcpu_sregs_set(vm, VCPU_ID, &sregs);
+
+	vcpu_run(vm, VCPU_ID);
+
+	switch (get_ucall(vm, VCPU_ID, &uc)) {
+	case UCALL_DONE:
+		break;
+	case UCALL_ABORT:
+		TEST_FAIL("%s", (const char *)uc.args[0]);
+	default:
+		TEST_FAIL("Unexpected ucall: %lu", uc.cmd);
+	}
+}
diff --git a/tools/testing/selftests/kvm/x86_64/vmx_pmu_msrs_test.c b/tools/testing/selftests/kvm/x86_64/vmx_pmu_msrs_test.c
index 23051d84b907..2454a1f2ca0c 100644
--- a/tools/testing/selftests/kvm/x86_64/vmx_pmu_msrs_test.c
+++ b/tools/testing/selftests/kvm/x86_64/vmx_pmu_msrs_test.c
@@ -110,22 +110,5 @@ int main(int argc, char *argv[])
 	ret = _vcpu_set_msr(vm, 0, MSR_IA32_PERF_CAPABILITIES, PMU_CAP_LBR_FMT);
 	TEST_ASSERT(ret == 0, "Bad PERF_CAPABILITIES didn't fail.");
 
-	/* testcase 4, set capabilities when we don't have PDCM bit */
-	entry_1_0->ecx &= ~X86_FEATURE_PDCM;
-	vcpu_set_cpuid(vm, VCPU_ID, cpuid);
-	ret = _vcpu_set_msr(vm, 0, MSR_IA32_PERF_CAPABILITIES, host_cap.capabilities);
-	TEST_ASSERT(ret == 0, "Bad PERF_CAPABILITIES didn't fail.");
-
-	/* testcase 5, set capabilities when we don't have PMU version bits */
-	entry_1_0->ecx |= X86_FEATURE_PDCM;
-	eax.split.version_id = 0;
-	entry_1_0->ecx = eax.full;
-	vcpu_set_cpuid(vm, VCPU_ID, cpuid);
-	ret = _vcpu_set_msr(vm, 0, MSR_IA32_PERF_CAPABILITIES, PMU_CAP_FW_WRITES);
-	TEST_ASSERT(ret == 0, "Bad PERF_CAPABILITIES didn't fail.");
-
-	vcpu_set_msr(vm, 0, MSR_IA32_PERF_CAPABILITIES, 0);
-	ASSERT_EQ(vcpu_get_msr(vm, VCPU_ID, MSR_IA32_PERF_CAPABILITIES), 0);
-
 	kvm_vm_free(vm);
 }