Merge branch 'kvm-fixes' into 'next'

Pick up bugfixes from 5.9, otherwise various tests fail.

arch/arm64/kvm/hyp/nvhe/tlb.c

@@ -31,7 +31,14 @@ static void __tlb_switch_to_guest(struct kvm_s2_mmu *mmu,
 		isb();
 	}
 
+	/*
+	 * __load_guest_stage2() includes an ISB only when the AT
+	 * workaround is applied. Take care of the opposite condition,
+	 * ensuring that we always have an ISB, but not two ISBs back
+	 * to back.
+	 */
 	__load_guest_stage2(mmu);
+	asm(ALTERNATIVE("isb", "nop", ARM64_WORKAROUND_SPECULATIVE_AT));
 }
 
 static void __tlb_switch_to_host(struct tlb_inv_context *cxt)

arch/x86/kvm/svm/svm.c

@@ -2274,6 +2274,12 @@ static int iret_interception(struct vcpu_svm *svm)
 	return 1;
 }
 
+static int invd_interception(struct vcpu_svm *svm)
+{
+	/* Treat an INVD instruction as a NOP and just skip it. */
+	return kvm_skip_emulated_instruction(&svm->vcpu);
+}
+
 static int invlpg_interception(struct vcpu_svm *svm)
 {
 	if (!static_cpu_has(X86_FEATURE_DECODEASSISTS))
@@ -2891,7 +2897,7 @@ static int (*const svm_exit_handlers[])(struct vcpu_svm *svm) = {
 	[SVM_EXIT_RDPMC]			= rdpmc_interception,
 	[SVM_EXIT_CPUID]			= cpuid_interception,
 	[SVM_EXIT_IRET]                         = iret_interception,
-	[SVM_EXIT_INVD]                         = emulate_on_interception,
+	[SVM_EXIT_INVD]                         = invd_interception,
 	[SVM_EXIT_PAUSE]			= pause_interception,
 	[SVM_EXIT_HLT]				= halt_interception,
 	[SVM_EXIT_INVLPG]			= invlpg_interception,

arch/x86/kvm/vmx/vmx.c

@@ -128,6 +128,9 @@ static bool __read_mostly enable_preemption_timer = 1;
 module_param_named(preemption_timer, enable_preemption_timer, bool, S_IRUGO);
 #endif
 
+extern bool __read_mostly allow_smaller_maxphyaddr;
+module_param(allow_smaller_maxphyaddr, bool, S_IRUGO);
+
 #define KVM_VM_CR0_ALWAYS_OFF (X86_CR0_NW | X86_CR0_CD)
 #define KVM_VM_CR0_ALWAYS_ON_UNRESTRICTED_GUEST X86_CR0_NE
 #define KVM_VM_CR0_ALWAYS_ON				\
@@ -834,6 +837,18 @@ void update_exception_bitmap(struct kvm_vcpu *vcpu)
 	 */
 	if (is_guest_mode(vcpu))
 		eb |= get_vmcs12(vcpu)->exception_bitmap;
+        else {
+		/*
+		 * If EPT is enabled, #PF is only trapped if MAXPHYADDR is mismatched
+		 * between guest and host.  In that case we only care about present
+		 * faults.  For vmcs02, however, PFEC_MASK and PFEC_MATCH are set in
+		 * prepare_vmcs02_rare.
+		 */
+		bool selective_pf_trap = enable_ept && (eb & (1u << PF_VECTOR));
+		int mask = selective_pf_trap ? PFERR_PRESENT_MASK : 0;
+		vmcs_write32(PAGE_FAULT_ERROR_CODE_MASK, mask);
+		vmcs_write32(PAGE_FAULT_ERROR_CODE_MATCH, mask);
+	}
 
 	vmcs_write32(EXCEPTION_BITMAP, eb);
 }
@@ -4363,16 +4378,6 @@ static void init_vmcs(struct vcpu_vmx *vmx)
 		vmx->pt_desc.guest.output_mask = 0x7F;
 		vmcs_write64(GUEST_IA32_RTIT_CTL, 0);
 	}
-
-	/*
-	 * If EPT is enabled, #PF is only trapped if MAXPHYADDR is mismatched
-	 * between guest and host.  In that case we only care about present
-	 * faults.
-	 */
-	if (enable_ept) {
-		vmcs_write32(PAGE_FAULT_ERROR_CODE_MASK, PFERR_PRESENT_MASK);
-		vmcs_write32(PAGE_FAULT_ERROR_CODE_MATCH, PFERR_PRESENT_MASK);
-	}
 }
 
 static void vmx_vcpu_reset(struct kvm_vcpu *vcpu, bool init_event)
@@ -4814,6 +4819,7 @@ static int handle_exception_nmi(struct kvm_vcpu *vcpu)
 			 * EPT will cause page fault only if we need to
 			 * detect illegal GPAs.
 			 */
+			WARN_ON_ONCE(!allow_smaller_maxphyaddr);
 			kvm_fixup_and_inject_pf_error(vcpu, cr2, error_code);
 			return 1;
 		} else
@@ -5343,7 +5349,7 @@ static int handle_ept_violation(struct kvm_vcpu *vcpu)
 	 * would also use advanced VM-exit information for EPT violations to
 	 * reconstruct the page fault error code.
 	 */
-	if (unlikely(kvm_vcpu_is_illegal_gpa(vcpu, gpa)))
+	if (unlikely(allow_smaller_maxphyaddr && kvm_vcpu_is_illegal_gpa(vcpu, gpa)))
 		return kvm_emulate_instruction(vcpu, 0);
 
 	return kvm_mmu_page_fault(vcpu, gpa, error_code, NULL, 0);
@@ -8008,11 +8014,12 @@ static int __init vmx_init(void)
 	vmx_check_vmcs12_offsets();
 
 	/*
-	 * Intel processors don't have problems with
-	 * GUEST_MAXPHYADDR < HOST_MAXPHYADDR so enable
-	 * it for VMX by default
+	 * Shadow paging doesn't have a (further) performance penalty
+	 * from GUEST_MAXPHYADDR < HOST_MAXPHYADDR so enable it
+	 * by default
 	 */
-	allow_smaller_maxphyaddr = true;
+	if (!enable_ept)
+		allow_smaller_maxphyaddr = true;
 
 	return 0;
 }

arch/x86/kvm/vmx/vmx.h

@@ -470,7 +470,10 @@ static inline bool vmx_has_waitpkg(struct vcpu_vmx *vmx)
 
 static inline bool vmx_need_pf_intercept(struct kvm_vcpu *vcpu)
 {
-	return !enable_ept || cpuid_maxphyaddr(vcpu) < boot_cpu_data.x86_phys_bits;
+	if (!enable_ept)
+		return true;
+
+	return allow_smaller_maxphyaddr && cpuid_maxphyaddr(vcpu) < boot_cpu_data.x86_phys_bits;
 }
 
 static inline bool is_unrestricted_guest(struct kvm_vcpu *vcpu)

arch/x86/kvm/x86.c

@@ -194,7 +194,7 @@ static struct kvm_user_return_msrs __percpu *user_return_msrs;
 u64 __read_mostly host_efer;
 EXPORT_SYMBOL_GPL(host_efer);
 
-bool __read_mostly allow_smaller_maxphyaddr;
+bool __read_mostly allow_smaller_maxphyaddr = 0;
 EXPORT_SYMBOL_GPL(allow_smaller_maxphyaddr);
 
 static u64 __read_mostly host_xss;
@@ -982,6 +982,7 @@ int kvm_set_cr4(struct kvm_vcpu *vcpu, unsigned long cr4)
 	unsigned long old_cr4 = kvm_read_cr4(vcpu);
 	unsigned long pdptr_bits = X86_CR4_PGE | X86_CR4_PSE | X86_CR4_PAE |
 				   X86_CR4_SMEP;
+	unsigned long mmu_role_bits = pdptr_bits | X86_CR4_SMAP | X86_CR4_PKE;
 
 	if (kvm_valid_cr4(vcpu, cr4))
 		return 1;
@@ -1009,7 +1010,7 @@ int kvm_set_cr4(struct kvm_vcpu *vcpu, unsigned long cr4)
 	if (kvm_x86_ops.set_cr4(vcpu, cr4))
 		return 1;
 
-	if (((cr4 ^ old_cr4) & pdptr_bits) ||
+	if (((cr4 ^ old_cr4) & mmu_role_bits) ||
 	    (!(cr4 & X86_CR4_PCIDE) && (old_cr4 & X86_CR4_PCIDE)))
 		kvm_mmu_reset_context(vcpu);
 
@@ -3400,9 +3401,9 @@ int kvm_get_msr_common(struct kvm_vcpu *vcpu, struct msr_data *msr_info)
 		 * even when not intercepted. AMD manual doesn't explicitly
 		 * state this but appears to behave the same.
 		 *
-		 * Unconditionally return L1's TSC offset on userspace reads
-		 * so that userspace reads and writes always operate on L1's
-		 * offset, e.g. to ensure deterministic behavior for migration.
+		 * On userspace reads and writes, however, we unconditionally
+		 * return L1's TSC value to ensure backwards-compatible
+		 * behavior for migration.
 		 */
 		u64 tsc_offset = msr_info->host_initiated ? vcpu->arch.l1_tsc_offset :
 							    vcpu->arch.tsc_offset;

tools/testing/selftests/kvm/x86_64/debug_regs.c

@@ -73,7 +73,7 @@ int main(void)
 	int i;
 	/* Instruction lengths starting at ss_start */
 	int ss_size[4] = {
-		3,		/* xor */
+		2,		/* xor */
 		2,		/* cpuid */
 		5,		/* mov */
 		2,		/* rdmsr */