Merge tag 'for-linus' of git://git.kernel.org/pub/scm/virt/kvm/kvm

Pull kvm fixes from Paolo Bonzini:
 "This is a pretty large update.  I think it is roughly as big as what I
  usually had for the _whole_ rc period.

  There are a few bad bugs where the guest can OOPS or crash the host.
  We have also started looking at attack models for nested
  virtualization; bugs that usually result in the guest ring 0 crashing
  itself become more worrisome if you have nested virtualization,
  because the nested guest might bring down the non-nested guest as
  well.  For current uses of nested virtualization these do not really
  have a security impact, but you never know and bugs are bugs
  nevertheless.

  A lot of these bugs are in 3.17 too, resulting in a large number of
  stable@ Ccs.  I checked that all the patches apply there with no
  conflicts"

* tag 'for-linus' of git://git.kernel.org/pub/scm/virt/kvm/kvm:
  kvm: vfio: fix unregister kvm_device_ops of vfio
  KVM: x86: Wrong assertion on paging_tmpl.h
  kvm: fix excessive pages un-pinning in kvm_iommu_map error path.
  KVM: x86: PREFETCH and HINT_NOP should have SrcMem flag
  KVM: x86: Emulator does not decode clflush well
  KVM: emulate: avoid accessing NULL ctxt->memopp
  KVM: x86: Decoding guest instructions which cross page boundary may fail
  kvm: x86: don't kill guest on unknown exit reason
  kvm: vmx: handle invvpid vm exit gracefully
  KVM: x86: Handle errors when RIP is set during far jumps
  KVM: x86: Emulator fixes for eip canonical checks on near branches
  KVM: x86: Fix wrong masking on relative jump/call
  KVM: x86: Improve thread safety in pit
  KVM: x86: Prevent host from panicking on shared MSR writes.
  KVM: x86: Check non-canonical addresses upon WRMSR

arch/x86/include/asm/kvm_host.h

@@ -989,6 +989,20 @@ static inline void kvm_inject_gp(struct kvm_vcpu *vcpu, u32 error_code)
 	kvm_queue_exception_e(vcpu, GP_VECTOR, error_code);
 }
 
+static inline u64 get_canonical(u64 la)
+{
+	return ((int64_t)la << 16) >> 16;
+}
+
+static inline bool is_noncanonical_address(u64 la)
+{
+#ifdef CONFIG_X86_64
+	return get_canonical(la) != la;
+#else
+	return false;
+#endif
+}
+
 #define TSS_IOPB_BASE_OFFSET 0x66
 #define TSS_BASE_SIZE 0x68
 #define TSS_IOPB_SIZE (65536 / 8)
@@ -1050,7 +1064,7 @@ void kvm_arch_mmu_notifier_invalidate_page(struct kvm *kvm,
 					   unsigned long address);
 
 void kvm_define_shared_msr(unsigned index, u32 msr);
-void kvm_set_shared_msr(unsigned index, u64 val, u64 mask);
+int kvm_set_shared_msr(unsigned index, u64 val, u64 mask);
 
 bool kvm_is_linear_rip(struct kvm_vcpu *vcpu, unsigned long linear_rip);
 

arch/x86/include/uapi/asm/vmx.h

@@ -67,6 +67,7 @@
 #define EXIT_REASON_EPT_MISCONFIG       49
 #define EXIT_REASON_INVEPT              50
 #define EXIT_REASON_PREEMPTION_TIMER    52
+#define EXIT_REASON_INVVPID             53
 #define EXIT_REASON_WBINVD              54
 #define EXIT_REASON_XSETBV              55
 #define EXIT_REASON_APIC_WRITE          56
@@ -114,6 +115,7 @@
 	{ EXIT_REASON_EOI_INDUCED,           "EOI_INDUCED" }, \
 	{ EXIT_REASON_INVALID_STATE,         "INVALID_STATE" }, \
 	{ EXIT_REASON_INVD,                  "INVD" }, \
+	{ EXIT_REASON_INVVPID,               "INVVPID" }, \
 	{ EXIT_REASON_INVPCID,               "INVPCID" }
 
 #endif /* _UAPIVMX_H */

arch/x86/kvm/i8254.c

@@ -262,8 +262,10 @@ void __kvm_migrate_pit_timer(struct kvm_vcpu *vcpu)
 		return;
 
 	timer = &pit->pit_state.timer;
+	mutex_lock(&pit->pit_state.lock);
 	if (hrtimer_cancel(timer))
 		hrtimer_start_expires(timer, HRTIMER_MODE_ABS);
+	mutex_unlock(&pit->pit_state.lock);
 }
 
 static void destroy_pit_timer(struct kvm_pit *pit)

arch/x86/kvm/paging_tmpl.h

@@ -298,7 +298,7 @@ static int FNAME(walk_addr_generic)(struct guest_walker *walker,
 	}
 #endif
 	walker->max_level = walker->level;
-	ASSERT(!is_long_mode(vcpu) && is_pae(vcpu));
+	ASSERT(!(is_long_mode(vcpu) && !is_pae(vcpu)));
 
 	accessed_dirty = PT_GUEST_ACCESSED_MASK;
 	pt_access = pte_access = ACC_ALL;

arch/x86/kvm/svm.c

@@ -3251,7 +3251,7 @@ static int wrmsr_interception(struct vcpu_svm *svm)
 	msr.host_initiated = false;
 
 	svm->next_rip = kvm_rip_read(&svm->vcpu) + 2;
-	if (svm_set_msr(&svm->vcpu, &msr)) {
+	if (kvm_set_msr(&svm->vcpu, &msr)) {
 		trace_kvm_msr_write_ex(ecx, data);
 		kvm_inject_gp(&svm->vcpu, 0);
 	} else {
@@ -3551,9 +3551,9 @@ static int handle_exit(struct kvm_vcpu *vcpu)
 
 	if (exit_code >= ARRAY_SIZE(svm_exit_handlers)
 	    || !svm_exit_handlers[exit_code]) {
-		kvm_run->exit_reason = KVM_EXIT_UNKNOWN;
-		kvm_run->hw.hardware_exit_reason = exit_code;
-		return 0;
+		WARN_ONCE(1, "vmx: unexpected exit reason 0x%x\n", exit_code);
+		kvm_queue_exception(vcpu, UD_VECTOR);
+		return 1;
 	}
 
 	return svm_exit_handlers[exit_code](svm);

arch/x86/kvm/vmx.c

@@ -2659,12 +2659,15 @@ static int vmx_set_msr(struct kvm_vcpu *vcpu, struct msr_data *msr_info)
 	default:
 		msr = find_msr_entry(vmx, msr_index);
 		if (msr) {
+			u64 old_msr_data = msr->data;
 			msr->data = data;
 			if (msr - vmx->guest_msrs < vmx->save_nmsrs) {
 				preempt_disable();
-				kvm_set_shared_msr(msr->index, msr->data,
-						   msr->mask);
+				ret = kvm_set_shared_msr(msr->index, msr->data,
+							 msr->mask);
 				preempt_enable();
+				if (ret)
+					msr->data = old_msr_data;
 			}
 			break;
 		}
@@ -5291,7 +5294,7 @@ static int handle_wrmsr(struct kvm_vcpu *vcpu)
 	msr.data = data;
 	msr.index = ecx;
 	msr.host_initiated = false;
-	if (vmx_set_msr(vcpu, &msr) != 0) {
+	if (kvm_set_msr(vcpu, &msr) != 0) {
 		trace_kvm_msr_write_ex(ecx, data);
 		kvm_inject_gp(vcpu, 0);
 		return 1;
@@ -6743,6 +6746,12 @@ static int handle_invept(struct kvm_vcpu *vcpu)
 	return 1;
 }
 
+static int handle_invvpid(struct kvm_vcpu *vcpu)
+{
+	kvm_queue_exception(vcpu, UD_VECTOR);
+	return 1;
+}
+
 /*
  * The exit handlers return 1 if the exit was handled fully and guest execution
  * may resume.  Otherwise they set the kvm_run parameter to indicate what needs
@@ -6788,6 +6797,7 @@ static int (*const kvm_vmx_exit_handlers[])(struct kvm_vcpu *vcpu) = {
 	[EXIT_REASON_MWAIT_INSTRUCTION]	      = handle_mwait,
 	[EXIT_REASON_MONITOR_INSTRUCTION]     = handle_monitor,
 	[EXIT_REASON_INVEPT]                  = handle_invept,
+	[EXIT_REASON_INVVPID]                 = handle_invvpid,
 };
 
 static const int kvm_vmx_max_exit_handlers =
@@ -7023,7 +7033,7 @@ static bool nested_vmx_exit_handled(struct kvm_vcpu *vcpu)
 	case EXIT_REASON_VMPTRST: case EXIT_REASON_VMREAD:
 	case EXIT_REASON_VMRESUME: case EXIT_REASON_VMWRITE:
 	case EXIT_REASON_VMOFF: case EXIT_REASON_VMON:
-	case EXIT_REASON_INVEPT:
+	case EXIT_REASON_INVEPT: case EXIT_REASON_INVVPID:
 		/*
 		 * VMX instructions trap unconditionally. This allows L1 to
 		 * emulate them for its L2 guest, i.e., allows 3-level nesting!
@@ -7164,10 +7174,10 @@ static int vmx_handle_exit(struct kvm_vcpu *vcpu)
 	    && kvm_vmx_exit_handlers[exit_reason])
 		return kvm_vmx_exit_handlers[exit_reason](vcpu);
 	else {
-		vcpu->run->exit_reason = KVM_EXIT_UNKNOWN;
-		vcpu->run->hw.hardware_exit_reason = exit_reason;
+		WARN_ONCE(1, "vmx: unexpected exit reason 0x%x\n", exit_reason);
+		kvm_queue_exception(vcpu, UD_VECTOR);
+		return 1;
 	}
-	return 0;
 }
 
 static void update_cr8_intercept(struct kvm_vcpu *vcpu, int tpr, int irr)

arch/x86/kvm/x86.c

@@ -229,20 +229,25 @@ static void kvm_shared_msr_cpu_online(void)
 		shared_msr_update(i, shared_msrs_global.msrs[i]);
 }
 
-void kvm_set_shared_msr(unsigned slot, u64 value, u64 mask)
+int kvm_set_shared_msr(unsigned slot, u64 value, u64 mask)
 {
 	unsigned int cpu = smp_processor_id();
 	struct kvm_shared_msrs *smsr = per_cpu_ptr(shared_msrs, cpu);
+	int err;
 
 	if (((value ^ smsr->values[slot].curr) & mask) == 0)
-		return;
+		return 0;
 	smsr->values[slot].curr = value;
-	wrmsrl(shared_msrs_global.msrs[slot], value);
+	err = wrmsrl_safe(shared_msrs_global.msrs[slot], value);
+	if (err)
+		return 1;
+
 	if (!smsr->registered) {
 		smsr->urn.on_user_return = kvm_on_user_return;
 		user_return_notifier_register(&smsr->urn);
 		smsr->registered = true;
 	}
+	return 0;
 }
 EXPORT_SYMBOL_GPL(kvm_set_shared_msr);
 
@@ -987,7 +992,6 @@ void kvm_enable_efer_bits(u64 mask)
 }
 EXPORT_SYMBOL_GPL(kvm_enable_efer_bits);
 
-
 /*
  * Writes msr value into into the appropriate "register".
  * Returns 0 on success, non-0 otherwise.
@@ -995,8 +999,34 @@ EXPORT_SYMBOL_GPL(kvm_enable_efer_bits);
  */
 int kvm_set_msr(struct kvm_vcpu *vcpu, struct msr_data *msr)
 {
+	switch (msr->index) {
+	case MSR_FS_BASE:
+	case MSR_GS_BASE:
+	case MSR_KERNEL_GS_BASE:
+	case MSR_CSTAR:
+	case MSR_LSTAR:
+		if (is_noncanonical_address(msr->data))
+			return 1;
+		break;
+	case MSR_IA32_SYSENTER_EIP:
+	case MSR_IA32_SYSENTER_ESP:
+		/*
+		 * IA32_SYSENTER_ESP and IA32_SYSENTER_EIP cause #GP if
+		 * non-canonical address is written on Intel but not on
+		 * AMD (which ignores the top 32-bits, because it does
+		 * not implement 64-bit SYSENTER).
+		 *
+		 * 64-bit code should hence be able to write a non-canonical
+		 * value on AMD.  Making the address canonical ensures that
+		 * vmentry does not fail on Intel after writing a non-canonical
+		 * value, and that something deterministic happens if the guest
+		 * invokes 64-bit SYSENTER.
+		 */
+		msr->data = get_canonical(msr->data);
+	}
 	return kvm_x86_ops->set_msr(vcpu, msr);
 }
+EXPORT_SYMBOL_GPL(kvm_set_msr);
 
 /*
  * Adapt set_msr() to msr_io()'s calling convention

include/linux/kvm_host.h

@@ -1080,6 +1080,7 @@ void kvm_device_get(struct kvm_device *dev);
 void kvm_device_put(struct kvm_device *dev);
 struct kvm_device *kvm_device_from_filp(struct file *filp);
 int kvm_register_device_ops(struct kvm_device_ops *ops, u32 type);
+void kvm_unregister_device_ops(u32 type);
 
 extern struct kvm_device_ops kvm_mpic_ops;
 extern struct kvm_device_ops kvm_xics_ops;

virt/kvm/iommu.c

@@ -43,13 +43,13 @@ static void kvm_iommu_put_pages(struct kvm *kvm,
 				gfn_t base_gfn, unsigned long npages);
 
 static pfn_t kvm_pin_pages(struct kvm_memory_slot *slot, gfn_t gfn,
-			   unsigned long size)
+			   unsigned long npages)
 {
 	gfn_t end_gfn;
 	pfn_t pfn;
 
 	pfn     = gfn_to_pfn_memslot(slot, gfn);
-	end_gfn = gfn + (size >> PAGE_SHIFT);
+	end_gfn = gfn + npages;
 	gfn    += 1;
 
 	if (is_error_noslot_pfn(pfn))
@@ -119,7 +119,7 @@ int kvm_iommu_map_pages(struct kvm *kvm, struct kvm_memory_slot *slot)
 		 * Pin all pages we are about to map in memory. This is
 		 * important because we unmap and unpin in 4kb steps later.
 		 */
-		pfn = kvm_pin_pages(slot, gfn, page_size);
+		pfn = kvm_pin_pages(slot, gfn, page_size >> PAGE_SHIFT);
 		if (is_error_noslot_pfn(pfn)) {
 			gfn += 1;
 			continue;
@@ -131,7 +131,7 @@ int kvm_iommu_map_pages(struct kvm *kvm, struct kvm_memory_slot *slot)
 		if (r) {
 			printk(KERN_ERR "kvm_iommu_map_address:"
 			       "iommu failed to map pfn=%llx\n", pfn);
-			kvm_unpin_pages(kvm, pfn, page_size);
+			kvm_unpin_pages(kvm, pfn, page_size >> PAGE_SHIFT);
 			goto unmap_pages;
 		}
 

virt/kvm/kvm_main.c

@@ -2354,6 +2354,12 @@ int kvm_register_device_ops(struct kvm_device_ops *ops, u32 type)
 	return 0;
 }
 
+void kvm_unregister_device_ops(u32 type)
+{
+	if (kvm_device_ops_table[type] != NULL)
+		kvm_device_ops_table[type] = NULL;
+}
+
 static int kvm_ioctl_create_device(struct kvm *kvm,
 				   struct kvm_create_device *cd)
 {
@@ -3328,5 +3334,6 @@ void kvm_exit(void)
 	kvm_arch_exit();
 	kvm_irqfd_exit();
 	free_cpumask_var(cpus_hardware_enabled);
+	kvm_vfio_ops_exit();
 }
 EXPORT_SYMBOL_GPL(kvm_exit);

virt/kvm/vfio.c

@@ -283,3 +283,8 @@ int kvm_vfio_ops_init(void)
 {
 	return kvm_register_device_ops(&kvm_vfio_ops, KVM_DEV_TYPE_VFIO);
 }
+
+void kvm_vfio_ops_exit(void)
+{
+	kvm_unregister_device_ops(KVM_DEV_TYPE_VFIO);
+}

virt/kvm/vfio.h

@@ -3,11 +3,15 @@
 
 #ifdef CONFIG_KVM_VFIO
 int kvm_vfio_ops_init(void);
+void kvm_vfio_ops_exit(void);
 #else
 static inline int kvm_vfio_ops_init(void)
 {
 	return 0;
 }
+static inline void kvm_vfio_ops_exit(void)
+{
+}
 #endif
 
 #endif