5 * This work is licensed under the terms of the GNU GPL, version 2. See
6 * the COPYING file in the top-level directory.
9 #include <linux/types.h>
10 #include <linux/hardirq.h>
11 #include <linux/list.h>
12 #include <linux/mutex.h>
13 #include <linux/spinlock.h>
14 #include <linux/signal.h>
15 #include <linux/sched.h>
17 #include <linux/preempt.h>
18 #include <linux/msi.h>
19 #include <asm/signal.h>
21 #include <linux/kvm.h>
22 #include <linux/kvm_para.h>
24 #include <linux/kvm_types.h>
26 #include <asm/kvm_host.h>
29 * vcpu->requests bit members
31 #define KVM_REQ_TLB_FLUSH 0
32 #define KVM_REQ_MIGRATE_TIMER 1
33 #define KVM_REQ_REPORT_TPR_ACCESS 2
34 #define KVM_REQ_MMU_RELOAD 3
35 #define KVM_REQ_TRIPLE_FAULT 4
36 #define KVM_REQ_PENDING_TIMER 5
37 #define KVM_REQ_UNHALT 6
38 #define KVM_REQ_MMU_SYNC 7
39 #define KVM_REQ_KVMCLOCK_UPDATE 8
40 #define KVM_REQ_KICK 9
41 #define KVM_REQ_DEACTIVATE_FPU 10
43 #define KVM_USERSPACE_IRQ_SOURCE_ID 0
47 extern struct kmem_cache *kvm_vcpu_cache;
50 * It would be nice to use something smarter than a linear search, TBD...
51 * Thankfully we dont expect many devices to register (famous last words :),
52 * so until then it will suffice. At least its abstracted so we can change
57 #define NR_IOBUS_DEVS 6
58 struct kvm_io_device *devs[NR_IOBUS_DEVS];
67 int kvm_io_bus_write(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr,
68 int len, const void *val);
69 int kvm_io_bus_read(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr, int len,
71 int kvm_io_bus_register_dev(struct kvm *kvm, enum kvm_bus bus_idx,
72 struct kvm_io_device *dev);
73 int kvm_io_bus_unregister_dev(struct kvm *kvm, enum kvm_bus bus_idx,
74 struct kvm_io_device *dev);
78 #ifdef CONFIG_PREEMPT_NOTIFIERS
79 struct preempt_notifier preempt_notifier;
85 unsigned long requests;
86 unsigned long guest_debug;
94 struct kvm_vcpu_stat stat;
96 #ifdef CONFIG_HAS_IOMEM
98 int mmio_read_completed;
101 unsigned char mmio_data[8];
102 gpa_t mmio_phys_addr;
105 struct kvm_vcpu_arch arch;
108 struct kvm_memory_slot {
110 unsigned long npages;
113 unsigned long *dirty_bitmap;
115 unsigned long rmap_pde;
117 } *lpage_info[KVM_NR_PAGE_SIZES - 1];
118 unsigned long userspace_addr;
122 struct kvm_kernel_irq_routing_entry {
125 int (*set)(struct kvm_kernel_irq_routing_entry *e,
126 struct kvm *kvm, int irq_source_id, int level);
134 struct hlist_node link;
137 #ifdef __KVM_HAVE_IOAPIC
139 struct kvm_irq_routing_table {
140 int chip[KVM_NR_IRQCHIPS][KVM_IOAPIC_NUM_PINS];
141 struct kvm_kernel_irq_routing_entry *rt_entries;
144 * Array indexed by gsi. Each entry contains list of irq chips
145 * the gsi is connected to.
147 struct hlist_head map[0];
152 struct kvm_irq_routing_table {};
156 struct kvm_memslots {
158 struct kvm_memory_slot memslots[KVM_MEMORY_SLOTS +
159 KVM_PRIVATE_MEM_SLOTS];
164 spinlock_t requests_lock;
165 struct mutex slots_lock;
166 struct mm_struct *mm; /* userspace tied to this vm */
167 struct kvm_memslots *memslots;
168 struct srcu_struct srcu;
169 #ifdef CONFIG_KVM_APIC_ARCHITECTURE
171 struct kvm_vcpu *bsp_vcpu;
173 struct kvm_vcpu *vcpus[KVM_MAX_VCPUS];
174 atomic_t online_vcpus;
175 struct list_head vm_list;
177 struct kvm_io_bus *buses[KVM_NR_BUSES];
178 #ifdef CONFIG_HAVE_KVM_EVENTFD
181 struct list_head items;
183 struct list_head ioeventfds;
185 struct kvm_vm_stat stat;
186 struct kvm_arch arch;
187 atomic_t users_count;
188 #ifdef KVM_COALESCED_MMIO_PAGE_OFFSET
189 struct kvm_coalesced_mmio_dev *coalesced_mmio_dev;
190 struct kvm_coalesced_mmio_ring *coalesced_mmio_ring;
193 struct mutex irq_lock;
194 #ifdef CONFIG_HAVE_KVM_IRQCHIP
195 struct kvm_irq_routing_table *irq_routing;
196 struct hlist_head mask_notifier_list;
197 struct hlist_head irq_ack_notifier_list;
200 #ifdef KVM_ARCH_WANT_MMU_NOTIFIER
201 struct mmu_notifier mmu_notifier;
202 unsigned long mmu_notifier_seq;
203 long mmu_notifier_count;
207 /* The guest did something we don't support. */
208 #define pr_unimpl(vcpu, fmt, ...) \
210 if (printk_ratelimit()) \
211 printk(KERN_ERR "kvm: %i: cpu%i " fmt, \
212 current->tgid, (vcpu)->vcpu_id , ## __VA_ARGS__); \
215 #define kvm_printf(kvm, fmt ...) printk(KERN_DEBUG fmt)
216 #define vcpu_printf(vcpu, fmt...) kvm_printf(vcpu->kvm, fmt)
218 static inline struct kvm_vcpu *kvm_get_vcpu(struct kvm *kvm, int i)
221 return kvm->vcpus[i];
224 #define kvm_for_each_vcpu(idx, vcpup, kvm) \
225 for (idx = 0, vcpup = kvm_get_vcpu(kvm, idx); \
226 idx < atomic_read(&kvm->online_vcpus) && vcpup; \
227 vcpup = kvm_get_vcpu(kvm, ++idx))
229 int kvm_vcpu_init(struct kvm_vcpu *vcpu, struct kvm *kvm, unsigned id);
230 void kvm_vcpu_uninit(struct kvm_vcpu *vcpu);
232 void vcpu_load(struct kvm_vcpu *vcpu);
233 void vcpu_put(struct kvm_vcpu *vcpu);
235 int kvm_init(void *opaque, unsigned int vcpu_size,
236 struct module *module);
239 void kvm_get_kvm(struct kvm *kvm);
240 void kvm_put_kvm(struct kvm *kvm);
242 #define HPA_MSB ((sizeof(hpa_t) * 8) - 1)
243 #define HPA_ERR_MASK ((hpa_t)1 << HPA_MSB)
244 static inline int is_error_hpa(hpa_t hpa) { return hpa >> HPA_MSB; }
245 struct page *gva_to_page(struct kvm_vcpu *vcpu, gva_t gva);
247 extern struct page *bad_page;
248 extern pfn_t bad_pfn;
250 int is_error_page(struct page *page);
251 int is_error_pfn(pfn_t pfn);
252 int kvm_is_error_hva(unsigned long addr);
253 int kvm_set_memory_region(struct kvm *kvm,
254 struct kvm_userspace_memory_region *mem,
256 int __kvm_set_memory_region(struct kvm *kvm,
257 struct kvm_userspace_memory_region *mem,
259 int kvm_arch_prepare_memory_region(struct kvm *kvm,
260 struct kvm_memory_slot *memslot,
261 struct kvm_memory_slot old,
262 struct kvm_userspace_memory_region *mem,
264 void kvm_arch_commit_memory_region(struct kvm *kvm,
265 struct kvm_userspace_memory_region *mem,
266 struct kvm_memory_slot old,
268 void kvm_disable_largepages(void);
269 void kvm_arch_flush_shadow(struct kvm *kvm);
270 gfn_t unalias_gfn(struct kvm *kvm, gfn_t gfn);
271 gfn_t unalias_gfn_instantiation(struct kvm *kvm, gfn_t gfn);
273 struct page *gfn_to_page(struct kvm *kvm, gfn_t gfn);
274 unsigned long gfn_to_hva(struct kvm *kvm, gfn_t gfn);
275 void kvm_release_page_clean(struct page *page);
276 void kvm_release_page_dirty(struct page *page);
277 void kvm_set_page_dirty(struct page *page);
278 void kvm_set_page_accessed(struct page *page);
280 pfn_t gfn_to_pfn(struct kvm *kvm, gfn_t gfn);
281 pfn_t gfn_to_pfn_memslot(struct kvm *kvm,
282 struct kvm_memory_slot *slot, gfn_t gfn);
283 int memslot_id(struct kvm *kvm, gfn_t gfn);
284 void kvm_release_pfn_dirty(pfn_t);
285 void kvm_release_pfn_clean(pfn_t pfn);
286 void kvm_set_pfn_dirty(pfn_t pfn);
287 void kvm_set_pfn_accessed(pfn_t pfn);
288 void kvm_get_pfn(pfn_t pfn);
290 int kvm_read_guest_page(struct kvm *kvm, gfn_t gfn, void *data, int offset,
292 int kvm_read_guest_atomic(struct kvm *kvm, gpa_t gpa, void *data,
294 int kvm_read_guest(struct kvm *kvm, gpa_t gpa, void *data, unsigned long len);
295 int kvm_write_guest_page(struct kvm *kvm, gfn_t gfn, const void *data,
296 int offset, int len);
297 int kvm_write_guest(struct kvm *kvm, gpa_t gpa, const void *data,
299 int kvm_clear_guest_page(struct kvm *kvm, gfn_t gfn, int offset, int len);
300 int kvm_clear_guest(struct kvm *kvm, gpa_t gpa, unsigned long len);
301 struct kvm_memory_slot *gfn_to_memslot(struct kvm *kvm, gfn_t gfn);
302 int kvm_is_visible_gfn(struct kvm *kvm, gfn_t gfn);
303 void mark_page_dirty(struct kvm *kvm, gfn_t gfn);
305 void kvm_vcpu_block(struct kvm_vcpu *vcpu);
306 void kvm_vcpu_on_spin(struct kvm_vcpu *vcpu);
307 void kvm_resched(struct kvm_vcpu *vcpu);
308 void kvm_load_guest_fpu(struct kvm_vcpu *vcpu);
309 void kvm_put_guest_fpu(struct kvm_vcpu *vcpu);
310 void kvm_flush_remote_tlbs(struct kvm *kvm);
311 void kvm_reload_remote_mmus(struct kvm *kvm);
313 long kvm_arch_dev_ioctl(struct file *filp,
314 unsigned int ioctl, unsigned long arg);
315 long kvm_arch_vcpu_ioctl(struct file *filp,
316 unsigned int ioctl, unsigned long arg);
318 int kvm_dev_ioctl_check_extension(long ext);
320 int kvm_get_dirty_log(struct kvm *kvm,
321 struct kvm_dirty_log *log, int *is_dirty);
322 int kvm_vm_ioctl_get_dirty_log(struct kvm *kvm,
323 struct kvm_dirty_log *log);
325 int kvm_vm_ioctl_set_memory_region(struct kvm *kvm,
327 kvm_userspace_memory_region *mem,
329 long kvm_arch_vm_ioctl(struct file *filp,
330 unsigned int ioctl, unsigned long arg);
332 int kvm_arch_vcpu_ioctl_get_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu);
333 int kvm_arch_vcpu_ioctl_set_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu);
335 int kvm_arch_vcpu_ioctl_translate(struct kvm_vcpu *vcpu,
336 struct kvm_translation *tr);
338 int kvm_arch_vcpu_ioctl_get_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs);
339 int kvm_arch_vcpu_ioctl_set_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs);
340 int kvm_arch_vcpu_ioctl_get_sregs(struct kvm_vcpu *vcpu,
341 struct kvm_sregs *sregs);
342 int kvm_arch_vcpu_ioctl_set_sregs(struct kvm_vcpu *vcpu,
343 struct kvm_sregs *sregs);
344 int kvm_arch_vcpu_ioctl_get_mpstate(struct kvm_vcpu *vcpu,
345 struct kvm_mp_state *mp_state);
346 int kvm_arch_vcpu_ioctl_set_mpstate(struct kvm_vcpu *vcpu,
347 struct kvm_mp_state *mp_state);
348 int kvm_arch_vcpu_ioctl_set_guest_debug(struct kvm_vcpu *vcpu,
349 struct kvm_guest_debug *dbg);
350 int kvm_arch_vcpu_ioctl_run(struct kvm_vcpu *vcpu, struct kvm_run *kvm_run);
352 int kvm_arch_init(void *opaque);
353 void kvm_arch_exit(void);
355 int kvm_arch_vcpu_init(struct kvm_vcpu *vcpu);
356 void kvm_arch_vcpu_uninit(struct kvm_vcpu *vcpu);
358 void kvm_arch_vcpu_free(struct kvm_vcpu *vcpu);
359 void kvm_arch_vcpu_load(struct kvm_vcpu *vcpu, int cpu);
360 void kvm_arch_vcpu_put(struct kvm_vcpu *vcpu);
361 struct kvm_vcpu *kvm_arch_vcpu_create(struct kvm *kvm, unsigned int id);
362 int kvm_arch_vcpu_setup(struct kvm_vcpu *vcpu);
363 void kvm_arch_vcpu_destroy(struct kvm_vcpu *vcpu);
365 int kvm_arch_vcpu_reset(struct kvm_vcpu *vcpu);
366 int kvm_arch_hardware_enable(void *garbage);
367 void kvm_arch_hardware_disable(void *garbage);
368 int kvm_arch_hardware_setup(void);
369 void kvm_arch_hardware_unsetup(void);
370 void kvm_arch_check_processor_compat(void *rtn);
371 int kvm_arch_vcpu_runnable(struct kvm_vcpu *vcpu);
373 void kvm_free_physmem(struct kvm *kvm);
375 struct kvm *kvm_arch_create_vm(void);
376 void kvm_arch_destroy_vm(struct kvm *kvm);
377 void kvm_free_all_assigned_devices(struct kvm *kvm);
378 void kvm_arch_sync_events(struct kvm *kvm);
380 int kvm_cpu_has_pending_timer(struct kvm_vcpu *vcpu);
381 void kvm_vcpu_kick(struct kvm_vcpu *vcpu);
383 int kvm_is_mmio_pfn(pfn_t pfn);
385 struct kvm_irq_ack_notifier {
386 struct hlist_node link;
388 void (*irq_acked)(struct kvm_irq_ack_notifier *kian);
391 #define KVM_ASSIGNED_MSIX_PENDING 0x1
392 struct kvm_guest_msix_entry {
398 struct kvm_assigned_dev_kernel {
399 struct kvm_irq_ack_notifier ack_notifier;
400 struct work_struct interrupt_work;
401 struct list_head list;
405 unsigned int entries_nr;
407 bool host_irq_disabled;
408 struct msix_entry *host_msix_entries;
410 struct kvm_guest_msix_entry *guest_msix_entries;
411 unsigned long irq_requested_type;
416 spinlock_t assigned_dev_lock;
419 struct kvm_irq_mask_notifier {
420 void (*func)(struct kvm_irq_mask_notifier *kimn, bool masked);
422 struct hlist_node link;
425 void kvm_register_irq_mask_notifier(struct kvm *kvm, int irq,
426 struct kvm_irq_mask_notifier *kimn);
427 void kvm_unregister_irq_mask_notifier(struct kvm *kvm, int irq,
428 struct kvm_irq_mask_notifier *kimn);
429 void kvm_fire_mask_notifiers(struct kvm *kvm, int irq, bool mask);
431 #ifdef __KVM_HAVE_IOAPIC
432 void kvm_get_intr_delivery_bitmask(struct kvm_ioapic *ioapic,
433 union kvm_ioapic_redirect_entry *entry,
434 unsigned long *deliver_bitmask);
436 int kvm_set_irq(struct kvm *kvm, int irq_source_id, u32 irq, int level);
437 void kvm_notify_acked_irq(struct kvm *kvm, unsigned irqchip, unsigned pin);
438 void kvm_register_irq_ack_notifier(struct kvm *kvm,
439 struct kvm_irq_ack_notifier *kian);
440 void kvm_unregister_irq_ack_notifier(struct kvm *kvm,
441 struct kvm_irq_ack_notifier *kian);
442 int kvm_request_irq_source_id(struct kvm *kvm);
443 void kvm_free_irq_source_id(struct kvm *kvm, int irq_source_id);
445 /* For vcpu->arch.iommu_flags */
446 #define KVM_IOMMU_CACHE_COHERENCY 0x1
448 #ifdef CONFIG_IOMMU_API
449 int kvm_iommu_map_pages(struct kvm *kvm, struct kvm_memory_slot *slot);
450 int kvm_iommu_map_guest(struct kvm *kvm);
451 int kvm_iommu_unmap_guest(struct kvm *kvm);
452 int kvm_assign_device(struct kvm *kvm,
453 struct kvm_assigned_dev_kernel *assigned_dev);
454 int kvm_deassign_device(struct kvm *kvm,
455 struct kvm_assigned_dev_kernel *assigned_dev);
456 #else /* CONFIG_IOMMU_API */
457 static inline int kvm_iommu_map_pages(struct kvm *kvm,
459 unsigned long npages)
464 static inline int kvm_iommu_map_guest(struct kvm *kvm)
469 static inline int kvm_iommu_unmap_guest(struct kvm *kvm)
474 static inline int kvm_assign_device(struct kvm *kvm,
475 struct kvm_assigned_dev_kernel *assigned_dev)
480 static inline int kvm_deassign_device(struct kvm *kvm,
481 struct kvm_assigned_dev_kernel *assigned_dev)
485 #endif /* CONFIG_IOMMU_API */
487 static inline void kvm_guest_enter(void)
489 account_system_vtime(current);
490 current->flags |= PF_VCPU;
493 static inline void kvm_guest_exit(void)
495 account_system_vtime(current);
496 current->flags &= ~PF_VCPU;
499 static inline gpa_t gfn_to_gpa(gfn_t gfn)
501 return (gpa_t)gfn << PAGE_SHIFT;
504 static inline hpa_t pfn_to_hpa(pfn_t pfn)
506 return (hpa_t)pfn << PAGE_SHIFT;
509 static inline void kvm_migrate_timers(struct kvm_vcpu *vcpu)
511 set_bit(KVM_REQ_MIGRATE_TIMER, &vcpu->requests);
519 struct kvm_stats_debugfs_item {
522 enum kvm_stat_kind kind;
523 struct dentry *dentry;
525 extern struct kvm_stats_debugfs_item debugfs_entries[];
526 extern struct dentry *kvm_debugfs_dir;
528 #ifdef KVM_ARCH_WANT_MMU_NOTIFIER
529 static inline int mmu_notifier_retry(struct kvm_vcpu *vcpu, unsigned long mmu_seq)
531 if (unlikely(vcpu->kvm->mmu_notifier_count))
534 * Both reads happen under the mmu_lock and both values are
535 * modified under mmu_lock, so there's no need of smb_rmb()
536 * here in between, otherwise mmu_notifier_count should be
537 * read before mmu_notifier_seq, see
538 * mmu_notifier_invalidate_range_end write side.
540 if (vcpu->kvm->mmu_notifier_seq != mmu_seq)
546 #ifndef KVM_ARCH_HAS_UNALIAS_INSTANTIATION
547 #define unalias_gfn_instantiation unalias_gfn
550 #ifdef CONFIG_HAVE_KVM_IRQCHIP
552 #define KVM_MAX_IRQ_ROUTES 1024
554 int kvm_setup_default_irq_routing(struct kvm *kvm);
555 int kvm_set_irq_routing(struct kvm *kvm,
556 const struct kvm_irq_routing_entry *entries,
559 void kvm_free_irq_routing(struct kvm *kvm);
563 static inline void kvm_free_irq_routing(struct kvm *kvm) {}
567 #ifdef CONFIG_HAVE_KVM_EVENTFD
569 void kvm_eventfd_init(struct kvm *kvm);
570 int kvm_irqfd(struct kvm *kvm, int fd, int gsi, int flags);
571 void kvm_irqfd_release(struct kvm *kvm);
572 int kvm_ioeventfd(struct kvm *kvm, struct kvm_ioeventfd *args);
576 static inline void kvm_eventfd_init(struct kvm *kvm) {}
577 static inline int kvm_irqfd(struct kvm *kvm, int fd, int gsi, int flags)
582 static inline void kvm_irqfd_release(struct kvm *kvm) {}
583 static inline int kvm_ioeventfd(struct kvm *kvm, struct kvm_ioeventfd *args)
588 #endif /* CONFIG_HAVE_KVM_EVENTFD */
590 #ifdef CONFIG_KVM_APIC_ARCHITECTURE
591 static inline bool kvm_vcpu_is_bsp(struct kvm_vcpu *vcpu)
593 return vcpu->kvm->bsp_vcpu_id == vcpu->vcpu_id;
597 #ifdef __KVM_HAVE_DEVICE_ASSIGNMENT
599 long kvm_vm_ioctl_assigned_device(struct kvm *kvm, unsigned ioctl,
604 static inline long kvm_vm_ioctl_assigned_device(struct kvm *kvm, unsigned ioctl,