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
42 #define KVM_USERSPACE_IRQ_SOURCE_ID 0
46 extern struct kmem_cache *kvm_vcpu_cache;
49 * It would be nice to use something smarter than a linear search, TBD...
50 * Thankfully we dont expect many devices to register (famous last words :),
51 * so until then it will suffice. At least its abstracted so we can change
56 #define NR_IOBUS_DEVS 6
57 struct kvm_io_device *devs[NR_IOBUS_DEVS];
66 int kvm_io_bus_write(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr,
67 int len, const void *val);
68 int kvm_io_bus_read(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr, int len,
70 int kvm_io_bus_register_dev(struct kvm *kvm, enum kvm_bus bus_idx,
71 struct kvm_io_device *dev);
72 int kvm_io_bus_unregister_dev(struct kvm *kvm, enum kvm_bus bus_idx,
73 struct kvm_io_device *dev);
77 #ifdef CONFIG_PREEMPT_NOTIFIERS
78 struct preempt_notifier preempt_notifier;
84 unsigned long requests;
85 unsigned long guest_debug;
93 struct kvm_vcpu_stat stat;
95 #ifdef CONFIG_HAS_IOMEM
97 int mmio_read_completed;
100 unsigned char mmio_data[8];
101 gpa_t mmio_phys_addr;
104 struct kvm_vcpu_arch arch;
107 struct kvm_memory_slot {
109 unsigned long npages;
112 unsigned long *dirty_bitmap;
114 unsigned long rmap_pde;
116 } *lpage_info[KVM_NR_PAGE_SIZES - 1];
117 unsigned long userspace_addr;
121 struct kvm_kernel_irq_routing_entry {
124 int (*set)(struct kvm_kernel_irq_routing_entry *e,
125 struct kvm *kvm, int irq_source_id, int level);
133 struct hlist_node link;
136 #ifdef __KVM_HAVE_IOAPIC
138 struct kvm_irq_routing_table {
139 int chip[KVM_NR_IRQCHIPS][KVM_IOAPIC_NUM_PINS];
140 struct kvm_kernel_irq_routing_entry *rt_entries;
143 * Array indexed by gsi. Each entry contains list of irq chips
144 * the gsi is connected to.
146 struct hlist_head map[0];
151 struct kvm_irq_routing_table {};
155 struct kvm_memslots {
157 struct kvm_memory_slot memslots[KVM_MEMORY_SLOTS +
158 KVM_PRIVATE_MEM_SLOTS];
163 spinlock_t requests_lock;
164 struct rw_semaphore slots_lock;
165 struct mm_struct *mm; /* userspace tied to this vm */
166 struct kvm_memslots *memslots;
167 struct srcu_struct srcu;
168 #ifdef CONFIG_KVM_APIC_ARCHITECTURE
170 struct kvm_vcpu *bsp_vcpu;
172 struct kvm_vcpu *vcpus[KVM_MAX_VCPUS];
173 atomic_t online_vcpus;
174 struct list_head vm_list;
176 struct kvm_io_bus *buses[KVM_NR_BUSES];
177 #ifdef CONFIG_HAVE_KVM_EVENTFD
180 struct list_head items;
182 struct list_head ioeventfds;
184 struct kvm_vm_stat stat;
185 struct kvm_arch arch;
186 atomic_t users_count;
187 #ifdef KVM_COALESCED_MMIO_PAGE_OFFSET
188 struct kvm_coalesced_mmio_dev *coalesced_mmio_dev;
189 struct kvm_coalesced_mmio_ring *coalesced_mmio_ring;
192 struct mutex irq_lock;
193 #ifdef CONFIG_HAVE_KVM_IRQCHIP
194 struct kvm_irq_routing_table *irq_routing;
195 struct hlist_head mask_notifier_list;
196 struct hlist_head irq_ack_notifier_list;
199 #ifdef KVM_ARCH_WANT_MMU_NOTIFIER
200 struct mmu_notifier mmu_notifier;
201 unsigned long mmu_notifier_seq;
202 long mmu_notifier_count;
206 /* The guest did something we don't support. */
207 #define pr_unimpl(vcpu, fmt, ...) \
209 if (printk_ratelimit()) \
210 printk(KERN_ERR "kvm: %i: cpu%i " fmt, \
211 current->tgid, (vcpu)->vcpu_id , ## __VA_ARGS__); \
214 #define kvm_printf(kvm, fmt ...) printk(KERN_DEBUG fmt)
215 #define vcpu_printf(vcpu, fmt...) kvm_printf(vcpu->kvm, fmt)
217 static inline struct kvm_vcpu *kvm_get_vcpu(struct kvm *kvm, int i)
220 return kvm->vcpus[i];
223 #define kvm_for_each_vcpu(idx, vcpup, kvm) \
224 for (idx = 0, vcpup = kvm_get_vcpu(kvm, idx); \
225 idx < atomic_read(&kvm->online_vcpus) && vcpup; \
226 vcpup = kvm_get_vcpu(kvm, ++idx))
228 int kvm_vcpu_init(struct kvm_vcpu *vcpu, struct kvm *kvm, unsigned id);
229 void kvm_vcpu_uninit(struct kvm_vcpu *vcpu);
231 void vcpu_load(struct kvm_vcpu *vcpu);
232 void vcpu_put(struct kvm_vcpu *vcpu);
234 int kvm_init(void *opaque, unsigned int vcpu_size,
235 struct module *module);
238 void kvm_get_kvm(struct kvm *kvm);
239 void kvm_put_kvm(struct kvm *kvm);
241 #define HPA_MSB ((sizeof(hpa_t) * 8) - 1)
242 #define HPA_ERR_MASK ((hpa_t)1 << HPA_MSB)
243 static inline int is_error_hpa(hpa_t hpa) { return hpa >> HPA_MSB; }
244 struct page *gva_to_page(struct kvm_vcpu *vcpu, gva_t gva);
246 extern struct page *bad_page;
247 extern pfn_t bad_pfn;
249 int is_error_page(struct page *page);
250 int is_error_pfn(pfn_t pfn);
251 int kvm_is_error_hva(unsigned long addr);
252 int kvm_set_memory_region(struct kvm *kvm,
253 struct kvm_userspace_memory_region *mem,
255 int __kvm_set_memory_region(struct kvm *kvm,
256 struct kvm_userspace_memory_region *mem,
258 int kvm_arch_prepare_memory_region(struct kvm *kvm,
259 struct kvm_memory_slot *memslot,
260 struct kvm_memory_slot old,
261 struct kvm_userspace_memory_region *mem,
263 void kvm_arch_commit_memory_region(struct kvm *kvm,
264 struct kvm_userspace_memory_region *mem,
265 struct kvm_memory_slot old,
267 void kvm_disable_largepages(void);
268 void kvm_arch_flush_shadow(struct kvm *kvm);
269 gfn_t unalias_gfn(struct kvm *kvm, gfn_t gfn);
270 gfn_t unalias_gfn_instantiation(struct kvm *kvm, gfn_t gfn);
272 struct page *gfn_to_page(struct kvm *kvm, gfn_t gfn);
273 unsigned long gfn_to_hva(struct kvm *kvm, gfn_t gfn);
274 void kvm_release_page_clean(struct page *page);
275 void kvm_release_page_dirty(struct page *page);
276 void kvm_set_page_dirty(struct page *page);
277 void kvm_set_page_accessed(struct page *page);
279 pfn_t gfn_to_pfn(struct kvm *kvm, gfn_t gfn);
280 pfn_t gfn_to_pfn_memslot(struct kvm *kvm,
281 struct kvm_memory_slot *slot, gfn_t gfn);
282 int memslot_id(struct kvm *kvm, gfn_t gfn);
283 void kvm_release_pfn_dirty(pfn_t);
284 void kvm_release_pfn_clean(pfn_t pfn);
285 void kvm_set_pfn_dirty(pfn_t pfn);
286 void kvm_set_pfn_accessed(pfn_t pfn);
287 void kvm_get_pfn(pfn_t pfn);
289 int kvm_read_guest_page(struct kvm *kvm, gfn_t gfn, void *data, int offset,
291 int kvm_read_guest_atomic(struct kvm *kvm, gpa_t gpa, void *data,
293 int kvm_read_guest(struct kvm *kvm, gpa_t gpa, void *data, unsigned long len);
294 int kvm_write_guest_page(struct kvm *kvm, gfn_t gfn, const void *data,
295 int offset, int len);
296 int kvm_write_guest(struct kvm *kvm, gpa_t gpa, const void *data,
298 int kvm_clear_guest_page(struct kvm *kvm, gfn_t gfn, int offset, int len);
299 int kvm_clear_guest(struct kvm *kvm, gpa_t gpa, unsigned long len);
300 struct kvm_memory_slot *gfn_to_memslot(struct kvm *kvm, gfn_t gfn);
301 int kvm_is_visible_gfn(struct kvm *kvm, gfn_t gfn);
302 void mark_page_dirty(struct kvm *kvm, gfn_t gfn);
304 void kvm_vcpu_block(struct kvm_vcpu *vcpu);
305 void kvm_vcpu_on_spin(struct kvm_vcpu *vcpu);
306 void kvm_resched(struct kvm_vcpu *vcpu);
307 void kvm_load_guest_fpu(struct kvm_vcpu *vcpu);
308 void kvm_put_guest_fpu(struct kvm_vcpu *vcpu);
309 void kvm_flush_remote_tlbs(struct kvm *kvm);
310 void kvm_reload_remote_mmus(struct kvm *kvm);
312 long kvm_arch_dev_ioctl(struct file *filp,
313 unsigned int ioctl, unsigned long arg);
314 long kvm_arch_vcpu_ioctl(struct file *filp,
315 unsigned int ioctl, unsigned long arg);
317 int kvm_dev_ioctl_check_extension(long ext);
319 int kvm_get_dirty_log(struct kvm *kvm,
320 struct kvm_dirty_log *log, int *is_dirty);
321 int kvm_vm_ioctl_get_dirty_log(struct kvm *kvm,
322 struct kvm_dirty_log *log);
324 int kvm_vm_ioctl_set_memory_region(struct kvm *kvm,
326 kvm_userspace_memory_region *mem,
328 long kvm_arch_vm_ioctl(struct file *filp,
329 unsigned int ioctl, unsigned long arg);
331 int kvm_arch_vcpu_ioctl_get_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu);
332 int kvm_arch_vcpu_ioctl_set_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu);
334 int kvm_arch_vcpu_ioctl_translate(struct kvm_vcpu *vcpu,
335 struct kvm_translation *tr);
337 int kvm_arch_vcpu_ioctl_get_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs);
338 int kvm_arch_vcpu_ioctl_set_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs);
339 int kvm_arch_vcpu_ioctl_get_sregs(struct kvm_vcpu *vcpu,
340 struct kvm_sregs *sregs);
341 int kvm_arch_vcpu_ioctl_set_sregs(struct kvm_vcpu *vcpu,
342 struct kvm_sregs *sregs);
343 int kvm_arch_vcpu_ioctl_get_mpstate(struct kvm_vcpu *vcpu,
344 struct kvm_mp_state *mp_state);
345 int kvm_arch_vcpu_ioctl_set_mpstate(struct kvm_vcpu *vcpu,
346 struct kvm_mp_state *mp_state);
347 int kvm_arch_vcpu_ioctl_set_guest_debug(struct kvm_vcpu *vcpu,
348 struct kvm_guest_debug *dbg);
349 int kvm_arch_vcpu_ioctl_run(struct kvm_vcpu *vcpu, struct kvm_run *kvm_run);
351 int kvm_arch_init(void *opaque);
352 void kvm_arch_exit(void);
354 int kvm_arch_vcpu_init(struct kvm_vcpu *vcpu);
355 void kvm_arch_vcpu_uninit(struct kvm_vcpu *vcpu);
357 void kvm_arch_vcpu_free(struct kvm_vcpu *vcpu);
358 void kvm_arch_vcpu_load(struct kvm_vcpu *vcpu, int cpu);
359 void kvm_arch_vcpu_put(struct kvm_vcpu *vcpu);
360 struct kvm_vcpu *kvm_arch_vcpu_create(struct kvm *kvm, unsigned int id);
361 int kvm_arch_vcpu_setup(struct kvm_vcpu *vcpu);
362 void kvm_arch_vcpu_destroy(struct kvm_vcpu *vcpu);
364 int kvm_arch_vcpu_reset(struct kvm_vcpu *vcpu);
365 int kvm_arch_hardware_enable(void *garbage);
366 void kvm_arch_hardware_disable(void *garbage);
367 int kvm_arch_hardware_setup(void);
368 void kvm_arch_hardware_unsetup(void);
369 void kvm_arch_check_processor_compat(void *rtn);
370 int kvm_arch_vcpu_runnable(struct kvm_vcpu *vcpu);
372 void kvm_free_physmem(struct kvm *kvm);
374 struct kvm *kvm_arch_create_vm(void);
375 void kvm_arch_destroy_vm(struct kvm *kvm);
376 void kvm_free_all_assigned_devices(struct kvm *kvm);
377 void kvm_arch_sync_events(struct kvm *kvm);
379 int kvm_cpu_has_pending_timer(struct kvm_vcpu *vcpu);
380 void kvm_vcpu_kick(struct kvm_vcpu *vcpu);
382 int kvm_is_mmio_pfn(pfn_t pfn);
384 struct kvm_irq_ack_notifier {
385 struct hlist_node link;
387 void (*irq_acked)(struct kvm_irq_ack_notifier *kian);
390 #define KVM_ASSIGNED_MSIX_PENDING 0x1
391 struct kvm_guest_msix_entry {
397 struct kvm_assigned_dev_kernel {
398 struct kvm_irq_ack_notifier ack_notifier;
399 struct work_struct interrupt_work;
400 struct list_head list;
404 unsigned int entries_nr;
406 bool host_irq_disabled;
407 struct msix_entry *host_msix_entries;
409 struct kvm_guest_msix_entry *guest_msix_entries;
410 unsigned long irq_requested_type;
415 spinlock_t assigned_dev_lock;
418 struct kvm_irq_mask_notifier {
419 void (*func)(struct kvm_irq_mask_notifier *kimn, bool masked);
421 struct hlist_node link;
424 void kvm_register_irq_mask_notifier(struct kvm *kvm, int irq,
425 struct kvm_irq_mask_notifier *kimn);
426 void kvm_unregister_irq_mask_notifier(struct kvm *kvm, int irq,
427 struct kvm_irq_mask_notifier *kimn);
428 void kvm_fire_mask_notifiers(struct kvm *kvm, int irq, bool mask);
430 #ifdef __KVM_HAVE_IOAPIC
431 void kvm_get_intr_delivery_bitmask(struct kvm_ioapic *ioapic,
432 union kvm_ioapic_redirect_entry *entry,
433 unsigned long *deliver_bitmask);
435 int kvm_set_irq(struct kvm *kvm, int irq_source_id, u32 irq, int level);
436 void kvm_notify_acked_irq(struct kvm *kvm, unsigned irqchip, unsigned pin);
437 void kvm_register_irq_ack_notifier(struct kvm *kvm,
438 struct kvm_irq_ack_notifier *kian);
439 void kvm_unregister_irq_ack_notifier(struct kvm *kvm,
440 struct kvm_irq_ack_notifier *kian);
441 int kvm_request_irq_source_id(struct kvm *kvm);
442 void kvm_free_irq_source_id(struct kvm *kvm, int irq_source_id);
444 /* For vcpu->arch.iommu_flags */
445 #define KVM_IOMMU_CACHE_COHERENCY 0x1
447 #ifdef CONFIG_IOMMU_API
448 int kvm_iommu_map_pages(struct kvm *kvm, struct kvm_memory_slot *slot);
449 int kvm_iommu_map_guest(struct kvm *kvm);
450 int kvm_iommu_unmap_guest(struct kvm *kvm);
451 int kvm_assign_device(struct kvm *kvm,
452 struct kvm_assigned_dev_kernel *assigned_dev);
453 int kvm_deassign_device(struct kvm *kvm,
454 struct kvm_assigned_dev_kernel *assigned_dev);
455 #else /* CONFIG_IOMMU_API */
456 static inline int kvm_iommu_map_pages(struct kvm *kvm,
458 unsigned long npages)
463 static inline int kvm_iommu_map_guest(struct kvm *kvm)
468 static inline int kvm_iommu_unmap_guest(struct kvm *kvm)
473 static inline int kvm_assign_device(struct kvm *kvm,
474 struct kvm_assigned_dev_kernel *assigned_dev)
479 static inline int kvm_deassign_device(struct kvm *kvm,
480 struct kvm_assigned_dev_kernel *assigned_dev)
484 #endif /* CONFIG_IOMMU_API */
486 static inline void kvm_guest_enter(void)
488 account_system_vtime(current);
489 current->flags |= PF_VCPU;
492 static inline void kvm_guest_exit(void)
494 account_system_vtime(current);
495 current->flags &= ~PF_VCPU;
498 static inline gpa_t gfn_to_gpa(gfn_t gfn)
500 return (gpa_t)gfn << PAGE_SHIFT;
503 static inline hpa_t pfn_to_hpa(pfn_t pfn)
505 return (hpa_t)pfn << PAGE_SHIFT;
508 static inline void kvm_migrate_timers(struct kvm_vcpu *vcpu)
510 set_bit(KVM_REQ_MIGRATE_TIMER, &vcpu->requests);
518 struct kvm_stats_debugfs_item {
521 enum kvm_stat_kind kind;
522 struct dentry *dentry;
524 extern struct kvm_stats_debugfs_item debugfs_entries[];
525 extern struct dentry *kvm_debugfs_dir;
527 #ifdef KVM_ARCH_WANT_MMU_NOTIFIER
528 static inline int mmu_notifier_retry(struct kvm_vcpu *vcpu, unsigned long mmu_seq)
530 if (unlikely(vcpu->kvm->mmu_notifier_count))
533 * Both reads happen under the mmu_lock and both values are
534 * modified under mmu_lock, so there's no need of smb_rmb()
535 * here in between, otherwise mmu_notifier_count should be
536 * read before mmu_notifier_seq, see
537 * mmu_notifier_invalidate_range_end write side.
539 if (vcpu->kvm->mmu_notifier_seq != mmu_seq)
545 #ifndef KVM_ARCH_HAS_UNALIAS_INSTANTIATION
546 #define unalias_gfn_instantiation unalias_gfn
549 #ifdef CONFIG_HAVE_KVM_IRQCHIP
551 #define KVM_MAX_IRQ_ROUTES 1024
553 int kvm_setup_default_irq_routing(struct kvm *kvm);
554 int kvm_set_irq_routing(struct kvm *kvm,
555 const struct kvm_irq_routing_entry *entries,
558 void kvm_free_irq_routing(struct kvm *kvm);
562 static inline void kvm_free_irq_routing(struct kvm *kvm) {}
566 #ifdef CONFIG_HAVE_KVM_EVENTFD
568 void kvm_eventfd_init(struct kvm *kvm);
569 int kvm_irqfd(struct kvm *kvm, int fd, int gsi, int flags);
570 void kvm_irqfd_release(struct kvm *kvm);
571 int kvm_ioeventfd(struct kvm *kvm, struct kvm_ioeventfd *args);
575 static inline void kvm_eventfd_init(struct kvm *kvm) {}
576 static inline int kvm_irqfd(struct kvm *kvm, int fd, int gsi, int flags)
581 static inline void kvm_irqfd_release(struct kvm *kvm) {}
582 static inline int kvm_ioeventfd(struct kvm *kvm, struct kvm_ioeventfd *args)
587 #endif /* CONFIG_HAVE_KVM_EVENTFD */
589 #ifdef CONFIG_KVM_APIC_ARCHITECTURE
590 static inline bool kvm_vcpu_is_bsp(struct kvm_vcpu *vcpu)
592 return vcpu->kvm->bsp_vcpu_id == vcpu->vcpu_id;
596 #ifdef __KVM_HAVE_DEVICE_ASSIGNMENT
598 long kvm_vm_ioctl_assigned_device(struct kvm *kvm, unsigned ioctl,
603 static inline long kvm_vm_ioctl_assigned_device(struct kvm *kvm, unsigned ioctl,