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];
60 void kvm_io_bus_init(struct kvm_io_bus *bus);
61 void kvm_io_bus_destroy(struct kvm_io_bus *bus);
62 int kvm_io_bus_write(struct kvm_io_bus *bus, gpa_t addr, int len,
64 int kvm_io_bus_read(struct kvm_io_bus *bus, gpa_t addr, int len,
66 int __kvm_io_bus_register_dev(struct kvm_io_bus *bus,
67 struct kvm_io_device *dev);
68 int kvm_io_bus_register_dev(struct kvm *kvm, struct kvm_io_bus *bus,
69 struct kvm_io_device *dev);
70 void __kvm_io_bus_unregister_dev(struct kvm_io_bus *bus,
71 struct kvm_io_device *dev);
72 void kvm_io_bus_unregister_dev(struct kvm *kvm, struct kvm_io_bus *bus,
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;
91 struct kvm_vcpu_stat stat;
93 #ifdef CONFIG_HAS_IOMEM
95 int mmio_read_completed;
98 unsigned char mmio_data[8];
102 struct kvm_vcpu_arch arch;
105 struct kvm_memory_slot {
107 unsigned long npages;
110 unsigned long *dirty_bitmap;
112 unsigned long rmap_pde;
114 } *lpage_info[KVM_NR_PAGE_SIZES - 1];
115 unsigned long userspace_addr;
119 struct kvm_kernel_irq_routing_entry {
122 int (*set)(struct kvm_kernel_irq_routing_entry *e,
123 struct kvm *kvm, int irq_source_id, int level);
131 struct hlist_node link;
134 #ifdef __KVM_HAVE_IOAPIC
136 struct kvm_irq_routing_table {
137 int chip[KVM_NR_IRQCHIPS][KVM_IOAPIC_NUM_PINS];
138 struct kvm_kernel_irq_routing_entry *rt_entries;
141 * Array indexed by gsi. Each entry contains list of irq chips
142 * the gsi is connected to.
144 struct hlist_head map[0];
149 struct kvm_irq_routing_table {};
153 struct kvm_memslots {
155 struct kvm_memory_slot memslots[KVM_MEMORY_SLOTS +
156 KVM_PRIVATE_MEM_SLOTS];
161 spinlock_t requests_lock;
162 struct rw_semaphore slots_lock;
163 struct mm_struct *mm; /* userspace tied to this vm */
164 struct kvm_memslots *memslots;
165 #ifdef CONFIG_KVM_APIC_ARCHITECTURE
167 struct kvm_vcpu *bsp_vcpu;
169 struct kvm_vcpu *vcpus[KVM_MAX_VCPUS];
170 atomic_t online_vcpus;
171 struct list_head vm_list;
173 struct kvm_io_bus mmio_bus;
174 struct kvm_io_bus pio_bus;
175 #ifdef CONFIG_HAVE_KVM_EVENTFD
178 struct list_head items;
180 struct list_head ioeventfds;
182 struct kvm_vm_stat stat;
183 struct kvm_arch arch;
184 atomic_t users_count;
185 #ifdef KVM_COALESCED_MMIO_PAGE_OFFSET
186 struct kvm_coalesced_mmio_dev *coalesced_mmio_dev;
187 struct kvm_coalesced_mmio_ring *coalesced_mmio_ring;
190 struct mutex irq_lock;
191 #ifdef CONFIG_HAVE_KVM_IRQCHIP
192 struct kvm_irq_routing_table *irq_routing;
193 struct hlist_head mask_notifier_list;
194 struct hlist_head irq_ack_notifier_list;
197 #ifdef KVM_ARCH_WANT_MMU_NOTIFIER
198 struct mmu_notifier mmu_notifier;
199 unsigned long mmu_notifier_seq;
200 long mmu_notifier_count;
204 /* The guest did something we don't support. */
205 #define pr_unimpl(vcpu, fmt, ...) \
207 if (printk_ratelimit()) \
208 printk(KERN_ERR "kvm: %i: cpu%i " fmt, \
209 current->tgid, (vcpu)->vcpu_id , ## __VA_ARGS__); \
212 #define kvm_printf(kvm, fmt ...) printk(KERN_DEBUG fmt)
213 #define vcpu_printf(vcpu, fmt...) kvm_printf(vcpu->kvm, fmt)
215 static inline struct kvm_vcpu *kvm_get_vcpu(struct kvm *kvm, int i)
218 return kvm->vcpus[i];
221 #define kvm_for_each_vcpu(idx, vcpup, kvm) \
222 for (idx = 0, vcpup = kvm_get_vcpu(kvm, idx); \
223 idx < atomic_read(&kvm->online_vcpus) && vcpup; \
224 vcpup = kvm_get_vcpu(kvm, ++idx))
226 int kvm_vcpu_init(struct kvm_vcpu *vcpu, struct kvm *kvm, unsigned id);
227 void kvm_vcpu_uninit(struct kvm_vcpu *vcpu);
229 void vcpu_load(struct kvm_vcpu *vcpu);
230 void vcpu_put(struct kvm_vcpu *vcpu);
232 int kvm_init(void *opaque, unsigned int vcpu_size,
233 struct module *module);
236 void kvm_get_kvm(struct kvm *kvm);
237 void kvm_put_kvm(struct kvm *kvm);
239 #define HPA_MSB ((sizeof(hpa_t) * 8) - 1)
240 #define HPA_ERR_MASK ((hpa_t)1 << HPA_MSB)
241 static inline int is_error_hpa(hpa_t hpa) { return hpa >> HPA_MSB; }
242 struct page *gva_to_page(struct kvm_vcpu *vcpu, gva_t gva);
244 extern struct page *bad_page;
245 extern pfn_t bad_pfn;
247 int is_error_page(struct page *page);
248 int is_error_pfn(pfn_t pfn);
249 int kvm_is_error_hva(unsigned long addr);
250 int kvm_set_memory_region(struct kvm *kvm,
251 struct kvm_userspace_memory_region *mem,
253 int __kvm_set_memory_region(struct kvm *kvm,
254 struct kvm_userspace_memory_region *mem,
256 int kvm_arch_prepare_memory_region(struct kvm *kvm,
257 struct kvm_memory_slot *memslot,
258 struct kvm_memory_slot old,
259 struct kvm_userspace_memory_region *mem,
261 void kvm_arch_commit_memory_region(struct kvm *kvm,
262 struct kvm_userspace_memory_region *mem,
263 struct kvm_memory_slot old,
265 void kvm_disable_largepages(void);
266 void kvm_arch_flush_shadow(struct kvm *kvm);
267 gfn_t unalias_gfn(struct kvm *kvm, gfn_t gfn);
268 struct page *gfn_to_page(struct kvm *kvm, gfn_t gfn);
269 unsigned long gfn_to_hva(struct kvm *kvm, gfn_t gfn);
270 void kvm_release_page_clean(struct page *page);
271 void kvm_release_page_dirty(struct page *page);
272 void kvm_set_page_dirty(struct page *page);
273 void kvm_set_page_accessed(struct page *page);
275 pfn_t gfn_to_pfn(struct kvm *kvm, gfn_t gfn);
276 pfn_t gfn_to_pfn_memslot(struct kvm *kvm,
277 struct kvm_memory_slot *slot, gfn_t gfn);
278 void kvm_release_pfn_dirty(pfn_t);
279 void kvm_release_pfn_clean(pfn_t pfn);
280 void kvm_set_pfn_dirty(pfn_t pfn);
281 void kvm_set_pfn_accessed(pfn_t pfn);
282 void kvm_get_pfn(pfn_t pfn);
284 int kvm_read_guest_page(struct kvm *kvm, gfn_t gfn, void *data, int offset,
286 int kvm_read_guest_atomic(struct kvm *kvm, gpa_t gpa, void *data,
288 int kvm_read_guest(struct kvm *kvm, gpa_t gpa, void *data, unsigned long len);
289 int kvm_write_guest_page(struct kvm *kvm, gfn_t gfn, const void *data,
290 int offset, int len);
291 int kvm_write_guest(struct kvm *kvm, gpa_t gpa, const void *data,
293 int kvm_clear_guest_page(struct kvm *kvm, gfn_t gfn, int offset, int len);
294 int kvm_clear_guest(struct kvm *kvm, gpa_t gpa, unsigned long len);
295 struct kvm_memory_slot *gfn_to_memslot(struct kvm *kvm, gfn_t gfn);
296 int kvm_is_visible_gfn(struct kvm *kvm, gfn_t gfn);
297 void mark_page_dirty(struct kvm *kvm, gfn_t gfn);
299 void kvm_vcpu_block(struct kvm_vcpu *vcpu);
300 void kvm_vcpu_on_spin(struct kvm_vcpu *vcpu);
301 void kvm_resched(struct kvm_vcpu *vcpu);
302 void kvm_load_guest_fpu(struct kvm_vcpu *vcpu);
303 void kvm_put_guest_fpu(struct kvm_vcpu *vcpu);
304 void kvm_flush_remote_tlbs(struct kvm *kvm);
305 void kvm_reload_remote_mmus(struct kvm *kvm);
307 long kvm_arch_dev_ioctl(struct file *filp,
308 unsigned int ioctl, unsigned long arg);
309 long kvm_arch_vcpu_ioctl(struct file *filp,
310 unsigned int ioctl, unsigned long arg);
312 int kvm_dev_ioctl_check_extension(long ext);
314 int kvm_get_dirty_log(struct kvm *kvm,
315 struct kvm_dirty_log *log, int *is_dirty);
316 int kvm_vm_ioctl_get_dirty_log(struct kvm *kvm,
317 struct kvm_dirty_log *log);
319 int kvm_vm_ioctl_set_memory_region(struct kvm *kvm,
321 kvm_userspace_memory_region *mem,
323 long kvm_arch_vm_ioctl(struct file *filp,
324 unsigned int ioctl, unsigned long arg);
326 int kvm_arch_vcpu_ioctl_get_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu);
327 int kvm_arch_vcpu_ioctl_set_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu);
329 int kvm_arch_vcpu_ioctl_translate(struct kvm_vcpu *vcpu,
330 struct kvm_translation *tr);
332 int kvm_arch_vcpu_ioctl_get_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs);
333 int kvm_arch_vcpu_ioctl_set_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs);
334 int kvm_arch_vcpu_ioctl_get_sregs(struct kvm_vcpu *vcpu,
335 struct kvm_sregs *sregs);
336 int kvm_arch_vcpu_ioctl_set_sregs(struct kvm_vcpu *vcpu,
337 struct kvm_sregs *sregs);
338 int kvm_arch_vcpu_ioctl_get_mpstate(struct kvm_vcpu *vcpu,
339 struct kvm_mp_state *mp_state);
340 int kvm_arch_vcpu_ioctl_set_mpstate(struct kvm_vcpu *vcpu,
341 struct kvm_mp_state *mp_state);
342 int kvm_arch_vcpu_ioctl_set_guest_debug(struct kvm_vcpu *vcpu,
343 struct kvm_guest_debug *dbg);
344 int kvm_arch_vcpu_ioctl_run(struct kvm_vcpu *vcpu, struct kvm_run *kvm_run);
346 int kvm_arch_init(void *opaque);
347 void kvm_arch_exit(void);
349 int kvm_arch_vcpu_init(struct kvm_vcpu *vcpu);
350 void kvm_arch_vcpu_uninit(struct kvm_vcpu *vcpu);
352 void kvm_arch_vcpu_free(struct kvm_vcpu *vcpu);
353 void kvm_arch_vcpu_load(struct kvm_vcpu *vcpu, int cpu);
354 void kvm_arch_vcpu_put(struct kvm_vcpu *vcpu);
355 struct kvm_vcpu *kvm_arch_vcpu_create(struct kvm *kvm, unsigned int id);
356 int kvm_arch_vcpu_setup(struct kvm_vcpu *vcpu);
357 void kvm_arch_vcpu_destroy(struct kvm_vcpu *vcpu);
359 int kvm_arch_vcpu_reset(struct kvm_vcpu *vcpu);
360 int kvm_arch_hardware_enable(void *garbage);
361 void kvm_arch_hardware_disable(void *garbage);
362 int kvm_arch_hardware_setup(void);
363 void kvm_arch_hardware_unsetup(void);
364 void kvm_arch_check_processor_compat(void *rtn);
365 int kvm_arch_vcpu_runnable(struct kvm_vcpu *vcpu);
367 void kvm_free_physmem(struct kvm *kvm);
369 struct kvm *kvm_arch_create_vm(void);
370 void kvm_arch_destroy_vm(struct kvm *kvm);
371 void kvm_free_all_assigned_devices(struct kvm *kvm);
372 void kvm_arch_sync_events(struct kvm *kvm);
374 int kvm_cpu_has_pending_timer(struct kvm_vcpu *vcpu);
375 void kvm_vcpu_kick(struct kvm_vcpu *vcpu);
377 int kvm_is_mmio_pfn(pfn_t pfn);
379 struct kvm_irq_ack_notifier {
380 struct hlist_node link;
382 void (*irq_acked)(struct kvm_irq_ack_notifier *kian);
385 #define KVM_ASSIGNED_MSIX_PENDING 0x1
386 struct kvm_guest_msix_entry {
392 struct kvm_assigned_dev_kernel {
393 struct kvm_irq_ack_notifier ack_notifier;
394 struct work_struct interrupt_work;
395 struct list_head list;
399 unsigned int entries_nr;
401 bool host_irq_disabled;
402 struct msix_entry *host_msix_entries;
404 struct kvm_guest_msix_entry *guest_msix_entries;
405 unsigned long irq_requested_type;
410 spinlock_t assigned_dev_lock;
413 struct kvm_irq_mask_notifier {
414 void (*func)(struct kvm_irq_mask_notifier *kimn, bool masked);
416 struct hlist_node link;
419 void kvm_register_irq_mask_notifier(struct kvm *kvm, int irq,
420 struct kvm_irq_mask_notifier *kimn);
421 void kvm_unregister_irq_mask_notifier(struct kvm *kvm, int irq,
422 struct kvm_irq_mask_notifier *kimn);
423 void kvm_fire_mask_notifiers(struct kvm *kvm, int irq, bool mask);
425 #ifdef __KVM_HAVE_IOAPIC
426 void kvm_get_intr_delivery_bitmask(struct kvm_ioapic *ioapic,
427 union kvm_ioapic_redirect_entry *entry,
428 unsigned long *deliver_bitmask);
430 int kvm_set_irq(struct kvm *kvm, int irq_source_id, u32 irq, int level);
431 void kvm_notify_acked_irq(struct kvm *kvm, unsigned irqchip, unsigned pin);
432 void kvm_register_irq_ack_notifier(struct kvm *kvm,
433 struct kvm_irq_ack_notifier *kian);
434 void kvm_unregister_irq_ack_notifier(struct kvm *kvm,
435 struct kvm_irq_ack_notifier *kian);
436 int kvm_request_irq_source_id(struct kvm *kvm);
437 void kvm_free_irq_source_id(struct kvm *kvm, int irq_source_id);
439 /* For vcpu->arch.iommu_flags */
440 #define KVM_IOMMU_CACHE_COHERENCY 0x1
442 #ifdef CONFIG_IOMMU_API
443 int kvm_iommu_map_pages(struct kvm *kvm, struct kvm_memory_slot *slot);
444 int kvm_iommu_map_guest(struct kvm *kvm);
445 int kvm_iommu_unmap_guest(struct kvm *kvm);
446 int kvm_assign_device(struct kvm *kvm,
447 struct kvm_assigned_dev_kernel *assigned_dev);
448 int kvm_deassign_device(struct kvm *kvm,
449 struct kvm_assigned_dev_kernel *assigned_dev);
450 #else /* CONFIG_IOMMU_API */
451 static inline int kvm_iommu_map_pages(struct kvm *kvm,
453 unsigned long npages)
458 static inline int kvm_iommu_map_guest(struct kvm *kvm)
463 static inline int kvm_iommu_unmap_guest(struct kvm *kvm)
468 static inline int kvm_assign_device(struct kvm *kvm,
469 struct kvm_assigned_dev_kernel *assigned_dev)
474 static inline int kvm_deassign_device(struct kvm *kvm,
475 struct kvm_assigned_dev_kernel *assigned_dev)
479 #endif /* CONFIG_IOMMU_API */
481 static inline void kvm_guest_enter(void)
483 account_system_vtime(current);
484 current->flags |= PF_VCPU;
487 static inline void kvm_guest_exit(void)
489 account_system_vtime(current);
490 current->flags &= ~PF_VCPU;
493 static inline int memslot_id(struct kvm *kvm, struct kvm_memory_slot *slot)
495 return slot - kvm->memslots->memslots;
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 #ifdef CONFIG_HAVE_KVM_IRQCHIP
547 #define KVM_MAX_IRQ_ROUTES 1024
549 int kvm_setup_default_irq_routing(struct kvm *kvm);
550 int kvm_set_irq_routing(struct kvm *kvm,
551 const struct kvm_irq_routing_entry *entries,
554 void kvm_free_irq_routing(struct kvm *kvm);
558 static inline void kvm_free_irq_routing(struct kvm *kvm) {}
562 #ifdef CONFIG_HAVE_KVM_EVENTFD
564 void kvm_eventfd_init(struct kvm *kvm);
565 int kvm_irqfd(struct kvm *kvm, int fd, int gsi, int flags);
566 void kvm_irqfd_release(struct kvm *kvm);
567 int kvm_ioeventfd(struct kvm *kvm, struct kvm_ioeventfd *args);
571 static inline void kvm_eventfd_init(struct kvm *kvm) {}
572 static inline int kvm_irqfd(struct kvm *kvm, int fd, int gsi, int flags)
577 static inline void kvm_irqfd_release(struct kvm *kvm) {}
578 static inline int kvm_ioeventfd(struct kvm *kvm, struct kvm_ioeventfd *args)
583 #endif /* CONFIG_HAVE_KVM_EVENTFD */
585 #ifdef CONFIG_KVM_APIC_ARCHITECTURE
586 static inline bool kvm_vcpu_is_bsp(struct kvm_vcpu *vcpu)
588 return vcpu->kvm->bsp_vcpu_id == vcpu->vcpu_id;
592 #ifdef __KVM_HAVE_DEVICE_ASSIGNMENT
594 long kvm_vm_ioctl_assigned_device(struct kvm *kvm, unsigned ioctl,
599 static inline long kvm_vm_ioctl_assigned_device(struct kvm *kvm, unsigned ioctl,