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 200
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;
109 * Some of the bitops functions do not support too long bitmaps.
110 * This number must be determined not to exceed such limits.
112 #define KVM_MEM_MAX_NR_PAGES ((1UL << 31) - 1)
114 struct kvm_memory_slot {
116 unsigned long npages;
119 unsigned long *dirty_bitmap;
121 unsigned long rmap_pde;
123 } *lpage_info[KVM_NR_PAGE_SIZES - 1];
124 unsigned long userspace_addr;
128 static inline unsigned long kvm_dirty_bitmap_bytes(struct kvm_memory_slot *memslot)
130 return ALIGN(memslot->npages, BITS_PER_LONG) / 8;
133 struct kvm_kernel_irq_routing_entry {
136 int (*set)(struct kvm_kernel_irq_routing_entry *e,
137 struct kvm *kvm, int irq_source_id, int level);
145 struct hlist_node link;
148 #ifdef __KVM_HAVE_IOAPIC
150 struct kvm_irq_routing_table {
151 int chip[KVM_NR_IRQCHIPS][KVM_IOAPIC_NUM_PINS];
152 struct kvm_kernel_irq_routing_entry *rt_entries;
155 * Array indexed by gsi. Each entry contains list of irq chips
156 * the gsi is connected to.
158 struct hlist_head map[0];
163 struct kvm_irq_routing_table {};
167 struct kvm_memslots {
169 struct kvm_memory_slot memslots[KVM_MEMORY_SLOTS +
170 KVM_PRIVATE_MEM_SLOTS];
175 raw_spinlock_t requests_lock;
176 struct mutex slots_lock;
177 struct mm_struct *mm; /* userspace tied to this vm */
178 struct kvm_memslots *memslots;
179 struct srcu_struct srcu;
180 #ifdef CONFIG_KVM_APIC_ARCHITECTURE
182 struct kvm_vcpu *bsp_vcpu;
184 struct kvm_vcpu *vcpus[KVM_MAX_VCPUS];
185 atomic_t online_vcpus;
186 struct list_head vm_list;
188 struct kvm_io_bus *buses[KVM_NR_BUSES];
189 #ifdef CONFIG_HAVE_KVM_EVENTFD
192 struct list_head items;
194 struct list_head ioeventfds;
196 struct kvm_vm_stat stat;
197 struct kvm_arch arch;
198 atomic_t users_count;
199 #ifdef KVM_COALESCED_MMIO_PAGE_OFFSET
200 struct kvm_coalesced_mmio_dev *coalesced_mmio_dev;
201 struct kvm_coalesced_mmio_ring *coalesced_mmio_ring;
204 struct mutex irq_lock;
205 #ifdef CONFIG_HAVE_KVM_IRQCHIP
206 struct kvm_irq_routing_table *irq_routing;
207 struct hlist_head mask_notifier_list;
208 struct hlist_head irq_ack_notifier_list;
211 #ifdef KVM_ARCH_WANT_MMU_NOTIFIER
212 struct mmu_notifier mmu_notifier;
213 unsigned long mmu_notifier_seq;
214 long mmu_notifier_count;
218 /* The guest did something we don't support. */
219 #define pr_unimpl(vcpu, fmt, ...) \
221 if (printk_ratelimit()) \
222 printk(KERN_ERR "kvm: %i: cpu%i " fmt, \
223 current->tgid, (vcpu)->vcpu_id , ## __VA_ARGS__); \
226 #define kvm_printf(kvm, fmt ...) printk(KERN_DEBUG fmt)
227 #define vcpu_printf(vcpu, fmt...) kvm_printf(vcpu->kvm, fmt)
229 static inline struct kvm_vcpu *kvm_get_vcpu(struct kvm *kvm, int i)
232 return kvm->vcpus[i];
235 #define kvm_for_each_vcpu(idx, vcpup, kvm) \
236 for (idx = 0, vcpup = kvm_get_vcpu(kvm, idx); \
237 idx < atomic_read(&kvm->online_vcpus) && vcpup; \
238 vcpup = kvm_get_vcpu(kvm, ++idx))
240 int kvm_vcpu_init(struct kvm_vcpu *vcpu, struct kvm *kvm, unsigned id);
241 void kvm_vcpu_uninit(struct kvm_vcpu *vcpu);
243 void vcpu_load(struct kvm_vcpu *vcpu);
244 void vcpu_put(struct kvm_vcpu *vcpu);
246 int kvm_init(void *opaque, unsigned int vcpu_size,
247 struct module *module);
250 void kvm_get_kvm(struct kvm *kvm);
251 void kvm_put_kvm(struct kvm *kvm);
253 static inline struct kvm_memslots *kvm_memslots(struct kvm *kvm)
255 return rcu_dereference_check(kvm->memslots,
256 srcu_read_lock_held(&kvm->srcu)
257 || lockdep_is_held(&kvm->slots_lock));
260 #define HPA_MSB ((sizeof(hpa_t) * 8) - 1)
261 #define HPA_ERR_MASK ((hpa_t)1 << HPA_MSB)
262 static inline int is_error_hpa(hpa_t hpa) { return hpa >> HPA_MSB; }
263 struct page *gva_to_page(struct kvm_vcpu *vcpu, gva_t gva);
265 extern struct page *bad_page;
266 extern pfn_t bad_pfn;
268 int is_error_page(struct page *page);
269 int is_error_pfn(pfn_t pfn);
270 int kvm_is_error_hva(unsigned long addr);
271 int kvm_set_memory_region(struct kvm *kvm,
272 struct kvm_userspace_memory_region *mem,
274 int __kvm_set_memory_region(struct kvm *kvm,
275 struct kvm_userspace_memory_region *mem,
277 int kvm_arch_prepare_memory_region(struct kvm *kvm,
278 struct kvm_memory_slot *memslot,
279 struct kvm_memory_slot old,
280 struct kvm_userspace_memory_region *mem,
282 void kvm_arch_commit_memory_region(struct kvm *kvm,
283 struct kvm_userspace_memory_region *mem,
284 struct kvm_memory_slot old,
286 void kvm_disable_largepages(void);
287 void kvm_arch_flush_shadow(struct kvm *kvm);
288 gfn_t unalias_gfn(struct kvm *kvm, gfn_t gfn);
289 gfn_t unalias_gfn_instantiation(struct kvm *kvm, gfn_t gfn);
291 struct page *gfn_to_page(struct kvm *kvm, gfn_t gfn);
292 unsigned long gfn_to_hva(struct kvm *kvm, gfn_t gfn);
293 void kvm_release_page_clean(struct page *page);
294 void kvm_release_page_dirty(struct page *page);
295 void kvm_set_page_dirty(struct page *page);
296 void kvm_set_page_accessed(struct page *page);
298 pfn_t gfn_to_pfn(struct kvm *kvm, gfn_t gfn);
299 pfn_t gfn_to_pfn_memslot(struct kvm *kvm,
300 struct kvm_memory_slot *slot, gfn_t gfn);
301 int memslot_id(struct kvm *kvm, gfn_t gfn);
302 void kvm_release_pfn_dirty(pfn_t);
303 void kvm_release_pfn_clean(pfn_t pfn);
304 void kvm_set_pfn_dirty(pfn_t pfn);
305 void kvm_set_pfn_accessed(pfn_t pfn);
306 void kvm_get_pfn(pfn_t pfn);
308 int kvm_read_guest_page(struct kvm *kvm, gfn_t gfn, void *data, int offset,
310 int kvm_read_guest_atomic(struct kvm *kvm, gpa_t gpa, void *data,
312 int kvm_read_guest(struct kvm *kvm, gpa_t gpa, void *data, unsigned long len);
313 int kvm_write_guest_page(struct kvm *kvm, gfn_t gfn, const void *data,
314 int offset, int len);
315 int kvm_write_guest(struct kvm *kvm, gpa_t gpa, const void *data,
317 int kvm_clear_guest_page(struct kvm *kvm, gfn_t gfn, int offset, int len);
318 int kvm_clear_guest(struct kvm *kvm, gpa_t gpa, unsigned long len);
319 struct kvm_memory_slot *gfn_to_memslot(struct kvm *kvm, gfn_t gfn);
320 int kvm_is_visible_gfn(struct kvm *kvm, gfn_t gfn);
321 unsigned long kvm_host_page_size(struct kvm *kvm, gfn_t gfn);
322 void mark_page_dirty(struct kvm *kvm, gfn_t gfn);
324 void kvm_vcpu_block(struct kvm_vcpu *vcpu);
325 void kvm_vcpu_on_spin(struct kvm_vcpu *vcpu);
326 void kvm_resched(struct kvm_vcpu *vcpu);
327 void kvm_load_guest_fpu(struct kvm_vcpu *vcpu);
328 void kvm_put_guest_fpu(struct kvm_vcpu *vcpu);
329 void kvm_flush_remote_tlbs(struct kvm *kvm);
330 void kvm_reload_remote_mmus(struct kvm *kvm);
332 long kvm_arch_dev_ioctl(struct file *filp,
333 unsigned int ioctl, unsigned long arg);
334 long kvm_arch_vcpu_ioctl(struct file *filp,
335 unsigned int ioctl, unsigned long arg);
337 int kvm_dev_ioctl_check_extension(long ext);
339 int kvm_get_dirty_log(struct kvm *kvm,
340 struct kvm_dirty_log *log, int *is_dirty);
341 int kvm_vm_ioctl_get_dirty_log(struct kvm *kvm,
342 struct kvm_dirty_log *log);
344 int kvm_vm_ioctl_set_memory_region(struct kvm *kvm,
346 kvm_userspace_memory_region *mem,
348 long kvm_arch_vm_ioctl(struct file *filp,
349 unsigned int ioctl, unsigned long arg);
351 int kvm_arch_vcpu_ioctl_get_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu);
352 int kvm_arch_vcpu_ioctl_set_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu);
354 int kvm_arch_vcpu_ioctl_translate(struct kvm_vcpu *vcpu,
355 struct kvm_translation *tr);
357 int kvm_arch_vcpu_ioctl_get_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs);
358 int kvm_arch_vcpu_ioctl_set_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs);
359 int kvm_arch_vcpu_ioctl_get_sregs(struct kvm_vcpu *vcpu,
360 struct kvm_sregs *sregs);
361 int kvm_arch_vcpu_ioctl_set_sregs(struct kvm_vcpu *vcpu,
362 struct kvm_sregs *sregs);
363 int kvm_arch_vcpu_ioctl_get_mpstate(struct kvm_vcpu *vcpu,
364 struct kvm_mp_state *mp_state);
365 int kvm_arch_vcpu_ioctl_set_mpstate(struct kvm_vcpu *vcpu,
366 struct kvm_mp_state *mp_state);
367 int kvm_arch_vcpu_ioctl_set_guest_debug(struct kvm_vcpu *vcpu,
368 struct kvm_guest_debug *dbg);
369 int kvm_arch_vcpu_ioctl_run(struct kvm_vcpu *vcpu, struct kvm_run *kvm_run);
371 int kvm_arch_init(void *opaque);
372 void kvm_arch_exit(void);
374 int kvm_arch_vcpu_init(struct kvm_vcpu *vcpu);
375 void kvm_arch_vcpu_uninit(struct kvm_vcpu *vcpu);
377 void kvm_arch_vcpu_free(struct kvm_vcpu *vcpu);
378 void kvm_arch_vcpu_load(struct kvm_vcpu *vcpu, int cpu);
379 void kvm_arch_vcpu_put(struct kvm_vcpu *vcpu);
380 struct kvm_vcpu *kvm_arch_vcpu_create(struct kvm *kvm, unsigned int id);
381 int kvm_arch_vcpu_setup(struct kvm_vcpu *vcpu);
382 void kvm_arch_vcpu_destroy(struct kvm_vcpu *vcpu);
384 int kvm_arch_vcpu_reset(struct kvm_vcpu *vcpu);
385 int kvm_arch_hardware_enable(void *garbage);
386 void kvm_arch_hardware_disable(void *garbage);
387 int kvm_arch_hardware_setup(void);
388 void kvm_arch_hardware_unsetup(void);
389 void kvm_arch_check_processor_compat(void *rtn);
390 int kvm_arch_vcpu_runnable(struct kvm_vcpu *vcpu);
392 void kvm_free_physmem(struct kvm *kvm);
394 struct kvm *kvm_arch_create_vm(void);
395 void kvm_arch_destroy_vm(struct kvm *kvm);
396 void kvm_free_all_assigned_devices(struct kvm *kvm);
397 void kvm_arch_sync_events(struct kvm *kvm);
399 int kvm_cpu_has_pending_timer(struct kvm_vcpu *vcpu);
400 void kvm_vcpu_kick(struct kvm_vcpu *vcpu);
402 int kvm_is_mmio_pfn(pfn_t pfn);
404 struct kvm_irq_ack_notifier {
405 struct hlist_node link;
407 void (*irq_acked)(struct kvm_irq_ack_notifier *kian);
410 #define KVM_ASSIGNED_MSIX_PENDING 0x1
411 struct kvm_guest_msix_entry {
417 struct kvm_assigned_dev_kernel {
418 struct kvm_irq_ack_notifier ack_notifier;
419 struct work_struct interrupt_work;
420 struct list_head list;
425 unsigned int entries_nr;
427 bool host_irq_disabled;
428 struct msix_entry *host_msix_entries;
430 struct kvm_guest_msix_entry *guest_msix_entries;
431 unsigned long irq_requested_type;
436 spinlock_t assigned_dev_lock;
439 struct kvm_irq_mask_notifier {
440 void (*func)(struct kvm_irq_mask_notifier *kimn, bool masked);
442 struct hlist_node link;
445 void kvm_register_irq_mask_notifier(struct kvm *kvm, int irq,
446 struct kvm_irq_mask_notifier *kimn);
447 void kvm_unregister_irq_mask_notifier(struct kvm *kvm, int irq,
448 struct kvm_irq_mask_notifier *kimn);
449 void kvm_fire_mask_notifiers(struct kvm *kvm, int irq, bool mask);
451 #ifdef __KVM_HAVE_IOAPIC
452 void kvm_get_intr_delivery_bitmask(struct kvm_ioapic *ioapic,
453 union kvm_ioapic_redirect_entry *entry,
454 unsigned long *deliver_bitmask);
456 int kvm_set_irq(struct kvm *kvm, int irq_source_id, u32 irq, int level);
457 void kvm_notify_acked_irq(struct kvm *kvm, unsigned irqchip, unsigned pin);
458 void kvm_register_irq_ack_notifier(struct kvm *kvm,
459 struct kvm_irq_ack_notifier *kian);
460 void kvm_unregister_irq_ack_notifier(struct kvm *kvm,
461 struct kvm_irq_ack_notifier *kian);
462 int kvm_request_irq_source_id(struct kvm *kvm);
463 void kvm_free_irq_source_id(struct kvm *kvm, int irq_source_id);
465 /* For vcpu->arch.iommu_flags */
466 #define KVM_IOMMU_CACHE_COHERENCY 0x1
468 #ifdef CONFIG_IOMMU_API
469 int kvm_iommu_map_pages(struct kvm *kvm, struct kvm_memory_slot *slot);
470 int kvm_iommu_map_guest(struct kvm *kvm);
471 int kvm_iommu_unmap_guest(struct kvm *kvm);
472 int kvm_assign_device(struct kvm *kvm,
473 struct kvm_assigned_dev_kernel *assigned_dev);
474 int kvm_deassign_device(struct kvm *kvm,
475 struct kvm_assigned_dev_kernel *assigned_dev);
476 #else /* CONFIG_IOMMU_API */
477 static inline int kvm_iommu_map_pages(struct kvm *kvm,
479 unsigned long npages)
484 static inline int kvm_iommu_map_guest(struct kvm *kvm)
489 static inline int kvm_iommu_unmap_guest(struct kvm *kvm)
494 static inline int kvm_assign_device(struct kvm *kvm,
495 struct kvm_assigned_dev_kernel *assigned_dev)
500 static inline int kvm_deassign_device(struct kvm *kvm,
501 struct kvm_assigned_dev_kernel *assigned_dev)
505 #endif /* CONFIG_IOMMU_API */
507 static inline void kvm_guest_enter(void)
509 account_system_vtime(current);
510 current->flags |= PF_VCPU;
513 static inline void kvm_guest_exit(void)
515 account_system_vtime(current);
516 current->flags &= ~PF_VCPU;
519 static inline gpa_t gfn_to_gpa(gfn_t gfn)
521 return (gpa_t)gfn << PAGE_SHIFT;
524 static inline hpa_t pfn_to_hpa(pfn_t pfn)
526 return (hpa_t)pfn << PAGE_SHIFT;
529 static inline void kvm_migrate_timers(struct kvm_vcpu *vcpu)
531 set_bit(KVM_REQ_MIGRATE_TIMER, &vcpu->requests);
539 struct kvm_stats_debugfs_item {
542 enum kvm_stat_kind kind;
543 struct dentry *dentry;
545 extern struct kvm_stats_debugfs_item debugfs_entries[];
546 extern struct dentry *kvm_debugfs_dir;
548 #ifdef KVM_ARCH_WANT_MMU_NOTIFIER
549 static inline int mmu_notifier_retry(struct kvm_vcpu *vcpu, unsigned long mmu_seq)
551 if (unlikely(vcpu->kvm->mmu_notifier_count))
554 * Both reads happen under the mmu_lock and both values are
555 * modified under mmu_lock, so there's no need of smb_rmb()
556 * here in between, otherwise mmu_notifier_count should be
557 * read before mmu_notifier_seq, see
558 * mmu_notifier_invalidate_range_end write side.
560 if (vcpu->kvm->mmu_notifier_seq != mmu_seq)
566 #ifndef KVM_ARCH_HAS_UNALIAS_INSTANTIATION
567 #define unalias_gfn_instantiation unalias_gfn
570 #ifdef CONFIG_HAVE_KVM_IRQCHIP
572 #define KVM_MAX_IRQ_ROUTES 1024
574 int kvm_setup_default_irq_routing(struct kvm *kvm);
575 int kvm_set_irq_routing(struct kvm *kvm,
576 const struct kvm_irq_routing_entry *entries,
579 void kvm_free_irq_routing(struct kvm *kvm);
583 static inline void kvm_free_irq_routing(struct kvm *kvm) {}
587 #ifdef CONFIG_HAVE_KVM_EVENTFD
589 void kvm_eventfd_init(struct kvm *kvm);
590 int kvm_irqfd(struct kvm *kvm, int fd, int gsi, int flags);
591 void kvm_irqfd_release(struct kvm *kvm);
592 int kvm_ioeventfd(struct kvm *kvm, struct kvm_ioeventfd *args);
596 static inline void kvm_eventfd_init(struct kvm *kvm) {}
597 static inline int kvm_irqfd(struct kvm *kvm, int fd, int gsi, int flags)
602 static inline void kvm_irqfd_release(struct kvm *kvm) {}
603 static inline int kvm_ioeventfd(struct kvm *kvm, struct kvm_ioeventfd *args)
608 #endif /* CONFIG_HAVE_KVM_EVENTFD */
610 #ifdef CONFIG_KVM_APIC_ARCHITECTURE
611 static inline bool kvm_vcpu_is_bsp(struct kvm_vcpu *vcpu)
613 return vcpu->kvm->bsp_vcpu_id == vcpu->vcpu_id;
617 #ifdef __KVM_HAVE_DEVICE_ASSIGNMENT
619 long kvm_vm_ioctl_assigned_device(struct kvm *kvm, unsigned ioctl,
624 static inline long kvm_vm_ioctl_assigned_device(struct kvm *kvm, unsigned ioctl,