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_CLOCK_UPDATE 8
40 #define KVM_REQ_KICK 9
41 #define KVM_REQ_DEACTIVATE_FPU 10
42 #define KVM_REQ_EVENT 11
43 #define KVM_REQ_APF_HALT 12
45 #define KVM_USERSPACE_IRQ_SOURCE_ID 0
49 extern struct kmem_cache *kvm_vcpu_cache;
52 * It would be nice to use something smarter than a linear search, TBD...
53 * Thankfully we dont expect many devices to register (famous last words :),
54 * so until then it will suffice. At least its abstracted so we can change
59 #define NR_IOBUS_DEVS 200
60 struct kvm_io_device *devs[NR_IOBUS_DEVS];
69 int kvm_io_bus_write(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr,
70 int len, const void *val);
71 int kvm_io_bus_read(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr, int len,
73 int kvm_io_bus_register_dev(struct kvm *kvm, enum kvm_bus bus_idx,
74 struct kvm_io_device *dev);
75 int kvm_io_bus_unregister_dev(struct kvm *kvm, enum kvm_bus bus_idx,
76 struct kvm_io_device *dev);
78 #ifdef CONFIG_KVM_ASYNC_PF
80 struct work_struct work;
81 struct list_head link;
82 struct list_head queue;
83 struct kvm_vcpu *vcpu;
87 struct kvm_arch_async_pf arch;
92 void kvm_clear_async_pf_completion_queue(struct kvm_vcpu *vcpu);
93 void kvm_check_async_pf_completion(struct kvm_vcpu *vcpu);
94 int kvm_setup_async_pf(struct kvm_vcpu *vcpu, gva_t gva, gfn_t gfn,
95 struct kvm_arch_async_pf *arch);
100 #ifdef CONFIG_PREEMPT_NOTIFIERS
101 struct preempt_notifier preempt_notifier;
108 unsigned long requests;
109 unsigned long guest_debug;
113 int guest_fpu_loaded, guest_xcr0_loaded;
114 wait_queue_head_t wq;
117 struct kvm_vcpu_stat stat;
119 #ifdef CONFIG_HAS_IOMEM
121 int mmio_read_completed;
124 unsigned char mmio_data[8];
125 gpa_t mmio_phys_addr;
128 #ifdef CONFIG_KVM_ASYNC_PF
131 struct list_head queue;
132 struct list_head done;
137 struct kvm_vcpu_arch arch;
141 * Some of the bitops functions do not support too long bitmaps.
142 * This number must be determined not to exceed such limits.
144 #define KVM_MEM_MAX_NR_PAGES ((1UL << 31) - 1)
146 struct kvm_memory_slot {
148 unsigned long npages;
151 unsigned long *dirty_bitmap;
153 unsigned long rmap_pde;
155 } *lpage_info[KVM_NR_PAGE_SIZES - 1];
156 unsigned long userspace_addr;
161 static inline unsigned long kvm_dirty_bitmap_bytes(struct kvm_memory_slot *memslot)
163 return ALIGN(memslot->npages, BITS_PER_LONG) / 8;
166 struct kvm_kernel_irq_routing_entry {
169 int (*set)(struct kvm_kernel_irq_routing_entry *e,
170 struct kvm *kvm, int irq_source_id, int level);
178 struct hlist_node link;
181 #ifdef __KVM_HAVE_IOAPIC
183 struct kvm_irq_routing_table {
184 int chip[KVM_NR_IRQCHIPS][KVM_IOAPIC_NUM_PINS];
185 struct kvm_kernel_irq_routing_entry *rt_entries;
188 * Array indexed by gsi. Each entry contains list of irq chips
189 * the gsi is connected to.
191 struct hlist_head map[0];
196 struct kvm_irq_routing_table {};
200 struct kvm_memslots {
202 struct kvm_memory_slot memslots[KVM_MEMORY_SLOTS +
203 KVM_PRIVATE_MEM_SLOTS];
208 raw_spinlock_t requests_lock;
209 struct mutex slots_lock;
210 struct mm_struct *mm; /* userspace tied to this vm */
211 struct kvm_memslots *memslots;
212 struct srcu_struct srcu;
213 #ifdef CONFIG_KVM_APIC_ARCHITECTURE
215 struct kvm_vcpu *bsp_vcpu;
217 struct kvm_vcpu *vcpus[KVM_MAX_VCPUS];
218 atomic_t online_vcpus;
219 struct list_head vm_list;
221 struct kvm_io_bus *buses[KVM_NR_BUSES];
222 #ifdef CONFIG_HAVE_KVM_EVENTFD
225 struct list_head items;
227 struct list_head ioeventfds;
229 struct kvm_vm_stat stat;
230 struct kvm_arch arch;
231 atomic_t users_count;
232 #ifdef KVM_COALESCED_MMIO_PAGE_OFFSET
233 struct kvm_coalesced_mmio_dev *coalesced_mmio_dev;
234 struct kvm_coalesced_mmio_ring *coalesced_mmio_ring;
237 struct mutex irq_lock;
238 #ifdef CONFIG_HAVE_KVM_IRQCHIP
239 struct kvm_irq_routing_table __rcu *irq_routing;
240 struct hlist_head mask_notifier_list;
241 struct hlist_head irq_ack_notifier_list;
244 #ifdef KVM_ARCH_WANT_MMU_NOTIFIER
245 struct mmu_notifier mmu_notifier;
246 unsigned long mmu_notifier_seq;
247 long mmu_notifier_count;
251 /* The guest did something we don't support. */
252 #define pr_unimpl(vcpu, fmt, ...) \
254 if (printk_ratelimit()) \
255 printk(KERN_ERR "kvm: %i: cpu%i " fmt, \
256 current->tgid, (vcpu)->vcpu_id , ## __VA_ARGS__); \
259 #define kvm_printf(kvm, fmt ...) printk(KERN_DEBUG fmt)
260 #define vcpu_printf(vcpu, fmt...) kvm_printf(vcpu->kvm, fmt)
262 static inline struct kvm_vcpu *kvm_get_vcpu(struct kvm *kvm, int i)
265 return kvm->vcpus[i];
268 #define kvm_for_each_vcpu(idx, vcpup, kvm) \
269 for (idx = 0, vcpup = kvm_get_vcpu(kvm, idx); \
270 idx < atomic_read(&kvm->online_vcpus) && vcpup; \
271 vcpup = kvm_get_vcpu(kvm, ++idx))
273 int kvm_vcpu_init(struct kvm_vcpu *vcpu, struct kvm *kvm, unsigned id);
274 void kvm_vcpu_uninit(struct kvm_vcpu *vcpu);
276 void vcpu_load(struct kvm_vcpu *vcpu);
277 void vcpu_put(struct kvm_vcpu *vcpu);
279 int kvm_init(void *opaque, unsigned vcpu_size, unsigned vcpu_align,
280 struct module *module);
283 void kvm_get_kvm(struct kvm *kvm);
284 void kvm_put_kvm(struct kvm *kvm);
286 static inline struct kvm_memslots *kvm_memslots(struct kvm *kvm)
288 return rcu_dereference_check(kvm->memslots,
289 srcu_read_lock_held(&kvm->srcu)
290 || lockdep_is_held(&kvm->slots_lock));
293 #define HPA_MSB ((sizeof(hpa_t) * 8) - 1)
294 #define HPA_ERR_MASK ((hpa_t)1 << HPA_MSB)
295 static inline int is_error_hpa(hpa_t hpa) { return hpa >> HPA_MSB; }
297 extern struct page *bad_page;
298 extern pfn_t bad_pfn;
300 int is_error_page(struct page *page);
301 int is_error_pfn(pfn_t pfn);
302 int is_hwpoison_pfn(pfn_t pfn);
303 int is_fault_pfn(pfn_t pfn);
304 int kvm_is_error_hva(unsigned long addr);
305 int kvm_set_memory_region(struct kvm *kvm,
306 struct kvm_userspace_memory_region *mem,
308 int __kvm_set_memory_region(struct kvm *kvm,
309 struct kvm_userspace_memory_region *mem,
311 int kvm_arch_prepare_memory_region(struct kvm *kvm,
312 struct kvm_memory_slot *memslot,
313 struct kvm_memory_slot old,
314 struct kvm_userspace_memory_region *mem,
316 void kvm_arch_commit_memory_region(struct kvm *kvm,
317 struct kvm_userspace_memory_region *mem,
318 struct kvm_memory_slot old,
320 void kvm_disable_largepages(void);
321 void kvm_arch_flush_shadow(struct kvm *kvm);
323 int gfn_to_page_many_atomic(struct kvm *kvm, gfn_t gfn, struct page **pages,
326 struct page *gfn_to_page(struct kvm *kvm, gfn_t gfn);
327 unsigned long gfn_to_hva(struct kvm *kvm, gfn_t gfn);
328 void kvm_release_page_clean(struct page *page);
329 void kvm_release_page_dirty(struct page *page);
330 void kvm_set_page_dirty(struct page *page);
331 void kvm_set_page_accessed(struct page *page);
333 pfn_t hva_to_pfn_atomic(struct kvm *kvm, unsigned long addr);
334 pfn_t gfn_to_pfn_atomic(struct kvm *kvm, gfn_t gfn);
335 pfn_t gfn_to_pfn_async(struct kvm *kvm, gfn_t gfn, bool *async);
336 pfn_t gfn_to_pfn(struct kvm *kvm, gfn_t gfn);
337 pfn_t gfn_to_pfn_memslot(struct kvm *kvm,
338 struct kvm_memory_slot *slot, gfn_t gfn);
339 int memslot_id(struct kvm *kvm, gfn_t gfn);
340 void kvm_release_pfn_dirty(pfn_t);
341 void kvm_release_pfn_clean(pfn_t pfn);
342 void kvm_set_pfn_dirty(pfn_t pfn);
343 void kvm_set_pfn_accessed(pfn_t pfn);
344 void kvm_get_pfn(pfn_t pfn);
346 int kvm_read_guest_page(struct kvm *kvm, gfn_t gfn, void *data, int offset,
348 int kvm_read_guest_atomic(struct kvm *kvm, gpa_t gpa, void *data,
350 int kvm_read_guest(struct kvm *kvm, gpa_t gpa, void *data, unsigned long len);
351 int kvm_write_guest_page(struct kvm *kvm, gfn_t gfn, const void *data,
352 int offset, int len);
353 int kvm_write_guest(struct kvm *kvm, gpa_t gpa, const void *data,
355 int kvm_clear_guest_page(struct kvm *kvm, gfn_t gfn, int offset, int len);
356 int kvm_clear_guest(struct kvm *kvm, gpa_t gpa, unsigned long len);
357 struct kvm_memory_slot *gfn_to_memslot(struct kvm *kvm, gfn_t gfn);
358 int kvm_is_visible_gfn(struct kvm *kvm, gfn_t gfn);
359 unsigned long kvm_host_page_size(struct kvm *kvm, gfn_t gfn);
360 void mark_page_dirty(struct kvm *kvm, gfn_t gfn);
362 void kvm_vcpu_block(struct kvm_vcpu *vcpu);
363 void kvm_vcpu_on_spin(struct kvm_vcpu *vcpu);
364 void kvm_resched(struct kvm_vcpu *vcpu);
365 void kvm_load_guest_fpu(struct kvm_vcpu *vcpu);
366 void kvm_put_guest_fpu(struct kvm_vcpu *vcpu);
367 void kvm_flush_remote_tlbs(struct kvm *kvm);
368 void kvm_reload_remote_mmus(struct kvm *kvm);
370 long kvm_arch_dev_ioctl(struct file *filp,
371 unsigned int ioctl, unsigned long arg);
372 long kvm_arch_vcpu_ioctl(struct file *filp,
373 unsigned int ioctl, unsigned long arg);
375 int kvm_dev_ioctl_check_extension(long ext);
377 int kvm_get_dirty_log(struct kvm *kvm,
378 struct kvm_dirty_log *log, int *is_dirty);
379 int kvm_vm_ioctl_get_dirty_log(struct kvm *kvm,
380 struct kvm_dirty_log *log);
382 int kvm_vm_ioctl_set_memory_region(struct kvm *kvm,
384 kvm_userspace_memory_region *mem,
386 long kvm_arch_vm_ioctl(struct file *filp,
387 unsigned int ioctl, unsigned long arg);
389 int kvm_arch_vcpu_ioctl_get_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu);
390 int kvm_arch_vcpu_ioctl_set_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu);
392 int kvm_arch_vcpu_ioctl_translate(struct kvm_vcpu *vcpu,
393 struct kvm_translation *tr);
395 int kvm_arch_vcpu_ioctl_get_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs);
396 int kvm_arch_vcpu_ioctl_set_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs);
397 int kvm_arch_vcpu_ioctl_get_sregs(struct kvm_vcpu *vcpu,
398 struct kvm_sregs *sregs);
399 int kvm_arch_vcpu_ioctl_set_sregs(struct kvm_vcpu *vcpu,
400 struct kvm_sregs *sregs);
401 int kvm_arch_vcpu_ioctl_get_mpstate(struct kvm_vcpu *vcpu,
402 struct kvm_mp_state *mp_state);
403 int kvm_arch_vcpu_ioctl_set_mpstate(struct kvm_vcpu *vcpu,
404 struct kvm_mp_state *mp_state);
405 int kvm_arch_vcpu_ioctl_set_guest_debug(struct kvm_vcpu *vcpu,
406 struct kvm_guest_debug *dbg);
407 int kvm_arch_vcpu_ioctl_run(struct kvm_vcpu *vcpu, struct kvm_run *kvm_run);
409 int kvm_arch_init(void *opaque);
410 void kvm_arch_exit(void);
412 int kvm_arch_vcpu_init(struct kvm_vcpu *vcpu);
413 void kvm_arch_vcpu_uninit(struct kvm_vcpu *vcpu);
415 void kvm_arch_vcpu_free(struct kvm_vcpu *vcpu);
416 void kvm_arch_vcpu_load(struct kvm_vcpu *vcpu, int cpu);
417 void kvm_arch_vcpu_put(struct kvm_vcpu *vcpu);
418 struct kvm_vcpu *kvm_arch_vcpu_create(struct kvm *kvm, unsigned int id);
419 int kvm_arch_vcpu_setup(struct kvm_vcpu *vcpu);
420 void kvm_arch_vcpu_destroy(struct kvm_vcpu *vcpu);
422 int kvm_arch_vcpu_reset(struct kvm_vcpu *vcpu);
423 int kvm_arch_hardware_enable(void *garbage);
424 void kvm_arch_hardware_disable(void *garbage);
425 int kvm_arch_hardware_setup(void);
426 void kvm_arch_hardware_unsetup(void);
427 void kvm_arch_check_processor_compat(void *rtn);
428 int kvm_arch_vcpu_runnable(struct kvm_vcpu *vcpu);
430 void kvm_free_physmem(struct kvm *kvm);
432 struct kvm *kvm_arch_create_vm(void);
433 void kvm_arch_destroy_vm(struct kvm *kvm);
434 void kvm_free_all_assigned_devices(struct kvm *kvm);
435 void kvm_arch_sync_events(struct kvm *kvm);
437 int kvm_cpu_has_pending_timer(struct kvm_vcpu *vcpu);
438 void kvm_vcpu_kick(struct kvm_vcpu *vcpu);
440 int kvm_is_mmio_pfn(pfn_t pfn);
442 struct kvm_irq_ack_notifier {
443 struct hlist_node link;
445 void (*irq_acked)(struct kvm_irq_ack_notifier *kian);
448 #define KVM_ASSIGNED_MSIX_PENDING 0x1
449 struct kvm_guest_msix_entry {
455 struct kvm_assigned_dev_kernel {
456 struct kvm_irq_ack_notifier ack_notifier;
457 struct work_struct interrupt_work;
458 struct list_head list;
463 unsigned int entries_nr;
465 bool host_irq_disabled;
466 struct msix_entry *host_msix_entries;
468 struct kvm_guest_msix_entry *guest_msix_entries;
469 unsigned long irq_requested_type;
474 spinlock_t assigned_dev_lock;
477 struct kvm_irq_mask_notifier {
478 void (*func)(struct kvm_irq_mask_notifier *kimn, bool masked);
480 struct hlist_node link;
483 void kvm_register_irq_mask_notifier(struct kvm *kvm, int irq,
484 struct kvm_irq_mask_notifier *kimn);
485 void kvm_unregister_irq_mask_notifier(struct kvm *kvm, int irq,
486 struct kvm_irq_mask_notifier *kimn);
487 void kvm_fire_mask_notifiers(struct kvm *kvm, unsigned irqchip, unsigned pin,
490 #ifdef __KVM_HAVE_IOAPIC
491 void kvm_get_intr_delivery_bitmask(struct kvm_ioapic *ioapic,
492 union kvm_ioapic_redirect_entry *entry,
493 unsigned long *deliver_bitmask);
495 int kvm_set_irq(struct kvm *kvm, int irq_source_id, u32 irq, int level);
496 void kvm_notify_acked_irq(struct kvm *kvm, unsigned irqchip, unsigned pin);
497 void kvm_register_irq_ack_notifier(struct kvm *kvm,
498 struct kvm_irq_ack_notifier *kian);
499 void kvm_unregister_irq_ack_notifier(struct kvm *kvm,
500 struct kvm_irq_ack_notifier *kian);
501 int kvm_request_irq_source_id(struct kvm *kvm);
502 void kvm_free_irq_source_id(struct kvm *kvm, int irq_source_id);
504 /* For vcpu->arch.iommu_flags */
505 #define KVM_IOMMU_CACHE_COHERENCY 0x1
507 #ifdef CONFIG_IOMMU_API
508 int kvm_iommu_map_pages(struct kvm *kvm, struct kvm_memory_slot *slot);
509 int kvm_iommu_map_guest(struct kvm *kvm);
510 int kvm_iommu_unmap_guest(struct kvm *kvm);
511 int kvm_assign_device(struct kvm *kvm,
512 struct kvm_assigned_dev_kernel *assigned_dev);
513 int kvm_deassign_device(struct kvm *kvm,
514 struct kvm_assigned_dev_kernel *assigned_dev);
515 #else /* CONFIG_IOMMU_API */
516 static inline int kvm_iommu_map_pages(struct kvm *kvm,
517 struct kvm_memory_slot *slot)
522 static inline int kvm_iommu_map_guest(struct kvm *kvm)
527 static inline int kvm_iommu_unmap_guest(struct kvm *kvm)
532 static inline int kvm_assign_device(struct kvm *kvm,
533 struct kvm_assigned_dev_kernel *assigned_dev)
538 static inline int kvm_deassign_device(struct kvm *kvm,
539 struct kvm_assigned_dev_kernel *assigned_dev)
543 #endif /* CONFIG_IOMMU_API */
545 static inline void kvm_guest_enter(void)
547 account_system_vtime(current);
548 current->flags |= PF_VCPU;
551 static inline void kvm_guest_exit(void)
553 account_system_vtime(current);
554 current->flags &= ~PF_VCPU;
557 static inline unsigned long gfn_to_hva_memslot(struct kvm_memory_slot *slot,
560 return slot->userspace_addr + (gfn - slot->base_gfn) * PAGE_SIZE;
563 static inline gpa_t gfn_to_gpa(gfn_t gfn)
565 return (gpa_t)gfn << PAGE_SHIFT;
568 static inline gfn_t gpa_to_gfn(gpa_t gpa)
570 return (gfn_t)(gpa >> PAGE_SHIFT);
573 static inline hpa_t pfn_to_hpa(pfn_t pfn)
575 return (hpa_t)pfn << PAGE_SHIFT;
578 static inline void kvm_migrate_timers(struct kvm_vcpu *vcpu)
580 set_bit(KVM_REQ_MIGRATE_TIMER, &vcpu->requests);
588 struct kvm_stats_debugfs_item {
591 enum kvm_stat_kind kind;
592 struct dentry *dentry;
594 extern struct kvm_stats_debugfs_item debugfs_entries[];
595 extern struct dentry *kvm_debugfs_dir;
597 #ifdef KVM_ARCH_WANT_MMU_NOTIFIER
598 static inline int mmu_notifier_retry(struct kvm_vcpu *vcpu, unsigned long mmu_seq)
600 if (unlikely(vcpu->kvm->mmu_notifier_count))
603 * Both reads happen under the mmu_lock and both values are
604 * modified under mmu_lock, so there's no need of smb_rmb()
605 * here in between, otherwise mmu_notifier_count should be
606 * read before mmu_notifier_seq, see
607 * mmu_notifier_invalidate_range_end write side.
609 if (vcpu->kvm->mmu_notifier_seq != mmu_seq)
615 #ifdef CONFIG_HAVE_KVM_IRQCHIP
617 #define KVM_MAX_IRQ_ROUTES 1024
619 int kvm_setup_default_irq_routing(struct kvm *kvm);
620 int kvm_set_irq_routing(struct kvm *kvm,
621 const struct kvm_irq_routing_entry *entries,
624 void kvm_free_irq_routing(struct kvm *kvm);
628 static inline void kvm_free_irq_routing(struct kvm *kvm) {}
632 #ifdef CONFIG_HAVE_KVM_EVENTFD
634 void kvm_eventfd_init(struct kvm *kvm);
635 int kvm_irqfd(struct kvm *kvm, int fd, int gsi, int flags);
636 void kvm_irqfd_release(struct kvm *kvm);
637 int kvm_ioeventfd(struct kvm *kvm, struct kvm_ioeventfd *args);
641 static inline void kvm_eventfd_init(struct kvm *kvm) {}
642 static inline int kvm_irqfd(struct kvm *kvm, int fd, int gsi, int flags)
647 static inline void kvm_irqfd_release(struct kvm *kvm) {}
648 static inline int kvm_ioeventfd(struct kvm *kvm, struct kvm_ioeventfd *args)
653 #endif /* CONFIG_HAVE_KVM_EVENTFD */
655 #ifdef CONFIG_KVM_APIC_ARCHITECTURE
656 static inline bool kvm_vcpu_is_bsp(struct kvm_vcpu *vcpu)
658 return vcpu->kvm->bsp_vcpu_id == vcpu->vcpu_id;
662 #ifdef __KVM_HAVE_DEVICE_ASSIGNMENT
664 long kvm_vm_ioctl_assigned_device(struct kvm *kvm, unsigned ioctl,
669 static inline long kvm_vm_ioctl_assigned_device(struct kvm *kvm, unsigned ioctl,
677 static inline void kvm_make_request(int req, struct kvm_vcpu *vcpu)
679 set_bit(req, &vcpu->requests);
682 static inline bool kvm_make_check_request(int req, struct kvm_vcpu *vcpu)
684 return test_and_set_bit(req, &vcpu->requests);
687 static inline bool kvm_check_request(int req, struct kvm_vcpu *vcpu)
689 if (test_bit(req, &vcpu->requests)) {
690 clear_bit(req, &vcpu->requests);