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;
86 unsigned long requests;
87 unsigned long guest_debug;
91 int guest_fpu_loaded, guest_xcr0_loaded;
95 struct kvm_vcpu_stat stat;
97 #ifdef CONFIG_HAS_IOMEM
99 int mmio_read_completed;
102 unsigned char mmio_data[8];
103 gpa_t mmio_phys_addr;
106 struct kvm_vcpu_arch arch;
110 * Some of the bitops functions do not support too long bitmaps.
111 * This number must be determined not to exceed such limits.
113 #define KVM_MEM_MAX_NR_PAGES ((1UL << 31) - 1)
115 struct kvm_memory_slot {
117 unsigned long npages;
120 unsigned long *dirty_bitmap;
122 unsigned long rmap_pde;
124 } *lpage_info[KVM_NR_PAGE_SIZES - 1];
125 unsigned long userspace_addr;
129 static inline unsigned long kvm_dirty_bitmap_bytes(struct kvm_memory_slot *memslot)
131 return ALIGN(memslot->npages, BITS_PER_LONG) / 8;
134 struct kvm_kernel_irq_routing_entry {
137 int (*set)(struct kvm_kernel_irq_routing_entry *e,
138 struct kvm *kvm, int irq_source_id, int level);
146 struct hlist_node link;
149 #ifdef __KVM_HAVE_IOAPIC
151 struct kvm_irq_routing_table {
152 int chip[KVM_NR_IRQCHIPS][KVM_IOAPIC_NUM_PINS];
153 struct kvm_kernel_irq_routing_entry *rt_entries;
156 * Array indexed by gsi. Each entry contains list of irq chips
157 * the gsi is connected to.
159 struct hlist_head map[0];
164 struct kvm_irq_routing_table {};
168 struct kvm_memslots {
170 struct kvm_memory_slot memslots[KVM_MEMORY_SLOTS +
171 KVM_PRIVATE_MEM_SLOTS];
176 raw_spinlock_t requests_lock;
177 struct mutex slots_lock;
178 struct mm_struct *mm; /* userspace tied to this vm */
179 struct kvm_memslots *memslots;
180 struct srcu_struct srcu;
181 #ifdef CONFIG_KVM_APIC_ARCHITECTURE
183 struct kvm_vcpu *bsp_vcpu;
185 struct kvm_vcpu *vcpus[KVM_MAX_VCPUS];
186 atomic_t online_vcpus;
187 struct list_head vm_list;
189 struct kvm_io_bus *buses[KVM_NR_BUSES];
190 #ifdef CONFIG_HAVE_KVM_EVENTFD
193 struct list_head items;
195 struct list_head ioeventfds;
197 struct kvm_vm_stat stat;
198 struct kvm_arch arch;
199 atomic_t users_count;
200 #ifdef KVM_COALESCED_MMIO_PAGE_OFFSET
201 struct kvm_coalesced_mmio_dev *coalesced_mmio_dev;
202 struct kvm_coalesced_mmio_ring *coalesced_mmio_ring;
205 struct mutex irq_lock;
206 #ifdef CONFIG_HAVE_KVM_IRQCHIP
207 struct kvm_irq_routing_table *irq_routing;
208 struct hlist_head mask_notifier_list;
209 struct hlist_head irq_ack_notifier_list;
212 #ifdef KVM_ARCH_WANT_MMU_NOTIFIER
213 struct mmu_notifier mmu_notifier;
214 unsigned long mmu_notifier_seq;
215 long mmu_notifier_count;
219 /* The guest did something we don't support. */
220 #define pr_unimpl(vcpu, fmt, ...) \
222 if (printk_ratelimit()) \
223 printk(KERN_ERR "kvm: %i: cpu%i " fmt, \
224 current->tgid, (vcpu)->vcpu_id , ## __VA_ARGS__); \
227 #define kvm_printf(kvm, fmt ...) printk(KERN_DEBUG fmt)
228 #define vcpu_printf(vcpu, fmt...) kvm_printf(vcpu->kvm, fmt)
230 static inline struct kvm_vcpu *kvm_get_vcpu(struct kvm *kvm, int i)
233 return kvm->vcpus[i];
236 #define kvm_for_each_vcpu(idx, vcpup, kvm) \
237 for (idx = 0, vcpup = kvm_get_vcpu(kvm, idx); \
238 idx < atomic_read(&kvm->online_vcpus) && vcpup; \
239 vcpup = kvm_get_vcpu(kvm, ++idx))
241 int kvm_vcpu_init(struct kvm_vcpu *vcpu, struct kvm *kvm, unsigned id);
242 void kvm_vcpu_uninit(struct kvm_vcpu *vcpu);
244 void vcpu_load(struct kvm_vcpu *vcpu);
245 void vcpu_put(struct kvm_vcpu *vcpu);
247 int kvm_init(void *opaque, unsigned vcpu_size, unsigned vcpu_align,
248 struct module *module);
251 void kvm_get_kvm(struct kvm *kvm);
252 void kvm_put_kvm(struct kvm *kvm);
254 static inline struct kvm_memslots *kvm_memslots(struct kvm *kvm)
256 return rcu_dereference_check(kvm->memslots,
257 srcu_read_lock_held(&kvm->srcu)
258 || lockdep_is_held(&kvm->slots_lock));
261 #define HPA_MSB ((sizeof(hpa_t) * 8) - 1)
262 #define HPA_ERR_MASK ((hpa_t)1 << HPA_MSB)
263 static inline int is_error_hpa(hpa_t hpa) { return hpa >> HPA_MSB; }
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 is_hwpoison_pfn(pfn_t pfn);
271 int kvm_is_error_hva(unsigned long addr);
272 int kvm_set_memory_region(struct kvm *kvm,
273 struct kvm_userspace_memory_region *mem,
275 int __kvm_set_memory_region(struct kvm *kvm,
276 struct kvm_userspace_memory_region *mem,
278 int kvm_arch_prepare_memory_region(struct kvm *kvm,
279 struct kvm_memory_slot *memslot,
280 struct kvm_memory_slot old,
281 struct kvm_userspace_memory_region *mem,
283 void kvm_arch_commit_memory_region(struct kvm *kvm,
284 struct kvm_userspace_memory_region *mem,
285 struct kvm_memory_slot old,
287 void kvm_disable_largepages(void);
288 void kvm_arch_flush_shadow(struct kvm *kvm);
290 struct page *gfn_to_page(struct kvm *kvm, gfn_t gfn);
291 unsigned long gfn_to_hva(struct kvm *kvm, gfn_t gfn);
292 void kvm_release_page_clean(struct page *page);
293 void kvm_release_page_dirty(struct page *page);
294 void kvm_set_page_dirty(struct page *page);
295 void kvm_set_page_accessed(struct page *page);
297 pfn_t gfn_to_pfn(struct kvm *kvm, gfn_t gfn);
298 pfn_t gfn_to_pfn_memslot(struct kvm *kvm,
299 struct kvm_memory_slot *slot, gfn_t gfn);
300 int memslot_id(struct kvm *kvm, gfn_t gfn);
301 void kvm_release_pfn_dirty(pfn_t);
302 void kvm_release_pfn_clean(pfn_t pfn);
303 void kvm_set_pfn_dirty(pfn_t pfn);
304 void kvm_set_pfn_accessed(pfn_t pfn);
305 void kvm_get_pfn(pfn_t pfn);
307 int kvm_read_guest_page(struct kvm *kvm, gfn_t gfn, void *data, int offset,
309 int kvm_read_guest_atomic(struct kvm *kvm, gpa_t gpa, void *data,
311 int kvm_read_guest(struct kvm *kvm, gpa_t gpa, void *data, unsigned long len);
312 int kvm_write_guest_page(struct kvm *kvm, gfn_t gfn, const void *data,
313 int offset, int len);
314 int kvm_write_guest(struct kvm *kvm, gpa_t gpa, const void *data,
316 int kvm_clear_guest_page(struct kvm *kvm, gfn_t gfn, int offset, int len);
317 int kvm_clear_guest(struct kvm *kvm, gpa_t gpa, unsigned long len);
318 struct kvm_memory_slot *gfn_to_memslot(struct kvm *kvm, gfn_t gfn);
319 int kvm_is_visible_gfn(struct kvm *kvm, gfn_t gfn);
320 unsigned long kvm_host_page_size(struct kvm *kvm, gfn_t gfn);
321 void mark_page_dirty(struct kvm *kvm, gfn_t gfn);
323 void kvm_vcpu_block(struct kvm_vcpu *vcpu);
324 void kvm_vcpu_on_spin(struct kvm_vcpu *vcpu);
325 void kvm_resched(struct kvm_vcpu *vcpu);
326 void kvm_load_guest_fpu(struct kvm_vcpu *vcpu);
327 void kvm_put_guest_fpu(struct kvm_vcpu *vcpu);
328 void kvm_flush_remote_tlbs(struct kvm *kvm);
329 void kvm_reload_remote_mmus(struct kvm *kvm);
331 long kvm_arch_dev_ioctl(struct file *filp,
332 unsigned int ioctl, unsigned long arg);
333 long kvm_arch_vcpu_ioctl(struct file *filp,
334 unsigned int ioctl, unsigned long arg);
336 int kvm_dev_ioctl_check_extension(long ext);
338 int kvm_get_dirty_log(struct kvm *kvm,
339 struct kvm_dirty_log *log, int *is_dirty);
340 int kvm_vm_ioctl_get_dirty_log(struct kvm *kvm,
341 struct kvm_dirty_log *log);
343 int kvm_vm_ioctl_set_memory_region(struct kvm *kvm,
345 kvm_userspace_memory_region *mem,
347 long kvm_arch_vm_ioctl(struct file *filp,
348 unsigned int ioctl, unsigned long arg);
350 int kvm_arch_vcpu_ioctl_get_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu);
351 int kvm_arch_vcpu_ioctl_set_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu);
353 int kvm_arch_vcpu_ioctl_translate(struct kvm_vcpu *vcpu,
354 struct kvm_translation *tr);
356 int kvm_arch_vcpu_ioctl_get_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs);
357 int kvm_arch_vcpu_ioctl_set_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs);
358 int kvm_arch_vcpu_ioctl_get_sregs(struct kvm_vcpu *vcpu,
359 struct kvm_sregs *sregs);
360 int kvm_arch_vcpu_ioctl_set_sregs(struct kvm_vcpu *vcpu,
361 struct kvm_sregs *sregs);
362 int kvm_arch_vcpu_ioctl_get_mpstate(struct kvm_vcpu *vcpu,
363 struct kvm_mp_state *mp_state);
364 int kvm_arch_vcpu_ioctl_set_mpstate(struct kvm_vcpu *vcpu,
365 struct kvm_mp_state *mp_state);
366 int kvm_arch_vcpu_ioctl_set_guest_debug(struct kvm_vcpu *vcpu,
367 struct kvm_guest_debug *dbg);
368 int kvm_arch_vcpu_ioctl_run(struct kvm_vcpu *vcpu, struct kvm_run *kvm_run);
370 int kvm_arch_init(void *opaque);
371 void kvm_arch_exit(void);
373 int kvm_arch_vcpu_init(struct kvm_vcpu *vcpu);
374 void kvm_arch_vcpu_uninit(struct kvm_vcpu *vcpu);
376 void kvm_arch_vcpu_free(struct kvm_vcpu *vcpu);
377 void kvm_arch_vcpu_load(struct kvm_vcpu *vcpu, int cpu);
378 void kvm_arch_vcpu_put(struct kvm_vcpu *vcpu);
379 struct kvm_vcpu *kvm_arch_vcpu_create(struct kvm *kvm, unsigned int id);
380 int kvm_arch_vcpu_setup(struct kvm_vcpu *vcpu);
381 void kvm_arch_vcpu_destroy(struct kvm_vcpu *vcpu);
383 int kvm_arch_vcpu_reset(struct kvm_vcpu *vcpu);
384 int kvm_arch_hardware_enable(void *garbage);
385 void kvm_arch_hardware_disable(void *garbage);
386 int kvm_arch_hardware_setup(void);
387 void kvm_arch_hardware_unsetup(void);
388 void kvm_arch_check_processor_compat(void *rtn);
389 int kvm_arch_vcpu_runnable(struct kvm_vcpu *vcpu);
391 void kvm_free_physmem(struct kvm *kvm);
393 struct kvm *kvm_arch_create_vm(void);
394 void kvm_arch_destroy_vm(struct kvm *kvm);
395 void kvm_free_all_assigned_devices(struct kvm *kvm);
396 void kvm_arch_sync_events(struct kvm *kvm);
398 int kvm_cpu_has_pending_timer(struct kvm_vcpu *vcpu);
399 void kvm_vcpu_kick(struct kvm_vcpu *vcpu);
401 int kvm_is_mmio_pfn(pfn_t pfn);
403 struct kvm_irq_ack_notifier {
404 struct hlist_node link;
406 void (*irq_acked)(struct kvm_irq_ack_notifier *kian);
409 #define KVM_ASSIGNED_MSIX_PENDING 0x1
410 struct kvm_guest_msix_entry {
416 struct kvm_assigned_dev_kernel {
417 struct kvm_irq_ack_notifier ack_notifier;
418 struct work_struct interrupt_work;
419 struct list_head list;
424 unsigned int entries_nr;
426 bool host_irq_disabled;
427 struct msix_entry *host_msix_entries;
429 struct kvm_guest_msix_entry *guest_msix_entries;
430 unsigned long irq_requested_type;
435 spinlock_t assigned_dev_lock;
438 struct kvm_irq_mask_notifier {
439 void (*func)(struct kvm_irq_mask_notifier *kimn, bool masked);
441 struct hlist_node link;
444 void kvm_register_irq_mask_notifier(struct kvm *kvm, int irq,
445 struct kvm_irq_mask_notifier *kimn);
446 void kvm_unregister_irq_mask_notifier(struct kvm *kvm, int irq,
447 struct kvm_irq_mask_notifier *kimn);
448 void kvm_fire_mask_notifiers(struct kvm *kvm, int irq, bool mask);
450 #ifdef __KVM_HAVE_IOAPIC
451 void kvm_get_intr_delivery_bitmask(struct kvm_ioapic *ioapic,
452 union kvm_ioapic_redirect_entry *entry,
453 unsigned long *deliver_bitmask);
455 int kvm_set_irq(struct kvm *kvm, int irq_source_id, u32 irq, int level);
456 void kvm_notify_acked_irq(struct kvm *kvm, unsigned irqchip, unsigned pin);
457 void kvm_register_irq_ack_notifier(struct kvm *kvm,
458 struct kvm_irq_ack_notifier *kian);
459 void kvm_unregister_irq_ack_notifier(struct kvm *kvm,
460 struct kvm_irq_ack_notifier *kian);
461 int kvm_request_irq_source_id(struct kvm *kvm);
462 void kvm_free_irq_source_id(struct kvm *kvm, int irq_source_id);
464 /* For vcpu->arch.iommu_flags */
465 #define KVM_IOMMU_CACHE_COHERENCY 0x1
467 #ifdef CONFIG_IOMMU_API
468 int kvm_iommu_map_pages(struct kvm *kvm, struct kvm_memory_slot *slot);
469 int kvm_iommu_map_guest(struct kvm *kvm);
470 int kvm_iommu_unmap_guest(struct kvm *kvm);
471 int kvm_assign_device(struct kvm *kvm,
472 struct kvm_assigned_dev_kernel *assigned_dev);
473 int kvm_deassign_device(struct kvm *kvm,
474 struct kvm_assigned_dev_kernel *assigned_dev);
475 #else /* CONFIG_IOMMU_API */
476 static inline int kvm_iommu_map_pages(struct kvm *kvm,
478 unsigned long npages)
483 static inline int kvm_iommu_map_guest(struct kvm *kvm)
488 static inline int kvm_iommu_unmap_guest(struct kvm *kvm)
493 static inline int kvm_assign_device(struct kvm *kvm,
494 struct kvm_assigned_dev_kernel *assigned_dev)
499 static inline int kvm_deassign_device(struct kvm *kvm,
500 struct kvm_assigned_dev_kernel *assigned_dev)
504 #endif /* CONFIG_IOMMU_API */
506 static inline void kvm_guest_enter(void)
508 account_system_vtime(current);
509 current->flags |= PF_VCPU;
512 static inline void kvm_guest_exit(void)
514 account_system_vtime(current);
515 current->flags &= ~PF_VCPU;
518 static inline gpa_t gfn_to_gpa(gfn_t gfn)
520 return (gpa_t)gfn << PAGE_SHIFT;
523 static inline hpa_t pfn_to_hpa(pfn_t pfn)
525 return (hpa_t)pfn << PAGE_SHIFT;
528 static inline void kvm_migrate_timers(struct kvm_vcpu *vcpu)
530 set_bit(KVM_REQ_MIGRATE_TIMER, &vcpu->requests);
538 struct kvm_stats_debugfs_item {
541 enum kvm_stat_kind kind;
542 struct dentry *dentry;
544 extern struct kvm_stats_debugfs_item debugfs_entries[];
545 extern struct dentry *kvm_debugfs_dir;
547 #ifdef KVM_ARCH_WANT_MMU_NOTIFIER
548 static inline int mmu_notifier_retry(struct kvm_vcpu *vcpu, unsigned long mmu_seq)
550 if (unlikely(vcpu->kvm->mmu_notifier_count))
553 * Both reads happen under the mmu_lock and both values are
554 * modified under mmu_lock, so there's no need of smb_rmb()
555 * here in between, otherwise mmu_notifier_count should be
556 * read before mmu_notifier_seq, see
557 * mmu_notifier_invalidate_range_end write side.
559 if (vcpu->kvm->mmu_notifier_seq != mmu_seq)
565 #ifdef CONFIG_HAVE_KVM_IRQCHIP
567 #define KVM_MAX_IRQ_ROUTES 1024
569 int kvm_setup_default_irq_routing(struct kvm *kvm);
570 int kvm_set_irq_routing(struct kvm *kvm,
571 const struct kvm_irq_routing_entry *entries,
574 void kvm_free_irq_routing(struct kvm *kvm);
578 static inline void kvm_free_irq_routing(struct kvm *kvm) {}
582 #ifdef CONFIG_HAVE_KVM_EVENTFD
584 void kvm_eventfd_init(struct kvm *kvm);
585 int kvm_irqfd(struct kvm *kvm, int fd, int gsi, int flags);
586 void kvm_irqfd_release(struct kvm *kvm);
587 int kvm_ioeventfd(struct kvm *kvm, struct kvm_ioeventfd *args);
591 static inline void kvm_eventfd_init(struct kvm *kvm) {}
592 static inline int kvm_irqfd(struct kvm *kvm, int fd, int gsi, int flags)
597 static inline void kvm_irqfd_release(struct kvm *kvm) {}
598 static inline int kvm_ioeventfd(struct kvm *kvm, struct kvm_ioeventfd *args)
603 #endif /* CONFIG_HAVE_KVM_EVENTFD */
605 #ifdef CONFIG_KVM_APIC_ARCHITECTURE
606 static inline bool kvm_vcpu_is_bsp(struct kvm_vcpu *vcpu)
608 return vcpu->kvm->bsp_vcpu_id == vcpu->vcpu_id;
612 #ifdef __KVM_HAVE_DEVICE_ASSIGNMENT
614 long kvm_vm_ioctl_assigned_device(struct kvm *kvm, unsigned ioctl,
619 static inline long kvm_vm_ioctl_assigned_device(struct kvm *kvm, unsigned ioctl,
627 static inline void kvm_make_request(int req, struct kvm_vcpu *vcpu)
629 set_bit(req, &vcpu->requests);
632 static inline bool kvm_make_check_request(int req, struct kvm_vcpu *vcpu)
634 return test_and_set_bit(req, &vcpu->requests);
637 static inline bool kvm_check_request(int req, struct kvm_vcpu *vcpu)
639 if (test_bit(req, &vcpu->requests)) {
640 clear_bit(req, &vcpu->requests);