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>
16 #include <linux/bug.h>
18 #include <linux/mmu_notifier.h>
19 #include <linux/preempt.h>
20 #include <linux/msi.h>
21 #include <linux/slab.h>
22 #include <linux/rcupdate.h>
23 #include <linux/ratelimit.h>
24 #include <linux/err.h>
25 #include <linux/irqflags.h>
26 #include <linux/context_tracking.h>
27 #include <asm/signal.h>
29 #include <linux/kvm.h>
30 #include <linux/kvm_para.h>
32 #include <linux/kvm_types.h>
34 #include <asm/kvm_host.h>
37 #define KVM_MMIO_SIZE 8
41 * The bit 16 ~ bit 31 of kvm_memory_region::flags are internally used
42 * in kvm, other bits are visible for userspace which are defined in
43 * include/linux/kvm_h.
45 #define KVM_MEMSLOT_INVALID (1UL << 16)
47 /* Two fragments for cross MMIO pages. */
48 #define KVM_MAX_MMIO_FRAGMENTS 2
51 * For the normal pfn, the highest 12 bits should be zero,
52 * so we can mask bit 62 ~ bit 52 to indicate the error pfn,
53 * mask bit 63 to indicate the noslot pfn.
55 #define KVM_PFN_ERR_MASK (0x7ffULL << 52)
56 #define KVM_PFN_ERR_NOSLOT_MASK (0xfffULL << 52)
57 #define KVM_PFN_NOSLOT (0x1ULL << 63)
59 #define KVM_PFN_ERR_FAULT (KVM_PFN_ERR_MASK)
60 #define KVM_PFN_ERR_HWPOISON (KVM_PFN_ERR_MASK + 1)
61 #define KVM_PFN_ERR_RO_FAULT (KVM_PFN_ERR_MASK + 2)
64 * error pfns indicate that the gfn is in slot but faild to
65 * translate it to pfn on host.
67 static inline bool is_error_pfn(pfn_t pfn)
69 return !!(pfn & KVM_PFN_ERR_MASK);
73 * error_noslot pfns indicate that the gfn can not be
74 * translated to pfn - it is not in slot or failed to
75 * translate it to pfn.
77 static inline bool is_error_noslot_pfn(pfn_t pfn)
79 return !!(pfn & KVM_PFN_ERR_NOSLOT_MASK);
82 /* noslot pfn indicates that the gfn is not in slot. */
83 static inline bool is_noslot_pfn(pfn_t pfn)
85 return pfn == KVM_PFN_NOSLOT;
88 #define KVM_HVA_ERR_BAD (PAGE_OFFSET)
89 #define KVM_HVA_ERR_RO_BAD (PAGE_OFFSET + PAGE_SIZE)
91 static inline bool kvm_is_error_hva(unsigned long addr)
93 return addr >= PAGE_OFFSET;
96 #define KVM_ERR_PTR_BAD_PAGE (ERR_PTR(-ENOENT))
98 static inline bool is_error_page(struct page *page)
104 * vcpu->requests bit members
106 #define KVM_REQ_TLB_FLUSH 0
107 #define KVM_REQ_MIGRATE_TIMER 1
108 #define KVM_REQ_REPORT_TPR_ACCESS 2
109 #define KVM_REQ_MMU_RELOAD 3
110 #define KVM_REQ_TRIPLE_FAULT 4
111 #define KVM_REQ_PENDING_TIMER 5
112 #define KVM_REQ_UNHALT 6
113 #define KVM_REQ_MMU_SYNC 7
114 #define KVM_REQ_CLOCK_UPDATE 8
115 #define KVM_REQ_KICK 9
116 #define KVM_REQ_DEACTIVATE_FPU 10
117 #define KVM_REQ_EVENT 11
118 #define KVM_REQ_APF_HALT 12
119 #define KVM_REQ_STEAL_UPDATE 13
120 #define KVM_REQ_NMI 14
121 #define KVM_REQ_PMU 15
122 #define KVM_REQ_PMI 16
123 #define KVM_REQ_WATCHDOG 17
124 #define KVM_REQ_MASTERCLOCK_UPDATE 18
125 #define KVM_REQ_MCLOCK_INPROGRESS 19
126 #define KVM_REQ_EPR_EXIT 20
127 #define KVM_REQ_SCAN_IOAPIC 21
129 #define KVM_USERSPACE_IRQ_SOURCE_ID 0
130 #define KVM_IRQFD_RESAMPLE_IRQ_SOURCE_ID 1
134 extern struct kmem_cache *kvm_vcpu_cache;
136 extern spinlock_t kvm_lock;
137 extern struct list_head vm_list;
139 struct kvm_io_range {
142 struct kvm_io_device *dev;
145 #define NR_IOBUS_DEVS 1000
149 struct kvm_io_range range[];
155 KVM_VIRTIO_CCW_NOTIFY_BUS,
159 int kvm_io_bus_write(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr,
160 int len, const void *val);
161 int kvm_io_bus_read(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr, int len,
163 int kvm_io_bus_register_dev(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr,
164 int len, struct kvm_io_device *dev);
165 int kvm_io_bus_unregister_dev(struct kvm *kvm, enum kvm_bus bus_idx,
166 struct kvm_io_device *dev);
168 #ifdef CONFIG_KVM_ASYNC_PF
169 struct kvm_async_pf {
170 struct work_struct work;
171 struct list_head link;
172 struct list_head queue;
173 struct kvm_vcpu *vcpu;
174 struct mm_struct *mm;
177 struct kvm_arch_async_pf arch;
181 void kvm_clear_async_pf_completion_queue(struct kvm_vcpu *vcpu);
182 void kvm_check_async_pf_completion(struct kvm_vcpu *vcpu);
183 int kvm_setup_async_pf(struct kvm_vcpu *vcpu, gva_t gva, unsigned long hva,
184 struct kvm_arch_async_pf *arch);
185 int kvm_async_pf_wakeup_all(struct kvm_vcpu *vcpu);
192 READING_SHADOW_PAGE_TABLES,
196 * Sometimes a large or cross-page mmio needs to be broken up into separate
197 * exits for userspace servicing.
199 struct kvm_mmio_fragment {
207 #ifdef CONFIG_PREEMPT_NOTIFIERS
208 struct preempt_notifier preempt_notifier;
214 unsigned long requests;
215 unsigned long guest_debug;
221 int guest_fpu_loaded, guest_xcr0_loaded;
222 wait_queue_head_t wq;
226 struct kvm_vcpu_stat stat;
228 #ifdef CONFIG_HAS_IOMEM
230 int mmio_read_completed;
232 int mmio_cur_fragment;
233 int mmio_nr_fragments;
234 struct kvm_mmio_fragment mmio_fragments[KVM_MAX_MMIO_FRAGMENTS];
237 #ifdef CONFIG_KVM_ASYNC_PF
240 struct list_head queue;
241 struct list_head done;
246 #ifdef CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT
248 * Cpu relax intercept or pause loop exit optimization
249 * in_spin_loop: set when a vcpu does a pause loop exit
250 * or cpu relax intercepted.
251 * dy_eligible: indicates whether vcpu is eligible for directed yield.
259 struct kvm_vcpu_arch arch;
262 static inline int kvm_vcpu_exiting_guest_mode(struct kvm_vcpu *vcpu)
264 return cmpxchg(&vcpu->mode, IN_GUEST_MODE, EXITING_GUEST_MODE);
268 * Some of the bitops functions do not support too long bitmaps.
269 * This number must be determined not to exceed such limits.
271 #define KVM_MEM_MAX_NR_PAGES ((1UL << 31) - 1)
273 struct kvm_memory_slot {
275 unsigned long npages;
276 unsigned long *dirty_bitmap;
277 struct kvm_arch_memory_slot arch;
278 unsigned long userspace_addr;
283 static inline unsigned long kvm_dirty_bitmap_bytes(struct kvm_memory_slot *memslot)
285 return ALIGN(memslot->npages, BITS_PER_LONG) / 8;
288 struct kvm_kernel_irq_routing_entry {
291 int (*set)(struct kvm_kernel_irq_routing_entry *e,
292 struct kvm *kvm, int irq_source_id, int level,
301 struct hlist_node link;
304 struct kvm_irq_routing_table;
306 #ifndef KVM_PRIVATE_MEM_SLOTS
307 #define KVM_PRIVATE_MEM_SLOTS 0
310 #ifndef KVM_MEM_SLOTS_NUM
311 #define KVM_MEM_SLOTS_NUM (KVM_USER_MEM_SLOTS + KVM_PRIVATE_MEM_SLOTS)
316 * memslots are not sorted by id anymore, please use id_to_memslot()
317 * to get the memslot by its id.
319 struct kvm_memslots {
321 struct kvm_memory_slot memslots[KVM_MEM_SLOTS_NUM];
322 /* The mapping table from slot id to the index in memslots[]. */
323 short id_to_index[KVM_MEM_SLOTS_NUM];
328 struct mutex slots_lock;
329 struct mm_struct *mm; /* userspace tied to this vm */
330 struct kvm_memslots *memslots;
331 struct srcu_struct srcu;
332 struct srcu_struct irq_srcu;
333 #ifdef CONFIG_KVM_APIC_ARCHITECTURE
336 struct kvm_vcpu *vcpus[KVM_MAX_VCPUS];
337 atomic_t online_vcpus;
338 int last_boosted_vcpu;
339 struct list_head vm_list;
341 struct kvm_io_bus *buses[KVM_NR_BUSES];
342 #ifdef CONFIG_HAVE_KVM_EVENTFD
345 struct list_head items;
346 struct list_head resampler_list;
347 struct mutex resampler_lock;
349 struct list_head ioeventfds;
351 struct kvm_vm_stat stat;
352 struct kvm_arch arch;
353 atomic_t users_count;
354 #ifdef KVM_COALESCED_MMIO_PAGE_OFFSET
355 struct kvm_coalesced_mmio_ring *coalesced_mmio_ring;
356 spinlock_t ring_lock;
357 struct list_head coalesced_zones;
360 struct mutex irq_lock;
361 #ifdef CONFIG_HAVE_KVM_IRQCHIP
363 * Update side is protected by irq_lock.
365 struct kvm_irq_routing_table __rcu *irq_routing;
366 struct hlist_head mask_notifier_list;
368 #ifdef CONFIG_HAVE_KVM_IRQFD
369 struct hlist_head irq_ack_notifier_list;
372 #if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
373 struct mmu_notifier mmu_notifier;
374 unsigned long mmu_notifier_seq;
375 long mmu_notifier_count;
378 struct list_head devices;
381 #define kvm_err(fmt, ...) \
382 pr_err("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
383 #define kvm_info(fmt, ...) \
384 pr_info("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
385 #define kvm_debug(fmt, ...) \
386 pr_debug("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
387 #define kvm_pr_unimpl(fmt, ...) \
388 pr_err_ratelimited("kvm [%i]: " fmt, \
389 task_tgid_nr(current), ## __VA_ARGS__)
391 /* The guest did something we don't support. */
392 #define vcpu_unimpl(vcpu, fmt, ...) \
393 kvm_pr_unimpl("vcpu%i " fmt, (vcpu)->vcpu_id, ## __VA_ARGS__)
395 static inline struct kvm_vcpu *kvm_get_vcpu(struct kvm *kvm, int i)
398 return kvm->vcpus[i];
401 #define kvm_for_each_vcpu(idx, vcpup, kvm) \
403 idx < atomic_read(&kvm->online_vcpus) && \
404 (vcpup = kvm_get_vcpu(kvm, idx)) != NULL; \
407 #define kvm_for_each_memslot(memslot, slots) \
408 for (memslot = &slots->memslots[0]; \
409 memslot < slots->memslots + KVM_MEM_SLOTS_NUM && memslot->npages;\
412 int kvm_vcpu_init(struct kvm_vcpu *vcpu, struct kvm *kvm, unsigned id);
413 void kvm_vcpu_uninit(struct kvm_vcpu *vcpu);
415 int __must_check vcpu_load(struct kvm_vcpu *vcpu);
416 void vcpu_put(struct kvm_vcpu *vcpu);
418 #ifdef CONFIG_HAVE_KVM_IRQFD
419 int kvm_irqfd_init(void);
420 void kvm_irqfd_exit(void);
422 static inline int kvm_irqfd_init(void)
427 static inline void kvm_irqfd_exit(void)
431 int kvm_init(void *opaque, unsigned vcpu_size, unsigned vcpu_align,
432 struct module *module);
435 void kvm_get_kvm(struct kvm *kvm);
436 void kvm_put_kvm(struct kvm *kvm);
438 static inline struct kvm_memslots *kvm_memslots(struct kvm *kvm)
440 return rcu_dereference_check(kvm->memslots,
441 srcu_read_lock_held(&kvm->srcu)
442 || lockdep_is_held(&kvm->slots_lock));
445 static inline struct kvm_memory_slot *
446 id_to_memslot(struct kvm_memslots *slots, int id)
448 int index = slots->id_to_index[id];
449 struct kvm_memory_slot *slot;
451 slot = &slots->memslots[index];
453 WARN_ON(slot->id != id);
458 * KVM_SET_USER_MEMORY_REGION ioctl allows the following operations:
459 * - create a new memory slot
460 * - delete an existing memory slot
461 * - modify an existing memory slot
462 * -- move it in the guest physical memory space
463 * -- just change its flags
465 * Since flags can be changed by some of these operations, the following
466 * differentiation is the best we can do for __kvm_set_memory_region():
475 int kvm_set_memory_region(struct kvm *kvm,
476 struct kvm_userspace_memory_region *mem);
477 int __kvm_set_memory_region(struct kvm *kvm,
478 struct kvm_userspace_memory_region *mem);
479 void kvm_arch_free_memslot(struct kvm *kvm, struct kvm_memory_slot *free,
480 struct kvm_memory_slot *dont);
481 int kvm_arch_create_memslot(struct kvm *kvm, struct kvm_memory_slot *slot,
482 unsigned long npages);
483 void kvm_arch_memslots_updated(struct kvm *kvm);
484 int kvm_arch_prepare_memory_region(struct kvm *kvm,
485 struct kvm_memory_slot *memslot,
486 struct kvm_userspace_memory_region *mem,
487 enum kvm_mr_change change);
488 void kvm_arch_commit_memory_region(struct kvm *kvm,
489 struct kvm_userspace_memory_region *mem,
490 const struct kvm_memory_slot *old,
491 enum kvm_mr_change change);
492 bool kvm_largepages_enabled(void);
493 void kvm_disable_largepages(void);
494 /* flush all memory translations */
495 void kvm_arch_flush_shadow_all(struct kvm *kvm);
496 /* flush memory translations pointing to 'slot' */
497 void kvm_arch_flush_shadow_memslot(struct kvm *kvm,
498 struct kvm_memory_slot *slot);
500 int gfn_to_page_many_atomic(struct kvm *kvm, gfn_t gfn, struct page **pages,
503 struct page *gfn_to_page(struct kvm *kvm, gfn_t gfn);
504 unsigned long gfn_to_hva(struct kvm *kvm, gfn_t gfn);
505 unsigned long gfn_to_hva_prot(struct kvm *kvm, gfn_t gfn, bool *writable);
506 unsigned long gfn_to_hva_memslot(struct kvm_memory_slot *slot, gfn_t gfn);
507 unsigned long gfn_to_hva_memslot_prot(struct kvm_memory_slot *slot, gfn_t gfn,
509 void kvm_release_page_clean(struct page *page);
510 void kvm_release_page_dirty(struct page *page);
511 void kvm_set_page_accessed(struct page *page);
513 pfn_t gfn_to_pfn_atomic(struct kvm *kvm, gfn_t gfn);
514 pfn_t gfn_to_pfn_async(struct kvm *kvm, gfn_t gfn, bool *async,
515 bool write_fault, bool *writable);
516 pfn_t gfn_to_pfn(struct kvm *kvm, gfn_t gfn);
517 pfn_t gfn_to_pfn_prot(struct kvm *kvm, gfn_t gfn, bool write_fault,
519 pfn_t gfn_to_pfn_memslot(struct kvm_memory_slot *slot, gfn_t gfn);
520 pfn_t gfn_to_pfn_memslot_atomic(struct kvm_memory_slot *slot, gfn_t gfn);
522 void kvm_release_pfn_clean(pfn_t pfn);
523 void kvm_set_pfn_dirty(pfn_t pfn);
524 void kvm_set_pfn_accessed(pfn_t pfn);
525 void kvm_get_pfn(pfn_t pfn);
527 int kvm_read_guest_page(struct kvm *kvm, gfn_t gfn, void *data, int offset,
529 int kvm_read_guest_atomic(struct kvm *kvm, gpa_t gpa, void *data,
531 int kvm_read_guest(struct kvm *kvm, gpa_t gpa, void *data, unsigned long len);
532 int kvm_read_guest_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
533 void *data, unsigned long len);
534 int kvm_write_guest_page(struct kvm *kvm, gfn_t gfn, const void *data,
535 int offset, int len);
536 int kvm_write_guest(struct kvm *kvm, gpa_t gpa, const void *data,
538 int kvm_write_guest_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
539 void *data, unsigned long len);
540 int kvm_gfn_to_hva_cache_init(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
541 gpa_t gpa, unsigned long len);
542 int kvm_clear_guest_page(struct kvm *kvm, gfn_t gfn, int offset, int len);
543 int kvm_clear_guest(struct kvm *kvm, gpa_t gpa, unsigned long len);
544 struct kvm_memory_slot *gfn_to_memslot(struct kvm *kvm, gfn_t gfn);
545 int kvm_is_visible_gfn(struct kvm *kvm, gfn_t gfn);
546 unsigned long kvm_host_page_size(struct kvm *kvm, gfn_t gfn);
547 void mark_page_dirty(struct kvm *kvm, gfn_t gfn);
549 void kvm_vcpu_block(struct kvm_vcpu *vcpu);
550 void kvm_vcpu_kick(struct kvm_vcpu *vcpu);
551 bool kvm_vcpu_yield_to(struct kvm_vcpu *target);
552 void kvm_vcpu_on_spin(struct kvm_vcpu *vcpu);
553 void kvm_load_guest_fpu(struct kvm_vcpu *vcpu);
554 void kvm_put_guest_fpu(struct kvm_vcpu *vcpu);
556 void kvm_flush_remote_tlbs(struct kvm *kvm);
557 void kvm_reload_remote_mmus(struct kvm *kvm);
558 void kvm_make_mclock_inprogress_request(struct kvm *kvm);
559 void kvm_make_scan_ioapic_request(struct kvm *kvm);
561 long kvm_arch_dev_ioctl(struct file *filp,
562 unsigned int ioctl, unsigned long arg);
563 long kvm_arch_vcpu_ioctl(struct file *filp,
564 unsigned int ioctl, unsigned long arg);
565 int kvm_arch_vcpu_fault(struct kvm_vcpu *vcpu, struct vm_fault *vmf);
567 int kvm_vm_ioctl_check_extension(struct kvm *kvm, long ext);
569 int kvm_get_dirty_log(struct kvm *kvm,
570 struct kvm_dirty_log *log, int *is_dirty);
571 int kvm_vm_ioctl_get_dirty_log(struct kvm *kvm,
572 struct kvm_dirty_log *log);
574 int kvm_vm_ioctl_irq_line(struct kvm *kvm, struct kvm_irq_level *irq_level,
576 long kvm_arch_vm_ioctl(struct file *filp,
577 unsigned int ioctl, unsigned long arg);
579 int kvm_arch_vcpu_ioctl_get_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu);
580 int kvm_arch_vcpu_ioctl_set_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu);
582 int kvm_arch_vcpu_ioctl_translate(struct kvm_vcpu *vcpu,
583 struct kvm_translation *tr);
585 int kvm_arch_vcpu_ioctl_get_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs);
586 int kvm_arch_vcpu_ioctl_set_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs);
587 int kvm_arch_vcpu_ioctl_get_sregs(struct kvm_vcpu *vcpu,
588 struct kvm_sregs *sregs);
589 int kvm_arch_vcpu_ioctl_set_sregs(struct kvm_vcpu *vcpu,
590 struct kvm_sregs *sregs);
591 int kvm_arch_vcpu_ioctl_get_mpstate(struct kvm_vcpu *vcpu,
592 struct kvm_mp_state *mp_state);
593 int kvm_arch_vcpu_ioctl_set_mpstate(struct kvm_vcpu *vcpu,
594 struct kvm_mp_state *mp_state);
595 int kvm_arch_vcpu_ioctl_set_guest_debug(struct kvm_vcpu *vcpu,
596 struct kvm_guest_debug *dbg);
597 int kvm_arch_vcpu_ioctl_run(struct kvm_vcpu *vcpu, struct kvm_run *kvm_run);
599 int kvm_arch_init(void *opaque);
600 void kvm_arch_exit(void);
602 int kvm_arch_vcpu_init(struct kvm_vcpu *vcpu);
603 void kvm_arch_vcpu_uninit(struct kvm_vcpu *vcpu);
605 void kvm_arch_sched_in(struct kvm_vcpu *vcpu, int cpu);
607 void kvm_arch_vcpu_free(struct kvm_vcpu *vcpu);
608 void kvm_arch_vcpu_load(struct kvm_vcpu *vcpu, int cpu);
609 void kvm_arch_vcpu_put(struct kvm_vcpu *vcpu);
610 struct kvm_vcpu *kvm_arch_vcpu_create(struct kvm *kvm, unsigned int id);
611 int kvm_arch_vcpu_setup(struct kvm_vcpu *vcpu);
612 int kvm_arch_vcpu_postcreate(struct kvm_vcpu *vcpu);
613 void kvm_arch_vcpu_destroy(struct kvm_vcpu *vcpu);
615 int kvm_arch_hardware_enable(void *garbage);
616 void kvm_arch_hardware_disable(void *garbage);
617 int kvm_arch_hardware_setup(void);
618 void kvm_arch_hardware_unsetup(void);
619 void kvm_arch_check_processor_compat(void *rtn);
620 int kvm_arch_vcpu_runnable(struct kvm_vcpu *vcpu);
621 int kvm_arch_vcpu_should_kick(struct kvm_vcpu *vcpu);
623 void *kvm_kvzalloc(unsigned long size);
624 void kvm_kvfree(const void *addr);
626 #ifndef __KVM_HAVE_ARCH_VM_ALLOC
627 static inline struct kvm *kvm_arch_alloc_vm(void)
629 return kzalloc(sizeof(struct kvm), GFP_KERNEL);
632 static inline void kvm_arch_free_vm(struct kvm *kvm)
638 static inline wait_queue_head_t *kvm_arch_vcpu_wq(struct kvm_vcpu *vcpu)
640 #ifdef __KVM_HAVE_ARCH_WQP
641 return vcpu->arch.wqp;
647 int kvm_arch_init_vm(struct kvm *kvm, unsigned long type);
648 void kvm_arch_destroy_vm(struct kvm *kvm);
649 void kvm_arch_sync_events(struct kvm *kvm);
651 int kvm_cpu_has_pending_timer(struct kvm_vcpu *vcpu);
652 void kvm_vcpu_kick(struct kvm_vcpu *vcpu);
654 bool kvm_is_mmio_pfn(pfn_t pfn);
656 struct kvm_irq_ack_notifier {
657 struct hlist_node link;
659 void (*irq_acked)(struct kvm_irq_ack_notifier *kian);
662 struct kvm_assigned_dev_kernel {
663 struct kvm_irq_ack_notifier ack_notifier;
664 struct list_head list;
669 unsigned int entries_nr;
671 bool host_irq_disabled;
673 struct msix_entry *host_msix_entries;
675 struct msix_entry *guest_msix_entries;
676 unsigned long irq_requested_type;
681 spinlock_t intx_lock;
682 spinlock_t intx_mask_lock;
684 struct pci_saved_state *pci_saved_state;
687 struct kvm_irq_mask_notifier {
688 void (*func)(struct kvm_irq_mask_notifier *kimn, bool masked);
690 struct hlist_node link;
693 void kvm_register_irq_mask_notifier(struct kvm *kvm, int irq,
694 struct kvm_irq_mask_notifier *kimn);
695 void kvm_unregister_irq_mask_notifier(struct kvm *kvm, int irq,
696 struct kvm_irq_mask_notifier *kimn);
697 void kvm_fire_mask_notifiers(struct kvm *kvm, unsigned irqchip, unsigned pin,
700 int kvm_irq_map_gsi(struct kvm *kvm,
701 struct kvm_kernel_irq_routing_entry *entries, int gsi);
702 int kvm_irq_map_chip_pin(struct kvm *kvm, unsigned irqchip, unsigned pin);
704 int kvm_set_irq(struct kvm *kvm, int irq_source_id, u32 irq, int level,
706 int kvm_set_irq_inatomic(struct kvm *kvm, int irq_source_id, u32 irq, int level);
707 int kvm_set_msi(struct kvm_kernel_irq_routing_entry *irq_entry, struct kvm *kvm,
708 int irq_source_id, int level, bool line_status);
709 bool kvm_irq_has_notifier(struct kvm *kvm, unsigned irqchip, unsigned pin);
710 void kvm_notify_acked_irq(struct kvm *kvm, unsigned irqchip, unsigned pin);
711 void kvm_register_irq_ack_notifier(struct kvm *kvm,
712 struct kvm_irq_ack_notifier *kian);
713 void kvm_unregister_irq_ack_notifier(struct kvm *kvm,
714 struct kvm_irq_ack_notifier *kian);
715 int kvm_request_irq_source_id(struct kvm *kvm);
716 void kvm_free_irq_source_id(struct kvm *kvm, int irq_source_id);
718 /* For vcpu->arch.iommu_flags */
719 #define KVM_IOMMU_CACHE_COHERENCY 0x1
721 #ifdef CONFIG_KVM_DEVICE_ASSIGNMENT
722 int kvm_iommu_map_pages(struct kvm *kvm, struct kvm_memory_slot *slot);
723 void kvm_iommu_unmap_pages(struct kvm *kvm, struct kvm_memory_slot *slot);
724 int kvm_iommu_map_guest(struct kvm *kvm);
725 int kvm_iommu_unmap_guest(struct kvm *kvm);
726 int kvm_assign_device(struct kvm *kvm,
727 struct kvm_assigned_dev_kernel *assigned_dev);
728 int kvm_deassign_device(struct kvm *kvm,
729 struct kvm_assigned_dev_kernel *assigned_dev);
731 static inline int kvm_iommu_map_pages(struct kvm *kvm,
732 struct kvm_memory_slot *slot)
737 static inline void kvm_iommu_unmap_pages(struct kvm *kvm,
738 struct kvm_memory_slot *slot)
742 static inline int kvm_iommu_unmap_guest(struct kvm *kvm)
748 static inline void kvm_guest_enter(void)
752 BUG_ON(preemptible());
754 local_irq_save(flags);
756 local_irq_restore(flags);
758 /* KVM does not hold any references to rcu protected data when it
759 * switches CPU into a guest mode. In fact switching to a guest mode
760 * is very similar to exiting to userspace from rcu point of view. In
761 * addition CPU may stay in a guest mode for quite a long time (up to
762 * one time slice). Lets treat guest mode as quiescent state, just like
763 * we do with user-mode execution.
765 rcu_virt_note_context_switch(smp_processor_id());
768 static inline void kvm_guest_exit(void)
772 local_irq_save(flags);
774 local_irq_restore(flags);
778 * search_memslots() and __gfn_to_memslot() are here because they are
779 * used in non-modular code in arch/powerpc/kvm/book3s_hv_rm_mmu.c.
780 * gfn_to_memslot() itself isn't here as an inline because that would
781 * bloat other code too much.
783 static inline struct kvm_memory_slot *
784 search_memslots(struct kvm_memslots *slots, gfn_t gfn)
786 struct kvm_memory_slot *memslot;
788 kvm_for_each_memslot(memslot, slots)
789 if (gfn >= memslot->base_gfn &&
790 gfn < memslot->base_gfn + memslot->npages)
796 static inline struct kvm_memory_slot *
797 __gfn_to_memslot(struct kvm_memslots *slots, gfn_t gfn)
799 return search_memslots(slots, gfn);
802 static inline unsigned long
803 __gfn_to_hva_memslot(struct kvm_memory_slot *slot, gfn_t gfn)
805 return slot->userspace_addr + (gfn - slot->base_gfn) * PAGE_SIZE;
808 static inline int memslot_id(struct kvm *kvm, gfn_t gfn)
810 return gfn_to_memslot(kvm, gfn)->id;
814 hva_to_gfn_memslot(unsigned long hva, struct kvm_memory_slot *slot)
816 gfn_t gfn_offset = (hva - slot->userspace_addr) >> PAGE_SHIFT;
818 return slot->base_gfn + gfn_offset;
821 static inline gpa_t gfn_to_gpa(gfn_t gfn)
823 return (gpa_t)gfn << PAGE_SHIFT;
826 static inline gfn_t gpa_to_gfn(gpa_t gpa)
828 return (gfn_t)(gpa >> PAGE_SHIFT);
831 static inline hpa_t pfn_to_hpa(pfn_t pfn)
833 return (hpa_t)pfn << PAGE_SHIFT;
836 static inline bool kvm_is_error_gpa(struct kvm *kvm, gpa_t gpa)
838 unsigned long hva = gfn_to_hva(kvm, gpa_to_gfn(gpa));
840 return kvm_is_error_hva(hva);
843 static inline void kvm_migrate_timers(struct kvm_vcpu *vcpu)
845 set_bit(KVM_REQ_MIGRATE_TIMER, &vcpu->requests);
853 struct kvm_stats_debugfs_item {
856 enum kvm_stat_kind kind;
857 struct dentry *dentry;
859 extern struct kvm_stats_debugfs_item debugfs_entries[];
860 extern struct dentry *kvm_debugfs_dir;
862 #if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
863 static inline int mmu_notifier_retry(struct kvm *kvm, unsigned long mmu_seq)
865 if (unlikely(kvm->mmu_notifier_count))
868 * Ensure the read of mmu_notifier_count happens before the read
869 * of mmu_notifier_seq. This interacts with the smp_wmb() in
870 * mmu_notifier_invalidate_range_end to make sure that the caller
871 * either sees the old (non-zero) value of mmu_notifier_count or
872 * the new (incremented) value of mmu_notifier_seq.
873 * PowerPC Book3s HV KVM calls this under a per-page lock
874 * rather than under kvm->mmu_lock, for scalability, so
875 * can't rely on kvm->mmu_lock to keep things ordered.
878 if (kvm->mmu_notifier_seq != mmu_seq)
884 #ifdef CONFIG_HAVE_KVM_IRQ_ROUTING
886 #define KVM_MAX_IRQ_ROUTES 1024
888 int kvm_setup_default_irq_routing(struct kvm *kvm);
889 int kvm_set_irq_routing(struct kvm *kvm,
890 const struct kvm_irq_routing_entry *entries,
893 int kvm_set_routing_entry(struct kvm_kernel_irq_routing_entry *e,
894 const struct kvm_irq_routing_entry *ue);
895 void kvm_free_irq_routing(struct kvm *kvm);
899 static inline void kvm_free_irq_routing(struct kvm *kvm) {}
903 int kvm_send_userspace_msi(struct kvm *kvm, struct kvm_msi *msi);
905 #ifdef CONFIG_HAVE_KVM_EVENTFD
907 void kvm_eventfd_init(struct kvm *kvm);
908 int kvm_ioeventfd(struct kvm *kvm, struct kvm_ioeventfd *args);
910 #ifdef CONFIG_HAVE_KVM_IRQFD
911 int kvm_irqfd(struct kvm *kvm, struct kvm_irqfd *args);
912 void kvm_irqfd_release(struct kvm *kvm);
913 void kvm_irq_routing_update(struct kvm *);
915 static inline int kvm_irqfd(struct kvm *kvm, struct kvm_irqfd *args)
920 static inline void kvm_irqfd_release(struct kvm *kvm) {}
925 static inline void kvm_eventfd_init(struct kvm *kvm) {}
927 static inline int kvm_irqfd(struct kvm *kvm, struct kvm_irqfd *args)
932 static inline void kvm_irqfd_release(struct kvm *kvm) {}
934 #ifdef CONFIG_HAVE_KVM_IRQCHIP
935 static inline void kvm_irq_routing_update(struct kvm *kvm)
940 static inline int kvm_ioeventfd(struct kvm *kvm, struct kvm_ioeventfd *args)
945 #endif /* CONFIG_HAVE_KVM_EVENTFD */
947 #ifdef CONFIG_KVM_APIC_ARCHITECTURE
948 static inline bool kvm_vcpu_is_bsp(struct kvm_vcpu *vcpu)
950 return vcpu->kvm->bsp_vcpu_id == vcpu->vcpu_id;
953 bool kvm_vcpu_compatible(struct kvm_vcpu *vcpu);
957 static inline bool kvm_vcpu_compatible(struct kvm_vcpu *vcpu) { return true; }
961 #ifdef CONFIG_KVM_DEVICE_ASSIGNMENT
963 long kvm_vm_ioctl_assigned_device(struct kvm *kvm, unsigned ioctl,
966 void kvm_free_all_assigned_devices(struct kvm *kvm);
970 static inline long kvm_vm_ioctl_assigned_device(struct kvm *kvm, unsigned ioctl,
976 static inline void kvm_free_all_assigned_devices(struct kvm *kvm) {}
980 static inline void kvm_make_request(int req, struct kvm_vcpu *vcpu)
982 set_bit(req, &vcpu->requests);
985 static inline bool kvm_check_request(int req, struct kvm_vcpu *vcpu)
987 if (test_bit(req, &vcpu->requests)) {
988 clear_bit(req, &vcpu->requests);
995 extern bool kvm_rebooting;
997 struct kvm_device_ops;
1000 struct kvm_device_ops *ops;
1003 struct list_head vm_node;
1006 /* create, destroy, and name are mandatory */
1007 struct kvm_device_ops {
1009 int (*create)(struct kvm_device *dev, u32 type);
1012 * Destroy is responsible for freeing dev.
1014 * Destroy may be called before or after destructors are called
1015 * on emulated I/O regions, depending on whether a reference is
1016 * held by a vcpu or other kvm component that gets destroyed
1017 * after the emulated I/O.
1019 void (*destroy)(struct kvm_device *dev);
1021 int (*set_attr)(struct kvm_device *dev, struct kvm_device_attr *attr);
1022 int (*get_attr)(struct kvm_device *dev, struct kvm_device_attr *attr);
1023 int (*has_attr)(struct kvm_device *dev, struct kvm_device_attr *attr);
1024 long (*ioctl)(struct kvm_device *dev, unsigned int ioctl,
1028 void kvm_device_get(struct kvm_device *dev);
1029 void kvm_device_put(struct kvm_device *dev);
1030 struct kvm_device *kvm_device_from_filp(struct file *filp);
1032 extern struct kvm_device_ops kvm_mpic_ops;
1033 extern struct kvm_device_ops kvm_xics_ops;
1034 extern struct kvm_device_ops kvm_vfio_ops;
1035 extern struct kvm_device_ops kvm_arm_vgic_v2_ops;
1037 #ifdef CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT
1039 static inline void kvm_vcpu_set_in_spin_loop(struct kvm_vcpu *vcpu, bool val)
1041 vcpu->spin_loop.in_spin_loop = val;
1043 static inline void kvm_vcpu_set_dy_eligible(struct kvm_vcpu *vcpu, bool val)
1045 vcpu->spin_loop.dy_eligible = val;
1048 #else /* !CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT */
1050 static inline void kvm_vcpu_set_in_spin_loop(struct kvm_vcpu *vcpu, bool val)
1054 static inline void kvm_vcpu_set_dy_eligible(struct kvm_vcpu *vcpu, bool val)
1057 #endif /* CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT */