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 <asm/signal.h>
27 #include <linux/kvm.h>
28 #include <linux/kvm_para.h>
30 #include <linux/kvm_types.h>
32 #include <asm/kvm_host.h>
35 #define KVM_MMIO_SIZE 8
39 * The bit 16 ~ bit 31 of kvm_memory_region::flags are internally used
40 * in kvm, other bits are visible for userspace which are defined in
41 * include/linux/kvm_h.
43 #define KVM_MEMSLOT_INVALID (1UL << 16)
46 * If we support unaligned MMIO, at most one fragment will be split into two:
48 #ifdef KVM_UNALIGNED_MMIO
49 # define KVM_EXTRA_MMIO_FRAGMENTS 1
51 # define KVM_EXTRA_MMIO_FRAGMENTS 0
54 #define KVM_USER_MMIO_SIZE 8
56 #define KVM_MAX_MMIO_FRAGMENTS \
57 (KVM_MMIO_SIZE / KVM_USER_MMIO_SIZE + KVM_EXTRA_MMIO_FRAGMENTS)
60 * For the normal pfn, the highest 12 bits should be zero,
61 * so we can mask these bits to indicate the error.
63 #define KVM_PFN_ERR_MASK (0xfffULL << 52)
65 #define KVM_PFN_ERR_FAULT (KVM_PFN_ERR_MASK)
66 #define KVM_PFN_ERR_HWPOISON (KVM_PFN_ERR_MASK + 1)
67 #define KVM_PFN_ERR_BAD (KVM_PFN_ERR_MASK + 2)
68 #define KVM_PFN_ERR_RO_FAULT (KVM_PFN_ERR_MASK + 3)
70 static inline bool is_error_pfn(pfn_t pfn)
72 return !!(pfn & KVM_PFN_ERR_MASK);
75 static inline bool is_noslot_pfn(pfn_t pfn)
77 return pfn == KVM_PFN_ERR_BAD;
80 static inline bool is_invalid_pfn(pfn_t pfn)
82 return !is_noslot_pfn(pfn) && is_error_pfn(pfn);
85 #define KVM_HVA_ERR_BAD (PAGE_OFFSET)
86 #define KVM_HVA_ERR_RO_BAD (PAGE_OFFSET + PAGE_SIZE)
88 static inline bool kvm_is_error_hva(unsigned long addr)
90 return addr >= PAGE_OFFSET;
93 #define KVM_ERR_PTR_BAD_PAGE (ERR_PTR(-ENOENT))
95 static inline bool is_error_page(struct page *page)
101 * vcpu->requests bit members
103 #define KVM_REQ_TLB_FLUSH 0
104 #define KVM_REQ_MIGRATE_TIMER 1
105 #define KVM_REQ_REPORT_TPR_ACCESS 2
106 #define KVM_REQ_MMU_RELOAD 3
107 #define KVM_REQ_TRIPLE_FAULT 4
108 #define KVM_REQ_PENDING_TIMER 5
109 #define KVM_REQ_UNHALT 6
110 #define KVM_REQ_MMU_SYNC 7
111 #define KVM_REQ_CLOCK_UPDATE 8
112 #define KVM_REQ_KICK 9
113 #define KVM_REQ_DEACTIVATE_FPU 10
114 #define KVM_REQ_EVENT 11
115 #define KVM_REQ_APF_HALT 12
116 #define KVM_REQ_STEAL_UPDATE 13
117 #define KVM_REQ_NMI 14
118 #define KVM_REQ_IMMEDIATE_EXIT 15
119 #define KVM_REQ_PMU 16
120 #define KVM_REQ_PMI 17
121 #define KVM_REQ_WATCHDOG 18
123 #define KVM_USERSPACE_IRQ_SOURCE_ID 0
124 #define KVM_IRQFD_RESAMPLE_IRQ_SOURCE_ID 1
128 extern struct kmem_cache *kvm_vcpu_cache;
130 struct kvm_io_range {
133 struct kvm_io_device *dev;
136 #define NR_IOBUS_DEVS 1000
140 struct kvm_io_range range[];
149 int kvm_io_bus_write(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr,
150 int len, const void *val);
151 int kvm_io_bus_read(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr, int len,
153 int kvm_io_bus_register_dev(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr,
154 int len, struct kvm_io_device *dev);
155 int kvm_io_bus_unregister_dev(struct kvm *kvm, enum kvm_bus bus_idx,
156 struct kvm_io_device *dev);
158 #ifdef CONFIG_KVM_ASYNC_PF
159 struct kvm_async_pf {
160 struct work_struct work;
161 struct list_head link;
162 struct list_head queue;
163 struct kvm_vcpu *vcpu;
164 struct mm_struct *mm;
167 struct kvm_arch_async_pf arch;
172 void kvm_clear_async_pf_completion_queue(struct kvm_vcpu *vcpu);
173 void kvm_check_async_pf_completion(struct kvm_vcpu *vcpu);
174 int kvm_setup_async_pf(struct kvm_vcpu *vcpu, gva_t gva, gfn_t gfn,
175 struct kvm_arch_async_pf *arch);
176 int kvm_async_pf_wakeup_all(struct kvm_vcpu *vcpu);
183 READING_SHADOW_PAGE_TABLES,
187 * Sometimes a large or cross-page mmio needs to be broken up into separate
188 * exits for userspace servicing.
190 struct kvm_mmio_fragment {
198 #ifdef CONFIG_PREEMPT_NOTIFIERS
199 struct preempt_notifier preempt_notifier;
205 unsigned long requests;
206 unsigned long guest_debug;
212 int guest_fpu_loaded, guest_xcr0_loaded;
213 wait_queue_head_t wq;
217 struct kvm_vcpu_stat stat;
219 #ifdef CONFIG_HAS_IOMEM
221 int mmio_read_completed;
223 int mmio_cur_fragment;
224 int mmio_nr_fragments;
225 struct kvm_mmio_fragment mmio_fragments[KVM_MAX_MMIO_FRAGMENTS];
228 #ifdef CONFIG_KVM_ASYNC_PF
231 struct list_head queue;
232 struct list_head done;
237 #ifdef CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT
239 * Cpu relax intercept or pause loop exit optimization
240 * in_spin_loop: set when a vcpu does a pause loop exit
241 * or cpu relax intercepted.
242 * dy_eligible: indicates whether vcpu is eligible for directed yield.
249 struct kvm_vcpu_arch arch;
252 static inline int kvm_vcpu_exiting_guest_mode(struct kvm_vcpu *vcpu)
254 return cmpxchg(&vcpu->mode, IN_GUEST_MODE, EXITING_GUEST_MODE);
258 * Some of the bitops functions do not support too long bitmaps.
259 * This number must be determined not to exceed such limits.
261 #define KVM_MEM_MAX_NR_PAGES ((1UL << 31) - 1)
263 struct kvm_memory_slot {
265 unsigned long npages;
267 unsigned long *dirty_bitmap;
268 struct kvm_arch_memory_slot arch;
269 unsigned long userspace_addr;
274 static inline unsigned long kvm_dirty_bitmap_bytes(struct kvm_memory_slot *memslot)
276 return ALIGN(memslot->npages, BITS_PER_LONG) / 8;
279 struct kvm_kernel_irq_routing_entry {
282 int (*set)(struct kvm_kernel_irq_routing_entry *e,
283 struct kvm *kvm, int irq_source_id, int level);
291 struct hlist_node link;
294 #ifdef __KVM_HAVE_IOAPIC
296 struct kvm_irq_routing_table {
297 int chip[KVM_NR_IRQCHIPS][KVM_IOAPIC_NUM_PINS];
298 struct kvm_kernel_irq_routing_entry *rt_entries;
301 * Array indexed by gsi. Each entry contains list of irq chips
302 * the gsi is connected to.
304 struct hlist_head map[0];
309 struct kvm_irq_routing_table {};
313 #ifndef KVM_MEM_SLOTS_NUM
314 #define KVM_MEM_SLOTS_NUM (KVM_MEMORY_SLOTS + KVM_PRIVATE_MEM_SLOTS)
319 * memslots are not sorted by id anymore, please use id_to_memslot()
320 * to get the memslot by its id.
322 struct kvm_memslots {
324 struct kvm_memory_slot memslots[KVM_MEM_SLOTS_NUM];
325 /* The mapping table from slot id to the index in memslots[]. */
326 int id_to_index[KVM_MEM_SLOTS_NUM];
331 struct mutex slots_lock;
332 struct mm_struct *mm; /* userspace tied to this vm */
333 struct kvm_memslots *memslots;
334 struct srcu_struct srcu;
335 #ifdef CONFIG_KVM_APIC_ARCHITECTURE
338 struct kvm_vcpu *vcpus[KVM_MAX_VCPUS];
339 atomic_t online_vcpus;
340 int last_boosted_vcpu;
341 struct list_head vm_list;
343 struct kvm_io_bus *buses[KVM_NR_BUSES];
344 #ifdef CONFIG_HAVE_KVM_EVENTFD
347 struct list_head items;
348 struct list_head resampler_list;
349 struct mutex resampler_lock;
351 struct list_head ioeventfds;
353 struct kvm_vm_stat stat;
354 struct kvm_arch arch;
355 atomic_t users_count;
356 #ifdef KVM_COALESCED_MMIO_PAGE_OFFSET
357 struct kvm_coalesced_mmio_ring *coalesced_mmio_ring;
358 spinlock_t ring_lock;
359 struct list_head coalesced_zones;
362 struct mutex irq_lock;
363 #ifdef CONFIG_HAVE_KVM_IRQCHIP
365 * Update side is protected by irq_lock and,
366 * if configured, irqfds.lock.
368 struct kvm_irq_routing_table __rcu *irq_routing;
369 struct hlist_head mask_notifier_list;
370 struct hlist_head irq_ack_notifier_list;
373 #if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
374 struct mmu_notifier mmu_notifier;
375 unsigned long mmu_notifier_seq;
376 long mmu_notifier_count;
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 int kvm_init(void *opaque, unsigned vcpu_size, unsigned vcpu_align,
419 struct module *module);
422 void kvm_get_kvm(struct kvm *kvm);
423 void kvm_put_kvm(struct kvm *kvm);
424 void update_memslots(struct kvm_memslots *slots, struct kvm_memory_slot *new);
426 static inline struct kvm_memslots *kvm_memslots(struct kvm *kvm)
428 return rcu_dereference_check(kvm->memslots,
429 srcu_read_lock_held(&kvm->srcu)
430 || lockdep_is_held(&kvm->slots_lock));
433 static inline struct kvm_memory_slot *
434 id_to_memslot(struct kvm_memslots *slots, int id)
436 int index = slots->id_to_index[id];
437 struct kvm_memory_slot *slot;
439 slot = &slots->memslots[index];
441 WARN_ON(slot->id != id);
445 int kvm_set_memory_region(struct kvm *kvm,
446 struct kvm_userspace_memory_region *mem,
448 int __kvm_set_memory_region(struct kvm *kvm,
449 struct kvm_userspace_memory_region *mem,
451 void kvm_arch_free_memslot(struct kvm_memory_slot *free,
452 struct kvm_memory_slot *dont);
453 int kvm_arch_create_memslot(struct kvm_memory_slot *slot, unsigned long npages);
454 int kvm_arch_prepare_memory_region(struct kvm *kvm,
455 struct kvm_memory_slot *memslot,
456 struct kvm_memory_slot old,
457 struct kvm_userspace_memory_region *mem,
459 void kvm_arch_commit_memory_region(struct kvm *kvm,
460 struct kvm_userspace_memory_region *mem,
461 struct kvm_memory_slot old,
463 bool kvm_largepages_enabled(void);
464 void kvm_disable_largepages(void);
465 /* flush all memory translations */
466 void kvm_arch_flush_shadow_all(struct kvm *kvm);
467 /* flush memory translations pointing to 'slot' */
468 void kvm_arch_flush_shadow_memslot(struct kvm *kvm,
469 struct kvm_memory_slot *slot);
471 int gfn_to_page_many_atomic(struct kvm *kvm, gfn_t gfn, struct page **pages,
474 struct page *gfn_to_page(struct kvm *kvm, gfn_t gfn);
475 unsigned long gfn_to_hva(struct kvm *kvm, gfn_t gfn);
476 unsigned long gfn_to_hva_memslot(struct kvm_memory_slot *slot, gfn_t gfn);
477 void kvm_release_page_clean(struct page *page);
478 void kvm_release_page_dirty(struct page *page);
479 void kvm_set_page_dirty(struct page *page);
480 void kvm_set_page_accessed(struct page *page);
482 pfn_t gfn_to_pfn_atomic(struct kvm *kvm, gfn_t gfn);
483 pfn_t gfn_to_pfn_async(struct kvm *kvm, gfn_t gfn, bool *async,
484 bool write_fault, bool *writable);
485 pfn_t gfn_to_pfn(struct kvm *kvm, gfn_t gfn);
486 pfn_t gfn_to_pfn_prot(struct kvm *kvm, gfn_t gfn, bool write_fault,
488 pfn_t gfn_to_pfn_memslot(struct kvm_memory_slot *slot, gfn_t gfn);
489 pfn_t gfn_to_pfn_memslot_atomic(struct kvm_memory_slot *slot, gfn_t gfn);
491 void kvm_release_pfn_dirty(pfn_t pfn);
492 void kvm_release_pfn_clean(pfn_t pfn);
493 void kvm_set_pfn_dirty(pfn_t pfn);
494 void kvm_set_pfn_accessed(pfn_t pfn);
495 void kvm_get_pfn(pfn_t pfn);
497 int kvm_read_guest_page(struct kvm *kvm, gfn_t gfn, void *data, int offset,
499 int kvm_read_guest_atomic(struct kvm *kvm, gpa_t gpa, void *data,
501 int kvm_read_guest(struct kvm *kvm, gpa_t gpa, void *data, unsigned long len);
502 int kvm_read_guest_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
503 void *data, unsigned long len);
504 int kvm_write_guest_page(struct kvm *kvm, gfn_t gfn, const void *data,
505 int offset, int len);
506 int kvm_write_guest(struct kvm *kvm, gpa_t gpa, const void *data,
508 int kvm_write_guest_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
509 void *data, unsigned long len);
510 int kvm_gfn_to_hva_cache_init(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
512 int kvm_clear_guest_page(struct kvm *kvm, gfn_t gfn, int offset, int len);
513 int kvm_clear_guest(struct kvm *kvm, gpa_t gpa, unsigned long len);
514 struct kvm_memory_slot *gfn_to_memslot(struct kvm *kvm, gfn_t gfn);
515 int kvm_is_visible_gfn(struct kvm *kvm, gfn_t gfn);
516 unsigned long kvm_host_page_size(struct kvm *kvm, gfn_t gfn);
517 void mark_page_dirty(struct kvm *kvm, gfn_t gfn);
518 void mark_page_dirty_in_slot(struct kvm *kvm, struct kvm_memory_slot *memslot,
521 void kvm_vcpu_block(struct kvm_vcpu *vcpu);
522 void kvm_vcpu_kick(struct kvm_vcpu *vcpu);
523 bool kvm_vcpu_yield_to(struct kvm_vcpu *target);
524 void kvm_vcpu_on_spin(struct kvm_vcpu *vcpu);
525 void kvm_resched(struct kvm_vcpu *vcpu);
526 void kvm_load_guest_fpu(struct kvm_vcpu *vcpu);
527 void kvm_put_guest_fpu(struct kvm_vcpu *vcpu);
529 void kvm_flush_remote_tlbs(struct kvm *kvm);
530 void kvm_reload_remote_mmus(struct kvm *kvm);
532 long kvm_arch_dev_ioctl(struct file *filp,
533 unsigned int ioctl, unsigned long arg);
534 long kvm_arch_vcpu_ioctl(struct file *filp,
535 unsigned int ioctl, unsigned long arg);
536 int kvm_arch_vcpu_fault(struct kvm_vcpu *vcpu, struct vm_fault *vmf);
538 int kvm_dev_ioctl_check_extension(long ext);
540 int kvm_get_dirty_log(struct kvm *kvm,
541 struct kvm_dirty_log *log, int *is_dirty);
542 int kvm_vm_ioctl_get_dirty_log(struct kvm *kvm,
543 struct kvm_dirty_log *log);
545 int kvm_vm_ioctl_set_memory_region(struct kvm *kvm,
547 kvm_userspace_memory_region *mem,
549 int kvm_vm_ioctl_irq_line(struct kvm *kvm, struct kvm_irq_level *irq_level);
550 long kvm_arch_vm_ioctl(struct file *filp,
551 unsigned int ioctl, unsigned long arg);
553 int kvm_arch_vcpu_ioctl_get_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu);
554 int kvm_arch_vcpu_ioctl_set_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu);
556 int kvm_arch_vcpu_ioctl_translate(struct kvm_vcpu *vcpu,
557 struct kvm_translation *tr);
559 int kvm_arch_vcpu_ioctl_get_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs);
560 int kvm_arch_vcpu_ioctl_set_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs);
561 int kvm_arch_vcpu_ioctl_get_sregs(struct kvm_vcpu *vcpu,
562 struct kvm_sregs *sregs);
563 int kvm_arch_vcpu_ioctl_set_sregs(struct kvm_vcpu *vcpu,
564 struct kvm_sregs *sregs);
565 int kvm_arch_vcpu_ioctl_get_mpstate(struct kvm_vcpu *vcpu,
566 struct kvm_mp_state *mp_state);
567 int kvm_arch_vcpu_ioctl_set_mpstate(struct kvm_vcpu *vcpu,
568 struct kvm_mp_state *mp_state);
569 int kvm_arch_vcpu_ioctl_set_guest_debug(struct kvm_vcpu *vcpu,
570 struct kvm_guest_debug *dbg);
571 int kvm_arch_vcpu_ioctl_run(struct kvm_vcpu *vcpu, struct kvm_run *kvm_run);
573 int kvm_arch_init(void *opaque);
574 void kvm_arch_exit(void);
576 int kvm_arch_vcpu_init(struct kvm_vcpu *vcpu);
577 void kvm_arch_vcpu_uninit(struct kvm_vcpu *vcpu);
579 void kvm_arch_vcpu_free(struct kvm_vcpu *vcpu);
580 void kvm_arch_vcpu_load(struct kvm_vcpu *vcpu, int cpu);
581 void kvm_arch_vcpu_put(struct kvm_vcpu *vcpu);
582 struct kvm_vcpu *kvm_arch_vcpu_create(struct kvm *kvm, unsigned int id);
583 int kvm_arch_vcpu_setup(struct kvm_vcpu *vcpu);
584 void kvm_arch_vcpu_destroy(struct kvm_vcpu *vcpu);
586 int kvm_arch_vcpu_reset(struct kvm_vcpu *vcpu);
587 int kvm_arch_hardware_enable(void *garbage);
588 void kvm_arch_hardware_disable(void *garbage);
589 int kvm_arch_hardware_setup(void);
590 void kvm_arch_hardware_unsetup(void);
591 void kvm_arch_check_processor_compat(void *rtn);
592 int kvm_arch_vcpu_runnable(struct kvm_vcpu *vcpu);
593 int kvm_arch_vcpu_should_kick(struct kvm_vcpu *vcpu);
595 void kvm_free_physmem(struct kvm *kvm);
597 void *kvm_kvzalloc(unsigned long size);
598 void kvm_kvfree(const void *addr);
600 #ifndef __KVM_HAVE_ARCH_VM_ALLOC
601 static inline struct kvm *kvm_arch_alloc_vm(void)
603 return kzalloc(sizeof(struct kvm), GFP_KERNEL);
606 static inline void kvm_arch_free_vm(struct kvm *kvm)
612 static inline wait_queue_head_t *kvm_arch_vcpu_wq(struct kvm_vcpu *vcpu)
614 #ifdef __KVM_HAVE_ARCH_WQP
615 return vcpu->arch.wqp;
621 int kvm_arch_init_vm(struct kvm *kvm, unsigned long type);
622 void kvm_arch_destroy_vm(struct kvm *kvm);
623 void kvm_free_all_assigned_devices(struct kvm *kvm);
624 void kvm_arch_sync_events(struct kvm *kvm);
626 int kvm_cpu_has_pending_timer(struct kvm_vcpu *vcpu);
627 void kvm_vcpu_kick(struct kvm_vcpu *vcpu);
629 bool kvm_is_mmio_pfn(pfn_t pfn);
631 struct kvm_irq_ack_notifier {
632 struct hlist_node link;
634 void (*irq_acked)(struct kvm_irq_ack_notifier *kian);
637 struct kvm_assigned_dev_kernel {
638 struct kvm_irq_ack_notifier ack_notifier;
639 struct list_head list;
644 unsigned int entries_nr;
646 bool host_irq_disabled;
648 struct msix_entry *host_msix_entries;
650 struct msix_entry *guest_msix_entries;
651 unsigned long irq_requested_type;
656 spinlock_t intx_lock;
657 spinlock_t intx_mask_lock;
659 struct pci_saved_state *pci_saved_state;
662 struct kvm_irq_mask_notifier {
663 void (*func)(struct kvm_irq_mask_notifier *kimn, bool masked);
665 struct hlist_node link;
668 void kvm_register_irq_mask_notifier(struct kvm *kvm, int irq,
669 struct kvm_irq_mask_notifier *kimn);
670 void kvm_unregister_irq_mask_notifier(struct kvm *kvm, int irq,
671 struct kvm_irq_mask_notifier *kimn);
672 void kvm_fire_mask_notifiers(struct kvm *kvm, unsigned irqchip, unsigned pin,
675 #ifdef __KVM_HAVE_IOAPIC
676 void kvm_get_intr_delivery_bitmask(struct kvm_ioapic *ioapic,
677 union kvm_ioapic_redirect_entry *entry,
678 unsigned long *deliver_bitmask);
680 int kvm_set_irq(struct kvm *kvm, int irq_source_id, u32 irq, int level);
681 int kvm_set_msi(struct kvm_kernel_irq_routing_entry *irq_entry, struct kvm *kvm,
682 int irq_source_id, int level);
683 void kvm_notify_acked_irq(struct kvm *kvm, unsigned irqchip, unsigned pin);
684 void kvm_register_irq_ack_notifier(struct kvm *kvm,
685 struct kvm_irq_ack_notifier *kian);
686 void kvm_unregister_irq_ack_notifier(struct kvm *kvm,
687 struct kvm_irq_ack_notifier *kian);
688 int kvm_request_irq_source_id(struct kvm *kvm);
689 void kvm_free_irq_source_id(struct kvm *kvm, int irq_source_id);
691 /* For vcpu->arch.iommu_flags */
692 #define KVM_IOMMU_CACHE_COHERENCY 0x1
694 #ifdef CONFIG_IOMMU_API
695 int kvm_iommu_map_pages(struct kvm *kvm, struct kvm_memory_slot *slot);
696 void kvm_iommu_unmap_pages(struct kvm *kvm, struct kvm_memory_slot *slot);
697 int kvm_iommu_map_guest(struct kvm *kvm);
698 int kvm_iommu_unmap_guest(struct kvm *kvm);
699 int kvm_assign_device(struct kvm *kvm,
700 struct kvm_assigned_dev_kernel *assigned_dev);
701 int kvm_deassign_device(struct kvm *kvm,
702 struct kvm_assigned_dev_kernel *assigned_dev);
703 #else /* CONFIG_IOMMU_API */
704 static inline int kvm_iommu_map_pages(struct kvm *kvm,
705 struct kvm_memory_slot *slot)
710 static inline void kvm_iommu_unmap_pages(struct kvm *kvm,
711 struct kvm_memory_slot *slot)
715 static inline int kvm_iommu_map_guest(struct kvm *kvm)
720 static inline int kvm_iommu_unmap_guest(struct kvm *kvm)
725 static inline int kvm_assign_device(struct kvm *kvm,
726 struct kvm_assigned_dev_kernel *assigned_dev)
731 static inline int kvm_deassign_device(struct kvm *kvm,
732 struct kvm_assigned_dev_kernel *assigned_dev)
736 #endif /* CONFIG_IOMMU_API */
738 static inline void kvm_guest_enter(void)
740 BUG_ON(preemptible());
741 vtime_account(current);
742 current->flags |= PF_VCPU;
743 /* KVM does not hold any references to rcu protected data when it
744 * switches CPU into a guest mode. In fact switching to a guest mode
745 * is very similar to exiting to userspase from rcu point of view. In
746 * addition CPU may stay in a guest mode for quite a long time (up to
747 * one time slice). Lets treat guest mode as quiescent state, just like
748 * we do with user-mode execution.
750 rcu_virt_note_context_switch(smp_processor_id());
753 static inline void kvm_guest_exit(void)
755 vtime_account(current);
756 current->flags &= ~PF_VCPU;
760 * search_memslots() and __gfn_to_memslot() are here because they are
761 * used in non-modular code in arch/powerpc/kvm/book3s_hv_rm_mmu.c.
762 * gfn_to_memslot() itself isn't here as an inline because that would
763 * bloat other code too much.
765 static inline struct kvm_memory_slot *
766 search_memslots(struct kvm_memslots *slots, gfn_t gfn)
768 struct kvm_memory_slot *memslot;
770 kvm_for_each_memslot(memslot, slots)
771 if (gfn >= memslot->base_gfn &&
772 gfn < memslot->base_gfn + memslot->npages)
778 static inline struct kvm_memory_slot *
779 __gfn_to_memslot(struct kvm_memslots *slots, gfn_t gfn)
781 return search_memslots(slots, gfn);
784 static inline unsigned long
785 __gfn_to_hva_memslot(struct kvm_memory_slot *slot, gfn_t gfn)
787 return slot->userspace_addr + (gfn - slot->base_gfn) * PAGE_SIZE;
790 static inline int memslot_id(struct kvm *kvm, gfn_t gfn)
792 return gfn_to_memslot(kvm, gfn)->id;
795 static inline gfn_t gfn_to_index(gfn_t gfn, gfn_t base_gfn, int level)
797 /* KVM_HPAGE_GFN_SHIFT(PT_PAGE_TABLE_LEVEL) must be 0. */
798 return (gfn >> KVM_HPAGE_GFN_SHIFT(level)) -
799 (base_gfn >> KVM_HPAGE_GFN_SHIFT(level));
803 hva_to_gfn_memslot(unsigned long hva, struct kvm_memory_slot *slot)
805 gfn_t gfn_offset = (hva - slot->userspace_addr) >> PAGE_SHIFT;
807 return slot->base_gfn + gfn_offset;
810 static inline gpa_t gfn_to_gpa(gfn_t gfn)
812 return (gpa_t)gfn << PAGE_SHIFT;
815 static inline gfn_t gpa_to_gfn(gpa_t gpa)
817 return (gfn_t)(gpa >> PAGE_SHIFT);
820 static inline hpa_t pfn_to_hpa(pfn_t pfn)
822 return (hpa_t)pfn << PAGE_SHIFT;
825 static inline void kvm_migrate_timers(struct kvm_vcpu *vcpu)
827 set_bit(KVM_REQ_MIGRATE_TIMER, &vcpu->requests);
835 struct kvm_stats_debugfs_item {
838 enum kvm_stat_kind kind;
839 struct dentry *dentry;
841 extern struct kvm_stats_debugfs_item debugfs_entries[];
842 extern struct dentry *kvm_debugfs_dir;
844 #if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
845 static inline int mmu_notifier_retry(struct kvm_vcpu *vcpu, unsigned long mmu_seq)
847 if (unlikely(vcpu->kvm->mmu_notifier_count))
850 * Ensure the read of mmu_notifier_count happens before the read
851 * of mmu_notifier_seq. This interacts with the smp_wmb() in
852 * mmu_notifier_invalidate_range_end to make sure that the caller
853 * either sees the old (non-zero) value of mmu_notifier_count or
854 * the new (incremented) value of mmu_notifier_seq.
855 * PowerPC Book3s HV KVM calls this under a per-page lock
856 * rather than under kvm->mmu_lock, for scalability, so
857 * can't rely on kvm->mmu_lock to keep things ordered.
860 if (vcpu->kvm->mmu_notifier_seq != mmu_seq)
866 #ifdef KVM_CAP_IRQ_ROUTING
868 #define KVM_MAX_IRQ_ROUTES 1024
870 int kvm_setup_default_irq_routing(struct kvm *kvm);
871 int kvm_set_irq_routing(struct kvm *kvm,
872 const struct kvm_irq_routing_entry *entries,
875 void kvm_free_irq_routing(struct kvm *kvm);
877 int kvm_send_userspace_msi(struct kvm *kvm, struct kvm_msi *msi);
881 static inline void kvm_free_irq_routing(struct kvm *kvm) {}
885 #ifdef CONFIG_HAVE_KVM_EVENTFD
887 void kvm_eventfd_init(struct kvm *kvm);
888 int kvm_irqfd(struct kvm *kvm, struct kvm_irqfd *args);
889 void kvm_irqfd_release(struct kvm *kvm);
890 void kvm_irq_routing_update(struct kvm *, struct kvm_irq_routing_table *);
891 int kvm_ioeventfd(struct kvm *kvm, struct kvm_ioeventfd *args);
895 static inline void kvm_eventfd_init(struct kvm *kvm) {}
897 static inline int kvm_irqfd(struct kvm *kvm, struct kvm_irqfd *args)
902 static inline void kvm_irqfd_release(struct kvm *kvm) {}
904 #ifdef CONFIG_HAVE_KVM_IRQCHIP
905 static inline void kvm_irq_routing_update(struct kvm *kvm,
906 struct kvm_irq_routing_table *irq_rt)
908 rcu_assign_pointer(kvm->irq_routing, irq_rt);
912 static inline int kvm_ioeventfd(struct kvm *kvm, struct kvm_ioeventfd *args)
917 #endif /* CONFIG_HAVE_KVM_EVENTFD */
919 #ifdef CONFIG_KVM_APIC_ARCHITECTURE
920 static inline bool kvm_vcpu_is_bsp(struct kvm_vcpu *vcpu)
922 return vcpu->kvm->bsp_vcpu_id == vcpu->vcpu_id;
925 bool kvm_vcpu_compatible(struct kvm_vcpu *vcpu);
929 static inline bool kvm_vcpu_compatible(struct kvm_vcpu *vcpu) { return true; }
933 #ifdef __KVM_HAVE_DEVICE_ASSIGNMENT
935 long kvm_vm_ioctl_assigned_device(struct kvm *kvm, unsigned ioctl,
940 static inline long kvm_vm_ioctl_assigned_device(struct kvm *kvm, unsigned ioctl,
948 static inline void kvm_make_request(int req, struct kvm_vcpu *vcpu)
950 set_bit(req, &vcpu->requests);
953 static inline bool kvm_check_request(int req, struct kvm_vcpu *vcpu)
955 if (test_bit(req, &vcpu->requests)) {
956 clear_bit(req, &vcpu->requests);
963 #ifdef CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT
965 static inline void kvm_vcpu_set_in_spin_loop(struct kvm_vcpu *vcpu, bool val)
967 vcpu->spin_loop.in_spin_loop = val;
969 static inline void kvm_vcpu_set_dy_eligible(struct kvm_vcpu *vcpu, bool val)
971 vcpu->spin_loop.dy_eligible = val;
974 #else /* !CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT */
976 static inline void kvm_vcpu_set_in_spin_loop(struct kvm_vcpu *vcpu, bool val)
980 static inline void kvm_vcpu_set_dy_eligible(struct kvm_vcpu *vcpu, bool val)
984 static inline bool kvm_vcpu_eligible_for_directed_yield(struct kvm_vcpu *vcpu)
989 #endif /* CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT */