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
122 #define KVM_USERSPACE_IRQ_SOURCE_ID 0
126 extern struct kmem_cache *kvm_vcpu_cache;
128 struct kvm_io_range {
131 struct kvm_io_device *dev;
134 #define NR_IOBUS_DEVS 1000
138 struct kvm_io_range range[];
147 int kvm_io_bus_write(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr,
148 int len, const void *val);
149 int kvm_io_bus_read(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr, int len,
151 int kvm_io_bus_register_dev(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr,
152 int len, struct kvm_io_device *dev);
153 int kvm_io_bus_unregister_dev(struct kvm *kvm, enum kvm_bus bus_idx,
154 struct kvm_io_device *dev);
156 #ifdef CONFIG_KVM_ASYNC_PF
157 struct kvm_async_pf {
158 struct work_struct work;
159 struct list_head link;
160 struct list_head queue;
161 struct kvm_vcpu *vcpu;
162 struct mm_struct *mm;
165 struct kvm_arch_async_pf arch;
170 void kvm_clear_async_pf_completion_queue(struct kvm_vcpu *vcpu);
171 void kvm_check_async_pf_completion(struct kvm_vcpu *vcpu);
172 int kvm_setup_async_pf(struct kvm_vcpu *vcpu, gva_t gva, gfn_t gfn,
173 struct kvm_arch_async_pf *arch);
174 int kvm_async_pf_wakeup_all(struct kvm_vcpu *vcpu);
181 READING_SHADOW_PAGE_TABLES,
185 * Sometimes a large or cross-page mmio needs to be broken up into separate
186 * exits for userspace servicing.
188 struct kvm_mmio_fragment {
196 #ifdef CONFIG_PREEMPT_NOTIFIERS
197 struct preempt_notifier preempt_notifier;
203 unsigned long requests;
204 unsigned long guest_debug;
210 int guest_fpu_loaded, guest_xcr0_loaded;
211 wait_queue_head_t wq;
215 struct kvm_vcpu_stat stat;
217 #ifdef CONFIG_HAS_IOMEM
219 int mmio_read_completed;
221 int mmio_cur_fragment;
222 int mmio_nr_fragments;
223 struct kvm_mmio_fragment mmio_fragments[KVM_MAX_MMIO_FRAGMENTS];
226 #ifdef CONFIG_KVM_ASYNC_PF
229 struct list_head queue;
230 struct list_head done;
235 #ifdef CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT
237 * Cpu relax intercept or pause loop exit optimization
238 * in_spin_loop: set when a vcpu does a pause loop exit
239 * or cpu relax intercepted.
240 * dy_eligible: indicates whether vcpu is eligible for directed yield.
247 struct kvm_vcpu_arch arch;
250 static inline int kvm_vcpu_exiting_guest_mode(struct kvm_vcpu *vcpu)
252 return cmpxchg(&vcpu->mode, IN_GUEST_MODE, EXITING_GUEST_MODE);
256 * Some of the bitops functions do not support too long bitmaps.
257 * This number must be determined not to exceed such limits.
259 #define KVM_MEM_MAX_NR_PAGES ((1UL << 31) - 1)
261 struct kvm_memory_slot {
263 unsigned long npages;
265 unsigned long *dirty_bitmap;
266 struct kvm_arch_memory_slot arch;
267 unsigned long userspace_addr;
272 static inline unsigned long kvm_dirty_bitmap_bytes(struct kvm_memory_slot *memslot)
274 return ALIGN(memslot->npages, BITS_PER_LONG) / 8;
277 struct kvm_kernel_irq_routing_entry {
280 int (*set)(struct kvm_kernel_irq_routing_entry *e,
281 struct kvm *kvm, int irq_source_id, int level);
289 struct hlist_node link;
292 #ifdef __KVM_HAVE_IOAPIC
294 struct kvm_irq_routing_table {
295 int chip[KVM_NR_IRQCHIPS][KVM_IOAPIC_NUM_PINS];
296 struct kvm_kernel_irq_routing_entry *rt_entries;
299 * Array indexed by gsi. Each entry contains list of irq chips
300 * the gsi is connected to.
302 struct hlist_head map[0];
307 struct kvm_irq_routing_table {};
311 #ifndef KVM_MEM_SLOTS_NUM
312 #define KVM_MEM_SLOTS_NUM (KVM_MEMORY_SLOTS + KVM_PRIVATE_MEM_SLOTS)
317 * memslots are not sorted by id anymore, please use id_to_memslot()
318 * to get the memslot by its id.
320 struct kvm_memslots {
322 struct kvm_memory_slot memslots[KVM_MEM_SLOTS_NUM];
323 /* The mapping table from slot id to the index in memslots[]. */
324 int id_to_index[KVM_MEM_SLOTS_NUM];
329 struct mutex slots_lock;
330 struct mm_struct *mm; /* userspace tied to this vm */
331 struct kvm_memslots *memslots;
332 struct srcu_struct 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;
347 struct list_head ioeventfds;
349 struct kvm_vm_stat stat;
350 struct kvm_arch arch;
351 atomic_t users_count;
352 #ifdef KVM_COALESCED_MMIO_PAGE_OFFSET
353 struct kvm_coalesced_mmio_ring *coalesced_mmio_ring;
354 spinlock_t ring_lock;
355 struct list_head coalesced_zones;
358 struct mutex irq_lock;
359 #ifdef CONFIG_HAVE_KVM_IRQCHIP
361 * Update side is protected by irq_lock and,
362 * if configured, irqfds.lock.
364 struct kvm_irq_routing_table __rcu *irq_routing;
365 struct hlist_head mask_notifier_list;
366 struct hlist_head irq_ack_notifier_list;
369 #if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
370 struct mmu_notifier mmu_notifier;
371 unsigned long mmu_notifier_seq;
372 long mmu_notifier_count;
377 #define kvm_err(fmt, ...) \
378 pr_err("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
379 #define kvm_info(fmt, ...) \
380 pr_info("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
381 #define kvm_debug(fmt, ...) \
382 pr_debug("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
383 #define kvm_pr_unimpl(fmt, ...) \
384 pr_err_ratelimited("kvm [%i]: " fmt, \
385 task_tgid_nr(current), ## __VA_ARGS__)
387 /* The guest did something we don't support. */
388 #define vcpu_unimpl(vcpu, fmt, ...) \
389 kvm_pr_unimpl("vcpu%i " fmt, (vcpu)->vcpu_id, ## __VA_ARGS__)
391 static inline struct kvm_vcpu *kvm_get_vcpu(struct kvm *kvm, int i)
394 return kvm->vcpus[i];
397 #define kvm_for_each_vcpu(idx, vcpup, kvm) \
399 idx < atomic_read(&kvm->online_vcpus) && \
400 (vcpup = kvm_get_vcpu(kvm, idx)) != NULL; \
403 #define kvm_for_each_memslot(memslot, slots) \
404 for (memslot = &slots->memslots[0]; \
405 memslot < slots->memslots + KVM_MEM_SLOTS_NUM && memslot->npages;\
408 int kvm_vcpu_init(struct kvm_vcpu *vcpu, struct kvm *kvm, unsigned id);
409 void kvm_vcpu_uninit(struct kvm_vcpu *vcpu);
411 int __must_check vcpu_load(struct kvm_vcpu *vcpu);
412 void vcpu_put(struct kvm_vcpu *vcpu);
414 int kvm_init(void *opaque, unsigned vcpu_size, unsigned vcpu_align,
415 struct module *module);
418 void kvm_get_kvm(struct kvm *kvm);
419 void kvm_put_kvm(struct kvm *kvm);
420 void update_memslots(struct kvm_memslots *slots, struct kvm_memory_slot *new);
422 static inline struct kvm_memslots *kvm_memslots(struct kvm *kvm)
424 return rcu_dereference_check(kvm->memslots,
425 srcu_read_lock_held(&kvm->srcu)
426 || lockdep_is_held(&kvm->slots_lock));
429 static inline struct kvm_memory_slot *
430 id_to_memslot(struct kvm_memslots *slots, int id)
432 int index = slots->id_to_index[id];
433 struct kvm_memory_slot *slot;
435 slot = &slots->memslots[index];
437 WARN_ON(slot->id != id);
441 int kvm_set_memory_region(struct kvm *kvm,
442 struct kvm_userspace_memory_region *mem,
444 int __kvm_set_memory_region(struct kvm *kvm,
445 struct kvm_userspace_memory_region *mem,
447 void kvm_arch_free_memslot(struct kvm_memory_slot *free,
448 struct kvm_memory_slot *dont);
449 int kvm_arch_create_memslot(struct kvm_memory_slot *slot, unsigned long npages);
450 int kvm_arch_prepare_memory_region(struct kvm *kvm,
451 struct kvm_memory_slot *memslot,
452 struct kvm_memory_slot old,
453 struct kvm_userspace_memory_region *mem,
455 void kvm_arch_commit_memory_region(struct kvm *kvm,
456 struct kvm_userspace_memory_region *mem,
457 struct kvm_memory_slot old,
459 bool kvm_largepages_enabled(void);
460 void kvm_disable_largepages(void);
461 /* flush all memory translations */
462 void kvm_arch_flush_shadow_all(struct kvm *kvm);
463 /* flush memory translations pointing to 'slot' */
464 void kvm_arch_flush_shadow_memslot(struct kvm *kvm,
465 struct kvm_memory_slot *slot);
467 int gfn_to_page_many_atomic(struct kvm *kvm, gfn_t gfn, struct page **pages,
470 struct page *gfn_to_page(struct kvm *kvm, gfn_t gfn);
471 unsigned long gfn_to_hva(struct kvm *kvm, gfn_t gfn);
472 unsigned long gfn_to_hva_memslot(struct kvm_memory_slot *slot, gfn_t gfn);
473 void kvm_release_page_clean(struct page *page);
474 void kvm_release_page_dirty(struct page *page);
475 void kvm_set_page_dirty(struct page *page);
476 void kvm_set_page_accessed(struct page *page);
478 pfn_t gfn_to_pfn_atomic(struct kvm *kvm, gfn_t gfn);
479 pfn_t gfn_to_pfn_async(struct kvm *kvm, gfn_t gfn, bool *async,
480 bool write_fault, bool *writable);
481 pfn_t gfn_to_pfn(struct kvm *kvm, gfn_t gfn);
482 pfn_t gfn_to_pfn_prot(struct kvm *kvm, gfn_t gfn, bool write_fault,
484 pfn_t gfn_to_pfn_memslot(struct kvm_memory_slot *slot, gfn_t gfn);
485 pfn_t gfn_to_pfn_memslot_atomic(struct kvm_memory_slot *slot, gfn_t gfn);
487 void kvm_release_pfn_dirty(pfn_t pfn);
488 void kvm_release_pfn_clean(pfn_t pfn);
489 void kvm_set_pfn_dirty(pfn_t pfn);
490 void kvm_set_pfn_accessed(pfn_t pfn);
491 void kvm_get_pfn(pfn_t pfn);
493 int kvm_read_guest_page(struct kvm *kvm, gfn_t gfn, void *data, int offset,
495 int kvm_read_guest_atomic(struct kvm *kvm, gpa_t gpa, void *data,
497 int kvm_read_guest(struct kvm *kvm, gpa_t gpa, void *data, unsigned long len);
498 int kvm_read_guest_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
499 void *data, unsigned long len);
500 int kvm_write_guest_page(struct kvm *kvm, gfn_t gfn, const void *data,
501 int offset, int len);
502 int kvm_write_guest(struct kvm *kvm, gpa_t gpa, const void *data,
504 int kvm_write_guest_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
505 void *data, unsigned long len);
506 int kvm_gfn_to_hva_cache_init(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
508 int kvm_clear_guest_page(struct kvm *kvm, gfn_t gfn, int offset, int len);
509 int kvm_clear_guest(struct kvm *kvm, gpa_t gpa, unsigned long len);
510 struct kvm_memory_slot *gfn_to_memslot(struct kvm *kvm, gfn_t gfn);
511 int kvm_is_visible_gfn(struct kvm *kvm, gfn_t gfn);
512 unsigned long kvm_host_page_size(struct kvm *kvm, gfn_t gfn);
513 void mark_page_dirty(struct kvm *kvm, gfn_t gfn);
514 void mark_page_dirty_in_slot(struct kvm *kvm, struct kvm_memory_slot *memslot,
517 void kvm_vcpu_block(struct kvm_vcpu *vcpu);
518 void kvm_vcpu_kick(struct kvm_vcpu *vcpu);
519 bool kvm_vcpu_yield_to(struct kvm_vcpu *target);
520 void kvm_vcpu_on_spin(struct kvm_vcpu *vcpu);
521 void kvm_resched(struct kvm_vcpu *vcpu);
522 void kvm_load_guest_fpu(struct kvm_vcpu *vcpu);
523 void kvm_put_guest_fpu(struct kvm_vcpu *vcpu);
525 void kvm_flush_remote_tlbs(struct kvm *kvm);
526 void kvm_reload_remote_mmus(struct kvm *kvm);
528 long kvm_arch_dev_ioctl(struct file *filp,
529 unsigned int ioctl, unsigned long arg);
530 long kvm_arch_vcpu_ioctl(struct file *filp,
531 unsigned int ioctl, unsigned long arg);
532 int kvm_arch_vcpu_fault(struct kvm_vcpu *vcpu, struct vm_fault *vmf);
534 int kvm_dev_ioctl_check_extension(long ext);
536 int kvm_get_dirty_log(struct kvm *kvm,
537 struct kvm_dirty_log *log, int *is_dirty);
538 int kvm_vm_ioctl_get_dirty_log(struct kvm *kvm,
539 struct kvm_dirty_log *log);
541 int kvm_vm_ioctl_set_memory_region(struct kvm *kvm,
543 kvm_userspace_memory_region *mem,
545 int kvm_vm_ioctl_irq_line(struct kvm *kvm, struct kvm_irq_level *irq_level);
546 long kvm_arch_vm_ioctl(struct file *filp,
547 unsigned int ioctl, unsigned long arg);
549 int kvm_arch_vcpu_ioctl_get_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu);
550 int kvm_arch_vcpu_ioctl_set_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu);
552 int kvm_arch_vcpu_ioctl_translate(struct kvm_vcpu *vcpu,
553 struct kvm_translation *tr);
555 int kvm_arch_vcpu_ioctl_get_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs);
556 int kvm_arch_vcpu_ioctl_set_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs);
557 int kvm_arch_vcpu_ioctl_get_sregs(struct kvm_vcpu *vcpu,
558 struct kvm_sregs *sregs);
559 int kvm_arch_vcpu_ioctl_set_sregs(struct kvm_vcpu *vcpu,
560 struct kvm_sregs *sregs);
561 int kvm_arch_vcpu_ioctl_get_mpstate(struct kvm_vcpu *vcpu,
562 struct kvm_mp_state *mp_state);
563 int kvm_arch_vcpu_ioctl_set_mpstate(struct kvm_vcpu *vcpu,
564 struct kvm_mp_state *mp_state);
565 int kvm_arch_vcpu_ioctl_set_guest_debug(struct kvm_vcpu *vcpu,
566 struct kvm_guest_debug *dbg);
567 int kvm_arch_vcpu_ioctl_run(struct kvm_vcpu *vcpu, struct kvm_run *kvm_run);
569 int kvm_arch_init(void *opaque);
570 void kvm_arch_exit(void);
572 int kvm_arch_vcpu_init(struct kvm_vcpu *vcpu);
573 void kvm_arch_vcpu_uninit(struct kvm_vcpu *vcpu);
575 void kvm_arch_vcpu_free(struct kvm_vcpu *vcpu);
576 void kvm_arch_vcpu_load(struct kvm_vcpu *vcpu, int cpu);
577 void kvm_arch_vcpu_put(struct kvm_vcpu *vcpu);
578 struct kvm_vcpu *kvm_arch_vcpu_create(struct kvm *kvm, unsigned int id);
579 int kvm_arch_vcpu_setup(struct kvm_vcpu *vcpu);
580 void kvm_arch_vcpu_destroy(struct kvm_vcpu *vcpu);
582 int kvm_arch_vcpu_reset(struct kvm_vcpu *vcpu);
583 int kvm_arch_hardware_enable(void *garbage);
584 void kvm_arch_hardware_disable(void *garbage);
585 int kvm_arch_hardware_setup(void);
586 void kvm_arch_hardware_unsetup(void);
587 void kvm_arch_check_processor_compat(void *rtn);
588 int kvm_arch_vcpu_runnable(struct kvm_vcpu *vcpu);
589 int kvm_arch_vcpu_should_kick(struct kvm_vcpu *vcpu);
591 void kvm_free_physmem(struct kvm *kvm);
593 void *kvm_kvzalloc(unsigned long size);
594 void kvm_kvfree(const void *addr);
596 #ifndef __KVM_HAVE_ARCH_VM_ALLOC
597 static inline struct kvm *kvm_arch_alloc_vm(void)
599 return kzalloc(sizeof(struct kvm), GFP_KERNEL);
602 static inline void kvm_arch_free_vm(struct kvm *kvm)
608 static inline wait_queue_head_t *kvm_arch_vcpu_wq(struct kvm_vcpu *vcpu)
610 #ifdef __KVM_HAVE_ARCH_WQP
611 return vcpu->arch.wqp;
617 int kvm_arch_init_vm(struct kvm *kvm, unsigned long type);
618 void kvm_arch_destroy_vm(struct kvm *kvm);
619 void kvm_free_all_assigned_devices(struct kvm *kvm);
620 void kvm_arch_sync_events(struct kvm *kvm);
622 int kvm_cpu_has_pending_timer(struct kvm_vcpu *vcpu);
623 void kvm_vcpu_kick(struct kvm_vcpu *vcpu);
625 bool kvm_is_mmio_pfn(pfn_t pfn);
627 struct kvm_irq_ack_notifier {
628 struct hlist_node link;
630 void (*irq_acked)(struct kvm_irq_ack_notifier *kian);
633 struct kvm_assigned_dev_kernel {
634 struct kvm_irq_ack_notifier ack_notifier;
635 struct list_head list;
640 unsigned int entries_nr;
642 bool host_irq_disabled;
644 struct msix_entry *host_msix_entries;
646 struct msix_entry *guest_msix_entries;
647 unsigned long irq_requested_type;
652 spinlock_t intx_lock;
653 spinlock_t intx_mask_lock;
655 struct pci_saved_state *pci_saved_state;
658 struct kvm_irq_mask_notifier {
659 void (*func)(struct kvm_irq_mask_notifier *kimn, bool masked);
661 struct hlist_node link;
664 void kvm_register_irq_mask_notifier(struct kvm *kvm, int irq,
665 struct kvm_irq_mask_notifier *kimn);
666 void kvm_unregister_irq_mask_notifier(struct kvm *kvm, int irq,
667 struct kvm_irq_mask_notifier *kimn);
668 void kvm_fire_mask_notifiers(struct kvm *kvm, unsigned irqchip, unsigned pin,
671 #ifdef __KVM_HAVE_IOAPIC
672 void kvm_get_intr_delivery_bitmask(struct kvm_ioapic *ioapic,
673 union kvm_ioapic_redirect_entry *entry,
674 unsigned long *deliver_bitmask);
676 int kvm_set_irq(struct kvm *kvm, int irq_source_id, u32 irq, int level);
677 int kvm_set_msi(struct kvm_kernel_irq_routing_entry *irq_entry, struct kvm *kvm,
678 int irq_source_id, int level);
679 void kvm_notify_acked_irq(struct kvm *kvm, unsigned irqchip, unsigned pin);
680 void kvm_register_irq_ack_notifier(struct kvm *kvm,
681 struct kvm_irq_ack_notifier *kian);
682 void kvm_unregister_irq_ack_notifier(struct kvm *kvm,
683 struct kvm_irq_ack_notifier *kian);
684 int kvm_request_irq_source_id(struct kvm *kvm);
685 void kvm_free_irq_source_id(struct kvm *kvm, int irq_source_id);
687 /* For vcpu->arch.iommu_flags */
688 #define KVM_IOMMU_CACHE_COHERENCY 0x1
690 #ifdef CONFIG_IOMMU_API
691 int kvm_iommu_map_pages(struct kvm *kvm, struct kvm_memory_slot *slot);
692 void kvm_iommu_unmap_pages(struct kvm *kvm, struct kvm_memory_slot *slot);
693 int kvm_iommu_map_guest(struct kvm *kvm);
694 int kvm_iommu_unmap_guest(struct kvm *kvm);
695 int kvm_assign_device(struct kvm *kvm,
696 struct kvm_assigned_dev_kernel *assigned_dev);
697 int kvm_deassign_device(struct kvm *kvm,
698 struct kvm_assigned_dev_kernel *assigned_dev);
699 #else /* CONFIG_IOMMU_API */
700 static inline int kvm_iommu_map_pages(struct kvm *kvm,
701 struct kvm_memory_slot *slot)
706 static inline void kvm_iommu_unmap_pages(struct kvm *kvm,
707 struct kvm_memory_slot *slot)
711 static inline int kvm_iommu_map_guest(struct kvm *kvm)
716 static inline int kvm_iommu_unmap_guest(struct kvm *kvm)
721 static inline int kvm_assign_device(struct kvm *kvm,
722 struct kvm_assigned_dev_kernel *assigned_dev)
727 static inline int kvm_deassign_device(struct kvm *kvm,
728 struct kvm_assigned_dev_kernel *assigned_dev)
732 #endif /* CONFIG_IOMMU_API */
734 static inline void kvm_guest_enter(void)
736 BUG_ON(preemptible());
737 account_system_vtime(current);
738 current->flags |= PF_VCPU;
739 /* KVM does not hold any references to rcu protected data when it
740 * switches CPU into a guest mode. In fact switching to a guest mode
741 * is very similar to exiting to userspase from rcu point of view. In
742 * addition CPU may stay in a guest mode for quite a long time (up to
743 * one time slice). Lets treat guest mode as quiescent state, just like
744 * we do with user-mode execution.
746 rcu_virt_note_context_switch(smp_processor_id());
749 static inline void kvm_guest_exit(void)
751 account_system_vtime(current);
752 current->flags &= ~PF_VCPU;
756 * search_memslots() and __gfn_to_memslot() are here because they are
757 * used in non-modular code in arch/powerpc/kvm/book3s_hv_rm_mmu.c.
758 * gfn_to_memslot() itself isn't here as an inline because that would
759 * bloat other code too much.
761 static inline struct kvm_memory_slot *
762 search_memslots(struct kvm_memslots *slots, gfn_t gfn)
764 struct kvm_memory_slot *memslot;
766 kvm_for_each_memslot(memslot, slots)
767 if (gfn >= memslot->base_gfn &&
768 gfn < memslot->base_gfn + memslot->npages)
774 static inline struct kvm_memory_slot *
775 __gfn_to_memslot(struct kvm_memslots *slots, gfn_t gfn)
777 return search_memslots(slots, gfn);
780 static inline unsigned long
781 __gfn_to_hva_memslot(struct kvm_memory_slot *slot, gfn_t gfn)
783 return slot->userspace_addr + (gfn - slot->base_gfn) * PAGE_SIZE;
786 static inline int memslot_id(struct kvm *kvm, gfn_t gfn)
788 return gfn_to_memslot(kvm, gfn)->id;
791 static inline gfn_t gfn_to_index(gfn_t gfn, gfn_t base_gfn, int level)
793 /* KVM_HPAGE_GFN_SHIFT(PT_PAGE_TABLE_LEVEL) must be 0. */
794 return (gfn >> KVM_HPAGE_GFN_SHIFT(level)) -
795 (base_gfn >> KVM_HPAGE_GFN_SHIFT(level));
799 hva_to_gfn_memslot(unsigned long hva, struct kvm_memory_slot *slot)
801 gfn_t gfn_offset = (hva - slot->userspace_addr) >> PAGE_SHIFT;
803 return slot->base_gfn + gfn_offset;
806 static inline gpa_t gfn_to_gpa(gfn_t gfn)
808 return (gpa_t)gfn << PAGE_SHIFT;
811 static inline gfn_t gpa_to_gfn(gpa_t gpa)
813 return (gfn_t)(gpa >> PAGE_SHIFT);
816 static inline hpa_t pfn_to_hpa(pfn_t pfn)
818 return (hpa_t)pfn << PAGE_SHIFT;
821 static inline void kvm_migrate_timers(struct kvm_vcpu *vcpu)
823 set_bit(KVM_REQ_MIGRATE_TIMER, &vcpu->requests);
831 struct kvm_stats_debugfs_item {
834 enum kvm_stat_kind kind;
835 struct dentry *dentry;
837 extern struct kvm_stats_debugfs_item debugfs_entries[];
838 extern struct dentry *kvm_debugfs_dir;
840 #if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
841 static inline int mmu_notifier_retry(struct kvm_vcpu *vcpu, unsigned long mmu_seq)
843 if (unlikely(vcpu->kvm->mmu_notifier_count))
846 * Ensure the read of mmu_notifier_count happens before the read
847 * of mmu_notifier_seq. This interacts with the smp_wmb() in
848 * mmu_notifier_invalidate_range_end to make sure that the caller
849 * either sees the old (non-zero) value of mmu_notifier_count or
850 * the new (incremented) value of mmu_notifier_seq.
851 * PowerPC Book3s HV KVM calls this under a per-page lock
852 * rather than under kvm->mmu_lock, for scalability, so
853 * can't rely on kvm->mmu_lock to keep things ordered.
856 if (vcpu->kvm->mmu_notifier_seq != mmu_seq)
862 #ifdef KVM_CAP_IRQ_ROUTING
864 #define KVM_MAX_IRQ_ROUTES 1024
866 int kvm_setup_default_irq_routing(struct kvm *kvm);
867 int kvm_set_irq_routing(struct kvm *kvm,
868 const struct kvm_irq_routing_entry *entries,
871 void kvm_free_irq_routing(struct kvm *kvm);
873 int kvm_send_userspace_msi(struct kvm *kvm, struct kvm_msi *msi);
877 static inline void kvm_free_irq_routing(struct kvm *kvm) {}
881 #ifdef CONFIG_HAVE_KVM_EVENTFD
883 void kvm_eventfd_init(struct kvm *kvm);
884 int kvm_irqfd(struct kvm *kvm, struct kvm_irqfd *args);
885 void kvm_irqfd_release(struct kvm *kvm);
886 void kvm_irq_routing_update(struct kvm *, struct kvm_irq_routing_table *);
887 int kvm_ioeventfd(struct kvm *kvm, struct kvm_ioeventfd *args);
891 static inline void kvm_eventfd_init(struct kvm *kvm) {}
893 static inline int kvm_irqfd(struct kvm *kvm, struct kvm_irqfd *args)
898 static inline void kvm_irqfd_release(struct kvm *kvm) {}
900 #ifdef CONFIG_HAVE_KVM_IRQCHIP
901 static inline void kvm_irq_routing_update(struct kvm *kvm,
902 struct kvm_irq_routing_table *irq_rt)
904 rcu_assign_pointer(kvm->irq_routing, irq_rt);
908 static inline int kvm_ioeventfd(struct kvm *kvm, struct kvm_ioeventfd *args)
913 #endif /* CONFIG_HAVE_KVM_EVENTFD */
915 #ifdef CONFIG_KVM_APIC_ARCHITECTURE
916 static inline bool kvm_vcpu_is_bsp(struct kvm_vcpu *vcpu)
918 return vcpu->kvm->bsp_vcpu_id == vcpu->vcpu_id;
921 bool kvm_vcpu_compatible(struct kvm_vcpu *vcpu);
925 static inline bool kvm_vcpu_compatible(struct kvm_vcpu *vcpu) { return true; }
929 #ifdef __KVM_HAVE_DEVICE_ASSIGNMENT
931 long kvm_vm_ioctl_assigned_device(struct kvm *kvm, unsigned ioctl,
936 static inline long kvm_vm_ioctl_assigned_device(struct kvm *kvm, unsigned ioctl,
944 static inline void kvm_make_request(int req, struct kvm_vcpu *vcpu)
946 set_bit(req, &vcpu->requests);
949 static inline bool kvm_check_request(int req, struct kvm_vcpu *vcpu)
951 if (test_bit(req, &vcpu->requests)) {
952 clear_bit(req, &vcpu->requests);
959 #ifdef CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT
961 static inline void kvm_vcpu_set_in_spin_loop(struct kvm_vcpu *vcpu, bool val)
963 vcpu->spin_loop.in_spin_loop = val;
965 static inline void kvm_vcpu_set_dy_eligible(struct kvm_vcpu *vcpu, bool val)
967 vcpu->spin_loop.dy_eligible = val;
970 #else /* !CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT */
972 static inline void kvm_vcpu_set_in_spin_loop(struct kvm_vcpu *vcpu, bool val)
976 static inline void kvm_vcpu_set_dy_eligible(struct kvm_vcpu *vcpu, bool val)
980 static inline bool kvm_vcpu_eligible_for_directed_yield(struct kvm_vcpu *vcpu)
985 #endif /* CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT */