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;
89 * architectures with KVM_HVA_ERR_BAD other than PAGE_OFFSET (e.g. s390)
90 * provide own defines and kvm_is_error_hva
92 #ifndef KVM_HVA_ERR_BAD
94 #define KVM_HVA_ERR_BAD (PAGE_OFFSET)
95 #define KVM_HVA_ERR_RO_BAD (PAGE_OFFSET + PAGE_SIZE)
97 static inline bool kvm_is_error_hva(unsigned long addr)
99 return addr >= PAGE_OFFSET;
104 #define KVM_ERR_PTR_BAD_PAGE (ERR_PTR(-ENOENT))
106 static inline bool is_error_page(struct page *page)
112 * vcpu->requests bit members
114 #define KVM_REQ_TLB_FLUSH 0
115 #define KVM_REQ_MIGRATE_TIMER 1
116 #define KVM_REQ_REPORT_TPR_ACCESS 2
117 #define KVM_REQ_MMU_RELOAD 3
118 #define KVM_REQ_TRIPLE_FAULT 4
119 #define KVM_REQ_PENDING_TIMER 5
120 #define KVM_REQ_UNHALT 6
121 #define KVM_REQ_MMU_SYNC 7
122 #define KVM_REQ_CLOCK_UPDATE 8
123 #define KVM_REQ_KICK 9
124 #define KVM_REQ_DEACTIVATE_FPU 10
125 #define KVM_REQ_EVENT 11
126 #define KVM_REQ_APF_HALT 12
127 #define KVM_REQ_STEAL_UPDATE 13
128 #define KVM_REQ_NMI 14
129 #define KVM_REQ_PMU 15
130 #define KVM_REQ_PMI 16
131 #define KVM_REQ_WATCHDOG 17
132 #define KVM_REQ_MASTERCLOCK_UPDATE 18
133 #define KVM_REQ_MCLOCK_INPROGRESS 19
134 #define KVM_REQ_EPR_EXIT 20
135 #define KVM_REQ_SCAN_IOAPIC 21
136 #define KVM_REQ_GLOBAL_CLOCK_UPDATE 22
137 #define KVM_REQ_ENABLE_IBS 23
138 #define KVM_REQ_DISABLE_IBS 24
140 #define KVM_USERSPACE_IRQ_SOURCE_ID 0
141 #define KVM_IRQFD_RESAMPLE_IRQ_SOURCE_ID 1
145 extern struct kmem_cache *kvm_vcpu_cache;
147 extern spinlock_t kvm_lock;
148 extern struct list_head vm_list;
150 struct kvm_io_range {
153 struct kvm_io_device *dev;
156 #define NR_IOBUS_DEVS 1000
161 struct kvm_io_range range[];
167 KVM_VIRTIO_CCW_NOTIFY_BUS,
172 int kvm_io_bus_write(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr,
173 int len, const void *val);
174 int kvm_io_bus_write_cookie(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr,
175 int len, const void *val, long cookie);
176 int kvm_io_bus_read(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr, int len,
178 int kvm_io_bus_register_dev(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr,
179 int len, struct kvm_io_device *dev);
180 int kvm_io_bus_unregister_dev(struct kvm *kvm, enum kvm_bus bus_idx,
181 struct kvm_io_device *dev);
183 #ifdef CONFIG_KVM_ASYNC_PF
184 struct kvm_async_pf {
185 struct work_struct work;
186 struct list_head link;
187 struct list_head queue;
188 struct kvm_vcpu *vcpu;
189 struct mm_struct *mm;
192 struct kvm_arch_async_pf arch;
196 void kvm_clear_async_pf_completion_queue(struct kvm_vcpu *vcpu);
197 void kvm_check_async_pf_completion(struct kvm_vcpu *vcpu);
198 int kvm_setup_async_pf(struct kvm_vcpu *vcpu, gva_t gva, unsigned long hva,
199 struct kvm_arch_async_pf *arch);
200 int kvm_async_pf_wakeup_all(struct kvm_vcpu *vcpu);
207 READING_SHADOW_PAGE_TABLES,
211 * Sometimes a large or cross-page mmio needs to be broken up into separate
212 * exits for userspace servicing.
214 struct kvm_mmio_fragment {
222 #ifdef CONFIG_PREEMPT_NOTIFIERS
223 struct preempt_notifier preempt_notifier;
229 unsigned long requests;
230 unsigned long guest_debug;
236 int guest_fpu_loaded, guest_xcr0_loaded;
237 wait_queue_head_t wq;
241 struct kvm_vcpu_stat stat;
243 #ifdef CONFIG_HAS_IOMEM
245 int mmio_read_completed;
247 int mmio_cur_fragment;
248 int mmio_nr_fragments;
249 struct kvm_mmio_fragment mmio_fragments[KVM_MAX_MMIO_FRAGMENTS];
252 #ifdef CONFIG_KVM_ASYNC_PF
255 struct list_head queue;
256 struct list_head done;
261 #ifdef CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT
263 * Cpu relax intercept or pause loop exit optimization
264 * in_spin_loop: set when a vcpu does a pause loop exit
265 * or cpu relax intercepted.
266 * dy_eligible: indicates whether vcpu is eligible for directed yield.
274 struct kvm_vcpu_arch arch;
277 static inline int kvm_vcpu_exiting_guest_mode(struct kvm_vcpu *vcpu)
279 return cmpxchg(&vcpu->mode, IN_GUEST_MODE, EXITING_GUEST_MODE);
283 * Some of the bitops functions do not support too long bitmaps.
284 * This number must be determined not to exceed such limits.
286 #define KVM_MEM_MAX_NR_PAGES ((1UL << 31) - 1)
288 struct kvm_memory_slot {
290 unsigned long npages;
291 unsigned long *dirty_bitmap;
292 struct kvm_arch_memory_slot arch;
293 unsigned long userspace_addr;
298 static inline unsigned long kvm_dirty_bitmap_bytes(struct kvm_memory_slot *memslot)
300 return ALIGN(memslot->npages, BITS_PER_LONG) / 8;
303 struct kvm_s390_adapter_int {
311 struct kvm_kernel_irq_routing_entry {
314 int (*set)(struct kvm_kernel_irq_routing_entry *e,
315 struct kvm *kvm, int irq_source_id, int level,
323 struct kvm_s390_adapter_int adapter;
325 struct hlist_node link;
328 struct kvm_irq_routing_table;
330 #ifndef KVM_PRIVATE_MEM_SLOTS
331 #define KVM_PRIVATE_MEM_SLOTS 0
334 #ifndef KVM_MEM_SLOTS_NUM
335 #define KVM_MEM_SLOTS_NUM (KVM_USER_MEM_SLOTS + KVM_PRIVATE_MEM_SLOTS)
340 * memslots are not sorted by id anymore, please use id_to_memslot()
341 * to get the memslot by its id.
343 struct kvm_memslots {
345 struct kvm_memory_slot memslots[KVM_MEM_SLOTS_NUM];
346 /* The mapping table from slot id to the index in memslots[]. */
347 short id_to_index[KVM_MEM_SLOTS_NUM];
352 struct mutex slots_lock;
353 struct mm_struct *mm; /* userspace tied to this vm */
354 struct kvm_memslots *memslots;
355 struct srcu_struct srcu;
356 struct srcu_struct irq_srcu;
357 #ifdef CONFIG_KVM_APIC_ARCHITECTURE
360 struct kvm_vcpu *vcpus[KVM_MAX_VCPUS];
361 atomic_t online_vcpus;
362 int last_boosted_vcpu;
363 struct list_head vm_list;
365 struct kvm_io_bus *buses[KVM_NR_BUSES];
366 #ifdef CONFIG_HAVE_KVM_EVENTFD
369 struct list_head items;
370 struct list_head resampler_list;
371 struct mutex resampler_lock;
373 struct list_head ioeventfds;
375 struct kvm_vm_stat stat;
376 struct kvm_arch arch;
377 atomic_t users_count;
378 #ifdef KVM_COALESCED_MMIO_PAGE_OFFSET
379 struct kvm_coalesced_mmio_ring *coalesced_mmio_ring;
380 spinlock_t ring_lock;
381 struct list_head coalesced_zones;
384 struct mutex irq_lock;
385 #ifdef CONFIG_HAVE_KVM_IRQCHIP
387 * Update side is protected by irq_lock.
389 struct kvm_irq_routing_table __rcu *irq_routing;
390 struct hlist_head mask_notifier_list;
391 struct hlist_head irq_ack_notifier_list;
394 #if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
395 struct mmu_notifier mmu_notifier;
396 unsigned long mmu_notifier_seq;
397 long mmu_notifier_count;
400 struct list_head devices;
403 #define kvm_err(fmt, ...) \
404 pr_err("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
405 #define kvm_info(fmt, ...) \
406 pr_info("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
407 #define kvm_debug(fmt, ...) \
408 pr_debug("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
409 #define kvm_pr_unimpl(fmt, ...) \
410 pr_err_ratelimited("kvm [%i]: " fmt, \
411 task_tgid_nr(current), ## __VA_ARGS__)
413 /* The guest did something we don't support. */
414 #define vcpu_unimpl(vcpu, fmt, ...) \
415 kvm_pr_unimpl("vcpu%i " fmt, (vcpu)->vcpu_id, ## __VA_ARGS__)
417 static inline struct kvm_vcpu *kvm_get_vcpu(struct kvm *kvm, int i)
420 return kvm->vcpus[i];
423 #define kvm_for_each_vcpu(idx, vcpup, kvm) \
425 idx < atomic_read(&kvm->online_vcpus) && \
426 (vcpup = kvm_get_vcpu(kvm, idx)) != NULL; \
429 #define kvm_for_each_memslot(memslot, slots) \
430 for (memslot = &slots->memslots[0]; \
431 memslot < slots->memslots + KVM_MEM_SLOTS_NUM && memslot->npages;\
434 int kvm_vcpu_init(struct kvm_vcpu *vcpu, struct kvm *kvm, unsigned id);
435 void kvm_vcpu_uninit(struct kvm_vcpu *vcpu);
437 int __must_check vcpu_load(struct kvm_vcpu *vcpu);
438 void vcpu_put(struct kvm_vcpu *vcpu);
440 #ifdef CONFIG_HAVE_KVM_IRQFD
441 int kvm_irqfd_init(void);
442 void kvm_irqfd_exit(void);
444 static inline int kvm_irqfd_init(void)
449 static inline void kvm_irqfd_exit(void)
453 int kvm_init(void *opaque, unsigned vcpu_size, unsigned vcpu_align,
454 struct module *module);
457 void kvm_get_kvm(struct kvm *kvm);
458 void kvm_put_kvm(struct kvm *kvm);
460 static inline struct kvm_memslots *kvm_memslots(struct kvm *kvm)
462 return rcu_dereference_check(kvm->memslots,
463 srcu_read_lock_held(&kvm->srcu)
464 || lockdep_is_held(&kvm->slots_lock));
467 static inline struct kvm_memory_slot *
468 id_to_memslot(struct kvm_memslots *slots, int id)
470 int index = slots->id_to_index[id];
471 struct kvm_memory_slot *slot;
473 slot = &slots->memslots[index];
475 WARN_ON(slot->id != id);
480 * KVM_SET_USER_MEMORY_REGION ioctl allows the following operations:
481 * - create a new memory slot
482 * - delete an existing memory slot
483 * - modify an existing memory slot
484 * -- move it in the guest physical memory space
485 * -- just change its flags
487 * Since flags can be changed by some of these operations, the following
488 * differentiation is the best we can do for __kvm_set_memory_region():
497 int kvm_set_memory_region(struct kvm *kvm,
498 struct kvm_userspace_memory_region *mem);
499 int __kvm_set_memory_region(struct kvm *kvm,
500 struct kvm_userspace_memory_region *mem);
501 void kvm_arch_free_memslot(struct kvm *kvm, struct kvm_memory_slot *free,
502 struct kvm_memory_slot *dont);
503 int kvm_arch_create_memslot(struct kvm *kvm, struct kvm_memory_slot *slot,
504 unsigned long npages);
505 void kvm_arch_memslots_updated(struct kvm *kvm);
506 int kvm_arch_prepare_memory_region(struct kvm *kvm,
507 struct kvm_memory_slot *memslot,
508 struct kvm_userspace_memory_region *mem,
509 enum kvm_mr_change change);
510 void kvm_arch_commit_memory_region(struct kvm *kvm,
511 struct kvm_userspace_memory_region *mem,
512 const struct kvm_memory_slot *old,
513 enum kvm_mr_change change);
514 bool kvm_largepages_enabled(void);
515 void kvm_disable_largepages(void);
516 /* flush all memory translations */
517 void kvm_arch_flush_shadow_all(struct kvm *kvm);
518 /* flush memory translations pointing to 'slot' */
519 void kvm_arch_flush_shadow_memslot(struct kvm *kvm,
520 struct kvm_memory_slot *slot);
522 int gfn_to_page_many_atomic(struct kvm *kvm, gfn_t gfn, struct page **pages,
525 struct page *gfn_to_page(struct kvm *kvm, gfn_t gfn);
526 unsigned long gfn_to_hva(struct kvm *kvm, gfn_t gfn);
527 unsigned long gfn_to_hva_prot(struct kvm *kvm, gfn_t gfn, bool *writable);
528 unsigned long gfn_to_hva_memslot(struct kvm_memory_slot *slot, gfn_t gfn);
529 void kvm_release_page_clean(struct page *page);
530 void kvm_release_page_dirty(struct page *page);
531 void kvm_set_page_accessed(struct page *page);
533 pfn_t gfn_to_pfn_atomic(struct kvm *kvm, gfn_t gfn);
534 pfn_t gfn_to_pfn_async(struct kvm *kvm, gfn_t gfn, bool *async,
535 bool write_fault, bool *writable);
536 pfn_t gfn_to_pfn(struct kvm *kvm, gfn_t gfn);
537 pfn_t gfn_to_pfn_prot(struct kvm *kvm, gfn_t gfn, bool write_fault,
539 pfn_t gfn_to_pfn_memslot(struct kvm_memory_slot *slot, gfn_t gfn);
540 pfn_t gfn_to_pfn_memslot_atomic(struct kvm_memory_slot *slot, gfn_t gfn);
542 void kvm_release_pfn_clean(pfn_t pfn);
543 void kvm_set_pfn_dirty(pfn_t pfn);
544 void kvm_set_pfn_accessed(pfn_t pfn);
545 void kvm_get_pfn(pfn_t pfn);
547 int kvm_read_guest_page(struct kvm *kvm, gfn_t gfn, void *data, int offset,
549 int kvm_read_guest_atomic(struct kvm *kvm, gpa_t gpa, void *data,
551 int kvm_read_guest(struct kvm *kvm, gpa_t gpa, void *data, unsigned long len);
552 int kvm_read_guest_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
553 void *data, unsigned long len);
554 int kvm_write_guest_page(struct kvm *kvm, gfn_t gfn, const void *data,
555 int offset, int len);
556 int kvm_write_guest(struct kvm *kvm, gpa_t gpa, const void *data,
558 int kvm_write_guest_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
559 void *data, unsigned long len);
560 int kvm_gfn_to_hva_cache_init(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
561 gpa_t gpa, unsigned long len);
562 int kvm_clear_guest_page(struct kvm *kvm, gfn_t gfn, int offset, int len);
563 int kvm_clear_guest(struct kvm *kvm, gpa_t gpa, unsigned long len);
564 struct kvm_memory_slot *gfn_to_memslot(struct kvm *kvm, gfn_t gfn);
565 int kvm_is_visible_gfn(struct kvm *kvm, gfn_t gfn);
566 unsigned long kvm_host_page_size(struct kvm *kvm, gfn_t gfn);
567 void mark_page_dirty(struct kvm *kvm, gfn_t gfn);
569 void kvm_vcpu_block(struct kvm_vcpu *vcpu);
570 void kvm_vcpu_kick(struct kvm_vcpu *vcpu);
571 int kvm_vcpu_yield_to(struct kvm_vcpu *target);
572 void kvm_vcpu_on_spin(struct kvm_vcpu *vcpu);
573 void kvm_load_guest_fpu(struct kvm_vcpu *vcpu);
574 void kvm_put_guest_fpu(struct kvm_vcpu *vcpu);
576 void kvm_flush_remote_tlbs(struct kvm *kvm);
577 void kvm_reload_remote_mmus(struct kvm *kvm);
578 void kvm_make_mclock_inprogress_request(struct kvm *kvm);
579 void kvm_make_scan_ioapic_request(struct kvm *kvm);
581 long kvm_arch_dev_ioctl(struct file *filp,
582 unsigned int ioctl, unsigned long arg);
583 long kvm_arch_vcpu_ioctl(struct file *filp,
584 unsigned int ioctl, unsigned long arg);
585 int kvm_arch_vcpu_fault(struct kvm_vcpu *vcpu, struct vm_fault *vmf);
587 int kvm_vm_ioctl_check_extension(struct kvm *kvm, long ext);
589 int kvm_get_dirty_log(struct kvm *kvm,
590 struct kvm_dirty_log *log, int *is_dirty);
591 int kvm_vm_ioctl_get_dirty_log(struct kvm *kvm,
592 struct kvm_dirty_log *log);
594 int kvm_vm_ioctl_irq_line(struct kvm *kvm, struct kvm_irq_level *irq_level,
596 long kvm_arch_vm_ioctl(struct file *filp,
597 unsigned int ioctl, unsigned long arg);
599 int kvm_arch_vcpu_ioctl_get_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu);
600 int kvm_arch_vcpu_ioctl_set_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu);
602 int kvm_arch_vcpu_ioctl_translate(struct kvm_vcpu *vcpu,
603 struct kvm_translation *tr);
605 int kvm_arch_vcpu_ioctl_get_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs);
606 int kvm_arch_vcpu_ioctl_set_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs);
607 int kvm_arch_vcpu_ioctl_get_sregs(struct kvm_vcpu *vcpu,
608 struct kvm_sregs *sregs);
609 int kvm_arch_vcpu_ioctl_set_sregs(struct kvm_vcpu *vcpu,
610 struct kvm_sregs *sregs);
611 int kvm_arch_vcpu_ioctl_get_mpstate(struct kvm_vcpu *vcpu,
612 struct kvm_mp_state *mp_state);
613 int kvm_arch_vcpu_ioctl_set_mpstate(struct kvm_vcpu *vcpu,
614 struct kvm_mp_state *mp_state);
615 int kvm_arch_vcpu_ioctl_set_guest_debug(struct kvm_vcpu *vcpu,
616 struct kvm_guest_debug *dbg);
617 int kvm_arch_vcpu_ioctl_run(struct kvm_vcpu *vcpu, struct kvm_run *kvm_run);
619 int kvm_arch_init(void *opaque);
620 void kvm_arch_exit(void);
622 int kvm_arch_vcpu_init(struct kvm_vcpu *vcpu);
623 void kvm_arch_vcpu_uninit(struct kvm_vcpu *vcpu);
625 void kvm_arch_vcpu_free(struct kvm_vcpu *vcpu);
626 void kvm_arch_vcpu_load(struct kvm_vcpu *vcpu, int cpu);
627 void kvm_arch_vcpu_put(struct kvm_vcpu *vcpu);
628 struct kvm_vcpu *kvm_arch_vcpu_create(struct kvm *kvm, unsigned int id);
629 int kvm_arch_vcpu_setup(struct kvm_vcpu *vcpu);
630 int kvm_arch_vcpu_postcreate(struct kvm_vcpu *vcpu);
631 void kvm_arch_vcpu_destroy(struct kvm_vcpu *vcpu);
633 int kvm_arch_hardware_enable(void *garbage);
634 void kvm_arch_hardware_disable(void *garbage);
635 int kvm_arch_hardware_setup(void);
636 void kvm_arch_hardware_unsetup(void);
637 void kvm_arch_check_processor_compat(void *rtn);
638 int kvm_arch_vcpu_runnable(struct kvm_vcpu *vcpu);
639 int kvm_arch_vcpu_should_kick(struct kvm_vcpu *vcpu);
641 void *kvm_kvzalloc(unsigned long size);
642 void kvm_kvfree(const void *addr);
644 #ifndef __KVM_HAVE_ARCH_VM_ALLOC
645 static inline struct kvm *kvm_arch_alloc_vm(void)
647 return kzalloc(sizeof(struct kvm), GFP_KERNEL);
650 static inline void kvm_arch_free_vm(struct kvm *kvm)
656 #ifdef __KVM_HAVE_ARCH_NONCOHERENT_DMA
657 void kvm_arch_register_noncoherent_dma(struct kvm *kvm);
658 void kvm_arch_unregister_noncoherent_dma(struct kvm *kvm);
659 bool kvm_arch_has_noncoherent_dma(struct kvm *kvm);
661 static inline void kvm_arch_register_noncoherent_dma(struct kvm *kvm)
665 static inline void kvm_arch_unregister_noncoherent_dma(struct kvm *kvm)
669 static inline bool kvm_arch_has_noncoherent_dma(struct kvm *kvm)
675 static inline wait_queue_head_t *kvm_arch_vcpu_wq(struct kvm_vcpu *vcpu)
677 #ifdef __KVM_HAVE_ARCH_WQP
678 return vcpu->arch.wqp;
684 int kvm_arch_init_vm(struct kvm *kvm, unsigned long type);
685 void kvm_arch_destroy_vm(struct kvm *kvm);
686 void kvm_arch_sync_events(struct kvm *kvm);
688 int kvm_cpu_has_pending_timer(struct kvm_vcpu *vcpu);
689 void kvm_vcpu_kick(struct kvm_vcpu *vcpu);
691 bool kvm_is_mmio_pfn(pfn_t pfn);
693 struct kvm_irq_ack_notifier {
694 struct hlist_node link;
696 void (*irq_acked)(struct kvm_irq_ack_notifier *kian);
699 struct kvm_assigned_dev_kernel {
700 struct kvm_irq_ack_notifier ack_notifier;
701 struct list_head list;
706 unsigned int entries_nr;
708 bool host_irq_disabled;
710 struct msix_entry *host_msix_entries;
712 struct msix_entry *guest_msix_entries;
713 unsigned long irq_requested_type;
718 spinlock_t intx_lock;
719 spinlock_t intx_mask_lock;
721 struct pci_saved_state *pci_saved_state;
724 struct kvm_irq_mask_notifier {
725 void (*func)(struct kvm_irq_mask_notifier *kimn, bool masked);
727 struct hlist_node link;
730 void kvm_register_irq_mask_notifier(struct kvm *kvm, int irq,
731 struct kvm_irq_mask_notifier *kimn);
732 void kvm_unregister_irq_mask_notifier(struct kvm *kvm, int irq,
733 struct kvm_irq_mask_notifier *kimn);
734 void kvm_fire_mask_notifiers(struct kvm *kvm, unsigned irqchip, unsigned pin,
737 int kvm_irq_map_gsi(struct kvm *kvm,
738 struct kvm_kernel_irq_routing_entry *entries, int gsi);
739 int kvm_irq_map_chip_pin(struct kvm *kvm, unsigned irqchip, unsigned pin);
741 int kvm_set_irq(struct kvm *kvm, int irq_source_id, u32 irq, int level,
743 int kvm_set_irq_inatomic(struct kvm *kvm, int irq_source_id, u32 irq, int level);
744 int kvm_set_msi(struct kvm_kernel_irq_routing_entry *irq_entry, struct kvm *kvm,
745 int irq_source_id, int level, bool line_status);
746 bool kvm_irq_has_notifier(struct kvm *kvm, unsigned irqchip, unsigned pin);
747 void kvm_notify_acked_irq(struct kvm *kvm, unsigned irqchip, unsigned pin);
748 void kvm_register_irq_ack_notifier(struct kvm *kvm,
749 struct kvm_irq_ack_notifier *kian);
750 void kvm_unregister_irq_ack_notifier(struct kvm *kvm,
751 struct kvm_irq_ack_notifier *kian);
752 int kvm_request_irq_source_id(struct kvm *kvm);
753 void kvm_free_irq_source_id(struct kvm *kvm, int irq_source_id);
755 #ifdef CONFIG_KVM_DEVICE_ASSIGNMENT
756 int kvm_iommu_map_pages(struct kvm *kvm, struct kvm_memory_slot *slot);
757 void kvm_iommu_unmap_pages(struct kvm *kvm, struct kvm_memory_slot *slot);
758 int kvm_iommu_map_guest(struct kvm *kvm);
759 int kvm_iommu_unmap_guest(struct kvm *kvm);
760 int kvm_assign_device(struct kvm *kvm,
761 struct kvm_assigned_dev_kernel *assigned_dev);
762 int kvm_deassign_device(struct kvm *kvm,
763 struct kvm_assigned_dev_kernel *assigned_dev);
765 static inline int kvm_iommu_map_pages(struct kvm *kvm,
766 struct kvm_memory_slot *slot)
771 static inline void kvm_iommu_unmap_pages(struct kvm *kvm,
772 struct kvm_memory_slot *slot)
776 static inline int kvm_iommu_unmap_guest(struct kvm *kvm)
782 static inline void kvm_guest_enter(void)
786 BUG_ON(preemptible());
788 local_irq_save(flags);
790 local_irq_restore(flags);
792 /* KVM does not hold any references to rcu protected data when it
793 * switches CPU into a guest mode. In fact switching to a guest mode
794 * is very similar to exiting to userspace from rcu point of view. In
795 * addition CPU may stay in a guest mode for quite a long time (up to
796 * one time slice). Lets treat guest mode as quiescent state, just like
797 * we do with user-mode execution.
799 rcu_virt_note_context_switch(smp_processor_id());
802 static inline void kvm_guest_exit(void)
806 local_irq_save(flags);
808 local_irq_restore(flags);
812 * search_memslots() and __gfn_to_memslot() are here because they are
813 * used in non-modular code in arch/powerpc/kvm/book3s_hv_rm_mmu.c.
814 * gfn_to_memslot() itself isn't here as an inline because that would
815 * bloat other code too much.
817 static inline struct kvm_memory_slot *
818 search_memslots(struct kvm_memslots *slots, gfn_t gfn)
820 struct kvm_memory_slot *memslot;
822 kvm_for_each_memslot(memslot, slots)
823 if (gfn >= memslot->base_gfn &&
824 gfn < memslot->base_gfn + memslot->npages)
830 static inline struct kvm_memory_slot *
831 __gfn_to_memslot(struct kvm_memslots *slots, gfn_t gfn)
833 return search_memslots(slots, gfn);
836 static inline unsigned long
837 __gfn_to_hva_memslot(struct kvm_memory_slot *slot, gfn_t gfn)
839 return slot->userspace_addr + (gfn - slot->base_gfn) * PAGE_SIZE;
842 static inline int memslot_id(struct kvm *kvm, gfn_t gfn)
844 return gfn_to_memslot(kvm, gfn)->id;
848 hva_to_gfn_memslot(unsigned long hva, struct kvm_memory_slot *slot)
850 gfn_t gfn_offset = (hva - slot->userspace_addr) >> PAGE_SHIFT;
852 return slot->base_gfn + gfn_offset;
855 static inline gpa_t gfn_to_gpa(gfn_t gfn)
857 return (gpa_t)gfn << PAGE_SHIFT;
860 static inline gfn_t gpa_to_gfn(gpa_t gpa)
862 return (gfn_t)(gpa >> PAGE_SHIFT);
865 static inline hpa_t pfn_to_hpa(pfn_t pfn)
867 return (hpa_t)pfn << PAGE_SHIFT;
870 static inline bool kvm_is_error_gpa(struct kvm *kvm, gpa_t gpa)
872 unsigned long hva = gfn_to_hva(kvm, gpa_to_gfn(gpa));
874 return kvm_is_error_hva(hva);
877 static inline void kvm_migrate_timers(struct kvm_vcpu *vcpu)
879 set_bit(KVM_REQ_MIGRATE_TIMER, &vcpu->requests);
887 struct kvm_stats_debugfs_item {
890 enum kvm_stat_kind kind;
891 struct dentry *dentry;
893 extern struct kvm_stats_debugfs_item debugfs_entries[];
894 extern struct dentry *kvm_debugfs_dir;
896 #if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
897 static inline int mmu_notifier_retry(struct kvm *kvm, unsigned long mmu_seq)
899 if (unlikely(kvm->mmu_notifier_count))
902 * Ensure the read of mmu_notifier_count happens before the read
903 * of mmu_notifier_seq. This interacts with the smp_wmb() in
904 * mmu_notifier_invalidate_range_end to make sure that the caller
905 * either sees the old (non-zero) value of mmu_notifier_count or
906 * the new (incremented) value of mmu_notifier_seq.
907 * PowerPC Book3s HV KVM calls this under a per-page lock
908 * rather than under kvm->mmu_lock, for scalability, so
909 * can't rely on kvm->mmu_lock to keep things ordered.
912 if (kvm->mmu_notifier_seq != mmu_seq)
918 #ifdef CONFIG_HAVE_KVM_IRQ_ROUTING
921 #define KVM_MAX_IRQ_ROUTES 4096 //FIXME: we can have more than that...
923 #define KVM_MAX_IRQ_ROUTES 1024
926 int kvm_setup_default_irq_routing(struct kvm *kvm);
927 int kvm_set_irq_routing(struct kvm *kvm,
928 const struct kvm_irq_routing_entry *entries,
931 int kvm_set_routing_entry(struct kvm_kernel_irq_routing_entry *e,
932 const struct kvm_irq_routing_entry *ue);
933 void kvm_free_irq_routing(struct kvm *kvm);
937 static inline void kvm_free_irq_routing(struct kvm *kvm) {}
941 int kvm_send_userspace_msi(struct kvm *kvm, struct kvm_msi *msi);
943 #ifdef CONFIG_HAVE_KVM_EVENTFD
945 void kvm_eventfd_init(struct kvm *kvm);
946 int kvm_ioeventfd(struct kvm *kvm, struct kvm_ioeventfd *args);
948 #ifdef CONFIG_HAVE_KVM_IRQFD
949 int kvm_irqfd(struct kvm *kvm, struct kvm_irqfd *args);
950 void kvm_irqfd_release(struct kvm *kvm);
951 void kvm_irq_routing_update(struct kvm *);
953 static inline int kvm_irqfd(struct kvm *kvm, struct kvm_irqfd *args)
958 static inline void kvm_irqfd_release(struct kvm *kvm) {}
963 static inline void kvm_eventfd_init(struct kvm *kvm) {}
965 static inline int kvm_irqfd(struct kvm *kvm, struct kvm_irqfd *args)
970 static inline void kvm_irqfd_release(struct kvm *kvm) {}
972 #ifdef CONFIG_HAVE_KVM_IRQCHIP
973 static inline void kvm_irq_routing_update(struct kvm *kvm)
978 static inline int kvm_ioeventfd(struct kvm *kvm, struct kvm_ioeventfd *args)
983 #endif /* CONFIG_HAVE_KVM_EVENTFD */
985 #ifdef CONFIG_KVM_APIC_ARCHITECTURE
986 static inline bool kvm_vcpu_is_bsp(struct kvm_vcpu *vcpu)
988 return vcpu->kvm->bsp_vcpu_id == vcpu->vcpu_id;
991 bool kvm_vcpu_compatible(struct kvm_vcpu *vcpu);
995 static inline bool kvm_vcpu_compatible(struct kvm_vcpu *vcpu) { return true; }
999 #ifdef CONFIG_KVM_DEVICE_ASSIGNMENT
1001 long kvm_vm_ioctl_assigned_device(struct kvm *kvm, unsigned ioctl,
1004 void kvm_free_all_assigned_devices(struct kvm *kvm);
1008 static inline long kvm_vm_ioctl_assigned_device(struct kvm *kvm, unsigned ioctl,
1014 static inline void kvm_free_all_assigned_devices(struct kvm *kvm) {}
1018 static inline void kvm_make_request(int req, struct kvm_vcpu *vcpu)
1020 set_bit(req, &vcpu->requests);
1023 static inline bool kvm_check_request(int req, struct kvm_vcpu *vcpu)
1025 if (test_bit(req, &vcpu->requests)) {
1026 clear_bit(req, &vcpu->requests);
1033 extern bool kvm_rebooting;
1035 struct kvm_device_ops;
1038 struct kvm_device_ops *ops;
1041 struct list_head vm_node;
1044 /* create, destroy, and name are mandatory */
1045 struct kvm_device_ops {
1047 int (*create)(struct kvm_device *dev, u32 type);
1050 * Destroy is responsible for freeing dev.
1052 * Destroy may be called before or after destructors are called
1053 * on emulated I/O regions, depending on whether a reference is
1054 * held by a vcpu or other kvm component that gets destroyed
1055 * after the emulated I/O.
1057 void (*destroy)(struct kvm_device *dev);
1059 int (*set_attr)(struct kvm_device *dev, struct kvm_device_attr *attr);
1060 int (*get_attr)(struct kvm_device *dev, struct kvm_device_attr *attr);
1061 int (*has_attr)(struct kvm_device *dev, struct kvm_device_attr *attr);
1062 long (*ioctl)(struct kvm_device *dev, unsigned int ioctl,
1066 void kvm_device_get(struct kvm_device *dev);
1067 void kvm_device_put(struct kvm_device *dev);
1068 struct kvm_device *kvm_device_from_filp(struct file *filp);
1070 extern struct kvm_device_ops kvm_mpic_ops;
1071 extern struct kvm_device_ops kvm_xics_ops;
1072 extern struct kvm_device_ops kvm_vfio_ops;
1073 extern struct kvm_device_ops kvm_arm_vgic_v2_ops;
1074 extern struct kvm_device_ops kvm_flic_ops;
1076 #ifdef CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT
1078 static inline void kvm_vcpu_set_in_spin_loop(struct kvm_vcpu *vcpu, bool val)
1080 vcpu->spin_loop.in_spin_loop = val;
1082 static inline void kvm_vcpu_set_dy_eligible(struct kvm_vcpu *vcpu, bool val)
1084 vcpu->spin_loop.dy_eligible = val;
1087 #else /* !CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT */
1089 static inline void kvm_vcpu_set_in_spin_loop(struct kvm_vcpu *vcpu, bool val)
1093 static inline void kvm_vcpu_set_dy_eligible(struct kvm_vcpu *vcpu, bool val)
1096 #endif /* CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT */