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 <linux/irqbypass.h>
28 #include <asm/signal.h>
30 #include <linux/kvm.h>
31 #include <linux/kvm_para.h>
33 #include <linux/kvm_types.h>
35 #include <asm/kvm_host.h>
38 * The bit 16 ~ bit 31 of kvm_memory_region::flags are internally used
39 * in kvm, other bits are visible for userspace which are defined in
40 * include/linux/kvm_h.
42 #define KVM_MEMSLOT_INVALID (1UL << 16)
43 #define KVM_MEMSLOT_INCOHERENT (1UL << 17)
45 /* Two fragments for cross MMIO pages. */
46 #define KVM_MAX_MMIO_FRAGMENTS 2
48 #ifndef KVM_ADDRESS_SPACE_NUM
49 #define KVM_ADDRESS_SPACE_NUM 1
53 * For the normal pfn, the highest 12 bits should be zero,
54 * so we can mask bit 62 ~ bit 52 to indicate the error pfn,
55 * mask bit 63 to indicate the noslot pfn.
57 #define KVM_PFN_ERR_MASK (0x7ffULL << 52)
58 #define KVM_PFN_ERR_NOSLOT_MASK (0xfffULL << 52)
59 #define KVM_PFN_NOSLOT (0x1ULL << 63)
61 #define KVM_PFN_ERR_FAULT (KVM_PFN_ERR_MASK)
62 #define KVM_PFN_ERR_HWPOISON (KVM_PFN_ERR_MASK + 1)
63 #define KVM_PFN_ERR_RO_FAULT (KVM_PFN_ERR_MASK + 2)
66 * error pfns indicate that the gfn is in slot but faild to
67 * translate it to pfn on host.
69 static inline bool is_error_pfn(pfn_t pfn)
71 return !!(pfn & KVM_PFN_ERR_MASK);
75 * error_noslot pfns indicate that the gfn can not be
76 * translated to pfn - it is not in slot or failed to
77 * translate it to pfn.
79 static inline bool is_error_noslot_pfn(pfn_t pfn)
81 return !!(pfn & KVM_PFN_ERR_NOSLOT_MASK);
84 /* noslot pfn indicates that the gfn is not in slot. */
85 static inline bool is_noslot_pfn(pfn_t pfn)
87 return pfn == KVM_PFN_NOSLOT;
91 * architectures with KVM_HVA_ERR_BAD other than PAGE_OFFSET (e.g. s390)
92 * provide own defines and kvm_is_error_hva
94 #ifndef KVM_HVA_ERR_BAD
96 #define KVM_HVA_ERR_BAD (PAGE_OFFSET)
97 #define KVM_HVA_ERR_RO_BAD (PAGE_OFFSET + PAGE_SIZE)
99 static inline bool kvm_is_error_hva(unsigned long addr)
101 return addr >= PAGE_OFFSET;
106 #define KVM_ERR_PTR_BAD_PAGE (ERR_PTR(-ENOENT))
108 static inline bool is_error_page(struct page *page)
114 * vcpu->requests bit members
116 #define KVM_REQ_TLB_FLUSH 0
117 #define KVM_REQ_MIGRATE_TIMER 1
118 #define KVM_REQ_REPORT_TPR_ACCESS 2
119 #define KVM_REQ_MMU_RELOAD 3
120 #define KVM_REQ_TRIPLE_FAULT 4
121 #define KVM_REQ_PENDING_TIMER 5
122 #define KVM_REQ_UNHALT 6
123 #define KVM_REQ_MMU_SYNC 7
124 #define KVM_REQ_CLOCK_UPDATE 8
125 #define KVM_REQ_KICK 9
126 #define KVM_REQ_DEACTIVATE_FPU 10
127 #define KVM_REQ_EVENT 11
128 #define KVM_REQ_APF_HALT 12
129 #define KVM_REQ_STEAL_UPDATE 13
130 #define KVM_REQ_NMI 14
131 #define KVM_REQ_PMU 15
132 #define KVM_REQ_PMI 16
133 #define KVM_REQ_WATCHDOG 17
134 #define KVM_REQ_MASTERCLOCK_UPDATE 18
135 #define KVM_REQ_MCLOCK_INPROGRESS 19
136 #define KVM_REQ_EPR_EXIT 20
137 #define KVM_REQ_SCAN_IOAPIC 21
138 #define KVM_REQ_GLOBAL_CLOCK_UPDATE 22
139 #define KVM_REQ_ENABLE_IBS 23
140 #define KVM_REQ_DISABLE_IBS 24
141 #define KVM_REQ_APIC_PAGE_RELOAD 25
142 #define KVM_REQ_SMI 26
143 #define KVM_REQ_HV_CRASH 27
144 #define KVM_REQ_IOAPIC_EOI_EXIT 28
145 #define KVM_REQ_HV_RESET 29
146 #define KVM_REQ_HV_EXIT 30
147 #define KVM_REQ_HV_STIMER 31
149 #define KVM_USERSPACE_IRQ_SOURCE_ID 0
150 #define KVM_IRQFD_RESAMPLE_IRQ_SOURCE_ID 1
152 extern struct kmem_cache *kvm_vcpu_cache;
154 extern spinlock_t kvm_lock;
155 extern struct list_head vm_list;
157 struct kvm_io_range {
160 struct kvm_io_device *dev;
163 #define NR_IOBUS_DEVS 1000
168 struct kvm_io_range range[];
174 KVM_VIRTIO_CCW_NOTIFY_BUS,
179 int kvm_io_bus_write(struct kvm_vcpu *vcpu, enum kvm_bus bus_idx, gpa_t addr,
180 int len, const void *val);
181 int kvm_io_bus_write_cookie(struct kvm_vcpu *vcpu, enum kvm_bus bus_idx,
182 gpa_t addr, int len, const void *val, long cookie);
183 int kvm_io_bus_read(struct kvm_vcpu *vcpu, enum kvm_bus bus_idx, gpa_t addr,
185 int kvm_io_bus_register_dev(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr,
186 int len, struct kvm_io_device *dev);
187 int kvm_io_bus_unregister_dev(struct kvm *kvm, enum kvm_bus bus_idx,
188 struct kvm_io_device *dev);
190 #ifdef CONFIG_KVM_ASYNC_PF
191 struct kvm_async_pf {
192 struct work_struct work;
193 struct list_head link;
194 struct list_head queue;
195 struct kvm_vcpu *vcpu;
196 struct mm_struct *mm;
199 struct kvm_arch_async_pf arch;
203 void kvm_clear_async_pf_completion_queue(struct kvm_vcpu *vcpu);
204 void kvm_check_async_pf_completion(struct kvm_vcpu *vcpu);
205 int kvm_setup_async_pf(struct kvm_vcpu *vcpu, gva_t gva, unsigned long hva,
206 struct kvm_arch_async_pf *arch);
207 int kvm_async_pf_wakeup_all(struct kvm_vcpu *vcpu);
214 READING_SHADOW_PAGE_TABLES,
218 * Sometimes a large or cross-page mmio needs to be broken up into separate
219 * exits for userspace servicing.
221 struct kvm_mmio_fragment {
229 #ifdef CONFIG_PREEMPT_NOTIFIERS
230 struct preempt_notifier preempt_notifier;
236 unsigned long requests;
237 unsigned long guest_debug;
240 struct list_head blocked_vcpu_list;
246 int guest_fpu_loaded, guest_xcr0_loaded;
247 unsigned char fpu_counter;
248 wait_queue_head_t wq;
252 struct kvm_vcpu_stat stat;
253 unsigned int halt_poll_ns;
255 #ifdef CONFIG_HAS_IOMEM
257 int mmio_read_completed;
259 int mmio_cur_fragment;
260 int mmio_nr_fragments;
261 struct kvm_mmio_fragment mmio_fragments[KVM_MAX_MMIO_FRAGMENTS];
264 #ifdef CONFIG_KVM_ASYNC_PF
267 struct list_head queue;
268 struct list_head done;
273 #ifdef CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT
275 * Cpu relax intercept or pause loop exit optimization
276 * in_spin_loop: set when a vcpu does a pause loop exit
277 * or cpu relax intercepted.
278 * dy_eligible: indicates whether vcpu is eligible for directed yield.
286 struct kvm_vcpu_arch arch;
289 static inline int kvm_vcpu_exiting_guest_mode(struct kvm_vcpu *vcpu)
291 return cmpxchg(&vcpu->mode, IN_GUEST_MODE, EXITING_GUEST_MODE);
295 * Some of the bitops functions do not support too long bitmaps.
296 * This number must be determined not to exceed such limits.
298 #define KVM_MEM_MAX_NR_PAGES ((1UL << 31) - 1)
300 struct kvm_memory_slot {
302 unsigned long npages;
303 unsigned long *dirty_bitmap;
304 struct kvm_arch_memory_slot arch;
305 unsigned long userspace_addr;
310 static inline unsigned long kvm_dirty_bitmap_bytes(struct kvm_memory_slot *memslot)
312 return ALIGN(memslot->npages, BITS_PER_LONG) / 8;
315 struct kvm_s390_adapter_int {
328 struct kvm_kernel_irq_routing_entry {
331 int (*set)(struct kvm_kernel_irq_routing_entry *e,
332 struct kvm *kvm, int irq_source_id, int level,
340 struct kvm_s390_adapter_int adapter;
341 struct kvm_hv_sint hv_sint;
343 struct hlist_node link;
346 #ifdef CONFIG_HAVE_KVM_IRQ_ROUTING
347 struct kvm_irq_routing_table {
348 int chip[KVM_NR_IRQCHIPS][KVM_IRQCHIP_NUM_PINS];
351 * Array indexed by gsi. Each entry contains list of irq chips
352 * the gsi is connected to.
354 struct hlist_head map[0];
358 #ifndef KVM_PRIVATE_MEM_SLOTS
359 #define KVM_PRIVATE_MEM_SLOTS 0
362 #ifndef KVM_MEM_SLOTS_NUM
363 #define KVM_MEM_SLOTS_NUM (KVM_USER_MEM_SLOTS + KVM_PRIVATE_MEM_SLOTS)
366 #ifndef __KVM_VCPU_MULTIPLE_ADDRESS_SPACE
367 static inline int kvm_arch_vcpu_memslots_id(struct kvm_vcpu *vcpu)
375 * memslots are not sorted by id anymore, please use id_to_memslot()
376 * to get the memslot by its id.
378 struct kvm_memslots {
380 struct kvm_memory_slot memslots[KVM_MEM_SLOTS_NUM];
381 /* The mapping table from slot id to the index in memslots[]. */
382 short id_to_index[KVM_MEM_SLOTS_NUM];
389 struct mutex slots_lock;
390 struct mm_struct *mm; /* userspace tied to this vm */
391 struct kvm_memslots *memslots[KVM_ADDRESS_SPACE_NUM];
392 struct srcu_struct srcu;
393 struct srcu_struct irq_srcu;
394 struct kvm_vcpu *vcpus[KVM_MAX_VCPUS];
395 atomic_t online_vcpus;
396 int last_boosted_vcpu;
397 struct list_head vm_list;
399 struct kvm_io_bus *buses[KVM_NR_BUSES];
400 #ifdef CONFIG_HAVE_KVM_EVENTFD
403 struct list_head items;
404 struct list_head resampler_list;
405 struct mutex resampler_lock;
407 struct list_head ioeventfds;
409 struct kvm_vm_stat stat;
410 struct kvm_arch arch;
411 atomic_t users_count;
412 #ifdef KVM_COALESCED_MMIO_PAGE_OFFSET
413 struct kvm_coalesced_mmio_ring *coalesced_mmio_ring;
414 spinlock_t ring_lock;
415 struct list_head coalesced_zones;
418 struct mutex irq_lock;
419 #ifdef CONFIG_HAVE_KVM_IRQCHIP
421 * Update side is protected by irq_lock.
423 struct kvm_irq_routing_table __rcu *irq_routing;
425 #ifdef CONFIG_HAVE_KVM_IRQFD
426 struct hlist_head irq_ack_notifier_list;
429 #if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
430 struct mmu_notifier mmu_notifier;
431 unsigned long mmu_notifier_seq;
432 long mmu_notifier_count;
435 struct list_head devices;
438 #define kvm_err(fmt, ...) \
439 pr_err("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
440 #define kvm_info(fmt, ...) \
441 pr_info("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
442 #define kvm_debug(fmt, ...) \
443 pr_debug("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
444 #define kvm_pr_unimpl(fmt, ...) \
445 pr_err_ratelimited("kvm [%i]: " fmt, \
446 task_tgid_nr(current), ## __VA_ARGS__)
448 /* The guest did something we don't support. */
449 #define vcpu_unimpl(vcpu, fmt, ...) \
450 kvm_pr_unimpl("vcpu%i, guest rIP: 0x%lx " fmt, \
451 (vcpu)->vcpu_id, kvm_rip_read(vcpu), ## __VA_ARGS__)
453 #define vcpu_debug(vcpu, fmt, ...) \
454 kvm_debug("vcpu%i " fmt, (vcpu)->vcpu_id, ## __VA_ARGS__)
455 #define vcpu_err(vcpu, fmt, ...) \
456 kvm_err("vcpu%i " fmt, (vcpu)->vcpu_id, ## __VA_ARGS__)
458 static inline struct kvm_vcpu *kvm_get_vcpu(struct kvm *kvm, int i)
460 /* Pairs with smp_wmb() in kvm_vm_ioctl_create_vcpu, in case
461 * the caller has read kvm->online_vcpus before (as is the case
462 * for kvm_for_each_vcpu, for example).
465 return kvm->vcpus[i];
468 #define kvm_for_each_vcpu(idx, vcpup, kvm) \
470 idx < atomic_read(&kvm->online_vcpus) && \
471 (vcpup = kvm_get_vcpu(kvm, idx)) != NULL; \
474 static inline struct kvm_vcpu *kvm_get_vcpu_by_id(struct kvm *kvm, int id)
476 struct kvm_vcpu *vcpu;
479 if (id < 0 || id >= KVM_MAX_VCPUS)
481 vcpu = kvm_get_vcpu(kvm, id);
482 if (vcpu && vcpu->vcpu_id == id)
484 kvm_for_each_vcpu(i, vcpu, kvm)
485 if (vcpu->vcpu_id == id)
490 #define kvm_for_each_memslot(memslot, slots) \
491 for (memslot = &slots->memslots[0]; \
492 memslot < slots->memslots + KVM_MEM_SLOTS_NUM && memslot->npages;\
495 int kvm_vcpu_init(struct kvm_vcpu *vcpu, struct kvm *kvm, unsigned id);
496 void kvm_vcpu_uninit(struct kvm_vcpu *vcpu);
498 int __must_check vcpu_load(struct kvm_vcpu *vcpu);
499 void vcpu_put(struct kvm_vcpu *vcpu);
501 #ifdef __KVM_HAVE_IOAPIC
502 void kvm_vcpu_request_scan_ioapic(struct kvm *kvm);
503 void kvm_arch_post_irq_routing_update(struct kvm *kvm);
505 static inline void kvm_vcpu_request_scan_ioapic(struct kvm *kvm)
508 static inline void kvm_arch_post_irq_routing_update(struct kvm *kvm)
513 #ifdef CONFIG_HAVE_KVM_IRQFD
514 int kvm_irqfd_init(void);
515 void kvm_irqfd_exit(void);
517 static inline int kvm_irqfd_init(void)
522 static inline void kvm_irqfd_exit(void)
526 int kvm_init(void *opaque, unsigned vcpu_size, unsigned vcpu_align,
527 struct module *module);
530 void kvm_get_kvm(struct kvm *kvm);
531 void kvm_put_kvm(struct kvm *kvm);
533 static inline struct kvm_memslots *__kvm_memslots(struct kvm *kvm, int as_id)
535 return rcu_dereference_check(kvm->memslots[as_id],
536 srcu_read_lock_held(&kvm->srcu)
537 || lockdep_is_held(&kvm->slots_lock));
540 static inline struct kvm_memslots *kvm_memslots(struct kvm *kvm)
542 return __kvm_memslots(kvm, 0);
545 static inline struct kvm_memslots *kvm_vcpu_memslots(struct kvm_vcpu *vcpu)
547 int as_id = kvm_arch_vcpu_memslots_id(vcpu);
549 return __kvm_memslots(vcpu->kvm, as_id);
552 static inline struct kvm_memory_slot *
553 id_to_memslot(struct kvm_memslots *slots, int id)
555 int index = slots->id_to_index[id];
556 struct kvm_memory_slot *slot;
558 slot = &slots->memslots[index];
560 WARN_ON(slot->id != id);
565 * KVM_SET_USER_MEMORY_REGION ioctl allows the following operations:
566 * - create a new memory slot
567 * - delete an existing memory slot
568 * - modify an existing memory slot
569 * -- move it in the guest physical memory space
570 * -- just change its flags
572 * Since flags can be changed by some of these operations, the following
573 * differentiation is the best we can do for __kvm_set_memory_region():
582 int kvm_set_memory_region(struct kvm *kvm,
583 const struct kvm_userspace_memory_region *mem);
584 int __kvm_set_memory_region(struct kvm *kvm,
585 const struct kvm_userspace_memory_region *mem);
586 void kvm_arch_free_memslot(struct kvm *kvm, struct kvm_memory_slot *free,
587 struct kvm_memory_slot *dont);
588 int kvm_arch_create_memslot(struct kvm *kvm, struct kvm_memory_slot *slot,
589 unsigned long npages);
590 void kvm_arch_memslots_updated(struct kvm *kvm, struct kvm_memslots *slots);
591 int kvm_arch_prepare_memory_region(struct kvm *kvm,
592 struct kvm_memory_slot *memslot,
593 const struct kvm_userspace_memory_region *mem,
594 enum kvm_mr_change change);
595 void kvm_arch_commit_memory_region(struct kvm *kvm,
596 const struct kvm_userspace_memory_region *mem,
597 const struct kvm_memory_slot *old,
598 const struct kvm_memory_slot *new,
599 enum kvm_mr_change change);
600 bool kvm_largepages_enabled(void);
601 void kvm_disable_largepages(void);
602 /* flush all memory translations */
603 void kvm_arch_flush_shadow_all(struct kvm *kvm);
604 /* flush memory translations pointing to 'slot' */
605 void kvm_arch_flush_shadow_memslot(struct kvm *kvm,
606 struct kvm_memory_slot *slot);
608 int gfn_to_page_many_atomic(struct kvm_memory_slot *slot, gfn_t gfn,
609 struct page **pages, int nr_pages);
611 struct page *gfn_to_page(struct kvm *kvm, gfn_t gfn);
612 unsigned long gfn_to_hva(struct kvm *kvm, gfn_t gfn);
613 unsigned long gfn_to_hva_prot(struct kvm *kvm, gfn_t gfn, bool *writable);
614 unsigned long gfn_to_hva_memslot(struct kvm_memory_slot *slot, gfn_t gfn);
615 unsigned long gfn_to_hva_memslot_prot(struct kvm_memory_slot *slot, gfn_t gfn,
617 void kvm_release_page_clean(struct page *page);
618 void kvm_release_page_dirty(struct page *page);
619 void kvm_set_page_accessed(struct page *page);
621 pfn_t gfn_to_pfn_atomic(struct kvm *kvm, gfn_t gfn);
622 pfn_t gfn_to_pfn(struct kvm *kvm, gfn_t gfn);
623 pfn_t gfn_to_pfn_prot(struct kvm *kvm, gfn_t gfn, bool write_fault,
625 pfn_t gfn_to_pfn_memslot(struct kvm_memory_slot *slot, gfn_t gfn);
626 pfn_t gfn_to_pfn_memslot_atomic(struct kvm_memory_slot *slot, gfn_t gfn);
627 pfn_t __gfn_to_pfn_memslot(struct kvm_memory_slot *slot, gfn_t gfn, bool atomic,
628 bool *async, bool write_fault, bool *writable);
630 void kvm_release_pfn_clean(pfn_t pfn);
631 void kvm_set_pfn_dirty(pfn_t pfn);
632 void kvm_set_pfn_accessed(pfn_t pfn);
633 void kvm_get_pfn(pfn_t pfn);
635 int kvm_read_guest_page(struct kvm *kvm, gfn_t gfn, void *data, int offset,
637 int kvm_read_guest_atomic(struct kvm *kvm, gpa_t gpa, void *data,
639 int kvm_read_guest(struct kvm *kvm, gpa_t gpa, void *data, unsigned long len);
640 int kvm_read_guest_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
641 void *data, unsigned long len);
642 int kvm_write_guest_page(struct kvm *kvm, gfn_t gfn, const void *data,
643 int offset, int len);
644 int kvm_write_guest(struct kvm *kvm, gpa_t gpa, const void *data,
646 int kvm_write_guest_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
647 void *data, unsigned long len);
648 int kvm_gfn_to_hva_cache_init(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
649 gpa_t gpa, unsigned long len);
650 int kvm_clear_guest_page(struct kvm *kvm, gfn_t gfn, int offset, int len);
651 int kvm_clear_guest(struct kvm *kvm, gpa_t gpa, unsigned long len);
652 struct kvm_memory_slot *gfn_to_memslot(struct kvm *kvm, gfn_t gfn);
653 bool kvm_is_visible_gfn(struct kvm *kvm, gfn_t gfn);
654 unsigned long kvm_host_page_size(struct kvm *kvm, gfn_t gfn);
655 void mark_page_dirty(struct kvm *kvm, gfn_t gfn);
657 struct kvm_memslots *kvm_vcpu_memslots(struct kvm_vcpu *vcpu);
658 struct kvm_memory_slot *kvm_vcpu_gfn_to_memslot(struct kvm_vcpu *vcpu, gfn_t gfn);
659 pfn_t kvm_vcpu_gfn_to_pfn_atomic(struct kvm_vcpu *vcpu, gfn_t gfn);
660 pfn_t kvm_vcpu_gfn_to_pfn(struct kvm_vcpu *vcpu, gfn_t gfn);
661 struct page *kvm_vcpu_gfn_to_page(struct kvm_vcpu *vcpu, gfn_t gfn);
662 unsigned long kvm_vcpu_gfn_to_hva(struct kvm_vcpu *vcpu, gfn_t gfn);
663 unsigned long kvm_vcpu_gfn_to_hva_prot(struct kvm_vcpu *vcpu, gfn_t gfn, bool *writable);
664 int kvm_vcpu_read_guest_page(struct kvm_vcpu *vcpu, gfn_t gfn, void *data, int offset,
666 int kvm_vcpu_read_guest_atomic(struct kvm_vcpu *vcpu, gpa_t gpa, void *data,
668 int kvm_vcpu_read_guest(struct kvm_vcpu *vcpu, gpa_t gpa, void *data,
670 int kvm_vcpu_write_guest_page(struct kvm_vcpu *vcpu, gfn_t gfn, const void *data,
671 int offset, int len);
672 int kvm_vcpu_write_guest(struct kvm_vcpu *vcpu, gpa_t gpa, const void *data,
674 void kvm_vcpu_mark_page_dirty(struct kvm_vcpu *vcpu, gfn_t gfn);
676 void kvm_vcpu_block(struct kvm_vcpu *vcpu);
677 void kvm_arch_vcpu_blocking(struct kvm_vcpu *vcpu);
678 void kvm_arch_vcpu_unblocking(struct kvm_vcpu *vcpu);
679 void kvm_vcpu_kick(struct kvm_vcpu *vcpu);
680 int kvm_vcpu_yield_to(struct kvm_vcpu *target);
681 void kvm_vcpu_on_spin(struct kvm_vcpu *vcpu);
682 void kvm_load_guest_fpu(struct kvm_vcpu *vcpu);
683 void kvm_put_guest_fpu(struct kvm_vcpu *vcpu);
685 void kvm_flush_remote_tlbs(struct kvm *kvm);
686 void kvm_reload_remote_mmus(struct kvm *kvm);
687 void kvm_make_mclock_inprogress_request(struct kvm *kvm);
688 void kvm_make_scan_ioapic_request(struct kvm *kvm);
689 bool kvm_make_all_cpus_request(struct kvm *kvm, unsigned int req);
691 long kvm_arch_dev_ioctl(struct file *filp,
692 unsigned int ioctl, unsigned long arg);
693 long kvm_arch_vcpu_ioctl(struct file *filp,
694 unsigned int ioctl, unsigned long arg);
695 int kvm_arch_vcpu_fault(struct kvm_vcpu *vcpu, struct vm_fault *vmf);
697 int kvm_vm_ioctl_check_extension(struct kvm *kvm, long ext);
699 int kvm_get_dirty_log(struct kvm *kvm,
700 struct kvm_dirty_log *log, int *is_dirty);
702 int kvm_get_dirty_log_protect(struct kvm *kvm,
703 struct kvm_dirty_log *log, bool *is_dirty);
705 void kvm_arch_mmu_enable_log_dirty_pt_masked(struct kvm *kvm,
706 struct kvm_memory_slot *slot,
710 int kvm_vm_ioctl_get_dirty_log(struct kvm *kvm,
711 struct kvm_dirty_log *log);
713 int kvm_vm_ioctl_irq_line(struct kvm *kvm, struct kvm_irq_level *irq_level,
715 long kvm_arch_vm_ioctl(struct file *filp,
716 unsigned int ioctl, unsigned long arg);
718 int kvm_arch_vcpu_ioctl_get_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu);
719 int kvm_arch_vcpu_ioctl_set_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu);
721 int kvm_arch_vcpu_ioctl_translate(struct kvm_vcpu *vcpu,
722 struct kvm_translation *tr);
724 int kvm_arch_vcpu_ioctl_get_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs);
725 int kvm_arch_vcpu_ioctl_set_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs);
726 int kvm_arch_vcpu_ioctl_get_sregs(struct kvm_vcpu *vcpu,
727 struct kvm_sregs *sregs);
728 int kvm_arch_vcpu_ioctl_set_sregs(struct kvm_vcpu *vcpu,
729 struct kvm_sregs *sregs);
730 int kvm_arch_vcpu_ioctl_get_mpstate(struct kvm_vcpu *vcpu,
731 struct kvm_mp_state *mp_state);
732 int kvm_arch_vcpu_ioctl_set_mpstate(struct kvm_vcpu *vcpu,
733 struct kvm_mp_state *mp_state);
734 int kvm_arch_vcpu_ioctl_set_guest_debug(struct kvm_vcpu *vcpu,
735 struct kvm_guest_debug *dbg);
736 int kvm_arch_vcpu_ioctl_run(struct kvm_vcpu *vcpu, struct kvm_run *kvm_run);
738 int kvm_arch_init(void *opaque);
739 void kvm_arch_exit(void);
741 int kvm_arch_vcpu_init(struct kvm_vcpu *vcpu);
742 void kvm_arch_vcpu_uninit(struct kvm_vcpu *vcpu);
744 void kvm_arch_sched_in(struct kvm_vcpu *vcpu, int cpu);
746 void kvm_arch_vcpu_free(struct kvm_vcpu *vcpu);
747 void kvm_arch_vcpu_load(struct kvm_vcpu *vcpu, int cpu);
748 void kvm_arch_vcpu_put(struct kvm_vcpu *vcpu);
749 struct kvm_vcpu *kvm_arch_vcpu_create(struct kvm *kvm, unsigned int id);
750 int kvm_arch_vcpu_setup(struct kvm_vcpu *vcpu);
751 void kvm_arch_vcpu_postcreate(struct kvm_vcpu *vcpu);
752 void kvm_arch_vcpu_destroy(struct kvm_vcpu *vcpu);
754 int kvm_arch_hardware_enable(void);
755 void kvm_arch_hardware_disable(void);
756 int kvm_arch_hardware_setup(void);
757 void kvm_arch_hardware_unsetup(void);
758 void kvm_arch_check_processor_compat(void *rtn);
759 int kvm_arch_vcpu_runnable(struct kvm_vcpu *vcpu);
760 int kvm_arch_vcpu_should_kick(struct kvm_vcpu *vcpu);
762 void *kvm_kvzalloc(unsigned long size);
764 #ifndef __KVM_HAVE_ARCH_VM_ALLOC
765 static inline struct kvm *kvm_arch_alloc_vm(void)
767 return kzalloc(sizeof(struct kvm), GFP_KERNEL);
770 static inline void kvm_arch_free_vm(struct kvm *kvm)
776 #ifdef __KVM_HAVE_ARCH_NONCOHERENT_DMA
777 void kvm_arch_register_noncoherent_dma(struct kvm *kvm);
778 void kvm_arch_unregister_noncoherent_dma(struct kvm *kvm);
779 bool kvm_arch_has_noncoherent_dma(struct kvm *kvm);
781 static inline void kvm_arch_register_noncoherent_dma(struct kvm *kvm)
785 static inline void kvm_arch_unregister_noncoherent_dma(struct kvm *kvm)
789 static inline bool kvm_arch_has_noncoherent_dma(struct kvm *kvm)
794 #ifdef __KVM_HAVE_ARCH_ASSIGNED_DEVICE
795 void kvm_arch_start_assignment(struct kvm *kvm);
796 void kvm_arch_end_assignment(struct kvm *kvm);
797 bool kvm_arch_has_assigned_device(struct kvm *kvm);
799 static inline void kvm_arch_start_assignment(struct kvm *kvm)
803 static inline void kvm_arch_end_assignment(struct kvm *kvm)
807 static inline bool kvm_arch_has_assigned_device(struct kvm *kvm)
813 static inline wait_queue_head_t *kvm_arch_vcpu_wq(struct kvm_vcpu *vcpu)
815 #ifdef __KVM_HAVE_ARCH_WQP
816 return vcpu->arch.wqp;
822 #ifdef __KVM_HAVE_ARCH_INTC_INITIALIZED
824 * returns true if the virtual interrupt controller is initialized and
825 * ready to accept virtual IRQ. On some architectures the virtual interrupt
826 * controller is dynamically instantiated and this is not always true.
828 bool kvm_arch_intc_initialized(struct kvm *kvm);
830 static inline bool kvm_arch_intc_initialized(struct kvm *kvm)
836 int kvm_arch_init_vm(struct kvm *kvm, unsigned long type);
837 void kvm_arch_destroy_vm(struct kvm *kvm);
838 void kvm_arch_sync_events(struct kvm *kvm);
840 int kvm_cpu_has_pending_timer(struct kvm_vcpu *vcpu);
841 void kvm_vcpu_kick(struct kvm_vcpu *vcpu);
843 bool kvm_is_reserved_pfn(pfn_t pfn);
845 struct kvm_irq_ack_notifier {
846 struct hlist_node link;
848 void (*irq_acked)(struct kvm_irq_ack_notifier *kian);
851 int kvm_irq_map_gsi(struct kvm *kvm,
852 struct kvm_kernel_irq_routing_entry *entries, int gsi);
853 int kvm_irq_map_chip_pin(struct kvm *kvm, unsigned irqchip, unsigned pin);
855 int kvm_set_irq(struct kvm *kvm, int irq_source_id, u32 irq, int level,
857 int kvm_set_msi(struct kvm_kernel_irq_routing_entry *irq_entry, struct kvm *kvm,
858 int irq_source_id, int level, bool line_status);
859 int kvm_arch_set_irq_inatomic(struct kvm_kernel_irq_routing_entry *e,
860 struct kvm *kvm, int irq_source_id,
861 int level, bool line_status);
862 bool kvm_irq_has_notifier(struct kvm *kvm, unsigned irqchip, unsigned pin);
863 void kvm_notify_acked_gsi(struct kvm *kvm, int gsi);
864 void kvm_notify_acked_irq(struct kvm *kvm, unsigned irqchip, unsigned pin);
865 void kvm_register_irq_ack_notifier(struct kvm *kvm,
866 struct kvm_irq_ack_notifier *kian);
867 void kvm_unregister_irq_ack_notifier(struct kvm *kvm,
868 struct kvm_irq_ack_notifier *kian);
869 int kvm_request_irq_source_id(struct kvm *kvm);
870 void kvm_free_irq_source_id(struct kvm *kvm, int irq_source_id);
872 #ifdef CONFIG_KVM_DEVICE_ASSIGNMENT
873 int kvm_iommu_map_pages(struct kvm *kvm, struct kvm_memory_slot *slot);
874 void kvm_iommu_unmap_pages(struct kvm *kvm, struct kvm_memory_slot *slot);
876 static inline int kvm_iommu_map_pages(struct kvm *kvm,
877 struct kvm_memory_slot *slot)
882 static inline void kvm_iommu_unmap_pages(struct kvm *kvm,
883 struct kvm_memory_slot *slot)
888 /* must be called with irqs disabled */
889 static inline void __kvm_guest_enter(void)
892 /* KVM does not hold any references to rcu protected data when it
893 * switches CPU into a guest mode. In fact switching to a guest mode
894 * is very similar to exiting to userspace from rcu point of view. In
895 * addition CPU may stay in a guest mode for quite a long time (up to
896 * one time slice). Lets treat guest mode as quiescent state, just like
897 * we do with user-mode execution.
899 if (!context_tracking_cpu_is_enabled())
900 rcu_virt_note_context_switch(smp_processor_id());
903 /* must be called with irqs disabled */
904 static inline void __kvm_guest_exit(void)
909 static inline void kvm_guest_enter(void)
913 local_irq_save(flags);
915 local_irq_restore(flags);
918 static inline void kvm_guest_exit(void)
922 local_irq_save(flags);
924 local_irq_restore(flags);
928 * search_memslots() and __gfn_to_memslot() are here because they are
929 * used in non-modular code in arch/powerpc/kvm/book3s_hv_rm_mmu.c.
930 * gfn_to_memslot() itself isn't here as an inline because that would
931 * bloat other code too much.
933 static inline struct kvm_memory_slot *
934 search_memslots(struct kvm_memslots *slots, gfn_t gfn)
936 int start = 0, end = slots->used_slots;
937 int slot = atomic_read(&slots->lru_slot);
938 struct kvm_memory_slot *memslots = slots->memslots;
940 if (gfn >= memslots[slot].base_gfn &&
941 gfn < memslots[slot].base_gfn + memslots[slot].npages)
942 return &memslots[slot];
944 while (start < end) {
945 slot = start + (end - start) / 2;
947 if (gfn >= memslots[slot].base_gfn)
953 if (gfn >= memslots[start].base_gfn &&
954 gfn < memslots[start].base_gfn + memslots[start].npages) {
955 atomic_set(&slots->lru_slot, start);
956 return &memslots[start];
962 static inline struct kvm_memory_slot *
963 __gfn_to_memslot(struct kvm_memslots *slots, gfn_t gfn)
965 return search_memslots(slots, gfn);
968 static inline unsigned long
969 __gfn_to_hva_memslot(struct kvm_memory_slot *slot, gfn_t gfn)
971 return slot->userspace_addr + (gfn - slot->base_gfn) * PAGE_SIZE;
974 static inline int memslot_id(struct kvm *kvm, gfn_t gfn)
976 return gfn_to_memslot(kvm, gfn)->id;
980 hva_to_gfn_memslot(unsigned long hva, struct kvm_memory_slot *slot)
982 gfn_t gfn_offset = (hva - slot->userspace_addr) >> PAGE_SHIFT;
984 return slot->base_gfn + gfn_offset;
987 static inline gpa_t gfn_to_gpa(gfn_t gfn)
989 return (gpa_t)gfn << PAGE_SHIFT;
992 static inline gfn_t gpa_to_gfn(gpa_t gpa)
994 return (gfn_t)(gpa >> PAGE_SHIFT);
997 static inline hpa_t pfn_to_hpa(pfn_t pfn)
999 return (hpa_t)pfn << PAGE_SHIFT;
1002 static inline bool kvm_is_error_gpa(struct kvm *kvm, gpa_t gpa)
1004 unsigned long hva = gfn_to_hva(kvm, gpa_to_gfn(gpa));
1006 return kvm_is_error_hva(hva);
1009 static inline void kvm_migrate_timers(struct kvm_vcpu *vcpu)
1011 set_bit(KVM_REQ_MIGRATE_TIMER, &vcpu->requests);
1014 enum kvm_stat_kind {
1019 struct kvm_stats_debugfs_item {
1022 enum kvm_stat_kind kind;
1024 extern struct kvm_stats_debugfs_item debugfs_entries[];
1025 extern struct dentry *kvm_debugfs_dir;
1027 #if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
1028 static inline int mmu_notifier_retry(struct kvm *kvm, unsigned long mmu_seq)
1030 if (unlikely(kvm->mmu_notifier_count))
1033 * Ensure the read of mmu_notifier_count happens before the read
1034 * of mmu_notifier_seq. This interacts with the smp_wmb() in
1035 * mmu_notifier_invalidate_range_end to make sure that the caller
1036 * either sees the old (non-zero) value of mmu_notifier_count or
1037 * the new (incremented) value of mmu_notifier_seq.
1038 * PowerPC Book3s HV KVM calls this under a per-page lock
1039 * rather than under kvm->mmu_lock, for scalability, so
1040 * can't rely on kvm->mmu_lock to keep things ordered.
1043 if (kvm->mmu_notifier_seq != mmu_seq)
1049 #ifdef CONFIG_HAVE_KVM_IRQ_ROUTING
1052 #define KVM_MAX_IRQ_ROUTES 4096 //FIXME: we can have more than that...
1054 #define KVM_MAX_IRQ_ROUTES 1024
1057 int kvm_setup_default_irq_routing(struct kvm *kvm);
1058 int kvm_setup_empty_irq_routing(struct kvm *kvm);
1059 int kvm_set_irq_routing(struct kvm *kvm,
1060 const struct kvm_irq_routing_entry *entries,
1063 int kvm_set_routing_entry(struct kvm_kernel_irq_routing_entry *e,
1064 const struct kvm_irq_routing_entry *ue);
1065 void kvm_free_irq_routing(struct kvm *kvm);
1069 static inline void kvm_free_irq_routing(struct kvm *kvm) {}
1073 int kvm_send_userspace_msi(struct kvm *kvm, struct kvm_msi *msi);
1075 #ifdef CONFIG_HAVE_KVM_EVENTFD
1077 void kvm_eventfd_init(struct kvm *kvm);
1078 int kvm_ioeventfd(struct kvm *kvm, struct kvm_ioeventfd *args);
1080 #ifdef CONFIG_HAVE_KVM_IRQFD
1081 int kvm_irqfd(struct kvm *kvm, struct kvm_irqfd *args);
1082 void kvm_irqfd_release(struct kvm *kvm);
1083 void kvm_irq_routing_update(struct kvm *);
1085 static inline int kvm_irqfd(struct kvm *kvm, struct kvm_irqfd *args)
1090 static inline void kvm_irqfd_release(struct kvm *kvm) {}
1095 static inline void kvm_eventfd_init(struct kvm *kvm) {}
1097 static inline int kvm_irqfd(struct kvm *kvm, struct kvm_irqfd *args)
1102 static inline void kvm_irqfd_release(struct kvm *kvm) {}
1104 #ifdef CONFIG_HAVE_KVM_IRQCHIP
1105 static inline void kvm_irq_routing_update(struct kvm *kvm)
1109 void kvm_arch_irq_routing_update(struct kvm *kvm);
1111 static inline int kvm_ioeventfd(struct kvm *kvm, struct kvm_ioeventfd *args)
1116 #endif /* CONFIG_HAVE_KVM_EVENTFD */
1118 #ifdef CONFIG_KVM_APIC_ARCHITECTURE
1119 bool kvm_vcpu_compatible(struct kvm_vcpu *vcpu);
1121 static inline bool kvm_vcpu_compatible(struct kvm_vcpu *vcpu) { return true; }
1124 static inline void kvm_make_request(int req, struct kvm_vcpu *vcpu)
1126 set_bit(req, &vcpu->requests);
1129 static inline bool kvm_check_request(int req, struct kvm_vcpu *vcpu)
1131 if (test_bit(req, &vcpu->requests)) {
1132 clear_bit(req, &vcpu->requests);
1139 extern bool kvm_rebooting;
1142 struct kvm_device_ops *ops;
1145 struct list_head vm_node;
1148 /* create, destroy, and name are mandatory */
1149 struct kvm_device_ops {
1151 int (*create)(struct kvm_device *dev, u32 type);
1154 * Destroy is responsible for freeing dev.
1156 * Destroy may be called before or after destructors are called
1157 * on emulated I/O regions, depending on whether a reference is
1158 * held by a vcpu or other kvm component that gets destroyed
1159 * after the emulated I/O.
1161 void (*destroy)(struct kvm_device *dev);
1163 int (*set_attr)(struct kvm_device *dev, struct kvm_device_attr *attr);
1164 int (*get_attr)(struct kvm_device *dev, struct kvm_device_attr *attr);
1165 int (*has_attr)(struct kvm_device *dev, struct kvm_device_attr *attr);
1166 long (*ioctl)(struct kvm_device *dev, unsigned int ioctl,
1170 void kvm_device_get(struct kvm_device *dev);
1171 void kvm_device_put(struct kvm_device *dev);
1172 struct kvm_device *kvm_device_from_filp(struct file *filp);
1173 int kvm_register_device_ops(struct kvm_device_ops *ops, u32 type);
1174 void kvm_unregister_device_ops(u32 type);
1176 extern struct kvm_device_ops kvm_mpic_ops;
1177 extern struct kvm_device_ops kvm_xics_ops;
1178 extern struct kvm_device_ops kvm_arm_vgic_v2_ops;
1179 extern struct kvm_device_ops kvm_arm_vgic_v3_ops;
1181 #ifdef CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT
1183 static inline void kvm_vcpu_set_in_spin_loop(struct kvm_vcpu *vcpu, bool val)
1185 vcpu->spin_loop.in_spin_loop = val;
1187 static inline void kvm_vcpu_set_dy_eligible(struct kvm_vcpu *vcpu, bool val)
1189 vcpu->spin_loop.dy_eligible = val;
1192 #else /* !CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT */
1194 static inline void kvm_vcpu_set_in_spin_loop(struct kvm_vcpu *vcpu, bool val)
1198 static inline void kvm_vcpu_set_dy_eligible(struct kvm_vcpu *vcpu, bool val)
1201 #endif /* CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT */
1203 #ifdef CONFIG_HAVE_KVM_IRQ_BYPASS
1204 int kvm_arch_irq_bypass_add_producer(struct irq_bypass_consumer *,
1205 struct irq_bypass_producer *);
1206 void kvm_arch_irq_bypass_del_producer(struct irq_bypass_consumer *,
1207 struct irq_bypass_producer *);
1208 void kvm_arch_irq_bypass_stop(struct irq_bypass_consumer *);
1209 void kvm_arch_irq_bypass_start(struct irq_bypass_consumer *);
1210 int kvm_arch_update_irqfd_routing(struct kvm *kvm, unsigned int host_irq,
1211 uint32_t guest_irq, bool set);
1212 #endif /* CONFIG_HAVE_KVM_IRQ_BYPASS */