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_MMU_RELOAD 3
118 #define KVM_REQ_PENDING_TIMER 5
119 #define KVM_REQ_UNHALT 6
121 /* x86-specific requests */
122 #define KVM_REQ_MIGRATE_TIMER 1
123 #define KVM_REQ_REPORT_TPR_ACCESS 2
124 #define KVM_REQ_TRIPLE_FAULT 4
125 #define KVM_REQ_MMU_SYNC 7
126 #define KVM_REQ_CLOCK_UPDATE 8
128 #define KVM_REQ_DEACTIVATE_FPU 10
129 #define KVM_REQ_EVENT 11
130 #define KVM_REQ_APF_HALT 12
131 #define KVM_REQ_STEAL_UPDATE 13
132 #define KVM_REQ_NMI 14
133 #define KVM_REQ_PMU 15
134 #define KVM_REQ_PMI 16
135 #define KVM_REQ_MASTERCLOCK_UPDATE 18
136 #define KVM_REQ_MCLOCK_INPROGRESS 19
137 #define KVM_REQ_SCAN_IOAPIC 21
138 #define KVM_REQ_GLOBAL_CLOCK_UPDATE 22
139 #define KVM_REQ_APIC_PAGE_RELOAD 25
140 #define KVM_REQ_SMI 26
141 #define KVM_REQ_HV_CRASH 27
142 #define KVM_REQ_IOAPIC_EOI_EXIT 28
143 #define KVM_REQ_HV_RESET 29
144 #define KVM_REQ_HV_EXIT 30
145 #define KVM_REQ_HV_STIMER 31
147 /* PPC-specific requests */
148 #define KVM_REQ_WATCHDOG 17
149 #define KVM_REQ_EPR_EXIT 20
151 /* s390-specific requests */
152 #define KVM_REQ_ENABLE_IBS 23
153 #define KVM_REQ_DISABLE_IBS 24
155 #define KVM_USERSPACE_IRQ_SOURCE_ID 0
156 #define KVM_IRQFD_RESAMPLE_IRQ_SOURCE_ID 1
158 extern struct kmem_cache *kvm_vcpu_cache;
160 extern spinlock_t kvm_lock;
161 extern struct list_head vm_list;
163 struct kvm_io_range {
166 struct kvm_io_device *dev;
169 #define NR_IOBUS_DEVS 1000
174 struct kvm_io_range range[];
180 KVM_VIRTIO_CCW_NOTIFY_BUS,
185 int kvm_io_bus_write(struct kvm_vcpu *vcpu, enum kvm_bus bus_idx, gpa_t addr,
186 int len, const void *val);
187 int kvm_io_bus_write_cookie(struct kvm_vcpu *vcpu, enum kvm_bus bus_idx,
188 gpa_t addr, int len, const void *val, long cookie);
189 int kvm_io_bus_read(struct kvm_vcpu *vcpu, enum kvm_bus bus_idx, gpa_t addr,
191 int kvm_io_bus_register_dev(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr,
192 int len, struct kvm_io_device *dev);
193 int kvm_io_bus_unregister_dev(struct kvm *kvm, enum kvm_bus bus_idx,
194 struct kvm_io_device *dev);
196 #ifdef CONFIG_KVM_ASYNC_PF
197 struct kvm_async_pf {
198 struct work_struct work;
199 struct list_head link;
200 struct list_head queue;
201 struct kvm_vcpu *vcpu;
202 struct mm_struct *mm;
205 struct kvm_arch_async_pf arch;
209 void kvm_clear_async_pf_completion_queue(struct kvm_vcpu *vcpu);
210 void kvm_check_async_pf_completion(struct kvm_vcpu *vcpu);
211 int kvm_setup_async_pf(struct kvm_vcpu *vcpu, gva_t gva, unsigned long hva,
212 struct kvm_arch_async_pf *arch);
213 int kvm_async_pf_wakeup_all(struct kvm_vcpu *vcpu);
220 READING_SHADOW_PAGE_TABLES,
224 * Sometimes a large or cross-page mmio needs to be broken up into separate
225 * exits for userspace servicing.
227 struct kvm_mmio_fragment {
235 #ifdef CONFIG_PREEMPT_NOTIFIERS
236 struct preempt_notifier preempt_notifier;
242 unsigned long requests;
243 unsigned long guest_debug;
246 struct list_head blocked_vcpu_list;
252 int guest_fpu_loaded, guest_xcr0_loaded;
253 unsigned char fpu_counter;
254 wait_queue_head_t wq;
258 struct kvm_vcpu_stat stat;
259 unsigned int halt_poll_ns;
261 #ifdef CONFIG_HAS_IOMEM
263 int mmio_read_completed;
265 int mmio_cur_fragment;
266 int mmio_nr_fragments;
267 struct kvm_mmio_fragment mmio_fragments[KVM_MAX_MMIO_FRAGMENTS];
270 #ifdef CONFIG_KVM_ASYNC_PF
273 struct list_head queue;
274 struct list_head done;
279 #ifdef CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT
281 * Cpu relax intercept or pause loop exit optimization
282 * in_spin_loop: set when a vcpu does a pause loop exit
283 * or cpu relax intercepted.
284 * dy_eligible: indicates whether vcpu is eligible for directed yield.
292 struct kvm_vcpu_arch arch;
295 static inline int kvm_vcpu_exiting_guest_mode(struct kvm_vcpu *vcpu)
297 return cmpxchg(&vcpu->mode, IN_GUEST_MODE, EXITING_GUEST_MODE);
301 * Some of the bitops functions do not support too long bitmaps.
302 * This number must be determined not to exceed such limits.
304 #define KVM_MEM_MAX_NR_PAGES ((1UL << 31) - 1)
306 struct kvm_memory_slot {
308 unsigned long npages;
309 unsigned long *dirty_bitmap;
310 struct kvm_arch_memory_slot arch;
311 unsigned long userspace_addr;
316 static inline unsigned long kvm_dirty_bitmap_bytes(struct kvm_memory_slot *memslot)
318 return ALIGN(memslot->npages, BITS_PER_LONG) / 8;
321 struct kvm_s390_adapter_int {
334 struct kvm_kernel_irq_routing_entry {
337 int (*set)(struct kvm_kernel_irq_routing_entry *e,
338 struct kvm *kvm, int irq_source_id, int level,
346 struct kvm_s390_adapter_int adapter;
347 struct kvm_hv_sint hv_sint;
349 struct hlist_node link;
352 #ifdef CONFIG_HAVE_KVM_IRQ_ROUTING
353 struct kvm_irq_routing_table {
354 int chip[KVM_NR_IRQCHIPS][KVM_IRQCHIP_NUM_PINS];
357 * Array indexed by gsi. Each entry contains list of irq chips
358 * the gsi is connected to.
360 struct hlist_head map[0];
364 #ifndef KVM_PRIVATE_MEM_SLOTS
365 #define KVM_PRIVATE_MEM_SLOTS 0
368 #ifndef KVM_MEM_SLOTS_NUM
369 #define KVM_MEM_SLOTS_NUM (KVM_USER_MEM_SLOTS + KVM_PRIVATE_MEM_SLOTS)
372 #ifndef __KVM_VCPU_MULTIPLE_ADDRESS_SPACE
373 static inline int kvm_arch_vcpu_memslots_id(struct kvm_vcpu *vcpu)
381 * memslots are not sorted by id anymore, please use id_to_memslot()
382 * to get the memslot by its id.
384 struct kvm_memslots {
386 struct kvm_memory_slot memslots[KVM_MEM_SLOTS_NUM];
387 /* The mapping table from slot id to the index in memslots[]. */
388 short id_to_index[KVM_MEM_SLOTS_NUM];
395 struct mutex slots_lock;
396 struct mm_struct *mm; /* userspace tied to this vm */
397 struct kvm_memslots *memslots[KVM_ADDRESS_SPACE_NUM];
398 struct srcu_struct srcu;
399 struct srcu_struct irq_srcu;
400 struct kvm_vcpu *vcpus[KVM_MAX_VCPUS];
401 atomic_t online_vcpus;
402 int last_boosted_vcpu;
403 struct list_head vm_list;
405 struct kvm_io_bus *buses[KVM_NR_BUSES];
406 #ifdef CONFIG_HAVE_KVM_EVENTFD
409 struct list_head items;
410 struct list_head resampler_list;
411 struct mutex resampler_lock;
413 struct list_head ioeventfds;
415 struct kvm_vm_stat stat;
416 struct kvm_arch arch;
417 atomic_t users_count;
418 #ifdef KVM_COALESCED_MMIO_PAGE_OFFSET
419 struct kvm_coalesced_mmio_ring *coalesced_mmio_ring;
420 spinlock_t ring_lock;
421 struct list_head coalesced_zones;
424 struct mutex irq_lock;
425 #ifdef CONFIG_HAVE_KVM_IRQCHIP
427 * Update side is protected by irq_lock.
429 struct kvm_irq_routing_table __rcu *irq_routing;
431 #ifdef CONFIG_HAVE_KVM_IRQFD
432 struct hlist_head irq_ack_notifier_list;
435 #if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
436 struct mmu_notifier mmu_notifier;
437 unsigned long mmu_notifier_seq;
438 long mmu_notifier_count;
441 struct list_head devices;
444 #define kvm_err(fmt, ...) \
445 pr_err("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
446 #define kvm_info(fmt, ...) \
447 pr_info("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
448 #define kvm_debug(fmt, ...) \
449 pr_debug("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
450 #define kvm_pr_unimpl(fmt, ...) \
451 pr_err_ratelimited("kvm [%i]: " fmt, \
452 task_tgid_nr(current), ## __VA_ARGS__)
454 /* The guest did something we don't support. */
455 #define vcpu_unimpl(vcpu, fmt, ...) \
456 kvm_pr_unimpl("vcpu%i, guest rIP: 0x%lx " fmt, \
457 (vcpu)->vcpu_id, kvm_rip_read(vcpu), ## __VA_ARGS__)
459 #define vcpu_debug(vcpu, fmt, ...) \
460 kvm_debug("vcpu%i " fmt, (vcpu)->vcpu_id, ## __VA_ARGS__)
461 #define vcpu_err(vcpu, fmt, ...) \
462 kvm_err("vcpu%i " fmt, (vcpu)->vcpu_id, ## __VA_ARGS__)
464 static inline struct kvm_vcpu *kvm_get_vcpu(struct kvm *kvm, int i)
466 /* Pairs with smp_wmb() in kvm_vm_ioctl_create_vcpu, in case
467 * the caller has read kvm->online_vcpus before (as is the case
468 * for kvm_for_each_vcpu, for example).
471 return kvm->vcpus[i];
474 #define kvm_for_each_vcpu(idx, vcpup, kvm) \
476 idx < atomic_read(&kvm->online_vcpus) && \
477 (vcpup = kvm_get_vcpu(kvm, idx)) != NULL; \
480 static inline struct kvm_vcpu *kvm_get_vcpu_by_id(struct kvm *kvm, int id)
482 struct kvm_vcpu *vcpu;
485 if (id < 0 || id >= KVM_MAX_VCPUS)
487 vcpu = kvm_get_vcpu(kvm, id);
488 if (vcpu && vcpu->vcpu_id == id)
490 kvm_for_each_vcpu(i, vcpu, kvm)
491 if (vcpu->vcpu_id == id)
496 #define kvm_for_each_memslot(memslot, slots) \
497 for (memslot = &slots->memslots[0]; \
498 memslot < slots->memslots + KVM_MEM_SLOTS_NUM && memslot->npages;\
501 int kvm_vcpu_init(struct kvm_vcpu *vcpu, struct kvm *kvm, unsigned id);
502 void kvm_vcpu_uninit(struct kvm_vcpu *vcpu);
504 int __must_check vcpu_load(struct kvm_vcpu *vcpu);
505 void vcpu_put(struct kvm_vcpu *vcpu);
507 #ifdef __KVM_HAVE_IOAPIC
508 void kvm_vcpu_request_scan_ioapic(struct kvm *kvm);
509 void kvm_arch_post_irq_routing_update(struct kvm *kvm);
511 static inline void kvm_vcpu_request_scan_ioapic(struct kvm *kvm)
514 static inline void kvm_arch_post_irq_routing_update(struct kvm *kvm)
519 #ifdef CONFIG_HAVE_KVM_IRQFD
520 int kvm_irqfd_init(void);
521 void kvm_irqfd_exit(void);
523 static inline int kvm_irqfd_init(void)
528 static inline void kvm_irqfd_exit(void)
532 int kvm_init(void *opaque, unsigned vcpu_size, unsigned vcpu_align,
533 struct module *module);
536 void kvm_get_kvm(struct kvm *kvm);
537 void kvm_put_kvm(struct kvm *kvm);
539 static inline struct kvm_memslots *__kvm_memslots(struct kvm *kvm, int as_id)
541 return rcu_dereference_check(kvm->memslots[as_id],
542 srcu_read_lock_held(&kvm->srcu)
543 || lockdep_is_held(&kvm->slots_lock));
546 static inline struct kvm_memslots *kvm_memslots(struct kvm *kvm)
548 return __kvm_memslots(kvm, 0);
551 static inline struct kvm_memslots *kvm_vcpu_memslots(struct kvm_vcpu *vcpu)
553 int as_id = kvm_arch_vcpu_memslots_id(vcpu);
555 return __kvm_memslots(vcpu->kvm, as_id);
558 static inline struct kvm_memory_slot *
559 id_to_memslot(struct kvm_memslots *slots, int id)
561 int index = slots->id_to_index[id];
562 struct kvm_memory_slot *slot;
564 slot = &slots->memslots[index];
566 WARN_ON(slot->id != id);
571 * KVM_SET_USER_MEMORY_REGION ioctl allows the following operations:
572 * - create a new memory slot
573 * - delete an existing memory slot
574 * - modify an existing memory slot
575 * -- move it in the guest physical memory space
576 * -- just change its flags
578 * Since flags can be changed by some of these operations, the following
579 * differentiation is the best we can do for __kvm_set_memory_region():
588 int kvm_set_memory_region(struct kvm *kvm,
589 const struct kvm_userspace_memory_region *mem);
590 int __kvm_set_memory_region(struct kvm *kvm,
591 const struct kvm_userspace_memory_region *mem);
592 void kvm_arch_free_memslot(struct kvm *kvm, struct kvm_memory_slot *free,
593 struct kvm_memory_slot *dont);
594 int kvm_arch_create_memslot(struct kvm *kvm, struct kvm_memory_slot *slot,
595 unsigned long npages);
596 void kvm_arch_memslots_updated(struct kvm *kvm, struct kvm_memslots *slots);
597 int kvm_arch_prepare_memory_region(struct kvm *kvm,
598 struct kvm_memory_slot *memslot,
599 const struct kvm_userspace_memory_region *mem,
600 enum kvm_mr_change change);
601 void kvm_arch_commit_memory_region(struct kvm *kvm,
602 const struct kvm_userspace_memory_region *mem,
603 const struct kvm_memory_slot *old,
604 const struct kvm_memory_slot *new,
605 enum kvm_mr_change change);
606 bool kvm_largepages_enabled(void);
607 void kvm_disable_largepages(void);
608 /* flush all memory translations */
609 void kvm_arch_flush_shadow_all(struct kvm *kvm);
610 /* flush memory translations pointing to 'slot' */
611 void kvm_arch_flush_shadow_memslot(struct kvm *kvm,
612 struct kvm_memory_slot *slot);
614 int gfn_to_page_many_atomic(struct kvm_memory_slot *slot, gfn_t gfn,
615 struct page **pages, int nr_pages);
617 struct page *gfn_to_page(struct kvm *kvm, gfn_t gfn);
618 unsigned long gfn_to_hva(struct kvm *kvm, gfn_t gfn);
619 unsigned long gfn_to_hva_prot(struct kvm *kvm, gfn_t gfn, bool *writable);
620 unsigned long gfn_to_hva_memslot(struct kvm_memory_slot *slot, gfn_t gfn);
621 unsigned long gfn_to_hva_memslot_prot(struct kvm_memory_slot *slot, gfn_t gfn,
623 void kvm_release_page_clean(struct page *page);
624 void kvm_release_page_dirty(struct page *page);
625 void kvm_set_page_accessed(struct page *page);
627 pfn_t gfn_to_pfn_atomic(struct kvm *kvm, gfn_t gfn);
628 pfn_t gfn_to_pfn(struct kvm *kvm, gfn_t gfn);
629 pfn_t gfn_to_pfn_prot(struct kvm *kvm, gfn_t gfn, bool write_fault,
631 pfn_t gfn_to_pfn_memslot(struct kvm_memory_slot *slot, gfn_t gfn);
632 pfn_t gfn_to_pfn_memslot_atomic(struct kvm_memory_slot *slot, gfn_t gfn);
633 pfn_t __gfn_to_pfn_memslot(struct kvm_memory_slot *slot, gfn_t gfn, bool atomic,
634 bool *async, bool write_fault, bool *writable);
636 void kvm_release_pfn_clean(pfn_t pfn);
637 void kvm_set_pfn_dirty(pfn_t pfn);
638 void kvm_set_pfn_accessed(pfn_t pfn);
639 void kvm_get_pfn(pfn_t pfn);
641 int kvm_read_guest_page(struct kvm *kvm, gfn_t gfn, void *data, int offset,
643 int kvm_read_guest_atomic(struct kvm *kvm, gpa_t gpa, void *data,
645 int kvm_read_guest(struct kvm *kvm, gpa_t gpa, void *data, unsigned long len);
646 int kvm_read_guest_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
647 void *data, unsigned long len);
648 int kvm_write_guest_page(struct kvm *kvm, gfn_t gfn, const void *data,
649 int offset, int len);
650 int kvm_write_guest(struct kvm *kvm, gpa_t gpa, const void *data,
652 int kvm_write_guest_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
653 void *data, unsigned long len);
654 int kvm_gfn_to_hva_cache_init(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
655 gpa_t gpa, unsigned long len);
656 int kvm_clear_guest_page(struct kvm *kvm, gfn_t gfn, int offset, int len);
657 int kvm_clear_guest(struct kvm *kvm, gpa_t gpa, unsigned long len);
658 struct kvm_memory_slot *gfn_to_memslot(struct kvm *kvm, gfn_t gfn);
659 bool kvm_is_visible_gfn(struct kvm *kvm, gfn_t gfn);
660 unsigned long kvm_host_page_size(struct kvm *kvm, gfn_t gfn);
661 void mark_page_dirty(struct kvm *kvm, gfn_t gfn);
663 struct kvm_memslots *kvm_vcpu_memslots(struct kvm_vcpu *vcpu);
664 struct kvm_memory_slot *kvm_vcpu_gfn_to_memslot(struct kvm_vcpu *vcpu, gfn_t gfn);
665 pfn_t kvm_vcpu_gfn_to_pfn_atomic(struct kvm_vcpu *vcpu, gfn_t gfn);
666 pfn_t kvm_vcpu_gfn_to_pfn(struct kvm_vcpu *vcpu, gfn_t gfn);
667 struct page *kvm_vcpu_gfn_to_page(struct kvm_vcpu *vcpu, gfn_t gfn);
668 unsigned long kvm_vcpu_gfn_to_hva(struct kvm_vcpu *vcpu, gfn_t gfn);
669 unsigned long kvm_vcpu_gfn_to_hva_prot(struct kvm_vcpu *vcpu, gfn_t gfn, bool *writable);
670 int kvm_vcpu_read_guest_page(struct kvm_vcpu *vcpu, gfn_t gfn, void *data, int offset,
672 int kvm_vcpu_read_guest_atomic(struct kvm_vcpu *vcpu, gpa_t gpa, void *data,
674 int kvm_vcpu_read_guest(struct kvm_vcpu *vcpu, gpa_t gpa, void *data,
676 int kvm_vcpu_write_guest_page(struct kvm_vcpu *vcpu, gfn_t gfn, const void *data,
677 int offset, int len);
678 int kvm_vcpu_write_guest(struct kvm_vcpu *vcpu, gpa_t gpa, const void *data,
680 void kvm_vcpu_mark_page_dirty(struct kvm_vcpu *vcpu, gfn_t gfn);
682 void kvm_vcpu_block(struct kvm_vcpu *vcpu);
683 void kvm_arch_vcpu_blocking(struct kvm_vcpu *vcpu);
684 void kvm_arch_vcpu_unblocking(struct kvm_vcpu *vcpu);
685 void kvm_vcpu_kick(struct kvm_vcpu *vcpu);
686 int kvm_vcpu_yield_to(struct kvm_vcpu *target);
687 void kvm_vcpu_on_spin(struct kvm_vcpu *vcpu);
688 void kvm_load_guest_fpu(struct kvm_vcpu *vcpu);
689 void kvm_put_guest_fpu(struct kvm_vcpu *vcpu);
691 void kvm_flush_remote_tlbs(struct kvm *kvm);
692 void kvm_reload_remote_mmus(struct kvm *kvm);
693 void kvm_make_mclock_inprogress_request(struct kvm *kvm);
694 void kvm_make_scan_ioapic_request(struct kvm *kvm);
695 bool kvm_make_all_cpus_request(struct kvm *kvm, unsigned int req);
697 long kvm_arch_dev_ioctl(struct file *filp,
698 unsigned int ioctl, unsigned long arg);
699 long kvm_arch_vcpu_ioctl(struct file *filp,
700 unsigned int ioctl, unsigned long arg);
701 int kvm_arch_vcpu_fault(struct kvm_vcpu *vcpu, struct vm_fault *vmf);
703 int kvm_vm_ioctl_check_extension(struct kvm *kvm, long ext);
705 int kvm_get_dirty_log(struct kvm *kvm,
706 struct kvm_dirty_log *log, int *is_dirty);
708 int kvm_get_dirty_log_protect(struct kvm *kvm,
709 struct kvm_dirty_log *log, bool *is_dirty);
711 void kvm_arch_mmu_enable_log_dirty_pt_masked(struct kvm *kvm,
712 struct kvm_memory_slot *slot,
716 int kvm_vm_ioctl_get_dirty_log(struct kvm *kvm,
717 struct kvm_dirty_log *log);
719 int kvm_vm_ioctl_irq_line(struct kvm *kvm, struct kvm_irq_level *irq_level,
721 long kvm_arch_vm_ioctl(struct file *filp,
722 unsigned int ioctl, unsigned long arg);
724 int kvm_arch_vcpu_ioctl_get_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu);
725 int kvm_arch_vcpu_ioctl_set_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu);
727 int kvm_arch_vcpu_ioctl_translate(struct kvm_vcpu *vcpu,
728 struct kvm_translation *tr);
730 int kvm_arch_vcpu_ioctl_get_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs);
731 int kvm_arch_vcpu_ioctl_set_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs);
732 int kvm_arch_vcpu_ioctl_get_sregs(struct kvm_vcpu *vcpu,
733 struct kvm_sregs *sregs);
734 int kvm_arch_vcpu_ioctl_set_sregs(struct kvm_vcpu *vcpu,
735 struct kvm_sregs *sregs);
736 int kvm_arch_vcpu_ioctl_get_mpstate(struct kvm_vcpu *vcpu,
737 struct kvm_mp_state *mp_state);
738 int kvm_arch_vcpu_ioctl_set_mpstate(struct kvm_vcpu *vcpu,
739 struct kvm_mp_state *mp_state);
740 int kvm_arch_vcpu_ioctl_set_guest_debug(struct kvm_vcpu *vcpu,
741 struct kvm_guest_debug *dbg);
742 int kvm_arch_vcpu_ioctl_run(struct kvm_vcpu *vcpu, struct kvm_run *kvm_run);
744 int kvm_arch_init(void *opaque);
745 void kvm_arch_exit(void);
747 int kvm_arch_vcpu_init(struct kvm_vcpu *vcpu);
748 void kvm_arch_vcpu_uninit(struct kvm_vcpu *vcpu);
750 void kvm_arch_sched_in(struct kvm_vcpu *vcpu, int cpu);
752 void kvm_arch_vcpu_free(struct kvm_vcpu *vcpu);
753 void kvm_arch_vcpu_load(struct kvm_vcpu *vcpu, int cpu);
754 void kvm_arch_vcpu_put(struct kvm_vcpu *vcpu);
755 struct kvm_vcpu *kvm_arch_vcpu_create(struct kvm *kvm, unsigned int id);
756 int kvm_arch_vcpu_setup(struct kvm_vcpu *vcpu);
757 void kvm_arch_vcpu_postcreate(struct kvm_vcpu *vcpu);
758 void kvm_arch_vcpu_destroy(struct kvm_vcpu *vcpu);
760 int kvm_arch_hardware_enable(void);
761 void kvm_arch_hardware_disable(void);
762 int kvm_arch_hardware_setup(void);
763 void kvm_arch_hardware_unsetup(void);
764 void kvm_arch_check_processor_compat(void *rtn);
765 int kvm_arch_vcpu_runnable(struct kvm_vcpu *vcpu);
766 int kvm_arch_vcpu_should_kick(struct kvm_vcpu *vcpu);
768 void *kvm_kvzalloc(unsigned long size);
770 #ifndef __KVM_HAVE_ARCH_VM_ALLOC
771 static inline struct kvm *kvm_arch_alloc_vm(void)
773 return kzalloc(sizeof(struct kvm), GFP_KERNEL);
776 static inline void kvm_arch_free_vm(struct kvm *kvm)
782 #ifdef __KVM_HAVE_ARCH_NONCOHERENT_DMA
783 void kvm_arch_register_noncoherent_dma(struct kvm *kvm);
784 void kvm_arch_unregister_noncoherent_dma(struct kvm *kvm);
785 bool kvm_arch_has_noncoherent_dma(struct kvm *kvm);
787 static inline void kvm_arch_register_noncoherent_dma(struct kvm *kvm)
791 static inline void kvm_arch_unregister_noncoherent_dma(struct kvm *kvm)
795 static inline bool kvm_arch_has_noncoherent_dma(struct kvm *kvm)
800 #ifdef __KVM_HAVE_ARCH_ASSIGNED_DEVICE
801 void kvm_arch_start_assignment(struct kvm *kvm);
802 void kvm_arch_end_assignment(struct kvm *kvm);
803 bool kvm_arch_has_assigned_device(struct kvm *kvm);
805 static inline void kvm_arch_start_assignment(struct kvm *kvm)
809 static inline void kvm_arch_end_assignment(struct kvm *kvm)
813 static inline bool kvm_arch_has_assigned_device(struct kvm *kvm)
819 static inline wait_queue_head_t *kvm_arch_vcpu_wq(struct kvm_vcpu *vcpu)
821 #ifdef __KVM_HAVE_ARCH_WQP
822 return vcpu->arch.wqp;
828 #ifdef __KVM_HAVE_ARCH_INTC_INITIALIZED
830 * returns true if the virtual interrupt controller is initialized and
831 * ready to accept virtual IRQ. On some architectures the virtual interrupt
832 * controller is dynamically instantiated and this is not always true.
834 bool kvm_arch_intc_initialized(struct kvm *kvm);
836 static inline bool kvm_arch_intc_initialized(struct kvm *kvm)
842 int kvm_arch_init_vm(struct kvm *kvm, unsigned long type);
843 void kvm_arch_destroy_vm(struct kvm *kvm);
844 void kvm_arch_sync_events(struct kvm *kvm);
846 int kvm_cpu_has_pending_timer(struct kvm_vcpu *vcpu);
847 void kvm_vcpu_kick(struct kvm_vcpu *vcpu);
849 bool kvm_is_reserved_pfn(pfn_t pfn);
851 struct kvm_irq_ack_notifier {
852 struct hlist_node link;
854 void (*irq_acked)(struct kvm_irq_ack_notifier *kian);
857 int kvm_irq_map_gsi(struct kvm *kvm,
858 struct kvm_kernel_irq_routing_entry *entries, int gsi);
859 int kvm_irq_map_chip_pin(struct kvm *kvm, unsigned irqchip, unsigned pin);
861 int kvm_set_irq(struct kvm *kvm, int irq_source_id, u32 irq, int level,
863 int kvm_set_msi(struct kvm_kernel_irq_routing_entry *irq_entry, struct kvm *kvm,
864 int irq_source_id, int level, bool line_status);
865 int kvm_arch_set_irq_inatomic(struct kvm_kernel_irq_routing_entry *e,
866 struct kvm *kvm, int irq_source_id,
867 int level, bool line_status);
868 bool kvm_irq_has_notifier(struct kvm *kvm, unsigned irqchip, unsigned pin);
869 void kvm_notify_acked_gsi(struct kvm *kvm, int gsi);
870 void kvm_notify_acked_irq(struct kvm *kvm, unsigned irqchip, unsigned pin);
871 void kvm_register_irq_ack_notifier(struct kvm *kvm,
872 struct kvm_irq_ack_notifier *kian);
873 void kvm_unregister_irq_ack_notifier(struct kvm *kvm,
874 struct kvm_irq_ack_notifier *kian);
875 int kvm_request_irq_source_id(struct kvm *kvm);
876 void kvm_free_irq_source_id(struct kvm *kvm, int irq_source_id);
878 #ifdef CONFIG_KVM_DEVICE_ASSIGNMENT
879 int kvm_iommu_map_pages(struct kvm *kvm, struct kvm_memory_slot *slot);
880 void kvm_iommu_unmap_pages(struct kvm *kvm, struct kvm_memory_slot *slot);
882 static inline int kvm_iommu_map_pages(struct kvm *kvm,
883 struct kvm_memory_slot *slot)
888 static inline void kvm_iommu_unmap_pages(struct kvm *kvm,
889 struct kvm_memory_slot *slot)
894 /* must be called with irqs disabled */
895 static inline void __kvm_guest_enter(void)
898 /* KVM does not hold any references to rcu protected data when it
899 * switches CPU into a guest mode. In fact switching to a guest mode
900 * is very similar to exiting to userspace from rcu point of view. In
901 * addition CPU may stay in a guest mode for quite a long time (up to
902 * one time slice). Lets treat guest mode as quiescent state, just like
903 * we do with user-mode execution.
905 if (!context_tracking_cpu_is_enabled())
906 rcu_virt_note_context_switch(smp_processor_id());
909 /* must be called with irqs disabled */
910 static inline void __kvm_guest_exit(void)
915 static inline void kvm_guest_enter(void)
919 local_irq_save(flags);
921 local_irq_restore(flags);
924 static inline void kvm_guest_exit(void)
928 local_irq_save(flags);
930 local_irq_restore(flags);
934 * search_memslots() and __gfn_to_memslot() are here because they are
935 * used in non-modular code in arch/powerpc/kvm/book3s_hv_rm_mmu.c.
936 * gfn_to_memslot() itself isn't here as an inline because that would
937 * bloat other code too much.
939 static inline struct kvm_memory_slot *
940 search_memslots(struct kvm_memslots *slots, gfn_t gfn)
942 int start = 0, end = slots->used_slots;
943 int slot = atomic_read(&slots->lru_slot);
944 struct kvm_memory_slot *memslots = slots->memslots;
946 if (gfn >= memslots[slot].base_gfn &&
947 gfn < memslots[slot].base_gfn + memslots[slot].npages)
948 return &memslots[slot];
950 while (start < end) {
951 slot = start + (end - start) / 2;
953 if (gfn >= memslots[slot].base_gfn)
959 if (gfn >= memslots[start].base_gfn &&
960 gfn < memslots[start].base_gfn + memslots[start].npages) {
961 atomic_set(&slots->lru_slot, start);
962 return &memslots[start];
968 static inline struct kvm_memory_slot *
969 __gfn_to_memslot(struct kvm_memslots *slots, gfn_t gfn)
971 return search_memslots(slots, gfn);
974 static inline unsigned long
975 __gfn_to_hva_memslot(struct kvm_memory_slot *slot, gfn_t gfn)
977 return slot->userspace_addr + (gfn - slot->base_gfn) * PAGE_SIZE;
980 static inline int memslot_id(struct kvm *kvm, gfn_t gfn)
982 return gfn_to_memslot(kvm, gfn)->id;
986 hva_to_gfn_memslot(unsigned long hva, struct kvm_memory_slot *slot)
988 gfn_t gfn_offset = (hva - slot->userspace_addr) >> PAGE_SHIFT;
990 return slot->base_gfn + gfn_offset;
993 static inline gpa_t gfn_to_gpa(gfn_t gfn)
995 return (gpa_t)gfn << PAGE_SHIFT;
998 static inline gfn_t gpa_to_gfn(gpa_t gpa)
1000 return (gfn_t)(gpa >> PAGE_SHIFT);
1003 static inline hpa_t pfn_to_hpa(pfn_t pfn)
1005 return (hpa_t)pfn << PAGE_SHIFT;
1008 static inline bool kvm_is_error_gpa(struct kvm *kvm, gpa_t gpa)
1010 unsigned long hva = gfn_to_hva(kvm, gpa_to_gfn(gpa));
1012 return kvm_is_error_hva(hva);
1015 static inline void kvm_migrate_timers(struct kvm_vcpu *vcpu)
1017 set_bit(KVM_REQ_MIGRATE_TIMER, &vcpu->requests);
1020 enum kvm_stat_kind {
1025 struct kvm_stats_debugfs_item {
1028 enum kvm_stat_kind kind;
1030 extern struct kvm_stats_debugfs_item debugfs_entries[];
1031 extern struct dentry *kvm_debugfs_dir;
1033 #if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
1034 static inline int mmu_notifier_retry(struct kvm *kvm, unsigned long mmu_seq)
1036 if (unlikely(kvm->mmu_notifier_count))
1039 * Ensure the read of mmu_notifier_count happens before the read
1040 * of mmu_notifier_seq. This interacts with the smp_wmb() in
1041 * mmu_notifier_invalidate_range_end to make sure that the caller
1042 * either sees the old (non-zero) value of mmu_notifier_count or
1043 * the new (incremented) value of mmu_notifier_seq.
1044 * PowerPC Book3s HV KVM calls this under a per-page lock
1045 * rather than under kvm->mmu_lock, for scalability, so
1046 * can't rely on kvm->mmu_lock to keep things ordered.
1049 if (kvm->mmu_notifier_seq != mmu_seq)
1055 #ifdef CONFIG_HAVE_KVM_IRQ_ROUTING
1058 #define KVM_MAX_IRQ_ROUTES 4096 //FIXME: we can have more than that...
1060 #define KVM_MAX_IRQ_ROUTES 1024
1063 int kvm_setup_default_irq_routing(struct kvm *kvm);
1064 int kvm_setup_empty_irq_routing(struct kvm *kvm);
1065 int kvm_set_irq_routing(struct kvm *kvm,
1066 const struct kvm_irq_routing_entry *entries,
1069 int kvm_set_routing_entry(struct kvm_kernel_irq_routing_entry *e,
1070 const struct kvm_irq_routing_entry *ue);
1071 void kvm_free_irq_routing(struct kvm *kvm);
1075 static inline void kvm_free_irq_routing(struct kvm *kvm) {}
1079 int kvm_send_userspace_msi(struct kvm *kvm, struct kvm_msi *msi);
1081 #ifdef CONFIG_HAVE_KVM_EVENTFD
1083 void kvm_eventfd_init(struct kvm *kvm);
1084 int kvm_ioeventfd(struct kvm *kvm, struct kvm_ioeventfd *args);
1086 #ifdef CONFIG_HAVE_KVM_IRQFD
1087 int kvm_irqfd(struct kvm *kvm, struct kvm_irqfd *args);
1088 void kvm_irqfd_release(struct kvm *kvm);
1089 void kvm_irq_routing_update(struct kvm *);
1091 static inline int kvm_irqfd(struct kvm *kvm, struct kvm_irqfd *args)
1096 static inline void kvm_irqfd_release(struct kvm *kvm) {}
1101 static inline void kvm_eventfd_init(struct kvm *kvm) {}
1103 static inline int kvm_irqfd(struct kvm *kvm, struct kvm_irqfd *args)
1108 static inline void kvm_irqfd_release(struct kvm *kvm) {}
1110 #ifdef CONFIG_HAVE_KVM_IRQCHIP
1111 static inline void kvm_irq_routing_update(struct kvm *kvm)
1115 void kvm_arch_irq_routing_update(struct kvm *kvm);
1117 static inline int kvm_ioeventfd(struct kvm *kvm, struct kvm_ioeventfd *args)
1122 #endif /* CONFIG_HAVE_KVM_EVENTFD */
1124 #ifdef CONFIG_KVM_APIC_ARCHITECTURE
1125 bool kvm_vcpu_compatible(struct kvm_vcpu *vcpu);
1127 static inline bool kvm_vcpu_compatible(struct kvm_vcpu *vcpu) { return true; }
1130 static inline void kvm_make_request(int req, struct kvm_vcpu *vcpu)
1132 set_bit(req, &vcpu->requests);
1135 static inline bool kvm_check_request(int req, struct kvm_vcpu *vcpu)
1137 if (test_bit(req, &vcpu->requests)) {
1138 clear_bit(req, &vcpu->requests);
1145 extern bool kvm_rebooting;
1148 struct kvm_device_ops *ops;
1151 struct list_head vm_node;
1154 /* create, destroy, and name are mandatory */
1155 struct kvm_device_ops {
1157 int (*create)(struct kvm_device *dev, u32 type);
1160 * Destroy is responsible for freeing dev.
1162 * Destroy may be called before or after destructors are called
1163 * on emulated I/O regions, depending on whether a reference is
1164 * held by a vcpu or other kvm component that gets destroyed
1165 * after the emulated I/O.
1167 void (*destroy)(struct kvm_device *dev);
1169 int (*set_attr)(struct kvm_device *dev, struct kvm_device_attr *attr);
1170 int (*get_attr)(struct kvm_device *dev, struct kvm_device_attr *attr);
1171 int (*has_attr)(struct kvm_device *dev, struct kvm_device_attr *attr);
1172 long (*ioctl)(struct kvm_device *dev, unsigned int ioctl,
1176 void kvm_device_get(struct kvm_device *dev);
1177 void kvm_device_put(struct kvm_device *dev);
1178 struct kvm_device *kvm_device_from_filp(struct file *filp);
1179 int kvm_register_device_ops(struct kvm_device_ops *ops, u32 type);
1180 void kvm_unregister_device_ops(u32 type);
1182 extern struct kvm_device_ops kvm_mpic_ops;
1183 extern struct kvm_device_ops kvm_xics_ops;
1184 extern struct kvm_device_ops kvm_arm_vgic_v2_ops;
1185 extern struct kvm_device_ops kvm_arm_vgic_v3_ops;
1187 #ifdef CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT
1189 static inline void kvm_vcpu_set_in_spin_loop(struct kvm_vcpu *vcpu, bool val)
1191 vcpu->spin_loop.in_spin_loop = val;
1193 static inline void kvm_vcpu_set_dy_eligible(struct kvm_vcpu *vcpu, bool val)
1195 vcpu->spin_loop.dy_eligible = val;
1198 #else /* !CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT */
1200 static inline void kvm_vcpu_set_in_spin_loop(struct kvm_vcpu *vcpu, bool val)
1204 static inline void kvm_vcpu_set_dy_eligible(struct kvm_vcpu *vcpu, bool val)
1207 #endif /* CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT */
1209 #ifdef CONFIG_HAVE_KVM_IRQ_BYPASS
1210 int kvm_arch_irq_bypass_add_producer(struct irq_bypass_consumer *,
1211 struct irq_bypass_producer *);
1212 void kvm_arch_irq_bypass_del_producer(struct irq_bypass_consumer *,
1213 struct irq_bypass_producer *);
1214 void kvm_arch_irq_bypass_stop(struct irq_bypass_consumer *);
1215 void kvm_arch_irq_bypass_start(struct irq_bypass_consumer *);
1216 int kvm_arch_update_irqfd_routing(struct kvm *kvm, unsigned int host_irq,
1217 uint32_t guest_irq, bool set);
1218 #endif /* CONFIG_HAVE_KVM_IRQ_BYPASS */