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 <linux/swait.h>
29 #include <asm/signal.h>
31 #include <linux/kvm.h>
32 #include <linux/kvm_para.h>
34 #include <linux/kvm_types.h>
36 #include <asm/kvm_host.h>
38 #ifndef KVM_MAX_VCPU_ID
39 #define KVM_MAX_VCPU_ID KVM_MAX_VCPUS
43 * The bit 16 ~ bit 31 of kvm_memory_region::flags are internally used
44 * in kvm, other bits are visible for userspace which are defined in
45 * include/linux/kvm_h.
47 #define KVM_MEMSLOT_INVALID (1UL << 16)
48 #define KVM_MEMSLOT_INCOHERENT (1UL << 17)
50 /* Two fragments for cross MMIO pages. */
51 #define KVM_MAX_MMIO_FRAGMENTS 2
53 #ifndef KVM_ADDRESS_SPACE_NUM
54 #define KVM_ADDRESS_SPACE_NUM 1
58 * For the normal pfn, the highest 12 bits should be zero,
59 * so we can mask bit 62 ~ bit 52 to indicate the error pfn,
60 * mask bit 63 to indicate the noslot pfn.
62 #define KVM_PFN_ERR_MASK (0x7ffULL << 52)
63 #define KVM_PFN_ERR_NOSLOT_MASK (0xfffULL << 52)
64 #define KVM_PFN_NOSLOT (0x1ULL << 63)
66 #define KVM_PFN_ERR_FAULT (KVM_PFN_ERR_MASK)
67 #define KVM_PFN_ERR_HWPOISON (KVM_PFN_ERR_MASK + 1)
68 #define KVM_PFN_ERR_RO_FAULT (KVM_PFN_ERR_MASK + 2)
71 * error pfns indicate that the gfn is in slot but faild to
72 * translate it to pfn on host.
74 static inline bool is_error_pfn(kvm_pfn_t pfn)
76 return !!(pfn & KVM_PFN_ERR_MASK);
80 * error_noslot pfns indicate that the gfn can not be
81 * translated to pfn - it is not in slot or failed to
82 * translate it to pfn.
84 static inline bool is_error_noslot_pfn(kvm_pfn_t pfn)
86 return !!(pfn & KVM_PFN_ERR_NOSLOT_MASK);
89 /* noslot pfn indicates that the gfn is not in slot. */
90 static inline bool is_noslot_pfn(kvm_pfn_t pfn)
92 return pfn == KVM_PFN_NOSLOT;
96 * architectures with KVM_HVA_ERR_BAD other than PAGE_OFFSET (e.g. s390)
97 * provide own defines and kvm_is_error_hva
99 #ifndef KVM_HVA_ERR_BAD
101 #define KVM_HVA_ERR_BAD (PAGE_OFFSET)
102 #define KVM_HVA_ERR_RO_BAD (PAGE_OFFSET + PAGE_SIZE)
104 static inline bool kvm_is_error_hva(unsigned long addr)
106 return addr >= PAGE_OFFSET;
111 #define KVM_ERR_PTR_BAD_PAGE (ERR_PTR(-ENOENT))
113 static inline bool is_error_page(struct page *page)
119 * Architecture-independent vcpu->requests bit members
120 * Bits 4-7 are reserved for more arch-independent bits.
122 #define KVM_REQ_TLB_FLUSH 0
123 #define KVM_REQ_MMU_RELOAD 1
124 #define KVM_REQ_PENDING_TIMER 2
125 #define KVM_REQ_UNHALT 3
127 #define KVM_USERSPACE_IRQ_SOURCE_ID 0
128 #define KVM_IRQFD_RESAMPLE_IRQ_SOURCE_ID 1
130 extern struct kmem_cache *kvm_vcpu_cache;
132 extern spinlock_t kvm_lock;
133 extern struct list_head vm_list;
135 struct kvm_io_range {
138 struct kvm_io_device *dev;
141 #define NR_IOBUS_DEVS 1000
146 struct kvm_io_range range[];
152 KVM_VIRTIO_CCW_NOTIFY_BUS,
157 int kvm_io_bus_write(struct kvm_vcpu *vcpu, enum kvm_bus bus_idx, gpa_t addr,
158 int len, const void *val);
159 int kvm_io_bus_write_cookie(struct kvm_vcpu *vcpu, enum kvm_bus bus_idx,
160 gpa_t addr, int len, const void *val, long cookie);
161 int kvm_io_bus_read(struct kvm_vcpu *vcpu, enum kvm_bus bus_idx, gpa_t addr,
163 int kvm_io_bus_register_dev(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr,
164 int len, struct kvm_io_device *dev);
165 int kvm_io_bus_unregister_dev(struct kvm *kvm, enum kvm_bus bus_idx,
166 struct kvm_io_device *dev);
167 struct kvm_io_device *kvm_io_bus_get_dev(struct kvm *kvm, enum kvm_bus bus_idx,
170 #ifdef CONFIG_KVM_ASYNC_PF
171 struct kvm_async_pf {
172 struct work_struct work;
173 struct list_head link;
174 struct list_head queue;
175 struct kvm_vcpu *vcpu;
176 struct mm_struct *mm;
179 struct kvm_arch_async_pf arch;
183 void kvm_clear_async_pf_completion_queue(struct kvm_vcpu *vcpu);
184 void kvm_check_async_pf_completion(struct kvm_vcpu *vcpu);
185 int kvm_setup_async_pf(struct kvm_vcpu *vcpu, gva_t gva, unsigned long hva,
186 struct kvm_arch_async_pf *arch);
187 int kvm_async_pf_wakeup_all(struct kvm_vcpu *vcpu);
194 READING_SHADOW_PAGE_TABLES,
198 * Sometimes a large or cross-page mmio needs to be broken up into separate
199 * exits for userspace servicing.
201 struct kvm_mmio_fragment {
209 #ifdef CONFIG_PREEMPT_NOTIFIERS
210 struct preempt_notifier preempt_notifier;
216 unsigned long requests;
217 unsigned long guest_debug;
220 struct list_head blocked_vcpu_list;
226 int guest_fpu_loaded, guest_xcr0_loaded;
227 unsigned char fpu_counter;
228 struct swait_queue_head wq;
232 struct kvm_vcpu_stat stat;
233 unsigned int halt_poll_ns;
236 #ifdef CONFIG_HAS_IOMEM
238 int mmio_read_completed;
240 int mmio_cur_fragment;
241 int mmio_nr_fragments;
242 struct kvm_mmio_fragment mmio_fragments[KVM_MAX_MMIO_FRAGMENTS];
245 #ifdef CONFIG_KVM_ASYNC_PF
248 struct list_head queue;
249 struct list_head done;
254 #ifdef CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT
256 * Cpu relax intercept or pause loop exit optimization
257 * in_spin_loop: set when a vcpu does a pause loop exit
258 * or cpu relax intercepted.
259 * dy_eligible: indicates whether vcpu is eligible for directed yield.
267 struct kvm_vcpu_arch arch;
268 struct dentry *debugfs_dentry;
271 static inline int kvm_vcpu_exiting_guest_mode(struct kvm_vcpu *vcpu)
273 return cmpxchg(&vcpu->mode, IN_GUEST_MODE, EXITING_GUEST_MODE);
277 * Some of the bitops functions do not support too long bitmaps.
278 * This number must be determined not to exceed such limits.
280 #define KVM_MEM_MAX_NR_PAGES ((1UL << 31) - 1)
282 struct kvm_memory_slot {
284 unsigned long npages;
285 unsigned long *dirty_bitmap;
286 struct kvm_arch_memory_slot arch;
287 unsigned long userspace_addr;
292 static inline unsigned long kvm_dirty_bitmap_bytes(struct kvm_memory_slot *memslot)
294 return ALIGN(memslot->npages, BITS_PER_LONG) / 8;
297 struct kvm_s390_adapter_int {
310 struct kvm_kernel_irq_routing_entry {
313 int (*set)(struct kvm_kernel_irq_routing_entry *e,
314 struct kvm *kvm, int irq_source_id, int level,
328 struct kvm_s390_adapter_int adapter;
329 struct kvm_hv_sint hv_sint;
331 struct hlist_node link;
334 #ifdef CONFIG_HAVE_KVM_IRQ_ROUTING
335 struct kvm_irq_routing_table {
336 int chip[KVM_NR_IRQCHIPS][KVM_IRQCHIP_NUM_PINS];
339 * Array indexed by gsi. Each entry contains list of irq chips
340 * the gsi is connected to.
342 struct hlist_head map[0];
346 #ifndef KVM_PRIVATE_MEM_SLOTS
347 #define KVM_PRIVATE_MEM_SLOTS 0
350 #ifndef KVM_MEM_SLOTS_NUM
351 #define KVM_MEM_SLOTS_NUM (KVM_USER_MEM_SLOTS + KVM_PRIVATE_MEM_SLOTS)
354 #ifndef __KVM_VCPU_MULTIPLE_ADDRESS_SPACE
355 static inline int kvm_arch_vcpu_memslots_id(struct kvm_vcpu *vcpu)
363 * memslots are not sorted by id anymore, please use id_to_memslot()
364 * to get the memslot by its id.
366 struct kvm_memslots {
368 struct kvm_memory_slot memslots[KVM_MEM_SLOTS_NUM];
369 /* The mapping table from slot id to the index in memslots[]. */
370 short id_to_index[KVM_MEM_SLOTS_NUM];
377 struct mutex slots_lock;
378 struct mm_struct *mm; /* userspace tied to this vm */
379 struct kvm_memslots *memslots[KVM_ADDRESS_SPACE_NUM];
380 struct srcu_struct srcu;
381 struct srcu_struct irq_srcu;
382 struct kvm_vcpu *vcpus[KVM_MAX_VCPUS];
385 * created_vcpus is protected by kvm->lock, and is incremented
386 * at the beginning of KVM_CREATE_VCPU. online_vcpus is only
387 * incremented after storing the kvm_vcpu pointer in vcpus,
388 * and is accessed atomically.
390 atomic_t online_vcpus;
392 int last_boosted_vcpu;
393 struct list_head vm_list;
395 struct kvm_io_bus *buses[KVM_NR_BUSES];
396 #ifdef CONFIG_HAVE_KVM_EVENTFD
399 struct list_head items;
400 struct list_head resampler_list;
401 struct mutex resampler_lock;
403 struct list_head ioeventfds;
405 struct kvm_vm_stat stat;
406 struct kvm_arch arch;
407 atomic_t users_count;
408 #ifdef KVM_COALESCED_MMIO_PAGE_OFFSET
409 struct kvm_coalesced_mmio_ring *coalesced_mmio_ring;
410 spinlock_t ring_lock;
411 struct list_head coalesced_zones;
414 struct mutex irq_lock;
415 #ifdef CONFIG_HAVE_KVM_IRQCHIP
417 * Update side is protected by irq_lock.
419 struct kvm_irq_routing_table __rcu *irq_routing;
421 #ifdef CONFIG_HAVE_KVM_IRQFD
422 struct hlist_head irq_ack_notifier_list;
425 #if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
426 struct mmu_notifier mmu_notifier;
427 unsigned long mmu_notifier_seq;
428 long mmu_notifier_count;
431 struct list_head devices;
432 struct dentry *debugfs_dentry;
433 struct kvm_stat_data **debugfs_stat_data;
436 #define kvm_err(fmt, ...) \
437 pr_err("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
438 #define kvm_info(fmt, ...) \
439 pr_info("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
440 #define kvm_debug(fmt, ...) \
441 pr_debug("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
442 #define kvm_pr_unimpl(fmt, ...) \
443 pr_err_ratelimited("kvm [%i]: " fmt, \
444 task_tgid_nr(current), ## __VA_ARGS__)
446 /* The guest did something we don't support. */
447 #define vcpu_unimpl(vcpu, fmt, ...) \
448 kvm_pr_unimpl("vcpu%i, guest rIP: 0x%lx " fmt, \
449 (vcpu)->vcpu_id, kvm_rip_read(vcpu), ## __VA_ARGS__)
451 #define vcpu_debug(vcpu, fmt, ...) \
452 kvm_debug("vcpu%i " fmt, (vcpu)->vcpu_id, ## __VA_ARGS__)
453 #define vcpu_err(vcpu, fmt, ...) \
454 kvm_err("vcpu%i " fmt, (vcpu)->vcpu_id, ## __VA_ARGS__)
456 static inline struct kvm_vcpu *kvm_get_vcpu(struct kvm *kvm, int i)
458 /* Pairs with smp_wmb() in kvm_vm_ioctl_create_vcpu, in case
459 * the caller has read kvm->online_vcpus before (as is the case
460 * for kvm_for_each_vcpu, for example).
463 return kvm->vcpus[i];
466 #define kvm_for_each_vcpu(idx, vcpup, kvm) \
468 idx < atomic_read(&kvm->online_vcpus) && \
469 (vcpup = kvm_get_vcpu(kvm, idx)) != NULL; \
472 static inline struct kvm_vcpu *kvm_get_vcpu_by_id(struct kvm *kvm, int id)
474 struct kvm_vcpu *vcpu = NULL;
479 if (id < KVM_MAX_VCPUS)
480 vcpu = kvm_get_vcpu(kvm, id);
481 if (vcpu && vcpu->vcpu_id == id)
483 kvm_for_each_vcpu(i, vcpu, kvm)
484 if (vcpu->vcpu_id == id)
489 #define kvm_for_each_memslot(memslot, slots) \
490 for (memslot = &slots->memslots[0]; \
491 memslot < slots->memslots + KVM_MEM_SLOTS_NUM && memslot->npages;\
494 int kvm_vcpu_init(struct kvm_vcpu *vcpu, struct kvm *kvm, unsigned id);
495 void kvm_vcpu_uninit(struct kvm_vcpu *vcpu);
497 int __must_check vcpu_load(struct kvm_vcpu *vcpu);
498 void vcpu_put(struct kvm_vcpu *vcpu);
500 #ifdef __KVM_HAVE_IOAPIC
501 void kvm_vcpu_request_scan_ioapic(struct kvm *kvm);
502 void kvm_arch_post_irq_routing_update(struct kvm *kvm);
504 static inline void kvm_vcpu_request_scan_ioapic(struct kvm *kvm)
507 static inline void kvm_arch_post_irq_routing_update(struct kvm *kvm)
512 #ifdef CONFIG_HAVE_KVM_IRQFD
513 int kvm_irqfd_init(void);
514 void kvm_irqfd_exit(void);
516 static inline int kvm_irqfd_init(void)
521 static inline void kvm_irqfd_exit(void)
525 int kvm_init(void *opaque, unsigned vcpu_size, unsigned vcpu_align,
526 struct module *module);
529 void kvm_get_kvm(struct kvm *kvm);
530 void kvm_put_kvm(struct kvm *kvm);
532 static inline struct kvm_memslots *__kvm_memslots(struct kvm *kvm, int as_id)
534 return rcu_dereference_check(kvm->memslots[as_id],
535 srcu_read_lock_held(&kvm->srcu)
536 || lockdep_is_held(&kvm->slots_lock));
539 static inline struct kvm_memslots *kvm_memslots(struct kvm *kvm)
541 return __kvm_memslots(kvm, 0);
544 static inline struct kvm_memslots *kvm_vcpu_memslots(struct kvm_vcpu *vcpu)
546 int as_id = kvm_arch_vcpu_memslots_id(vcpu);
548 return __kvm_memslots(vcpu->kvm, as_id);
551 static inline struct kvm_memory_slot *
552 id_to_memslot(struct kvm_memslots *slots, int id)
554 int index = slots->id_to_index[id];
555 struct kvm_memory_slot *slot;
557 slot = &slots->memslots[index];
559 WARN_ON(slot->id != id);
564 * KVM_SET_USER_MEMORY_REGION ioctl allows the following operations:
565 * - create a new memory slot
566 * - delete an existing memory slot
567 * - modify an existing memory slot
568 * -- move it in the guest physical memory space
569 * -- just change its flags
571 * Since flags can be changed by some of these operations, the following
572 * differentiation is the best we can do for __kvm_set_memory_region():
581 int kvm_set_memory_region(struct kvm *kvm,
582 const struct kvm_userspace_memory_region *mem);
583 int __kvm_set_memory_region(struct kvm *kvm,
584 const struct kvm_userspace_memory_region *mem);
585 void kvm_arch_free_memslot(struct kvm *kvm, struct kvm_memory_slot *free,
586 struct kvm_memory_slot *dont);
587 int kvm_arch_create_memslot(struct kvm *kvm, struct kvm_memory_slot *slot,
588 unsigned long npages);
589 void kvm_arch_memslots_updated(struct kvm *kvm, struct kvm_memslots *slots);
590 int kvm_arch_prepare_memory_region(struct kvm *kvm,
591 struct kvm_memory_slot *memslot,
592 const struct kvm_userspace_memory_region *mem,
593 enum kvm_mr_change change);
594 void kvm_arch_commit_memory_region(struct kvm *kvm,
595 const struct kvm_userspace_memory_region *mem,
596 const struct kvm_memory_slot *old,
597 const struct kvm_memory_slot *new,
598 enum kvm_mr_change change);
599 bool kvm_largepages_enabled(void);
600 void kvm_disable_largepages(void);
601 /* flush all memory translations */
602 void kvm_arch_flush_shadow_all(struct kvm *kvm);
603 /* flush memory translations pointing to 'slot' */
604 void kvm_arch_flush_shadow_memslot(struct kvm *kvm,
605 struct kvm_memory_slot *slot);
607 int gfn_to_page_many_atomic(struct kvm_memory_slot *slot, gfn_t gfn,
608 struct page **pages, int nr_pages);
610 struct page *gfn_to_page(struct kvm *kvm, gfn_t gfn);
611 unsigned long gfn_to_hva(struct kvm *kvm, gfn_t gfn);
612 unsigned long gfn_to_hva_prot(struct kvm *kvm, gfn_t gfn, bool *writable);
613 unsigned long gfn_to_hva_memslot(struct kvm_memory_slot *slot, gfn_t gfn);
614 unsigned long gfn_to_hva_memslot_prot(struct kvm_memory_slot *slot, gfn_t gfn,
616 void kvm_release_page_clean(struct page *page);
617 void kvm_release_page_dirty(struct page *page);
618 void kvm_set_page_accessed(struct page *page);
620 kvm_pfn_t gfn_to_pfn_atomic(struct kvm *kvm, gfn_t gfn);
621 kvm_pfn_t gfn_to_pfn(struct kvm *kvm, gfn_t gfn);
622 kvm_pfn_t gfn_to_pfn_prot(struct kvm *kvm, gfn_t gfn, bool write_fault,
624 kvm_pfn_t gfn_to_pfn_memslot(struct kvm_memory_slot *slot, gfn_t gfn);
625 kvm_pfn_t gfn_to_pfn_memslot_atomic(struct kvm_memory_slot *slot, gfn_t gfn);
626 kvm_pfn_t __gfn_to_pfn_memslot(struct kvm_memory_slot *slot, gfn_t gfn,
627 bool atomic, bool *async, bool write_fault,
630 void kvm_release_pfn_clean(kvm_pfn_t pfn);
631 void kvm_set_pfn_dirty(kvm_pfn_t pfn);
632 void kvm_set_pfn_accessed(kvm_pfn_t pfn);
633 void kvm_get_pfn(kvm_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 kvm_pfn_t kvm_vcpu_gfn_to_pfn_atomic(struct kvm_vcpu *vcpu, gfn_t gfn);
660 kvm_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_wake_up(struct kvm_vcpu *vcpu);
680 void kvm_vcpu_kick(struct kvm_vcpu *vcpu);
681 int kvm_vcpu_yield_to(struct kvm_vcpu *target);
682 void kvm_vcpu_on_spin(struct kvm_vcpu *vcpu);
683 void kvm_load_guest_fpu(struct kvm_vcpu *vcpu);
684 void kvm_put_guest_fpu(struct kvm_vcpu *vcpu);
686 void kvm_flush_remote_tlbs(struct kvm *kvm);
687 void kvm_reload_remote_mmus(struct kvm *kvm);
688 bool kvm_make_all_cpus_request(struct kvm *kvm, unsigned int req);
690 long kvm_arch_dev_ioctl(struct file *filp,
691 unsigned int ioctl, unsigned long arg);
692 long kvm_arch_vcpu_ioctl(struct file *filp,
693 unsigned int ioctl, unsigned long arg);
694 int kvm_arch_vcpu_fault(struct kvm_vcpu *vcpu, struct vm_fault *vmf);
696 int kvm_vm_ioctl_check_extension(struct kvm *kvm, long ext);
698 int kvm_get_dirty_log(struct kvm *kvm,
699 struct kvm_dirty_log *log, int *is_dirty);
701 int kvm_get_dirty_log_protect(struct kvm *kvm,
702 struct kvm_dirty_log *log, bool *is_dirty);
704 void kvm_arch_mmu_enable_log_dirty_pt_masked(struct kvm *kvm,
705 struct kvm_memory_slot *slot,
709 int kvm_vm_ioctl_get_dirty_log(struct kvm *kvm,
710 struct kvm_dirty_log *log);
712 int kvm_vm_ioctl_irq_line(struct kvm *kvm, struct kvm_irq_level *irq_level,
714 long kvm_arch_vm_ioctl(struct file *filp,
715 unsigned int ioctl, unsigned long arg);
717 int kvm_arch_vcpu_ioctl_get_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu);
718 int kvm_arch_vcpu_ioctl_set_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu);
720 int kvm_arch_vcpu_ioctl_translate(struct kvm_vcpu *vcpu,
721 struct kvm_translation *tr);
723 int kvm_arch_vcpu_ioctl_get_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs);
724 int kvm_arch_vcpu_ioctl_set_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs);
725 int kvm_arch_vcpu_ioctl_get_sregs(struct kvm_vcpu *vcpu,
726 struct kvm_sregs *sregs);
727 int kvm_arch_vcpu_ioctl_set_sregs(struct kvm_vcpu *vcpu,
728 struct kvm_sregs *sregs);
729 int kvm_arch_vcpu_ioctl_get_mpstate(struct kvm_vcpu *vcpu,
730 struct kvm_mp_state *mp_state);
731 int kvm_arch_vcpu_ioctl_set_mpstate(struct kvm_vcpu *vcpu,
732 struct kvm_mp_state *mp_state);
733 int kvm_arch_vcpu_ioctl_set_guest_debug(struct kvm_vcpu *vcpu,
734 struct kvm_guest_debug *dbg);
735 int kvm_arch_vcpu_ioctl_run(struct kvm_vcpu *vcpu, struct kvm_run *kvm_run);
737 int kvm_arch_init(void *opaque);
738 void kvm_arch_exit(void);
740 int kvm_arch_vcpu_init(struct kvm_vcpu *vcpu);
741 void kvm_arch_vcpu_uninit(struct kvm_vcpu *vcpu);
743 void kvm_arch_sched_in(struct kvm_vcpu *vcpu, int cpu);
745 void kvm_arch_vcpu_free(struct kvm_vcpu *vcpu);
746 void kvm_arch_vcpu_load(struct kvm_vcpu *vcpu, int cpu);
747 void kvm_arch_vcpu_put(struct kvm_vcpu *vcpu);
748 struct kvm_vcpu *kvm_arch_vcpu_create(struct kvm *kvm, unsigned int id);
749 int kvm_arch_vcpu_setup(struct kvm_vcpu *vcpu);
750 void kvm_arch_vcpu_postcreate(struct kvm_vcpu *vcpu);
751 void kvm_arch_vcpu_destroy(struct kvm_vcpu *vcpu);
753 bool kvm_arch_has_vcpu_debugfs(void);
754 int kvm_arch_create_vcpu_debugfs(struct kvm_vcpu *vcpu);
756 int kvm_arch_hardware_enable(void);
757 void kvm_arch_hardware_disable(void);
758 int kvm_arch_hardware_setup(void);
759 void kvm_arch_hardware_unsetup(void);
760 void kvm_arch_check_processor_compat(void *rtn);
761 int kvm_arch_vcpu_runnable(struct kvm_vcpu *vcpu);
762 int kvm_arch_vcpu_should_kick(struct kvm_vcpu *vcpu);
764 void *kvm_kvzalloc(unsigned long size);
766 #ifndef __KVM_HAVE_ARCH_VM_ALLOC
767 static inline struct kvm *kvm_arch_alloc_vm(void)
769 return kzalloc(sizeof(struct kvm), GFP_KERNEL);
772 static inline void kvm_arch_free_vm(struct kvm *kvm)
778 #ifdef __KVM_HAVE_ARCH_NONCOHERENT_DMA
779 void kvm_arch_register_noncoherent_dma(struct kvm *kvm);
780 void kvm_arch_unregister_noncoherent_dma(struct kvm *kvm);
781 bool kvm_arch_has_noncoherent_dma(struct kvm *kvm);
783 static inline void kvm_arch_register_noncoherent_dma(struct kvm *kvm)
787 static inline void kvm_arch_unregister_noncoherent_dma(struct kvm *kvm)
791 static inline bool kvm_arch_has_noncoherent_dma(struct kvm *kvm)
796 #ifdef __KVM_HAVE_ARCH_ASSIGNED_DEVICE
797 void kvm_arch_start_assignment(struct kvm *kvm);
798 void kvm_arch_end_assignment(struct kvm *kvm);
799 bool kvm_arch_has_assigned_device(struct kvm *kvm);
801 static inline void kvm_arch_start_assignment(struct kvm *kvm)
805 static inline void kvm_arch_end_assignment(struct kvm *kvm)
809 static inline bool kvm_arch_has_assigned_device(struct kvm *kvm)
815 static inline struct swait_queue_head *kvm_arch_vcpu_wq(struct kvm_vcpu *vcpu)
817 #ifdef __KVM_HAVE_ARCH_WQP
818 return vcpu->arch.wqp;
824 #ifdef __KVM_HAVE_ARCH_INTC_INITIALIZED
826 * returns true if the virtual interrupt controller is initialized and
827 * ready to accept virtual IRQ. On some architectures the virtual interrupt
828 * controller is dynamically instantiated and this is not always true.
830 bool kvm_arch_intc_initialized(struct kvm *kvm);
832 static inline bool kvm_arch_intc_initialized(struct kvm *kvm)
838 int kvm_arch_init_vm(struct kvm *kvm, unsigned long type);
839 void kvm_arch_destroy_vm(struct kvm *kvm);
840 void kvm_arch_sync_events(struct kvm *kvm);
842 int kvm_cpu_has_pending_timer(struct kvm_vcpu *vcpu);
843 void kvm_vcpu_kick(struct kvm_vcpu *vcpu);
845 bool kvm_is_reserved_pfn(kvm_pfn_t pfn);
847 struct kvm_irq_ack_notifier {
848 struct hlist_node link;
850 void (*irq_acked)(struct kvm_irq_ack_notifier *kian);
853 int kvm_irq_map_gsi(struct kvm *kvm,
854 struct kvm_kernel_irq_routing_entry *entries, int gsi);
855 int kvm_irq_map_chip_pin(struct kvm *kvm, unsigned irqchip, unsigned pin);
857 int kvm_set_irq(struct kvm *kvm, int irq_source_id, u32 irq, int level,
859 int kvm_set_msi(struct kvm_kernel_irq_routing_entry *irq_entry, struct kvm *kvm,
860 int irq_source_id, int level, bool line_status);
861 int kvm_arch_set_irq_inatomic(struct kvm_kernel_irq_routing_entry *e,
862 struct kvm *kvm, int irq_source_id,
863 int level, bool line_status);
864 bool kvm_irq_has_notifier(struct kvm *kvm, unsigned irqchip, unsigned pin);
865 void kvm_notify_acked_gsi(struct kvm *kvm, int gsi);
866 void kvm_notify_acked_irq(struct kvm *kvm, unsigned irqchip, unsigned pin);
867 void kvm_register_irq_ack_notifier(struct kvm *kvm,
868 struct kvm_irq_ack_notifier *kian);
869 void kvm_unregister_irq_ack_notifier(struct kvm *kvm,
870 struct kvm_irq_ack_notifier *kian);
871 int kvm_request_irq_source_id(struct kvm *kvm);
872 void kvm_free_irq_source_id(struct kvm *kvm, int irq_source_id);
874 #ifdef CONFIG_KVM_DEVICE_ASSIGNMENT
875 int kvm_iommu_map_pages(struct kvm *kvm, struct kvm_memory_slot *slot);
876 void kvm_iommu_unmap_pages(struct kvm *kvm, struct kvm_memory_slot *slot);
878 static inline int kvm_iommu_map_pages(struct kvm *kvm,
879 struct kvm_memory_slot *slot)
884 static inline void kvm_iommu_unmap_pages(struct kvm *kvm,
885 struct kvm_memory_slot *slot)
891 * search_memslots() and __gfn_to_memslot() are here because they are
892 * used in non-modular code in arch/powerpc/kvm/book3s_hv_rm_mmu.c.
893 * gfn_to_memslot() itself isn't here as an inline because that would
894 * bloat other code too much.
896 static inline struct kvm_memory_slot *
897 search_memslots(struct kvm_memslots *slots, gfn_t gfn)
899 int start = 0, end = slots->used_slots;
900 int slot = atomic_read(&slots->lru_slot);
901 struct kvm_memory_slot *memslots = slots->memslots;
903 if (gfn >= memslots[slot].base_gfn &&
904 gfn < memslots[slot].base_gfn + memslots[slot].npages)
905 return &memslots[slot];
907 while (start < end) {
908 slot = start + (end - start) / 2;
910 if (gfn >= memslots[slot].base_gfn)
916 if (gfn >= memslots[start].base_gfn &&
917 gfn < memslots[start].base_gfn + memslots[start].npages) {
918 atomic_set(&slots->lru_slot, start);
919 return &memslots[start];
925 static inline struct kvm_memory_slot *
926 __gfn_to_memslot(struct kvm_memslots *slots, gfn_t gfn)
928 return search_memslots(slots, gfn);
931 static inline unsigned long
932 __gfn_to_hva_memslot(struct kvm_memory_slot *slot, gfn_t gfn)
934 return slot->userspace_addr + (gfn - slot->base_gfn) * PAGE_SIZE;
937 static inline int memslot_id(struct kvm *kvm, gfn_t gfn)
939 return gfn_to_memslot(kvm, gfn)->id;
943 hva_to_gfn_memslot(unsigned long hva, struct kvm_memory_slot *slot)
945 gfn_t gfn_offset = (hva - slot->userspace_addr) >> PAGE_SHIFT;
947 return slot->base_gfn + gfn_offset;
950 static inline gpa_t gfn_to_gpa(gfn_t gfn)
952 return (gpa_t)gfn << PAGE_SHIFT;
955 static inline gfn_t gpa_to_gfn(gpa_t gpa)
957 return (gfn_t)(gpa >> PAGE_SHIFT);
960 static inline hpa_t pfn_to_hpa(kvm_pfn_t pfn)
962 return (hpa_t)pfn << PAGE_SHIFT;
965 static inline bool kvm_is_error_gpa(struct kvm *kvm, gpa_t gpa)
967 unsigned long hva = gfn_to_hva(kvm, gpa_to_gfn(gpa));
969 return kvm_is_error_hva(hva);
977 struct kvm_stat_data {
982 struct kvm_stats_debugfs_item {
985 enum kvm_stat_kind kind;
987 extern struct kvm_stats_debugfs_item debugfs_entries[];
988 extern struct dentry *kvm_debugfs_dir;
990 #if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
991 static inline int mmu_notifier_retry(struct kvm *kvm, unsigned long mmu_seq)
993 if (unlikely(kvm->mmu_notifier_count))
996 * Ensure the read of mmu_notifier_count happens before the read
997 * of mmu_notifier_seq. This interacts with the smp_wmb() in
998 * mmu_notifier_invalidate_range_end to make sure that the caller
999 * either sees the old (non-zero) value of mmu_notifier_count or
1000 * the new (incremented) value of mmu_notifier_seq.
1001 * PowerPC Book3s HV KVM calls this under a per-page lock
1002 * rather than under kvm->mmu_lock, for scalability, so
1003 * can't rely on kvm->mmu_lock to keep things ordered.
1006 if (kvm->mmu_notifier_seq != mmu_seq)
1012 #ifdef CONFIG_HAVE_KVM_IRQ_ROUTING
1015 #define KVM_MAX_IRQ_ROUTES 4096 //FIXME: we can have more than that...
1016 #elif defined(CONFIG_ARM64)
1017 #define KVM_MAX_IRQ_ROUTES 4096
1019 #define KVM_MAX_IRQ_ROUTES 1024
1022 int kvm_set_irq_routing(struct kvm *kvm,
1023 const struct kvm_irq_routing_entry *entries,
1026 int kvm_set_routing_entry(struct kvm *kvm,
1027 struct kvm_kernel_irq_routing_entry *e,
1028 const struct kvm_irq_routing_entry *ue);
1029 void kvm_free_irq_routing(struct kvm *kvm);
1033 static inline void kvm_free_irq_routing(struct kvm *kvm) {}
1037 int kvm_send_userspace_msi(struct kvm *kvm, struct kvm_msi *msi);
1039 #ifdef CONFIG_HAVE_KVM_EVENTFD
1041 void kvm_eventfd_init(struct kvm *kvm);
1042 int kvm_ioeventfd(struct kvm *kvm, struct kvm_ioeventfd *args);
1044 #ifdef CONFIG_HAVE_KVM_IRQFD
1045 int kvm_irqfd(struct kvm *kvm, struct kvm_irqfd *args);
1046 void kvm_irqfd_release(struct kvm *kvm);
1047 void kvm_irq_routing_update(struct kvm *);
1049 static inline int kvm_irqfd(struct kvm *kvm, struct kvm_irqfd *args)
1054 static inline void kvm_irqfd_release(struct kvm *kvm) {}
1059 static inline void kvm_eventfd_init(struct kvm *kvm) {}
1061 static inline int kvm_irqfd(struct kvm *kvm, struct kvm_irqfd *args)
1066 static inline void kvm_irqfd_release(struct kvm *kvm) {}
1068 #ifdef CONFIG_HAVE_KVM_IRQCHIP
1069 static inline void kvm_irq_routing_update(struct kvm *kvm)
1073 void kvm_arch_irq_routing_update(struct kvm *kvm);
1075 static inline int kvm_ioeventfd(struct kvm *kvm, struct kvm_ioeventfd *args)
1080 #endif /* CONFIG_HAVE_KVM_EVENTFD */
1082 static inline void kvm_make_request(int req, struct kvm_vcpu *vcpu)
1085 * Ensure the rest of the request is published to kvm_check_request's
1086 * caller. Paired with the smp_mb__after_atomic in kvm_check_request.
1089 set_bit(req, &vcpu->requests);
1092 static inline bool kvm_check_request(int req, struct kvm_vcpu *vcpu)
1094 if (test_bit(req, &vcpu->requests)) {
1095 clear_bit(req, &vcpu->requests);
1098 * Ensure the rest of the request is visible to kvm_check_request's
1099 * caller. Paired with the smp_wmb in kvm_make_request.
1101 smp_mb__after_atomic();
1108 extern bool kvm_rebooting;
1111 struct kvm_device_ops *ops;
1114 struct list_head vm_node;
1117 /* create, destroy, and name are mandatory */
1118 struct kvm_device_ops {
1122 * create is called holding kvm->lock and any operations not suitable
1123 * to do while holding the lock should be deferred to init (see
1126 int (*create)(struct kvm_device *dev, u32 type);
1129 * init is called after create if create is successful and is called
1130 * outside of holding kvm->lock.
1132 void (*init)(struct kvm_device *dev);
1135 * Destroy is responsible for freeing dev.
1137 * Destroy may be called before or after destructors are called
1138 * on emulated I/O regions, depending on whether a reference is
1139 * held by a vcpu or other kvm component that gets destroyed
1140 * after the emulated I/O.
1142 void (*destroy)(struct kvm_device *dev);
1144 int (*set_attr)(struct kvm_device *dev, struct kvm_device_attr *attr);
1145 int (*get_attr)(struct kvm_device *dev, struct kvm_device_attr *attr);
1146 int (*has_attr)(struct kvm_device *dev, struct kvm_device_attr *attr);
1147 long (*ioctl)(struct kvm_device *dev, unsigned int ioctl,
1151 void kvm_device_get(struct kvm_device *dev);
1152 void kvm_device_put(struct kvm_device *dev);
1153 struct kvm_device *kvm_device_from_filp(struct file *filp);
1154 int kvm_register_device_ops(struct kvm_device_ops *ops, u32 type);
1155 void kvm_unregister_device_ops(u32 type);
1157 extern struct kvm_device_ops kvm_mpic_ops;
1158 extern struct kvm_device_ops kvm_xics_ops;
1159 extern struct kvm_device_ops kvm_arm_vgic_v2_ops;
1160 extern struct kvm_device_ops kvm_arm_vgic_v3_ops;
1162 #ifdef CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT
1164 static inline void kvm_vcpu_set_in_spin_loop(struct kvm_vcpu *vcpu, bool val)
1166 vcpu->spin_loop.in_spin_loop = val;
1168 static inline void kvm_vcpu_set_dy_eligible(struct kvm_vcpu *vcpu, bool val)
1170 vcpu->spin_loop.dy_eligible = val;
1173 #else /* !CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT */
1175 static inline void kvm_vcpu_set_in_spin_loop(struct kvm_vcpu *vcpu, bool val)
1179 static inline void kvm_vcpu_set_dy_eligible(struct kvm_vcpu *vcpu, bool val)
1182 #endif /* CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT */
1184 #ifdef CONFIG_HAVE_KVM_IRQ_BYPASS
1185 bool kvm_arch_has_irq_bypass(void);
1186 int kvm_arch_irq_bypass_add_producer(struct irq_bypass_consumer *,
1187 struct irq_bypass_producer *);
1188 void kvm_arch_irq_bypass_del_producer(struct irq_bypass_consumer *,
1189 struct irq_bypass_producer *);
1190 void kvm_arch_irq_bypass_stop(struct irq_bypass_consumer *);
1191 void kvm_arch_irq_bypass_start(struct irq_bypass_consumer *);
1192 int kvm_arch_update_irqfd_routing(struct kvm *kvm, unsigned int host_irq,
1193 uint32_t guest_irq, bool set);
1194 #endif /* CONFIG_HAVE_KVM_IRQ_BYPASS */
1196 #ifdef CONFIG_HAVE_KVM_INVALID_WAKEUPS
1197 /* If we wakeup during the poll time, was it a sucessful poll? */
1198 static inline bool vcpu_valid_wakeup(struct kvm_vcpu *vcpu)
1200 return vcpu->valid_wakeup;
1204 static inline bool vcpu_valid_wakeup(struct kvm_vcpu *vcpu)
1208 #endif /* CONFIG_HAVE_KVM_INVALID_WAKEUPS */