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 <linux/nospec.h>
30 #include <asm/signal.h>
32 #include <linux/kvm.h>
33 #include <linux/kvm_para.h>
35 #include <linux/kvm_types.h>
37 #include <asm/kvm_host.h>
39 #ifndef KVM_MAX_VCPU_ID
40 #define KVM_MAX_VCPU_ID KVM_MAX_VCPUS
44 * The bit 16 ~ bit 31 of kvm_memory_region::flags are internally used
45 * in kvm, other bits are visible for userspace which are defined in
46 * include/linux/kvm_h.
48 #define KVM_MEMSLOT_INVALID (1UL << 16)
49 #define KVM_MEMSLOT_INCOHERENT (1UL << 17)
51 /* Two fragments for cross MMIO pages. */
52 #define KVM_MAX_MMIO_FRAGMENTS 2
54 #ifndef KVM_ADDRESS_SPACE_NUM
55 #define KVM_ADDRESS_SPACE_NUM 1
59 * For the normal pfn, the highest 12 bits should be zero,
60 * so we can mask bit 62 ~ bit 52 to indicate the error pfn,
61 * mask bit 63 to indicate the noslot pfn.
63 #define KVM_PFN_ERR_MASK (0x7ffULL << 52)
64 #define KVM_PFN_ERR_NOSLOT_MASK (0xfffULL << 52)
65 #define KVM_PFN_NOSLOT (0x1ULL << 63)
67 #define KVM_PFN_ERR_FAULT (KVM_PFN_ERR_MASK)
68 #define KVM_PFN_ERR_HWPOISON (KVM_PFN_ERR_MASK + 1)
69 #define KVM_PFN_ERR_RO_FAULT (KVM_PFN_ERR_MASK + 2)
72 * error pfns indicate that the gfn is in slot but faild to
73 * translate it to pfn on host.
75 static inline bool is_error_pfn(kvm_pfn_t pfn)
77 return !!(pfn & KVM_PFN_ERR_MASK);
81 * error_noslot pfns indicate that the gfn can not be
82 * translated to pfn - it is not in slot or failed to
83 * translate it to pfn.
85 static inline bool is_error_noslot_pfn(kvm_pfn_t pfn)
87 return !!(pfn & KVM_PFN_ERR_NOSLOT_MASK);
90 /* noslot pfn indicates that the gfn is not in slot. */
91 static inline bool is_noslot_pfn(kvm_pfn_t pfn)
93 return pfn == KVM_PFN_NOSLOT;
97 * architectures with KVM_HVA_ERR_BAD other than PAGE_OFFSET (e.g. s390)
98 * provide own defines and kvm_is_error_hva
100 #ifndef KVM_HVA_ERR_BAD
102 #define KVM_HVA_ERR_BAD (PAGE_OFFSET)
103 #define KVM_HVA_ERR_RO_BAD (PAGE_OFFSET + PAGE_SIZE)
105 static inline bool kvm_is_error_hva(unsigned long addr)
107 return addr >= PAGE_OFFSET;
112 #define KVM_ERR_PTR_BAD_PAGE (ERR_PTR(-ENOENT))
114 static inline bool is_error_page(struct page *page)
120 * Architecture-independent vcpu->requests bit members
121 * Bits 4-7 are reserved for more arch-independent bits.
123 #define KVM_REQ_TLB_FLUSH 0
124 #define KVM_REQ_MMU_RELOAD 1
125 #define KVM_REQ_PENDING_TIMER 2
126 #define KVM_REQ_UNHALT 3
128 #define KVM_USERSPACE_IRQ_SOURCE_ID 0
129 #define KVM_IRQFD_RESAMPLE_IRQ_SOURCE_ID 1
131 extern struct kmem_cache *kvm_vcpu_cache;
133 extern struct mutex kvm_lock;
134 extern struct list_head vm_list;
136 struct kvm_io_range {
139 struct kvm_io_device *dev;
142 #define NR_IOBUS_DEVS 1000
147 struct kvm_io_range range[];
153 KVM_VIRTIO_CCW_NOTIFY_BUS,
158 int kvm_io_bus_write(struct kvm_vcpu *vcpu, enum kvm_bus bus_idx, gpa_t addr,
159 int len, const void *val);
160 int kvm_io_bus_write_cookie(struct kvm_vcpu *vcpu, enum kvm_bus bus_idx,
161 gpa_t addr, int len, const void *val, long cookie);
162 int kvm_io_bus_read(struct kvm_vcpu *vcpu, enum kvm_bus bus_idx, gpa_t addr,
164 int kvm_io_bus_register_dev(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr,
165 int len, struct kvm_io_device *dev);
166 void kvm_io_bus_unregister_dev(struct kvm *kvm, enum kvm_bus bus_idx,
167 struct kvm_io_device *dev);
168 struct kvm_io_device *kvm_io_bus_get_dev(struct kvm *kvm, enum kvm_bus bus_idx,
171 #ifdef CONFIG_KVM_ASYNC_PF
172 struct kvm_async_pf {
173 struct work_struct work;
174 struct list_head link;
175 struct list_head queue;
176 struct kvm_vcpu *vcpu;
177 struct mm_struct *mm;
180 struct kvm_arch_async_pf arch;
184 void kvm_clear_async_pf_completion_queue(struct kvm_vcpu *vcpu);
185 void kvm_check_async_pf_completion(struct kvm_vcpu *vcpu);
186 int kvm_setup_async_pf(struct kvm_vcpu *vcpu, gva_t gva, unsigned long hva,
187 struct kvm_arch_async_pf *arch);
188 int kvm_async_pf_wakeup_all(struct kvm_vcpu *vcpu);
195 READING_SHADOW_PAGE_TABLES,
199 * Sometimes a large or cross-page mmio needs to be broken up into separate
200 * exits for userspace servicing.
202 struct kvm_mmio_fragment {
210 #ifdef CONFIG_PREEMPT_NOTIFIERS
211 struct preempt_notifier preempt_notifier;
217 unsigned long requests;
218 unsigned long guest_debug;
221 struct list_head blocked_vcpu_list;
227 int guest_fpu_loaded, guest_xcr0_loaded;
228 unsigned char fpu_counter;
229 struct swait_queue_head wq;
233 struct kvm_vcpu_stat stat;
234 unsigned int halt_poll_ns;
237 #ifdef CONFIG_HAS_IOMEM
239 int mmio_read_completed;
241 int mmio_cur_fragment;
242 int mmio_nr_fragments;
243 struct kvm_mmio_fragment mmio_fragments[KVM_MAX_MMIO_FRAGMENTS];
246 #ifdef CONFIG_KVM_ASYNC_PF
249 struct list_head queue;
250 struct list_head done;
255 #ifdef CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT
257 * Cpu relax intercept or pause loop exit optimization
258 * in_spin_loop: set when a vcpu does a pause loop exit
259 * or cpu relax intercepted.
260 * dy_eligible: indicates whether vcpu is eligible for directed yield.
268 struct kvm_vcpu_arch arch;
269 struct dentry *debugfs_dentry;
272 static inline int kvm_vcpu_exiting_guest_mode(struct kvm_vcpu *vcpu)
274 return cmpxchg(&vcpu->mode, IN_GUEST_MODE, EXITING_GUEST_MODE);
278 * Some of the bitops functions do not support too long bitmaps.
279 * This number must be determined not to exceed such limits.
281 #define KVM_MEM_MAX_NR_PAGES ((1UL << 31) - 1)
283 struct kvm_memory_slot {
285 unsigned long npages;
286 unsigned long *dirty_bitmap;
287 struct kvm_arch_memory_slot arch;
288 unsigned long userspace_addr;
293 static inline unsigned long kvm_dirty_bitmap_bytes(struct kvm_memory_slot *memslot)
295 return ALIGN(memslot->npages, BITS_PER_LONG) / 8;
298 struct kvm_s390_adapter_int {
311 struct kvm_kernel_irq_routing_entry {
314 int (*set)(struct kvm_kernel_irq_routing_entry *e,
315 struct kvm *kvm, int irq_source_id, int level,
329 struct kvm_s390_adapter_int adapter;
330 struct kvm_hv_sint hv_sint;
332 struct hlist_node link;
335 #ifdef CONFIG_HAVE_KVM_IRQ_ROUTING
336 struct kvm_irq_routing_table {
337 int chip[KVM_NR_IRQCHIPS][KVM_IRQCHIP_NUM_PINS];
340 * Array indexed by gsi. Each entry contains list of irq chips
341 * the gsi is connected to.
343 struct hlist_head map[0];
347 #ifndef KVM_PRIVATE_MEM_SLOTS
348 #define KVM_PRIVATE_MEM_SLOTS 0
351 #ifndef KVM_MEM_SLOTS_NUM
352 #define KVM_MEM_SLOTS_NUM (KVM_USER_MEM_SLOTS + KVM_PRIVATE_MEM_SLOTS)
355 #ifndef __KVM_VCPU_MULTIPLE_ADDRESS_SPACE
356 static inline int kvm_arch_vcpu_memslots_id(struct kvm_vcpu *vcpu)
364 * memslots are not sorted by id anymore, please use id_to_memslot()
365 * to get the memslot by its id.
367 struct kvm_memslots {
369 struct kvm_memory_slot memslots[KVM_MEM_SLOTS_NUM];
370 /* The mapping table from slot id to the index in memslots[]. */
371 short id_to_index[KVM_MEM_SLOTS_NUM];
378 struct mutex slots_lock;
379 struct mm_struct *mm; /* userspace tied to this vm */
380 struct kvm_memslots *memslots[KVM_ADDRESS_SPACE_NUM];
381 struct srcu_struct srcu;
382 struct srcu_struct irq_srcu;
383 struct kvm_vcpu *vcpus[KVM_MAX_VCPUS];
386 * created_vcpus is protected by kvm->lock, and is incremented
387 * at the beginning of KVM_CREATE_VCPU. online_vcpus is only
388 * incremented after storing the kvm_vcpu pointer in vcpus,
389 * and is accessed atomically.
391 atomic_t online_vcpus;
393 int last_boosted_vcpu;
394 struct list_head vm_list;
396 struct kvm_io_bus *buses[KVM_NR_BUSES];
397 #ifdef CONFIG_HAVE_KVM_EVENTFD
400 struct list_head items;
401 struct list_head resampler_list;
402 struct mutex resampler_lock;
404 struct list_head ioeventfds;
406 struct kvm_vm_stat stat;
407 struct kvm_arch arch;
408 atomic_t users_count;
409 #ifdef KVM_COALESCED_MMIO_PAGE_OFFSET
410 struct kvm_coalesced_mmio_ring *coalesced_mmio_ring;
411 spinlock_t ring_lock;
412 struct list_head coalesced_zones;
415 struct mutex irq_lock;
416 #ifdef CONFIG_HAVE_KVM_IRQCHIP
418 * Update side is protected by irq_lock.
420 struct kvm_irq_routing_table __rcu *irq_routing;
422 #ifdef CONFIG_HAVE_KVM_IRQFD
423 struct hlist_head irq_ack_notifier_list;
426 #if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
427 struct mmu_notifier mmu_notifier;
428 unsigned long mmu_notifier_seq;
429 long mmu_notifier_count;
432 struct list_head devices;
433 struct dentry *debugfs_dentry;
434 struct kvm_stat_data **debugfs_stat_data;
437 #define kvm_err(fmt, ...) \
438 pr_err("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
439 #define kvm_info(fmt, ...) \
440 pr_info("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
441 #define kvm_debug(fmt, ...) \
442 pr_debug("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
443 #define kvm_pr_unimpl(fmt, ...) \
444 pr_err_ratelimited("kvm [%i]: " fmt, \
445 task_tgid_nr(current), ## __VA_ARGS__)
447 /* The guest did something we don't support. */
448 #define vcpu_unimpl(vcpu, fmt, ...) \
449 kvm_pr_unimpl("vcpu%i, guest rIP: 0x%lx " fmt, \
450 (vcpu)->vcpu_id, kvm_rip_read(vcpu), ## __VA_ARGS__)
452 #define vcpu_debug(vcpu, fmt, ...) \
453 kvm_debug("vcpu%i " fmt, (vcpu)->vcpu_id, ## __VA_ARGS__)
454 #define vcpu_err(vcpu, fmt, ...) \
455 kvm_err("vcpu%i " fmt, (vcpu)->vcpu_id, ## __VA_ARGS__)
457 static inline struct kvm_vcpu *kvm_get_vcpu(struct kvm *kvm, int i)
459 /* Pairs with smp_wmb() in kvm_vm_ioctl_create_vcpu, in case
460 * the caller has read kvm->online_vcpus before (as is the case
461 * for kvm_for_each_vcpu, for example).
464 return kvm->vcpus[i];
467 #define kvm_for_each_vcpu(idx, vcpup, kvm) \
469 idx < atomic_read(&kvm->online_vcpus) && \
470 (vcpup = kvm_get_vcpu(kvm, idx)) != NULL; \
473 static inline struct kvm_vcpu *kvm_get_vcpu_by_id(struct kvm *kvm, int id)
475 struct kvm_vcpu *vcpu = NULL;
480 if (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 kvm_pfn_t gfn_to_pfn_atomic(struct kvm *kvm, gfn_t gfn);
622 kvm_pfn_t gfn_to_pfn(struct kvm *kvm, gfn_t gfn);
623 kvm_pfn_t gfn_to_pfn_prot(struct kvm *kvm, gfn_t gfn, bool write_fault,
625 kvm_pfn_t gfn_to_pfn_memslot(struct kvm_memory_slot *slot, gfn_t gfn);
626 kvm_pfn_t gfn_to_pfn_memslot_atomic(struct kvm_memory_slot *slot, gfn_t gfn);
627 kvm_pfn_t __gfn_to_pfn_memslot(struct kvm_memory_slot *slot, gfn_t gfn,
628 bool atomic, bool *async, bool write_fault,
631 void kvm_release_pfn_clean(kvm_pfn_t pfn);
632 void kvm_set_pfn_dirty(kvm_pfn_t pfn);
633 void kvm_set_pfn_accessed(kvm_pfn_t pfn);
634 void kvm_get_pfn(kvm_pfn_t pfn);
636 int kvm_read_guest_page(struct kvm *kvm, gfn_t gfn, void *data, int offset,
638 int kvm_read_guest_atomic(struct kvm *kvm, gpa_t gpa, void *data,
640 int kvm_read_guest(struct kvm *kvm, gpa_t gpa, void *data, unsigned long len);
641 int kvm_read_guest_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
642 void *data, unsigned long len);
643 int kvm_write_guest_page(struct kvm *kvm, gfn_t gfn, const void *data,
644 int offset, int len);
645 int kvm_write_guest(struct kvm *kvm, gpa_t gpa, const void *data,
647 int kvm_write_guest_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
648 void *data, unsigned long len);
649 int kvm_gfn_to_hva_cache_init(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
650 gpa_t gpa, unsigned long len);
651 int kvm_clear_guest_page(struct kvm *kvm, gfn_t gfn, int offset, int len);
652 int kvm_clear_guest(struct kvm *kvm, gpa_t gpa, unsigned long len);
653 struct kvm_memory_slot *gfn_to_memslot(struct kvm *kvm, gfn_t gfn);
654 bool kvm_is_visible_gfn(struct kvm *kvm, gfn_t gfn);
655 unsigned long kvm_host_page_size(struct kvm *kvm, gfn_t gfn);
656 void mark_page_dirty(struct kvm *kvm, gfn_t gfn);
658 struct kvm_memslots *kvm_vcpu_memslots(struct kvm_vcpu *vcpu);
659 struct kvm_memory_slot *kvm_vcpu_gfn_to_memslot(struct kvm_vcpu *vcpu, gfn_t gfn);
660 kvm_pfn_t kvm_vcpu_gfn_to_pfn_atomic(struct kvm_vcpu *vcpu, gfn_t gfn);
661 kvm_pfn_t kvm_vcpu_gfn_to_pfn(struct kvm_vcpu *vcpu, gfn_t gfn);
662 struct page *kvm_vcpu_gfn_to_page(struct kvm_vcpu *vcpu, gfn_t gfn);
663 unsigned long kvm_vcpu_gfn_to_hva(struct kvm_vcpu *vcpu, gfn_t gfn);
664 unsigned long kvm_vcpu_gfn_to_hva_prot(struct kvm_vcpu *vcpu, gfn_t gfn, bool *writable);
665 int kvm_vcpu_read_guest_page(struct kvm_vcpu *vcpu, gfn_t gfn, void *data, int offset,
667 int kvm_vcpu_read_guest_atomic(struct kvm_vcpu *vcpu, gpa_t gpa, void *data,
669 int kvm_vcpu_read_guest(struct kvm_vcpu *vcpu, gpa_t gpa, void *data,
671 int kvm_vcpu_write_guest_page(struct kvm_vcpu *vcpu, gfn_t gfn, const void *data,
672 int offset, int len);
673 int kvm_vcpu_write_guest(struct kvm_vcpu *vcpu, gpa_t gpa, const void *data,
675 void kvm_vcpu_mark_page_dirty(struct kvm_vcpu *vcpu, gfn_t gfn);
677 void kvm_vcpu_block(struct kvm_vcpu *vcpu);
678 void kvm_arch_vcpu_blocking(struct kvm_vcpu *vcpu);
679 void kvm_arch_vcpu_unblocking(struct kvm_vcpu *vcpu);
680 void kvm_vcpu_wake_up(struct kvm_vcpu *vcpu);
681 void kvm_vcpu_kick(struct kvm_vcpu *vcpu);
682 int kvm_vcpu_yield_to(struct kvm_vcpu *target);
683 void kvm_vcpu_on_spin(struct kvm_vcpu *vcpu);
684 void kvm_load_guest_fpu(struct kvm_vcpu *vcpu);
685 void kvm_put_guest_fpu(struct kvm_vcpu *vcpu);
687 void kvm_flush_remote_tlbs(struct kvm *kvm);
688 void kvm_reload_remote_mmus(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 bool kvm_arch_has_vcpu_debugfs(void);
755 int kvm_arch_create_vcpu_debugfs(struct kvm_vcpu *vcpu);
757 int kvm_arch_hardware_enable(void);
758 void kvm_arch_hardware_disable(void);
759 int kvm_arch_hardware_setup(void);
760 void kvm_arch_hardware_unsetup(void);
761 void kvm_arch_check_processor_compat(void *rtn);
762 int kvm_arch_vcpu_runnable(struct kvm_vcpu *vcpu);
763 int kvm_arch_vcpu_should_kick(struct kvm_vcpu *vcpu);
765 void *kvm_kvzalloc(unsigned long size);
767 #ifndef __KVM_HAVE_ARCH_VM_ALLOC
768 static inline struct kvm *kvm_arch_alloc_vm(void)
770 return kzalloc(sizeof(struct kvm), GFP_KERNEL);
773 static inline void kvm_arch_free_vm(struct kvm *kvm)
779 #ifdef __KVM_HAVE_ARCH_NONCOHERENT_DMA
780 void kvm_arch_register_noncoherent_dma(struct kvm *kvm);
781 void kvm_arch_unregister_noncoherent_dma(struct kvm *kvm);
782 bool kvm_arch_has_noncoherent_dma(struct kvm *kvm);
784 static inline void kvm_arch_register_noncoherent_dma(struct kvm *kvm)
788 static inline void kvm_arch_unregister_noncoherent_dma(struct kvm *kvm)
792 static inline bool kvm_arch_has_noncoherent_dma(struct kvm *kvm)
797 #ifdef __KVM_HAVE_ARCH_ASSIGNED_DEVICE
798 void kvm_arch_start_assignment(struct kvm *kvm);
799 void kvm_arch_end_assignment(struct kvm *kvm);
800 bool kvm_arch_has_assigned_device(struct kvm *kvm);
802 static inline void kvm_arch_start_assignment(struct kvm *kvm)
806 static inline void kvm_arch_end_assignment(struct kvm *kvm)
810 static inline bool kvm_arch_has_assigned_device(struct kvm *kvm)
816 static inline struct swait_queue_head *kvm_arch_vcpu_wq(struct kvm_vcpu *vcpu)
818 #ifdef __KVM_HAVE_ARCH_WQP
819 return vcpu->arch.wqp;
825 #ifdef __KVM_HAVE_ARCH_INTC_INITIALIZED
827 * returns true if the virtual interrupt controller is initialized and
828 * ready to accept virtual IRQ. On some architectures the virtual interrupt
829 * controller is dynamically instantiated and this is not always true.
831 bool kvm_arch_intc_initialized(struct kvm *kvm);
833 static inline bool kvm_arch_intc_initialized(struct kvm *kvm)
839 int kvm_arch_init_vm(struct kvm *kvm, unsigned long type);
840 void kvm_arch_destroy_vm(struct kvm *kvm);
841 void kvm_arch_sync_events(struct kvm *kvm);
843 int kvm_cpu_has_pending_timer(struct kvm_vcpu *vcpu);
844 void kvm_vcpu_kick(struct kvm_vcpu *vcpu);
846 bool kvm_is_reserved_pfn(kvm_pfn_t pfn);
847 bool kvm_is_zone_device_pfn(kvm_pfn_t pfn);
849 struct kvm_irq_ack_notifier {
850 struct hlist_node link;
852 void (*irq_acked)(struct kvm_irq_ack_notifier *kian);
855 int kvm_irq_map_gsi(struct kvm *kvm,
856 struct kvm_kernel_irq_routing_entry *entries, int gsi);
857 int kvm_irq_map_chip_pin(struct kvm *kvm, unsigned irqchip, unsigned pin);
859 int kvm_set_irq(struct kvm *kvm, int irq_source_id, u32 irq, int level,
861 int kvm_set_msi(struct kvm_kernel_irq_routing_entry *irq_entry, struct kvm *kvm,
862 int irq_source_id, int level, bool line_status);
863 int kvm_arch_set_irq_inatomic(struct kvm_kernel_irq_routing_entry *e,
864 struct kvm *kvm, int irq_source_id,
865 int level, bool line_status);
866 bool kvm_irq_has_notifier(struct kvm *kvm, unsigned irqchip, unsigned pin);
867 void kvm_notify_acked_gsi(struct kvm *kvm, int gsi);
868 void kvm_notify_acked_irq(struct kvm *kvm, unsigned irqchip, unsigned pin);
869 void kvm_register_irq_ack_notifier(struct kvm *kvm,
870 struct kvm_irq_ack_notifier *kian);
871 void kvm_unregister_irq_ack_notifier(struct kvm *kvm,
872 struct kvm_irq_ack_notifier *kian);
873 int kvm_request_irq_source_id(struct kvm *kvm);
874 void kvm_free_irq_source_id(struct kvm *kvm, int irq_source_id);
876 #ifdef CONFIG_KVM_DEVICE_ASSIGNMENT
877 int kvm_iommu_map_pages(struct kvm *kvm, struct kvm_memory_slot *slot);
878 void kvm_iommu_unmap_pages(struct kvm *kvm, struct kvm_memory_slot *slot);
880 static inline int kvm_iommu_map_pages(struct kvm *kvm,
881 struct kvm_memory_slot *slot)
886 static inline void kvm_iommu_unmap_pages(struct kvm *kvm,
887 struct kvm_memory_slot *slot)
893 * search_memslots() and __gfn_to_memslot() are here because they are
894 * used in non-modular code in arch/powerpc/kvm/book3s_hv_rm_mmu.c.
895 * gfn_to_memslot() itself isn't here as an inline because that would
896 * bloat other code too much.
898 static inline struct kvm_memory_slot *
899 search_memslots(struct kvm_memslots *slots, gfn_t gfn)
901 int start = 0, end = slots->used_slots;
902 int slot = atomic_read(&slots->lru_slot);
903 struct kvm_memory_slot *memslots = slots->memslots;
905 if (gfn >= memslots[slot].base_gfn &&
906 gfn < memslots[slot].base_gfn + memslots[slot].npages)
907 return &memslots[slot];
909 while (start < end) {
910 slot = start + (end - start) / 2;
912 if (gfn >= memslots[slot].base_gfn)
918 if (start < slots->used_slots && gfn >= memslots[start].base_gfn &&
919 gfn < memslots[start].base_gfn + memslots[start].npages) {
920 atomic_set(&slots->lru_slot, start);
921 return &memslots[start];
927 static inline struct kvm_memory_slot *
928 __gfn_to_memslot(struct kvm_memslots *slots, gfn_t gfn)
930 return search_memslots(slots, gfn);
933 static inline unsigned long
934 __gfn_to_hva_memslot(struct kvm_memory_slot *slot, gfn_t gfn)
937 * The index was checked originally in search_memslots. To avoid
938 * that a malicious guest builds a Spectre gadget out of e.g. page
939 * table walks, do not let the processor speculate loads outside
940 * the guest's registered memslots.
942 unsigned long offset = gfn - slot->base_gfn;
943 offset = array_index_nospec(offset, slot->npages);
944 return slot->userspace_addr + offset * PAGE_SIZE;
947 static inline int memslot_id(struct kvm *kvm, gfn_t gfn)
949 return gfn_to_memslot(kvm, gfn)->id;
953 hva_to_gfn_memslot(unsigned long hva, struct kvm_memory_slot *slot)
955 gfn_t gfn_offset = (hva - slot->userspace_addr) >> PAGE_SHIFT;
957 return slot->base_gfn + gfn_offset;
960 static inline gpa_t gfn_to_gpa(gfn_t gfn)
962 return (gpa_t)gfn << PAGE_SHIFT;
965 static inline gfn_t gpa_to_gfn(gpa_t gpa)
967 return (gfn_t)(gpa >> PAGE_SHIFT);
970 static inline hpa_t pfn_to_hpa(kvm_pfn_t pfn)
972 return (hpa_t)pfn << PAGE_SHIFT;
975 static inline bool kvm_is_error_gpa(struct kvm *kvm, gpa_t gpa)
977 unsigned long hva = gfn_to_hva(kvm, gpa_to_gfn(gpa));
979 return kvm_is_error_hva(hva);
987 struct kvm_stat_data {
992 struct kvm_stats_debugfs_item {
995 enum kvm_stat_kind kind;
997 extern struct kvm_stats_debugfs_item debugfs_entries[];
998 extern struct dentry *kvm_debugfs_dir;
1000 #if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
1001 static inline int mmu_notifier_retry(struct kvm *kvm, unsigned long mmu_seq)
1003 if (unlikely(kvm->mmu_notifier_count))
1006 * Ensure the read of mmu_notifier_count happens before the read
1007 * of mmu_notifier_seq. This interacts with the smp_wmb() in
1008 * mmu_notifier_invalidate_range_end to make sure that the caller
1009 * either sees the old (non-zero) value of mmu_notifier_count or
1010 * the new (incremented) value of mmu_notifier_seq.
1011 * PowerPC Book3s HV KVM calls this under a per-page lock
1012 * rather than under kvm->mmu_lock, for scalability, so
1013 * can't rely on kvm->mmu_lock to keep things ordered.
1016 if (kvm->mmu_notifier_seq != mmu_seq)
1022 #ifdef CONFIG_HAVE_KVM_IRQ_ROUTING
1025 #define KVM_MAX_IRQ_ROUTES 4096 //FIXME: we can have more than that...
1026 #elif defined(CONFIG_ARM64)
1027 #define KVM_MAX_IRQ_ROUTES 4096
1029 #define KVM_MAX_IRQ_ROUTES 1024
1032 int kvm_set_irq_routing(struct kvm *kvm,
1033 const struct kvm_irq_routing_entry *entries,
1036 int kvm_set_routing_entry(struct kvm *kvm,
1037 struct kvm_kernel_irq_routing_entry *e,
1038 const struct kvm_irq_routing_entry *ue);
1039 void kvm_free_irq_routing(struct kvm *kvm);
1043 static inline void kvm_free_irq_routing(struct kvm *kvm) {}
1047 int kvm_send_userspace_msi(struct kvm *kvm, struct kvm_msi *msi);
1049 #ifdef CONFIG_HAVE_KVM_EVENTFD
1051 void kvm_eventfd_init(struct kvm *kvm);
1052 int kvm_ioeventfd(struct kvm *kvm, struct kvm_ioeventfd *args);
1054 #ifdef CONFIG_HAVE_KVM_IRQFD
1055 int kvm_irqfd(struct kvm *kvm, struct kvm_irqfd *args);
1056 void kvm_irqfd_release(struct kvm *kvm);
1057 void kvm_irq_routing_update(struct kvm *);
1059 static inline int kvm_irqfd(struct kvm *kvm, struct kvm_irqfd *args)
1064 static inline void kvm_irqfd_release(struct kvm *kvm) {}
1069 static inline void kvm_eventfd_init(struct kvm *kvm) {}
1071 static inline int kvm_irqfd(struct kvm *kvm, struct kvm_irqfd *args)
1076 static inline void kvm_irqfd_release(struct kvm *kvm) {}
1078 #ifdef CONFIG_HAVE_KVM_IRQCHIP
1079 static inline void kvm_irq_routing_update(struct kvm *kvm)
1084 static inline int kvm_ioeventfd(struct kvm *kvm, struct kvm_ioeventfd *args)
1089 #endif /* CONFIG_HAVE_KVM_EVENTFD */
1091 void kvm_arch_irq_routing_update(struct kvm *kvm);
1093 static inline void kvm_make_request(int req, struct kvm_vcpu *vcpu)
1096 * Ensure the rest of the request is published to kvm_check_request's
1097 * caller. Paired with the smp_mb__after_atomic in kvm_check_request.
1100 set_bit(req, &vcpu->requests);
1103 static inline bool kvm_check_request(int req, struct kvm_vcpu *vcpu)
1105 if (test_bit(req, &vcpu->requests)) {
1106 clear_bit(req, &vcpu->requests);
1109 * Ensure the rest of the request is visible to kvm_check_request's
1110 * caller. Paired with the smp_wmb in kvm_make_request.
1112 smp_mb__after_atomic();
1119 extern bool kvm_rebooting;
1122 struct kvm_device_ops *ops;
1125 struct list_head vm_node;
1128 /* create, destroy, and name are mandatory */
1129 struct kvm_device_ops {
1133 * create is called holding kvm->lock and any operations not suitable
1134 * to do while holding the lock should be deferred to init (see
1137 int (*create)(struct kvm_device *dev, u32 type);
1140 * init is called after create if create is successful and is called
1141 * outside of holding kvm->lock.
1143 void (*init)(struct kvm_device *dev);
1146 * Destroy is responsible for freeing dev.
1148 * Destroy may be called before or after destructors are called
1149 * on emulated I/O regions, depending on whether a reference is
1150 * held by a vcpu or other kvm component that gets destroyed
1151 * after the emulated I/O.
1153 void (*destroy)(struct kvm_device *dev);
1155 int (*set_attr)(struct kvm_device *dev, struct kvm_device_attr *attr);
1156 int (*get_attr)(struct kvm_device *dev, struct kvm_device_attr *attr);
1157 int (*has_attr)(struct kvm_device *dev, struct kvm_device_attr *attr);
1158 long (*ioctl)(struct kvm_device *dev, unsigned int ioctl,
1162 void kvm_device_get(struct kvm_device *dev);
1163 void kvm_device_put(struct kvm_device *dev);
1164 struct kvm_device *kvm_device_from_filp(struct file *filp);
1165 int kvm_register_device_ops(struct kvm_device_ops *ops, u32 type);
1166 void kvm_unregister_device_ops(u32 type);
1168 extern struct kvm_device_ops kvm_mpic_ops;
1169 extern struct kvm_device_ops kvm_xics_ops;
1170 extern struct kvm_device_ops kvm_arm_vgic_v2_ops;
1171 extern struct kvm_device_ops kvm_arm_vgic_v3_ops;
1173 #ifdef CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT
1175 static inline void kvm_vcpu_set_in_spin_loop(struct kvm_vcpu *vcpu, bool val)
1177 vcpu->spin_loop.in_spin_loop = val;
1179 static inline void kvm_vcpu_set_dy_eligible(struct kvm_vcpu *vcpu, bool val)
1181 vcpu->spin_loop.dy_eligible = val;
1184 #else /* !CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT */
1186 static inline void kvm_vcpu_set_in_spin_loop(struct kvm_vcpu *vcpu, bool val)
1190 static inline void kvm_vcpu_set_dy_eligible(struct kvm_vcpu *vcpu, bool val)
1193 #endif /* CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT */
1195 #ifdef CONFIG_HAVE_KVM_IRQ_BYPASS
1196 bool kvm_arch_has_irq_bypass(void);
1197 int kvm_arch_irq_bypass_add_producer(struct irq_bypass_consumer *,
1198 struct irq_bypass_producer *);
1199 void kvm_arch_irq_bypass_del_producer(struct irq_bypass_consumer *,
1200 struct irq_bypass_producer *);
1201 void kvm_arch_irq_bypass_stop(struct irq_bypass_consumer *);
1202 void kvm_arch_irq_bypass_start(struct irq_bypass_consumer *);
1203 int kvm_arch_update_irqfd_routing(struct kvm *kvm, unsigned int host_irq,
1204 uint32_t guest_irq, bool set);
1205 #endif /* CONFIG_HAVE_KVM_IRQ_BYPASS */
1207 #ifdef CONFIG_HAVE_KVM_INVALID_WAKEUPS
1208 /* If we wakeup during the poll time, was it a sucessful poll? */
1209 static inline bool vcpu_valid_wakeup(struct kvm_vcpu *vcpu)
1211 return vcpu->valid_wakeup;
1215 static inline bool vcpu_valid_wakeup(struct kvm_vcpu *vcpu)
1219 #endif /* CONFIG_HAVE_KVM_INVALID_WAKEUPS */
1221 typedef int (*kvm_vm_thread_fn_t)(struct kvm *kvm, uintptr_t data);
1223 int kvm_vm_create_worker_thread(struct kvm *kvm, kvm_vm_thread_fn_t thread_fn,
1224 uintptr_t data, const char *name,
1225 struct task_struct **thread_ptr);