1 /* SPDX-License-Identifier: GPL-2.0-only */
6 #include <linux/types.h>
7 #include <linux/hardirq.h>
8 #include <linux/list.h>
9 #include <linux/mutex.h>
10 #include <linux/spinlock.h>
11 #include <linux/signal.h>
12 #include <linux/sched.h>
13 #include <linux/bug.h>
14 #include <linux/minmax.h>
16 #include <linux/mmu_notifier.h>
17 #include <linux/preempt.h>
18 #include <linux/msi.h>
19 #include <linux/slab.h>
20 #include <linux/vmalloc.h>
21 #include <linux/rcupdate.h>
22 #include <linux/ratelimit.h>
23 #include <linux/err.h>
24 #include <linux/irqflags.h>
25 #include <linux/context_tracking.h>
26 #include <linux/irqbypass.h>
27 #include <linux/rcuwait.h>
28 #include <linux/refcount.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>
38 #include <linux/kvm_dirty_ring.h>
40 #ifndef KVM_MAX_VCPU_ID
41 #define KVM_MAX_VCPU_ID KVM_MAX_VCPUS
45 * The bit 16 ~ bit 31 of kvm_memory_region::flags are internally used
46 * in kvm, other bits are visible for userspace which are defined in
47 * include/linux/kvm_h.
49 #define KVM_MEMSLOT_INVALID (1UL << 16)
52 * Bit 63 of the memslot generation number is an "update in-progress flag",
53 * e.g. is temporarily set for the duration of install_new_memslots().
54 * This flag effectively creates a unique generation number that is used to
55 * mark cached memslot data, e.g. MMIO accesses, as potentially being stale,
56 * i.e. may (or may not) have come from the previous memslots generation.
58 * This is necessary because the actual memslots update is not atomic with
59 * respect to the generation number update. Updating the generation number
60 * first would allow a vCPU to cache a spte from the old memslots using the
61 * new generation number, and updating the generation number after switching
62 * to the new memslots would allow cache hits using the old generation number
63 * to reference the defunct memslots.
65 * This mechanism is used to prevent getting hits in KVM's caches while a
66 * memslot update is in-progress, and to prevent cache hits *after* updating
67 * the actual generation number against accesses that were inserted into the
68 * cache *before* the memslots were updated.
70 #define KVM_MEMSLOT_GEN_UPDATE_IN_PROGRESS BIT_ULL(63)
72 /* Two fragments for cross MMIO pages. */
73 #define KVM_MAX_MMIO_FRAGMENTS 2
75 #ifndef KVM_ADDRESS_SPACE_NUM
76 #define KVM_ADDRESS_SPACE_NUM 1
80 * For the normal pfn, the highest 12 bits should be zero,
81 * so we can mask bit 62 ~ bit 52 to indicate the error pfn,
82 * mask bit 63 to indicate the noslot pfn.
84 #define KVM_PFN_ERR_MASK (0x7ffULL << 52)
85 #define KVM_PFN_ERR_NOSLOT_MASK (0xfffULL << 52)
86 #define KVM_PFN_NOSLOT (0x1ULL << 63)
88 #define KVM_PFN_ERR_FAULT (KVM_PFN_ERR_MASK)
89 #define KVM_PFN_ERR_HWPOISON (KVM_PFN_ERR_MASK + 1)
90 #define KVM_PFN_ERR_RO_FAULT (KVM_PFN_ERR_MASK + 2)
93 * error pfns indicate that the gfn is in slot but faild to
94 * translate it to pfn on host.
96 static inline bool is_error_pfn(kvm_pfn_t pfn)
98 return !!(pfn & KVM_PFN_ERR_MASK);
102 * error_noslot pfns indicate that the gfn can not be
103 * translated to pfn - it is not in slot or failed to
104 * translate it to pfn.
106 static inline bool is_error_noslot_pfn(kvm_pfn_t pfn)
108 return !!(pfn & KVM_PFN_ERR_NOSLOT_MASK);
111 /* noslot pfn indicates that the gfn is not in slot. */
112 static inline bool is_noslot_pfn(kvm_pfn_t pfn)
114 return pfn == KVM_PFN_NOSLOT;
118 * architectures with KVM_HVA_ERR_BAD other than PAGE_OFFSET (e.g. s390)
119 * provide own defines and kvm_is_error_hva
121 #ifndef KVM_HVA_ERR_BAD
123 #define KVM_HVA_ERR_BAD (PAGE_OFFSET)
124 #define KVM_HVA_ERR_RO_BAD (PAGE_OFFSET + PAGE_SIZE)
126 static inline bool kvm_is_error_hva(unsigned long addr)
128 return addr >= PAGE_OFFSET;
133 #define KVM_ERR_PTR_BAD_PAGE (ERR_PTR(-ENOENT))
135 static inline bool is_error_page(struct page *page)
140 #define KVM_REQUEST_MASK GENMASK(7,0)
141 #define KVM_REQUEST_NO_WAKEUP BIT(8)
142 #define KVM_REQUEST_WAIT BIT(9)
144 * Architecture-independent vcpu->requests bit members
145 * Bits 4-7 are reserved for more arch-independent bits.
147 #define KVM_REQ_TLB_FLUSH (0 | KVM_REQUEST_WAIT | KVM_REQUEST_NO_WAKEUP)
148 #define KVM_REQ_MMU_RELOAD (1 | KVM_REQUEST_WAIT | KVM_REQUEST_NO_WAKEUP)
149 #define KVM_REQ_PENDING_TIMER 2
150 #define KVM_REQ_UNHALT 3
151 #define KVM_REQUEST_ARCH_BASE 8
153 #define KVM_ARCH_REQ_FLAGS(nr, flags) ({ \
154 BUILD_BUG_ON((unsigned)(nr) >= (sizeof_field(struct kvm_vcpu, requests) * 8) - KVM_REQUEST_ARCH_BASE); \
155 (unsigned)(((nr) + KVM_REQUEST_ARCH_BASE) | (flags)); \
157 #define KVM_ARCH_REQ(nr) KVM_ARCH_REQ_FLAGS(nr, 0)
159 #define KVM_USERSPACE_IRQ_SOURCE_ID 0
160 #define KVM_IRQFD_RESAMPLE_IRQ_SOURCE_ID 1
162 extern struct mutex kvm_lock;
163 extern struct list_head vm_list;
165 struct kvm_io_range {
168 struct kvm_io_device *dev;
171 #define NR_IOBUS_DEVS 1000
176 struct kvm_io_range range[];
182 KVM_VIRTIO_CCW_NOTIFY_BUS,
187 int kvm_io_bus_write(struct kvm_vcpu *vcpu, enum kvm_bus bus_idx, gpa_t addr,
188 int len, const void *val);
189 int kvm_io_bus_write_cookie(struct kvm_vcpu *vcpu, enum kvm_bus bus_idx,
190 gpa_t addr, int len, const void *val, long cookie);
191 int kvm_io_bus_read(struct kvm_vcpu *vcpu, enum kvm_bus bus_idx, gpa_t addr,
193 int kvm_io_bus_register_dev(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr,
194 int len, struct kvm_io_device *dev);
195 void kvm_io_bus_unregister_dev(struct kvm *kvm, enum kvm_bus bus_idx,
196 struct kvm_io_device *dev);
197 struct kvm_io_device *kvm_io_bus_get_dev(struct kvm *kvm, enum kvm_bus bus_idx,
200 #ifdef CONFIG_KVM_ASYNC_PF
201 struct kvm_async_pf {
202 struct work_struct work;
203 struct list_head link;
204 struct list_head queue;
205 struct kvm_vcpu *vcpu;
206 struct mm_struct *mm;
209 struct kvm_arch_async_pf arch;
211 bool notpresent_injected;
214 void kvm_clear_async_pf_completion_queue(struct kvm_vcpu *vcpu);
215 void kvm_check_async_pf_completion(struct kvm_vcpu *vcpu);
216 bool kvm_setup_async_pf(struct kvm_vcpu *vcpu, gpa_t cr2_or_gpa,
217 unsigned long hva, struct kvm_arch_async_pf *arch);
218 int kvm_async_pf_wakeup_all(struct kvm_vcpu *vcpu);
225 READING_SHADOW_PAGE_TABLES,
228 #define KVM_UNMAPPED_PAGE ((void *) 0x500 + POISON_POINTER_DELTA)
230 struct kvm_host_map {
232 * Only valid if the 'pfn' is managed by the host kernel (i.e. There is
233 * a 'struct page' for it. When using mem= kernel parameter some memory
234 * can be used as guest memory but they are not managed by host
236 * If 'pfn' is not managed by the host kernel, this field is
237 * initialized to KVM_UNMAPPED_PAGE.
246 * Used to check if the mapping is valid or not. Never use 'kvm_host_map'
247 * directly to check for that.
249 static inline bool kvm_vcpu_mapped(struct kvm_host_map *map)
255 * Sometimes a large or cross-page mmio needs to be broken up into separate
256 * exits for userspace servicing.
258 struct kvm_mmio_fragment {
266 #ifdef CONFIG_PREEMPT_NOTIFIERS
267 struct preempt_notifier preempt_notifier;
270 int vcpu_id; /* id given by userspace at creation */
271 int vcpu_idx; /* index in kvm->vcpus array */
275 unsigned long guest_debug;
278 struct list_head blocked_vcpu_list;
284 struct pid __rcu *pid;
287 struct kvm_vcpu_stat stat;
288 unsigned int halt_poll_ns;
291 #ifdef CONFIG_HAS_IOMEM
293 int mmio_read_completed;
295 int mmio_cur_fragment;
296 int mmio_nr_fragments;
297 struct kvm_mmio_fragment mmio_fragments[KVM_MAX_MMIO_FRAGMENTS];
300 #ifdef CONFIG_KVM_ASYNC_PF
303 struct list_head queue;
304 struct list_head done;
309 #ifdef CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT
311 * Cpu relax intercept or pause loop exit optimization
312 * in_spin_loop: set when a vcpu does a pause loop exit
313 * or cpu relax intercepted.
314 * dy_eligible: indicates whether vcpu is eligible for directed yield.
323 struct kvm_vcpu_arch arch;
324 struct kvm_dirty_ring dirty_ring;
327 static inline int kvm_vcpu_exiting_guest_mode(struct kvm_vcpu *vcpu)
330 * The memory barrier ensures a previous write to vcpu->requests cannot
331 * be reordered with the read of vcpu->mode. It pairs with the general
332 * memory barrier following the write of vcpu->mode in VCPU RUN.
334 smp_mb__before_atomic();
335 return cmpxchg(&vcpu->mode, IN_GUEST_MODE, EXITING_GUEST_MODE);
339 * Some of the bitops functions do not support too long bitmaps.
340 * This number must be determined not to exceed such limits.
342 #define KVM_MEM_MAX_NR_PAGES ((1UL << 31) - 1)
344 struct kvm_memory_slot {
346 unsigned long npages;
347 unsigned long *dirty_bitmap;
348 struct kvm_arch_memory_slot arch;
349 unsigned long userspace_addr;
355 static inline bool kvm_slot_dirty_track_enabled(struct kvm_memory_slot *slot)
357 return slot->flags & KVM_MEM_LOG_DIRTY_PAGES;
360 static inline unsigned long kvm_dirty_bitmap_bytes(struct kvm_memory_slot *memslot)
362 return ALIGN(memslot->npages, BITS_PER_LONG) / 8;
365 static inline unsigned long *kvm_second_dirty_bitmap(struct kvm_memory_slot *memslot)
367 unsigned long len = kvm_dirty_bitmap_bytes(memslot);
369 return memslot->dirty_bitmap + len / sizeof(*memslot->dirty_bitmap);
372 #ifndef KVM_DIRTY_LOG_MANUAL_CAPS
373 #define KVM_DIRTY_LOG_MANUAL_CAPS KVM_DIRTY_LOG_MANUAL_PROTECT_ENABLE
376 struct kvm_s390_adapter_int {
389 struct kvm_kernel_irq_routing_entry {
392 int (*set)(struct kvm_kernel_irq_routing_entry *e,
393 struct kvm *kvm, int irq_source_id, int level,
407 struct kvm_s390_adapter_int adapter;
408 struct kvm_hv_sint hv_sint;
410 struct hlist_node link;
413 #ifdef CONFIG_HAVE_KVM_IRQ_ROUTING
414 struct kvm_irq_routing_table {
415 int chip[KVM_NR_IRQCHIPS][KVM_IRQCHIP_NUM_PINS];
418 * Array indexed by gsi. Each entry contains list of irq chips
419 * the gsi is connected to.
421 struct hlist_head map[];
425 #ifndef KVM_PRIVATE_MEM_SLOTS
426 #define KVM_PRIVATE_MEM_SLOTS 0
429 #define KVM_MEM_SLOTS_NUM SHRT_MAX
430 #define KVM_USER_MEM_SLOTS (KVM_MEM_SLOTS_NUM - KVM_PRIVATE_MEM_SLOTS)
432 #ifndef __KVM_VCPU_MULTIPLE_ADDRESS_SPACE
433 static inline int kvm_arch_vcpu_memslots_id(struct kvm_vcpu *vcpu)
441 * memslots are not sorted by id anymore, please use id_to_memslot()
442 * to get the memslot by its id.
444 struct kvm_memslots {
446 /* The mapping table from slot id to the index in memslots[]. */
447 short id_to_index[KVM_MEM_SLOTS_NUM];
450 struct kvm_memory_slot memslots[];
454 #ifdef KVM_HAVE_MMU_RWLOCK
458 #endif /* KVM_HAVE_MMU_RWLOCK */
460 struct mutex slots_lock;
461 struct mm_struct *mm; /* userspace tied to this vm */
462 struct kvm_memslots __rcu *memslots[KVM_ADDRESS_SPACE_NUM];
463 struct kvm_vcpu *vcpus[KVM_MAX_VCPUS];
466 * created_vcpus is protected by kvm->lock, and is incremented
467 * at the beginning of KVM_CREATE_VCPU. online_vcpus is only
468 * incremented after storing the kvm_vcpu pointer in vcpus,
469 * and is accessed atomically.
471 atomic_t online_vcpus;
473 int last_boosted_vcpu;
474 struct list_head vm_list;
476 struct kvm_io_bus __rcu *buses[KVM_NR_BUSES];
477 #ifdef CONFIG_HAVE_KVM_EVENTFD
480 struct list_head items;
481 struct list_head resampler_list;
482 struct mutex resampler_lock;
484 struct list_head ioeventfds;
486 struct kvm_vm_stat stat;
487 struct kvm_arch arch;
488 refcount_t users_count;
489 #ifdef CONFIG_KVM_MMIO
490 struct kvm_coalesced_mmio_ring *coalesced_mmio_ring;
491 spinlock_t ring_lock;
492 struct list_head coalesced_zones;
495 struct mutex irq_lock;
496 #ifdef CONFIG_HAVE_KVM_IRQCHIP
498 * Update side is protected by irq_lock.
500 struct kvm_irq_routing_table __rcu *irq_routing;
502 #ifdef CONFIG_HAVE_KVM_IRQFD
503 struct hlist_head irq_ack_notifier_list;
506 #if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
507 struct mmu_notifier mmu_notifier;
508 unsigned long mmu_notifier_seq;
509 long mmu_notifier_count;
510 unsigned long mmu_notifier_range_start;
511 unsigned long mmu_notifier_range_end;
514 struct list_head devices;
515 u64 manual_dirty_log_protect;
516 struct dentry *debugfs_dentry;
517 struct kvm_stat_data **debugfs_stat_data;
518 struct srcu_struct srcu;
519 struct srcu_struct irq_srcu;
521 unsigned int max_halt_poll_ns;
525 #define kvm_err(fmt, ...) \
526 pr_err("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
527 #define kvm_info(fmt, ...) \
528 pr_info("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
529 #define kvm_debug(fmt, ...) \
530 pr_debug("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
531 #define kvm_debug_ratelimited(fmt, ...) \
532 pr_debug_ratelimited("kvm [%i]: " fmt, task_pid_nr(current), \
534 #define kvm_pr_unimpl(fmt, ...) \
535 pr_err_ratelimited("kvm [%i]: " fmt, \
536 task_tgid_nr(current), ## __VA_ARGS__)
538 /* The guest did something we don't support. */
539 #define vcpu_unimpl(vcpu, fmt, ...) \
540 kvm_pr_unimpl("vcpu%i, guest rIP: 0x%lx " fmt, \
541 (vcpu)->vcpu_id, kvm_rip_read(vcpu), ## __VA_ARGS__)
543 #define vcpu_debug(vcpu, fmt, ...) \
544 kvm_debug("vcpu%i " fmt, (vcpu)->vcpu_id, ## __VA_ARGS__)
545 #define vcpu_debug_ratelimited(vcpu, fmt, ...) \
546 kvm_debug_ratelimited("vcpu%i " fmt, (vcpu)->vcpu_id, \
548 #define vcpu_err(vcpu, fmt, ...) \
549 kvm_err("vcpu%i " fmt, (vcpu)->vcpu_id, ## __VA_ARGS__)
551 static inline bool kvm_dirty_log_manual_protect_and_init_set(struct kvm *kvm)
553 return !!(kvm->manual_dirty_log_protect & KVM_DIRTY_LOG_INITIALLY_SET);
556 static inline struct kvm_io_bus *kvm_get_bus(struct kvm *kvm, enum kvm_bus idx)
558 return srcu_dereference_check(kvm->buses[idx], &kvm->srcu,
559 lockdep_is_held(&kvm->slots_lock) ||
560 !refcount_read(&kvm->users_count));
563 static inline struct kvm_vcpu *kvm_get_vcpu(struct kvm *kvm, int i)
565 int num_vcpus = atomic_read(&kvm->online_vcpus);
566 i = array_index_nospec(i, num_vcpus);
568 /* Pairs with smp_wmb() in kvm_vm_ioctl_create_vcpu. */
570 return kvm->vcpus[i];
573 #define kvm_for_each_vcpu(idx, vcpup, kvm) \
575 idx < atomic_read(&kvm->online_vcpus) && \
576 (vcpup = kvm_get_vcpu(kvm, idx)) != NULL; \
579 static inline struct kvm_vcpu *kvm_get_vcpu_by_id(struct kvm *kvm, int id)
581 struct kvm_vcpu *vcpu = NULL;
586 if (id < KVM_MAX_VCPUS)
587 vcpu = kvm_get_vcpu(kvm, id);
588 if (vcpu && vcpu->vcpu_id == id)
590 kvm_for_each_vcpu(i, vcpu, kvm)
591 if (vcpu->vcpu_id == id)
596 static inline int kvm_vcpu_get_idx(struct kvm_vcpu *vcpu)
598 return vcpu->vcpu_idx;
601 #define kvm_for_each_memslot(memslot, slots) \
602 for (memslot = &slots->memslots[0]; \
603 memslot < slots->memslots + slots->used_slots; memslot++) \
604 if (WARN_ON_ONCE(!memslot->npages)) { \
607 void kvm_vcpu_destroy(struct kvm_vcpu *vcpu);
609 void vcpu_load(struct kvm_vcpu *vcpu);
610 void vcpu_put(struct kvm_vcpu *vcpu);
612 #ifdef __KVM_HAVE_IOAPIC
613 void kvm_arch_post_irq_ack_notifier_list_update(struct kvm *kvm);
614 void kvm_arch_post_irq_routing_update(struct kvm *kvm);
616 static inline void kvm_arch_post_irq_ack_notifier_list_update(struct kvm *kvm)
619 static inline void kvm_arch_post_irq_routing_update(struct kvm *kvm)
624 #ifdef CONFIG_HAVE_KVM_IRQFD
625 int kvm_irqfd_init(void);
626 void kvm_irqfd_exit(void);
628 static inline int kvm_irqfd_init(void)
633 static inline void kvm_irqfd_exit(void)
637 int kvm_init(void *opaque, unsigned vcpu_size, unsigned vcpu_align,
638 struct module *module);
641 void kvm_get_kvm(struct kvm *kvm);
642 void kvm_put_kvm(struct kvm *kvm);
643 void kvm_put_kvm_no_destroy(struct kvm *kvm);
645 static inline struct kvm_memslots *__kvm_memslots(struct kvm *kvm, int as_id)
647 as_id = array_index_nospec(as_id, KVM_ADDRESS_SPACE_NUM);
648 return srcu_dereference_check(kvm->memslots[as_id], &kvm->srcu,
649 lockdep_is_held(&kvm->slots_lock) ||
650 !refcount_read(&kvm->users_count));
653 static inline struct kvm_memslots *kvm_memslots(struct kvm *kvm)
655 return __kvm_memslots(kvm, 0);
658 static inline struct kvm_memslots *kvm_vcpu_memslots(struct kvm_vcpu *vcpu)
660 int as_id = kvm_arch_vcpu_memslots_id(vcpu);
662 return __kvm_memslots(vcpu->kvm, as_id);
666 struct kvm_memory_slot *id_to_memslot(struct kvm_memslots *slots, int id)
668 int index = slots->id_to_index[id];
669 struct kvm_memory_slot *slot;
674 slot = &slots->memslots[index];
676 WARN_ON(slot->id != id);
681 * KVM_SET_USER_MEMORY_REGION ioctl allows the following operations:
682 * - create a new memory slot
683 * - delete an existing memory slot
684 * - modify an existing memory slot
685 * -- move it in the guest physical memory space
686 * -- just change its flags
688 * Since flags can be changed by some of these operations, the following
689 * differentiation is the best we can do for __kvm_set_memory_region():
698 int kvm_set_memory_region(struct kvm *kvm,
699 const struct kvm_userspace_memory_region *mem);
700 int __kvm_set_memory_region(struct kvm *kvm,
701 const struct kvm_userspace_memory_region *mem);
702 void kvm_arch_free_memslot(struct kvm *kvm, struct kvm_memory_slot *slot);
703 void kvm_arch_memslots_updated(struct kvm *kvm, u64 gen);
704 int kvm_arch_prepare_memory_region(struct kvm *kvm,
705 struct kvm_memory_slot *memslot,
706 const struct kvm_userspace_memory_region *mem,
707 enum kvm_mr_change change);
708 void kvm_arch_commit_memory_region(struct kvm *kvm,
709 const struct kvm_userspace_memory_region *mem,
710 struct kvm_memory_slot *old,
711 const struct kvm_memory_slot *new,
712 enum kvm_mr_change change);
713 /* flush all memory translations */
714 void kvm_arch_flush_shadow_all(struct kvm *kvm);
715 /* flush memory translations pointing to 'slot' */
716 void kvm_arch_flush_shadow_memslot(struct kvm *kvm,
717 struct kvm_memory_slot *slot);
719 int gfn_to_page_many_atomic(struct kvm_memory_slot *slot, gfn_t gfn,
720 struct page **pages, int nr_pages);
722 struct page *gfn_to_page(struct kvm *kvm, gfn_t gfn);
723 unsigned long gfn_to_hva(struct kvm *kvm, gfn_t gfn);
724 unsigned long gfn_to_hva_prot(struct kvm *kvm, gfn_t gfn, bool *writable);
725 unsigned long gfn_to_hva_memslot(struct kvm_memory_slot *slot, gfn_t gfn);
726 unsigned long gfn_to_hva_memslot_prot(struct kvm_memory_slot *slot, gfn_t gfn,
728 void kvm_release_page_clean(struct page *page);
729 void kvm_release_page_dirty(struct page *page);
730 void kvm_set_page_accessed(struct page *page);
732 kvm_pfn_t gfn_to_pfn(struct kvm *kvm, gfn_t gfn);
733 kvm_pfn_t gfn_to_pfn_prot(struct kvm *kvm, gfn_t gfn, bool write_fault,
735 kvm_pfn_t gfn_to_pfn_memslot(struct kvm_memory_slot *slot, gfn_t gfn);
736 kvm_pfn_t gfn_to_pfn_memslot_atomic(struct kvm_memory_slot *slot, gfn_t gfn);
737 kvm_pfn_t __gfn_to_pfn_memslot(struct kvm_memory_slot *slot, gfn_t gfn,
738 bool atomic, bool *async, bool write_fault,
739 bool *writable, hva_t *hva);
741 void kvm_release_pfn_clean(kvm_pfn_t pfn);
742 void kvm_release_pfn_dirty(kvm_pfn_t pfn);
743 void kvm_set_pfn_dirty(kvm_pfn_t pfn);
744 void kvm_set_pfn_accessed(kvm_pfn_t pfn);
745 void kvm_get_pfn(kvm_pfn_t pfn);
747 void kvm_release_pfn(kvm_pfn_t pfn, bool dirty, struct gfn_to_pfn_cache *cache);
748 int kvm_read_guest_page(struct kvm *kvm, gfn_t gfn, void *data, int offset,
750 int kvm_read_guest(struct kvm *kvm, gpa_t gpa, void *data, unsigned long len);
751 int kvm_read_guest_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
752 void *data, unsigned long len);
753 int kvm_read_guest_offset_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
754 void *data, unsigned int offset,
756 int kvm_write_guest_page(struct kvm *kvm, gfn_t gfn, const void *data,
757 int offset, int len);
758 int kvm_write_guest(struct kvm *kvm, gpa_t gpa, const void *data,
760 int kvm_write_guest_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
761 void *data, unsigned long len);
762 int kvm_write_guest_offset_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
763 void *data, unsigned int offset,
765 int kvm_gfn_to_hva_cache_init(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
766 gpa_t gpa, unsigned long len);
768 #define __kvm_get_guest(kvm, gfn, offset, v) \
770 unsigned long __addr = gfn_to_hva(kvm, gfn); \
771 typeof(v) __user *__uaddr = (typeof(__uaddr))(__addr + offset); \
772 int __ret = -EFAULT; \
774 if (!kvm_is_error_hva(__addr)) \
775 __ret = get_user(v, __uaddr); \
779 #define kvm_get_guest(kvm, gpa, v) \
782 struct kvm *__kvm = kvm; \
784 __kvm_get_guest(__kvm, __gpa >> PAGE_SHIFT, \
785 offset_in_page(__gpa), v); \
788 #define __kvm_put_guest(kvm, gfn, offset, v) \
790 unsigned long __addr = gfn_to_hva(kvm, gfn); \
791 typeof(v) __user *__uaddr = (typeof(__uaddr))(__addr + offset); \
792 int __ret = -EFAULT; \
794 if (!kvm_is_error_hva(__addr)) \
795 __ret = put_user(v, __uaddr); \
797 mark_page_dirty(kvm, gfn); \
801 #define kvm_put_guest(kvm, gpa, v) \
804 struct kvm *__kvm = kvm; \
806 __kvm_put_guest(__kvm, __gpa >> PAGE_SHIFT, \
807 offset_in_page(__gpa), v); \
810 int kvm_clear_guest(struct kvm *kvm, gpa_t gpa, unsigned long len);
811 struct kvm_memory_slot *gfn_to_memslot(struct kvm *kvm, gfn_t gfn);
812 bool kvm_is_visible_gfn(struct kvm *kvm, gfn_t gfn);
813 bool kvm_vcpu_is_visible_gfn(struct kvm_vcpu *vcpu, gfn_t gfn);
814 unsigned long kvm_host_page_size(struct kvm_vcpu *vcpu, gfn_t gfn);
815 void mark_page_dirty_in_slot(struct kvm *kvm, struct kvm_memory_slot *memslot, gfn_t gfn);
816 void mark_page_dirty(struct kvm *kvm, gfn_t gfn);
818 struct kvm_memslots *kvm_vcpu_memslots(struct kvm_vcpu *vcpu);
819 struct kvm_memory_slot *kvm_vcpu_gfn_to_memslot(struct kvm_vcpu *vcpu, gfn_t gfn);
820 kvm_pfn_t kvm_vcpu_gfn_to_pfn_atomic(struct kvm_vcpu *vcpu, gfn_t gfn);
821 kvm_pfn_t kvm_vcpu_gfn_to_pfn(struct kvm_vcpu *vcpu, gfn_t gfn);
822 int kvm_vcpu_map(struct kvm_vcpu *vcpu, gpa_t gpa, struct kvm_host_map *map);
823 int kvm_map_gfn(struct kvm_vcpu *vcpu, gfn_t gfn, struct kvm_host_map *map,
824 struct gfn_to_pfn_cache *cache, bool atomic);
825 struct page *kvm_vcpu_gfn_to_page(struct kvm_vcpu *vcpu, gfn_t gfn);
826 void kvm_vcpu_unmap(struct kvm_vcpu *vcpu, struct kvm_host_map *map, bool dirty);
827 int kvm_unmap_gfn(struct kvm_vcpu *vcpu, struct kvm_host_map *map,
828 struct gfn_to_pfn_cache *cache, bool dirty, bool atomic);
829 unsigned long kvm_vcpu_gfn_to_hva(struct kvm_vcpu *vcpu, gfn_t gfn);
830 unsigned long kvm_vcpu_gfn_to_hva_prot(struct kvm_vcpu *vcpu, gfn_t gfn, bool *writable);
831 int kvm_vcpu_read_guest_page(struct kvm_vcpu *vcpu, gfn_t gfn, void *data, int offset,
833 int kvm_vcpu_read_guest_atomic(struct kvm_vcpu *vcpu, gpa_t gpa, void *data,
835 int kvm_vcpu_read_guest(struct kvm_vcpu *vcpu, gpa_t gpa, void *data,
837 int kvm_vcpu_write_guest_page(struct kvm_vcpu *vcpu, gfn_t gfn, const void *data,
838 int offset, int len);
839 int kvm_vcpu_write_guest(struct kvm_vcpu *vcpu, gpa_t gpa, const void *data,
841 void kvm_vcpu_mark_page_dirty(struct kvm_vcpu *vcpu, gfn_t gfn);
843 void kvm_sigset_activate(struct kvm_vcpu *vcpu);
844 void kvm_sigset_deactivate(struct kvm_vcpu *vcpu);
846 void kvm_vcpu_block(struct kvm_vcpu *vcpu);
847 void kvm_arch_vcpu_blocking(struct kvm_vcpu *vcpu);
848 void kvm_arch_vcpu_unblocking(struct kvm_vcpu *vcpu);
849 bool kvm_vcpu_wake_up(struct kvm_vcpu *vcpu);
850 void kvm_vcpu_kick(struct kvm_vcpu *vcpu);
851 int kvm_vcpu_yield_to(struct kvm_vcpu *target);
852 void kvm_vcpu_on_spin(struct kvm_vcpu *vcpu, bool usermode_vcpu_not_eligible);
854 void kvm_flush_remote_tlbs(struct kvm *kvm);
855 void kvm_reload_remote_mmus(struct kvm *kvm);
857 #ifdef KVM_ARCH_NR_OBJS_PER_MEMORY_CACHE
858 int kvm_mmu_topup_memory_cache(struct kvm_mmu_memory_cache *mc, int min);
859 int kvm_mmu_memory_cache_nr_free_objects(struct kvm_mmu_memory_cache *mc);
860 void kvm_mmu_free_memory_cache(struct kvm_mmu_memory_cache *mc);
861 void *kvm_mmu_memory_cache_alloc(struct kvm_mmu_memory_cache *mc);
864 bool kvm_make_vcpus_request_mask(struct kvm *kvm, unsigned int req,
865 struct kvm_vcpu *except,
866 unsigned long *vcpu_bitmap, cpumask_var_t tmp);
867 bool kvm_make_all_cpus_request(struct kvm *kvm, unsigned int req);
868 bool kvm_make_all_cpus_request_except(struct kvm *kvm, unsigned int req,
869 struct kvm_vcpu *except);
870 bool kvm_make_cpus_request_mask(struct kvm *kvm, unsigned int req,
871 unsigned long *vcpu_bitmap);
873 long kvm_arch_dev_ioctl(struct file *filp,
874 unsigned int ioctl, unsigned long arg);
875 long kvm_arch_vcpu_ioctl(struct file *filp,
876 unsigned int ioctl, unsigned long arg);
877 vm_fault_t kvm_arch_vcpu_fault(struct kvm_vcpu *vcpu, struct vm_fault *vmf);
879 int kvm_vm_ioctl_check_extension(struct kvm *kvm, long ext);
881 void kvm_arch_mmu_enable_log_dirty_pt_masked(struct kvm *kvm,
882 struct kvm_memory_slot *slot,
885 void kvm_arch_sync_dirty_log(struct kvm *kvm, struct kvm_memory_slot *memslot);
887 #ifdef CONFIG_KVM_GENERIC_DIRTYLOG_READ_PROTECT
888 void kvm_arch_flush_remote_tlbs_memslot(struct kvm *kvm,
889 struct kvm_memory_slot *memslot);
890 #else /* !CONFIG_KVM_GENERIC_DIRTYLOG_READ_PROTECT */
891 int kvm_vm_ioctl_get_dirty_log(struct kvm *kvm, struct kvm_dirty_log *log);
892 int kvm_get_dirty_log(struct kvm *kvm, struct kvm_dirty_log *log,
893 int *is_dirty, struct kvm_memory_slot **memslot);
896 int kvm_vm_ioctl_irq_line(struct kvm *kvm, struct kvm_irq_level *irq_level,
898 int kvm_vm_ioctl_enable_cap(struct kvm *kvm,
899 struct kvm_enable_cap *cap);
900 long kvm_arch_vm_ioctl(struct file *filp,
901 unsigned int ioctl, unsigned long arg);
903 int kvm_arch_vcpu_ioctl_get_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu);
904 int kvm_arch_vcpu_ioctl_set_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu);
906 int kvm_arch_vcpu_ioctl_translate(struct kvm_vcpu *vcpu,
907 struct kvm_translation *tr);
909 int kvm_arch_vcpu_ioctl_get_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs);
910 int kvm_arch_vcpu_ioctl_set_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs);
911 int kvm_arch_vcpu_ioctl_get_sregs(struct kvm_vcpu *vcpu,
912 struct kvm_sregs *sregs);
913 int kvm_arch_vcpu_ioctl_set_sregs(struct kvm_vcpu *vcpu,
914 struct kvm_sregs *sregs);
915 int kvm_arch_vcpu_ioctl_get_mpstate(struct kvm_vcpu *vcpu,
916 struct kvm_mp_state *mp_state);
917 int kvm_arch_vcpu_ioctl_set_mpstate(struct kvm_vcpu *vcpu,
918 struct kvm_mp_state *mp_state);
919 int kvm_arch_vcpu_ioctl_set_guest_debug(struct kvm_vcpu *vcpu,
920 struct kvm_guest_debug *dbg);
921 int kvm_arch_vcpu_ioctl_run(struct kvm_vcpu *vcpu);
923 int kvm_arch_init(void *opaque);
924 void kvm_arch_exit(void);
926 void kvm_arch_sched_in(struct kvm_vcpu *vcpu, int cpu);
928 void kvm_arch_vcpu_load(struct kvm_vcpu *vcpu, int cpu);
929 void kvm_arch_vcpu_put(struct kvm_vcpu *vcpu);
930 int kvm_arch_vcpu_precreate(struct kvm *kvm, unsigned int id);
931 int kvm_arch_vcpu_create(struct kvm_vcpu *vcpu);
932 void kvm_arch_vcpu_postcreate(struct kvm_vcpu *vcpu);
933 void kvm_arch_vcpu_destroy(struct kvm_vcpu *vcpu);
935 #ifdef __KVM_HAVE_ARCH_VCPU_DEBUGFS
936 void kvm_arch_create_vcpu_debugfs(struct kvm_vcpu *vcpu, struct dentry *debugfs_dentry);
939 int kvm_arch_hardware_enable(void);
940 void kvm_arch_hardware_disable(void);
941 int kvm_arch_hardware_setup(void *opaque);
942 void kvm_arch_hardware_unsetup(void);
943 int kvm_arch_check_processor_compat(void *opaque);
944 int kvm_arch_vcpu_runnable(struct kvm_vcpu *vcpu);
945 bool kvm_arch_vcpu_in_kernel(struct kvm_vcpu *vcpu);
946 int kvm_arch_vcpu_should_kick(struct kvm_vcpu *vcpu);
947 bool kvm_arch_dy_runnable(struct kvm_vcpu *vcpu);
948 int kvm_arch_post_init_vm(struct kvm *kvm);
949 void kvm_arch_pre_destroy_vm(struct kvm *kvm);
951 #ifndef __KVM_HAVE_ARCH_VM_ALLOC
953 * All architectures that want to use vzalloc currently also
954 * need their own kvm_arch_alloc_vm implementation.
956 static inline struct kvm *kvm_arch_alloc_vm(void)
958 return kzalloc(sizeof(struct kvm), GFP_KERNEL);
961 static inline void kvm_arch_free_vm(struct kvm *kvm)
967 #ifndef __KVM_HAVE_ARCH_FLUSH_REMOTE_TLB
968 static inline int kvm_arch_flush_remote_tlb(struct kvm *kvm)
974 #ifdef __KVM_HAVE_ARCH_NONCOHERENT_DMA
975 void kvm_arch_register_noncoherent_dma(struct kvm *kvm);
976 void kvm_arch_unregister_noncoherent_dma(struct kvm *kvm);
977 bool kvm_arch_has_noncoherent_dma(struct kvm *kvm);
979 static inline void kvm_arch_register_noncoherent_dma(struct kvm *kvm)
983 static inline void kvm_arch_unregister_noncoherent_dma(struct kvm *kvm)
987 static inline bool kvm_arch_has_noncoherent_dma(struct kvm *kvm)
992 #ifdef __KVM_HAVE_ARCH_ASSIGNED_DEVICE
993 void kvm_arch_start_assignment(struct kvm *kvm);
994 void kvm_arch_end_assignment(struct kvm *kvm);
995 bool kvm_arch_has_assigned_device(struct kvm *kvm);
997 static inline void kvm_arch_start_assignment(struct kvm *kvm)
1001 static inline void kvm_arch_end_assignment(struct kvm *kvm)
1005 static inline bool kvm_arch_has_assigned_device(struct kvm *kvm)
1011 static inline struct rcuwait *kvm_arch_vcpu_get_wait(struct kvm_vcpu *vcpu)
1013 #ifdef __KVM_HAVE_ARCH_WQP
1014 return vcpu->arch.waitp;
1020 #ifdef __KVM_HAVE_ARCH_INTC_INITIALIZED
1022 * returns true if the virtual interrupt controller is initialized and
1023 * ready to accept virtual IRQ. On some architectures the virtual interrupt
1024 * controller is dynamically instantiated and this is not always true.
1026 bool kvm_arch_intc_initialized(struct kvm *kvm);
1028 static inline bool kvm_arch_intc_initialized(struct kvm *kvm)
1034 int kvm_arch_init_vm(struct kvm *kvm, unsigned long type);
1035 void kvm_arch_destroy_vm(struct kvm *kvm);
1036 void kvm_arch_sync_events(struct kvm *kvm);
1038 int kvm_cpu_has_pending_timer(struct kvm_vcpu *vcpu);
1040 bool kvm_is_reserved_pfn(kvm_pfn_t pfn);
1041 bool kvm_is_zone_device_pfn(kvm_pfn_t pfn);
1042 bool kvm_is_transparent_hugepage(kvm_pfn_t pfn);
1044 struct kvm_irq_ack_notifier {
1045 struct hlist_node link;
1047 void (*irq_acked)(struct kvm_irq_ack_notifier *kian);
1050 int kvm_irq_map_gsi(struct kvm *kvm,
1051 struct kvm_kernel_irq_routing_entry *entries, int gsi);
1052 int kvm_irq_map_chip_pin(struct kvm *kvm, unsigned irqchip, unsigned pin);
1054 int kvm_set_irq(struct kvm *kvm, int irq_source_id, u32 irq, int level,
1056 int kvm_set_msi(struct kvm_kernel_irq_routing_entry *irq_entry, struct kvm *kvm,
1057 int irq_source_id, int level, bool line_status);
1058 int kvm_arch_set_irq_inatomic(struct kvm_kernel_irq_routing_entry *e,
1059 struct kvm *kvm, int irq_source_id,
1060 int level, bool line_status);
1061 bool kvm_irq_has_notifier(struct kvm *kvm, unsigned irqchip, unsigned pin);
1062 void kvm_notify_acked_gsi(struct kvm *kvm, int gsi);
1063 void kvm_notify_acked_irq(struct kvm *kvm, unsigned irqchip, unsigned pin);
1064 void kvm_register_irq_ack_notifier(struct kvm *kvm,
1065 struct kvm_irq_ack_notifier *kian);
1066 void kvm_unregister_irq_ack_notifier(struct kvm *kvm,
1067 struct kvm_irq_ack_notifier *kian);
1068 int kvm_request_irq_source_id(struct kvm *kvm);
1069 void kvm_free_irq_source_id(struct kvm *kvm, int irq_source_id);
1070 bool kvm_arch_irqfd_allowed(struct kvm *kvm, struct kvm_irqfd *args);
1073 * search_memslots() and __gfn_to_memslot() are here because they are
1074 * used in non-modular code in arch/powerpc/kvm/book3s_hv_rm_mmu.c.
1075 * gfn_to_memslot() itself isn't here as an inline because that would
1076 * bloat other code too much.
1078 * IMPORTANT: Slots are sorted from highest GFN to lowest GFN!
1080 static inline struct kvm_memory_slot *
1081 search_memslots(struct kvm_memslots *slots, gfn_t gfn)
1083 int start = 0, end = slots->used_slots;
1084 int slot = atomic_read(&slots->lru_slot);
1085 struct kvm_memory_slot *memslots = slots->memslots;
1087 if (unlikely(!slots->used_slots))
1090 if (gfn >= memslots[slot].base_gfn &&
1091 gfn < memslots[slot].base_gfn + memslots[slot].npages)
1092 return &memslots[slot];
1094 while (start < end) {
1095 slot = start + (end - start) / 2;
1097 if (gfn >= memslots[slot].base_gfn)
1103 if (start < slots->used_slots && gfn >= memslots[start].base_gfn &&
1104 gfn < memslots[start].base_gfn + memslots[start].npages) {
1105 atomic_set(&slots->lru_slot, start);
1106 return &memslots[start];
1112 static inline struct kvm_memory_slot *
1113 __gfn_to_memslot(struct kvm_memslots *slots, gfn_t gfn)
1115 return search_memslots(slots, gfn);
1118 static inline unsigned long
1119 __gfn_to_hva_memslot(struct kvm_memory_slot *slot, gfn_t gfn)
1121 return slot->userspace_addr + (gfn - slot->base_gfn) * PAGE_SIZE;
1124 static inline int memslot_id(struct kvm *kvm, gfn_t gfn)
1126 return gfn_to_memslot(kvm, gfn)->id;
1130 hva_to_gfn_memslot(unsigned long hva, struct kvm_memory_slot *slot)
1132 gfn_t gfn_offset = (hva - slot->userspace_addr) >> PAGE_SHIFT;
1134 return slot->base_gfn + gfn_offset;
1137 static inline gpa_t gfn_to_gpa(gfn_t gfn)
1139 return (gpa_t)gfn << PAGE_SHIFT;
1142 static inline gfn_t gpa_to_gfn(gpa_t gpa)
1144 return (gfn_t)(gpa >> PAGE_SHIFT);
1147 static inline hpa_t pfn_to_hpa(kvm_pfn_t pfn)
1149 return (hpa_t)pfn << PAGE_SHIFT;
1152 static inline struct page *kvm_vcpu_gpa_to_page(struct kvm_vcpu *vcpu,
1155 return kvm_vcpu_gfn_to_page(vcpu, gpa_to_gfn(gpa));
1158 static inline bool kvm_is_error_gpa(struct kvm *kvm, gpa_t gpa)
1160 unsigned long hva = gfn_to_hva(kvm, gpa_to_gfn(gpa));
1162 return kvm_is_error_hva(hva);
1165 enum kvm_stat_kind {
1170 struct kvm_stat_data {
1172 struct kvm_stats_debugfs_item *dbgfs_item;
1175 struct kvm_stats_debugfs_item {
1178 enum kvm_stat_kind kind;
1182 #define KVM_DBGFS_GET_MODE(dbgfs_item) \
1183 ((dbgfs_item)->mode ? (dbgfs_item)->mode : 0644)
1185 #define VM_STAT(n, x, ...) \
1186 { n, offsetof(struct kvm, stat.x), KVM_STAT_VM, ## __VA_ARGS__ }
1187 #define VCPU_STAT(n, x, ...) \
1188 { n, offsetof(struct kvm_vcpu, stat.x), KVM_STAT_VCPU, ## __VA_ARGS__ }
1190 extern struct kvm_stats_debugfs_item debugfs_entries[];
1191 extern struct dentry *kvm_debugfs_dir;
1193 #if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
1194 static inline int mmu_notifier_retry(struct kvm *kvm, unsigned long mmu_seq)
1196 if (unlikely(kvm->mmu_notifier_count))
1199 * Ensure the read of mmu_notifier_count happens before the read
1200 * of mmu_notifier_seq. This interacts with the smp_wmb() in
1201 * mmu_notifier_invalidate_range_end to make sure that the caller
1202 * either sees the old (non-zero) value of mmu_notifier_count or
1203 * the new (incremented) value of mmu_notifier_seq.
1204 * PowerPC Book3s HV KVM calls this under a per-page lock
1205 * rather than under kvm->mmu_lock, for scalability, so
1206 * can't rely on kvm->mmu_lock to keep things ordered.
1209 if (kvm->mmu_notifier_seq != mmu_seq)
1214 static inline int mmu_notifier_retry_hva(struct kvm *kvm,
1215 unsigned long mmu_seq,
1218 lockdep_assert_held(&kvm->mmu_lock);
1220 * If mmu_notifier_count is non-zero, then the range maintained by
1221 * kvm_mmu_notifier_invalidate_range_start contains all addresses that
1222 * might be being invalidated. Note that it may include some false
1223 * positives, due to shortcuts when handing concurrent invalidations.
1225 if (unlikely(kvm->mmu_notifier_count) &&
1226 hva >= kvm->mmu_notifier_range_start &&
1227 hva < kvm->mmu_notifier_range_end)
1229 if (kvm->mmu_notifier_seq != mmu_seq)
1235 #ifdef CONFIG_HAVE_KVM_IRQ_ROUTING
1237 #define KVM_MAX_IRQ_ROUTES 4096 /* might need extension/rework in the future */
1239 bool kvm_arch_can_set_irq_routing(struct kvm *kvm);
1240 int kvm_set_irq_routing(struct kvm *kvm,
1241 const struct kvm_irq_routing_entry *entries,
1244 int kvm_set_routing_entry(struct kvm *kvm,
1245 struct kvm_kernel_irq_routing_entry *e,
1246 const struct kvm_irq_routing_entry *ue);
1247 void kvm_free_irq_routing(struct kvm *kvm);
1251 static inline void kvm_free_irq_routing(struct kvm *kvm) {}
1255 int kvm_send_userspace_msi(struct kvm *kvm, struct kvm_msi *msi);
1257 #ifdef CONFIG_HAVE_KVM_EVENTFD
1259 void kvm_eventfd_init(struct kvm *kvm);
1260 int kvm_ioeventfd(struct kvm *kvm, struct kvm_ioeventfd *args);
1262 #ifdef CONFIG_HAVE_KVM_IRQFD
1263 int kvm_irqfd(struct kvm *kvm, struct kvm_irqfd *args);
1264 void kvm_irqfd_release(struct kvm *kvm);
1265 void kvm_irq_routing_update(struct kvm *);
1267 static inline int kvm_irqfd(struct kvm *kvm, struct kvm_irqfd *args)
1272 static inline void kvm_irqfd_release(struct kvm *kvm) {}
1277 static inline void kvm_eventfd_init(struct kvm *kvm) {}
1279 static inline int kvm_irqfd(struct kvm *kvm, struct kvm_irqfd *args)
1284 static inline void kvm_irqfd_release(struct kvm *kvm) {}
1286 #ifdef CONFIG_HAVE_KVM_IRQCHIP
1287 static inline void kvm_irq_routing_update(struct kvm *kvm)
1292 static inline int kvm_ioeventfd(struct kvm *kvm, struct kvm_ioeventfd *args)
1297 #endif /* CONFIG_HAVE_KVM_EVENTFD */
1299 void kvm_arch_irq_routing_update(struct kvm *kvm);
1301 static inline void kvm_make_request(int req, struct kvm_vcpu *vcpu)
1304 * Ensure the rest of the request is published to kvm_check_request's
1305 * caller. Paired with the smp_mb__after_atomic in kvm_check_request.
1308 set_bit(req & KVM_REQUEST_MASK, (void *)&vcpu->requests);
1311 static inline bool kvm_request_pending(struct kvm_vcpu *vcpu)
1313 return READ_ONCE(vcpu->requests);
1316 static inline bool kvm_test_request(int req, struct kvm_vcpu *vcpu)
1318 return test_bit(req & KVM_REQUEST_MASK, (void *)&vcpu->requests);
1321 static inline void kvm_clear_request(int req, struct kvm_vcpu *vcpu)
1323 clear_bit(req & KVM_REQUEST_MASK, (void *)&vcpu->requests);
1326 static inline bool kvm_check_request(int req, struct kvm_vcpu *vcpu)
1328 if (kvm_test_request(req, vcpu)) {
1329 kvm_clear_request(req, vcpu);
1332 * Ensure the rest of the request is visible to kvm_check_request's
1333 * caller. Paired with the smp_wmb in kvm_make_request.
1335 smp_mb__after_atomic();
1342 extern bool kvm_rebooting;
1344 extern unsigned int halt_poll_ns;
1345 extern unsigned int halt_poll_ns_grow;
1346 extern unsigned int halt_poll_ns_grow_start;
1347 extern unsigned int halt_poll_ns_shrink;
1350 const struct kvm_device_ops *ops;
1353 struct list_head vm_node;
1356 /* create, destroy, and name are mandatory */
1357 struct kvm_device_ops {
1361 * create is called holding kvm->lock and any operations not suitable
1362 * to do while holding the lock should be deferred to init (see
1365 int (*create)(struct kvm_device *dev, u32 type);
1368 * init is called after create if create is successful and is called
1369 * outside of holding kvm->lock.
1371 void (*init)(struct kvm_device *dev);
1374 * Destroy is responsible for freeing dev.
1376 * Destroy may be called before or after destructors are called
1377 * on emulated I/O regions, depending on whether a reference is
1378 * held by a vcpu or other kvm component that gets destroyed
1379 * after the emulated I/O.
1381 void (*destroy)(struct kvm_device *dev);
1384 * Release is an alternative method to free the device. It is
1385 * called when the device file descriptor is closed. Once
1386 * release is called, the destroy method will not be called
1387 * anymore as the device is removed from the device list of
1388 * the VM. kvm->lock is held.
1390 void (*release)(struct kvm_device *dev);
1392 int (*set_attr)(struct kvm_device *dev, struct kvm_device_attr *attr);
1393 int (*get_attr)(struct kvm_device *dev, struct kvm_device_attr *attr);
1394 int (*has_attr)(struct kvm_device *dev, struct kvm_device_attr *attr);
1395 long (*ioctl)(struct kvm_device *dev, unsigned int ioctl,
1397 int (*mmap)(struct kvm_device *dev, struct vm_area_struct *vma);
1400 void kvm_device_get(struct kvm_device *dev);
1401 void kvm_device_put(struct kvm_device *dev);
1402 struct kvm_device *kvm_device_from_filp(struct file *filp);
1403 int kvm_register_device_ops(const struct kvm_device_ops *ops, u32 type);
1404 void kvm_unregister_device_ops(u32 type);
1406 extern struct kvm_device_ops kvm_mpic_ops;
1407 extern struct kvm_device_ops kvm_arm_vgic_v2_ops;
1408 extern struct kvm_device_ops kvm_arm_vgic_v3_ops;
1410 #ifdef CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT
1412 static inline void kvm_vcpu_set_in_spin_loop(struct kvm_vcpu *vcpu, bool val)
1414 vcpu->spin_loop.in_spin_loop = val;
1416 static inline void kvm_vcpu_set_dy_eligible(struct kvm_vcpu *vcpu, bool val)
1418 vcpu->spin_loop.dy_eligible = val;
1421 #else /* !CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT */
1423 static inline void kvm_vcpu_set_in_spin_loop(struct kvm_vcpu *vcpu, bool val)
1427 static inline void kvm_vcpu_set_dy_eligible(struct kvm_vcpu *vcpu, bool val)
1430 #endif /* CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT */
1432 static inline bool kvm_is_visible_memslot(struct kvm_memory_slot *memslot)
1434 return (memslot && memslot->id < KVM_USER_MEM_SLOTS &&
1435 !(memslot->flags & KVM_MEMSLOT_INVALID));
1438 struct kvm_vcpu *kvm_get_running_vcpu(void);
1439 struct kvm_vcpu * __percpu *kvm_get_running_vcpus(void);
1441 #ifdef CONFIG_HAVE_KVM_IRQ_BYPASS
1442 bool kvm_arch_has_irq_bypass(void);
1443 int kvm_arch_irq_bypass_add_producer(struct irq_bypass_consumer *,
1444 struct irq_bypass_producer *);
1445 void kvm_arch_irq_bypass_del_producer(struct irq_bypass_consumer *,
1446 struct irq_bypass_producer *);
1447 void kvm_arch_irq_bypass_stop(struct irq_bypass_consumer *);
1448 void kvm_arch_irq_bypass_start(struct irq_bypass_consumer *);
1449 int kvm_arch_update_irqfd_routing(struct kvm *kvm, unsigned int host_irq,
1450 uint32_t guest_irq, bool set);
1451 #endif /* CONFIG_HAVE_KVM_IRQ_BYPASS */
1453 #ifdef CONFIG_HAVE_KVM_INVALID_WAKEUPS
1454 /* If we wakeup during the poll time, was it a sucessful poll? */
1455 static inline bool vcpu_valid_wakeup(struct kvm_vcpu *vcpu)
1457 return vcpu->valid_wakeup;
1461 static inline bool vcpu_valid_wakeup(struct kvm_vcpu *vcpu)
1465 #endif /* CONFIG_HAVE_KVM_INVALID_WAKEUPS */
1467 #ifdef CONFIG_HAVE_KVM_NO_POLL
1468 /* Callback that tells if we must not poll */
1469 bool kvm_arch_no_poll(struct kvm_vcpu *vcpu);
1471 static inline bool kvm_arch_no_poll(struct kvm_vcpu *vcpu)
1475 #endif /* CONFIG_HAVE_KVM_NO_POLL */
1477 #ifdef CONFIG_HAVE_KVM_VCPU_ASYNC_IOCTL
1478 long kvm_arch_vcpu_async_ioctl(struct file *filp,
1479 unsigned int ioctl, unsigned long arg);
1481 static inline long kvm_arch_vcpu_async_ioctl(struct file *filp,
1485 return -ENOIOCTLCMD;
1487 #endif /* CONFIG_HAVE_KVM_VCPU_ASYNC_IOCTL */
1489 void kvm_arch_mmu_notifier_invalidate_range(struct kvm *kvm,
1490 unsigned long start, unsigned long end);
1492 #ifdef CONFIG_HAVE_KVM_VCPU_RUN_PID_CHANGE
1493 int kvm_arch_vcpu_run_pid_change(struct kvm_vcpu *vcpu);
1495 static inline int kvm_arch_vcpu_run_pid_change(struct kvm_vcpu *vcpu)
1499 #endif /* CONFIG_HAVE_KVM_VCPU_RUN_PID_CHANGE */
1501 typedef int (*kvm_vm_thread_fn_t)(struct kvm *kvm, uintptr_t data);
1503 int kvm_vm_create_worker_thread(struct kvm *kvm, kvm_vm_thread_fn_t thread_fn,
1504 uintptr_t data, const char *name,
1505 struct task_struct **thread_ptr);
1507 #ifdef CONFIG_KVM_XFER_TO_GUEST_WORK
1508 static inline void kvm_handle_signal_exit(struct kvm_vcpu *vcpu)
1510 vcpu->run->exit_reason = KVM_EXIT_INTR;
1511 vcpu->stat.signal_exits++;
1513 #endif /* CONFIG_KVM_XFER_TO_GUEST_WORK */
1516 * This defines how many reserved entries we want to keep before we
1517 * kick the vcpu to the userspace to avoid dirty ring full. This
1518 * value can be tuned to higher if e.g. PML is enabled on the host.
1520 #define KVM_DIRTY_RING_RSVD_ENTRIES 64
1522 /* Max number of entries allowed for each kvm dirty ring */
1523 #define KVM_DIRTY_RING_MAX_ENTRIES 65536