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);
221 #ifdef KVM_ARCH_WANT_MMU_NOTIFIER
222 #ifdef KVM_ARCH_WANT_NEW_MMU_NOTIFIER_APIS
223 struct kvm_gfn_range {
224 struct kvm_memory_slot *slot;
230 bool kvm_unmap_gfn_range(struct kvm *kvm, struct kvm_gfn_range *range);
231 bool kvm_age_gfn(struct kvm *kvm, struct kvm_gfn_range *range);
232 bool kvm_test_age_gfn(struct kvm *kvm, struct kvm_gfn_range *range);
233 bool kvm_set_spte_gfn(struct kvm *kvm, struct kvm_gfn_range *range);
235 int kvm_unmap_hva_range(struct kvm *kvm,
236 unsigned long start, unsigned long end, unsigned flags);
237 int kvm_set_spte_hva(struct kvm *kvm, unsigned long hva, pte_t pte);
238 int kvm_age_hva(struct kvm *kvm, unsigned long start, unsigned long end);
239 int kvm_test_age_hva(struct kvm *kvm, unsigned long hva);
240 #endif /* KVM_ARCH_WANT_NEW_MMU_NOTIFIER_APIS */
247 READING_SHADOW_PAGE_TABLES,
250 #define KVM_UNMAPPED_PAGE ((void *) 0x500 + POISON_POINTER_DELTA)
252 struct kvm_host_map {
254 * Only valid if the 'pfn' is managed by the host kernel (i.e. There is
255 * a 'struct page' for it. When using mem= kernel parameter some memory
256 * can be used as guest memory but they are not managed by host
258 * If 'pfn' is not managed by the host kernel, this field is
259 * initialized to KVM_UNMAPPED_PAGE.
268 * Used to check if the mapping is valid or not. Never use 'kvm_host_map'
269 * directly to check for that.
271 static inline bool kvm_vcpu_mapped(struct kvm_host_map *map)
277 * Sometimes a large or cross-page mmio needs to be broken up into separate
278 * exits for userspace servicing.
280 struct kvm_mmio_fragment {
288 #ifdef CONFIG_PREEMPT_NOTIFIERS
289 struct preempt_notifier preempt_notifier;
292 int vcpu_id; /* id given by userspace at creation */
293 int vcpu_idx; /* index in kvm->vcpus array */
297 unsigned long guest_debug;
300 struct list_head blocked_vcpu_list;
306 struct pid __rcu *pid;
309 struct kvm_vcpu_stat stat;
310 unsigned int halt_poll_ns;
313 #ifdef CONFIG_HAS_IOMEM
315 int mmio_read_completed;
317 int mmio_cur_fragment;
318 int mmio_nr_fragments;
319 struct kvm_mmio_fragment mmio_fragments[KVM_MAX_MMIO_FRAGMENTS];
322 #ifdef CONFIG_KVM_ASYNC_PF
325 struct list_head queue;
326 struct list_head done;
331 #ifdef CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT
333 * Cpu relax intercept or pause loop exit optimization
334 * in_spin_loop: set when a vcpu does a pause loop exit
335 * or cpu relax intercepted.
336 * dy_eligible: indicates whether vcpu is eligible for directed yield.
345 struct kvm_vcpu_arch arch;
346 struct kvm_dirty_ring dirty_ring;
349 static inline int kvm_vcpu_exiting_guest_mode(struct kvm_vcpu *vcpu)
352 * The memory barrier ensures a previous write to vcpu->requests cannot
353 * be reordered with the read of vcpu->mode. It pairs with the general
354 * memory barrier following the write of vcpu->mode in VCPU RUN.
356 smp_mb__before_atomic();
357 return cmpxchg(&vcpu->mode, IN_GUEST_MODE, EXITING_GUEST_MODE);
361 * Some of the bitops functions do not support too long bitmaps.
362 * This number must be determined not to exceed such limits.
364 #define KVM_MEM_MAX_NR_PAGES ((1UL << 31) - 1)
366 struct kvm_memory_slot {
368 unsigned long npages;
369 unsigned long *dirty_bitmap;
370 struct kvm_arch_memory_slot arch;
371 unsigned long userspace_addr;
377 static inline bool kvm_slot_dirty_track_enabled(struct kvm_memory_slot *slot)
379 return slot->flags & KVM_MEM_LOG_DIRTY_PAGES;
382 static inline unsigned long kvm_dirty_bitmap_bytes(struct kvm_memory_slot *memslot)
384 return ALIGN(memslot->npages, BITS_PER_LONG) / 8;
387 static inline unsigned long *kvm_second_dirty_bitmap(struct kvm_memory_slot *memslot)
389 unsigned long len = kvm_dirty_bitmap_bytes(memslot);
391 return memslot->dirty_bitmap + len / sizeof(*memslot->dirty_bitmap);
394 #ifndef KVM_DIRTY_LOG_MANUAL_CAPS
395 #define KVM_DIRTY_LOG_MANUAL_CAPS KVM_DIRTY_LOG_MANUAL_PROTECT_ENABLE
398 struct kvm_s390_adapter_int {
411 struct kvm_kernel_irq_routing_entry {
414 int (*set)(struct kvm_kernel_irq_routing_entry *e,
415 struct kvm *kvm, int irq_source_id, int level,
429 struct kvm_s390_adapter_int adapter;
430 struct kvm_hv_sint hv_sint;
432 struct hlist_node link;
435 #ifdef CONFIG_HAVE_KVM_IRQ_ROUTING
436 struct kvm_irq_routing_table {
437 int chip[KVM_NR_IRQCHIPS][KVM_IRQCHIP_NUM_PINS];
440 * Array indexed by gsi. Each entry contains list of irq chips
441 * the gsi is connected to.
443 struct hlist_head map[];
447 #ifndef KVM_PRIVATE_MEM_SLOTS
448 #define KVM_PRIVATE_MEM_SLOTS 0
451 #define KVM_MEM_SLOTS_NUM SHRT_MAX
452 #define KVM_USER_MEM_SLOTS (KVM_MEM_SLOTS_NUM - KVM_PRIVATE_MEM_SLOTS)
454 #ifndef __KVM_VCPU_MULTIPLE_ADDRESS_SPACE
455 static inline int kvm_arch_vcpu_memslots_id(struct kvm_vcpu *vcpu)
463 * memslots are not sorted by id anymore, please use id_to_memslot()
464 * to get the memslot by its id.
466 struct kvm_memslots {
468 /* The mapping table from slot id to the index in memslots[]. */
469 short id_to_index[KVM_MEM_SLOTS_NUM];
472 struct kvm_memory_slot memslots[];
476 #ifdef KVM_HAVE_MMU_RWLOCK
480 #endif /* KVM_HAVE_MMU_RWLOCK */
482 struct mutex slots_lock;
483 struct mm_struct *mm; /* userspace tied to this vm */
484 struct kvm_memslots __rcu *memslots[KVM_ADDRESS_SPACE_NUM];
485 struct kvm_vcpu *vcpus[KVM_MAX_VCPUS];
488 * created_vcpus is protected by kvm->lock, and is incremented
489 * at the beginning of KVM_CREATE_VCPU. online_vcpus is only
490 * incremented after storing the kvm_vcpu pointer in vcpus,
491 * and is accessed atomically.
493 atomic_t online_vcpus;
495 int last_boosted_vcpu;
496 struct list_head vm_list;
498 struct kvm_io_bus __rcu *buses[KVM_NR_BUSES];
499 #ifdef CONFIG_HAVE_KVM_EVENTFD
502 struct list_head items;
503 struct list_head resampler_list;
504 struct mutex resampler_lock;
506 struct list_head ioeventfds;
508 struct kvm_vm_stat stat;
509 struct kvm_arch arch;
510 refcount_t users_count;
511 #ifdef CONFIG_KVM_MMIO
512 struct kvm_coalesced_mmio_ring *coalesced_mmio_ring;
513 spinlock_t ring_lock;
514 struct list_head coalesced_zones;
517 struct mutex irq_lock;
518 #ifdef CONFIG_HAVE_KVM_IRQCHIP
520 * Update side is protected by irq_lock.
522 struct kvm_irq_routing_table __rcu *irq_routing;
524 #ifdef CONFIG_HAVE_KVM_IRQFD
525 struct hlist_head irq_ack_notifier_list;
528 #if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
529 struct mmu_notifier mmu_notifier;
530 unsigned long mmu_notifier_seq;
531 long mmu_notifier_count;
532 unsigned long mmu_notifier_range_start;
533 unsigned long mmu_notifier_range_end;
536 struct list_head devices;
537 u64 manual_dirty_log_protect;
538 struct dentry *debugfs_dentry;
539 struct kvm_stat_data **debugfs_stat_data;
540 struct srcu_struct srcu;
541 struct srcu_struct irq_srcu;
543 unsigned int max_halt_poll_ns;
547 #define kvm_err(fmt, ...) \
548 pr_err("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
549 #define kvm_info(fmt, ...) \
550 pr_info("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
551 #define kvm_debug(fmt, ...) \
552 pr_debug("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
553 #define kvm_debug_ratelimited(fmt, ...) \
554 pr_debug_ratelimited("kvm [%i]: " fmt, task_pid_nr(current), \
556 #define kvm_pr_unimpl(fmt, ...) \
557 pr_err_ratelimited("kvm [%i]: " fmt, \
558 task_tgid_nr(current), ## __VA_ARGS__)
560 /* The guest did something we don't support. */
561 #define vcpu_unimpl(vcpu, fmt, ...) \
562 kvm_pr_unimpl("vcpu%i, guest rIP: 0x%lx " fmt, \
563 (vcpu)->vcpu_id, kvm_rip_read(vcpu), ## __VA_ARGS__)
565 #define vcpu_debug(vcpu, fmt, ...) \
566 kvm_debug("vcpu%i " fmt, (vcpu)->vcpu_id, ## __VA_ARGS__)
567 #define vcpu_debug_ratelimited(vcpu, fmt, ...) \
568 kvm_debug_ratelimited("vcpu%i " fmt, (vcpu)->vcpu_id, \
570 #define vcpu_err(vcpu, fmt, ...) \
571 kvm_err("vcpu%i " fmt, (vcpu)->vcpu_id, ## __VA_ARGS__)
573 static inline bool kvm_dirty_log_manual_protect_and_init_set(struct kvm *kvm)
575 return !!(kvm->manual_dirty_log_protect & KVM_DIRTY_LOG_INITIALLY_SET);
578 static inline struct kvm_io_bus *kvm_get_bus(struct kvm *kvm, enum kvm_bus idx)
580 return srcu_dereference_check(kvm->buses[idx], &kvm->srcu,
581 lockdep_is_held(&kvm->slots_lock) ||
582 !refcount_read(&kvm->users_count));
585 static inline struct kvm_vcpu *kvm_get_vcpu(struct kvm *kvm, int i)
587 int num_vcpus = atomic_read(&kvm->online_vcpus);
588 i = array_index_nospec(i, num_vcpus);
590 /* Pairs with smp_wmb() in kvm_vm_ioctl_create_vcpu. */
592 return kvm->vcpus[i];
595 #define kvm_for_each_vcpu(idx, vcpup, kvm) \
597 idx < atomic_read(&kvm->online_vcpus) && \
598 (vcpup = kvm_get_vcpu(kvm, idx)) != NULL; \
601 static inline struct kvm_vcpu *kvm_get_vcpu_by_id(struct kvm *kvm, int id)
603 struct kvm_vcpu *vcpu = NULL;
608 if (id < KVM_MAX_VCPUS)
609 vcpu = kvm_get_vcpu(kvm, id);
610 if (vcpu && vcpu->vcpu_id == id)
612 kvm_for_each_vcpu(i, vcpu, kvm)
613 if (vcpu->vcpu_id == id)
618 static inline int kvm_vcpu_get_idx(struct kvm_vcpu *vcpu)
620 return vcpu->vcpu_idx;
623 #define kvm_for_each_memslot(memslot, slots) \
624 for (memslot = &slots->memslots[0]; \
625 memslot < slots->memslots + slots->used_slots; memslot++) \
626 if (WARN_ON_ONCE(!memslot->npages)) { \
629 void kvm_vcpu_destroy(struct kvm_vcpu *vcpu);
631 void vcpu_load(struct kvm_vcpu *vcpu);
632 void vcpu_put(struct kvm_vcpu *vcpu);
634 #ifdef __KVM_HAVE_IOAPIC
635 void kvm_arch_post_irq_ack_notifier_list_update(struct kvm *kvm);
636 void kvm_arch_post_irq_routing_update(struct kvm *kvm);
638 static inline void kvm_arch_post_irq_ack_notifier_list_update(struct kvm *kvm)
641 static inline void kvm_arch_post_irq_routing_update(struct kvm *kvm)
646 #ifdef CONFIG_HAVE_KVM_IRQFD
647 int kvm_irqfd_init(void);
648 void kvm_irqfd_exit(void);
650 static inline int kvm_irqfd_init(void)
655 static inline void kvm_irqfd_exit(void)
659 int kvm_init(void *opaque, unsigned vcpu_size, unsigned vcpu_align,
660 struct module *module);
663 void kvm_get_kvm(struct kvm *kvm);
664 void kvm_put_kvm(struct kvm *kvm);
665 void kvm_put_kvm_no_destroy(struct kvm *kvm);
667 static inline struct kvm_memslots *__kvm_memslots(struct kvm *kvm, int as_id)
669 as_id = array_index_nospec(as_id, KVM_ADDRESS_SPACE_NUM);
670 return srcu_dereference_check(kvm->memslots[as_id], &kvm->srcu,
671 lockdep_is_held(&kvm->slots_lock) ||
672 !refcount_read(&kvm->users_count));
675 static inline struct kvm_memslots *kvm_memslots(struct kvm *kvm)
677 return __kvm_memslots(kvm, 0);
680 static inline struct kvm_memslots *kvm_vcpu_memslots(struct kvm_vcpu *vcpu)
682 int as_id = kvm_arch_vcpu_memslots_id(vcpu);
684 return __kvm_memslots(vcpu->kvm, as_id);
688 struct kvm_memory_slot *id_to_memslot(struct kvm_memslots *slots, int id)
690 int index = slots->id_to_index[id];
691 struct kvm_memory_slot *slot;
696 slot = &slots->memslots[index];
698 WARN_ON(slot->id != id);
703 * KVM_SET_USER_MEMORY_REGION ioctl allows the following operations:
704 * - create a new memory slot
705 * - delete an existing memory slot
706 * - modify an existing memory slot
707 * -- move it in the guest physical memory space
708 * -- just change its flags
710 * Since flags can be changed by some of these operations, the following
711 * differentiation is the best we can do for __kvm_set_memory_region():
720 int kvm_set_memory_region(struct kvm *kvm,
721 const struct kvm_userspace_memory_region *mem);
722 int __kvm_set_memory_region(struct kvm *kvm,
723 const struct kvm_userspace_memory_region *mem);
724 void kvm_arch_free_memslot(struct kvm *kvm, struct kvm_memory_slot *slot);
725 void kvm_arch_memslots_updated(struct kvm *kvm, u64 gen);
726 int kvm_arch_prepare_memory_region(struct kvm *kvm,
727 struct kvm_memory_slot *memslot,
728 const struct kvm_userspace_memory_region *mem,
729 enum kvm_mr_change change);
730 void kvm_arch_commit_memory_region(struct kvm *kvm,
731 const struct kvm_userspace_memory_region *mem,
732 struct kvm_memory_slot *old,
733 const struct kvm_memory_slot *new,
734 enum kvm_mr_change change);
735 /* flush all memory translations */
736 void kvm_arch_flush_shadow_all(struct kvm *kvm);
737 /* flush memory translations pointing to 'slot' */
738 void kvm_arch_flush_shadow_memslot(struct kvm *kvm,
739 struct kvm_memory_slot *slot);
741 int gfn_to_page_many_atomic(struct kvm_memory_slot *slot, gfn_t gfn,
742 struct page **pages, int nr_pages);
744 struct page *gfn_to_page(struct kvm *kvm, gfn_t gfn);
745 unsigned long gfn_to_hva(struct kvm *kvm, gfn_t gfn);
746 unsigned long gfn_to_hva_prot(struct kvm *kvm, gfn_t gfn, bool *writable);
747 unsigned long gfn_to_hva_memslot(struct kvm_memory_slot *slot, gfn_t gfn);
748 unsigned long gfn_to_hva_memslot_prot(struct kvm_memory_slot *slot, gfn_t gfn,
750 void kvm_release_page_clean(struct page *page);
751 void kvm_release_page_dirty(struct page *page);
752 void kvm_set_page_accessed(struct page *page);
754 kvm_pfn_t gfn_to_pfn(struct kvm *kvm, gfn_t gfn);
755 kvm_pfn_t gfn_to_pfn_prot(struct kvm *kvm, gfn_t gfn, bool write_fault,
757 kvm_pfn_t gfn_to_pfn_memslot(struct kvm_memory_slot *slot, gfn_t gfn);
758 kvm_pfn_t gfn_to_pfn_memslot_atomic(struct kvm_memory_slot *slot, gfn_t gfn);
759 kvm_pfn_t __gfn_to_pfn_memslot(struct kvm_memory_slot *slot, gfn_t gfn,
760 bool atomic, bool *async, bool write_fault,
761 bool *writable, hva_t *hva);
763 void kvm_release_pfn_clean(kvm_pfn_t pfn);
764 void kvm_release_pfn_dirty(kvm_pfn_t pfn);
765 void kvm_set_pfn_dirty(kvm_pfn_t pfn);
766 void kvm_set_pfn_accessed(kvm_pfn_t pfn);
767 void kvm_get_pfn(kvm_pfn_t pfn);
769 void kvm_release_pfn(kvm_pfn_t pfn, bool dirty, struct gfn_to_pfn_cache *cache);
770 int kvm_read_guest_page(struct kvm *kvm, gfn_t gfn, void *data, int offset,
772 int kvm_read_guest(struct kvm *kvm, gpa_t gpa, void *data, unsigned long len);
773 int kvm_read_guest_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
774 void *data, unsigned long len);
775 int kvm_read_guest_offset_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
776 void *data, unsigned int offset,
778 int kvm_write_guest_page(struct kvm *kvm, gfn_t gfn, const void *data,
779 int offset, int len);
780 int kvm_write_guest(struct kvm *kvm, gpa_t gpa, const void *data,
782 int kvm_write_guest_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
783 void *data, unsigned long len);
784 int kvm_write_guest_offset_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
785 void *data, unsigned int offset,
787 int kvm_gfn_to_hva_cache_init(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
788 gpa_t gpa, unsigned long len);
790 #define __kvm_get_guest(kvm, gfn, offset, v) \
792 unsigned long __addr = gfn_to_hva(kvm, gfn); \
793 typeof(v) __user *__uaddr = (typeof(__uaddr))(__addr + offset); \
794 int __ret = -EFAULT; \
796 if (!kvm_is_error_hva(__addr)) \
797 __ret = get_user(v, __uaddr); \
801 #define kvm_get_guest(kvm, gpa, v) \
804 struct kvm *__kvm = kvm; \
806 __kvm_get_guest(__kvm, __gpa >> PAGE_SHIFT, \
807 offset_in_page(__gpa), v); \
810 #define __kvm_put_guest(kvm, gfn, offset, v) \
812 unsigned long __addr = gfn_to_hva(kvm, gfn); \
813 typeof(v) __user *__uaddr = (typeof(__uaddr))(__addr + offset); \
814 int __ret = -EFAULT; \
816 if (!kvm_is_error_hva(__addr)) \
817 __ret = put_user(v, __uaddr); \
819 mark_page_dirty(kvm, gfn); \
823 #define kvm_put_guest(kvm, gpa, v) \
826 struct kvm *__kvm = kvm; \
828 __kvm_put_guest(__kvm, __gpa >> PAGE_SHIFT, \
829 offset_in_page(__gpa), v); \
832 int kvm_clear_guest(struct kvm *kvm, gpa_t gpa, unsigned long len);
833 struct kvm_memory_slot *gfn_to_memslot(struct kvm *kvm, gfn_t gfn);
834 bool kvm_is_visible_gfn(struct kvm *kvm, gfn_t gfn);
835 bool kvm_vcpu_is_visible_gfn(struct kvm_vcpu *vcpu, gfn_t gfn);
836 unsigned long kvm_host_page_size(struct kvm_vcpu *vcpu, gfn_t gfn);
837 void mark_page_dirty_in_slot(struct kvm *kvm, struct kvm_memory_slot *memslot, gfn_t gfn);
838 void mark_page_dirty(struct kvm *kvm, gfn_t gfn);
840 struct kvm_memslots *kvm_vcpu_memslots(struct kvm_vcpu *vcpu);
841 struct kvm_memory_slot *kvm_vcpu_gfn_to_memslot(struct kvm_vcpu *vcpu, gfn_t gfn);
842 kvm_pfn_t kvm_vcpu_gfn_to_pfn_atomic(struct kvm_vcpu *vcpu, gfn_t gfn);
843 kvm_pfn_t kvm_vcpu_gfn_to_pfn(struct kvm_vcpu *vcpu, gfn_t gfn);
844 int kvm_vcpu_map(struct kvm_vcpu *vcpu, gpa_t gpa, struct kvm_host_map *map);
845 int kvm_map_gfn(struct kvm_vcpu *vcpu, gfn_t gfn, struct kvm_host_map *map,
846 struct gfn_to_pfn_cache *cache, bool atomic);
847 struct page *kvm_vcpu_gfn_to_page(struct kvm_vcpu *vcpu, gfn_t gfn);
848 void kvm_vcpu_unmap(struct kvm_vcpu *vcpu, struct kvm_host_map *map, bool dirty);
849 int kvm_unmap_gfn(struct kvm_vcpu *vcpu, struct kvm_host_map *map,
850 struct gfn_to_pfn_cache *cache, bool dirty, bool atomic);
851 unsigned long kvm_vcpu_gfn_to_hva(struct kvm_vcpu *vcpu, gfn_t gfn);
852 unsigned long kvm_vcpu_gfn_to_hva_prot(struct kvm_vcpu *vcpu, gfn_t gfn, bool *writable);
853 int kvm_vcpu_read_guest_page(struct kvm_vcpu *vcpu, gfn_t gfn, void *data, int offset,
855 int kvm_vcpu_read_guest_atomic(struct kvm_vcpu *vcpu, gpa_t gpa, void *data,
857 int kvm_vcpu_read_guest(struct kvm_vcpu *vcpu, gpa_t gpa, void *data,
859 int kvm_vcpu_write_guest_page(struct kvm_vcpu *vcpu, gfn_t gfn, const void *data,
860 int offset, int len);
861 int kvm_vcpu_write_guest(struct kvm_vcpu *vcpu, gpa_t gpa, const void *data,
863 void kvm_vcpu_mark_page_dirty(struct kvm_vcpu *vcpu, gfn_t gfn);
865 void kvm_sigset_activate(struct kvm_vcpu *vcpu);
866 void kvm_sigset_deactivate(struct kvm_vcpu *vcpu);
868 void kvm_vcpu_block(struct kvm_vcpu *vcpu);
869 void kvm_arch_vcpu_blocking(struct kvm_vcpu *vcpu);
870 void kvm_arch_vcpu_unblocking(struct kvm_vcpu *vcpu);
871 bool kvm_vcpu_wake_up(struct kvm_vcpu *vcpu);
872 void kvm_vcpu_kick(struct kvm_vcpu *vcpu);
873 int kvm_vcpu_yield_to(struct kvm_vcpu *target);
874 void kvm_vcpu_on_spin(struct kvm_vcpu *vcpu, bool usermode_vcpu_not_eligible);
876 void kvm_flush_remote_tlbs(struct kvm *kvm);
877 void kvm_reload_remote_mmus(struct kvm *kvm);
879 #ifdef KVM_ARCH_NR_OBJS_PER_MEMORY_CACHE
880 int kvm_mmu_topup_memory_cache(struct kvm_mmu_memory_cache *mc, int min);
881 int kvm_mmu_memory_cache_nr_free_objects(struct kvm_mmu_memory_cache *mc);
882 void kvm_mmu_free_memory_cache(struct kvm_mmu_memory_cache *mc);
883 void *kvm_mmu_memory_cache_alloc(struct kvm_mmu_memory_cache *mc);
886 bool kvm_make_vcpus_request_mask(struct kvm *kvm, unsigned int req,
887 struct kvm_vcpu *except,
888 unsigned long *vcpu_bitmap, cpumask_var_t tmp);
889 bool kvm_make_all_cpus_request(struct kvm *kvm, unsigned int req);
890 bool kvm_make_all_cpus_request_except(struct kvm *kvm, unsigned int req,
891 struct kvm_vcpu *except);
892 bool kvm_make_cpus_request_mask(struct kvm *kvm, unsigned int req,
893 unsigned long *vcpu_bitmap);
895 long kvm_arch_dev_ioctl(struct file *filp,
896 unsigned int ioctl, unsigned long arg);
897 long kvm_arch_vcpu_ioctl(struct file *filp,
898 unsigned int ioctl, unsigned long arg);
899 vm_fault_t kvm_arch_vcpu_fault(struct kvm_vcpu *vcpu, struct vm_fault *vmf);
901 int kvm_vm_ioctl_check_extension(struct kvm *kvm, long ext);
903 void kvm_arch_mmu_enable_log_dirty_pt_masked(struct kvm *kvm,
904 struct kvm_memory_slot *slot,
907 void kvm_arch_sync_dirty_log(struct kvm *kvm, struct kvm_memory_slot *memslot);
909 #ifdef CONFIG_KVM_GENERIC_DIRTYLOG_READ_PROTECT
910 void kvm_arch_flush_remote_tlbs_memslot(struct kvm *kvm,
911 const struct kvm_memory_slot *memslot);
912 #else /* !CONFIG_KVM_GENERIC_DIRTYLOG_READ_PROTECT */
913 int kvm_vm_ioctl_get_dirty_log(struct kvm *kvm, struct kvm_dirty_log *log);
914 int kvm_get_dirty_log(struct kvm *kvm, struct kvm_dirty_log *log,
915 int *is_dirty, struct kvm_memory_slot **memslot);
918 int kvm_vm_ioctl_irq_line(struct kvm *kvm, struct kvm_irq_level *irq_level,
920 int kvm_vm_ioctl_enable_cap(struct kvm *kvm,
921 struct kvm_enable_cap *cap);
922 long kvm_arch_vm_ioctl(struct file *filp,
923 unsigned int ioctl, unsigned long arg);
925 int kvm_arch_vcpu_ioctl_get_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu);
926 int kvm_arch_vcpu_ioctl_set_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu);
928 int kvm_arch_vcpu_ioctl_translate(struct kvm_vcpu *vcpu,
929 struct kvm_translation *tr);
931 int kvm_arch_vcpu_ioctl_get_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs);
932 int kvm_arch_vcpu_ioctl_set_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs);
933 int kvm_arch_vcpu_ioctl_get_sregs(struct kvm_vcpu *vcpu,
934 struct kvm_sregs *sregs);
935 int kvm_arch_vcpu_ioctl_set_sregs(struct kvm_vcpu *vcpu,
936 struct kvm_sregs *sregs);
937 int kvm_arch_vcpu_ioctl_get_mpstate(struct kvm_vcpu *vcpu,
938 struct kvm_mp_state *mp_state);
939 int kvm_arch_vcpu_ioctl_set_mpstate(struct kvm_vcpu *vcpu,
940 struct kvm_mp_state *mp_state);
941 int kvm_arch_vcpu_ioctl_set_guest_debug(struct kvm_vcpu *vcpu,
942 struct kvm_guest_debug *dbg);
943 int kvm_arch_vcpu_ioctl_run(struct kvm_vcpu *vcpu);
945 int kvm_arch_init(void *opaque);
946 void kvm_arch_exit(void);
948 void kvm_arch_sched_in(struct kvm_vcpu *vcpu, int cpu);
950 void kvm_arch_vcpu_load(struct kvm_vcpu *vcpu, int cpu);
951 void kvm_arch_vcpu_put(struct kvm_vcpu *vcpu);
952 int kvm_arch_vcpu_precreate(struct kvm *kvm, unsigned int id);
953 int kvm_arch_vcpu_create(struct kvm_vcpu *vcpu);
954 void kvm_arch_vcpu_postcreate(struct kvm_vcpu *vcpu);
955 void kvm_arch_vcpu_destroy(struct kvm_vcpu *vcpu);
957 #ifdef __KVM_HAVE_ARCH_VCPU_DEBUGFS
958 void kvm_arch_create_vcpu_debugfs(struct kvm_vcpu *vcpu, struct dentry *debugfs_dentry);
961 int kvm_arch_hardware_enable(void);
962 void kvm_arch_hardware_disable(void);
963 int kvm_arch_hardware_setup(void *opaque);
964 void kvm_arch_hardware_unsetup(void);
965 int kvm_arch_check_processor_compat(void *opaque);
966 int kvm_arch_vcpu_runnable(struct kvm_vcpu *vcpu);
967 bool kvm_arch_vcpu_in_kernel(struct kvm_vcpu *vcpu);
968 int kvm_arch_vcpu_should_kick(struct kvm_vcpu *vcpu);
969 bool kvm_arch_dy_runnable(struct kvm_vcpu *vcpu);
970 int kvm_arch_post_init_vm(struct kvm *kvm);
971 void kvm_arch_pre_destroy_vm(struct kvm *kvm);
973 #ifndef __KVM_HAVE_ARCH_VM_ALLOC
975 * All architectures that want to use vzalloc currently also
976 * need their own kvm_arch_alloc_vm implementation.
978 static inline struct kvm *kvm_arch_alloc_vm(void)
980 return kzalloc(sizeof(struct kvm), GFP_KERNEL);
983 static inline void kvm_arch_free_vm(struct kvm *kvm)
989 #ifndef __KVM_HAVE_ARCH_FLUSH_REMOTE_TLB
990 static inline int kvm_arch_flush_remote_tlb(struct kvm *kvm)
996 #ifdef __KVM_HAVE_ARCH_NONCOHERENT_DMA
997 void kvm_arch_register_noncoherent_dma(struct kvm *kvm);
998 void kvm_arch_unregister_noncoherent_dma(struct kvm *kvm);
999 bool kvm_arch_has_noncoherent_dma(struct kvm *kvm);
1001 static inline void kvm_arch_register_noncoherent_dma(struct kvm *kvm)
1005 static inline void kvm_arch_unregister_noncoherent_dma(struct kvm *kvm)
1009 static inline bool kvm_arch_has_noncoherent_dma(struct kvm *kvm)
1014 #ifdef __KVM_HAVE_ARCH_ASSIGNED_DEVICE
1015 void kvm_arch_start_assignment(struct kvm *kvm);
1016 void kvm_arch_end_assignment(struct kvm *kvm);
1017 bool kvm_arch_has_assigned_device(struct kvm *kvm);
1019 static inline void kvm_arch_start_assignment(struct kvm *kvm)
1023 static inline void kvm_arch_end_assignment(struct kvm *kvm)
1027 static inline bool kvm_arch_has_assigned_device(struct kvm *kvm)
1033 static inline struct rcuwait *kvm_arch_vcpu_get_wait(struct kvm_vcpu *vcpu)
1035 #ifdef __KVM_HAVE_ARCH_WQP
1036 return vcpu->arch.waitp;
1042 #ifdef __KVM_HAVE_ARCH_INTC_INITIALIZED
1044 * returns true if the virtual interrupt controller is initialized and
1045 * ready to accept virtual IRQ. On some architectures the virtual interrupt
1046 * controller is dynamically instantiated and this is not always true.
1048 bool kvm_arch_intc_initialized(struct kvm *kvm);
1050 static inline bool kvm_arch_intc_initialized(struct kvm *kvm)
1056 int kvm_arch_init_vm(struct kvm *kvm, unsigned long type);
1057 void kvm_arch_destroy_vm(struct kvm *kvm);
1058 void kvm_arch_sync_events(struct kvm *kvm);
1060 int kvm_cpu_has_pending_timer(struct kvm_vcpu *vcpu);
1062 bool kvm_is_reserved_pfn(kvm_pfn_t pfn);
1063 bool kvm_is_zone_device_pfn(kvm_pfn_t pfn);
1064 bool kvm_is_transparent_hugepage(kvm_pfn_t pfn);
1066 struct kvm_irq_ack_notifier {
1067 struct hlist_node link;
1069 void (*irq_acked)(struct kvm_irq_ack_notifier *kian);
1072 int kvm_irq_map_gsi(struct kvm *kvm,
1073 struct kvm_kernel_irq_routing_entry *entries, int gsi);
1074 int kvm_irq_map_chip_pin(struct kvm *kvm, unsigned irqchip, unsigned pin);
1076 int kvm_set_irq(struct kvm *kvm, int irq_source_id, u32 irq, int level,
1078 int kvm_set_msi(struct kvm_kernel_irq_routing_entry *irq_entry, struct kvm *kvm,
1079 int irq_source_id, int level, bool line_status);
1080 int kvm_arch_set_irq_inatomic(struct kvm_kernel_irq_routing_entry *e,
1081 struct kvm *kvm, int irq_source_id,
1082 int level, bool line_status);
1083 bool kvm_irq_has_notifier(struct kvm *kvm, unsigned irqchip, unsigned pin);
1084 void kvm_notify_acked_gsi(struct kvm *kvm, int gsi);
1085 void kvm_notify_acked_irq(struct kvm *kvm, unsigned irqchip, unsigned pin);
1086 void kvm_register_irq_ack_notifier(struct kvm *kvm,
1087 struct kvm_irq_ack_notifier *kian);
1088 void kvm_unregister_irq_ack_notifier(struct kvm *kvm,
1089 struct kvm_irq_ack_notifier *kian);
1090 int kvm_request_irq_source_id(struct kvm *kvm);
1091 void kvm_free_irq_source_id(struct kvm *kvm, int irq_source_id);
1092 bool kvm_arch_irqfd_allowed(struct kvm *kvm, struct kvm_irqfd *args);
1095 * search_memslots() and __gfn_to_memslot() are here because they are
1096 * used in non-modular code in arch/powerpc/kvm/book3s_hv_rm_mmu.c.
1097 * gfn_to_memslot() itself isn't here as an inline because that would
1098 * bloat other code too much.
1100 * IMPORTANT: Slots are sorted from highest GFN to lowest GFN!
1102 static inline struct kvm_memory_slot *
1103 search_memslots(struct kvm_memslots *slots, gfn_t gfn)
1105 int start = 0, end = slots->used_slots;
1106 int slot = atomic_read(&slots->lru_slot);
1107 struct kvm_memory_slot *memslots = slots->memslots;
1109 if (unlikely(!slots->used_slots))
1112 if (gfn >= memslots[slot].base_gfn &&
1113 gfn < memslots[slot].base_gfn + memslots[slot].npages)
1114 return &memslots[slot];
1116 while (start < end) {
1117 slot = start + (end - start) / 2;
1119 if (gfn >= memslots[slot].base_gfn)
1125 if (start < slots->used_slots && gfn >= memslots[start].base_gfn &&
1126 gfn < memslots[start].base_gfn + memslots[start].npages) {
1127 atomic_set(&slots->lru_slot, start);
1128 return &memslots[start];
1134 static inline struct kvm_memory_slot *
1135 __gfn_to_memslot(struct kvm_memslots *slots, gfn_t gfn)
1137 return search_memslots(slots, gfn);
1140 static inline unsigned long
1141 __gfn_to_hva_memslot(struct kvm_memory_slot *slot, gfn_t gfn)
1143 return slot->userspace_addr + (gfn - slot->base_gfn) * PAGE_SIZE;
1146 static inline int memslot_id(struct kvm *kvm, gfn_t gfn)
1148 return gfn_to_memslot(kvm, gfn)->id;
1152 hva_to_gfn_memslot(unsigned long hva, struct kvm_memory_slot *slot)
1154 gfn_t gfn_offset = (hva - slot->userspace_addr) >> PAGE_SHIFT;
1156 return slot->base_gfn + gfn_offset;
1159 static inline gpa_t gfn_to_gpa(gfn_t gfn)
1161 return (gpa_t)gfn << PAGE_SHIFT;
1164 static inline gfn_t gpa_to_gfn(gpa_t gpa)
1166 return (gfn_t)(gpa >> PAGE_SHIFT);
1169 static inline hpa_t pfn_to_hpa(kvm_pfn_t pfn)
1171 return (hpa_t)pfn << PAGE_SHIFT;
1174 static inline struct page *kvm_vcpu_gpa_to_page(struct kvm_vcpu *vcpu,
1177 return kvm_vcpu_gfn_to_page(vcpu, gpa_to_gfn(gpa));
1180 static inline bool kvm_is_error_gpa(struct kvm *kvm, gpa_t gpa)
1182 unsigned long hva = gfn_to_hva(kvm, gpa_to_gfn(gpa));
1184 return kvm_is_error_hva(hva);
1187 enum kvm_stat_kind {
1192 struct kvm_stat_data {
1194 struct kvm_stats_debugfs_item *dbgfs_item;
1197 struct kvm_stats_debugfs_item {
1200 enum kvm_stat_kind kind;
1204 #define KVM_DBGFS_GET_MODE(dbgfs_item) \
1205 ((dbgfs_item)->mode ? (dbgfs_item)->mode : 0644)
1207 #define VM_STAT(n, x, ...) \
1208 { n, offsetof(struct kvm, stat.x), KVM_STAT_VM, ## __VA_ARGS__ }
1209 #define VCPU_STAT(n, x, ...) \
1210 { n, offsetof(struct kvm_vcpu, stat.x), KVM_STAT_VCPU, ## __VA_ARGS__ }
1212 extern struct kvm_stats_debugfs_item debugfs_entries[];
1213 extern struct dentry *kvm_debugfs_dir;
1215 #if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
1216 static inline int mmu_notifier_retry(struct kvm *kvm, unsigned long mmu_seq)
1218 if (unlikely(kvm->mmu_notifier_count))
1221 * Ensure the read of mmu_notifier_count happens before the read
1222 * of mmu_notifier_seq. This interacts with the smp_wmb() in
1223 * mmu_notifier_invalidate_range_end to make sure that the caller
1224 * either sees the old (non-zero) value of mmu_notifier_count or
1225 * the new (incremented) value of mmu_notifier_seq.
1226 * PowerPC Book3s HV KVM calls this under a per-page lock
1227 * rather than under kvm->mmu_lock, for scalability, so
1228 * can't rely on kvm->mmu_lock to keep things ordered.
1231 if (kvm->mmu_notifier_seq != mmu_seq)
1236 static inline int mmu_notifier_retry_hva(struct kvm *kvm,
1237 unsigned long mmu_seq,
1240 lockdep_assert_held(&kvm->mmu_lock);
1242 * If mmu_notifier_count is non-zero, then the range maintained by
1243 * kvm_mmu_notifier_invalidate_range_start contains all addresses that
1244 * might be being invalidated. Note that it may include some false
1245 * positives, due to shortcuts when handing concurrent invalidations.
1247 if (unlikely(kvm->mmu_notifier_count) &&
1248 hva >= kvm->mmu_notifier_range_start &&
1249 hva < kvm->mmu_notifier_range_end)
1251 if (kvm->mmu_notifier_seq != mmu_seq)
1257 #ifdef CONFIG_HAVE_KVM_IRQ_ROUTING
1259 #define KVM_MAX_IRQ_ROUTES 4096 /* might need extension/rework in the future */
1261 bool kvm_arch_can_set_irq_routing(struct kvm *kvm);
1262 int kvm_set_irq_routing(struct kvm *kvm,
1263 const struct kvm_irq_routing_entry *entries,
1266 int kvm_set_routing_entry(struct kvm *kvm,
1267 struct kvm_kernel_irq_routing_entry *e,
1268 const struct kvm_irq_routing_entry *ue);
1269 void kvm_free_irq_routing(struct kvm *kvm);
1273 static inline void kvm_free_irq_routing(struct kvm *kvm) {}
1277 int kvm_send_userspace_msi(struct kvm *kvm, struct kvm_msi *msi);
1279 #ifdef CONFIG_HAVE_KVM_EVENTFD
1281 void kvm_eventfd_init(struct kvm *kvm);
1282 int kvm_ioeventfd(struct kvm *kvm, struct kvm_ioeventfd *args);
1284 #ifdef CONFIG_HAVE_KVM_IRQFD
1285 int kvm_irqfd(struct kvm *kvm, struct kvm_irqfd *args);
1286 void kvm_irqfd_release(struct kvm *kvm);
1287 void kvm_irq_routing_update(struct kvm *);
1289 static inline int kvm_irqfd(struct kvm *kvm, struct kvm_irqfd *args)
1294 static inline void kvm_irqfd_release(struct kvm *kvm) {}
1299 static inline void kvm_eventfd_init(struct kvm *kvm) {}
1301 static inline int kvm_irqfd(struct kvm *kvm, struct kvm_irqfd *args)
1306 static inline void kvm_irqfd_release(struct kvm *kvm) {}
1308 #ifdef CONFIG_HAVE_KVM_IRQCHIP
1309 static inline void kvm_irq_routing_update(struct kvm *kvm)
1314 static inline int kvm_ioeventfd(struct kvm *kvm, struct kvm_ioeventfd *args)
1319 #endif /* CONFIG_HAVE_KVM_EVENTFD */
1321 void kvm_arch_irq_routing_update(struct kvm *kvm);
1323 static inline void kvm_make_request(int req, struct kvm_vcpu *vcpu)
1326 * Ensure the rest of the request is published to kvm_check_request's
1327 * caller. Paired with the smp_mb__after_atomic in kvm_check_request.
1330 set_bit(req & KVM_REQUEST_MASK, (void *)&vcpu->requests);
1333 static inline bool kvm_request_pending(struct kvm_vcpu *vcpu)
1335 return READ_ONCE(vcpu->requests);
1338 static inline bool kvm_test_request(int req, struct kvm_vcpu *vcpu)
1340 return test_bit(req & KVM_REQUEST_MASK, (void *)&vcpu->requests);
1343 static inline void kvm_clear_request(int req, struct kvm_vcpu *vcpu)
1345 clear_bit(req & KVM_REQUEST_MASK, (void *)&vcpu->requests);
1348 static inline bool kvm_check_request(int req, struct kvm_vcpu *vcpu)
1350 if (kvm_test_request(req, vcpu)) {
1351 kvm_clear_request(req, vcpu);
1354 * Ensure the rest of the request is visible to kvm_check_request's
1355 * caller. Paired with the smp_wmb in kvm_make_request.
1357 smp_mb__after_atomic();
1364 extern bool kvm_rebooting;
1366 extern unsigned int halt_poll_ns;
1367 extern unsigned int halt_poll_ns_grow;
1368 extern unsigned int halt_poll_ns_grow_start;
1369 extern unsigned int halt_poll_ns_shrink;
1372 const struct kvm_device_ops *ops;
1375 struct list_head vm_node;
1378 /* create, destroy, and name are mandatory */
1379 struct kvm_device_ops {
1383 * create is called holding kvm->lock and any operations not suitable
1384 * to do while holding the lock should be deferred to init (see
1387 int (*create)(struct kvm_device *dev, u32 type);
1390 * init is called after create if create is successful and is called
1391 * outside of holding kvm->lock.
1393 void (*init)(struct kvm_device *dev);
1396 * Destroy is responsible for freeing dev.
1398 * Destroy may be called before or after destructors are called
1399 * on emulated I/O regions, depending on whether a reference is
1400 * held by a vcpu or other kvm component that gets destroyed
1401 * after the emulated I/O.
1403 void (*destroy)(struct kvm_device *dev);
1406 * Release is an alternative method to free the device. It is
1407 * called when the device file descriptor is closed. Once
1408 * release is called, the destroy method will not be called
1409 * anymore as the device is removed from the device list of
1410 * the VM. kvm->lock is held.
1412 void (*release)(struct kvm_device *dev);
1414 int (*set_attr)(struct kvm_device *dev, struct kvm_device_attr *attr);
1415 int (*get_attr)(struct kvm_device *dev, struct kvm_device_attr *attr);
1416 int (*has_attr)(struct kvm_device *dev, struct kvm_device_attr *attr);
1417 long (*ioctl)(struct kvm_device *dev, unsigned int ioctl,
1419 int (*mmap)(struct kvm_device *dev, struct vm_area_struct *vma);
1422 void kvm_device_get(struct kvm_device *dev);
1423 void kvm_device_put(struct kvm_device *dev);
1424 struct kvm_device *kvm_device_from_filp(struct file *filp);
1425 int kvm_register_device_ops(const struct kvm_device_ops *ops, u32 type);
1426 void kvm_unregister_device_ops(u32 type);
1428 extern struct kvm_device_ops kvm_mpic_ops;
1429 extern struct kvm_device_ops kvm_arm_vgic_v2_ops;
1430 extern struct kvm_device_ops kvm_arm_vgic_v3_ops;
1432 #ifdef CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT
1434 static inline void kvm_vcpu_set_in_spin_loop(struct kvm_vcpu *vcpu, bool val)
1436 vcpu->spin_loop.in_spin_loop = val;
1438 static inline void kvm_vcpu_set_dy_eligible(struct kvm_vcpu *vcpu, bool val)
1440 vcpu->spin_loop.dy_eligible = val;
1443 #else /* !CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT */
1445 static inline void kvm_vcpu_set_in_spin_loop(struct kvm_vcpu *vcpu, bool val)
1449 static inline void kvm_vcpu_set_dy_eligible(struct kvm_vcpu *vcpu, bool val)
1452 #endif /* CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT */
1454 static inline bool kvm_is_visible_memslot(struct kvm_memory_slot *memslot)
1456 return (memslot && memslot->id < KVM_USER_MEM_SLOTS &&
1457 !(memslot->flags & KVM_MEMSLOT_INVALID));
1460 struct kvm_vcpu *kvm_get_running_vcpu(void);
1461 struct kvm_vcpu * __percpu *kvm_get_running_vcpus(void);
1463 #ifdef CONFIG_HAVE_KVM_IRQ_BYPASS
1464 bool kvm_arch_has_irq_bypass(void);
1465 int kvm_arch_irq_bypass_add_producer(struct irq_bypass_consumer *,
1466 struct irq_bypass_producer *);
1467 void kvm_arch_irq_bypass_del_producer(struct irq_bypass_consumer *,
1468 struct irq_bypass_producer *);
1469 void kvm_arch_irq_bypass_stop(struct irq_bypass_consumer *);
1470 void kvm_arch_irq_bypass_start(struct irq_bypass_consumer *);
1471 int kvm_arch_update_irqfd_routing(struct kvm *kvm, unsigned int host_irq,
1472 uint32_t guest_irq, bool set);
1473 #endif /* CONFIG_HAVE_KVM_IRQ_BYPASS */
1475 #ifdef CONFIG_HAVE_KVM_INVALID_WAKEUPS
1476 /* If we wakeup during the poll time, was it a sucessful poll? */
1477 static inline bool vcpu_valid_wakeup(struct kvm_vcpu *vcpu)
1479 return vcpu->valid_wakeup;
1483 static inline bool vcpu_valid_wakeup(struct kvm_vcpu *vcpu)
1487 #endif /* CONFIG_HAVE_KVM_INVALID_WAKEUPS */
1489 #ifdef CONFIG_HAVE_KVM_NO_POLL
1490 /* Callback that tells if we must not poll */
1491 bool kvm_arch_no_poll(struct kvm_vcpu *vcpu);
1493 static inline bool kvm_arch_no_poll(struct kvm_vcpu *vcpu)
1497 #endif /* CONFIG_HAVE_KVM_NO_POLL */
1499 #ifdef CONFIG_HAVE_KVM_VCPU_ASYNC_IOCTL
1500 long kvm_arch_vcpu_async_ioctl(struct file *filp,
1501 unsigned int ioctl, unsigned long arg);
1503 static inline long kvm_arch_vcpu_async_ioctl(struct file *filp,
1507 return -ENOIOCTLCMD;
1509 #endif /* CONFIG_HAVE_KVM_VCPU_ASYNC_IOCTL */
1511 void kvm_arch_mmu_notifier_invalidate_range(struct kvm *kvm,
1512 unsigned long start, unsigned long end);
1514 #ifdef CONFIG_HAVE_KVM_VCPU_RUN_PID_CHANGE
1515 int kvm_arch_vcpu_run_pid_change(struct kvm_vcpu *vcpu);
1517 static inline int kvm_arch_vcpu_run_pid_change(struct kvm_vcpu *vcpu)
1521 #endif /* CONFIG_HAVE_KVM_VCPU_RUN_PID_CHANGE */
1523 typedef int (*kvm_vm_thread_fn_t)(struct kvm *kvm, uintptr_t data);
1525 int kvm_vm_create_worker_thread(struct kvm *kvm, kvm_vm_thread_fn_t thread_fn,
1526 uintptr_t data, const char *name,
1527 struct task_struct **thread_ptr);
1529 #ifdef CONFIG_KVM_XFER_TO_GUEST_WORK
1530 static inline void kvm_handle_signal_exit(struct kvm_vcpu *vcpu)
1532 vcpu->run->exit_reason = KVM_EXIT_INTR;
1533 vcpu->stat.signal_exits++;
1535 #endif /* CONFIG_KVM_XFER_TO_GUEST_WORK */
1538 * This defines how many reserved entries we want to keep before we
1539 * kick the vcpu to the userspace to avoid dirty ring full. This
1540 * value can be tuned to higher if e.g. PML is enabled on the host.
1542 #define KVM_DIRTY_RING_RSVD_ENTRIES 64
1544 /* Max number of entries allowed for each kvm dirty ring */
1545 #define KVM_DIRTY_RING_MAX_ENTRIES 65536