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>
15 #include <linux/mmu_notifier.h>
16 #include <linux/preempt.h>
17 #include <linux/msi.h>
18 #include <linux/slab.h>
19 #include <linux/vmalloc.h>
20 #include <linux/rcupdate.h>
21 #include <linux/ratelimit.h>
22 #include <linux/err.h>
23 #include <linux/irqflags.h>
24 #include <linux/context_tracking.h>
25 #include <linux/irqbypass.h>
26 #include <linux/rcuwait.h>
27 #include <linux/refcount.h>
28 #include <linux/nospec.h>
29 #include <asm/signal.h>
31 #include <linux/kvm.h>
32 #include <linux/kvm_para.h>
34 #include <linux/kvm_types.h>
36 #include <asm/kvm_host.h>
38 #ifndef KVM_MAX_VCPU_ID
39 #define KVM_MAX_VCPU_ID KVM_MAX_VCPUS
43 * The bit 16 ~ bit 31 of kvm_memory_region::flags are internally used
44 * in kvm, other bits are visible for userspace which are defined in
45 * include/linux/kvm_h.
47 #define KVM_MEMSLOT_INVALID (1UL << 16)
50 * Bit 63 of the memslot generation number is an "update in-progress flag",
51 * e.g. is temporarily set for the duration of install_new_memslots().
52 * This flag effectively creates a unique generation number that is used to
53 * mark cached memslot data, e.g. MMIO accesses, as potentially being stale,
54 * i.e. may (or may not) have come from the previous memslots generation.
56 * This is necessary because the actual memslots update is not atomic with
57 * respect to the generation number update. Updating the generation number
58 * first would allow a vCPU to cache a spte from the old memslots using the
59 * new generation number, and updating the generation number after switching
60 * to the new memslots would allow cache hits using the old generation number
61 * to reference the defunct memslots.
63 * This mechanism is used to prevent getting hits in KVM's caches while a
64 * memslot update is in-progress, and to prevent cache hits *after* updating
65 * the actual generation number against accesses that were inserted into the
66 * cache *before* the memslots were updated.
68 #define KVM_MEMSLOT_GEN_UPDATE_IN_PROGRESS BIT_ULL(63)
70 /* Two fragments for cross MMIO pages. */
71 #define KVM_MAX_MMIO_FRAGMENTS 2
73 #ifndef KVM_ADDRESS_SPACE_NUM
74 #define KVM_ADDRESS_SPACE_NUM 1
78 * For the normal pfn, the highest 12 bits should be zero,
79 * so we can mask bit 62 ~ bit 52 to indicate the error pfn,
80 * mask bit 63 to indicate the noslot pfn.
82 #define KVM_PFN_ERR_MASK (0x7ffULL << 52)
83 #define KVM_PFN_ERR_NOSLOT_MASK (0xfffULL << 52)
84 #define KVM_PFN_NOSLOT (0x1ULL << 63)
86 #define KVM_PFN_ERR_FAULT (KVM_PFN_ERR_MASK)
87 #define KVM_PFN_ERR_HWPOISON (KVM_PFN_ERR_MASK + 1)
88 #define KVM_PFN_ERR_RO_FAULT (KVM_PFN_ERR_MASK + 2)
91 * error pfns indicate that the gfn is in slot but faild to
92 * translate it to pfn on host.
94 static inline bool is_error_pfn(kvm_pfn_t pfn)
96 return !!(pfn & KVM_PFN_ERR_MASK);
100 * error_noslot pfns indicate that the gfn can not be
101 * translated to pfn - it is not in slot or failed to
102 * translate it to pfn.
104 static inline bool is_error_noslot_pfn(kvm_pfn_t pfn)
106 return !!(pfn & KVM_PFN_ERR_NOSLOT_MASK);
109 /* noslot pfn indicates that the gfn is not in slot. */
110 static inline bool is_noslot_pfn(kvm_pfn_t pfn)
112 return pfn == KVM_PFN_NOSLOT;
116 * architectures with KVM_HVA_ERR_BAD other than PAGE_OFFSET (e.g. s390)
117 * provide own defines and kvm_is_error_hva
119 #ifndef KVM_HVA_ERR_BAD
121 #define KVM_HVA_ERR_BAD (PAGE_OFFSET)
122 #define KVM_HVA_ERR_RO_BAD (PAGE_OFFSET + PAGE_SIZE)
124 static inline bool kvm_is_error_hva(unsigned long addr)
126 return addr >= PAGE_OFFSET;
131 #define KVM_ERR_PTR_BAD_PAGE (ERR_PTR(-ENOENT))
133 static inline bool is_error_page(struct page *page)
138 #define KVM_REQUEST_MASK GENMASK(7,0)
139 #define KVM_REQUEST_NO_WAKEUP BIT(8)
140 #define KVM_REQUEST_WAIT BIT(9)
142 * Architecture-independent vcpu->requests bit members
143 * Bits 4-7 are reserved for more arch-independent bits.
145 #define KVM_REQ_TLB_FLUSH (0 | KVM_REQUEST_WAIT | KVM_REQUEST_NO_WAKEUP)
146 #define KVM_REQ_MMU_RELOAD (1 | KVM_REQUEST_WAIT | KVM_REQUEST_NO_WAKEUP)
147 #define KVM_REQ_PENDING_TIMER 2
148 #define KVM_REQ_UNHALT 3
149 #define KVM_REQUEST_ARCH_BASE 8
151 #define KVM_ARCH_REQ_FLAGS(nr, flags) ({ \
152 BUILD_BUG_ON((unsigned)(nr) >= (sizeof_field(struct kvm_vcpu, requests) * 8) - KVM_REQUEST_ARCH_BASE); \
153 (unsigned)(((nr) + KVM_REQUEST_ARCH_BASE) | (flags)); \
155 #define KVM_ARCH_REQ(nr) KVM_ARCH_REQ_FLAGS(nr, 0)
157 #define KVM_USERSPACE_IRQ_SOURCE_ID 0
158 #define KVM_IRQFD_RESAMPLE_IRQ_SOURCE_ID 1
160 extern struct mutex kvm_lock;
161 extern struct list_head vm_list;
163 struct kvm_io_range {
166 struct kvm_io_device *dev;
169 #define NR_IOBUS_DEVS 1000
174 struct kvm_io_range range[];
180 KVM_VIRTIO_CCW_NOTIFY_BUS,
185 int kvm_io_bus_write(struct kvm_vcpu *vcpu, enum kvm_bus bus_idx, gpa_t addr,
186 int len, const void *val);
187 int kvm_io_bus_write_cookie(struct kvm_vcpu *vcpu, enum kvm_bus bus_idx,
188 gpa_t addr, int len, const void *val, long cookie);
189 int kvm_io_bus_read(struct kvm_vcpu *vcpu, enum kvm_bus bus_idx, gpa_t addr,
191 int kvm_io_bus_register_dev(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr,
192 int len, struct kvm_io_device *dev);
193 void kvm_io_bus_unregister_dev(struct kvm *kvm, enum kvm_bus bus_idx,
194 struct kvm_io_device *dev);
195 struct kvm_io_device *kvm_io_bus_get_dev(struct kvm *kvm, enum kvm_bus bus_idx,
198 #ifdef CONFIG_KVM_ASYNC_PF
199 struct kvm_async_pf {
200 struct work_struct work;
201 struct list_head link;
202 struct list_head queue;
203 struct kvm_vcpu *vcpu;
204 struct mm_struct *mm;
207 struct kvm_arch_async_pf arch;
209 bool notpresent_injected;
212 void kvm_clear_async_pf_completion_queue(struct kvm_vcpu *vcpu);
213 void kvm_check_async_pf_completion(struct kvm_vcpu *vcpu);
214 bool kvm_setup_async_pf(struct kvm_vcpu *vcpu, gpa_t cr2_or_gpa,
215 unsigned long hva, struct kvm_arch_async_pf *arch);
216 int kvm_async_pf_wakeup_all(struct kvm_vcpu *vcpu);
223 READING_SHADOW_PAGE_TABLES,
226 #define KVM_UNMAPPED_PAGE ((void *) 0x500 + POISON_POINTER_DELTA)
228 struct kvm_host_map {
230 * Only valid if the 'pfn' is managed by the host kernel (i.e. There is
231 * a 'struct page' for it. When using mem= kernel parameter some memory
232 * can be used as guest memory but they are not managed by host
234 * If 'pfn' is not managed by the host kernel, this field is
235 * initialized to KVM_UNMAPPED_PAGE.
244 * Used to check if the mapping is valid or not. Never use 'kvm_host_map'
245 * directly to check for that.
247 static inline bool kvm_vcpu_mapped(struct kvm_host_map *map)
253 * Sometimes a large or cross-page mmio needs to be broken up into separate
254 * exits for userspace servicing.
256 struct kvm_mmio_fragment {
264 #ifdef CONFIG_PREEMPT_NOTIFIERS
265 struct preempt_notifier preempt_notifier;
268 int vcpu_id; /* id given by userspace at creation */
269 int vcpu_idx; /* index in kvm->vcpus array */
273 unsigned long guest_debug;
276 struct list_head blocked_vcpu_list;
282 struct pid __rcu *pid;
285 struct kvm_vcpu_stat stat;
286 unsigned int halt_poll_ns;
289 #ifdef CONFIG_HAS_IOMEM
291 int mmio_read_completed;
293 int mmio_cur_fragment;
294 int mmio_nr_fragments;
295 struct kvm_mmio_fragment mmio_fragments[KVM_MAX_MMIO_FRAGMENTS];
298 #ifdef CONFIG_KVM_ASYNC_PF
301 struct list_head queue;
302 struct list_head done;
307 #ifdef CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT
309 * Cpu relax intercept or pause loop exit optimization
310 * in_spin_loop: set when a vcpu does a pause loop exit
311 * or cpu relax intercepted.
312 * dy_eligible: indicates whether vcpu is eligible for directed yield.
321 struct kvm_vcpu_arch arch;
324 static inline int kvm_vcpu_exiting_guest_mode(struct kvm_vcpu *vcpu)
327 * The memory barrier ensures a previous write to vcpu->requests cannot
328 * be reordered with the read of vcpu->mode. It pairs with the general
329 * memory barrier following the write of vcpu->mode in VCPU RUN.
331 smp_mb__before_atomic();
332 return cmpxchg(&vcpu->mode, IN_GUEST_MODE, EXITING_GUEST_MODE);
336 * Some of the bitops functions do not support too long bitmaps.
337 * This number must be determined not to exceed such limits.
339 #define KVM_MEM_MAX_NR_PAGES ((1UL << 31) - 1)
341 struct kvm_memory_slot {
343 unsigned long npages;
344 unsigned long *dirty_bitmap;
345 struct kvm_arch_memory_slot arch;
346 unsigned long userspace_addr;
352 static inline unsigned long kvm_dirty_bitmap_bytes(struct kvm_memory_slot *memslot)
354 return ALIGN(memslot->npages, BITS_PER_LONG) / 8;
357 static inline unsigned long *kvm_second_dirty_bitmap(struct kvm_memory_slot *memslot)
359 unsigned long len = kvm_dirty_bitmap_bytes(memslot);
361 return memslot->dirty_bitmap + len / sizeof(*memslot->dirty_bitmap);
364 #ifndef KVM_DIRTY_LOG_MANUAL_CAPS
365 #define KVM_DIRTY_LOG_MANUAL_CAPS KVM_DIRTY_LOG_MANUAL_PROTECT_ENABLE
368 struct kvm_s390_adapter_int {
381 struct kvm_kernel_irq_routing_entry {
384 int (*set)(struct kvm_kernel_irq_routing_entry *e,
385 struct kvm *kvm, int irq_source_id, int level,
399 struct kvm_s390_adapter_int adapter;
400 struct kvm_hv_sint hv_sint;
402 struct hlist_node link;
405 #ifdef CONFIG_HAVE_KVM_IRQ_ROUTING
406 struct kvm_irq_routing_table {
407 int chip[KVM_NR_IRQCHIPS][KVM_IRQCHIP_NUM_PINS];
410 * Array indexed by gsi. Each entry contains list of irq chips
411 * the gsi is connected to.
413 struct hlist_head map[];
417 #ifndef KVM_PRIVATE_MEM_SLOTS
418 #define KVM_PRIVATE_MEM_SLOTS 0
421 #ifndef KVM_MEM_SLOTS_NUM
422 #define KVM_MEM_SLOTS_NUM (KVM_USER_MEM_SLOTS + KVM_PRIVATE_MEM_SLOTS)
425 #ifndef __KVM_VCPU_MULTIPLE_ADDRESS_SPACE
426 static inline int kvm_arch_vcpu_memslots_id(struct kvm_vcpu *vcpu)
434 * memslots are not sorted by id anymore, please use id_to_memslot()
435 * to get the memslot by its id.
437 struct kvm_memslots {
439 /* The mapping table from slot id to the index in memslots[]. */
440 short id_to_index[KVM_MEM_SLOTS_NUM];
443 struct kvm_memory_slot memslots[];
448 struct mutex slots_lock;
449 struct mm_struct *mm; /* userspace tied to this vm */
450 struct kvm_memslots __rcu *memslots[KVM_ADDRESS_SPACE_NUM];
451 struct kvm_vcpu *vcpus[KVM_MAX_VCPUS];
454 * created_vcpus is protected by kvm->lock, and is incremented
455 * at the beginning of KVM_CREATE_VCPU. online_vcpus is only
456 * incremented after storing the kvm_vcpu pointer in vcpus,
457 * and is accessed atomically.
459 atomic_t online_vcpus;
461 int last_boosted_vcpu;
462 struct list_head vm_list;
464 struct kvm_io_bus __rcu *buses[KVM_NR_BUSES];
465 #ifdef CONFIG_HAVE_KVM_EVENTFD
468 struct list_head items;
469 struct list_head resampler_list;
470 struct mutex resampler_lock;
472 struct list_head ioeventfds;
474 struct kvm_vm_stat stat;
475 struct kvm_arch arch;
476 refcount_t users_count;
477 #ifdef CONFIG_KVM_MMIO
478 struct kvm_coalesced_mmio_ring *coalesced_mmio_ring;
479 spinlock_t ring_lock;
480 struct list_head coalesced_zones;
483 struct mutex irq_lock;
484 #ifdef CONFIG_HAVE_KVM_IRQCHIP
486 * Update side is protected by irq_lock.
488 struct kvm_irq_routing_table __rcu *irq_routing;
490 #ifdef CONFIG_HAVE_KVM_IRQFD
491 struct hlist_head irq_ack_notifier_list;
494 #if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
495 struct mmu_notifier mmu_notifier;
496 unsigned long mmu_notifier_seq;
497 long mmu_notifier_count;
500 struct list_head devices;
501 u64 manual_dirty_log_protect;
502 struct dentry *debugfs_dentry;
503 struct kvm_stat_data **debugfs_stat_data;
504 struct srcu_struct srcu;
505 struct srcu_struct irq_srcu;
507 unsigned int max_halt_poll_ns;
510 #define kvm_err(fmt, ...) \
511 pr_err("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
512 #define kvm_info(fmt, ...) \
513 pr_info("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
514 #define kvm_debug(fmt, ...) \
515 pr_debug("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
516 #define kvm_debug_ratelimited(fmt, ...) \
517 pr_debug_ratelimited("kvm [%i]: " fmt, task_pid_nr(current), \
519 #define kvm_pr_unimpl(fmt, ...) \
520 pr_err_ratelimited("kvm [%i]: " fmt, \
521 task_tgid_nr(current), ## __VA_ARGS__)
523 /* The guest did something we don't support. */
524 #define vcpu_unimpl(vcpu, fmt, ...) \
525 kvm_pr_unimpl("vcpu%i, guest rIP: 0x%lx " fmt, \
526 (vcpu)->vcpu_id, kvm_rip_read(vcpu), ## __VA_ARGS__)
528 #define vcpu_debug(vcpu, fmt, ...) \
529 kvm_debug("vcpu%i " fmt, (vcpu)->vcpu_id, ## __VA_ARGS__)
530 #define vcpu_debug_ratelimited(vcpu, fmt, ...) \
531 kvm_debug_ratelimited("vcpu%i " fmt, (vcpu)->vcpu_id, \
533 #define vcpu_err(vcpu, fmt, ...) \
534 kvm_err("vcpu%i " fmt, (vcpu)->vcpu_id, ## __VA_ARGS__)
536 static inline bool kvm_dirty_log_manual_protect_and_init_set(struct kvm *kvm)
538 return !!(kvm->manual_dirty_log_protect & KVM_DIRTY_LOG_INITIALLY_SET);
541 static inline struct kvm_io_bus *kvm_get_bus(struct kvm *kvm, enum kvm_bus idx)
543 return srcu_dereference_check(kvm->buses[idx], &kvm->srcu,
544 lockdep_is_held(&kvm->slots_lock) ||
545 !refcount_read(&kvm->users_count));
548 static inline struct kvm_vcpu *kvm_get_vcpu(struct kvm *kvm, int i)
550 int num_vcpus = atomic_read(&kvm->online_vcpus);
551 i = array_index_nospec(i, num_vcpus);
553 /* Pairs with smp_wmb() in kvm_vm_ioctl_create_vcpu. */
555 return kvm->vcpus[i];
558 #define kvm_for_each_vcpu(idx, vcpup, kvm) \
560 idx < atomic_read(&kvm->online_vcpus) && \
561 (vcpup = kvm_get_vcpu(kvm, idx)) != NULL; \
564 static inline struct kvm_vcpu *kvm_get_vcpu_by_id(struct kvm *kvm, int id)
566 struct kvm_vcpu *vcpu = NULL;
571 if (id < KVM_MAX_VCPUS)
572 vcpu = kvm_get_vcpu(kvm, id);
573 if (vcpu && vcpu->vcpu_id == id)
575 kvm_for_each_vcpu(i, vcpu, kvm)
576 if (vcpu->vcpu_id == id)
581 static inline int kvm_vcpu_get_idx(struct kvm_vcpu *vcpu)
583 return vcpu->vcpu_idx;
586 #define kvm_for_each_memslot(memslot, slots) \
587 for (memslot = &slots->memslots[0]; \
588 memslot < slots->memslots + slots->used_slots; memslot++) \
589 if (WARN_ON_ONCE(!memslot->npages)) { \
592 void kvm_vcpu_destroy(struct kvm_vcpu *vcpu);
594 void vcpu_load(struct kvm_vcpu *vcpu);
595 void vcpu_put(struct kvm_vcpu *vcpu);
597 #ifdef __KVM_HAVE_IOAPIC
598 void kvm_arch_post_irq_ack_notifier_list_update(struct kvm *kvm);
599 void kvm_arch_post_irq_routing_update(struct kvm *kvm);
601 static inline void kvm_arch_post_irq_ack_notifier_list_update(struct kvm *kvm)
604 static inline void kvm_arch_post_irq_routing_update(struct kvm *kvm)
609 #ifdef CONFIG_HAVE_KVM_IRQFD
610 int kvm_irqfd_init(void);
611 void kvm_irqfd_exit(void);
613 static inline int kvm_irqfd_init(void)
618 static inline void kvm_irqfd_exit(void)
622 int kvm_init(void *opaque, unsigned vcpu_size, unsigned vcpu_align,
623 struct module *module);
626 void kvm_get_kvm(struct kvm *kvm);
627 void kvm_put_kvm(struct kvm *kvm);
628 void kvm_put_kvm_no_destroy(struct kvm *kvm);
630 static inline struct kvm_memslots *__kvm_memslots(struct kvm *kvm, int as_id)
632 as_id = array_index_nospec(as_id, KVM_ADDRESS_SPACE_NUM);
633 return srcu_dereference_check(kvm->memslots[as_id], &kvm->srcu,
634 lockdep_is_held(&kvm->slots_lock) ||
635 !refcount_read(&kvm->users_count));
638 static inline struct kvm_memslots *kvm_memslots(struct kvm *kvm)
640 return __kvm_memslots(kvm, 0);
643 static inline struct kvm_memslots *kvm_vcpu_memslots(struct kvm_vcpu *vcpu)
645 int as_id = kvm_arch_vcpu_memslots_id(vcpu);
647 return __kvm_memslots(vcpu->kvm, as_id);
651 struct kvm_memory_slot *id_to_memslot(struct kvm_memslots *slots, int id)
653 int index = slots->id_to_index[id];
654 struct kvm_memory_slot *slot;
659 slot = &slots->memslots[index];
661 WARN_ON(slot->id != id);
666 * KVM_SET_USER_MEMORY_REGION ioctl allows the following operations:
667 * - create a new memory slot
668 * - delete an existing memory slot
669 * - modify an existing memory slot
670 * -- move it in the guest physical memory space
671 * -- just change its flags
673 * Since flags can be changed by some of these operations, the following
674 * differentiation is the best we can do for __kvm_set_memory_region():
683 int kvm_set_memory_region(struct kvm *kvm,
684 const struct kvm_userspace_memory_region *mem);
685 int __kvm_set_memory_region(struct kvm *kvm,
686 const struct kvm_userspace_memory_region *mem);
687 void kvm_arch_free_memslot(struct kvm *kvm, struct kvm_memory_slot *slot);
688 void kvm_arch_memslots_updated(struct kvm *kvm, u64 gen);
689 int kvm_arch_prepare_memory_region(struct kvm *kvm,
690 struct kvm_memory_slot *memslot,
691 const struct kvm_userspace_memory_region *mem,
692 enum kvm_mr_change change);
693 void kvm_arch_commit_memory_region(struct kvm *kvm,
694 const struct kvm_userspace_memory_region *mem,
695 struct kvm_memory_slot *old,
696 const struct kvm_memory_slot *new,
697 enum kvm_mr_change change);
698 /* flush all memory translations */
699 void kvm_arch_flush_shadow_all(struct kvm *kvm);
700 /* flush memory translations pointing to 'slot' */
701 void kvm_arch_flush_shadow_memslot(struct kvm *kvm,
702 struct kvm_memory_slot *slot);
704 int gfn_to_page_many_atomic(struct kvm_memory_slot *slot, gfn_t gfn,
705 struct page **pages, int nr_pages);
707 struct page *gfn_to_page(struct kvm *kvm, gfn_t gfn);
708 unsigned long gfn_to_hva(struct kvm *kvm, gfn_t gfn);
709 unsigned long gfn_to_hva_prot(struct kvm *kvm, gfn_t gfn, bool *writable);
710 unsigned long gfn_to_hva_memslot(struct kvm_memory_slot *slot, gfn_t gfn);
711 unsigned long gfn_to_hva_memslot_prot(struct kvm_memory_slot *slot, gfn_t gfn,
713 void kvm_release_page_clean(struct page *page);
714 void kvm_release_page_dirty(struct page *page);
715 void kvm_set_page_accessed(struct page *page);
717 kvm_pfn_t gfn_to_pfn(struct kvm *kvm, gfn_t gfn);
718 kvm_pfn_t gfn_to_pfn_prot(struct kvm *kvm, gfn_t gfn, bool write_fault,
720 kvm_pfn_t gfn_to_pfn_memslot(struct kvm_memory_slot *slot, gfn_t gfn);
721 kvm_pfn_t gfn_to_pfn_memslot_atomic(struct kvm_memory_slot *slot, gfn_t gfn);
722 kvm_pfn_t __gfn_to_pfn_memslot(struct kvm_memory_slot *slot, gfn_t gfn,
723 bool atomic, bool *async, bool write_fault,
726 void kvm_release_pfn_clean(kvm_pfn_t pfn);
727 void kvm_release_pfn_dirty(kvm_pfn_t pfn);
728 void kvm_set_pfn_dirty(kvm_pfn_t pfn);
729 void kvm_set_pfn_accessed(kvm_pfn_t pfn);
730 void kvm_get_pfn(kvm_pfn_t pfn);
732 void kvm_release_pfn(kvm_pfn_t pfn, bool dirty, struct gfn_to_pfn_cache *cache);
733 int kvm_read_guest_page(struct kvm *kvm, gfn_t gfn, void *data, int offset,
735 int kvm_read_guest(struct kvm *kvm, gpa_t gpa, void *data, unsigned long len);
736 int kvm_read_guest_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
737 void *data, unsigned long len);
738 int kvm_read_guest_offset_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
739 void *data, unsigned int offset,
741 int kvm_write_guest_page(struct kvm *kvm, gfn_t gfn, const void *data,
742 int offset, int len);
743 int kvm_write_guest(struct kvm *kvm, gpa_t gpa, const void *data,
745 int kvm_write_guest_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
746 void *data, unsigned long len);
747 int kvm_write_guest_offset_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
748 void *data, unsigned int offset,
750 int kvm_gfn_to_hva_cache_init(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
751 gpa_t gpa, unsigned long len);
753 #define __kvm_get_guest(kvm, gfn, offset, v) \
755 unsigned long __addr = gfn_to_hva(kvm, gfn); \
756 typeof(v) __user *__uaddr = (typeof(__uaddr))(__addr + offset); \
757 int __ret = -EFAULT; \
759 if (!kvm_is_error_hva(__addr)) \
760 __ret = get_user(v, __uaddr); \
764 #define kvm_get_guest(kvm, gpa, v) \
767 struct kvm *__kvm = kvm; \
769 __kvm_get_guest(__kvm, __gpa >> PAGE_SHIFT, \
770 offset_in_page(__gpa), v); \
773 #define __kvm_put_guest(kvm, gfn, offset, v) \
775 unsigned long __addr = gfn_to_hva(kvm, gfn); \
776 typeof(v) __user *__uaddr = (typeof(__uaddr))(__addr + offset); \
777 int __ret = -EFAULT; \
779 if (!kvm_is_error_hva(__addr)) \
780 __ret = put_user(v, __uaddr); \
782 mark_page_dirty(kvm, gfn); \
786 #define kvm_put_guest(kvm, gpa, v) \
789 struct kvm *__kvm = kvm; \
791 __kvm_put_guest(__kvm, __gpa >> PAGE_SHIFT, \
792 offset_in_page(__gpa), v); \
795 int kvm_clear_guest(struct kvm *kvm, gpa_t gpa, unsigned long len);
796 struct kvm_memory_slot *gfn_to_memslot(struct kvm *kvm, gfn_t gfn);
797 bool kvm_is_visible_gfn(struct kvm *kvm, gfn_t gfn);
798 bool kvm_vcpu_is_visible_gfn(struct kvm_vcpu *vcpu, gfn_t gfn);
799 unsigned long kvm_host_page_size(struct kvm_vcpu *vcpu, gfn_t gfn);
800 void mark_page_dirty_in_slot(struct kvm *kvm, struct kvm_memory_slot *memslot, gfn_t gfn);
801 void mark_page_dirty(struct kvm *kvm, gfn_t gfn);
803 struct kvm_memslots *kvm_vcpu_memslots(struct kvm_vcpu *vcpu);
804 struct kvm_memory_slot *kvm_vcpu_gfn_to_memslot(struct kvm_vcpu *vcpu, gfn_t gfn);
805 kvm_pfn_t kvm_vcpu_gfn_to_pfn_atomic(struct kvm_vcpu *vcpu, gfn_t gfn);
806 kvm_pfn_t kvm_vcpu_gfn_to_pfn(struct kvm_vcpu *vcpu, gfn_t gfn);
807 int kvm_vcpu_map(struct kvm_vcpu *vcpu, gpa_t gpa, struct kvm_host_map *map);
808 int kvm_map_gfn(struct kvm_vcpu *vcpu, gfn_t gfn, struct kvm_host_map *map,
809 struct gfn_to_pfn_cache *cache, bool atomic);
810 struct page *kvm_vcpu_gfn_to_page(struct kvm_vcpu *vcpu, gfn_t gfn);
811 void kvm_vcpu_unmap(struct kvm_vcpu *vcpu, struct kvm_host_map *map, bool dirty);
812 int kvm_unmap_gfn(struct kvm_vcpu *vcpu, struct kvm_host_map *map,
813 struct gfn_to_pfn_cache *cache, bool dirty, bool atomic);
814 unsigned long kvm_vcpu_gfn_to_hva(struct kvm_vcpu *vcpu, gfn_t gfn);
815 unsigned long kvm_vcpu_gfn_to_hva_prot(struct kvm_vcpu *vcpu, gfn_t gfn, bool *writable);
816 int kvm_vcpu_read_guest_page(struct kvm_vcpu *vcpu, gfn_t gfn, void *data, int offset,
818 int kvm_vcpu_read_guest_atomic(struct kvm_vcpu *vcpu, gpa_t gpa, void *data,
820 int kvm_vcpu_read_guest(struct kvm_vcpu *vcpu, gpa_t gpa, void *data,
822 int kvm_vcpu_write_guest_page(struct kvm_vcpu *vcpu, gfn_t gfn, const void *data,
823 int offset, int len);
824 int kvm_vcpu_write_guest(struct kvm_vcpu *vcpu, gpa_t gpa, const void *data,
826 void kvm_vcpu_mark_page_dirty(struct kvm_vcpu *vcpu, gfn_t gfn);
828 void kvm_sigset_activate(struct kvm_vcpu *vcpu);
829 void kvm_sigset_deactivate(struct kvm_vcpu *vcpu);
831 void kvm_vcpu_block(struct kvm_vcpu *vcpu);
832 void kvm_arch_vcpu_blocking(struct kvm_vcpu *vcpu);
833 void kvm_arch_vcpu_unblocking(struct kvm_vcpu *vcpu);
834 bool kvm_vcpu_wake_up(struct kvm_vcpu *vcpu);
835 void kvm_vcpu_kick(struct kvm_vcpu *vcpu);
836 int kvm_vcpu_yield_to(struct kvm_vcpu *target);
837 void kvm_vcpu_on_spin(struct kvm_vcpu *vcpu, bool usermode_vcpu_not_eligible);
839 void kvm_flush_remote_tlbs(struct kvm *kvm);
840 void kvm_reload_remote_mmus(struct kvm *kvm);
842 #ifdef KVM_ARCH_NR_OBJS_PER_MEMORY_CACHE
843 int kvm_mmu_topup_memory_cache(struct kvm_mmu_memory_cache *mc, int min);
844 int kvm_mmu_memory_cache_nr_free_objects(struct kvm_mmu_memory_cache *mc);
845 void kvm_mmu_free_memory_cache(struct kvm_mmu_memory_cache *mc);
846 void *kvm_mmu_memory_cache_alloc(struct kvm_mmu_memory_cache *mc);
849 bool kvm_make_vcpus_request_mask(struct kvm *kvm, unsigned int req,
850 struct kvm_vcpu *except,
851 unsigned long *vcpu_bitmap, cpumask_var_t tmp);
852 bool kvm_make_all_cpus_request(struct kvm *kvm, unsigned int req);
853 bool kvm_make_all_cpus_request_except(struct kvm *kvm, unsigned int req,
854 struct kvm_vcpu *except);
855 bool kvm_make_cpus_request_mask(struct kvm *kvm, unsigned int req,
856 unsigned long *vcpu_bitmap);
858 long kvm_arch_dev_ioctl(struct file *filp,
859 unsigned int ioctl, unsigned long arg);
860 long kvm_arch_vcpu_ioctl(struct file *filp,
861 unsigned int ioctl, unsigned long arg);
862 vm_fault_t kvm_arch_vcpu_fault(struct kvm_vcpu *vcpu, struct vm_fault *vmf);
864 int kvm_vm_ioctl_check_extension(struct kvm *kvm, long ext);
866 void kvm_arch_mmu_enable_log_dirty_pt_masked(struct kvm *kvm,
867 struct kvm_memory_slot *slot,
870 void kvm_arch_sync_dirty_log(struct kvm *kvm, struct kvm_memory_slot *memslot);
872 #ifdef CONFIG_KVM_GENERIC_DIRTYLOG_READ_PROTECT
873 void kvm_arch_flush_remote_tlbs_memslot(struct kvm *kvm,
874 struct kvm_memory_slot *memslot);
875 #else /* !CONFIG_KVM_GENERIC_DIRTYLOG_READ_PROTECT */
876 int kvm_vm_ioctl_get_dirty_log(struct kvm *kvm, struct kvm_dirty_log *log);
877 int kvm_get_dirty_log(struct kvm *kvm, struct kvm_dirty_log *log,
878 int *is_dirty, struct kvm_memory_slot **memslot);
881 int kvm_vm_ioctl_irq_line(struct kvm *kvm, struct kvm_irq_level *irq_level,
883 int kvm_vm_ioctl_enable_cap(struct kvm *kvm,
884 struct kvm_enable_cap *cap);
885 long kvm_arch_vm_ioctl(struct file *filp,
886 unsigned int ioctl, unsigned long arg);
888 int kvm_arch_vcpu_ioctl_get_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu);
889 int kvm_arch_vcpu_ioctl_set_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu);
891 int kvm_arch_vcpu_ioctl_translate(struct kvm_vcpu *vcpu,
892 struct kvm_translation *tr);
894 int kvm_arch_vcpu_ioctl_get_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs);
895 int kvm_arch_vcpu_ioctl_set_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs);
896 int kvm_arch_vcpu_ioctl_get_sregs(struct kvm_vcpu *vcpu,
897 struct kvm_sregs *sregs);
898 int kvm_arch_vcpu_ioctl_set_sregs(struct kvm_vcpu *vcpu,
899 struct kvm_sregs *sregs);
900 int kvm_arch_vcpu_ioctl_get_mpstate(struct kvm_vcpu *vcpu,
901 struct kvm_mp_state *mp_state);
902 int kvm_arch_vcpu_ioctl_set_mpstate(struct kvm_vcpu *vcpu,
903 struct kvm_mp_state *mp_state);
904 int kvm_arch_vcpu_ioctl_set_guest_debug(struct kvm_vcpu *vcpu,
905 struct kvm_guest_debug *dbg);
906 int kvm_arch_vcpu_ioctl_run(struct kvm_vcpu *vcpu);
908 int kvm_arch_init(void *opaque);
909 void kvm_arch_exit(void);
911 void kvm_arch_sched_in(struct kvm_vcpu *vcpu, int cpu);
913 void kvm_arch_vcpu_load(struct kvm_vcpu *vcpu, int cpu);
914 void kvm_arch_vcpu_put(struct kvm_vcpu *vcpu);
915 int kvm_arch_vcpu_precreate(struct kvm *kvm, unsigned int id);
916 int kvm_arch_vcpu_create(struct kvm_vcpu *vcpu);
917 void kvm_arch_vcpu_postcreate(struct kvm_vcpu *vcpu);
918 void kvm_arch_vcpu_destroy(struct kvm_vcpu *vcpu);
920 #ifdef __KVM_HAVE_ARCH_VCPU_DEBUGFS
921 void kvm_arch_create_vcpu_debugfs(struct kvm_vcpu *vcpu, struct dentry *debugfs_dentry);
924 int kvm_arch_hardware_enable(void);
925 void kvm_arch_hardware_disable(void);
926 int kvm_arch_hardware_setup(void *opaque);
927 void kvm_arch_hardware_unsetup(void);
928 int kvm_arch_check_processor_compat(void *opaque);
929 int kvm_arch_vcpu_runnable(struct kvm_vcpu *vcpu);
930 bool kvm_arch_vcpu_in_kernel(struct kvm_vcpu *vcpu);
931 int kvm_arch_vcpu_should_kick(struct kvm_vcpu *vcpu);
932 bool kvm_arch_dy_runnable(struct kvm_vcpu *vcpu);
933 int kvm_arch_post_init_vm(struct kvm *kvm);
934 void kvm_arch_pre_destroy_vm(struct kvm *kvm);
936 #ifndef __KVM_HAVE_ARCH_VM_ALLOC
938 * All architectures that want to use vzalloc currently also
939 * need their own kvm_arch_alloc_vm implementation.
941 static inline struct kvm *kvm_arch_alloc_vm(void)
943 return kzalloc(sizeof(struct kvm), GFP_KERNEL);
946 static inline void kvm_arch_free_vm(struct kvm *kvm)
952 #ifndef __KVM_HAVE_ARCH_FLUSH_REMOTE_TLB
953 static inline int kvm_arch_flush_remote_tlb(struct kvm *kvm)
959 #ifdef __KVM_HAVE_ARCH_NONCOHERENT_DMA
960 void kvm_arch_register_noncoherent_dma(struct kvm *kvm);
961 void kvm_arch_unregister_noncoherent_dma(struct kvm *kvm);
962 bool kvm_arch_has_noncoherent_dma(struct kvm *kvm);
964 static inline void kvm_arch_register_noncoherent_dma(struct kvm *kvm)
968 static inline void kvm_arch_unregister_noncoherent_dma(struct kvm *kvm)
972 static inline bool kvm_arch_has_noncoherent_dma(struct kvm *kvm)
977 #ifdef __KVM_HAVE_ARCH_ASSIGNED_DEVICE
978 void kvm_arch_start_assignment(struct kvm *kvm);
979 void kvm_arch_end_assignment(struct kvm *kvm);
980 bool kvm_arch_has_assigned_device(struct kvm *kvm);
982 static inline void kvm_arch_start_assignment(struct kvm *kvm)
986 static inline void kvm_arch_end_assignment(struct kvm *kvm)
990 static inline bool kvm_arch_has_assigned_device(struct kvm *kvm)
996 static inline struct rcuwait *kvm_arch_vcpu_get_wait(struct kvm_vcpu *vcpu)
998 #ifdef __KVM_HAVE_ARCH_WQP
999 return vcpu->arch.waitp;
1005 #ifdef __KVM_HAVE_ARCH_INTC_INITIALIZED
1007 * returns true if the virtual interrupt controller is initialized and
1008 * ready to accept virtual IRQ. On some architectures the virtual interrupt
1009 * controller is dynamically instantiated and this is not always true.
1011 bool kvm_arch_intc_initialized(struct kvm *kvm);
1013 static inline bool kvm_arch_intc_initialized(struct kvm *kvm)
1019 int kvm_arch_init_vm(struct kvm *kvm, unsigned long type);
1020 void kvm_arch_destroy_vm(struct kvm *kvm);
1021 void kvm_arch_sync_events(struct kvm *kvm);
1023 int kvm_cpu_has_pending_timer(struct kvm_vcpu *vcpu);
1025 bool kvm_is_reserved_pfn(kvm_pfn_t pfn);
1026 bool kvm_is_zone_device_pfn(kvm_pfn_t pfn);
1027 bool kvm_is_transparent_hugepage(kvm_pfn_t pfn);
1029 struct kvm_irq_ack_notifier {
1030 struct hlist_node link;
1032 void (*irq_acked)(struct kvm_irq_ack_notifier *kian);
1035 int kvm_irq_map_gsi(struct kvm *kvm,
1036 struct kvm_kernel_irq_routing_entry *entries, int gsi);
1037 int kvm_irq_map_chip_pin(struct kvm *kvm, unsigned irqchip, unsigned pin);
1039 int kvm_set_irq(struct kvm *kvm, int irq_source_id, u32 irq, int level,
1041 int kvm_set_msi(struct kvm_kernel_irq_routing_entry *irq_entry, struct kvm *kvm,
1042 int irq_source_id, int level, bool line_status);
1043 int kvm_arch_set_irq_inatomic(struct kvm_kernel_irq_routing_entry *e,
1044 struct kvm *kvm, int irq_source_id,
1045 int level, bool line_status);
1046 bool kvm_irq_has_notifier(struct kvm *kvm, unsigned irqchip, unsigned pin);
1047 void kvm_notify_acked_gsi(struct kvm *kvm, int gsi);
1048 void kvm_notify_acked_irq(struct kvm *kvm, unsigned irqchip, unsigned pin);
1049 void kvm_register_irq_ack_notifier(struct kvm *kvm,
1050 struct kvm_irq_ack_notifier *kian);
1051 void kvm_unregister_irq_ack_notifier(struct kvm *kvm,
1052 struct kvm_irq_ack_notifier *kian);
1053 int kvm_request_irq_source_id(struct kvm *kvm);
1054 void kvm_free_irq_source_id(struct kvm *kvm, int irq_source_id);
1055 bool kvm_arch_irqfd_allowed(struct kvm *kvm, struct kvm_irqfd *args);
1058 * search_memslots() and __gfn_to_memslot() are here because they are
1059 * used in non-modular code in arch/powerpc/kvm/book3s_hv_rm_mmu.c.
1060 * gfn_to_memslot() itself isn't here as an inline because that would
1061 * bloat other code too much.
1063 * IMPORTANT: Slots are sorted from highest GFN to lowest GFN!
1065 static inline struct kvm_memory_slot *
1066 search_memslots(struct kvm_memslots *slots, gfn_t gfn)
1068 int start = 0, end = slots->used_slots;
1069 int slot = atomic_read(&slots->lru_slot);
1070 struct kvm_memory_slot *memslots = slots->memslots;
1072 if (unlikely(!slots->used_slots))
1075 if (gfn >= memslots[slot].base_gfn &&
1076 gfn < memslots[slot].base_gfn + memslots[slot].npages)
1077 return &memslots[slot];
1079 while (start < end) {
1080 slot = start + (end - start) / 2;
1082 if (gfn >= memslots[slot].base_gfn)
1088 if (start < slots->used_slots && gfn >= memslots[start].base_gfn &&
1089 gfn < memslots[start].base_gfn + memslots[start].npages) {
1090 atomic_set(&slots->lru_slot, start);
1091 return &memslots[start];
1097 static inline struct kvm_memory_slot *
1098 __gfn_to_memslot(struct kvm_memslots *slots, gfn_t gfn)
1100 return search_memslots(slots, gfn);
1103 static inline unsigned long
1104 __gfn_to_hva_memslot(struct kvm_memory_slot *slot, gfn_t gfn)
1106 return slot->userspace_addr + (gfn - slot->base_gfn) * PAGE_SIZE;
1109 static inline int memslot_id(struct kvm *kvm, gfn_t gfn)
1111 return gfn_to_memslot(kvm, gfn)->id;
1115 hva_to_gfn_memslot(unsigned long hva, struct kvm_memory_slot *slot)
1117 gfn_t gfn_offset = (hva - slot->userspace_addr) >> PAGE_SHIFT;
1119 return slot->base_gfn + gfn_offset;
1122 static inline gpa_t gfn_to_gpa(gfn_t gfn)
1124 return (gpa_t)gfn << PAGE_SHIFT;
1127 static inline gfn_t gpa_to_gfn(gpa_t gpa)
1129 return (gfn_t)(gpa >> PAGE_SHIFT);
1132 static inline hpa_t pfn_to_hpa(kvm_pfn_t pfn)
1134 return (hpa_t)pfn << PAGE_SHIFT;
1137 static inline struct page *kvm_vcpu_gpa_to_page(struct kvm_vcpu *vcpu,
1140 return kvm_vcpu_gfn_to_page(vcpu, gpa_to_gfn(gpa));
1143 static inline bool kvm_is_error_gpa(struct kvm *kvm, gpa_t gpa)
1145 unsigned long hva = gfn_to_hva(kvm, gpa_to_gfn(gpa));
1147 return kvm_is_error_hva(hva);
1150 enum kvm_stat_kind {
1155 struct kvm_stat_data {
1157 struct kvm_stats_debugfs_item *dbgfs_item;
1160 struct kvm_stats_debugfs_item {
1163 enum kvm_stat_kind kind;
1167 #define KVM_DBGFS_GET_MODE(dbgfs_item) \
1168 ((dbgfs_item)->mode ? (dbgfs_item)->mode : 0644)
1170 #define VM_STAT(n, x, ...) \
1171 { n, offsetof(struct kvm, stat.x), KVM_STAT_VM, ## __VA_ARGS__ }
1172 #define VCPU_STAT(n, x, ...) \
1173 { n, offsetof(struct kvm_vcpu, stat.x), KVM_STAT_VCPU, ## __VA_ARGS__ }
1175 extern struct kvm_stats_debugfs_item debugfs_entries[];
1176 extern struct dentry *kvm_debugfs_dir;
1178 #if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
1179 static inline int mmu_notifier_retry(struct kvm *kvm, unsigned long mmu_seq)
1181 if (unlikely(kvm->mmu_notifier_count))
1184 * Ensure the read of mmu_notifier_count happens before the read
1185 * of mmu_notifier_seq. This interacts with the smp_wmb() in
1186 * mmu_notifier_invalidate_range_end to make sure that the caller
1187 * either sees the old (non-zero) value of mmu_notifier_count or
1188 * the new (incremented) value of mmu_notifier_seq.
1189 * PowerPC Book3s HV KVM calls this under a per-page lock
1190 * rather than under kvm->mmu_lock, for scalability, so
1191 * can't rely on kvm->mmu_lock to keep things ordered.
1194 if (kvm->mmu_notifier_seq != mmu_seq)
1200 #ifdef CONFIG_HAVE_KVM_IRQ_ROUTING
1202 #define KVM_MAX_IRQ_ROUTES 4096 /* might need extension/rework in the future */
1204 bool kvm_arch_can_set_irq_routing(struct kvm *kvm);
1205 int kvm_set_irq_routing(struct kvm *kvm,
1206 const struct kvm_irq_routing_entry *entries,
1209 int kvm_set_routing_entry(struct kvm *kvm,
1210 struct kvm_kernel_irq_routing_entry *e,
1211 const struct kvm_irq_routing_entry *ue);
1212 void kvm_free_irq_routing(struct kvm *kvm);
1216 static inline void kvm_free_irq_routing(struct kvm *kvm) {}
1220 int kvm_send_userspace_msi(struct kvm *kvm, struct kvm_msi *msi);
1222 #ifdef CONFIG_HAVE_KVM_EVENTFD
1224 void kvm_eventfd_init(struct kvm *kvm);
1225 int kvm_ioeventfd(struct kvm *kvm, struct kvm_ioeventfd *args);
1227 #ifdef CONFIG_HAVE_KVM_IRQFD
1228 int kvm_irqfd(struct kvm *kvm, struct kvm_irqfd *args);
1229 void kvm_irqfd_release(struct kvm *kvm);
1230 void kvm_irq_routing_update(struct kvm *);
1232 static inline int kvm_irqfd(struct kvm *kvm, struct kvm_irqfd *args)
1237 static inline void kvm_irqfd_release(struct kvm *kvm) {}
1242 static inline void kvm_eventfd_init(struct kvm *kvm) {}
1244 static inline int kvm_irqfd(struct kvm *kvm, struct kvm_irqfd *args)
1249 static inline void kvm_irqfd_release(struct kvm *kvm) {}
1251 #ifdef CONFIG_HAVE_KVM_IRQCHIP
1252 static inline void kvm_irq_routing_update(struct kvm *kvm)
1257 static inline int kvm_ioeventfd(struct kvm *kvm, struct kvm_ioeventfd *args)
1262 #endif /* CONFIG_HAVE_KVM_EVENTFD */
1264 void kvm_arch_irq_routing_update(struct kvm *kvm);
1266 static inline void kvm_make_request(int req, struct kvm_vcpu *vcpu)
1269 * Ensure the rest of the request is published to kvm_check_request's
1270 * caller. Paired with the smp_mb__after_atomic in kvm_check_request.
1273 set_bit(req & KVM_REQUEST_MASK, (void *)&vcpu->requests);
1276 static inline bool kvm_request_pending(struct kvm_vcpu *vcpu)
1278 return READ_ONCE(vcpu->requests);
1281 static inline bool kvm_test_request(int req, struct kvm_vcpu *vcpu)
1283 return test_bit(req & KVM_REQUEST_MASK, (void *)&vcpu->requests);
1286 static inline void kvm_clear_request(int req, struct kvm_vcpu *vcpu)
1288 clear_bit(req & KVM_REQUEST_MASK, (void *)&vcpu->requests);
1291 static inline bool kvm_check_request(int req, struct kvm_vcpu *vcpu)
1293 if (kvm_test_request(req, vcpu)) {
1294 kvm_clear_request(req, vcpu);
1297 * Ensure the rest of the request is visible to kvm_check_request's
1298 * caller. Paired with the smp_wmb in kvm_make_request.
1300 smp_mb__after_atomic();
1307 extern bool kvm_rebooting;
1309 extern unsigned int halt_poll_ns;
1310 extern unsigned int halt_poll_ns_grow;
1311 extern unsigned int halt_poll_ns_grow_start;
1312 extern unsigned int halt_poll_ns_shrink;
1315 const struct kvm_device_ops *ops;
1318 struct list_head vm_node;
1321 /* create, destroy, and name are mandatory */
1322 struct kvm_device_ops {
1326 * create is called holding kvm->lock and any operations not suitable
1327 * to do while holding the lock should be deferred to init (see
1330 int (*create)(struct kvm_device *dev, u32 type);
1333 * init is called after create if create is successful and is called
1334 * outside of holding kvm->lock.
1336 void (*init)(struct kvm_device *dev);
1339 * Destroy is responsible for freeing dev.
1341 * Destroy may be called before or after destructors are called
1342 * on emulated I/O regions, depending on whether a reference is
1343 * held by a vcpu or other kvm component that gets destroyed
1344 * after the emulated I/O.
1346 void (*destroy)(struct kvm_device *dev);
1349 * Release is an alternative method to free the device. It is
1350 * called when the device file descriptor is closed. Once
1351 * release is called, the destroy method will not be called
1352 * anymore as the device is removed from the device list of
1353 * the VM. kvm->lock is held.
1355 void (*release)(struct kvm_device *dev);
1357 int (*set_attr)(struct kvm_device *dev, struct kvm_device_attr *attr);
1358 int (*get_attr)(struct kvm_device *dev, struct kvm_device_attr *attr);
1359 int (*has_attr)(struct kvm_device *dev, struct kvm_device_attr *attr);
1360 long (*ioctl)(struct kvm_device *dev, unsigned int ioctl,
1362 int (*mmap)(struct kvm_device *dev, struct vm_area_struct *vma);
1365 void kvm_device_get(struct kvm_device *dev);
1366 void kvm_device_put(struct kvm_device *dev);
1367 struct kvm_device *kvm_device_from_filp(struct file *filp);
1368 int kvm_register_device_ops(const struct kvm_device_ops *ops, u32 type);
1369 void kvm_unregister_device_ops(u32 type);
1371 extern struct kvm_device_ops kvm_mpic_ops;
1372 extern struct kvm_device_ops kvm_arm_vgic_v2_ops;
1373 extern struct kvm_device_ops kvm_arm_vgic_v3_ops;
1375 #ifdef CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT
1377 static inline void kvm_vcpu_set_in_spin_loop(struct kvm_vcpu *vcpu, bool val)
1379 vcpu->spin_loop.in_spin_loop = val;
1381 static inline void kvm_vcpu_set_dy_eligible(struct kvm_vcpu *vcpu, bool val)
1383 vcpu->spin_loop.dy_eligible = val;
1386 #else /* !CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT */
1388 static inline void kvm_vcpu_set_in_spin_loop(struct kvm_vcpu *vcpu, bool val)
1392 static inline void kvm_vcpu_set_dy_eligible(struct kvm_vcpu *vcpu, bool val)
1395 #endif /* CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT */
1397 static inline bool kvm_is_visible_memslot(struct kvm_memory_slot *memslot)
1399 return (memslot && memslot->id < KVM_USER_MEM_SLOTS &&
1400 !(memslot->flags & KVM_MEMSLOT_INVALID));
1403 struct kvm_vcpu *kvm_get_running_vcpu(void);
1404 struct kvm_vcpu * __percpu *kvm_get_running_vcpus(void);
1406 #ifdef CONFIG_HAVE_KVM_IRQ_BYPASS
1407 bool kvm_arch_has_irq_bypass(void);
1408 int kvm_arch_irq_bypass_add_producer(struct irq_bypass_consumer *,
1409 struct irq_bypass_producer *);
1410 void kvm_arch_irq_bypass_del_producer(struct irq_bypass_consumer *,
1411 struct irq_bypass_producer *);
1412 void kvm_arch_irq_bypass_stop(struct irq_bypass_consumer *);
1413 void kvm_arch_irq_bypass_start(struct irq_bypass_consumer *);
1414 int kvm_arch_update_irqfd_routing(struct kvm *kvm, unsigned int host_irq,
1415 uint32_t guest_irq, bool set);
1416 #endif /* CONFIG_HAVE_KVM_IRQ_BYPASS */
1418 #ifdef CONFIG_HAVE_KVM_INVALID_WAKEUPS
1419 /* If we wakeup during the poll time, was it a sucessful poll? */
1420 static inline bool vcpu_valid_wakeup(struct kvm_vcpu *vcpu)
1422 return vcpu->valid_wakeup;
1426 static inline bool vcpu_valid_wakeup(struct kvm_vcpu *vcpu)
1430 #endif /* CONFIG_HAVE_KVM_INVALID_WAKEUPS */
1432 #ifdef CONFIG_HAVE_KVM_NO_POLL
1433 /* Callback that tells if we must not poll */
1434 bool kvm_arch_no_poll(struct kvm_vcpu *vcpu);
1436 static inline bool kvm_arch_no_poll(struct kvm_vcpu *vcpu)
1440 #endif /* CONFIG_HAVE_KVM_NO_POLL */
1442 #ifdef CONFIG_HAVE_KVM_VCPU_ASYNC_IOCTL
1443 long kvm_arch_vcpu_async_ioctl(struct file *filp,
1444 unsigned int ioctl, unsigned long arg);
1446 static inline long kvm_arch_vcpu_async_ioctl(struct file *filp,
1450 return -ENOIOCTLCMD;
1452 #endif /* CONFIG_HAVE_KVM_VCPU_ASYNC_IOCTL */
1454 void kvm_arch_mmu_notifier_invalidate_range(struct kvm *kvm,
1455 unsigned long start, unsigned long end);
1457 #ifdef CONFIG_HAVE_KVM_VCPU_RUN_PID_CHANGE
1458 int kvm_arch_vcpu_run_pid_change(struct kvm_vcpu *vcpu);
1460 static inline int kvm_arch_vcpu_run_pid_change(struct kvm_vcpu *vcpu)
1464 #endif /* CONFIG_HAVE_KVM_VCPU_RUN_PID_CHANGE */
1466 typedef int (*kvm_vm_thread_fn_t)(struct kvm *kvm, uintptr_t data);
1468 int kvm_vm_create_worker_thread(struct kvm *kvm, kvm_vm_thread_fn_t thread_fn,
1469 uintptr_t data, const char *name,
1470 struct task_struct **thread_ptr);
1472 #ifdef CONFIG_KVM_XFER_TO_GUEST_WORK
1473 static inline void kvm_handle_signal_exit(struct kvm_vcpu *vcpu)
1475 vcpu->run->exit_reason = KVM_EXIT_INTR;
1476 vcpu->stat.signal_exits++;
1478 #endif /* CONFIG_KVM_XFER_TO_GUEST_WORK */