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_REQ_VM_BUGGED (4 | KVM_REQUEST_WAIT | KVM_REQUEST_NO_WAKEUP)
150 #define KVM_REQUEST_ARCH_BASE 8
152 #define KVM_ARCH_REQ_FLAGS(nr, flags) ({ \
153 BUILD_BUG_ON((unsigned)(nr) >= (sizeof_field(struct kvm_vcpu, requests) * 8) - KVM_REQUEST_ARCH_BASE); \
154 (unsigned)(((nr) + KVM_REQUEST_ARCH_BASE) | (flags)); \
156 #define KVM_ARCH_REQ(nr) KVM_ARCH_REQ_FLAGS(nr, 0)
158 #define KVM_USERSPACE_IRQ_SOURCE_ID 0
159 #define KVM_IRQFD_RESAMPLE_IRQ_SOURCE_ID 1
161 extern struct mutex kvm_lock;
162 extern struct list_head vm_list;
164 struct kvm_io_range {
167 struct kvm_io_device *dev;
170 #define NR_IOBUS_DEVS 1000
175 struct kvm_io_range range[];
181 KVM_VIRTIO_CCW_NOTIFY_BUS,
186 int kvm_io_bus_write(struct kvm_vcpu *vcpu, enum kvm_bus bus_idx, gpa_t addr,
187 int len, const void *val);
188 int kvm_io_bus_write_cookie(struct kvm_vcpu *vcpu, enum kvm_bus bus_idx,
189 gpa_t addr, int len, const void *val, long cookie);
190 int kvm_io_bus_read(struct kvm_vcpu *vcpu, enum kvm_bus bus_idx, gpa_t addr,
192 int kvm_io_bus_register_dev(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr,
193 int len, struct kvm_io_device *dev);
194 int kvm_io_bus_unregister_dev(struct kvm *kvm, enum kvm_bus bus_idx,
195 struct kvm_io_device *dev);
196 struct kvm_io_device *kvm_io_bus_get_dev(struct kvm *kvm, enum kvm_bus bus_idx,
199 #ifdef CONFIG_KVM_ASYNC_PF
200 struct kvm_async_pf {
201 struct work_struct work;
202 struct list_head link;
203 struct list_head queue;
204 struct kvm_vcpu *vcpu;
205 struct mm_struct *mm;
208 struct kvm_arch_async_pf arch;
210 bool notpresent_injected;
213 void kvm_clear_async_pf_completion_queue(struct kvm_vcpu *vcpu);
214 void kvm_check_async_pf_completion(struct kvm_vcpu *vcpu);
215 bool kvm_setup_async_pf(struct kvm_vcpu *vcpu, gpa_t cr2_or_gpa,
216 unsigned long hva, struct kvm_arch_async_pf *arch);
217 int kvm_async_pf_wakeup_all(struct kvm_vcpu *vcpu);
224 READING_SHADOW_PAGE_TABLES,
227 #define KVM_UNMAPPED_PAGE ((void *) 0x500 + POISON_POINTER_DELTA)
229 struct kvm_host_map {
231 * Only valid if the 'pfn' is managed by the host kernel (i.e. There is
232 * a 'struct page' for it. When using mem= kernel parameter some memory
233 * can be used as guest memory but they are not managed by host
235 * If 'pfn' is not managed by the host kernel, this field is
236 * initialized to KVM_UNMAPPED_PAGE.
245 * Used to check if the mapping is valid or not. Never use 'kvm_host_map'
246 * directly to check for that.
248 static inline bool kvm_vcpu_mapped(struct kvm_host_map *map)
254 * Sometimes a large or cross-page mmio needs to be broken up into separate
255 * exits for userspace servicing.
257 struct kvm_mmio_fragment {
265 #ifdef CONFIG_PREEMPT_NOTIFIERS
266 struct preempt_notifier preempt_notifier;
269 int vcpu_id; /* id given by userspace at creation */
270 int vcpu_idx; /* index in kvm->vcpus array */
274 unsigned long guest_debug;
277 struct list_head blocked_vcpu_list;
283 struct pid __rcu *pid;
286 struct kvm_vcpu_stat stat;
287 unsigned int halt_poll_ns;
290 #ifdef CONFIG_HAS_IOMEM
292 int mmio_read_completed;
294 int mmio_cur_fragment;
295 int mmio_nr_fragments;
296 struct kvm_mmio_fragment mmio_fragments[KVM_MAX_MMIO_FRAGMENTS];
299 #ifdef CONFIG_KVM_ASYNC_PF
302 struct list_head queue;
303 struct list_head done;
308 #ifdef CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT
310 * Cpu relax intercept or pause loop exit optimization
311 * in_spin_loop: set when a vcpu does a pause loop exit
312 * or cpu relax intercepted.
313 * dy_eligible: indicates whether vcpu is eligible for directed yield.
322 struct kvm_vcpu_arch arch;
325 static inline int kvm_vcpu_exiting_guest_mode(struct kvm_vcpu *vcpu)
328 * The memory barrier ensures a previous write to vcpu->requests cannot
329 * be reordered with the read of vcpu->mode. It pairs with the general
330 * memory barrier following the write of vcpu->mode in VCPU RUN.
332 smp_mb__before_atomic();
333 return cmpxchg(&vcpu->mode, IN_GUEST_MODE, EXITING_GUEST_MODE);
337 * Some of the bitops functions do not support too long bitmaps.
338 * This number must be determined not to exceed such limits.
340 #define KVM_MEM_MAX_NR_PAGES ((1UL << 31) - 1)
342 struct kvm_memory_slot {
344 unsigned long npages;
345 unsigned long *dirty_bitmap;
346 struct kvm_arch_memory_slot arch;
347 unsigned long userspace_addr;
353 static inline unsigned long kvm_dirty_bitmap_bytes(struct kvm_memory_slot *memslot)
355 return ALIGN(memslot->npages, BITS_PER_LONG) / 8;
358 static inline unsigned long *kvm_second_dirty_bitmap(struct kvm_memory_slot *memslot)
360 unsigned long len = kvm_dirty_bitmap_bytes(memslot);
362 return memslot->dirty_bitmap + len / sizeof(*memslot->dirty_bitmap);
365 #ifndef KVM_DIRTY_LOG_MANUAL_CAPS
366 #define KVM_DIRTY_LOG_MANUAL_CAPS KVM_DIRTY_LOG_MANUAL_PROTECT_ENABLE
369 struct kvm_s390_adapter_int {
382 struct kvm_kernel_irq_routing_entry {
385 int (*set)(struct kvm_kernel_irq_routing_entry *e,
386 struct kvm *kvm, int irq_source_id, int level,
400 struct kvm_s390_adapter_int adapter;
401 struct kvm_hv_sint hv_sint;
403 struct hlist_node link;
406 #ifdef CONFIG_HAVE_KVM_IRQ_ROUTING
407 struct kvm_irq_routing_table {
408 int chip[KVM_NR_IRQCHIPS][KVM_IRQCHIP_NUM_PINS];
411 * Array indexed by gsi. Each entry contains list of irq chips
412 * the gsi is connected to.
414 struct hlist_head map[];
418 #ifndef KVM_PRIVATE_MEM_SLOTS
419 #define KVM_PRIVATE_MEM_SLOTS 0
422 #ifndef KVM_MEM_SLOTS_NUM
423 #define KVM_MEM_SLOTS_NUM (KVM_USER_MEM_SLOTS + KVM_PRIVATE_MEM_SLOTS)
426 #ifndef __KVM_VCPU_MULTIPLE_ADDRESS_SPACE
427 static inline int kvm_arch_vcpu_memslots_id(struct kvm_vcpu *vcpu)
435 * memslots are not sorted by id anymore, please use id_to_memslot()
436 * to get the memslot by its id.
438 struct kvm_memslots {
440 /* The mapping table from slot id to the index in memslots[]. */
441 short id_to_index[KVM_MEM_SLOTS_NUM];
444 struct kvm_memory_slot memslots[];
449 struct mutex slots_lock;
450 struct mm_struct *mm; /* userspace tied to this vm */
451 struct kvm_memslots __rcu *memslots[KVM_ADDRESS_SPACE_NUM];
452 struct kvm_vcpu *vcpus[KVM_MAX_VCPUS];
455 * created_vcpus is protected by kvm->lock, and is incremented
456 * at the beginning of KVM_CREATE_VCPU. online_vcpus is only
457 * incremented after storing the kvm_vcpu pointer in vcpus,
458 * and is accessed atomically.
460 atomic_t online_vcpus;
462 int last_boosted_vcpu;
463 struct list_head vm_list;
465 struct kvm_io_bus __rcu *buses[KVM_NR_BUSES];
466 #ifdef CONFIG_HAVE_KVM_EVENTFD
469 struct list_head items;
470 struct list_head resampler_list;
471 struct mutex resampler_lock;
473 struct list_head ioeventfds;
475 struct kvm_vm_stat stat;
476 struct kvm_arch arch;
477 refcount_t users_count;
478 #ifdef CONFIG_KVM_MMIO
479 struct kvm_coalesced_mmio_ring *coalesced_mmio_ring;
480 spinlock_t ring_lock;
481 struct list_head coalesced_zones;
484 struct mutex irq_lock;
485 #ifdef CONFIG_HAVE_KVM_IRQCHIP
487 * Update side is protected by irq_lock.
489 struct kvm_irq_routing_table __rcu *irq_routing;
491 #ifdef CONFIG_HAVE_KVM_IRQFD
492 struct hlist_head irq_ack_notifier_list;
495 #if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
496 struct mmu_notifier mmu_notifier;
497 unsigned long mmu_notifier_seq;
498 long mmu_notifier_count;
501 struct list_head devices;
502 u64 manual_dirty_log_protect;
503 struct dentry *debugfs_dentry;
504 struct kvm_stat_data **debugfs_stat_data;
505 struct srcu_struct srcu;
506 struct srcu_struct irq_srcu;
508 unsigned int max_halt_poll_ns;
512 #define kvm_err(fmt, ...) \
513 pr_err("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
514 #define kvm_info(fmt, ...) \
515 pr_info("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
516 #define kvm_debug(fmt, ...) \
517 pr_debug("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
518 #define kvm_debug_ratelimited(fmt, ...) \
519 pr_debug_ratelimited("kvm [%i]: " fmt, task_pid_nr(current), \
521 #define kvm_pr_unimpl(fmt, ...) \
522 pr_err_ratelimited("kvm [%i]: " fmt, \
523 task_tgid_nr(current), ## __VA_ARGS__)
525 /* The guest did something we don't support. */
526 #define vcpu_unimpl(vcpu, fmt, ...) \
527 kvm_pr_unimpl("vcpu%i, guest rIP: 0x%lx " fmt, \
528 (vcpu)->vcpu_id, kvm_rip_read(vcpu), ## __VA_ARGS__)
530 #define vcpu_debug(vcpu, fmt, ...) \
531 kvm_debug("vcpu%i " fmt, (vcpu)->vcpu_id, ## __VA_ARGS__)
532 #define vcpu_debug_ratelimited(vcpu, fmt, ...) \
533 kvm_debug_ratelimited("vcpu%i " fmt, (vcpu)->vcpu_id, \
535 #define vcpu_err(vcpu, fmt, ...) \
536 kvm_err("vcpu%i " fmt, (vcpu)->vcpu_id, ## __VA_ARGS__)
538 bool kvm_make_all_cpus_request(struct kvm *kvm, unsigned int req);
539 static inline void kvm_vm_bugged(struct kvm *kvm)
541 kvm->vm_bugged = true;
542 kvm_make_all_cpus_request(kvm, KVM_REQ_VM_BUGGED);
545 #define KVM_BUG(cond, kvm, fmt...) \
547 int __ret = (cond); \
549 if (WARN_ONCE(__ret && !(kvm)->vm_bugged, fmt)) \
550 kvm_vm_bugged(kvm); \
554 #define KVM_BUG_ON(cond, kvm) \
556 int __ret = (cond); \
558 if (WARN_ON_ONCE(__ret && !(kvm)->vm_bugged)) \
559 kvm_vm_bugged(kvm); \
563 static inline bool kvm_dirty_log_manual_protect_and_init_set(struct kvm *kvm)
565 return !!(kvm->manual_dirty_log_protect & KVM_DIRTY_LOG_INITIALLY_SET);
568 static inline struct kvm_io_bus *kvm_get_bus(struct kvm *kvm, enum kvm_bus idx)
570 return srcu_dereference_check(kvm->buses[idx], &kvm->srcu,
571 lockdep_is_held(&kvm->slots_lock) ||
572 !refcount_read(&kvm->users_count));
575 static inline struct kvm_vcpu *kvm_get_vcpu(struct kvm *kvm, int i)
577 int num_vcpus = atomic_read(&kvm->online_vcpus);
578 i = array_index_nospec(i, num_vcpus);
580 /* Pairs with smp_wmb() in kvm_vm_ioctl_create_vcpu. */
582 return kvm->vcpus[i];
585 #define kvm_for_each_vcpu(idx, vcpup, kvm) \
587 idx < atomic_read(&kvm->online_vcpus) && \
588 (vcpup = kvm_get_vcpu(kvm, idx)) != NULL; \
591 static inline struct kvm_vcpu *kvm_get_vcpu_by_id(struct kvm *kvm, int id)
593 struct kvm_vcpu *vcpu = NULL;
598 if (id < KVM_MAX_VCPUS)
599 vcpu = kvm_get_vcpu(kvm, id);
600 if (vcpu && vcpu->vcpu_id == id)
602 kvm_for_each_vcpu(i, vcpu, kvm)
603 if (vcpu->vcpu_id == id)
608 static inline int kvm_vcpu_get_idx(struct kvm_vcpu *vcpu)
610 return vcpu->vcpu_idx;
613 #define kvm_for_each_memslot(memslot, slots) \
614 for (memslot = &slots->memslots[0]; \
615 memslot < slots->memslots + slots->used_slots; memslot++) \
616 if (WARN_ON_ONCE(!memslot->npages)) { \
619 void kvm_vcpu_destroy(struct kvm_vcpu *vcpu);
621 void vcpu_load(struct kvm_vcpu *vcpu);
622 void vcpu_put(struct kvm_vcpu *vcpu);
624 #ifdef __KVM_HAVE_IOAPIC
625 void kvm_arch_post_irq_ack_notifier_list_update(struct kvm *kvm);
626 void kvm_arch_post_irq_routing_update(struct kvm *kvm);
628 static inline void kvm_arch_post_irq_ack_notifier_list_update(struct kvm *kvm)
631 static inline void kvm_arch_post_irq_routing_update(struct kvm *kvm)
636 #ifdef CONFIG_HAVE_KVM_IRQFD
637 int kvm_irqfd_init(void);
638 void kvm_irqfd_exit(void);
640 static inline int kvm_irqfd_init(void)
645 static inline void kvm_irqfd_exit(void)
649 int kvm_init(void *opaque, unsigned vcpu_size, unsigned vcpu_align,
650 struct module *module);
653 void kvm_get_kvm(struct kvm *kvm);
654 void kvm_put_kvm(struct kvm *kvm);
655 void kvm_put_kvm_no_destroy(struct kvm *kvm);
657 static inline struct kvm_memslots *__kvm_memslots(struct kvm *kvm, int as_id)
659 as_id = array_index_nospec(as_id, KVM_ADDRESS_SPACE_NUM);
660 return srcu_dereference_check(kvm->memslots[as_id], &kvm->srcu,
661 lockdep_is_held(&kvm->slots_lock) ||
662 !refcount_read(&kvm->users_count));
665 static inline struct kvm_memslots *kvm_memslots(struct kvm *kvm)
667 return __kvm_memslots(kvm, 0);
670 static inline struct kvm_memslots *kvm_vcpu_memslots(struct kvm_vcpu *vcpu)
672 int as_id = kvm_arch_vcpu_memslots_id(vcpu);
674 return __kvm_memslots(vcpu->kvm, as_id);
678 struct kvm_memory_slot *id_to_memslot(struct kvm_memslots *slots, int id)
680 int index = slots->id_to_index[id];
681 struct kvm_memory_slot *slot;
686 slot = &slots->memslots[index];
688 WARN_ON(slot->id != id);
693 * KVM_SET_USER_MEMORY_REGION ioctl allows the following operations:
694 * - create a new memory slot
695 * - delete an existing memory slot
696 * - modify an existing memory slot
697 * -- move it in the guest physical memory space
698 * -- just change its flags
700 * Since flags can be changed by some of these operations, the following
701 * differentiation is the best we can do for __kvm_set_memory_region():
710 int kvm_set_memory_region(struct kvm *kvm,
711 const struct kvm_userspace_memory_region *mem);
712 int __kvm_set_memory_region(struct kvm *kvm,
713 const struct kvm_userspace_memory_region *mem);
714 void kvm_arch_free_memslot(struct kvm *kvm, struct kvm_memory_slot *slot);
715 void kvm_arch_memslots_updated(struct kvm *kvm, u64 gen);
716 int kvm_arch_prepare_memory_region(struct kvm *kvm,
717 struct kvm_memory_slot *memslot,
718 const struct kvm_userspace_memory_region *mem,
719 enum kvm_mr_change change);
720 void kvm_arch_commit_memory_region(struct kvm *kvm,
721 const struct kvm_userspace_memory_region *mem,
722 struct kvm_memory_slot *old,
723 const struct kvm_memory_slot *new,
724 enum kvm_mr_change change);
725 /* flush all memory translations */
726 void kvm_arch_flush_shadow_all(struct kvm *kvm);
727 /* flush memory translations pointing to 'slot' */
728 void kvm_arch_flush_shadow_memslot(struct kvm *kvm,
729 struct kvm_memory_slot *slot);
731 int gfn_to_page_many_atomic(struct kvm_memory_slot *slot, gfn_t gfn,
732 struct page **pages, int nr_pages);
734 struct page *gfn_to_page(struct kvm *kvm, gfn_t gfn);
735 unsigned long gfn_to_hva(struct kvm *kvm, gfn_t gfn);
736 unsigned long gfn_to_hva_prot(struct kvm *kvm, gfn_t gfn, bool *writable);
737 unsigned long gfn_to_hva_memslot(struct kvm_memory_slot *slot, gfn_t gfn);
738 unsigned long gfn_to_hva_memslot_prot(struct kvm_memory_slot *slot, gfn_t gfn,
740 void kvm_release_page_clean(struct page *page);
741 void kvm_release_page_dirty(struct page *page);
742 void kvm_set_page_accessed(struct page *page);
744 kvm_pfn_t gfn_to_pfn(struct kvm *kvm, gfn_t gfn);
745 kvm_pfn_t gfn_to_pfn_prot(struct kvm *kvm, gfn_t gfn, bool write_fault,
747 kvm_pfn_t gfn_to_pfn_memslot(struct kvm_memory_slot *slot, gfn_t gfn);
748 kvm_pfn_t gfn_to_pfn_memslot_atomic(struct kvm_memory_slot *slot, gfn_t gfn);
749 kvm_pfn_t __gfn_to_pfn_memslot(struct kvm_memory_slot *slot, gfn_t gfn,
750 bool atomic, bool *async, bool write_fault,
753 void kvm_release_pfn_clean(kvm_pfn_t pfn);
754 void kvm_release_pfn_dirty(kvm_pfn_t pfn);
755 void kvm_set_pfn_dirty(kvm_pfn_t pfn);
756 void kvm_set_pfn_accessed(kvm_pfn_t pfn);
757 void kvm_get_pfn(kvm_pfn_t pfn);
759 void kvm_release_pfn(kvm_pfn_t pfn, bool dirty, struct gfn_to_pfn_cache *cache);
760 int kvm_read_guest_page(struct kvm *kvm, gfn_t gfn, void *data, int offset,
762 int kvm_read_guest(struct kvm *kvm, gpa_t gpa, void *data, unsigned long len);
763 int kvm_read_guest_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
764 void *data, unsigned long len);
765 int kvm_read_guest_offset_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
766 void *data, unsigned int offset,
768 int kvm_write_guest_page(struct kvm *kvm, gfn_t gfn, const void *data,
769 int offset, int len);
770 int kvm_write_guest(struct kvm *kvm, gpa_t gpa, const void *data,
772 int kvm_write_guest_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
773 void *data, unsigned long len);
774 int kvm_write_guest_offset_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
775 void *data, unsigned int offset,
777 int kvm_gfn_to_hva_cache_init(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
778 gpa_t gpa, unsigned long len);
780 #define __kvm_get_guest(kvm, gfn, offset, v) \
782 unsigned long __addr = gfn_to_hva(kvm, gfn); \
783 typeof(v) __user *__uaddr = (typeof(__uaddr))(__addr + offset); \
784 int __ret = -EFAULT; \
786 if (!kvm_is_error_hva(__addr)) \
787 __ret = get_user(v, __uaddr); \
791 #define kvm_get_guest(kvm, gpa, v) \
794 struct kvm *__kvm = kvm; \
796 __kvm_get_guest(__kvm, __gpa >> PAGE_SHIFT, \
797 offset_in_page(__gpa), v); \
800 #define __kvm_put_guest(kvm, gfn, offset, v) \
802 unsigned long __addr = gfn_to_hva(kvm, gfn); \
803 typeof(v) __user *__uaddr = (typeof(__uaddr))(__addr + offset); \
804 int __ret = -EFAULT; \
806 if (!kvm_is_error_hva(__addr)) \
807 __ret = put_user(v, __uaddr); \
809 mark_page_dirty(kvm, gfn); \
813 #define kvm_put_guest(kvm, gpa, v) \
816 struct kvm *__kvm = kvm; \
818 __kvm_put_guest(__kvm, __gpa >> PAGE_SHIFT, \
819 offset_in_page(__gpa), v); \
822 int kvm_clear_guest_page(struct kvm *kvm, gfn_t gfn, int offset, int len);
823 int kvm_clear_guest(struct kvm *kvm, gpa_t gpa, unsigned long len);
824 struct kvm_memory_slot *gfn_to_memslot(struct kvm *kvm, gfn_t gfn);
825 bool kvm_is_visible_gfn(struct kvm *kvm, gfn_t gfn);
826 bool kvm_vcpu_is_visible_gfn(struct kvm_vcpu *vcpu, gfn_t gfn);
827 unsigned long kvm_host_page_size(struct kvm_vcpu *vcpu, gfn_t gfn);
828 void mark_page_dirty_in_slot(struct kvm_memory_slot *memslot, gfn_t gfn);
829 void mark_page_dirty(struct kvm *kvm, gfn_t gfn);
831 struct kvm_memslots *kvm_vcpu_memslots(struct kvm_vcpu *vcpu);
832 struct kvm_memory_slot *kvm_vcpu_gfn_to_memslot(struct kvm_vcpu *vcpu, gfn_t gfn);
833 kvm_pfn_t kvm_vcpu_gfn_to_pfn_atomic(struct kvm_vcpu *vcpu, gfn_t gfn);
834 kvm_pfn_t kvm_vcpu_gfn_to_pfn(struct kvm_vcpu *vcpu, gfn_t gfn);
835 int kvm_vcpu_map(struct kvm_vcpu *vcpu, gpa_t gpa, struct kvm_host_map *map);
836 int kvm_map_gfn(struct kvm_vcpu *vcpu, gfn_t gfn, struct kvm_host_map *map,
837 struct gfn_to_pfn_cache *cache, bool atomic);
838 struct page *kvm_vcpu_gfn_to_page(struct kvm_vcpu *vcpu, gfn_t gfn);
839 void kvm_vcpu_unmap(struct kvm_vcpu *vcpu, struct kvm_host_map *map, bool dirty);
840 int kvm_unmap_gfn(struct kvm_vcpu *vcpu, struct kvm_host_map *map,
841 struct gfn_to_pfn_cache *cache, bool dirty, bool atomic);
842 unsigned long kvm_vcpu_gfn_to_hva(struct kvm_vcpu *vcpu, gfn_t gfn);
843 unsigned long kvm_vcpu_gfn_to_hva_prot(struct kvm_vcpu *vcpu, gfn_t gfn, bool *writable);
844 int kvm_vcpu_read_guest_page(struct kvm_vcpu *vcpu, gfn_t gfn, void *data, int offset,
846 int kvm_vcpu_read_guest_atomic(struct kvm_vcpu *vcpu, gpa_t gpa, void *data,
848 int kvm_vcpu_read_guest(struct kvm_vcpu *vcpu, gpa_t gpa, void *data,
850 int kvm_vcpu_write_guest_page(struct kvm_vcpu *vcpu, gfn_t gfn, const void *data,
851 int offset, int len);
852 int kvm_vcpu_write_guest(struct kvm_vcpu *vcpu, gpa_t gpa, const void *data,
854 void kvm_vcpu_mark_page_dirty(struct kvm_vcpu *vcpu, gfn_t gfn);
856 void kvm_sigset_activate(struct kvm_vcpu *vcpu);
857 void kvm_sigset_deactivate(struct kvm_vcpu *vcpu);
859 void kvm_vcpu_block(struct kvm_vcpu *vcpu);
860 void kvm_arch_vcpu_blocking(struct kvm_vcpu *vcpu);
861 void kvm_arch_vcpu_unblocking(struct kvm_vcpu *vcpu);
862 bool kvm_vcpu_wake_up(struct kvm_vcpu *vcpu);
863 void kvm_vcpu_kick(struct kvm_vcpu *vcpu);
864 int kvm_vcpu_yield_to(struct kvm_vcpu *target);
865 void kvm_vcpu_on_spin(struct kvm_vcpu *vcpu, bool usermode_vcpu_not_eligible);
867 void kvm_flush_remote_tlbs(struct kvm *kvm);
868 void kvm_reload_remote_mmus(struct kvm *kvm);
870 #ifdef KVM_ARCH_NR_OBJS_PER_MEMORY_CACHE
871 int kvm_mmu_topup_memory_cache(struct kvm_mmu_memory_cache *mc, int min);
872 int kvm_mmu_memory_cache_nr_free_objects(struct kvm_mmu_memory_cache *mc);
873 void kvm_mmu_free_memory_cache(struct kvm_mmu_memory_cache *mc);
874 void *kvm_mmu_memory_cache_alloc(struct kvm_mmu_memory_cache *mc);
877 bool kvm_make_vcpus_request_mask(struct kvm *kvm, unsigned int req,
878 struct kvm_vcpu *except,
879 unsigned long *vcpu_bitmap, cpumask_var_t tmp);
880 bool kvm_make_all_cpus_request_except(struct kvm *kvm, unsigned int req,
881 struct kvm_vcpu *except);
882 bool kvm_make_cpus_request_mask(struct kvm *kvm, unsigned int req,
883 unsigned long *vcpu_bitmap);
885 long kvm_arch_dev_ioctl(struct file *filp,
886 unsigned int ioctl, unsigned long arg);
887 long kvm_arch_vcpu_ioctl(struct file *filp,
888 unsigned int ioctl, unsigned long arg);
889 vm_fault_t kvm_arch_vcpu_fault(struct kvm_vcpu *vcpu, struct vm_fault *vmf);
891 int kvm_vm_ioctl_check_extension(struct kvm *kvm, long ext);
893 void kvm_arch_mmu_enable_log_dirty_pt_masked(struct kvm *kvm,
894 struct kvm_memory_slot *slot,
897 void kvm_arch_sync_dirty_log(struct kvm *kvm, struct kvm_memory_slot *memslot);
899 #ifdef CONFIG_KVM_GENERIC_DIRTYLOG_READ_PROTECT
900 void kvm_arch_flush_remote_tlbs_memslot(struct kvm *kvm,
901 struct kvm_memory_slot *memslot);
902 #else /* !CONFIG_KVM_GENERIC_DIRTYLOG_READ_PROTECT */
903 int kvm_vm_ioctl_get_dirty_log(struct kvm *kvm, struct kvm_dirty_log *log);
904 int kvm_get_dirty_log(struct kvm *kvm, struct kvm_dirty_log *log,
905 int *is_dirty, struct kvm_memory_slot **memslot);
908 int kvm_vm_ioctl_irq_line(struct kvm *kvm, struct kvm_irq_level *irq_level,
910 int kvm_vm_ioctl_enable_cap(struct kvm *kvm,
911 struct kvm_enable_cap *cap);
912 long kvm_arch_vm_ioctl(struct file *filp,
913 unsigned int ioctl, unsigned long arg);
914 long kvm_arch_vm_compat_ioctl(struct file *filp, unsigned int ioctl,
917 int kvm_arch_vcpu_ioctl_get_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu);
918 int kvm_arch_vcpu_ioctl_set_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu);
920 int kvm_arch_vcpu_ioctl_translate(struct kvm_vcpu *vcpu,
921 struct kvm_translation *tr);
923 int kvm_arch_vcpu_ioctl_get_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs);
924 int kvm_arch_vcpu_ioctl_set_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs);
925 int kvm_arch_vcpu_ioctl_get_sregs(struct kvm_vcpu *vcpu,
926 struct kvm_sregs *sregs);
927 int kvm_arch_vcpu_ioctl_set_sregs(struct kvm_vcpu *vcpu,
928 struct kvm_sregs *sregs);
929 int kvm_arch_vcpu_ioctl_get_mpstate(struct kvm_vcpu *vcpu,
930 struct kvm_mp_state *mp_state);
931 int kvm_arch_vcpu_ioctl_set_mpstate(struct kvm_vcpu *vcpu,
932 struct kvm_mp_state *mp_state);
933 int kvm_arch_vcpu_ioctl_set_guest_debug(struct kvm_vcpu *vcpu,
934 struct kvm_guest_debug *dbg);
935 int kvm_arch_vcpu_ioctl_run(struct kvm_vcpu *vcpu);
937 int kvm_arch_init(void *opaque);
938 void kvm_arch_exit(void);
940 void kvm_arch_sched_in(struct kvm_vcpu *vcpu, int cpu);
942 void kvm_arch_vcpu_load(struct kvm_vcpu *vcpu, int cpu);
943 void kvm_arch_vcpu_put(struct kvm_vcpu *vcpu);
944 int kvm_arch_vcpu_precreate(struct kvm *kvm, unsigned int id);
945 int kvm_arch_vcpu_create(struct kvm_vcpu *vcpu);
946 void kvm_arch_vcpu_postcreate(struct kvm_vcpu *vcpu);
947 void kvm_arch_vcpu_destroy(struct kvm_vcpu *vcpu);
949 #ifdef __KVM_HAVE_ARCH_VCPU_DEBUGFS
950 void kvm_arch_create_vcpu_debugfs(struct kvm_vcpu *vcpu, struct dentry *debugfs_dentry);
953 int kvm_arch_hardware_enable(void);
954 void kvm_arch_hardware_disable(void);
955 int kvm_arch_hardware_setup(void *opaque);
956 void kvm_arch_hardware_unsetup(void);
957 int kvm_arch_check_processor_compat(void *opaque);
958 int kvm_arch_vcpu_runnable(struct kvm_vcpu *vcpu);
959 bool kvm_arch_vcpu_in_kernel(struct kvm_vcpu *vcpu);
960 int kvm_arch_vcpu_should_kick(struct kvm_vcpu *vcpu);
961 bool kvm_arch_dy_runnable(struct kvm_vcpu *vcpu);
962 int kvm_arch_post_init_vm(struct kvm *kvm);
963 void kvm_arch_pre_destroy_vm(struct kvm *kvm);
965 #ifndef __KVM_HAVE_ARCH_VM_ALLOC
967 * All architectures that want to use vzalloc currently also
968 * need their own kvm_arch_alloc_vm implementation.
970 static inline struct kvm *kvm_arch_alloc_vm(void)
972 return kzalloc(sizeof(struct kvm), GFP_KERNEL);
975 static inline void kvm_arch_free_vm(struct kvm *kvm)
981 #ifndef __KVM_HAVE_ARCH_FLUSH_REMOTE_TLB
982 static inline int kvm_arch_flush_remote_tlb(struct kvm *kvm)
988 #ifdef __KVM_HAVE_ARCH_NONCOHERENT_DMA
989 void kvm_arch_register_noncoherent_dma(struct kvm *kvm);
990 void kvm_arch_unregister_noncoherent_dma(struct kvm *kvm);
991 bool kvm_arch_has_noncoherent_dma(struct kvm *kvm);
993 static inline void kvm_arch_register_noncoherent_dma(struct kvm *kvm)
997 static inline void kvm_arch_unregister_noncoherent_dma(struct kvm *kvm)
1001 static inline bool kvm_arch_has_noncoherent_dma(struct kvm *kvm)
1006 #ifdef __KVM_HAVE_ARCH_ASSIGNED_DEVICE
1007 void kvm_arch_start_assignment(struct kvm *kvm);
1008 void kvm_arch_end_assignment(struct kvm *kvm);
1009 bool kvm_arch_has_assigned_device(struct kvm *kvm);
1011 static inline void kvm_arch_start_assignment(struct kvm *kvm)
1015 static inline void kvm_arch_end_assignment(struct kvm *kvm)
1019 static __always_inline bool kvm_arch_has_assigned_device(struct kvm *kvm)
1025 static inline struct rcuwait *kvm_arch_vcpu_get_wait(struct kvm_vcpu *vcpu)
1027 #ifdef __KVM_HAVE_ARCH_WQP
1028 return vcpu->arch.waitp;
1034 #ifdef __KVM_HAVE_ARCH_INTC_INITIALIZED
1036 * returns true if the virtual interrupt controller is initialized and
1037 * ready to accept virtual IRQ. On some architectures the virtual interrupt
1038 * controller is dynamically instantiated and this is not always true.
1040 bool kvm_arch_intc_initialized(struct kvm *kvm);
1042 static inline bool kvm_arch_intc_initialized(struct kvm *kvm)
1048 int kvm_arch_init_vm(struct kvm *kvm, unsigned long type);
1049 void kvm_arch_destroy_vm(struct kvm *kvm);
1050 void kvm_arch_sync_events(struct kvm *kvm);
1052 int kvm_cpu_has_pending_timer(struct kvm_vcpu *vcpu);
1054 bool kvm_is_reserved_pfn(kvm_pfn_t pfn);
1055 bool kvm_is_zone_device_pfn(kvm_pfn_t pfn);
1056 bool kvm_is_transparent_hugepage(kvm_pfn_t pfn);
1058 struct kvm_irq_ack_notifier {
1059 struct hlist_node link;
1061 void (*irq_acked)(struct kvm_irq_ack_notifier *kian);
1064 int kvm_irq_map_gsi(struct kvm *kvm,
1065 struct kvm_kernel_irq_routing_entry *entries, int gsi);
1066 int kvm_irq_map_chip_pin(struct kvm *kvm, unsigned irqchip, unsigned pin);
1068 int kvm_set_irq(struct kvm *kvm, int irq_source_id, u32 irq, int level,
1070 int kvm_set_msi(struct kvm_kernel_irq_routing_entry *irq_entry, struct kvm *kvm,
1071 int irq_source_id, int level, bool line_status);
1072 int kvm_arch_set_irq_inatomic(struct kvm_kernel_irq_routing_entry *e,
1073 struct kvm *kvm, int irq_source_id,
1074 int level, bool line_status);
1075 bool kvm_irq_has_notifier(struct kvm *kvm, unsigned irqchip, unsigned pin);
1076 void kvm_notify_acked_gsi(struct kvm *kvm, int gsi);
1077 void kvm_notify_acked_irq(struct kvm *kvm, unsigned irqchip, unsigned pin);
1078 void kvm_register_irq_ack_notifier(struct kvm *kvm,
1079 struct kvm_irq_ack_notifier *kian);
1080 void kvm_unregister_irq_ack_notifier(struct kvm *kvm,
1081 struct kvm_irq_ack_notifier *kian);
1082 int kvm_request_irq_source_id(struct kvm *kvm);
1083 void kvm_free_irq_source_id(struct kvm *kvm, int irq_source_id);
1084 bool kvm_arch_irqfd_allowed(struct kvm *kvm, struct kvm_irqfd *args);
1087 * search_memslots() and __gfn_to_memslot() are here because they are
1088 * used in non-modular code in arch/powerpc/kvm/book3s_hv_rm_mmu.c.
1089 * gfn_to_memslot() itself isn't here as an inline because that would
1090 * bloat other code too much.
1092 * IMPORTANT: Slots are sorted from highest GFN to lowest GFN!
1094 static inline struct kvm_memory_slot *
1095 search_memslots(struct kvm_memslots *slots, gfn_t gfn)
1097 int start = 0, end = slots->used_slots;
1098 int slot = atomic_read(&slots->lru_slot);
1099 struct kvm_memory_slot *memslots = slots->memslots;
1101 if (unlikely(!slots->used_slots))
1104 if (gfn >= memslots[slot].base_gfn &&
1105 gfn < memslots[slot].base_gfn + memslots[slot].npages)
1106 return &memslots[slot];
1108 while (start < end) {
1109 slot = start + (end - start) / 2;
1111 if (gfn >= memslots[slot].base_gfn)
1117 if (start < slots->used_slots && gfn >= memslots[start].base_gfn &&
1118 gfn < memslots[start].base_gfn + memslots[start].npages) {
1119 atomic_set(&slots->lru_slot, start);
1120 return &memslots[start];
1126 static inline struct kvm_memory_slot *
1127 __gfn_to_memslot(struct kvm_memslots *slots, gfn_t gfn)
1129 return search_memslots(slots, gfn);
1132 static inline unsigned long
1133 __gfn_to_hva_memslot(struct kvm_memory_slot *slot, gfn_t gfn)
1136 * The index was checked originally in search_memslots. To avoid
1137 * that a malicious guest builds a Spectre gadget out of e.g. page
1138 * table walks, do not let the processor speculate loads outside
1139 * the guest's registered memslots.
1141 unsigned long offset = gfn - slot->base_gfn;
1142 offset = array_index_nospec(offset, slot->npages);
1143 return slot->userspace_addr + offset * 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)
1237 #ifdef CONFIG_HAVE_KVM_IRQ_ROUTING
1239 #define KVM_MAX_IRQ_ROUTES 4096 /* might need extension/rework in the future */
1241 bool kvm_arch_can_set_irq_routing(struct kvm *kvm);
1242 int kvm_set_irq_routing(struct kvm *kvm,
1243 const struct kvm_irq_routing_entry *entries,
1246 int kvm_set_routing_entry(struct kvm *kvm,
1247 struct kvm_kernel_irq_routing_entry *e,
1248 const struct kvm_irq_routing_entry *ue);
1249 void kvm_free_irq_routing(struct kvm *kvm);
1253 static inline void kvm_free_irq_routing(struct kvm *kvm) {}
1257 int kvm_send_userspace_msi(struct kvm *kvm, struct kvm_msi *msi);
1259 #ifdef CONFIG_HAVE_KVM_EVENTFD
1261 void kvm_eventfd_init(struct kvm *kvm);
1262 int kvm_ioeventfd(struct kvm *kvm, struct kvm_ioeventfd *args);
1264 #ifdef CONFIG_HAVE_KVM_IRQFD
1265 int kvm_irqfd(struct kvm *kvm, struct kvm_irqfd *args);
1266 void kvm_irqfd_release(struct kvm *kvm);
1267 void kvm_irq_routing_update(struct kvm *);
1269 static inline int kvm_irqfd(struct kvm *kvm, struct kvm_irqfd *args)
1274 static inline void kvm_irqfd_release(struct kvm *kvm) {}
1279 static inline void kvm_eventfd_init(struct kvm *kvm) {}
1281 static inline int kvm_irqfd(struct kvm *kvm, struct kvm_irqfd *args)
1286 static inline void kvm_irqfd_release(struct kvm *kvm) {}
1288 #ifdef CONFIG_HAVE_KVM_IRQCHIP
1289 static inline void kvm_irq_routing_update(struct kvm *kvm)
1294 static inline int kvm_ioeventfd(struct kvm *kvm, struct kvm_ioeventfd *args)
1299 #endif /* CONFIG_HAVE_KVM_EVENTFD */
1301 void kvm_arch_irq_routing_update(struct kvm *kvm);
1303 static inline void kvm_make_request(int req, struct kvm_vcpu *vcpu)
1306 * Ensure the rest of the request is published to kvm_check_request's
1307 * caller. Paired with the smp_mb__after_atomic in kvm_check_request.
1310 set_bit(req & KVM_REQUEST_MASK, (void *)&vcpu->requests);
1313 static inline bool kvm_request_pending(struct kvm_vcpu *vcpu)
1315 return READ_ONCE(vcpu->requests);
1318 static inline bool kvm_test_request(int req, struct kvm_vcpu *vcpu)
1320 return test_bit(req & KVM_REQUEST_MASK, (void *)&vcpu->requests);
1323 static inline void kvm_clear_request(int req, struct kvm_vcpu *vcpu)
1325 clear_bit(req & KVM_REQUEST_MASK, (void *)&vcpu->requests);
1328 static inline bool kvm_check_request(int req, struct kvm_vcpu *vcpu)
1330 if (kvm_test_request(req, vcpu)) {
1331 kvm_clear_request(req, vcpu);
1334 * Ensure the rest of the request is visible to kvm_check_request's
1335 * caller. Paired with the smp_wmb in kvm_make_request.
1337 smp_mb__after_atomic();
1344 extern bool kvm_rebooting;
1346 extern unsigned int halt_poll_ns;
1347 extern unsigned int halt_poll_ns_grow;
1348 extern unsigned int halt_poll_ns_grow_start;
1349 extern unsigned int halt_poll_ns_shrink;
1352 const struct kvm_device_ops *ops;
1355 struct list_head vm_node;
1358 /* create, destroy, and name are mandatory */
1359 struct kvm_device_ops {
1363 * create is called holding kvm->lock and any operations not suitable
1364 * to do while holding the lock should be deferred to init (see
1367 int (*create)(struct kvm_device *dev, u32 type);
1370 * init is called after create if create is successful and is called
1371 * outside of holding kvm->lock.
1373 void (*init)(struct kvm_device *dev);
1376 * Destroy is responsible for freeing dev.
1378 * Destroy may be called before or after destructors are called
1379 * on emulated I/O regions, depending on whether a reference is
1380 * held by a vcpu or other kvm component that gets destroyed
1381 * after the emulated I/O.
1383 void (*destroy)(struct kvm_device *dev);
1386 * Release is an alternative method to free the device. It is
1387 * called when the device file descriptor is closed. Once
1388 * release is called, the destroy method will not be called
1389 * anymore as the device is removed from the device list of
1390 * the VM. kvm->lock is held.
1392 void (*release)(struct kvm_device *dev);
1394 int (*set_attr)(struct kvm_device *dev, struct kvm_device_attr *attr);
1395 int (*get_attr)(struct kvm_device *dev, struct kvm_device_attr *attr);
1396 int (*has_attr)(struct kvm_device *dev, struct kvm_device_attr *attr);
1397 long (*ioctl)(struct kvm_device *dev, unsigned int ioctl,
1399 int (*mmap)(struct kvm_device *dev, struct vm_area_struct *vma);
1402 void kvm_device_get(struct kvm_device *dev);
1403 void kvm_device_put(struct kvm_device *dev);
1404 struct kvm_device *kvm_device_from_filp(struct file *filp);
1405 int kvm_register_device_ops(const struct kvm_device_ops *ops, u32 type);
1406 void kvm_unregister_device_ops(u32 type);
1408 extern struct kvm_device_ops kvm_mpic_ops;
1409 extern struct kvm_device_ops kvm_arm_vgic_v2_ops;
1410 extern struct kvm_device_ops kvm_arm_vgic_v3_ops;
1412 #ifdef CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT
1414 static inline void kvm_vcpu_set_in_spin_loop(struct kvm_vcpu *vcpu, bool val)
1416 vcpu->spin_loop.in_spin_loop = val;
1418 static inline void kvm_vcpu_set_dy_eligible(struct kvm_vcpu *vcpu, bool val)
1420 vcpu->spin_loop.dy_eligible = val;
1423 #else /* !CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT */
1425 static inline void kvm_vcpu_set_in_spin_loop(struct kvm_vcpu *vcpu, bool val)
1429 static inline void kvm_vcpu_set_dy_eligible(struct kvm_vcpu *vcpu, bool val)
1432 #endif /* CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT */
1434 static inline bool kvm_is_visible_memslot(struct kvm_memory_slot *memslot)
1436 return (memslot && memslot->id < KVM_USER_MEM_SLOTS &&
1437 !(memslot->flags & KVM_MEMSLOT_INVALID));
1440 struct kvm_vcpu *kvm_get_running_vcpu(void);
1441 struct kvm_vcpu * __percpu *kvm_get_running_vcpus(void);
1443 #ifdef CONFIG_HAVE_KVM_IRQ_BYPASS
1444 bool kvm_arch_has_irq_bypass(void);
1445 int kvm_arch_irq_bypass_add_producer(struct irq_bypass_consumer *,
1446 struct irq_bypass_producer *);
1447 void kvm_arch_irq_bypass_del_producer(struct irq_bypass_consumer *,
1448 struct irq_bypass_producer *);
1449 void kvm_arch_irq_bypass_stop(struct irq_bypass_consumer *);
1450 void kvm_arch_irq_bypass_start(struct irq_bypass_consumer *);
1451 int kvm_arch_update_irqfd_routing(struct kvm *kvm, unsigned int host_irq,
1452 uint32_t guest_irq, bool set);
1453 #endif /* CONFIG_HAVE_KVM_IRQ_BYPASS */
1455 #ifdef CONFIG_HAVE_KVM_INVALID_WAKEUPS
1456 /* If we wakeup during the poll time, was it a sucessful poll? */
1457 static inline bool vcpu_valid_wakeup(struct kvm_vcpu *vcpu)
1459 return vcpu->valid_wakeup;
1463 static inline bool vcpu_valid_wakeup(struct kvm_vcpu *vcpu)
1467 #endif /* CONFIG_HAVE_KVM_INVALID_WAKEUPS */
1469 #ifdef CONFIG_HAVE_KVM_NO_POLL
1470 /* Callback that tells if we must not poll */
1471 bool kvm_arch_no_poll(struct kvm_vcpu *vcpu);
1473 static inline bool kvm_arch_no_poll(struct kvm_vcpu *vcpu)
1477 #endif /* CONFIG_HAVE_KVM_NO_POLL */
1479 #ifdef CONFIG_HAVE_KVM_VCPU_ASYNC_IOCTL
1480 long kvm_arch_vcpu_async_ioctl(struct file *filp,
1481 unsigned int ioctl, unsigned long arg);
1483 static inline long kvm_arch_vcpu_async_ioctl(struct file *filp,
1487 return -ENOIOCTLCMD;
1489 #endif /* CONFIG_HAVE_KVM_VCPU_ASYNC_IOCTL */
1491 void kvm_arch_mmu_notifier_invalidate_range(struct kvm *kvm,
1492 unsigned long start, unsigned long end);
1494 void kvm_arch_guest_memory_reclaimed(struct kvm *kvm);
1496 #ifdef CONFIG_HAVE_KVM_VCPU_RUN_PID_CHANGE
1497 int kvm_arch_vcpu_run_pid_change(struct kvm_vcpu *vcpu);
1499 static inline int kvm_arch_vcpu_run_pid_change(struct kvm_vcpu *vcpu)
1503 #endif /* CONFIG_HAVE_KVM_VCPU_RUN_PID_CHANGE */
1505 typedef int (*kvm_vm_thread_fn_t)(struct kvm *kvm, uintptr_t data);
1507 int kvm_vm_create_worker_thread(struct kvm *kvm, kvm_vm_thread_fn_t thread_fn,
1508 uintptr_t data, const char *name,
1509 struct task_struct **thread_ptr);
1511 #ifdef CONFIG_KVM_XFER_TO_GUEST_WORK
1512 static inline void kvm_handle_signal_exit(struct kvm_vcpu *vcpu)
1514 vcpu->run->exit_reason = KVM_EXIT_INTR;
1515 vcpu->stat.signal_exits++;
1517 #endif /* CONFIG_KVM_XFER_TO_GUEST_WORK */