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>
37 #include <linux/kvm_dirty_ring.h>
39 #ifndef KVM_MAX_VCPU_ID
40 #define KVM_MAX_VCPU_ID KVM_MAX_VCPUS
44 * The bit 16 ~ bit 31 of kvm_memory_region::flags are internally used
45 * in kvm, other bits are visible for userspace which are defined in
46 * include/linux/kvm_h.
48 #define KVM_MEMSLOT_INVALID (1UL << 16)
51 * Bit 63 of the memslot generation number is an "update in-progress flag",
52 * e.g. is temporarily set for the duration of install_new_memslots().
53 * This flag effectively creates a unique generation number that is used to
54 * mark cached memslot data, e.g. MMIO accesses, as potentially being stale,
55 * i.e. may (or may not) have come from the previous memslots generation.
57 * This is necessary because the actual memslots update is not atomic with
58 * respect to the generation number update. Updating the generation number
59 * first would allow a vCPU to cache a spte from the old memslots using the
60 * new generation number, and updating the generation number after switching
61 * to the new memslots would allow cache hits using the old generation number
62 * to reference the defunct memslots.
64 * This mechanism is used to prevent getting hits in KVM's caches while a
65 * memslot update is in-progress, and to prevent cache hits *after* updating
66 * the actual generation number against accesses that were inserted into the
67 * cache *before* the memslots were updated.
69 #define KVM_MEMSLOT_GEN_UPDATE_IN_PROGRESS BIT_ULL(63)
71 /* Two fragments for cross MMIO pages. */
72 #define KVM_MAX_MMIO_FRAGMENTS 2
74 #ifndef KVM_ADDRESS_SPACE_NUM
75 #define KVM_ADDRESS_SPACE_NUM 1
79 * For the normal pfn, the highest 12 bits should be zero,
80 * so we can mask bit 62 ~ bit 52 to indicate the error pfn,
81 * mask bit 63 to indicate the noslot pfn.
83 #define KVM_PFN_ERR_MASK (0x7ffULL << 52)
84 #define KVM_PFN_ERR_NOSLOT_MASK (0xfffULL << 52)
85 #define KVM_PFN_NOSLOT (0x1ULL << 63)
87 #define KVM_PFN_ERR_FAULT (KVM_PFN_ERR_MASK)
88 #define KVM_PFN_ERR_HWPOISON (KVM_PFN_ERR_MASK + 1)
89 #define KVM_PFN_ERR_RO_FAULT (KVM_PFN_ERR_MASK + 2)
92 * error pfns indicate that the gfn is in slot but faild to
93 * translate it to pfn on host.
95 static inline bool is_error_pfn(kvm_pfn_t pfn)
97 return !!(pfn & KVM_PFN_ERR_MASK);
101 * error_noslot pfns indicate that the gfn can not be
102 * translated to pfn - it is not in slot or failed to
103 * translate it to pfn.
105 static inline bool is_error_noslot_pfn(kvm_pfn_t pfn)
107 return !!(pfn & KVM_PFN_ERR_NOSLOT_MASK);
110 /* noslot pfn indicates that the gfn is not in slot. */
111 static inline bool is_noslot_pfn(kvm_pfn_t pfn)
113 return pfn == KVM_PFN_NOSLOT;
117 * architectures with KVM_HVA_ERR_BAD other than PAGE_OFFSET (e.g. s390)
118 * provide own defines and kvm_is_error_hva
120 #ifndef KVM_HVA_ERR_BAD
122 #define KVM_HVA_ERR_BAD (PAGE_OFFSET)
123 #define KVM_HVA_ERR_RO_BAD (PAGE_OFFSET + PAGE_SIZE)
125 static inline bool kvm_is_error_hva(unsigned long addr)
127 return addr >= PAGE_OFFSET;
132 #define KVM_ERR_PTR_BAD_PAGE (ERR_PTR(-ENOENT))
134 static inline bool is_error_page(struct page *page)
139 #define KVM_REQUEST_MASK GENMASK(7,0)
140 #define KVM_REQUEST_NO_WAKEUP BIT(8)
141 #define KVM_REQUEST_WAIT BIT(9)
143 * Architecture-independent vcpu->requests bit members
144 * Bits 4-7 are reserved for more arch-independent bits.
146 #define KVM_REQ_TLB_FLUSH (0 | KVM_REQUEST_WAIT | KVM_REQUEST_NO_WAKEUP)
147 #define KVM_REQ_MMU_RELOAD (1 | KVM_REQUEST_WAIT | KVM_REQUEST_NO_WAKEUP)
148 #define KVM_REQ_PENDING_TIMER 2
149 #define KVM_REQ_UNHALT 3
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 void 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;
323 struct kvm_dirty_ring dirty_ring;
326 static inline int kvm_vcpu_exiting_guest_mode(struct kvm_vcpu *vcpu)
329 * The memory barrier ensures a previous write to vcpu->requests cannot
330 * be reordered with the read of vcpu->mode. It pairs with the general
331 * memory barrier following the write of vcpu->mode in VCPU RUN.
333 smp_mb__before_atomic();
334 return cmpxchg(&vcpu->mode, IN_GUEST_MODE, EXITING_GUEST_MODE);
338 * Some of the bitops functions do not support too long bitmaps.
339 * This number must be determined not to exceed such limits.
341 #define KVM_MEM_MAX_NR_PAGES ((1UL << 31) - 1)
343 struct kvm_memory_slot {
345 unsigned long npages;
346 unsigned long *dirty_bitmap;
347 struct kvm_arch_memory_slot arch;
348 unsigned long userspace_addr;
354 static inline bool kvm_slot_dirty_track_enabled(struct kvm_memory_slot *slot)
356 return slot->flags & KVM_MEM_LOG_DIRTY_PAGES;
359 static inline unsigned long kvm_dirty_bitmap_bytes(struct kvm_memory_slot *memslot)
361 return ALIGN(memslot->npages, BITS_PER_LONG) / 8;
364 static inline unsigned long *kvm_second_dirty_bitmap(struct kvm_memory_slot *memslot)
366 unsigned long len = kvm_dirty_bitmap_bytes(memslot);
368 return memslot->dirty_bitmap + len / sizeof(*memslot->dirty_bitmap);
371 #ifndef KVM_DIRTY_LOG_MANUAL_CAPS
372 #define KVM_DIRTY_LOG_MANUAL_CAPS KVM_DIRTY_LOG_MANUAL_PROTECT_ENABLE
375 struct kvm_s390_adapter_int {
388 struct kvm_kernel_irq_routing_entry {
391 int (*set)(struct kvm_kernel_irq_routing_entry *e,
392 struct kvm *kvm, int irq_source_id, int level,
406 struct kvm_s390_adapter_int adapter;
407 struct kvm_hv_sint hv_sint;
409 struct hlist_node link;
412 #ifdef CONFIG_HAVE_KVM_IRQ_ROUTING
413 struct kvm_irq_routing_table {
414 int chip[KVM_NR_IRQCHIPS][KVM_IRQCHIP_NUM_PINS];
417 * Array indexed by gsi. Each entry contains list of irq chips
418 * the gsi is connected to.
420 struct hlist_head map[];
424 #ifndef KVM_PRIVATE_MEM_SLOTS
425 #define KVM_PRIVATE_MEM_SLOTS 0
428 #ifndef KVM_MEM_SLOTS_NUM
429 #define KVM_MEM_SLOTS_NUM (KVM_USER_MEM_SLOTS + KVM_PRIVATE_MEM_SLOTS)
432 #ifndef __KVM_VCPU_MULTIPLE_ADDRESS_SPACE
433 static inline int kvm_arch_vcpu_memslots_id(struct kvm_vcpu *vcpu)
441 * memslots are not sorted by id anymore, please use id_to_memslot()
442 * to get the memslot by its id.
444 struct kvm_memslots {
446 /* The mapping table from slot id to the index in memslots[]. */
447 short id_to_index[KVM_MEM_SLOTS_NUM];
450 struct kvm_memory_slot memslots[];
455 struct mutex slots_lock;
456 struct mm_struct *mm; /* userspace tied to this vm */
457 struct kvm_memslots __rcu *memslots[KVM_ADDRESS_SPACE_NUM];
458 struct kvm_vcpu *vcpus[KVM_MAX_VCPUS];
461 * created_vcpus is protected by kvm->lock, and is incremented
462 * at the beginning of KVM_CREATE_VCPU. online_vcpus is only
463 * incremented after storing the kvm_vcpu pointer in vcpus,
464 * and is accessed atomically.
466 atomic_t online_vcpus;
468 int last_boosted_vcpu;
469 struct list_head vm_list;
471 struct kvm_io_bus __rcu *buses[KVM_NR_BUSES];
472 #ifdef CONFIG_HAVE_KVM_EVENTFD
475 struct list_head items;
476 struct list_head resampler_list;
477 struct mutex resampler_lock;
479 struct list_head ioeventfds;
481 struct kvm_vm_stat stat;
482 struct kvm_arch arch;
483 refcount_t users_count;
484 #ifdef CONFIG_KVM_MMIO
485 struct kvm_coalesced_mmio_ring *coalesced_mmio_ring;
486 spinlock_t ring_lock;
487 struct list_head coalesced_zones;
490 struct mutex irq_lock;
491 #ifdef CONFIG_HAVE_KVM_IRQCHIP
493 * Update side is protected by irq_lock.
495 struct kvm_irq_routing_table __rcu *irq_routing;
497 #ifdef CONFIG_HAVE_KVM_IRQFD
498 struct hlist_head irq_ack_notifier_list;
501 #if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
502 struct mmu_notifier mmu_notifier;
503 unsigned long mmu_notifier_seq;
504 long mmu_notifier_count;
507 struct list_head devices;
508 u64 manual_dirty_log_protect;
509 struct dentry *debugfs_dentry;
510 struct kvm_stat_data **debugfs_stat_data;
511 struct srcu_struct srcu;
512 struct srcu_struct irq_srcu;
514 unsigned int max_halt_poll_ns;
518 #define kvm_err(fmt, ...) \
519 pr_err("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
520 #define kvm_info(fmt, ...) \
521 pr_info("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
522 #define kvm_debug(fmt, ...) \
523 pr_debug("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
524 #define kvm_debug_ratelimited(fmt, ...) \
525 pr_debug_ratelimited("kvm [%i]: " fmt, task_pid_nr(current), \
527 #define kvm_pr_unimpl(fmt, ...) \
528 pr_err_ratelimited("kvm [%i]: " fmt, \
529 task_tgid_nr(current), ## __VA_ARGS__)
531 /* The guest did something we don't support. */
532 #define vcpu_unimpl(vcpu, fmt, ...) \
533 kvm_pr_unimpl("vcpu%i, guest rIP: 0x%lx " fmt, \
534 (vcpu)->vcpu_id, kvm_rip_read(vcpu), ## __VA_ARGS__)
536 #define vcpu_debug(vcpu, fmt, ...) \
537 kvm_debug("vcpu%i " fmt, (vcpu)->vcpu_id, ## __VA_ARGS__)
538 #define vcpu_debug_ratelimited(vcpu, fmt, ...) \
539 kvm_debug_ratelimited("vcpu%i " fmt, (vcpu)->vcpu_id, \
541 #define vcpu_err(vcpu, fmt, ...) \
542 kvm_err("vcpu%i " fmt, (vcpu)->vcpu_id, ## __VA_ARGS__)
544 static inline bool kvm_dirty_log_manual_protect_and_init_set(struct kvm *kvm)
546 return !!(kvm->manual_dirty_log_protect & KVM_DIRTY_LOG_INITIALLY_SET);
549 static inline struct kvm_io_bus *kvm_get_bus(struct kvm *kvm, enum kvm_bus idx)
551 return srcu_dereference_check(kvm->buses[idx], &kvm->srcu,
552 lockdep_is_held(&kvm->slots_lock) ||
553 !refcount_read(&kvm->users_count));
556 static inline struct kvm_vcpu *kvm_get_vcpu(struct kvm *kvm, int i)
558 int num_vcpus = atomic_read(&kvm->online_vcpus);
559 i = array_index_nospec(i, num_vcpus);
561 /* Pairs with smp_wmb() in kvm_vm_ioctl_create_vcpu. */
563 return kvm->vcpus[i];
566 #define kvm_for_each_vcpu(idx, vcpup, kvm) \
568 idx < atomic_read(&kvm->online_vcpus) && \
569 (vcpup = kvm_get_vcpu(kvm, idx)) != NULL; \
572 static inline struct kvm_vcpu *kvm_get_vcpu_by_id(struct kvm *kvm, int id)
574 struct kvm_vcpu *vcpu = NULL;
579 if (id < KVM_MAX_VCPUS)
580 vcpu = kvm_get_vcpu(kvm, id);
581 if (vcpu && vcpu->vcpu_id == id)
583 kvm_for_each_vcpu(i, vcpu, kvm)
584 if (vcpu->vcpu_id == id)
589 static inline int kvm_vcpu_get_idx(struct kvm_vcpu *vcpu)
591 return vcpu->vcpu_idx;
594 #define kvm_for_each_memslot(memslot, slots) \
595 for (memslot = &slots->memslots[0]; \
596 memslot < slots->memslots + slots->used_slots; memslot++) \
597 if (WARN_ON_ONCE(!memslot->npages)) { \
600 void kvm_vcpu_destroy(struct kvm_vcpu *vcpu);
602 void vcpu_load(struct kvm_vcpu *vcpu);
603 void vcpu_put(struct kvm_vcpu *vcpu);
605 #ifdef __KVM_HAVE_IOAPIC
606 void kvm_arch_post_irq_ack_notifier_list_update(struct kvm *kvm);
607 void kvm_arch_post_irq_routing_update(struct kvm *kvm);
609 static inline void kvm_arch_post_irq_ack_notifier_list_update(struct kvm *kvm)
612 static inline void kvm_arch_post_irq_routing_update(struct kvm *kvm)
617 #ifdef CONFIG_HAVE_KVM_IRQFD
618 int kvm_irqfd_init(void);
619 void kvm_irqfd_exit(void);
621 static inline int kvm_irqfd_init(void)
626 static inline void kvm_irqfd_exit(void)
630 int kvm_init(void *opaque, unsigned vcpu_size, unsigned vcpu_align,
631 struct module *module);
634 void kvm_get_kvm(struct kvm *kvm);
635 void kvm_put_kvm(struct kvm *kvm);
636 void kvm_put_kvm_no_destroy(struct kvm *kvm);
638 static inline struct kvm_memslots *__kvm_memslots(struct kvm *kvm, int as_id)
640 as_id = array_index_nospec(as_id, KVM_ADDRESS_SPACE_NUM);
641 return srcu_dereference_check(kvm->memslots[as_id], &kvm->srcu,
642 lockdep_is_held(&kvm->slots_lock) ||
643 !refcount_read(&kvm->users_count));
646 static inline struct kvm_memslots *kvm_memslots(struct kvm *kvm)
648 return __kvm_memslots(kvm, 0);
651 static inline struct kvm_memslots *kvm_vcpu_memslots(struct kvm_vcpu *vcpu)
653 int as_id = kvm_arch_vcpu_memslots_id(vcpu);
655 return __kvm_memslots(vcpu->kvm, as_id);
659 struct kvm_memory_slot *id_to_memslot(struct kvm_memslots *slots, int id)
661 int index = slots->id_to_index[id];
662 struct kvm_memory_slot *slot;
667 slot = &slots->memslots[index];
669 WARN_ON(slot->id != id);
674 * KVM_SET_USER_MEMORY_REGION ioctl allows the following operations:
675 * - create a new memory slot
676 * - delete an existing memory slot
677 * - modify an existing memory slot
678 * -- move it in the guest physical memory space
679 * -- just change its flags
681 * Since flags can be changed by some of these operations, the following
682 * differentiation is the best we can do for __kvm_set_memory_region():
691 int kvm_set_memory_region(struct kvm *kvm,
692 const struct kvm_userspace_memory_region *mem);
693 int __kvm_set_memory_region(struct kvm *kvm,
694 const struct kvm_userspace_memory_region *mem);
695 void kvm_arch_free_memslot(struct kvm *kvm, struct kvm_memory_slot *slot);
696 void kvm_arch_memslots_updated(struct kvm *kvm, u64 gen);
697 int kvm_arch_prepare_memory_region(struct kvm *kvm,
698 struct kvm_memory_slot *memslot,
699 const struct kvm_userspace_memory_region *mem,
700 enum kvm_mr_change change);
701 void kvm_arch_commit_memory_region(struct kvm *kvm,
702 const struct kvm_userspace_memory_region *mem,
703 struct kvm_memory_slot *old,
704 const struct kvm_memory_slot *new,
705 enum kvm_mr_change change);
706 /* flush all memory translations */
707 void kvm_arch_flush_shadow_all(struct kvm *kvm);
708 /* flush memory translations pointing to 'slot' */
709 void kvm_arch_flush_shadow_memslot(struct kvm *kvm,
710 struct kvm_memory_slot *slot);
712 int gfn_to_page_many_atomic(struct kvm_memory_slot *slot, gfn_t gfn,
713 struct page **pages, int nr_pages);
715 struct page *gfn_to_page(struct kvm *kvm, gfn_t gfn);
716 unsigned long gfn_to_hva(struct kvm *kvm, gfn_t gfn);
717 unsigned long gfn_to_hva_prot(struct kvm *kvm, gfn_t gfn, bool *writable);
718 unsigned long gfn_to_hva_memslot(struct kvm_memory_slot *slot, gfn_t gfn);
719 unsigned long gfn_to_hva_memslot_prot(struct kvm_memory_slot *slot, gfn_t gfn,
721 void kvm_release_page_clean(struct page *page);
722 void kvm_release_page_dirty(struct page *page);
723 void kvm_set_page_accessed(struct page *page);
725 kvm_pfn_t gfn_to_pfn(struct kvm *kvm, gfn_t gfn);
726 kvm_pfn_t gfn_to_pfn_prot(struct kvm *kvm, gfn_t gfn, bool write_fault,
728 kvm_pfn_t gfn_to_pfn_memslot(struct kvm_memory_slot *slot, gfn_t gfn);
729 kvm_pfn_t gfn_to_pfn_memslot_atomic(struct kvm_memory_slot *slot, gfn_t gfn);
730 kvm_pfn_t __gfn_to_pfn_memslot(struct kvm_memory_slot *slot, gfn_t gfn,
731 bool atomic, bool *async, bool write_fault,
734 void kvm_release_pfn_clean(kvm_pfn_t pfn);
735 void kvm_release_pfn_dirty(kvm_pfn_t pfn);
736 void kvm_set_pfn_dirty(kvm_pfn_t pfn);
737 void kvm_set_pfn_accessed(kvm_pfn_t pfn);
738 void kvm_get_pfn(kvm_pfn_t pfn);
740 void kvm_release_pfn(kvm_pfn_t pfn, bool dirty, struct gfn_to_pfn_cache *cache);
741 int kvm_read_guest_page(struct kvm *kvm, gfn_t gfn, void *data, int offset,
743 int kvm_read_guest(struct kvm *kvm, gpa_t gpa, void *data, unsigned long len);
744 int kvm_read_guest_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
745 void *data, unsigned long len);
746 int kvm_read_guest_offset_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
747 void *data, unsigned int offset,
749 int kvm_write_guest_page(struct kvm *kvm, gfn_t gfn, const void *data,
750 int offset, int len);
751 int kvm_write_guest(struct kvm *kvm, gpa_t gpa, const void *data,
753 int kvm_write_guest_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
754 void *data, unsigned long len);
755 int kvm_write_guest_offset_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
756 void *data, unsigned int offset,
758 int kvm_gfn_to_hva_cache_init(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
759 gpa_t gpa, unsigned long len);
761 #define __kvm_get_guest(kvm, gfn, offset, v) \
763 unsigned long __addr = gfn_to_hva(kvm, gfn); \
764 typeof(v) __user *__uaddr = (typeof(__uaddr))(__addr + offset); \
765 int __ret = -EFAULT; \
767 if (!kvm_is_error_hva(__addr)) \
768 __ret = get_user(v, __uaddr); \
772 #define kvm_get_guest(kvm, gpa, v) \
775 struct kvm *__kvm = kvm; \
777 __kvm_get_guest(__kvm, __gpa >> PAGE_SHIFT, \
778 offset_in_page(__gpa), v); \
781 #define __kvm_put_guest(kvm, gfn, offset, v) \
783 unsigned long __addr = gfn_to_hva(kvm, gfn); \
784 typeof(v) __user *__uaddr = (typeof(__uaddr))(__addr + offset); \
785 int __ret = -EFAULT; \
787 if (!kvm_is_error_hva(__addr)) \
788 __ret = put_user(v, __uaddr); \
790 mark_page_dirty(kvm, gfn); \
794 #define kvm_put_guest(kvm, gpa, v) \
797 struct kvm *__kvm = kvm; \
799 __kvm_put_guest(__kvm, __gpa >> PAGE_SHIFT, \
800 offset_in_page(__gpa), v); \
803 int kvm_clear_guest(struct kvm *kvm, gpa_t gpa, unsigned long len);
804 struct kvm_memory_slot *gfn_to_memslot(struct kvm *kvm, gfn_t gfn);
805 bool kvm_is_visible_gfn(struct kvm *kvm, gfn_t gfn);
806 bool kvm_vcpu_is_visible_gfn(struct kvm_vcpu *vcpu, gfn_t gfn);
807 unsigned long kvm_host_page_size(struct kvm_vcpu *vcpu, gfn_t gfn);
808 void mark_page_dirty_in_slot(struct kvm *kvm, struct kvm_memory_slot *memslot, gfn_t gfn);
809 void mark_page_dirty(struct kvm *kvm, gfn_t gfn);
811 struct kvm_memslots *kvm_vcpu_memslots(struct kvm_vcpu *vcpu);
812 struct kvm_memory_slot *kvm_vcpu_gfn_to_memslot(struct kvm_vcpu *vcpu, gfn_t gfn);
813 kvm_pfn_t kvm_vcpu_gfn_to_pfn_atomic(struct kvm_vcpu *vcpu, gfn_t gfn);
814 kvm_pfn_t kvm_vcpu_gfn_to_pfn(struct kvm_vcpu *vcpu, gfn_t gfn);
815 int kvm_vcpu_map(struct kvm_vcpu *vcpu, gpa_t gpa, struct kvm_host_map *map);
816 int kvm_map_gfn(struct kvm_vcpu *vcpu, gfn_t gfn, struct kvm_host_map *map,
817 struct gfn_to_pfn_cache *cache, bool atomic);
818 struct page *kvm_vcpu_gfn_to_page(struct kvm_vcpu *vcpu, gfn_t gfn);
819 void kvm_vcpu_unmap(struct kvm_vcpu *vcpu, struct kvm_host_map *map, bool dirty);
820 int kvm_unmap_gfn(struct kvm_vcpu *vcpu, struct kvm_host_map *map,
821 struct gfn_to_pfn_cache *cache, bool dirty, bool atomic);
822 unsigned long kvm_vcpu_gfn_to_hva(struct kvm_vcpu *vcpu, gfn_t gfn);
823 unsigned long kvm_vcpu_gfn_to_hva_prot(struct kvm_vcpu *vcpu, gfn_t gfn, bool *writable);
824 int kvm_vcpu_read_guest_page(struct kvm_vcpu *vcpu, gfn_t gfn, void *data, int offset,
826 int kvm_vcpu_read_guest_atomic(struct kvm_vcpu *vcpu, gpa_t gpa, void *data,
828 int kvm_vcpu_read_guest(struct kvm_vcpu *vcpu, gpa_t gpa, void *data,
830 int kvm_vcpu_write_guest_page(struct kvm_vcpu *vcpu, gfn_t gfn, const void *data,
831 int offset, int len);
832 int kvm_vcpu_write_guest(struct kvm_vcpu *vcpu, gpa_t gpa, const void *data,
834 void kvm_vcpu_mark_page_dirty(struct kvm_vcpu *vcpu, gfn_t gfn);
836 void kvm_sigset_activate(struct kvm_vcpu *vcpu);
837 void kvm_sigset_deactivate(struct kvm_vcpu *vcpu);
839 void kvm_vcpu_block(struct kvm_vcpu *vcpu);
840 void kvm_arch_vcpu_blocking(struct kvm_vcpu *vcpu);
841 void kvm_arch_vcpu_unblocking(struct kvm_vcpu *vcpu);
842 bool kvm_vcpu_wake_up(struct kvm_vcpu *vcpu);
843 void kvm_vcpu_kick(struct kvm_vcpu *vcpu);
844 int kvm_vcpu_yield_to(struct kvm_vcpu *target);
845 void kvm_vcpu_on_spin(struct kvm_vcpu *vcpu, bool usermode_vcpu_not_eligible);
847 void kvm_flush_remote_tlbs(struct kvm *kvm);
848 void kvm_reload_remote_mmus(struct kvm *kvm);
850 #ifdef KVM_ARCH_NR_OBJS_PER_MEMORY_CACHE
851 int kvm_mmu_topup_memory_cache(struct kvm_mmu_memory_cache *mc, int min);
852 int kvm_mmu_memory_cache_nr_free_objects(struct kvm_mmu_memory_cache *mc);
853 void kvm_mmu_free_memory_cache(struct kvm_mmu_memory_cache *mc);
854 void *kvm_mmu_memory_cache_alloc(struct kvm_mmu_memory_cache *mc);
857 bool kvm_make_vcpus_request_mask(struct kvm *kvm, unsigned int req,
858 struct kvm_vcpu *except,
859 unsigned long *vcpu_bitmap, cpumask_var_t tmp);
860 bool kvm_make_all_cpus_request(struct kvm *kvm, unsigned int req);
861 bool kvm_make_all_cpus_request_except(struct kvm *kvm, unsigned int req,
862 struct kvm_vcpu *except);
863 bool kvm_make_cpus_request_mask(struct kvm *kvm, unsigned int req,
864 unsigned long *vcpu_bitmap);
866 long kvm_arch_dev_ioctl(struct file *filp,
867 unsigned int ioctl, unsigned long arg);
868 long kvm_arch_vcpu_ioctl(struct file *filp,
869 unsigned int ioctl, unsigned long arg);
870 vm_fault_t kvm_arch_vcpu_fault(struct kvm_vcpu *vcpu, struct vm_fault *vmf);
872 int kvm_vm_ioctl_check_extension(struct kvm *kvm, long ext);
874 void kvm_arch_mmu_enable_log_dirty_pt_masked(struct kvm *kvm,
875 struct kvm_memory_slot *slot,
878 void kvm_arch_sync_dirty_log(struct kvm *kvm, struct kvm_memory_slot *memslot);
880 #ifdef CONFIG_KVM_GENERIC_DIRTYLOG_READ_PROTECT
881 void kvm_arch_flush_remote_tlbs_memslot(struct kvm *kvm,
882 struct kvm_memory_slot *memslot);
883 #else /* !CONFIG_KVM_GENERIC_DIRTYLOG_READ_PROTECT */
884 int kvm_vm_ioctl_get_dirty_log(struct kvm *kvm, struct kvm_dirty_log *log);
885 int kvm_get_dirty_log(struct kvm *kvm, struct kvm_dirty_log *log,
886 int *is_dirty, struct kvm_memory_slot **memslot);
889 int kvm_vm_ioctl_irq_line(struct kvm *kvm, struct kvm_irq_level *irq_level,
891 int kvm_vm_ioctl_enable_cap(struct kvm *kvm,
892 struct kvm_enable_cap *cap);
893 long kvm_arch_vm_ioctl(struct file *filp,
894 unsigned int ioctl, unsigned long arg);
896 int kvm_arch_vcpu_ioctl_get_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu);
897 int kvm_arch_vcpu_ioctl_set_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu);
899 int kvm_arch_vcpu_ioctl_translate(struct kvm_vcpu *vcpu,
900 struct kvm_translation *tr);
902 int kvm_arch_vcpu_ioctl_get_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs);
903 int kvm_arch_vcpu_ioctl_set_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs);
904 int kvm_arch_vcpu_ioctl_get_sregs(struct kvm_vcpu *vcpu,
905 struct kvm_sregs *sregs);
906 int kvm_arch_vcpu_ioctl_set_sregs(struct kvm_vcpu *vcpu,
907 struct kvm_sregs *sregs);
908 int kvm_arch_vcpu_ioctl_get_mpstate(struct kvm_vcpu *vcpu,
909 struct kvm_mp_state *mp_state);
910 int kvm_arch_vcpu_ioctl_set_mpstate(struct kvm_vcpu *vcpu,
911 struct kvm_mp_state *mp_state);
912 int kvm_arch_vcpu_ioctl_set_guest_debug(struct kvm_vcpu *vcpu,
913 struct kvm_guest_debug *dbg);
914 int kvm_arch_vcpu_ioctl_run(struct kvm_vcpu *vcpu);
916 int kvm_arch_init(void *opaque);
917 void kvm_arch_exit(void);
919 void kvm_arch_sched_in(struct kvm_vcpu *vcpu, int cpu);
921 void kvm_arch_vcpu_load(struct kvm_vcpu *vcpu, int cpu);
922 void kvm_arch_vcpu_put(struct kvm_vcpu *vcpu);
923 int kvm_arch_vcpu_precreate(struct kvm *kvm, unsigned int id);
924 int kvm_arch_vcpu_create(struct kvm_vcpu *vcpu);
925 void kvm_arch_vcpu_postcreate(struct kvm_vcpu *vcpu);
926 void kvm_arch_vcpu_destroy(struct kvm_vcpu *vcpu);
928 #ifdef __KVM_HAVE_ARCH_VCPU_DEBUGFS
929 void kvm_arch_create_vcpu_debugfs(struct kvm_vcpu *vcpu, struct dentry *debugfs_dentry);
932 int kvm_arch_hardware_enable(void);
933 void kvm_arch_hardware_disable(void);
934 int kvm_arch_hardware_setup(void *opaque);
935 void kvm_arch_hardware_unsetup(void);
936 int kvm_arch_check_processor_compat(void *opaque);
937 int kvm_arch_vcpu_runnable(struct kvm_vcpu *vcpu);
938 bool kvm_arch_vcpu_in_kernel(struct kvm_vcpu *vcpu);
939 int kvm_arch_vcpu_should_kick(struct kvm_vcpu *vcpu);
940 bool kvm_arch_dy_runnable(struct kvm_vcpu *vcpu);
941 int kvm_arch_post_init_vm(struct kvm *kvm);
942 void kvm_arch_pre_destroy_vm(struct kvm *kvm);
944 #ifndef __KVM_HAVE_ARCH_VM_ALLOC
946 * All architectures that want to use vzalloc currently also
947 * need their own kvm_arch_alloc_vm implementation.
949 static inline struct kvm *kvm_arch_alloc_vm(void)
951 return kzalloc(sizeof(struct kvm), GFP_KERNEL);
954 static inline void kvm_arch_free_vm(struct kvm *kvm)
960 #ifndef __KVM_HAVE_ARCH_FLUSH_REMOTE_TLB
961 static inline int kvm_arch_flush_remote_tlb(struct kvm *kvm)
967 #ifdef __KVM_HAVE_ARCH_NONCOHERENT_DMA
968 void kvm_arch_register_noncoherent_dma(struct kvm *kvm);
969 void kvm_arch_unregister_noncoherent_dma(struct kvm *kvm);
970 bool kvm_arch_has_noncoherent_dma(struct kvm *kvm);
972 static inline void kvm_arch_register_noncoherent_dma(struct kvm *kvm)
976 static inline void kvm_arch_unregister_noncoherent_dma(struct kvm *kvm)
980 static inline bool kvm_arch_has_noncoherent_dma(struct kvm *kvm)
985 #ifdef __KVM_HAVE_ARCH_ASSIGNED_DEVICE
986 void kvm_arch_start_assignment(struct kvm *kvm);
987 void kvm_arch_end_assignment(struct kvm *kvm);
988 bool kvm_arch_has_assigned_device(struct kvm *kvm);
990 static inline void kvm_arch_start_assignment(struct kvm *kvm)
994 static inline void kvm_arch_end_assignment(struct kvm *kvm)
998 static inline bool kvm_arch_has_assigned_device(struct kvm *kvm)
1004 static inline struct rcuwait *kvm_arch_vcpu_get_wait(struct kvm_vcpu *vcpu)
1006 #ifdef __KVM_HAVE_ARCH_WQP
1007 return vcpu->arch.waitp;
1013 #ifdef __KVM_HAVE_ARCH_INTC_INITIALIZED
1015 * returns true if the virtual interrupt controller is initialized and
1016 * ready to accept virtual IRQ. On some architectures the virtual interrupt
1017 * controller is dynamically instantiated and this is not always true.
1019 bool kvm_arch_intc_initialized(struct kvm *kvm);
1021 static inline bool kvm_arch_intc_initialized(struct kvm *kvm)
1027 int kvm_arch_init_vm(struct kvm *kvm, unsigned long type);
1028 void kvm_arch_destroy_vm(struct kvm *kvm);
1029 void kvm_arch_sync_events(struct kvm *kvm);
1031 int kvm_cpu_has_pending_timer(struct kvm_vcpu *vcpu);
1033 bool kvm_is_reserved_pfn(kvm_pfn_t pfn);
1034 bool kvm_is_zone_device_pfn(kvm_pfn_t pfn);
1035 bool kvm_is_transparent_hugepage(kvm_pfn_t pfn);
1037 struct kvm_irq_ack_notifier {
1038 struct hlist_node link;
1040 void (*irq_acked)(struct kvm_irq_ack_notifier *kian);
1043 int kvm_irq_map_gsi(struct kvm *kvm,
1044 struct kvm_kernel_irq_routing_entry *entries, int gsi);
1045 int kvm_irq_map_chip_pin(struct kvm *kvm, unsigned irqchip, unsigned pin);
1047 int kvm_set_irq(struct kvm *kvm, int irq_source_id, u32 irq, int level,
1049 int kvm_set_msi(struct kvm_kernel_irq_routing_entry *irq_entry, struct kvm *kvm,
1050 int irq_source_id, int level, bool line_status);
1051 int kvm_arch_set_irq_inatomic(struct kvm_kernel_irq_routing_entry *e,
1052 struct kvm *kvm, int irq_source_id,
1053 int level, bool line_status);
1054 bool kvm_irq_has_notifier(struct kvm *kvm, unsigned irqchip, unsigned pin);
1055 void kvm_notify_acked_gsi(struct kvm *kvm, int gsi);
1056 void kvm_notify_acked_irq(struct kvm *kvm, unsigned irqchip, unsigned pin);
1057 void kvm_register_irq_ack_notifier(struct kvm *kvm,
1058 struct kvm_irq_ack_notifier *kian);
1059 void kvm_unregister_irq_ack_notifier(struct kvm *kvm,
1060 struct kvm_irq_ack_notifier *kian);
1061 int kvm_request_irq_source_id(struct kvm *kvm);
1062 void kvm_free_irq_source_id(struct kvm *kvm, int irq_source_id);
1063 bool kvm_arch_irqfd_allowed(struct kvm *kvm, struct kvm_irqfd *args);
1066 * search_memslots() and __gfn_to_memslot() are here because they are
1067 * used in non-modular code in arch/powerpc/kvm/book3s_hv_rm_mmu.c.
1068 * gfn_to_memslot() itself isn't here as an inline because that would
1069 * bloat other code too much.
1071 * IMPORTANT: Slots are sorted from highest GFN to lowest GFN!
1073 static inline struct kvm_memory_slot *
1074 search_memslots(struct kvm_memslots *slots, gfn_t gfn)
1076 int start = 0, end = slots->used_slots;
1077 int slot = atomic_read(&slots->lru_slot);
1078 struct kvm_memory_slot *memslots = slots->memslots;
1080 if (unlikely(!slots->used_slots))
1083 if (gfn >= memslots[slot].base_gfn &&
1084 gfn < memslots[slot].base_gfn + memslots[slot].npages)
1085 return &memslots[slot];
1087 while (start < end) {
1088 slot = start + (end - start) / 2;
1090 if (gfn >= memslots[slot].base_gfn)
1096 if (start < slots->used_slots && gfn >= memslots[start].base_gfn &&
1097 gfn < memslots[start].base_gfn + memslots[start].npages) {
1098 atomic_set(&slots->lru_slot, start);
1099 return &memslots[start];
1105 static inline struct kvm_memory_slot *
1106 __gfn_to_memslot(struct kvm_memslots *slots, gfn_t gfn)
1108 return search_memslots(slots, gfn);
1111 static inline unsigned long
1112 __gfn_to_hva_memslot(struct kvm_memory_slot *slot, gfn_t gfn)
1114 return slot->userspace_addr + (gfn - slot->base_gfn) * PAGE_SIZE;
1117 static inline int memslot_id(struct kvm *kvm, gfn_t gfn)
1119 return gfn_to_memslot(kvm, gfn)->id;
1123 hva_to_gfn_memslot(unsigned long hva, struct kvm_memory_slot *slot)
1125 gfn_t gfn_offset = (hva - slot->userspace_addr) >> PAGE_SHIFT;
1127 return slot->base_gfn + gfn_offset;
1130 static inline gpa_t gfn_to_gpa(gfn_t gfn)
1132 return (gpa_t)gfn << PAGE_SHIFT;
1135 static inline gfn_t gpa_to_gfn(gpa_t gpa)
1137 return (gfn_t)(gpa >> PAGE_SHIFT);
1140 static inline hpa_t pfn_to_hpa(kvm_pfn_t pfn)
1142 return (hpa_t)pfn << PAGE_SHIFT;
1145 static inline struct page *kvm_vcpu_gpa_to_page(struct kvm_vcpu *vcpu,
1148 return kvm_vcpu_gfn_to_page(vcpu, gpa_to_gfn(gpa));
1151 static inline bool kvm_is_error_gpa(struct kvm *kvm, gpa_t gpa)
1153 unsigned long hva = gfn_to_hva(kvm, gpa_to_gfn(gpa));
1155 return kvm_is_error_hva(hva);
1158 enum kvm_stat_kind {
1163 struct kvm_stat_data {
1165 struct kvm_stats_debugfs_item *dbgfs_item;
1168 struct kvm_stats_debugfs_item {
1171 enum kvm_stat_kind kind;
1175 #define KVM_DBGFS_GET_MODE(dbgfs_item) \
1176 ((dbgfs_item)->mode ? (dbgfs_item)->mode : 0644)
1178 #define VM_STAT(n, x, ...) \
1179 { n, offsetof(struct kvm, stat.x), KVM_STAT_VM, ## __VA_ARGS__ }
1180 #define VCPU_STAT(n, x, ...) \
1181 { n, offsetof(struct kvm_vcpu, stat.x), KVM_STAT_VCPU, ## __VA_ARGS__ }
1183 extern struct kvm_stats_debugfs_item debugfs_entries[];
1184 extern struct dentry *kvm_debugfs_dir;
1186 #if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
1187 static inline int mmu_notifier_retry(struct kvm *kvm, unsigned long mmu_seq)
1189 if (unlikely(kvm->mmu_notifier_count))
1192 * Ensure the read of mmu_notifier_count happens before the read
1193 * of mmu_notifier_seq. This interacts with the smp_wmb() in
1194 * mmu_notifier_invalidate_range_end to make sure that the caller
1195 * either sees the old (non-zero) value of mmu_notifier_count or
1196 * the new (incremented) value of mmu_notifier_seq.
1197 * PowerPC Book3s HV KVM calls this under a per-page lock
1198 * rather than under kvm->mmu_lock, for scalability, so
1199 * can't rely on kvm->mmu_lock to keep things ordered.
1202 if (kvm->mmu_notifier_seq != mmu_seq)
1208 #ifdef CONFIG_HAVE_KVM_IRQ_ROUTING
1210 #define KVM_MAX_IRQ_ROUTES 4096 /* might need extension/rework in the future */
1212 bool kvm_arch_can_set_irq_routing(struct kvm *kvm);
1213 int kvm_set_irq_routing(struct kvm *kvm,
1214 const struct kvm_irq_routing_entry *entries,
1217 int kvm_set_routing_entry(struct kvm *kvm,
1218 struct kvm_kernel_irq_routing_entry *e,
1219 const struct kvm_irq_routing_entry *ue);
1220 void kvm_free_irq_routing(struct kvm *kvm);
1224 static inline void kvm_free_irq_routing(struct kvm *kvm) {}
1228 int kvm_send_userspace_msi(struct kvm *kvm, struct kvm_msi *msi);
1230 #ifdef CONFIG_HAVE_KVM_EVENTFD
1232 void kvm_eventfd_init(struct kvm *kvm);
1233 int kvm_ioeventfd(struct kvm *kvm, struct kvm_ioeventfd *args);
1235 #ifdef CONFIG_HAVE_KVM_IRQFD
1236 int kvm_irqfd(struct kvm *kvm, struct kvm_irqfd *args);
1237 void kvm_irqfd_release(struct kvm *kvm);
1238 void kvm_irq_routing_update(struct kvm *);
1240 static inline int kvm_irqfd(struct kvm *kvm, struct kvm_irqfd *args)
1245 static inline void kvm_irqfd_release(struct kvm *kvm) {}
1250 static inline void kvm_eventfd_init(struct kvm *kvm) {}
1252 static inline int kvm_irqfd(struct kvm *kvm, struct kvm_irqfd *args)
1257 static inline void kvm_irqfd_release(struct kvm *kvm) {}
1259 #ifdef CONFIG_HAVE_KVM_IRQCHIP
1260 static inline void kvm_irq_routing_update(struct kvm *kvm)
1265 static inline int kvm_ioeventfd(struct kvm *kvm, struct kvm_ioeventfd *args)
1270 #endif /* CONFIG_HAVE_KVM_EVENTFD */
1272 void kvm_arch_irq_routing_update(struct kvm *kvm);
1274 static inline void kvm_make_request(int req, struct kvm_vcpu *vcpu)
1277 * Ensure the rest of the request is published to kvm_check_request's
1278 * caller. Paired with the smp_mb__after_atomic in kvm_check_request.
1281 set_bit(req & KVM_REQUEST_MASK, (void *)&vcpu->requests);
1284 static inline bool kvm_request_pending(struct kvm_vcpu *vcpu)
1286 return READ_ONCE(vcpu->requests);
1289 static inline bool kvm_test_request(int req, struct kvm_vcpu *vcpu)
1291 return test_bit(req & KVM_REQUEST_MASK, (void *)&vcpu->requests);
1294 static inline void kvm_clear_request(int req, struct kvm_vcpu *vcpu)
1296 clear_bit(req & KVM_REQUEST_MASK, (void *)&vcpu->requests);
1299 static inline bool kvm_check_request(int req, struct kvm_vcpu *vcpu)
1301 if (kvm_test_request(req, vcpu)) {
1302 kvm_clear_request(req, vcpu);
1305 * Ensure the rest of the request is visible to kvm_check_request's
1306 * caller. Paired with the smp_wmb in kvm_make_request.
1308 smp_mb__after_atomic();
1315 extern bool kvm_rebooting;
1317 extern unsigned int halt_poll_ns;
1318 extern unsigned int halt_poll_ns_grow;
1319 extern unsigned int halt_poll_ns_grow_start;
1320 extern unsigned int halt_poll_ns_shrink;
1323 const struct kvm_device_ops *ops;
1326 struct list_head vm_node;
1329 /* create, destroy, and name are mandatory */
1330 struct kvm_device_ops {
1334 * create is called holding kvm->lock and any operations not suitable
1335 * to do while holding the lock should be deferred to init (see
1338 int (*create)(struct kvm_device *dev, u32 type);
1341 * init is called after create if create is successful and is called
1342 * outside of holding kvm->lock.
1344 void (*init)(struct kvm_device *dev);
1347 * Destroy is responsible for freeing dev.
1349 * Destroy may be called before or after destructors are called
1350 * on emulated I/O regions, depending on whether a reference is
1351 * held by a vcpu or other kvm component that gets destroyed
1352 * after the emulated I/O.
1354 void (*destroy)(struct kvm_device *dev);
1357 * Release is an alternative method to free the device. It is
1358 * called when the device file descriptor is closed. Once
1359 * release is called, the destroy method will not be called
1360 * anymore as the device is removed from the device list of
1361 * the VM. kvm->lock is held.
1363 void (*release)(struct kvm_device *dev);
1365 int (*set_attr)(struct kvm_device *dev, struct kvm_device_attr *attr);
1366 int (*get_attr)(struct kvm_device *dev, struct kvm_device_attr *attr);
1367 int (*has_attr)(struct kvm_device *dev, struct kvm_device_attr *attr);
1368 long (*ioctl)(struct kvm_device *dev, unsigned int ioctl,
1370 int (*mmap)(struct kvm_device *dev, struct vm_area_struct *vma);
1373 void kvm_device_get(struct kvm_device *dev);
1374 void kvm_device_put(struct kvm_device *dev);
1375 struct kvm_device *kvm_device_from_filp(struct file *filp);
1376 int kvm_register_device_ops(const struct kvm_device_ops *ops, u32 type);
1377 void kvm_unregister_device_ops(u32 type);
1379 extern struct kvm_device_ops kvm_mpic_ops;
1380 extern struct kvm_device_ops kvm_arm_vgic_v2_ops;
1381 extern struct kvm_device_ops kvm_arm_vgic_v3_ops;
1383 #ifdef CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT
1385 static inline void kvm_vcpu_set_in_spin_loop(struct kvm_vcpu *vcpu, bool val)
1387 vcpu->spin_loop.in_spin_loop = val;
1389 static inline void kvm_vcpu_set_dy_eligible(struct kvm_vcpu *vcpu, bool val)
1391 vcpu->spin_loop.dy_eligible = val;
1394 #else /* !CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT */
1396 static inline void kvm_vcpu_set_in_spin_loop(struct kvm_vcpu *vcpu, bool val)
1400 static inline void kvm_vcpu_set_dy_eligible(struct kvm_vcpu *vcpu, bool val)
1403 #endif /* CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT */
1405 static inline bool kvm_is_visible_memslot(struct kvm_memory_slot *memslot)
1407 return (memslot && memslot->id < KVM_USER_MEM_SLOTS &&
1408 !(memslot->flags & KVM_MEMSLOT_INVALID));
1411 struct kvm_vcpu *kvm_get_running_vcpu(void);
1412 struct kvm_vcpu * __percpu *kvm_get_running_vcpus(void);
1414 #ifdef CONFIG_HAVE_KVM_IRQ_BYPASS
1415 bool kvm_arch_has_irq_bypass(void);
1416 int kvm_arch_irq_bypass_add_producer(struct irq_bypass_consumer *,
1417 struct irq_bypass_producer *);
1418 void kvm_arch_irq_bypass_del_producer(struct irq_bypass_consumer *,
1419 struct irq_bypass_producer *);
1420 void kvm_arch_irq_bypass_stop(struct irq_bypass_consumer *);
1421 void kvm_arch_irq_bypass_start(struct irq_bypass_consumer *);
1422 int kvm_arch_update_irqfd_routing(struct kvm *kvm, unsigned int host_irq,
1423 uint32_t guest_irq, bool set);
1424 #endif /* CONFIG_HAVE_KVM_IRQ_BYPASS */
1426 #ifdef CONFIG_HAVE_KVM_INVALID_WAKEUPS
1427 /* If we wakeup during the poll time, was it a sucessful poll? */
1428 static inline bool vcpu_valid_wakeup(struct kvm_vcpu *vcpu)
1430 return vcpu->valid_wakeup;
1434 static inline bool vcpu_valid_wakeup(struct kvm_vcpu *vcpu)
1438 #endif /* CONFIG_HAVE_KVM_INVALID_WAKEUPS */
1440 #ifdef CONFIG_HAVE_KVM_NO_POLL
1441 /* Callback that tells if we must not poll */
1442 bool kvm_arch_no_poll(struct kvm_vcpu *vcpu);
1444 static inline bool kvm_arch_no_poll(struct kvm_vcpu *vcpu)
1448 #endif /* CONFIG_HAVE_KVM_NO_POLL */
1450 #ifdef CONFIG_HAVE_KVM_VCPU_ASYNC_IOCTL
1451 long kvm_arch_vcpu_async_ioctl(struct file *filp,
1452 unsigned int ioctl, unsigned long arg);
1454 static inline long kvm_arch_vcpu_async_ioctl(struct file *filp,
1458 return -ENOIOCTLCMD;
1460 #endif /* CONFIG_HAVE_KVM_VCPU_ASYNC_IOCTL */
1462 void kvm_arch_mmu_notifier_invalidate_range(struct kvm *kvm,
1463 unsigned long start, unsigned long end);
1465 #ifdef CONFIG_HAVE_KVM_VCPU_RUN_PID_CHANGE
1466 int kvm_arch_vcpu_run_pid_change(struct kvm_vcpu *vcpu);
1468 static inline int kvm_arch_vcpu_run_pid_change(struct kvm_vcpu *vcpu)
1472 #endif /* CONFIG_HAVE_KVM_VCPU_RUN_PID_CHANGE */
1474 typedef int (*kvm_vm_thread_fn_t)(struct kvm *kvm, uintptr_t data);
1476 int kvm_vm_create_worker_thread(struct kvm *kvm, kvm_vm_thread_fn_t thread_fn,
1477 uintptr_t data, const char *name,
1478 struct task_struct **thread_ptr);
1480 #ifdef CONFIG_KVM_XFER_TO_GUEST_WORK
1481 static inline void kvm_handle_signal_exit(struct kvm_vcpu *vcpu)
1483 vcpu->run->exit_reason = KVM_EXIT_INTR;
1484 vcpu->stat.signal_exits++;
1486 #endif /* CONFIG_KVM_XFER_TO_GUEST_WORK */
1489 * This defines how many reserved entries we want to keep before we
1490 * kick the vcpu to the userspace to avoid dirty ring full. This
1491 * value can be tuned to higher if e.g. PML is enabled on the host.
1493 #define KVM_DIRTY_RING_RSVD_ENTRIES 64
1495 /* Max number of entries allowed for each kvm dirty ring */
1496 #define KVM_DIRTY_RING_MAX_ENTRIES 65536