1 /* SPDX-License-Identifier: GPL-2.0-only */
6 #include <linux/types.h>
7 #include <linux/hardirq.h>
8 #include <linux/list.h>
9 #include <linux/mutex.h>
10 #include <linux/spinlock.h>
11 #include <linux/signal.h>
12 #include <linux/sched.h>
13 #include <linux/bug.h>
14 #include <linux/minmax.h>
16 #include <linux/mmu_notifier.h>
17 #include <linux/preempt.h>
18 #include <linux/msi.h>
19 #include <linux/slab.h>
20 #include <linux/vmalloc.h>
21 #include <linux/rcupdate.h>
22 #include <linux/ratelimit.h>
23 #include <linux/err.h>
24 #include <linux/irqflags.h>
25 #include <linux/context_tracking.h>
26 #include <linux/irqbypass.h>
27 #include <linux/rcuwait.h>
28 #include <linux/refcount.h>
29 #include <linux/nospec.h>
30 #include <asm/signal.h>
32 #include <linux/kvm.h>
33 #include <linux/kvm_para.h>
35 #include <linux/kvm_types.h>
37 #include <asm/kvm_host.h>
38 #include <linux/kvm_dirty_ring.h>
40 #ifndef KVM_MAX_VCPU_ID
41 #define KVM_MAX_VCPU_ID KVM_MAX_VCPUS
45 * The bit 16 ~ bit 31 of kvm_memory_region::flags are internally used
46 * in kvm, other bits are visible for userspace which are defined in
47 * include/linux/kvm_h.
49 #define KVM_MEMSLOT_INVALID (1UL << 16)
52 * Bit 63 of the memslot generation number is an "update in-progress flag",
53 * e.g. is temporarily set for the duration of install_new_memslots().
54 * This flag effectively creates a unique generation number that is used to
55 * mark cached memslot data, e.g. MMIO accesses, as potentially being stale,
56 * i.e. may (or may not) have come from the previous memslots generation.
58 * This is necessary because the actual memslots update is not atomic with
59 * respect to the generation number update. Updating the generation number
60 * first would allow a vCPU to cache a spte from the old memslots using the
61 * new generation number, and updating the generation number after switching
62 * to the new memslots would allow cache hits using the old generation number
63 * to reference the defunct memslots.
65 * This mechanism is used to prevent getting hits in KVM's caches while a
66 * memslot update is in-progress, and to prevent cache hits *after* updating
67 * the actual generation number against accesses that were inserted into the
68 * cache *before* the memslots were updated.
70 #define KVM_MEMSLOT_GEN_UPDATE_IN_PROGRESS BIT_ULL(63)
72 /* Two fragments for cross MMIO pages. */
73 #define KVM_MAX_MMIO_FRAGMENTS 2
75 #ifndef KVM_ADDRESS_SPACE_NUM
76 #define KVM_ADDRESS_SPACE_NUM 1
80 * For the normal pfn, the highest 12 bits should be zero,
81 * so we can mask bit 62 ~ bit 52 to indicate the error pfn,
82 * mask bit 63 to indicate the noslot pfn.
84 #define KVM_PFN_ERR_MASK (0x7ffULL << 52)
85 #define KVM_PFN_ERR_NOSLOT_MASK (0xfffULL << 52)
86 #define KVM_PFN_NOSLOT (0x1ULL << 63)
88 #define KVM_PFN_ERR_FAULT (KVM_PFN_ERR_MASK)
89 #define KVM_PFN_ERR_HWPOISON (KVM_PFN_ERR_MASK + 1)
90 #define KVM_PFN_ERR_RO_FAULT (KVM_PFN_ERR_MASK + 2)
93 * error pfns indicate that the gfn is in slot but faild to
94 * translate it to pfn on host.
96 static inline bool is_error_pfn(kvm_pfn_t pfn)
98 return !!(pfn & KVM_PFN_ERR_MASK);
102 * error_noslot pfns indicate that the gfn can not be
103 * translated to pfn - it is not in slot or failed to
104 * translate it to pfn.
106 static inline bool is_error_noslot_pfn(kvm_pfn_t pfn)
108 return !!(pfn & KVM_PFN_ERR_NOSLOT_MASK);
111 /* noslot pfn indicates that the gfn is not in slot. */
112 static inline bool is_noslot_pfn(kvm_pfn_t pfn)
114 return pfn == KVM_PFN_NOSLOT;
118 * architectures with KVM_HVA_ERR_BAD other than PAGE_OFFSET (e.g. s390)
119 * provide own defines and kvm_is_error_hva
121 #ifndef KVM_HVA_ERR_BAD
123 #define KVM_HVA_ERR_BAD (PAGE_OFFSET)
124 #define KVM_HVA_ERR_RO_BAD (PAGE_OFFSET + PAGE_SIZE)
126 static inline bool kvm_is_error_hva(unsigned long addr)
128 return addr >= PAGE_OFFSET;
133 #define KVM_ERR_PTR_BAD_PAGE (ERR_PTR(-ENOENT))
135 static inline bool is_error_page(struct page *page)
140 #define KVM_REQUEST_MASK GENMASK(7,0)
141 #define KVM_REQUEST_NO_WAKEUP BIT(8)
142 #define KVM_REQUEST_WAIT BIT(9)
144 * Architecture-independent vcpu->requests bit members
145 * Bits 4-7 are reserved for more arch-independent bits.
147 #define KVM_REQ_TLB_FLUSH (0 | KVM_REQUEST_WAIT | KVM_REQUEST_NO_WAKEUP)
148 #define KVM_REQ_MMU_RELOAD (1 | KVM_REQUEST_WAIT | KVM_REQUEST_NO_WAKEUP)
149 #define KVM_REQ_PENDING_TIMER 2
150 #define KVM_REQ_UNHALT 3
151 #define KVM_REQUEST_ARCH_BASE 8
153 #define KVM_ARCH_REQ_FLAGS(nr, flags) ({ \
154 BUILD_BUG_ON((unsigned)(nr) >= (sizeof_field(struct kvm_vcpu, requests) * 8) - KVM_REQUEST_ARCH_BASE); \
155 (unsigned)(((nr) + KVM_REQUEST_ARCH_BASE) | (flags)); \
157 #define KVM_ARCH_REQ(nr) KVM_ARCH_REQ_FLAGS(nr, 0)
159 #define KVM_USERSPACE_IRQ_SOURCE_ID 0
160 #define KVM_IRQFD_RESAMPLE_IRQ_SOURCE_ID 1
162 extern struct mutex kvm_lock;
163 extern struct list_head vm_list;
165 struct kvm_io_range {
168 struct kvm_io_device *dev;
171 #define NR_IOBUS_DEVS 1000
176 struct kvm_io_range range[];
182 KVM_VIRTIO_CCW_NOTIFY_BUS,
187 int kvm_io_bus_write(struct kvm_vcpu *vcpu, enum kvm_bus bus_idx, gpa_t addr,
188 int len, const void *val);
189 int kvm_io_bus_write_cookie(struct kvm_vcpu *vcpu, enum kvm_bus bus_idx,
190 gpa_t addr, int len, const void *val, long cookie);
191 int kvm_io_bus_read(struct kvm_vcpu *vcpu, enum kvm_bus bus_idx, gpa_t addr,
193 int kvm_io_bus_register_dev(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr,
194 int len, struct kvm_io_device *dev);
195 int kvm_io_bus_unregister_dev(struct kvm *kvm, enum kvm_bus bus_idx,
196 struct kvm_io_device *dev);
197 struct kvm_io_device *kvm_io_bus_get_dev(struct kvm *kvm, enum kvm_bus bus_idx,
200 #ifdef CONFIG_KVM_ASYNC_PF
201 struct kvm_async_pf {
202 struct work_struct work;
203 struct list_head link;
204 struct list_head queue;
205 struct kvm_vcpu *vcpu;
206 struct mm_struct *mm;
209 struct kvm_arch_async_pf arch;
211 bool notpresent_injected;
214 void kvm_clear_async_pf_completion_queue(struct kvm_vcpu *vcpu);
215 void kvm_check_async_pf_completion(struct kvm_vcpu *vcpu);
216 bool kvm_setup_async_pf(struct kvm_vcpu *vcpu, gpa_t cr2_or_gpa,
217 unsigned long hva, struct kvm_arch_async_pf *arch);
218 int kvm_async_pf_wakeup_all(struct kvm_vcpu *vcpu);
221 #ifdef KVM_ARCH_WANT_MMU_NOTIFIER
222 struct kvm_gfn_range {
223 struct kvm_memory_slot *slot;
229 bool kvm_unmap_gfn_range(struct kvm *kvm, struct kvm_gfn_range *range);
230 bool kvm_age_gfn(struct kvm *kvm, struct kvm_gfn_range *range);
231 bool kvm_test_age_gfn(struct kvm *kvm, struct kvm_gfn_range *range);
232 bool kvm_set_spte_gfn(struct kvm *kvm, struct kvm_gfn_range *range);
239 READING_SHADOW_PAGE_TABLES,
242 #define KVM_UNMAPPED_PAGE ((void *) 0x500 + POISON_POINTER_DELTA)
244 struct kvm_host_map {
246 * Only valid if the 'pfn' is managed by the host kernel (i.e. There is
247 * a 'struct page' for it. When using mem= kernel parameter some memory
248 * can be used as guest memory but they are not managed by host
250 * If 'pfn' is not managed by the host kernel, this field is
251 * initialized to KVM_UNMAPPED_PAGE.
260 * Used to check if the mapping is valid or not. Never use 'kvm_host_map'
261 * directly to check for that.
263 static inline bool kvm_vcpu_mapped(struct kvm_host_map *map)
269 * Sometimes a large or cross-page mmio needs to be broken up into separate
270 * exits for userspace servicing.
272 struct kvm_mmio_fragment {
280 #ifdef CONFIG_PREEMPT_NOTIFIERS
281 struct preempt_notifier preempt_notifier;
284 int vcpu_id; /* id given by userspace at creation */
285 int vcpu_idx; /* index in kvm->vcpus array */
289 unsigned long guest_debug;
292 struct list_head blocked_vcpu_list;
298 struct pid __rcu *pid;
301 struct kvm_vcpu_stat stat;
302 unsigned int halt_poll_ns;
305 #ifdef CONFIG_HAS_IOMEM
307 int mmio_read_completed;
309 int mmio_cur_fragment;
310 int mmio_nr_fragments;
311 struct kvm_mmio_fragment mmio_fragments[KVM_MAX_MMIO_FRAGMENTS];
314 #ifdef CONFIG_KVM_ASYNC_PF
317 struct list_head queue;
318 struct list_head done;
323 #ifdef CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT
325 * Cpu relax intercept or pause loop exit optimization
326 * in_spin_loop: set when a vcpu does a pause loop exit
327 * or cpu relax intercepted.
328 * dy_eligible: indicates whether vcpu is eligible for directed yield.
337 struct kvm_vcpu_arch arch;
338 struct kvm_dirty_ring dirty_ring;
341 /* must be called with irqs disabled */
342 static __always_inline void guest_enter_irqoff(void)
345 * This is running in ioctl context so its safe to assume that it's the
346 * stime pending cputime to flush.
348 instrumentation_begin();
349 vtime_account_guest_enter();
350 instrumentation_end();
353 * KVM does not hold any references to rcu protected data when it
354 * switches CPU into a guest mode. In fact switching to a guest mode
355 * is very similar to exiting to userspace from rcu point of view. In
356 * addition CPU may stay in a guest mode for quite a long time (up to
357 * one time slice). Lets treat guest mode as quiescent state, just like
358 * we do with user-mode execution.
360 if (!context_tracking_guest_enter()) {
361 instrumentation_begin();
362 rcu_virt_note_context_switch(smp_processor_id());
363 instrumentation_end();
367 static __always_inline void guest_exit_irqoff(void)
369 context_tracking_guest_exit();
371 instrumentation_begin();
372 /* Flush the guest cputime we spent on the guest */
373 vtime_account_guest_exit();
374 instrumentation_end();
377 static inline void guest_exit(void)
381 local_irq_save(flags);
383 local_irq_restore(flags);
386 static inline int kvm_vcpu_exiting_guest_mode(struct kvm_vcpu *vcpu)
389 * The memory barrier ensures a previous write to vcpu->requests cannot
390 * be reordered with the read of vcpu->mode. It pairs with the general
391 * memory barrier following the write of vcpu->mode in VCPU RUN.
393 smp_mb__before_atomic();
394 return cmpxchg(&vcpu->mode, IN_GUEST_MODE, EXITING_GUEST_MODE);
398 * Some of the bitops functions do not support too long bitmaps.
399 * This number must be determined not to exceed such limits.
401 #define KVM_MEM_MAX_NR_PAGES ((1UL << 31) - 1)
403 struct kvm_memory_slot {
405 unsigned long npages;
406 unsigned long *dirty_bitmap;
407 struct kvm_arch_memory_slot arch;
408 unsigned long userspace_addr;
414 static inline bool kvm_slot_dirty_track_enabled(struct kvm_memory_slot *slot)
416 return slot->flags & KVM_MEM_LOG_DIRTY_PAGES;
419 static inline unsigned long kvm_dirty_bitmap_bytes(struct kvm_memory_slot *memslot)
421 return ALIGN(memslot->npages, BITS_PER_LONG) / 8;
424 static inline unsigned long *kvm_second_dirty_bitmap(struct kvm_memory_slot *memslot)
426 unsigned long len = kvm_dirty_bitmap_bytes(memslot);
428 return memslot->dirty_bitmap + len / sizeof(*memslot->dirty_bitmap);
431 #ifndef KVM_DIRTY_LOG_MANUAL_CAPS
432 #define KVM_DIRTY_LOG_MANUAL_CAPS KVM_DIRTY_LOG_MANUAL_PROTECT_ENABLE
435 struct kvm_s390_adapter_int {
448 struct kvm_kernel_irq_routing_entry {
451 int (*set)(struct kvm_kernel_irq_routing_entry *e,
452 struct kvm *kvm, int irq_source_id, int level,
466 struct kvm_s390_adapter_int adapter;
467 struct kvm_hv_sint hv_sint;
469 struct hlist_node link;
472 #ifdef CONFIG_HAVE_KVM_IRQ_ROUTING
473 struct kvm_irq_routing_table {
474 int chip[KVM_NR_IRQCHIPS][KVM_IRQCHIP_NUM_PINS];
477 * Array indexed by gsi. Each entry contains list of irq chips
478 * the gsi is connected to.
480 struct hlist_head map[];
484 #ifndef KVM_PRIVATE_MEM_SLOTS
485 #define KVM_PRIVATE_MEM_SLOTS 0
488 #define KVM_MEM_SLOTS_NUM SHRT_MAX
489 #define KVM_USER_MEM_SLOTS (KVM_MEM_SLOTS_NUM - KVM_PRIVATE_MEM_SLOTS)
491 #ifndef __KVM_VCPU_MULTIPLE_ADDRESS_SPACE
492 static inline int kvm_arch_vcpu_memslots_id(struct kvm_vcpu *vcpu)
500 * memslots are not sorted by id anymore, please use id_to_memslot()
501 * to get the memslot by its id.
503 struct kvm_memslots {
505 /* The mapping table from slot id to the index in memslots[]. */
506 short id_to_index[KVM_MEM_SLOTS_NUM];
509 struct kvm_memory_slot memslots[];
513 #ifdef KVM_HAVE_MMU_RWLOCK
517 #endif /* KVM_HAVE_MMU_RWLOCK */
519 struct mutex slots_lock;
520 struct mm_struct *mm; /* userspace tied to this vm */
521 struct kvm_memslots __rcu *memslots[KVM_ADDRESS_SPACE_NUM];
522 struct kvm_vcpu *vcpus[KVM_MAX_VCPUS];
525 * created_vcpus is protected by kvm->lock, and is incremented
526 * at the beginning of KVM_CREATE_VCPU. online_vcpus is only
527 * incremented after storing the kvm_vcpu pointer in vcpus,
528 * and is accessed atomically.
530 atomic_t online_vcpus;
532 int last_boosted_vcpu;
533 struct list_head vm_list;
535 struct kvm_io_bus __rcu *buses[KVM_NR_BUSES];
536 #ifdef CONFIG_HAVE_KVM_EVENTFD
539 struct list_head items;
540 struct list_head resampler_list;
541 struct mutex resampler_lock;
543 struct list_head ioeventfds;
545 struct kvm_vm_stat stat;
546 struct kvm_arch arch;
547 refcount_t users_count;
548 #ifdef CONFIG_KVM_MMIO
549 struct kvm_coalesced_mmio_ring *coalesced_mmio_ring;
550 spinlock_t ring_lock;
551 struct list_head coalesced_zones;
554 struct mutex irq_lock;
555 #ifdef CONFIG_HAVE_KVM_IRQCHIP
557 * Update side is protected by irq_lock.
559 struct kvm_irq_routing_table __rcu *irq_routing;
561 #ifdef CONFIG_HAVE_KVM_IRQFD
562 struct hlist_head irq_ack_notifier_list;
565 #if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
566 struct mmu_notifier mmu_notifier;
567 unsigned long mmu_notifier_seq;
568 long mmu_notifier_count;
569 unsigned long mmu_notifier_range_start;
570 unsigned long mmu_notifier_range_end;
573 struct list_head devices;
574 u64 manual_dirty_log_protect;
575 struct dentry *debugfs_dentry;
576 struct kvm_stat_data **debugfs_stat_data;
577 struct srcu_struct srcu;
578 struct srcu_struct irq_srcu;
580 unsigned int max_halt_poll_ns;
584 #define kvm_err(fmt, ...) \
585 pr_err("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
586 #define kvm_info(fmt, ...) \
587 pr_info("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
588 #define kvm_debug(fmt, ...) \
589 pr_debug("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
590 #define kvm_debug_ratelimited(fmt, ...) \
591 pr_debug_ratelimited("kvm [%i]: " fmt, task_pid_nr(current), \
593 #define kvm_pr_unimpl(fmt, ...) \
594 pr_err_ratelimited("kvm [%i]: " fmt, \
595 task_tgid_nr(current), ## __VA_ARGS__)
597 /* The guest did something we don't support. */
598 #define vcpu_unimpl(vcpu, fmt, ...) \
599 kvm_pr_unimpl("vcpu%i, guest rIP: 0x%lx " fmt, \
600 (vcpu)->vcpu_id, kvm_rip_read(vcpu), ## __VA_ARGS__)
602 #define vcpu_debug(vcpu, fmt, ...) \
603 kvm_debug("vcpu%i " fmt, (vcpu)->vcpu_id, ## __VA_ARGS__)
604 #define vcpu_debug_ratelimited(vcpu, fmt, ...) \
605 kvm_debug_ratelimited("vcpu%i " fmt, (vcpu)->vcpu_id, \
607 #define vcpu_err(vcpu, fmt, ...) \
608 kvm_err("vcpu%i " fmt, (vcpu)->vcpu_id, ## __VA_ARGS__)
610 static inline bool kvm_dirty_log_manual_protect_and_init_set(struct kvm *kvm)
612 return !!(kvm->manual_dirty_log_protect & KVM_DIRTY_LOG_INITIALLY_SET);
615 static inline struct kvm_io_bus *kvm_get_bus(struct kvm *kvm, enum kvm_bus idx)
617 return srcu_dereference_check(kvm->buses[idx], &kvm->srcu,
618 lockdep_is_held(&kvm->slots_lock) ||
619 !refcount_read(&kvm->users_count));
622 static inline struct kvm_vcpu *kvm_get_vcpu(struct kvm *kvm, int i)
624 int num_vcpus = atomic_read(&kvm->online_vcpus);
625 i = array_index_nospec(i, num_vcpus);
627 /* Pairs with smp_wmb() in kvm_vm_ioctl_create_vcpu. */
629 return kvm->vcpus[i];
632 #define kvm_for_each_vcpu(idx, vcpup, kvm) \
634 idx < atomic_read(&kvm->online_vcpus) && \
635 (vcpup = kvm_get_vcpu(kvm, idx)) != NULL; \
638 static inline struct kvm_vcpu *kvm_get_vcpu_by_id(struct kvm *kvm, int id)
640 struct kvm_vcpu *vcpu = NULL;
645 if (id < KVM_MAX_VCPUS)
646 vcpu = kvm_get_vcpu(kvm, id);
647 if (vcpu && vcpu->vcpu_id == id)
649 kvm_for_each_vcpu(i, vcpu, kvm)
650 if (vcpu->vcpu_id == id)
655 static inline int kvm_vcpu_get_idx(struct kvm_vcpu *vcpu)
657 return vcpu->vcpu_idx;
660 #define kvm_for_each_memslot(memslot, slots) \
661 for (memslot = &slots->memslots[0]; \
662 memslot < slots->memslots + slots->used_slots; memslot++) \
663 if (WARN_ON_ONCE(!memslot->npages)) { \
666 void kvm_vcpu_destroy(struct kvm_vcpu *vcpu);
668 void vcpu_load(struct kvm_vcpu *vcpu);
669 void vcpu_put(struct kvm_vcpu *vcpu);
671 #ifdef __KVM_HAVE_IOAPIC
672 void kvm_arch_post_irq_ack_notifier_list_update(struct kvm *kvm);
673 void kvm_arch_post_irq_routing_update(struct kvm *kvm);
675 static inline void kvm_arch_post_irq_ack_notifier_list_update(struct kvm *kvm)
678 static inline void kvm_arch_post_irq_routing_update(struct kvm *kvm)
683 #ifdef CONFIG_HAVE_KVM_IRQFD
684 int kvm_irqfd_init(void);
685 void kvm_irqfd_exit(void);
687 static inline int kvm_irqfd_init(void)
692 static inline void kvm_irqfd_exit(void)
696 int kvm_init(void *opaque, unsigned vcpu_size, unsigned vcpu_align,
697 struct module *module);
700 void kvm_get_kvm(struct kvm *kvm);
701 void kvm_put_kvm(struct kvm *kvm);
702 bool file_is_kvm(struct file *file);
703 void kvm_put_kvm_no_destroy(struct kvm *kvm);
705 static inline struct kvm_memslots *__kvm_memslots(struct kvm *kvm, int as_id)
707 as_id = array_index_nospec(as_id, KVM_ADDRESS_SPACE_NUM);
708 return srcu_dereference_check(kvm->memslots[as_id], &kvm->srcu,
709 lockdep_is_held(&kvm->slots_lock) ||
710 !refcount_read(&kvm->users_count));
713 static inline struct kvm_memslots *kvm_memslots(struct kvm *kvm)
715 return __kvm_memslots(kvm, 0);
718 static inline struct kvm_memslots *kvm_vcpu_memslots(struct kvm_vcpu *vcpu)
720 int as_id = kvm_arch_vcpu_memslots_id(vcpu);
722 return __kvm_memslots(vcpu->kvm, as_id);
726 struct kvm_memory_slot *id_to_memslot(struct kvm_memslots *slots, int id)
728 int index = slots->id_to_index[id];
729 struct kvm_memory_slot *slot;
734 slot = &slots->memslots[index];
736 WARN_ON(slot->id != id);
741 * KVM_SET_USER_MEMORY_REGION ioctl allows the following operations:
742 * - create a new memory slot
743 * - delete an existing memory slot
744 * - modify an existing memory slot
745 * -- move it in the guest physical memory space
746 * -- just change its flags
748 * Since flags can be changed by some of these operations, the following
749 * differentiation is the best we can do for __kvm_set_memory_region():
758 int kvm_set_memory_region(struct kvm *kvm,
759 const struct kvm_userspace_memory_region *mem);
760 int __kvm_set_memory_region(struct kvm *kvm,
761 const struct kvm_userspace_memory_region *mem);
762 void kvm_arch_free_memslot(struct kvm *kvm, struct kvm_memory_slot *slot);
763 void kvm_arch_memslots_updated(struct kvm *kvm, u64 gen);
764 int kvm_arch_prepare_memory_region(struct kvm *kvm,
765 struct kvm_memory_slot *memslot,
766 const struct kvm_userspace_memory_region *mem,
767 enum kvm_mr_change change);
768 void kvm_arch_commit_memory_region(struct kvm *kvm,
769 const struct kvm_userspace_memory_region *mem,
770 struct kvm_memory_slot *old,
771 const struct kvm_memory_slot *new,
772 enum kvm_mr_change change);
773 /* flush all memory translations */
774 void kvm_arch_flush_shadow_all(struct kvm *kvm);
775 /* flush memory translations pointing to 'slot' */
776 void kvm_arch_flush_shadow_memslot(struct kvm *kvm,
777 struct kvm_memory_slot *slot);
779 int gfn_to_page_many_atomic(struct kvm_memory_slot *slot, gfn_t gfn,
780 struct page **pages, int nr_pages);
782 struct page *gfn_to_page(struct kvm *kvm, gfn_t gfn);
783 unsigned long gfn_to_hva(struct kvm *kvm, gfn_t gfn);
784 unsigned long gfn_to_hva_prot(struct kvm *kvm, gfn_t gfn, bool *writable);
785 unsigned long gfn_to_hva_memslot(struct kvm_memory_slot *slot, gfn_t gfn);
786 unsigned long gfn_to_hva_memslot_prot(struct kvm_memory_slot *slot, gfn_t gfn,
788 void kvm_release_page_clean(struct page *page);
789 void kvm_release_page_dirty(struct page *page);
790 void kvm_set_page_accessed(struct page *page);
792 kvm_pfn_t gfn_to_pfn(struct kvm *kvm, gfn_t gfn);
793 kvm_pfn_t gfn_to_pfn_prot(struct kvm *kvm, gfn_t gfn, bool write_fault,
795 kvm_pfn_t gfn_to_pfn_memslot(struct kvm_memory_slot *slot, gfn_t gfn);
796 kvm_pfn_t gfn_to_pfn_memslot_atomic(struct kvm_memory_slot *slot, gfn_t gfn);
797 kvm_pfn_t __gfn_to_pfn_memslot(struct kvm_memory_slot *slot, gfn_t gfn,
798 bool atomic, bool *async, bool write_fault,
799 bool *writable, hva_t *hva);
801 void kvm_release_pfn_clean(kvm_pfn_t pfn);
802 void kvm_release_pfn_dirty(kvm_pfn_t pfn);
803 void kvm_set_pfn_dirty(kvm_pfn_t pfn);
804 void kvm_set_pfn_accessed(kvm_pfn_t pfn);
805 void kvm_get_pfn(kvm_pfn_t pfn);
807 void kvm_release_pfn(kvm_pfn_t pfn, bool dirty, struct gfn_to_pfn_cache *cache);
808 int kvm_read_guest_page(struct kvm *kvm, gfn_t gfn, void *data, int offset,
810 int kvm_read_guest(struct kvm *kvm, gpa_t gpa, void *data, unsigned long len);
811 int kvm_read_guest_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
812 void *data, unsigned long len);
813 int kvm_read_guest_offset_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
814 void *data, unsigned int offset,
816 int kvm_write_guest_page(struct kvm *kvm, gfn_t gfn, const void *data,
817 int offset, int len);
818 int kvm_write_guest(struct kvm *kvm, gpa_t gpa, const void *data,
820 int kvm_write_guest_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
821 void *data, unsigned long len);
822 int kvm_write_guest_offset_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
823 void *data, unsigned int offset,
825 int kvm_gfn_to_hva_cache_init(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
826 gpa_t gpa, unsigned long len);
828 #define __kvm_get_guest(kvm, gfn, offset, v) \
830 unsigned long __addr = gfn_to_hva(kvm, gfn); \
831 typeof(v) __user *__uaddr = (typeof(__uaddr))(__addr + offset); \
832 int __ret = -EFAULT; \
834 if (!kvm_is_error_hva(__addr)) \
835 __ret = get_user(v, __uaddr); \
839 #define kvm_get_guest(kvm, gpa, v) \
842 struct kvm *__kvm = kvm; \
844 __kvm_get_guest(__kvm, __gpa >> PAGE_SHIFT, \
845 offset_in_page(__gpa), v); \
848 #define __kvm_put_guest(kvm, gfn, offset, v) \
850 unsigned long __addr = gfn_to_hva(kvm, gfn); \
851 typeof(v) __user *__uaddr = (typeof(__uaddr))(__addr + offset); \
852 int __ret = -EFAULT; \
854 if (!kvm_is_error_hva(__addr)) \
855 __ret = put_user(v, __uaddr); \
857 mark_page_dirty(kvm, gfn); \
861 #define kvm_put_guest(kvm, gpa, v) \
864 struct kvm *__kvm = kvm; \
866 __kvm_put_guest(__kvm, __gpa >> PAGE_SHIFT, \
867 offset_in_page(__gpa), v); \
870 int kvm_clear_guest(struct kvm *kvm, gpa_t gpa, unsigned long len);
871 struct kvm_memory_slot *gfn_to_memslot(struct kvm *kvm, gfn_t gfn);
872 bool kvm_is_visible_gfn(struct kvm *kvm, gfn_t gfn);
873 bool kvm_vcpu_is_visible_gfn(struct kvm_vcpu *vcpu, gfn_t gfn);
874 unsigned long kvm_host_page_size(struct kvm_vcpu *vcpu, gfn_t gfn);
875 void mark_page_dirty_in_slot(struct kvm *kvm, struct kvm_memory_slot *memslot, gfn_t gfn);
876 void mark_page_dirty(struct kvm *kvm, gfn_t gfn);
878 struct kvm_memslots *kvm_vcpu_memslots(struct kvm_vcpu *vcpu);
879 struct kvm_memory_slot *kvm_vcpu_gfn_to_memslot(struct kvm_vcpu *vcpu, gfn_t gfn);
880 kvm_pfn_t kvm_vcpu_gfn_to_pfn_atomic(struct kvm_vcpu *vcpu, gfn_t gfn);
881 kvm_pfn_t kvm_vcpu_gfn_to_pfn(struct kvm_vcpu *vcpu, gfn_t gfn);
882 int kvm_vcpu_map(struct kvm_vcpu *vcpu, gpa_t gpa, struct kvm_host_map *map);
883 int kvm_map_gfn(struct kvm_vcpu *vcpu, gfn_t gfn, struct kvm_host_map *map,
884 struct gfn_to_pfn_cache *cache, bool atomic);
885 struct page *kvm_vcpu_gfn_to_page(struct kvm_vcpu *vcpu, gfn_t gfn);
886 void kvm_vcpu_unmap(struct kvm_vcpu *vcpu, struct kvm_host_map *map, bool dirty);
887 int kvm_unmap_gfn(struct kvm_vcpu *vcpu, struct kvm_host_map *map,
888 struct gfn_to_pfn_cache *cache, bool dirty, bool atomic);
889 unsigned long kvm_vcpu_gfn_to_hva(struct kvm_vcpu *vcpu, gfn_t gfn);
890 unsigned long kvm_vcpu_gfn_to_hva_prot(struct kvm_vcpu *vcpu, gfn_t gfn, bool *writable);
891 int kvm_vcpu_read_guest_page(struct kvm_vcpu *vcpu, gfn_t gfn, void *data, int offset,
893 int kvm_vcpu_read_guest_atomic(struct kvm_vcpu *vcpu, gpa_t gpa, void *data,
895 int kvm_vcpu_read_guest(struct kvm_vcpu *vcpu, gpa_t gpa, void *data,
897 int kvm_vcpu_write_guest_page(struct kvm_vcpu *vcpu, gfn_t gfn, const void *data,
898 int offset, int len);
899 int kvm_vcpu_write_guest(struct kvm_vcpu *vcpu, gpa_t gpa, const void *data,
901 void kvm_vcpu_mark_page_dirty(struct kvm_vcpu *vcpu, gfn_t gfn);
903 void kvm_sigset_activate(struct kvm_vcpu *vcpu);
904 void kvm_sigset_deactivate(struct kvm_vcpu *vcpu);
906 void kvm_vcpu_block(struct kvm_vcpu *vcpu);
907 void kvm_arch_vcpu_blocking(struct kvm_vcpu *vcpu);
908 void kvm_arch_vcpu_unblocking(struct kvm_vcpu *vcpu);
909 bool kvm_vcpu_wake_up(struct kvm_vcpu *vcpu);
910 void kvm_vcpu_kick(struct kvm_vcpu *vcpu);
911 int kvm_vcpu_yield_to(struct kvm_vcpu *target);
912 void kvm_vcpu_on_spin(struct kvm_vcpu *vcpu, bool usermode_vcpu_not_eligible);
914 void kvm_flush_remote_tlbs(struct kvm *kvm);
915 void kvm_reload_remote_mmus(struct kvm *kvm);
917 #ifdef KVM_ARCH_NR_OBJS_PER_MEMORY_CACHE
918 int kvm_mmu_topup_memory_cache(struct kvm_mmu_memory_cache *mc, int min);
919 int kvm_mmu_memory_cache_nr_free_objects(struct kvm_mmu_memory_cache *mc);
920 void kvm_mmu_free_memory_cache(struct kvm_mmu_memory_cache *mc);
921 void *kvm_mmu_memory_cache_alloc(struct kvm_mmu_memory_cache *mc);
924 bool kvm_make_vcpus_request_mask(struct kvm *kvm, unsigned int req,
925 struct kvm_vcpu *except,
926 unsigned long *vcpu_bitmap, cpumask_var_t tmp);
927 bool kvm_make_all_cpus_request(struct kvm *kvm, unsigned int req);
928 bool kvm_make_all_cpus_request_except(struct kvm *kvm, unsigned int req,
929 struct kvm_vcpu *except);
930 bool kvm_make_cpus_request_mask(struct kvm *kvm, unsigned int req,
931 unsigned long *vcpu_bitmap);
933 long kvm_arch_dev_ioctl(struct file *filp,
934 unsigned int ioctl, unsigned long arg);
935 long kvm_arch_vcpu_ioctl(struct file *filp,
936 unsigned int ioctl, unsigned long arg);
937 vm_fault_t kvm_arch_vcpu_fault(struct kvm_vcpu *vcpu, struct vm_fault *vmf);
939 int kvm_vm_ioctl_check_extension(struct kvm *kvm, long ext);
941 void kvm_arch_mmu_enable_log_dirty_pt_masked(struct kvm *kvm,
942 struct kvm_memory_slot *slot,
945 void kvm_arch_sync_dirty_log(struct kvm *kvm, struct kvm_memory_slot *memslot);
947 #ifdef CONFIG_KVM_GENERIC_DIRTYLOG_READ_PROTECT
948 void kvm_arch_flush_remote_tlbs_memslot(struct kvm *kvm,
949 const struct kvm_memory_slot *memslot);
950 #else /* !CONFIG_KVM_GENERIC_DIRTYLOG_READ_PROTECT */
951 int kvm_vm_ioctl_get_dirty_log(struct kvm *kvm, struct kvm_dirty_log *log);
952 int kvm_get_dirty_log(struct kvm *kvm, struct kvm_dirty_log *log,
953 int *is_dirty, struct kvm_memory_slot **memslot);
956 int kvm_vm_ioctl_irq_line(struct kvm *kvm, struct kvm_irq_level *irq_level,
958 int kvm_vm_ioctl_enable_cap(struct kvm *kvm,
959 struct kvm_enable_cap *cap);
960 long kvm_arch_vm_ioctl(struct file *filp,
961 unsigned int ioctl, unsigned long arg);
963 int kvm_arch_vcpu_ioctl_get_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu);
964 int kvm_arch_vcpu_ioctl_set_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu);
966 int kvm_arch_vcpu_ioctl_translate(struct kvm_vcpu *vcpu,
967 struct kvm_translation *tr);
969 int kvm_arch_vcpu_ioctl_get_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs);
970 int kvm_arch_vcpu_ioctl_set_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs);
971 int kvm_arch_vcpu_ioctl_get_sregs(struct kvm_vcpu *vcpu,
972 struct kvm_sregs *sregs);
973 int kvm_arch_vcpu_ioctl_set_sregs(struct kvm_vcpu *vcpu,
974 struct kvm_sregs *sregs);
975 int kvm_arch_vcpu_ioctl_get_mpstate(struct kvm_vcpu *vcpu,
976 struct kvm_mp_state *mp_state);
977 int kvm_arch_vcpu_ioctl_set_mpstate(struct kvm_vcpu *vcpu,
978 struct kvm_mp_state *mp_state);
979 int kvm_arch_vcpu_ioctl_set_guest_debug(struct kvm_vcpu *vcpu,
980 struct kvm_guest_debug *dbg);
981 int kvm_arch_vcpu_ioctl_run(struct kvm_vcpu *vcpu);
983 int kvm_arch_init(void *opaque);
984 void kvm_arch_exit(void);
986 void kvm_arch_sched_in(struct kvm_vcpu *vcpu, int cpu);
988 void kvm_arch_vcpu_load(struct kvm_vcpu *vcpu, int cpu);
989 void kvm_arch_vcpu_put(struct kvm_vcpu *vcpu);
990 int kvm_arch_vcpu_precreate(struct kvm *kvm, unsigned int id);
991 int kvm_arch_vcpu_create(struct kvm_vcpu *vcpu);
992 void kvm_arch_vcpu_postcreate(struct kvm_vcpu *vcpu);
993 void kvm_arch_vcpu_destroy(struct kvm_vcpu *vcpu);
995 #ifdef __KVM_HAVE_ARCH_VCPU_DEBUGFS
996 void kvm_arch_create_vcpu_debugfs(struct kvm_vcpu *vcpu, struct dentry *debugfs_dentry);
999 int kvm_arch_hardware_enable(void);
1000 void kvm_arch_hardware_disable(void);
1001 int kvm_arch_hardware_setup(void *opaque);
1002 void kvm_arch_hardware_unsetup(void);
1003 int kvm_arch_check_processor_compat(void *opaque);
1004 int kvm_arch_vcpu_runnable(struct kvm_vcpu *vcpu);
1005 bool kvm_arch_vcpu_in_kernel(struct kvm_vcpu *vcpu);
1006 int kvm_arch_vcpu_should_kick(struct kvm_vcpu *vcpu);
1007 bool kvm_arch_dy_runnable(struct kvm_vcpu *vcpu);
1008 bool kvm_arch_dy_has_pending_interrupt(struct kvm_vcpu *vcpu);
1009 int kvm_arch_post_init_vm(struct kvm *kvm);
1010 void kvm_arch_pre_destroy_vm(struct kvm *kvm);
1012 #ifndef __KVM_HAVE_ARCH_VM_ALLOC
1014 * All architectures that want to use vzalloc currently also
1015 * need their own kvm_arch_alloc_vm implementation.
1017 static inline struct kvm *kvm_arch_alloc_vm(void)
1019 return kzalloc(sizeof(struct kvm), GFP_KERNEL);
1022 static inline void kvm_arch_free_vm(struct kvm *kvm)
1028 #ifndef __KVM_HAVE_ARCH_FLUSH_REMOTE_TLB
1029 static inline int kvm_arch_flush_remote_tlb(struct kvm *kvm)
1035 #ifdef __KVM_HAVE_ARCH_NONCOHERENT_DMA
1036 void kvm_arch_register_noncoherent_dma(struct kvm *kvm);
1037 void kvm_arch_unregister_noncoherent_dma(struct kvm *kvm);
1038 bool kvm_arch_has_noncoherent_dma(struct kvm *kvm);
1040 static inline void kvm_arch_register_noncoherent_dma(struct kvm *kvm)
1044 static inline void kvm_arch_unregister_noncoherent_dma(struct kvm *kvm)
1048 static inline bool kvm_arch_has_noncoherent_dma(struct kvm *kvm)
1053 #ifdef __KVM_HAVE_ARCH_ASSIGNED_DEVICE
1054 void kvm_arch_start_assignment(struct kvm *kvm);
1055 void kvm_arch_end_assignment(struct kvm *kvm);
1056 bool kvm_arch_has_assigned_device(struct kvm *kvm);
1058 static inline void kvm_arch_start_assignment(struct kvm *kvm)
1062 static inline void kvm_arch_end_assignment(struct kvm *kvm)
1066 static inline bool kvm_arch_has_assigned_device(struct kvm *kvm)
1072 static inline struct rcuwait *kvm_arch_vcpu_get_wait(struct kvm_vcpu *vcpu)
1074 #ifdef __KVM_HAVE_ARCH_WQP
1075 return vcpu->arch.waitp;
1081 #ifdef __KVM_HAVE_ARCH_INTC_INITIALIZED
1083 * returns true if the virtual interrupt controller is initialized and
1084 * ready to accept virtual IRQ. On some architectures the virtual interrupt
1085 * controller is dynamically instantiated and this is not always true.
1087 bool kvm_arch_intc_initialized(struct kvm *kvm);
1089 static inline bool kvm_arch_intc_initialized(struct kvm *kvm)
1095 int kvm_arch_init_vm(struct kvm *kvm, unsigned long type);
1096 void kvm_arch_destroy_vm(struct kvm *kvm);
1097 void kvm_arch_sync_events(struct kvm *kvm);
1099 int kvm_cpu_has_pending_timer(struct kvm_vcpu *vcpu);
1101 bool kvm_is_reserved_pfn(kvm_pfn_t pfn);
1102 bool kvm_is_zone_device_pfn(kvm_pfn_t pfn);
1103 bool kvm_is_transparent_hugepage(kvm_pfn_t pfn);
1105 struct kvm_irq_ack_notifier {
1106 struct hlist_node link;
1108 void (*irq_acked)(struct kvm_irq_ack_notifier *kian);
1111 int kvm_irq_map_gsi(struct kvm *kvm,
1112 struct kvm_kernel_irq_routing_entry *entries, int gsi);
1113 int kvm_irq_map_chip_pin(struct kvm *kvm, unsigned irqchip, unsigned pin);
1115 int kvm_set_irq(struct kvm *kvm, int irq_source_id, u32 irq, int level,
1117 int kvm_set_msi(struct kvm_kernel_irq_routing_entry *irq_entry, struct kvm *kvm,
1118 int irq_source_id, int level, bool line_status);
1119 int kvm_arch_set_irq_inatomic(struct kvm_kernel_irq_routing_entry *e,
1120 struct kvm *kvm, int irq_source_id,
1121 int level, bool line_status);
1122 bool kvm_irq_has_notifier(struct kvm *kvm, unsigned irqchip, unsigned pin);
1123 void kvm_notify_acked_gsi(struct kvm *kvm, int gsi);
1124 void kvm_notify_acked_irq(struct kvm *kvm, unsigned irqchip, unsigned pin);
1125 void kvm_register_irq_ack_notifier(struct kvm *kvm,
1126 struct kvm_irq_ack_notifier *kian);
1127 void kvm_unregister_irq_ack_notifier(struct kvm *kvm,
1128 struct kvm_irq_ack_notifier *kian);
1129 int kvm_request_irq_source_id(struct kvm *kvm);
1130 void kvm_free_irq_source_id(struct kvm *kvm, int irq_source_id);
1131 bool kvm_arch_irqfd_allowed(struct kvm *kvm, struct kvm_irqfd *args);
1134 * search_memslots() and __gfn_to_memslot() are here because they are
1135 * used in non-modular code in arch/powerpc/kvm/book3s_hv_rm_mmu.c.
1136 * gfn_to_memslot() itself isn't here as an inline because that would
1137 * bloat other code too much.
1139 * IMPORTANT: Slots are sorted from highest GFN to lowest GFN!
1141 static inline struct kvm_memory_slot *
1142 search_memslots(struct kvm_memslots *slots, gfn_t gfn)
1144 int start = 0, end = slots->used_slots;
1145 int slot = atomic_read(&slots->lru_slot);
1146 struct kvm_memory_slot *memslots = slots->memslots;
1148 if (unlikely(!slots->used_slots))
1151 if (gfn >= memslots[slot].base_gfn &&
1152 gfn < memslots[slot].base_gfn + memslots[slot].npages)
1153 return &memslots[slot];
1155 while (start < end) {
1156 slot = start + (end - start) / 2;
1158 if (gfn >= memslots[slot].base_gfn)
1164 if (start < slots->used_slots && gfn >= memslots[start].base_gfn &&
1165 gfn < memslots[start].base_gfn + memslots[start].npages) {
1166 atomic_set(&slots->lru_slot, start);
1167 return &memslots[start];
1173 static inline struct kvm_memory_slot *
1174 __gfn_to_memslot(struct kvm_memslots *slots, gfn_t gfn)
1176 return search_memslots(slots, gfn);
1179 static inline unsigned long
1180 __gfn_to_hva_memslot(const struct kvm_memory_slot *slot, gfn_t gfn)
1182 return slot->userspace_addr + (gfn - slot->base_gfn) * PAGE_SIZE;
1185 static inline int memslot_id(struct kvm *kvm, gfn_t gfn)
1187 return gfn_to_memslot(kvm, gfn)->id;
1191 hva_to_gfn_memslot(unsigned long hva, struct kvm_memory_slot *slot)
1193 gfn_t gfn_offset = (hva - slot->userspace_addr) >> PAGE_SHIFT;
1195 return slot->base_gfn + gfn_offset;
1198 static inline gpa_t gfn_to_gpa(gfn_t gfn)
1200 return (gpa_t)gfn << PAGE_SHIFT;
1203 static inline gfn_t gpa_to_gfn(gpa_t gpa)
1205 return (gfn_t)(gpa >> PAGE_SHIFT);
1208 static inline hpa_t pfn_to_hpa(kvm_pfn_t pfn)
1210 return (hpa_t)pfn << PAGE_SHIFT;
1213 static inline struct page *kvm_vcpu_gpa_to_page(struct kvm_vcpu *vcpu,
1216 return kvm_vcpu_gfn_to_page(vcpu, gpa_to_gfn(gpa));
1219 static inline bool kvm_is_error_gpa(struct kvm *kvm, gpa_t gpa)
1221 unsigned long hva = gfn_to_hva(kvm, gpa_to_gfn(gpa));
1223 return kvm_is_error_hva(hva);
1226 enum kvm_stat_kind {
1231 struct kvm_stat_data {
1233 struct kvm_stats_debugfs_item *dbgfs_item;
1236 struct kvm_stats_debugfs_item {
1239 enum kvm_stat_kind kind;
1243 #define KVM_DBGFS_GET_MODE(dbgfs_item) \
1244 ((dbgfs_item)->mode ? (dbgfs_item)->mode : 0644)
1246 #define VM_STAT(n, x, ...) \
1247 { n, offsetof(struct kvm, stat.x), KVM_STAT_VM, ## __VA_ARGS__ }
1248 #define VCPU_STAT(n, x, ...) \
1249 { n, offsetof(struct kvm_vcpu, stat.x), KVM_STAT_VCPU, ## __VA_ARGS__ }
1251 extern struct kvm_stats_debugfs_item debugfs_entries[];
1252 extern struct dentry *kvm_debugfs_dir;
1254 #if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
1255 static inline int mmu_notifier_retry(struct kvm *kvm, unsigned long mmu_seq)
1257 if (unlikely(kvm->mmu_notifier_count))
1260 * Ensure the read of mmu_notifier_count happens before the read
1261 * of mmu_notifier_seq. This interacts with the smp_wmb() in
1262 * mmu_notifier_invalidate_range_end to make sure that the caller
1263 * either sees the old (non-zero) value of mmu_notifier_count or
1264 * the new (incremented) value of mmu_notifier_seq.
1265 * PowerPC Book3s HV KVM calls this under a per-page lock
1266 * rather than under kvm->mmu_lock, for scalability, so
1267 * can't rely on kvm->mmu_lock to keep things ordered.
1270 if (kvm->mmu_notifier_seq != mmu_seq)
1275 static inline int mmu_notifier_retry_hva(struct kvm *kvm,
1276 unsigned long mmu_seq,
1279 lockdep_assert_held(&kvm->mmu_lock);
1281 * If mmu_notifier_count is non-zero, then the range maintained by
1282 * kvm_mmu_notifier_invalidate_range_start contains all addresses that
1283 * might be being invalidated. Note that it may include some false
1284 * positives, due to shortcuts when handing concurrent invalidations.
1286 if (unlikely(kvm->mmu_notifier_count) &&
1287 hva >= kvm->mmu_notifier_range_start &&
1288 hva < kvm->mmu_notifier_range_end)
1290 if (kvm->mmu_notifier_seq != mmu_seq)
1296 #ifdef CONFIG_HAVE_KVM_IRQ_ROUTING
1298 #define KVM_MAX_IRQ_ROUTES 4096 /* might need extension/rework in the future */
1300 bool kvm_arch_can_set_irq_routing(struct kvm *kvm);
1301 int kvm_set_irq_routing(struct kvm *kvm,
1302 const struct kvm_irq_routing_entry *entries,
1305 int kvm_set_routing_entry(struct kvm *kvm,
1306 struct kvm_kernel_irq_routing_entry *e,
1307 const struct kvm_irq_routing_entry *ue);
1308 void kvm_free_irq_routing(struct kvm *kvm);
1312 static inline void kvm_free_irq_routing(struct kvm *kvm) {}
1316 int kvm_send_userspace_msi(struct kvm *kvm, struct kvm_msi *msi);
1318 #ifdef CONFIG_HAVE_KVM_EVENTFD
1320 void kvm_eventfd_init(struct kvm *kvm);
1321 int kvm_ioeventfd(struct kvm *kvm, struct kvm_ioeventfd *args);
1323 #ifdef CONFIG_HAVE_KVM_IRQFD
1324 int kvm_irqfd(struct kvm *kvm, struct kvm_irqfd *args);
1325 void kvm_irqfd_release(struct kvm *kvm);
1326 void kvm_irq_routing_update(struct kvm *);
1328 static inline int kvm_irqfd(struct kvm *kvm, struct kvm_irqfd *args)
1333 static inline void kvm_irqfd_release(struct kvm *kvm) {}
1338 static inline void kvm_eventfd_init(struct kvm *kvm) {}
1340 static inline int kvm_irqfd(struct kvm *kvm, struct kvm_irqfd *args)
1345 static inline void kvm_irqfd_release(struct kvm *kvm) {}
1347 #ifdef CONFIG_HAVE_KVM_IRQCHIP
1348 static inline void kvm_irq_routing_update(struct kvm *kvm)
1353 static inline int kvm_ioeventfd(struct kvm *kvm, struct kvm_ioeventfd *args)
1358 #endif /* CONFIG_HAVE_KVM_EVENTFD */
1360 void kvm_arch_irq_routing_update(struct kvm *kvm);
1362 static inline void kvm_make_request(int req, struct kvm_vcpu *vcpu)
1365 * Ensure the rest of the request is published to kvm_check_request's
1366 * caller. Paired with the smp_mb__after_atomic in kvm_check_request.
1369 set_bit(req & KVM_REQUEST_MASK, (void *)&vcpu->requests);
1372 static inline bool kvm_request_pending(struct kvm_vcpu *vcpu)
1374 return READ_ONCE(vcpu->requests);
1377 static inline bool kvm_test_request(int req, struct kvm_vcpu *vcpu)
1379 return test_bit(req & KVM_REQUEST_MASK, (void *)&vcpu->requests);
1382 static inline void kvm_clear_request(int req, struct kvm_vcpu *vcpu)
1384 clear_bit(req & KVM_REQUEST_MASK, (void *)&vcpu->requests);
1387 static inline bool kvm_check_request(int req, struct kvm_vcpu *vcpu)
1389 if (kvm_test_request(req, vcpu)) {
1390 kvm_clear_request(req, vcpu);
1393 * Ensure the rest of the request is visible to kvm_check_request's
1394 * caller. Paired with the smp_wmb in kvm_make_request.
1396 smp_mb__after_atomic();
1403 extern bool kvm_rebooting;
1405 extern unsigned int halt_poll_ns;
1406 extern unsigned int halt_poll_ns_grow;
1407 extern unsigned int halt_poll_ns_grow_start;
1408 extern unsigned int halt_poll_ns_shrink;
1411 const struct kvm_device_ops *ops;
1414 struct list_head vm_node;
1417 /* create, destroy, and name are mandatory */
1418 struct kvm_device_ops {
1422 * create is called holding kvm->lock and any operations not suitable
1423 * to do while holding the lock should be deferred to init (see
1426 int (*create)(struct kvm_device *dev, u32 type);
1429 * init is called after create if create is successful and is called
1430 * outside of holding kvm->lock.
1432 void (*init)(struct kvm_device *dev);
1435 * Destroy is responsible for freeing dev.
1437 * Destroy may be called before or after destructors are called
1438 * on emulated I/O regions, depending on whether a reference is
1439 * held by a vcpu or other kvm component that gets destroyed
1440 * after the emulated I/O.
1442 void (*destroy)(struct kvm_device *dev);
1445 * Release is an alternative method to free the device. It is
1446 * called when the device file descriptor is closed. Once
1447 * release is called, the destroy method will not be called
1448 * anymore as the device is removed from the device list of
1449 * the VM. kvm->lock is held.
1451 void (*release)(struct kvm_device *dev);
1453 int (*set_attr)(struct kvm_device *dev, struct kvm_device_attr *attr);
1454 int (*get_attr)(struct kvm_device *dev, struct kvm_device_attr *attr);
1455 int (*has_attr)(struct kvm_device *dev, struct kvm_device_attr *attr);
1456 long (*ioctl)(struct kvm_device *dev, unsigned int ioctl,
1458 int (*mmap)(struct kvm_device *dev, struct vm_area_struct *vma);
1461 void kvm_device_get(struct kvm_device *dev);
1462 void kvm_device_put(struct kvm_device *dev);
1463 struct kvm_device *kvm_device_from_filp(struct file *filp);
1464 int kvm_register_device_ops(const struct kvm_device_ops *ops, u32 type);
1465 void kvm_unregister_device_ops(u32 type);
1467 extern struct kvm_device_ops kvm_mpic_ops;
1468 extern struct kvm_device_ops kvm_arm_vgic_v2_ops;
1469 extern struct kvm_device_ops kvm_arm_vgic_v3_ops;
1471 #ifdef CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT
1473 static inline void kvm_vcpu_set_in_spin_loop(struct kvm_vcpu *vcpu, bool val)
1475 vcpu->spin_loop.in_spin_loop = val;
1477 static inline void kvm_vcpu_set_dy_eligible(struct kvm_vcpu *vcpu, bool val)
1479 vcpu->spin_loop.dy_eligible = val;
1482 #else /* !CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT */
1484 static inline void kvm_vcpu_set_in_spin_loop(struct kvm_vcpu *vcpu, bool val)
1488 static inline void kvm_vcpu_set_dy_eligible(struct kvm_vcpu *vcpu, bool val)
1491 #endif /* CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT */
1493 static inline bool kvm_is_visible_memslot(struct kvm_memory_slot *memslot)
1495 return (memslot && memslot->id < KVM_USER_MEM_SLOTS &&
1496 !(memslot->flags & KVM_MEMSLOT_INVALID));
1499 struct kvm_vcpu *kvm_get_running_vcpu(void);
1500 struct kvm_vcpu * __percpu *kvm_get_running_vcpus(void);
1502 #ifdef CONFIG_HAVE_KVM_IRQ_BYPASS
1503 bool kvm_arch_has_irq_bypass(void);
1504 int kvm_arch_irq_bypass_add_producer(struct irq_bypass_consumer *,
1505 struct irq_bypass_producer *);
1506 void kvm_arch_irq_bypass_del_producer(struct irq_bypass_consumer *,
1507 struct irq_bypass_producer *);
1508 void kvm_arch_irq_bypass_stop(struct irq_bypass_consumer *);
1509 void kvm_arch_irq_bypass_start(struct irq_bypass_consumer *);
1510 int kvm_arch_update_irqfd_routing(struct kvm *kvm, unsigned int host_irq,
1511 uint32_t guest_irq, bool set);
1512 #endif /* CONFIG_HAVE_KVM_IRQ_BYPASS */
1514 #ifdef CONFIG_HAVE_KVM_INVALID_WAKEUPS
1515 /* If we wakeup during the poll time, was it a sucessful poll? */
1516 static inline bool vcpu_valid_wakeup(struct kvm_vcpu *vcpu)
1518 return vcpu->valid_wakeup;
1522 static inline bool vcpu_valid_wakeup(struct kvm_vcpu *vcpu)
1526 #endif /* CONFIG_HAVE_KVM_INVALID_WAKEUPS */
1528 #ifdef CONFIG_HAVE_KVM_NO_POLL
1529 /* Callback that tells if we must not poll */
1530 bool kvm_arch_no_poll(struct kvm_vcpu *vcpu);
1532 static inline bool kvm_arch_no_poll(struct kvm_vcpu *vcpu)
1536 #endif /* CONFIG_HAVE_KVM_NO_POLL */
1538 #ifdef CONFIG_HAVE_KVM_VCPU_ASYNC_IOCTL
1539 long kvm_arch_vcpu_async_ioctl(struct file *filp,
1540 unsigned int ioctl, unsigned long arg);
1542 static inline long kvm_arch_vcpu_async_ioctl(struct file *filp,
1546 return -ENOIOCTLCMD;
1548 #endif /* CONFIG_HAVE_KVM_VCPU_ASYNC_IOCTL */
1550 void kvm_arch_mmu_notifier_invalidate_range(struct kvm *kvm,
1551 unsigned long start, unsigned long end);
1553 #ifdef CONFIG_HAVE_KVM_VCPU_RUN_PID_CHANGE
1554 int kvm_arch_vcpu_run_pid_change(struct kvm_vcpu *vcpu);
1556 static inline int kvm_arch_vcpu_run_pid_change(struct kvm_vcpu *vcpu)
1560 #endif /* CONFIG_HAVE_KVM_VCPU_RUN_PID_CHANGE */
1562 typedef int (*kvm_vm_thread_fn_t)(struct kvm *kvm, uintptr_t data);
1564 int kvm_vm_create_worker_thread(struct kvm *kvm, kvm_vm_thread_fn_t thread_fn,
1565 uintptr_t data, const char *name,
1566 struct task_struct **thread_ptr);
1568 #ifdef CONFIG_KVM_XFER_TO_GUEST_WORK
1569 static inline void kvm_handle_signal_exit(struct kvm_vcpu *vcpu)
1571 vcpu->run->exit_reason = KVM_EXIT_INTR;
1572 vcpu->stat.signal_exits++;
1574 #endif /* CONFIG_KVM_XFER_TO_GUEST_WORK */
1577 * This defines how many reserved entries we want to keep before we
1578 * kick the vcpu to the userspace to avoid dirty ring full. This
1579 * value can be tuned to higher if e.g. PML is enabled on the host.
1581 #define KVM_DIRTY_RING_RSVD_ENTRIES 64
1583 /* Max number of entries allowed for each kvm dirty ring */
1584 #define KVM_DIRTY_RING_MAX_ENTRIES 65536