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/sched/stat.h>
14 #include <linux/bug.h>
15 #include <linux/minmax.h>
17 #include <linux/mmu_notifier.h>
18 #include <linux/preempt.h>
19 #include <linux/msi.h>
20 #include <linux/slab.h>
21 #include <linux/vmalloc.h>
22 #include <linux/rcupdate.h>
23 #include <linux/ratelimit.h>
24 #include <linux/err.h>
25 #include <linux/irqflags.h>
26 #include <linux/context_tracking.h>
27 #include <linux/irqbypass.h>
28 #include <linux/rcuwait.h>
29 #include <linux/refcount.h>
30 #include <linux/nospec.h>
31 #include <linux/notifier.h>
32 #include <linux/hashtable.h>
33 #include <linux/interval_tree.h>
34 #include <linux/rbtree.h>
35 #include <linux/xarray.h>
36 #include <asm/signal.h>
38 #include <linux/kvm.h>
39 #include <linux/kvm_para.h>
41 #include <linux/kvm_types.h>
43 #include <asm/kvm_host.h>
44 #include <linux/kvm_dirty_ring.h>
46 #ifndef KVM_MAX_VCPU_IDS
47 #define KVM_MAX_VCPU_IDS KVM_MAX_VCPUS
51 * The bit 16 ~ bit 31 of kvm_memory_region::flags are internally used
52 * in kvm, other bits are visible for userspace which are defined in
53 * include/linux/kvm_h.
55 #define KVM_MEMSLOT_INVALID (1UL << 16)
58 * Bit 63 of the memslot generation number is an "update in-progress flag",
59 * e.g. is temporarily set for the duration of install_new_memslots().
60 * This flag effectively creates a unique generation number that is used to
61 * mark cached memslot data, e.g. MMIO accesses, as potentially being stale,
62 * i.e. may (or may not) have come from the previous memslots generation.
64 * This is necessary because the actual memslots update is not atomic with
65 * respect to the generation number update. Updating the generation number
66 * first would allow a vCPU to cache a spte from the old memslots using the
67 * new generation number, and updating the generation number after switching
68 * to the new memslots would allow cache hits using the old generation number
69 * to reference the defunct memslots.
71 * This mechanism is used to prevent getting hits in KVM's caches while a
72 * memslot update is in-progress, and to prevent cache hits *after* updating
73 * the actual generation number against accesses that were inserted into the
74 * cache *before* the memslots were updated.
76 #define KVM_MEMSLOT_GEN_UPDATE_IN_PROGRESS BIT_ULL(63)
78 /* Two fragments for cross MMIO pages. */
79 #define KVM_MAX_MMIO_FRAGMENTS 2
81 #ifndef KVM_ADDRESS_SPACE_NUM
82 #define KVM_ADDRESS_SPACE_NUM 1
86 * For the normal pfn, the highest 12 bits should be zero,
87 * so we can mask bit 62 ~ bit 52 to indicate the error pfn,
88 * mask bit 63 to indicate the noslot pfn.
90 #define KVM_PFN_ERR_MASK (0x7ffULL << 52)
91 #define KVM_PFN_ERR_NOSLOT_MASK (0xfffULL << 52)
92 #define KVM_PFN_NOSLOT (0x1ULL << 63)
94 #define KVM_PFN_ERR_FAULT (KVM_PFN_ERR_MASK)
95 #define KVM_PFN_ERR_HWPOISON (KVM_PFN_ERR_MASK + 1)
96 #define KVM_PFN_ERR_RO_FAULT (KVM_PFN_ERR_MASK + 2)
99 * error pfns indicate that the gfn is in slot but faild to
100 * translate it to pfn on host.
102 static inline bool is_error_pfn(kvm_pfn_t pfn)
104 return !!(pfn & KVM_PFN_ERR_MASK);
108 * error_noslot pfns indicate that the gfn can not be
109 * translated to pfn - it is not in slot or failed to
110 * translate it to pfn.
112 static inline bool is_error_noslot_pfn(kvm_pfn_t pfn)
114 return !!(pfn & KVM_PFN_ERR_NOSLOT_MASK);
117 /* noslot pfn indicates that the gfn is not in slot. */
118 static inline bool is_noslot_pfn(kvm_pfn_t pfn)
120 return pfn == KVM_PFN_NOSLOT;
124 * architectures with KVM_HVA_ERR_BAD other than PAGE_OFFSET (e.g. s390)
125 * provide own defines and kvm_is_error_hva
127 #ifndef KVM_HVA_ERR_BAD
129 #define KVM_HVA_ERR_BAD (PAGE_OFFSET)
130 #define KVM_HVA_ERR_RO_BAD (PAGE_OFFSET + PAGE_SIZE)
132 static inline bool kvm_is_error_hva(unsigned long addr)
134 return addr >= PAGE_OFFSET;
139 #define KVM_ERR_PTR_BAD_PAGE (ERR_PTR(-ENOENT))
141 static inline bool is_error_page(struct page *page)
146 #define KVM_REQUEST_MASK GENMASK(7,0)
147 #define KVM_REQUEST_NO_WAKEUP BIT(8)
148 #define KVM_REQUEST_WAIT BIT(9)
150 * Architecture-independent vcpu->requests bit members
151 * Bits 4-7 are reserved for more arch-independent bits.
153 #define KVM_REQ_TLB_FLUSH (0 | KVM_REQUEST_WAIT | KVM_REQUEST_NO_WAKEUP)
154 #define KVM_REQ_MMU_RELOAD (1 | KVM_REQUEST_WAIT | KVM_REQUEST_NO_WAKEUP)
155 #define KVM_REQ_UNBLOCK 2
156 #define KVM_REQ_UNHALT 3
157 #define KVM_REQ_VM_DEAD (4 | KVM_REQUEST_WAIT | KVM_REQUEST_NO_WAKEUP)
158 #define KVM_REQ_GPC_INVALIDATE (5 | KVM_REQUEST_WAIT | KVM_REQUEST_NO_WAKEUP)
159 #define KVM_REQUEST_ARCH_BASE 8
161 #define KVM_ARCH_REQ_FLAGS(nr, flags) ({ \
162 BUILD_BUG_ON((unsigned)(nr) >= (sizeof_field(struct kvm_vcpu, requests) * 8) - KVM_REQUEST_ARCH_BASE); \
163 (unsigned)(((nr) + KVM_REQUEST_ARCH_BASE) | (flags)); \
165 #define KVM_ARCH_REQ(nr) KVM_ARCH_REQ_FLAGS(nr, 0)
167 bool kvm_make_vcpus_request_mask(struct kvm *kvm, unsigned int req,
168 unsigned long *vcpu_bitmap);
169 bool kvm_make_all_cpus_request(struct kvm *kvm, unsigned int req);
170 bool kvm_make_all_cpus_request_except(struct kvm *kvm, unsigned int req,
171 struct kvm_vcpu *except);
172 bool kvm_make_cpus_request_mask(struct kvm *kvm, unsigned int req,
173 unsigned long *vcpu_bitmap);
175 #define KVM_USERSPACE_IRQ_SOURCE_ID 0
176 #define KVM_IRQFD_RESAMPLE_IRQ_SOURCE_ID 1
178 extern struct mutex kvm_lock;
179 extern struct list_head vm_list;
181 struct kvm_io_range {
184 struct kvm_io_device *dev;
187 #define NR_IOBUS_DEVS 1000
192 struct kvm_io_range range[];
198 KVM_VIRTIO_CCW_NOTIFY_BUS,
203 int kvm_io_bus_write(struct kvm_vcpu *vcpu, enum kvm_bus bus_idx, gpa_t addr,
204 int len, const void *val);
205 int kvm_io_bus_write_cookie(struct kvm_vcpu *vcpu, enum kvm_bus bus_idx,
206 gpa_t addr, int len, const void *val, long cookie);
207 int kvm_io_bus_read(struct kvm_vcpu *vcpu, enum kvm_bus bus_idx, gpa_t addr,
209 int kvm_io_bus_register_dev(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr,
210 int len, struct kvm_io_device *dev);
211 int kvm_io_bus_unregister_dev(struct kvm *kvm, enum kvm_bus bus_idx,
212 struct kvm_io_device *dev);
213 struct kvm_io_device *kvm_io_bus_get_dev(struct kvm *kvm, enum kvm_bus bus_idx,
216 #ifdef CONFIG_KVM_ASYNC_PF
217 struct kvm_async_pf {
218 struct work_struct work;
219 struct list_head link;
220 struct list_head queue;
221 struct kvm_vcpu *vcpu;
222 struct mm_struct *mm;
225 struct kvm_arch_async_pf arch;
227 bool notpresent_injected;
230 void kvm_clear_async_pf_completion_queue(struct kvm_vcpu *vcpu);
231 void kvm_check_async_pf_completion(struct kvm_vcpu *vcpu);
232 bool kvm_setup_async_pf(struct kvm_vcpu *vcpu, gpa_t cr2_or_gpa,
233 unsigned long hva, struct kvm_arch_async_pf *arch);
234 int kvm_async_pf_wakeup_all(struct kvm_vcpu *vcpu);
237 #ifdef KVM_ARCH_WANT_MMU_NOTIFIER
238 struct kvm_gfn_range {
239 struct kvm_memory_slot *slot;
245 bool kvm_unmap_gfn_range(struct kvm *kvm, struct kvm_gfn_range *range);
246 bool kvm_age_gfn(struct kvm *kvm, struct kvm_gfn_range *range);
247 bool kvm_test_age_gfn(struct kvm *kvm, struct kvm_gfn_range *range);
248 bool kvm_set_spte_gfn(struct kvm *kvm, struct kvm_gfn_range *range);
255 READING_SHADOW_PAGE_TABLES,
258 #define KVM_UNMAPPED_PAGE ((void *) 0x500 + POISON_POINTER_DELTA)
260 struct kvm_host_map {
262 * Only valid if the 'pfn' is managed by the host kernel (i.e. There is
263 * a 'struct page' for it. When using mem= kernel parameter some memory
264 * can be used as guest memory but they are not managed by host
266 * If 'pfn' is not managed by the host kernel, this field is
267 * initialized to KVM_UNMAPPED_PAGE.
276 * Used to check if the mapping is valid or not. Never use 'kvm_host_map'
277 * directly to check for that.
279 static inline bool kvm_vcpu_mapped(struct kvm_host_map *map)
284 static inline bool kvm_vcpu_can_poll(ktime_t cur, ktime_t stop)
286 return single_task_running() && !need_resched() && ktime_before(cur, stop);
290 * Sometimes a large or cross-page mmio needs to be broken up into separate
291 * exits for userspace servicing.
293 struct kvm_mmio_fragment {
301 #ifdef CONFIG_PREEMPT_NOTIFIERS
302 struct preempt_notifier preempt_notifier;
305 int vcpu_id; /* id given by userspace at creation */
306 int vcpu_idx; /* index in kvm->vcpus array */
310 unsigned long guest_debug;
313 struct list_head blocked_vcpu_list;
318 #ifndef __KVM_HAVE_ARCH_WQP
321 struct pid __rcu *pid;
324 unsigned int halt_poll_ns;
327 #ifdef CONFIG_HAS_IOMEM
329 int mmio_read_completed;
331 int mmio_cur_fragment;
332 int mmio_nr_fragments;
333 struct kvm_mmio_fragment mmio_fragments[KVM_MAX_MMIO_FRAGMENTS];
336 #ifdef CONFIG_KVM_ASYNC_PF
339 struct list_head queue;
340 struct list_head done;
345 #ifdef CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT
347 * Cpu relax intercept or pause loop exit optimization
348 * in_spin_loop: set when a vcpu does a pause loop exit
349 * or cpu relax intercepted.
350 * dy_eligible: indicates whether vcpu is eligible for directed yield.
359 struct kvm_vcpu_arch arch;
360 struct kvm_vcpu_stat stat;
361 char stats_id[KVM_STATS_NAME_SIZE];
362 struct kvm_dirty_ring dirty_ring;
365 * The most recently used memslot by this vCPU and the slots generation
366 * for which it is valid.
367 * No wraparound protection is needed since generations won't overflow in
368 * thousands of years, even assuming 1M memslot operations per second.
370 struct kvm_memory_slot *last_used_slot;
371 u64 last_used_slot_gen;
374 /* must be called with irqs disabled */
375 static __always_inline void guest_enter_irqoff(void)
378 * This is running in ioctl context so its safe to assume that it's the
379 * stime pending cputime to flush.
381 instrumentation_begin();
382 vtime_account_guest_enter();
383 instrumentation_end();
386 * KVM does not hold any references to rcu protected data when it
387 * switches CPU into a guest mode. In fact switching to a guest mode
388 * is very similar to exiting to userspace from rcu point of view. In
389 * addition CPU may stay in a guest mode for quite a long time (up to
390 * one time slice). Lets treat guest mode as quiescent state, just like
391 * we do with user-mode execution.
393 if (!context_tracking_guest_enter()) {
394 instrumentation_begin();
395 rcu_virt_note_context_switch(smp_processor_id());
396 instrumentation_end();
400 static __always_inline void guest_exit_irqoff(void)
402 context_tracking_guest_exit();
404 instrumentation_begin();
405 /* Flush the guest cputime we spent on the guest */
406 vtime_account_guest_exit();
407 instrumentation_end();
410 static inline void guest_exit(void)
414 local_irq_save(flags);
416 local_irq_restore(flags);
419 static inline int kvm_vcpu_exiting_guest_mode(struct kvm_vcpu *vcpu)
422 * The memory barrier ensures a previous write to vcpu->requests cannot
423 * be reordered with the read of vcpu->mode. It pairs with the general
424 * memory barrier following the write of vcpu->mode in VCPU RUN.
426 smp_mb__before_atomic();
427 return cmpxchg(&vcpu->mode, IN_GUEST_MODE, EXITING_GUEST_MODE);
431 * Some of the bitops functions do not support too long bitmaps.
432 * This number must be determined not to exceed such limits.
434 #define KVM_MEM_MAX_NR_PAGES ((1UL << 31) - 1)
437 * Since at idle each memslot belongs to two memslot sets it has to contain
438 * two embedded nodes for each data structure that it forms a part of.
440 * Two memslot sets (one active and one inactive) are necessary so the VM
441 * continues to run on one memslot set while the other is being modified.
443 * These two memslot sets normally point to the same set of memslots.
444 * They can, however, be desynchronized when performing a memslot management
445 * operation by replacing the memslot to be modified by its copy.
446 * After the operation is complete, both memslot sets once again point to
447 * the same, common set of memslot data.
449 * The memslots themselves are independent of each other so they can be
450 * individually added or deleted.
452 struct kvm_memory_slot {
453 struct hlist_node id_node[2];
454 struct interval_tree_node hva_node[2];
455 struct rb_node gfn_node[2];
457 unsigned long npages;
458 unsigned long *dirty_bitmap;
459 struct kvm_arch_memory_slot arch;
460 unsigned long userspace_addr;
466 static inline bool kvm_slot_dirty_track_enabled(const struct kvm_memory_slot *slot)
468 return slot->flags & KVM_MEM_LOG_DIRTY_PAGES;
471 static inline unsigned long kvm_dirty_bitmap_bytes(struct kvm_memory_slot *memslot)
473 return ALIGN(memslot->npages, BITS_PER_LONG) / 8;
476 static inline unsigned long *kvm_second_dirty_bitmap(struct kvm_memory_slot *memslot)
478 unsigned long len = kvm_dirty_bitmap_bytes(memslot);
480 return memslot->dirty_bitmap + len / sizeof(*memslot->dirty_bitmap);
483 #ifndef KVM_DIRTY_LOG_MANUAL_CAPS
484 #define KVM_DIRTY_LOG_MANUAL_CAPS KVM_DIRTY_LOG_MANUAL_PROTECT_ENABLE
487 struct kvm_s390_adapter_int {
500 struct kvm_xen_evtchn {
506 struct kvm_kernel_irq_routing_entry {
509 int (*set)(struct kvm_kernel_irq_routing_entry *e,
510 struct kvm *kvm, int irq_source_id, int level,
524 struct kvm_s390_adapter_int adapter;
525 struct kvm_hv_sint hv_sint;
526 struct kvm_xen_evtchn xen_evtchn;
528 struct hlist_node link;
531 #ifdef CONFIG_HAVE_KVM_IRQ_ROUTING
532 struct kvm_irq_routing_table {
533 int chip[KVM_NR_IRQCHIPS][KVM_IRQCHIP_NUM_PINS];
536 * Array indexed by gsi. Each entry contains list of irq chips
537 * the gsi is connected to.
539 struct hlist_head map[];
543 #ifndef KVM_PRIVATE_MEM_SLOTS
544 #define KVM_PRIVATE_MEM_SLOTS 0
547 #define KVM_MEM_SLOTS_NUM SHRT_MAX
548 #define KVM_USER_MEM_SLOTS (KVM_MEM_SLOTS_NUM - KVM_PRIVATE_MEM_SLOTS)
550 #ifndef __KVM_VCPU_MULTIPLE_ADDRESS_SPACE
551 static inline int kvm_arch_vcpu_memslots_id(struct kvm_vcpu *vcpu)
557 struct kvm_memslots {
559 atomic_long_t last_used_slot;
560 struct rb_root_cached hva_tree;
561 struct rb_root gfn_tree;
563 * The mapping table from slot id to memslot.
565 * 7-bit bucket count matches the size of the old id to index array for
566 * 512 slots, while giving good performance with this slot count.
567 * Higher bucket counts bring only small performance improvements but
568 * always result in higher memory usage (even for lower memslot counts).
570 DECLARE_HASHTABLE(id_hash, 7);
575 #ifdef KVM_HAVE_MMU_RWLOCK
579 #endif /* KVM_HAVE_MMU_RWLOCK */
581 struct mutex slots_lock;
584 * Protects the arch-specific fields of struct kvm_memory_slots in
585 * use by the VM. To be used under the slots_lock (above) or in a
586 * kvm->srcu critical section where acquiring the slots_lock would
587 * lead to deadlock with the synchronize_srcu in
588 * install_new_memslots.
590 struct mutex slots_arch_lock;
591 struct mm_struct *mm; /* userspace tied to this vm */
592 unsigned long nr_memslot_pages;
593 /* The two memslot sets - active and inactive (per address space) */
594 struct kvm_memslots __memslots[KVM_ADDRESS_SPACE_NUM][2];
595 /* The current active memslot set for each address space */
596 struct kvm_memslots __rcu *memslots[KVM_ADDRESS_SPACE_NUM];
597 struct xarray vcpu_array;
599 /* Used to wait for completion of MMU notifiers. */
600 spinlock_t mn_invalidate_lock;
601 unsigned long mn_active_invalidate_count;
602 struct rcuwait mn_memslots_update_rcuwait;
604 /* For management / invalidation of gfn_to_pfn_caches */
606 struct list_head gpc_list;
609 * created_vcpus is protected by kvm->lock, and is incremented
610 * at the beginning of KVM_CREATE_VCPU. online_vcpus is only
611 * incremented after storing the kvm_vcpu pointer in vcpus,
612 * and is accessed atomically.
614 atomic_t online_vcpus;
616 int last_boosted_vcpu;
617 struct list_head vm_list;
619 struct kvm_io_bus __rcu *buses[KVM_NR_BUSES];
620 #ifdef CONFIG_HAVE_KVM_EVENTFD
623 struct list_head items;
624 struct list_head resampler_list;
625 struct mutex resampler_lock;
627 struct list_head ioeventfds;
629 struct kvm_vm_stat stat;
630 struct kvm_arch arch;
631 refcount_t users_count;
632 #ifdef CONFIG_KVM_MMIO
633 struct kvm_coalesced_mmio_ring *coalesced_mmio_ring;
634 spinlock_t ring_lock;
635 struct list_head coalesced_zones;
638 struct mutex irq_lock;
639 #ifdef CONFIG_HAVE_KVM_IRQCHIP
641 * Update side is protected by irq_lock.
643 struct kvm_irq_routing_table __rcu *irq_routing;
645 #ifdef CONFIG_HAVE_KVM_IRQFD
646 struct hlist_head irq_ack_notifier_list;
649 #if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
650 struct mmu_notifier mmu_notifier;
651 unsigned long mmu_notifier_seq;
652 long mmu_notifier_count;
653 unsigned long mmu_notifier_range_start;
654 unsigned long mmu_notifier_range_end;
656 struct list_head devices;
657 u64 manual_dirty_log_protect;
658 struct dentry *debugfs_dentry;
659 struct kvm_stat_data **debugfs_stat_data;
660 struct srcu_struct srcu;
661 struct srcu_struct irq_srcu;
663 unsigned int max_halt_poll_ns;
668 #ifdef CONFIG_HAVE_KVM_PM_NOTIFIER
669 struct notifier_block pm_notifier;
671 char stats_id[KVM_STATS_NAME_SIZE];
674 #define kvm_err(fmt, ...) \
675 pr_err("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
676 #define kvm_info(fmt, ...) \
677 pr_info("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
678 #define kvm_debug(fmt, ...) \
679 pr_debug("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
680 #define kvm_debug_ratelimited(fmt, ...) \
681 pr_debug_ratelimited("kvm [%i]: " fmt, task_pid_nr(current), \
683 #define kvm_pr_unimpl(fmt, ...) \
684 pr_err_ratelimited("kvm [%i]: " fmt, \
685 task_tgid_nr(current), ## __VA_ARGS__)
687 /* The guest did something we don't support. */
688 #define vcpu_unimpl(vcpu, fmt, ...) \
689 kvm_pr_unimpl("vcpu%i, guest rIP: 0x%lx " fmt, \
690 (vcpu)->vcpu_id, kvm_rip_read(vcpu), ## __VA_ARGS__)
692 #define vcpu_debug(vcpu, fmt, ...) \
693 kvm_debug("vcpu%i " fmt, (vcpu)->vcpu_id, ## __VA_ARGS__)
694 #define vcpu_debug_ratelimited(vcpu, fmt, ...) \
695 kvm_debug_ratelimited("vcpu%i " fmt, (vcpu)->vcpu_id, \
697 #define vcpu_err(vcpu, fmt, ...) \
698 kvm_err("vcpu%i " fmt, (vcpu)->vcpu_id, ## __VA_ARGS__)
700 static inline void kvm_vm_dead(struct kvm *kvm)
703 kvm_make_all_cpus_request(kvm, KVM_REQ_VM_DEAD);
706 static inline void kvm_vm_bugged(struct kvm *kvm)
708 kvm->vm_bugged = true;
713 #define KVM_BUG(cond, kvm, fmt...) \
715 int __ret = (cond); \
717 if (WARN_ONCE(__ret && !(kvm)->vm_bugged, fmt)) \
718 kvm_vm_bugged(kvm); \
722 #define KVM_BUG_ON(cond, kvm) \
724 int __ret = (cond); \
726 if (WARN_ON_ONCE(__ret && !(kvm)->vm_bugged)) \
727 kvm_vm_bugged(kvm); \
731 static inline bool kvm_dirty_log_manual_protect_and_init_set(struct kvm *kvm)
733 return !!(kvm->manual_dirty_log_protect & KVM_DIRTY_LOG_INITIALLY_SET);
736 static inline struct kvm_io_bus *kvm_get_bus(struct kvm *kvm, enum kvm_bus idx)
738 return srcu_dereference_check(kvm->buses[idx], &kvm->srcu,
739 lockdep_is_held(&kvm->slots_lock) ||
740 !refcount_read(&kvm->users_count));
743 static inline struct kvm_vcpu *kvm_get_vcpu(struct kvm *kvm, int i)
745 int num_vcpus = atomic_read(&kvm->online_vcpus);
746 i = array_index_nospec(i, num_vcpus);
748 /* Pairs with smp_wmb() in kvm_vm_ioctl_create_vcpu. */
750 return xa_load(&kvm->vcpu_array, i);
753 #define kvm_for_each_vcpu(idx, vcpup, kvm) \
754 xa_for_each_range(&kvm->vcpu_array, idx, vcpup, 0, \
755 (atomic_read(&kvm->online_vcpus) - 1))
757 static inline struct kvm_vcpu *kvm_get_vcpu_by_id(struct kvm *kvm, int id)
759 struct kvm_vcpu *vcpu = NULL;
764 if (id < KVM_MAX_VCPUS)
765 vcpu = kvm_get_vcpu(kvm, id);
766 if (vcpu && vcpu->vcpu_id == id)
768 kvm_for_each_vcpu(i, vcpu, kvm)
769 if (vcpu->vcpu_id == id)
774 static inline int kvm_vcpu_get_idx(struct kvm_vcpu *vcpu)
776 return vcpu->vcpu_idx;
779 void kvm_destroy_vcpus(struct kvm *kvm);
781 void vcpu_load(struct kvm_vcpu *vcpu);
782 void vcpu_put(struct kvm_vcpu *vcpu);
784 #ifdef __KVM_HAVE_IOAPIC
785 void kvm_arch_post_irq_ack_notifier_list_update(struct kvm *kvm);
786 void kvm_arch_post_irq_routing_update(struct kvm *kvm);
788 static inline void kvm_arch_post_irq_ack_notifier_list_update(struct kvm *kvm)
791 static inline void kvm_arch_post_irq_routing_update(struct kvm *kvm)
796 #ifdef CONFIG_HAVE_KVM_IRQFD
797 int kvm_irqfd_init(void);
798 void kvm_irqfd_exit(void);
800 static inline int kvm_irqfd_init(void)
805 static inline void kvm_irqfd_exit(void)
809 int kvm_init(void *opaque, unsigned vcpu_size, unsigned vcpu_align,
810 struct module *module);
813 void kvm_get_kvm(struct kvm *kvm);
814 bool kvm_get_kvm_safe(struct kvm *kvm);
815 void kvm_put_kvm(struct kvm *kvm);
816 bool file_is_kvm(struct file *file);
817 void kvm_put_kvm_no_destroy(struct kvm *kvm);
819 static inline struct kvm_memslots *__kvm_memslots(struct kvm *kvm, int as_id)
821 as_id = array_index_nospec(as_id, KVM_ADDRESS_SPACE_NUM);
822 return srcu_dereference_check(kvm->memslots[as_id], &kvm->srcu,
823 lockdep_is_held(&kvm->slots_lock) ||
824 !refcount_read(&kvm->users_count));
827 static inline struct kvm_memslots *kvm_memslots(struct kvm *kvm)
829 return __kvm_memslots(kvm, 0);
832 static inline struct kvm_memslots *kvm_vcpu_memslots(struct kvm_vcpu *vcpu)
834 int as_id = kvm_arch_vcpu_memslots_id(vcpu);
836 return __kvm_memslots(vcpu->kvm, as_id);
839 static inline bool kvm_memslots_empty(struct kvm_memslots *slots)
841 return RB_EMPTY_ROOT(&slots->gfn_tree);
844 #define kvm_for_each_memslot(memslot, bkt, slots) \
845 hash_for_each(slots->id_hash, bkt, memslot, id_node[slots->node_idx]) \
846 if (WARN_ON_ONCE(!memslot->npages)) { \
850 struct kvm_memory_slot *id_to_memslot(struct kvm_memslots *slots, int id)
852 struct kvm_memory_slot *slot;
853 int idx = slots->node_idx;
855 hash_for_each_possible(slots->id_hash, slot, id_node[idx], id) {
863 /* Iterator used for walking memslots that overlap a gfn range. */
864 struct kvm_memslot_iter {
865 struct kvm_memslots *slots;
866 struct rb_node *node;
867 struct kvm_memory_slot *slot;
870 static inline void kvm_memslot_iter_next(struct kvm_memslot_iter *iter)
872 iter->node = rb_next(iter->node);
876 iter->slot = container_of(iter->node, struct kvm_memory_slot, gfn_node[iter->slots->node_idx]);
879 static inline void kvm_memslot_iter_start(struct kvm_memslot_iter *iter,
880 struct kvm_memslots *slots,
883 int idx = slots->node_idx;
885 struct kvm_memory_slot *slot;
890 * Find the so called "upper bound" of a key - the first node that has
891 * its key strictly greater than the searched one (the start gfn in our case).
894 for (tmp = slots->gfn_tree.rb_node; tmp; ) {
895 slot = container_of(tmp, struct kvm_memory_slot, gfn_node[idx]);
896 if (start < slot->base_gfn) {
905 * Find the slot with the lowest gfn that can possibly intersect with
906 * the range, so we'll ideally have slot start <= range start
910 * A NULL previous node means that the very first slot
911 * already has a higher start gfn.
912 * In this case slot start > range start.
914 tmp = rb_prev(iter->node);
918 /* a NULL node below means no slots */
919 iter->node = rb_last(&slots->gfn_tree);
923 iter->slot = container_of(iter->node, struct kvm_memory_slot, gfn_node[idx]);
926 * It is possible in the slot start < range start case that the
927 * found slot ends before or at range start (slot end <= range start)
928 * and so it does not overlap the requested range.
930 * In such non-overlapping case the next slot (if it exists) will
931 * already have slot start > range start, otherwise the logic above
932 * would have found it instead of the current slot.
934 if (iter->slot->base_gfn + iter->slot->npages <= start)
935 kvm_memslot_iter_next(iter);
939 static inline bool kvm_memslot_iter_is_valid(struct kvm_memslot_iter *iter, gfn_t end)
945 * If this slot starts beyond or at the end of the range so does
948 return iter->slot->base_gfn < end;
951 /* Iterate over each memslot at least partially intersecting [start, end) range */
952 #define kvm_for_each_memslot_in_gfn_range(iter, slots, start, end) \
953 for (kvm_memslot_iter_start(iter, slots, start); \
954 kvm_memslot_iter_is_valid(iter, end); \
955 kvm_memslot_iter_next(iter))
958 * KVM_SET_USER_MEMORY_REGION ioctl allows the following operations:
959 * - create a new memory slot
960 * - delete an existing memory slot
961 * - modify an existing memory slot
962 * -- move it in the guest physical memory space
963 * -- just change its flags
965 * Since flags can be changed by some of these operations, the following
966 * differentiation is the best we can do for __kvm_set_memory_region():
975 int kvm_set_memory_region(struct kvm *kvm,
976 const struct kvm_userspace_memory_region *mem);
977 int __kvm_set_memory_region(struct kvm *kvm,
978 const struct kvm_userspace_memory_region *mem);
979 void kvm_arch_free_memslot(struct kvm *kvm, struct kvm_memory_slot *slot);
980 void kvm_arch_memslots_updated(struct kvm *kvm, u64 gen);
981 int kvm_arch_prepare_memory_region(struct kvm *kvm,
982 const struct kvm_memory_slot *old,
983 struct kvm_memory_slot *new,
984 enum kvm_mr_change change);
985 void kvm_arch_commit_memory_region(struct kvm *kvm,
986 struct kvm_memory_slot *old,
987 const struct kvm_memory_slot *new,
988 enum kvm_mr_change change);
989 /* flush all memory translations */
990 void kvm_arch_flush_shadow_all(struct kvm *kvm);
991 /* flush memory translations pointing to 'slot' */
992 void kvm_arch_flush_shadow_memslot(struct kvm *kvm,
993 struct kvm_memory_slot *slot);
995 int gfn_to_page_many_atomic(struct kvm_memory_slot *slot, gfn_t gfn,
996 struct page **pages, int nr_pages);
998 struct page *gfn_to_page(struct kvm *kvm, gfn_t gfn);
999 unsigned long gfn_to_hva(struct kvm *kvm, gfn_t gfn);
1000 unsigned long gfn_to_hva_prot(struct kvm *kvm, gfn_t gfn, bool *writable);
1001 unsigned long gfn_to_hva_memslot(struct kvm_memory_slot *slot, gfn_t gfn);
1002 unsigned long gfn_to_hva_memslot_prot(struct kvm_memory_slot *slot, gfn_t gfn,
1004 void kvm_release_page_clean(struct page *page);
1005 void kvm_release_page_dirty(struct page *page);
1006 void kvm_set_page_accessed(struct page *page);
1008 kvm_pfn_t gfn_to_pfn(struct kvm *kvm, gfn_t gfn);
1009 kvm_pfn_t gfn_to_pfn_prot(struct kvm *kvm, gfn_t gfn, bool write_fault,
1011 kvm_pfn_t gfn_to_pfn_memslot(const struct kvm_memory_slot *slot, gfn_t gfn);
1012 kvm_pfn_t gfn_to_pfn_memslot_atomic(const struct kvm_memory_slot *slot, gfn_t gfn);
1013 kvm_pfn_t __gfn_to_pfn_memslot(const struct kvm_memory_slot *slot, gfn_t gfn,
1014 bool atomic, bool *async, bool write_fault,
1015 bool *writable, hva_t *hva);
1017 void kvm_release_pfn_clean(kvm_pfn_t pfn);
1018 void kvm_release_pfn_dirty(kvm_pfn_t pfn);
1019 void kvm_set_pfn_dirty(kvm_pfn_t pfn);
1020 void kvm_set_pfn_accessed(kvm_pfn_t pfn);
1022 void kvm_release_pfn(kvm_pfn_t pfn, bool dirty);
1023 int kvm_read_guest_page(struct kvm *kvm, gfn_t gfn, void *data, int offset,
1025 int kvm_read_guest(struct kvm *kvm, gpa_t gpa, void *data, unsigned long len);
1026 int kvm_read_guest_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
1027 void *data, unsigned long len);
1028 int kvm_read_guest_offset_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
1029 void *data, unsigned int offset,
1031 int kvm_write_guest_page(struct kvm *kvm, gfn_t gfn, const void *data,
1032 int offset, int len);
1033 int kvm_write_guest(struct kvm *kvm, gpa_t gpa, const void *data,
1035 int kvm_write_guest_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
1036 void *data, unsigned long len);
1037 int kvm_write_guest_offset_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
1038 void *data, unsigned int offset,
1040 int kvm_gfn_to_hva_cache_init(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
1041 gpa_t gpa, unsigned long len);
1043 #define __kvm_get_guest(kvm, gfn, offset, v) \
1045 unsigned long __addr = gfn_to_hva(kvm, gfn); \
1046 typeof(v) __user *__uaddr = (typeof(__uaddr))(__addr + offset); \
1047 int __ret = -EFAULT; \
1049 if (!kvm_is_error_hva(__addr)) \
1050 __ret = get_user(v, __uaddr); \
1054 #define kvm_get_guest(kvm, gpa, v) \
1056 gpa_t __gpa = gpa; \
1057 struct kvm *__kvm = kvm; \
1059 __kvm_get_guest(__kvm, __gpa >> PAGE_SHIFT, \
1060 offset_in_page(__gpa), v); \
1063 #define __kvm_put_guest(kvm, gfn, offset, v) \
1065 unsigned long __addr = gfn_to_hva(kvm, gfn); \
1066 typeof(v) __user *__uaddr = (typeof(__uaddr))(__addr + offset); \
1067 int __ret = -EFAULT; \
1069 if (!kvm_is_error_hva(__addr)) \
1070 __ret = put_user(v, __uaddr); \
1072 mark_page_dirty(kvm, gfn); \
1076 #define kvm_put_guest(kvm, gpa, v) \
1078 gpa_t __gpa = gpa; \
1079 struct kvm *__kvm = kvm; \
1081 __kvm_put_guest(__kvm, __gpa >> PAGE_SHIFT, \
1082 offset_in_page(__gpa), v); \
1085 int kvm_clear_guest(struct kvm *kvm, gpa_t gpa, unsigned long len);
1086 struct kvm_memory_slot *gfn_to_memslot(struct kvm *kvm, gfn_t gfn);
1087 bool kvm_is_visible_gfn(struct kvm *kvm, gfn_t gfn);
1088 bool kvm_vcpu_is_visible_gfn(struct kvm_vcpu *vcpu, gfn_t gfn);
1089 unsigned long kvm_host_page_size(struct kvm_vcpu *vcpu, gfn_t gfn);
1090 void mark_page_dirty_in_slot(struct kvm *kvm, const struct kvm_memory_slot *memslot, gfn_t gfn);
1091 void mark_page_dirty(struct kvm *kvm, gfn_t gfn);
1093 struct kvm_memslots *kvm_vcpu_memslots(struct kvm_vcpu *vcpu);
1094 struct kvm_memory_slot *kvm_vcpu_gfn_to_memslot(struct kvm_vcpu *vcpu, gfn_t gfn);
1095 kvm_pfn_t kvm_vcpu_gfn_to_pfn_atomic(struct kvm_vcpu *vcpu, gfn_t gfn);
1096 kvm_pfn_t kvm_vcpu_gfn_to_pfn(struct kvm_vcpu *vcpu, gfn_t gfn);
1097 int kvm_vcpu_map(struct kvm_vcpu *vcpu, gpa_t gpa, struct kvm_host_map *map);
1098 struct page *kvm_vcpu_gfn_to_page(struct kvm_vcpu *vcpu, gfn_t gfn);
1099 void kvm_vcpu_unmap(struct kvm_vcpu *vcpu, struct kvm_host_map *map, bool dirty);
1100 unsigned long kvm_vcpu_gfn_to_hva(struct kvm_vcpu *vcpu, gfn_t gfn);
1101 unsigned long kvm_vcpu_gfn_to_hva_prot(struct kvm_vcpu *vcpu, gfn_t gfn, bool *writable);
1102 int kvm_vcpu_read_guest_page(struct kvm_vcpu *vcpu, gfn_t gfn, void *data, int offset,
1104 int kvm_vcpu_read_guest_atomic(struct kvm_vcpu *vcpu, gpa_t gpa, void *data,
1106 int kvm_vcpu_read_guest(struct kvm_vcpu *vcpu, gpa_t gpa, void *data,
1108 int kvm_vcpu_write_guest_page(struct kvm_vcpu *vcpu, gfn_t gfn, const void *data,
1109 int offset, int len);
1110 int kvm_vcpu_write_guest(struct kvm_vcpu *vcpu, gpa_t gpa, const void *data,
1112 void kvm_vcpu_mark_page_dirty(struct kvm_vcpu *vcpu, gfn_t gfn);
1115 * kvm_gfn_to_pfn_cache_init - prepare a cached kernel mapping and HPA for a
1116 * given guest physical address.
1118 * @kvm: pointer to kvm instance.
1119 * @gpc: struct gfn_to_pfn_cache object.
1120 * @vcpu: vCPU to be used for marking pages dirty and to be woken on
1122 * @guest_uses_pa: indicates that the resulting host physical PFN is used while
1123 * @vcpu is IN_GUEST_MODE so invalidations should wake it.
1124 * @kernel_map: requests a kernel virtual mapping (kmap / memremap).
1125 * @gpa: guest physical address to map.
1126 * @len: sanity check; the range being access must fit a single page.
1127 * @dirty: mark the cache dirty immediately.
1129 * @return: 0 for success.
1130 * -EINVAL for a mapping which would cross a page boundary.
1131 * -EFAULT for an untranslatable guest physical address.
1133 * This primes a gfn_to_pfn_cache and links it into the @kvm's list for
1134 * invalidations to be processed. Invalidation callbacks to @vcpu using
1135 * %KVM_REQ_GPC_INVALIDATE will occur only for MMU notifiers, not for KVM
1136 * memslot changes. Callers are required to use kvm_gfn_to_pfn_cache_check()
1137 * to ensure that the cache is valid before accessing the target page.
1139 int kvm_gfn_to_pfn_cache_init(struct kvm *kvm, struct gfn_to_pfn_cache *gpc,
1140 struct kvm_vcpu *vcpu, bool guest_uses_pa,
1141 bool kernel_map, gpa_t gpa, unsigned long len,
1145 * kvm_gfn_to_pfn_cache_check - check validity of a gfn_to_pfn_cache.
1147 * @kvm: pointer to kvm instance.
1148 * @gpc: struct gfn_to_pfn_cache object.
1149 * @gpa: current guest physical address to map.
1150 * @len: sanity check; the range being access must fit a single page.
1151 * @dirty: mark the cache dirty immediately.
1153 * @return: %true if the cache is still valid and the address matches.
1154 * %false if the cache is not valid.
1156 * Callers outside IN_GUEST_MODE context should hold a read lock on @gpc->lock
1157 * while calling this function, and then continue to hold the lock until the
1158 * access is complete.
1160 * Callers in IN_GUEST_MODE may do so without locking, although they should
1161 * still hold a read lock on kvm->scru for the memslot checks.
1163 bool kvm_gfn_to_pfn_cache_check(struct kvm *kvm, struct gfn_to_pfn_cache *gpc,
1164 gpa_t gpa, unsigned long len);
1167 * kvm_gfn_to_pfn_cache_refresh - update a previously initialized cache.
1169 * @kvm: pointer to kvm instance.
1170 * @gpc: struct gfn_to_pfn_cache object.
1171 * @gpa: updated guest physical address to map.
1172 * @len: sanity check; the range being access must fit a single page.
1173 * @dirty: mark the cache dirty immediately.
1175 * @return: 0 for success.
1176 * -EINVAL for a mapping which would cross a page boundary.
1177 * -EFAULT for an untranslatable guest physical address.
1179 * This will attempt to refresh a gfn_to_pfn_cache. Note that a successful
1180 * returm from this function does not mean the page can be immediately
1181 * accessed because it may have raced with an invalidation. Callers must
1182 * still lock and check the cache status, as this function does not return
1183 * with the lock still held to permit access.
1185 int kvm_gfn_to_pfn_cache_refresh(struct kvm *kvm, struct gfn_to_pfn_cache *gpc,
1186 gpa_t gpa, unsigned long len, bool dirty);
1189 * kvm_gfn_to_pfn_cache_unmap - temporarily unmap a gfn_to_pfn_cache.
1191 * @kvm: pointer to kvm instance.
1192 * @gpc: struct gfn_to_pfn_cache object.
1194 * This unmaps the referenced page and marks it dirty, if appropriate. The
1195 * cache is left in the invalid state but at least the mapping from GPA to
1196 * userspace HVA will remain cached and can be reused on a subsequent
1199 void kvm_gfn_to_pfn_cache_unmap(struct kvm *kvm, struct gfn_to_pfn_cache *gpc);
1202 * kvm_gfn_to_pfn_cache_destroy - destroy and unlink a gfn_to_pfn_cache.
1204 * @kvm: pointer to kvm instance.
1205 * @gpc: struct gfn_to_pfn_cache object.
1207 * This removes a cache from the @kvm's list to be processed on MMU notifier
1210 void kvm_gfn_to_pfn_cache_destroy(struct kvm *kvm, struct gfn_to_pfn_cache *gpc);
1212 void kvm_sigset_activate(struct kvm_vcpu *vcpu);
1213 void kvm_sigset_deactivate(struct kvm_vcpu *vcpu);
1215 void kvm_vcpu_halt(struct kvm_vcpu *vcpu);
1216 bool kvm_vcpu_block(struct kvm_vcpu *vcpu);
1217 void kvm_arch_vcpu_blocking(struct kvm_vcpu *vcpu);
1218 void kvm_arch_vcpu_unblocking(struct kvm_vcpu *vcpu);
1219 bool kvm_vcpu_wake_up(struct kvm_vcpu *vcpu);
1220 void kvm_vcpu_kick(struct kvm_vcpu *vcpu);
1221 int kvm_vcpu_yield_to(struct kvm_vcpu *target);
1222 void kvm_vcpu_on_spin(struct kvm_vcpu *vcpu, bool usermode_vcpu_not_eligible);
1224 void kvm_flush_remote_tlbs(struct kvm *kvm);
1225 void kvm_reload_remote_mmus(struct kvm *kvm);
1227 #ifdef KVM_ARCH_NR_OBJS_PER_MEMORY_CACHE
1228 int kvm_mmu_topup_memory_cache(struct kvm_mmu_memory_cache *mc, int min);
1229 int kvm_mmu_memory_cache_nr_free_objects(struct kvm_mmu_memory_cache *mc);
1230 void kvm_mmu_free_memory_cache(struct kvm_mmu_memory_cache *mc);
1231 void *kvm_mmu_memory_cache_alloc(struct kvm_mmu_memory_cache *mc);
1234 void kvm_inc_notifier_count(struct kvm *kvm, unsigned long start,
1236 void kvm_dec_notifier_count(struct kvm *kvm, unsigned long start,
1239 long kvm_arch_dev_ioctl(struct file *filp,
1240 unsigned int ioctl, unsigned long arg);
1241 long kvm_arch_vcpu_ioctl(struct file *filp,
1242 unsigned int ioctl, unsigned long arg);
1243 vm_fault_t kvm_arch_vcpu_fault(struct kvm_vcpu *vcpu, struct vm_fault *vmf);
1245 int kvm_vm_ioctl_check_extension(struct kvm *kvm, long ext);
1247 void kvm_arch_mmu_enable_log_dirty_pt_masked(struct kvm *kvm,
1248 struct kvm_memory_slot *slot,
1250 unsigned long mask);
1251 void kvm_arch_sync_dirty_log(struct kvm *kvm, struct kvm_memory_slot *memslot);
1253 #ifdef CONFIG_KVM_GENERIC_DIRTYLOG_READ_PROTECT
1254 void kvm_arch_flush_remote_tlbs_memslot(struct kvm *kvm,
1255 const struct kvm_memory_slot *memslot);
1256 #else /* !CONFIG_KVM_GENERIC_DIRTYLOG_READ_PROTECT */
1257 int kvm_vm_ioctl_get_dirty_log(struct kvm *kvm, struct kvm_dirty_log *log);
1258 int kvm_get_dirty_log(struct kvm *kvm, struct kvm_dirty_log *log,
1259 int *is_dirty, struct kvm_memory_slot **memslot);
1262 int kvm_vm_ioctl_irq_line(struct kvm *kvm, struct kvm_irq_level *irq_level,
1264 int kvm_vm_ioctl_enable_cap(struct kvm *kvm,
1265 struct kvm_enable_cap *cap);
1266 long kvm_arch_vm_ioctl(struct file *filp,
1267 unsigned int ioctl, unsigned long arg);
1269 int kvm_arch_vcpu_ioctl_get_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu);
1270 int kvm_arch_vcpu_ioctl_set_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu);
1272 int kvm_arch_vcpu_ioctl_translate(struct kvm_vcpu *vcpu,
1273 struct kvm_translation *tr);
1275 int kvm_arch_vcpu_ioctl_get_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs);
1276 int kvm_arch_vcpu_ioctl_set_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs);
1277 int kvm_arch_vcpu_ioctl_get_sregs(struct kvm_vcpu *vcpu,
1278 struct kvm_sregs *sregs);
1279 int kvm_arch_vcpu_ioctl_set_sregs(struct kvm_vcpu *vcpu,
1280 struct kvm_sregs *sregs);
1281 int kvm_arch_vcpu_ioctl_get_mpstate(struct kvm_vcpu *vcpu,
1282 struct kvm_mp_state *mp_state);
1283 int kvm_arch_vcpu_ioctl_set_mpstate(struct kvm_vcpu *vcpu,
1284 struct kvm_mp_state *mp_state);
1285 int kvm_arch_vcpu_ioctl_set_guest_debug(struct kvm_vcpu *vcpu,
1286 struct kvm_guest_debug *dbg);
1287 int kvm_arch_vcpu_ioctl_run(struct kvm_vcpu *vcpu);
1289 int kvm_arch_init(void *opaque);
1290 void kvm_arch_exit(void);
1292 void kvm_arch_sched_in(struct kvm_vcpu *vcpu, int cpu);
1294 void kvm_arch_vcpu_load(struct kvm_vcpu *vcpu, int cpu);
1295 void kvm_arch_vcpu_put(struct kvm_vcpu *vcpu);
1296 int kvm_arch_vcpu_precreate(struct kvm *kvm, unsigned int id);
1297 int kvm_arch_vcpu_create(struct kvm_vcpu *vcpu);
1298 void kvm_arch_vcpu_postcreate(struct kvm_vcpu *vcpu);
1299 void kvm_arch_vcpu_destroy(struct kvm_vcpu *vcpu);
1301 #ifdef CONFIG_HAVE_KVM_PM_NOTIFIER
1302 int kvm_arch_pm_notifier(struct kvm *kvm, unsigned long state);
1305 #ifdef __KVM_HAVE_ARCH_VCPU_DEBUGFS
1306 void kvm_arch_create_vcpu_debugfs(struct kvm_vcpu *vcpu, struct dentry *debugfs_dentry);
1309 int kvm_arch_hardware_enable(void);
1310 void kvm_arch_hardware_disable(void);
1311 int kvm_arch_hardware_setup(void *opaque);
1312 void kvm_arch_hardware_unsetup(void);
1313 int kvm_arch_check_processor_compat(void *opaque);
1314 int kvm_arch_vcpu_runnable(struct kvm_vcpu *vcpu);
1315 bool kvm_arch_vcpu_in_kernel(struct kvm_vcpu *vcpu);
1316 int kvm_arch_vcpu_should_kick(struct kvm_vcpu *vcpu);
1317 bool kvm_arch_dy_runnable(struct kvm_vcpu *vcpu);
1318 bool kvm_arch_dy_has_pending_interrupt(struct kvm_vcpu *vcpu);
1319 int kvm_arch_post_init_vm(struct kvm *kvm);
1320 void kvm_arch_pre_destroy_vm(struct kvm *kvm);
1321 int kvm_arch_create_vm_debugfs(struct kvm *kvm);
1323 #ifndef __KVM_HAVE_ARCH_VM_ALLOC
1325 * All architectures that want to use vzalloc currently also
1326 * need their own kvm_arch_alloc_vm implementation.
1328 static inline struct kvm *kvm_arch_alloc_vm(void)
1330 return kzalloc(sizeof(struct kvm), GFP_KERNEL);
1334 static inline void __kvm_arch_free_vm(struct kvm *kvm)
1339 #ifndef __KVM_HAVE_ARCH_VM_FREE
1340 static inline void kvm_arch_free_vm(struct kvm *kvm)
1342 __kvm_arch_free_vm(kvm);
1346 #ifndef __KVM_HAVE_ARCH_FLUSH_REMOTE_TLB
1347 static inline int kvm_arch_flush_remote_tlb(struct kvm *kvm)
1353 #ifdef __KVM_HAVE_ARCH_NONCOHERENT_DMA
1354 void kvm_arch_register_noncoherent_dma(struct kvm *kvm);
1355 void kvm_arch_unregister_noncoherent_dma(struct kvm *kvm);
1356 bool kvm_arch_has_noncoherent_dma(struct kvm *kvm);
1358 static inline void kvm_arch_register_noncoherent_dma(struct kvm *kvm)
1362 static inline void kvm_arch_unregister_noncoherent_dma(struct kvm *kvm)
1366 static inline bool kvm_arch_has_noncoherent_dma(struct kvm *kvm)
1371 #ifdef __KVM_HAVE_ARCH_ASSIGNED_DEVICE
1372 void kvm_arch_start_assignment(struct kvm *kvm);
1373 void kvm_arch_end_assignment(struct kvm *kvm);
1374 bool kvm_arch_has_assigned_device(struct kvm *kvm);
1376 static inline void kvm_arch_start_assignment(struct kvm *kvm)
1380 static inline void kvm_arch_end_assignment(struct kvm *kvm)
1384 static inline bool kvm_arch_has_assigned_device(struct kvm *kvm)
1390 static inline struct rcuwait *kvm_arch_vcpu_get_wait(struct kvm_vcpu *vcpu)
1392 #ifdef __KVM_HAVE_ARCH_WQP
1393 return vcpu->arch.waitp;
1400 * Wake a vCPU if necessary, but don't do any stats/metadata updates. Returns
1401 * true if the vCPU was blocking and was awakened, false otherwise.
1403 static inline bool __kvm_vcpu_wake_up(struct kvm_vcpu *vcpu)
1405 return !!rcuwait_wake_up(kvm_arch_vcpu_get_wait(vcpu));
1408 static inline bool kvm_vcpu_is_blocking(struct kvm_vcpu *vcpu)
1410 return rcuwait_active(kvm_arch_vcpu_get_wait(vcpu));
1413 #ifdef __KVM_HAVE_ARCH_INTC_INITIALIZED
1415 * returns true if the virtual interrupt controller is initialized and
1416 * ready to accept virtual IRQ. On some architectures the virtual interrupt
1417 * controller is dynamically instantiated and this is not always true.
1419 bool kvm_arch_intc_initialized(struct kvm *kvm);
1421 static inline bool kvm_arch_intc_initialized(struct kvm *kvm)
1427 #ifdef CONFIG_GUEST_PERF_EVENTS
1428 unsigned long kvm_arch_vcpu_get_ip(struct kvm_vcpu *vcpu);
1430 void kvm_register_perf_callbacks(unsigned int (*pt_intr_handler)(void));
1431 void kvm_unregister_perf_callbacks(void);
1433 static inline void kvm_register_perf_callbacks(void *ign) {}
1434 static inline void kvm_unregister_perf_callbacks(void) {}
1435 #endif /* CONFIG_GUEST_PERF_EVENTS */
1437 int kvm_arch_init_vm(struct kvm *kvm, unsigned long type);
1438 void kvm_arch_destroy_vm(struct kvm *kvm);
1439 void kvm_arch_sync_events(struct kvm *kvm);
1441 int kvm_cpu_has_pending_timer(struct kvm_vcpu *vcpu);
1443 bool kvm_is_reserved_pfn(kvm_pfn_t pfn);
1444 bool kvm_is_zone_device_pfn(kvm_pfn_t pfn);
1446 struct kvm_irq_ack_notifier {
1447 struct hlist_node link;
1449 void (*irq_acked)(struct kvm_irq_ack_notifier *kian);
1452 int kvm_irq_map_gsi(struct kvm *kvm,
1453 struct kvm_kernel_irq_routing_entry *entries, int gsi);
1454 int kvm_irq_map_chip_pin(struct kvm *kvm, unsigned irqchip, unsigned pin);
1456 int kvm_set_irq(struct kvm *kvm, int irq_source_id, u32 irq, int level,
1458 int kvm_set_msi(struct kvm_kernel_irq_routing_entry *irq_entry, struct kvm *kvm,
1459 int irq_source_id, int level, bool line_status);
1460 int kvm_arch_set_irq_inatomic(struct kvm_kernel_irq_routing_entry *e,
1461 struct kvm *kvm, int irq_source_id,
1462 int level, bool line_status);
1463 bool kvm_irq_has_notifier(struct kvm *kvm, unsigned irqchip, unsigned pin);
1464 void kvm_notify_acked_gsi(struct kvm *kvm, int gsi);
1465 void kvm_notify_acked_irq(struct kvm *kvm, unsigned irqchip, unsigned pin);
1466 void kvm_register_irq_ack_notifier(struct kvm *kvm,
1467 struct kvm_irq_ack_notifier *kian);
1468 void kvm_unregister_irq_ack_notifier(struct kvm *kvm,
1469 struct kvm_irq_ack_notifier *kian);
1470 int kvm_request_irq_source_id(struct kvm *kvm);
1471 void kvm_free_irq_source_id(struct kvm *kvm, int irq_source_id);
1472 bool kvm_arch_irqfd_allowed(struct kvm *kvm, struct kvm_irqfd *args);
1475 * Returns a pointer to the memslot if it contains gfn.
1476 * Otherwise returns NULL.
1478 static inline struct kvm_memory_slot *
1479 try_get_memslot(struct kvm_memory_slot *slot, gfn_t gfn)
1484 if (gfn >= slot->base_gfn && gfn < slot->base_gfn + slot->npages)
1491 * Returns a pointer to the memslot that contains gfn. Otherwise returns NULL.
1493 * With "approx" set returns the memslot also when the address falls
1494 * in a hole. In that case one of the memslots bordering the hole is
1497 static inline struct kvm_memory_slot *
1498 search_memslots(struct kvm_memslots *slots, gfn_t gfn, bool approx)
1500 struct kvm_memory_slot *slot;
1501 struct rb_node *node;
1502 int idx = slots->node_idx;
1505 for (node = slots->gfn_tree.rb_node; node; ) {
1506 slot = container_of(node, struct kvm_memory_slot, gfn_node[idx]);
1507 if (gfn >= slot->base_gfn) {
1508 if (gfn < slot->base_gfn + slot->npages)
1510 node = node->rb_right;
1512 node = node->rb_left;
1515 return approx ? slot : NULL;
1518 static inline struct kvm_memory_slot *
1519 ____gfn_to_memslot(struct kvm_memslots *slots, gfn_t gfn, bool approx)
1521 struct kvm_memory_slot *slot;
1523 slot = (struct kvm_memory_slot *)atomic_long_read(&slots->last_used_slot);
1524 slot = try_get_memslot(slot, gfn);
1528 slot = search_memslots(slots, gfn, approx);
1530 atomic_long_set(&slots->last_used_slot, (unsigned long)slot);
1538 * __gfn_to_memslot() and its descendants are here to allow arch code to inline
1539 * the lookups in hot paths. gfn_to_memslot() itself isn't here as an inline
1540 * because that would bloat other code too much.
1542 static inline struct kvm_memory_slot *
1543 __gfn_to_memslot(struct kvm_memslots *slots, gfn_t gfn)
1545 return ____gfn_to_memslot(slots, gfn, false);
1548 static inline unsigned long
1549 __gfn_to_hva_memslot(const struct kvm_memory_slot *slot, gfn_t gfn)
1552 * The index was checked originally in search_memslots. To avoid
1553 * that a malicious guest builds a Spectre gadget out of e.g. page
1554 * table walks, do not let the processor speculate loads outside
1555 * the guest's registered memslots.
1557 unsigned long offset = gfn - slot->base_gfn;
1558 offset = array_index_nospec(offset, slot->npages);
1559 return slot->userspace_addr + offset * PAGE_SIZE;
1562 static inline int memslot_id(struct kvm *kvm, gfn_t gfn)
1564 return gfn_to_memslot(kvm, gfn)->id;
1568 hva_to_gfn_memslot(unsigned long hva, struct kvm_memory_slot *slot)
1570 gfn_t gfn_offset = (hva - slot->userspace_addr) >> PAGE_SHIFT;
1572 return slot->base_gfn + gfn_offset;
1575 static inline gpa_t gfn_to_gpa(gfn_t gfn)
1577 return (gpa_t)gfn << PAGE_SHIFT;
1580 static inline gfn_t gpa_to_gfn(gpa_t gpa)
1582 return (gfn_t)(gpa >> PAGE_SHIFT);
1585 static inline hpa_t pfn_to_hpa(kvm_pfn_t pfn)
1587 return (hpa_t)pfn << PAGE_SHIFT;
1590 static inline struct page *kvm_vcpu_gpa_to_page(struct kvm_vcpu *vcpu,
1593 return kvm_vcpu_gfn_to_page(vcpu, gpa_to_gfn(gpa));
1596 static inline bool kvm_is_error_gpa(struct kvm *kvm, gpa_t gpa)
1598 unsigned long hva = gfn_to_hva(kvm, gpa_to_gfn(gpa));
1600 return kvm_is_error_hva(hva);
1603 enum kvm_stat_kind {
1608 struct kvm_stat_data {
1610 const struct _kvm_stats_desc *desc;
1611 enum kvm_stat_kind kind;
1614 struct _kvm_stats_desc {
1615 struct kvm_stats_desc desc;
1616 char name[KVM_STATS_NAME_SIZE];
1619 #define STATS_DESC_COMMON(type, unit, base, exp, sz, bsz) \
1620 .flags = type | unit | base | \
1621 BUILD_BUG_ON_ZERO(type & ~KVM_STATS_TYPE_MASK) | \
1622 BUILD_BUG_ON_ZERO(unit & ~KVM_STATS_UNIT_MASK) | \
1623 BUILD_BUG_ON_ZERO(base & ~KVM_STATS_BASE_MASK), \
1628 #define VM_GENERIC_STATS_DESC(stat, type, unit, base, exp, sz, bsz) \
1631 STATS_DESC_COMMON(type, unit, base, exp, sz, bsz), \
1632 .offset = offsetof(struct kvm_vm_stat, generic.stat) \
1636 #define VCPU_GENERIC_STATS_DESC(stat, type, unit, base, exp, sz, bsz) \
1639 STATS_DESC_COMMON(type, unit, base, exp, sz, bsz), \
1640 .offset = offsetof(struct kvm_vcpu_stat, generic.stat) \
1644 #define VM_STATS_DESC(stat, type, unit, base, exp, sz, bsz) \
1647 STATS_DESC_COMMON(type, unit, base, exp, sz, bsz), \
1648 .offset = offsetof(struct kvm_vm_stat, stat) \
1652 #define VCPU_STATS_DESC(stat, type, unit, base, exp, sz, bsz) \
1655 STATS_DESC_COMMON(type, unit, base, exp, sz, bsz), \
1656 .offset = offsetof(struct kvm_vcpu_stat, stat) \
1660 /* SCOPE: VM, VM_GENERIC, VCPU, VCPU_GENERIC */
1661 #define STATS_DESC(SCOPE, stat, type, unit, base, exp, sz, bsz) \
1662 SCOPE##_STATS_DESC(stat, type, unit, base, exp, sz, bsz)
1664 #define STATS_DESC_CUMULATIVE(SCOPE, name, unit, base, exponent) \
1665 STATS_DESC(SCOPE, name, KVM_STATS_TYPE_CUMULATIVE, \
1666 unit, base, exponent, 1, 0)
1667 #define STATS_DESC_INSTANT(SCOPE, name, unit, base, exponent) \
1668 STATS_DESC(SCOPE, name, KVM_STATS_TYPE_INSTANT, \
1669 unit, base, exponent, 1, 0)
1670 #define STATS_DESC_PEAK(SCOPE, name, unit, base, exponent) \
1671 STATS_DESC(SCOPE, name, KVM_STATS_TYPE_PEAK, \
1672 unit, base, exponent, 1, 0)
1673 #define STATS_DESC_LINEAR_HIST(SCOPE, name, unit, base, exponent, sz, bsz) \
1674 STATS_DESC(SCOPE, name, KVM_STATS_TYPE_LINEAR_HIST, \
1675 unit, base, exponent, sz, bsz)
1676 #define STATS_DESC_LOG_HIST(SCOPE, name, unit, base, exponent, sz) \
1677 STATS_DESC(SCOPE, name, KVM_STATS_TYPE_LOG_HIST, \
1678 unit, base, exponent, sz, 0)
1680 /* Cumulative counter, read/write */
1681 #define STATS_DESC_COUNTER(SCOPE, name) \
1682 STATS_DESC_CUMULATIVE(SCOPE, name, KVM_STATS_UNIT_NONE, \
1683 KVM_STATS_BASE_POW10, 0)
1684 /* Instantaneous counter, read only */
1685 #define STATS_DESC_ICOUNTER(SCOPE, name) \
1686 STATS_DESC_INSTANT(SCOPE, name, KVM_STATS_UNIT_NONE, \
1687 KVM_STATS_BASE_POW10, 0)
1688 /* Peak counter, read/write */
1689 #define STATS_DESC_PCOUNTER(SCOPE, name) \
1690 STATS_DESC_PEAK(SCOPE, name, KVM_STATS_UNIT_NONE, \
1691 KVM_STATS_BASE_POW10, 0)
1693 /* Cumulative time in nanosecond */
1694 #define STATS_DESC_TIME_NSEC(SCOPE, name) \
1695 STATS_DESC_CUMULATIVE(SCOPE, name, KVM_STATS_UNIT_SECONDS, \
1696 KVM_STATS_BASE_POW10, -9)
1697 /* Linear histogram for time in nanosecond */
1698 #define STATS_DESC_LINHIST_TIME_NSEC(SCOPE, name, sz, bsz) \
1699 STATS_DESC_LINEAR_HIST(SCOPE, name, KVM_STATS_UNIT_SECONDS, \
1700 KVM_STATS_BASE_POW10, -9, sz, bsz)
1701 /* Logarithmic histogram for time in nanosecond */
1702 #define STATS_DESC_LOGHIST_TIME_NSEC(SCOPE, name, sz) \
1703 STATS_DESC_LOG_HIST(SCOPE, name, KVM_STATS_UNIT_SECONDS, \
1704 KVM_STATS_BASE_POW10, -9, sz)
1706 #define KVM_GENERIC_VM_STATS() \
1707 STATS_DESC_COUNTER(VM_GENERIC, remote_tlb_flush), \
1708 STATS_DESC_COUNTER(VM_GENERIC, remote_tlb_flush_requests)
1710 #define KVM_GENERIC_VCPU_STATS() \
1711 STATS_DESC_COUNTER(VCPU_GENERIC, halt_successful_poll), \
1712 STATS_DESC_COUNTER(VCPU_GENERIC, halt_attempted_poll), \
1713 STATS_DESC_COUNTER(VCPU_GENERIC, halt_poll_invalid), \
1714 STATS_DESC_COUNTER(VCPU_GENERIC, halt_wakeup), \
1715 STATS_DESC_TIME_NSEC(VCPU_GENERIC, halt_poll_success_ns), \
1716 STATS_DESC_TIME_NSEC(VCPU_GENERIC, halt_poll_fail_ns), \
1717 STATS_DESC_TIME_NSEC(VCPU_GENERIC, halt_wait_ns), \
1718 STATS_DESC_LOGHIST_TIME_NSEC(VCPU_GENERIC, halt_poll_success_hist, \
1719 HALT_POLL_HIST_COUNT), \
1720 STATS_DESC_LOGHIST_TIME_NSEC(VCPU_GENERIC, halt_poll_fail_hist, \
1721 HALT_POLL_HIST_COUNT), \
1722 STATS_DESC_LOGHIST_TIME_NSEC(VCPU_GENERIC, halt_wait_hist, \
1723 HALT_POLL_HIST_COUNT), \
1724 STATS_DESC_ICOUNTER(VCPU_GENERIC, blocking)
1726 extern struct dentry *kvm_debugfs_dir;
1728 ssize_t kvm_stats_read(char *id, const struct kvm_stats_header *header,
1729 const struct _kvm_stats_desc *desc,
1730 void *stats, size_t size_stats,
1731 char __user *user_buffer, size_t size, loff_t *offset);
1734 * kvm_stats_linear_hist_update() - Update bucket value for linear histogram
1737 * @data: start address of the stats data
1738 * @size: the number of bucket of the stats data
1739 * @value: the new value used to update the linear histogram's bucket
1740 * @bucket_size: the size (width) of a bucket
1742 static inline void kvm_stats_linear_hist_update(u64 *data, size_t size,
1743 u64 value, size_t bucket_size)
1745 size_t index = div64_u64(value, bucket_size);
1747 index = min(index, size - 1);
1752 * kvm_stats_log_hist_update() - Update bucket value for logarithmic histogram
1755 * @data: start address of the stats data
1756 * @size: the number of bucket of the stats data
1757 * @value: the new value used to update the logarithmic histogram's bucket
1759 static inline void kvm_stats_log_hist_update(u64 *data, size_t size, u64 value)
1761 size_t index = fls64(value);
1763 index = min(index, size - 1);
1767 #define KVM_STATS_LINEAR_HIST_UPDATE(array, value, bsize) \
1768 kvm_stats_linear_hist_update(array, ARRAY_SIZE(array), value, bsize)
1769 #define KVM_STATS_LOG_HIST_UPDATE(array, value) \
1770 kvm_stats_log_hist_update(array, ARRAY_SIZE(array), value)
1773 extern const struct kvm_stats_header kvm_vm_stats_header;
1774 extern const struct _kvm_stats_desc kvm_vm_stats_desc[];
1775 extern const struct kvm_stats_header kvm_vcpu_stats_header;
1776 extern const struct _kvm_stats_desc kvm_vcpu_stats_desc[];
1778 #if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
1779 static inline int mmu_notifier_retry(struct kvm *kvm, unsigned long mmu_seq)
1781 if (unlikely(kvm->mmu_notifier_count))
1784 * Ensure the read of mmu_notifier_count happens before the read
1785 * of mmu_notifier_seq. This interacts with the smp_wmb() in
1786 * mmu_notifier_invalidate_range_end to make sure that the caller
1787 * either sees the old (non-zero) value of mmu_notifier_count or
1788 * the new (incremented) value of mmu_notifier_seq.
1789 * PowerPC Book3s HV KVM calls this under a per-page lock
1790 * rather than under kvm->mmu_lock, for scalability, so
1791 * can't rely on kvm->mmu_lock to keep things ordered.
1794 if (kvm->mmu_notifier_seq != mmu_seq)
1799 static inline int mmu_notifier_retry_hva(struct kvm *kvm,
1800 unsigned long mmu_seq,
1803 lockdep_assert_held(&kvm->mmu_lock);
1805 * If mmu_notifier_count is non-zero, then the range maintained by
1806 * kvm_mmu_notifier_invalidate_range_start contains all addresses that
1807 * might be being invalidated. Note that it may include some false
1808 * positives, due to shortcuts when handing concurrent invalidations.
1810 if (unlikely(kvm->mmu_notifier_count) &&
1811 hva >= kvm->mmu_notifier_range_start &&
1812 hva < kvm->mmu_notifier_range_end)
1814 if (kvm->mmu_notifier_seq != mmu_seq)
1820 #ifdef CONFIG_HAVE_KVM_IRQ_ROUTING
1822 #define KVM_MAX_IRQ_ROUTES 4096 /* might need extension/rework in the future */
1824 bool kvm_arch_can_set_irq_routing(struct kvm *kvm);
1825 int kvm_set_irq_routing(struct kvm *kvm,
1826 const struct kvm_irq_routing_entry *entries,
1829 int kvm_set_routing_entry(struct kvm *kvm,
1830 struct kvm_kernel_irq_routing_entry *e,
1831 const struct kvm_irq_routing_entry *ue);
1832 void kvm_free_irq_routing(struct kvm *kvm);
1836 static inline void kvm_free_irq_routing(struct kvm *kvm) {}
1840 int kvm_send_userspace_msi(struct kvm *kvm, struct kvm_msi *msi);
1842 #ifdef CONFIG_HAVE_KVM_EVENTFD
1844 void kvm_eventfd_init(struct kvm *kvm);
1845 int kvm_ioeventfd(struct kvm *kvm, struct kvm_ioeventfd *args);
1847 #ifdef CONFIG_HAVE_KVM_IRQFD
1848 int kvm_irqfd(struct kvm *kvm, struct kvm_irqfd *args);
1849 void kvm_irqfd_release(struct kvm *kvm);
1850 void kvm_irq_routing_update(struct kvm *);
1852 static inline int kvm_irqfd(struct kvm *kvm, struct kvm_irqfd *args)
1857 static inline void kvm_irqfd_release(struct kvm *kvm) {}
1862 static inline void kvm_eventfd_init(struct kvm *kvm) {}
1864 static inline int kvm_irqfd(struct kvm *kvm, struct kvm_irqfd *args)
1869 static inline void kvm_irqfd_release(struct kvm *kvm) {}
1871 #ifdef CONFIG_HAVE_KVM_IRQCHIP
1872 static inline void kvm_irq_routing_update(struct kvm *kvm)
1877 static inline int kvm_ioeventfd(struct kvm *kvm, struct kvm_ioeventfd *args)
1882 #endif /* CONFIG_HAVE_KVM_EVENTFD */
1884 void kvm_arch_irq_routing_update(struct kvm *kvm);
1886 static inline void kvm_make_request(int req, struct kvm_vcpu *vcpu)
1889 * Ensure the rest of the request is published to kvm_check_request's
1890 * caller. Paired with the smp_mb__after_atomic in kvm_check_request.
1893 set_bit(req & KVM_REQUEST_MASK, (void *)&vcpu->requests);
1896 static inline bool kvm_request_pending(struct kvm_vcpu *vcpu)
1898 return READ_ONCE(vcpu->requests);
1901 static inline bool kvm_test_request(int req, struct kvm_vcpu *vcpu)
1903 return test_bit(req & KVM_REQUEST_MASK, (void *)&vcpu->requests);
1906 static inline void kvm_clear_request(int req, struct kvm_vcpu *vcpu)
1908 clear_bit(req & KVM_REQUEST_MASK, (void *)&vcpu->requests);
1911 static inline bool kvm_check_request(int req, struct kvm_vcpu *vcpu)
1913 if (kvm_test_request(req, vcpu)) {
1914 kvm_clear_request(req, vcpu);
1917 * Ensure the rest of the request is visible to kvm_check_request's
1918 * caller. Paired with the smp_wmb in kvm_make_request.
1920 smp_mb__after_atomic();
1927 extern bool kvm_rebooting;
1929 extern unsigned int halt_poll_ns;
1930 extern unsigned int halt_poll_ns_grow;
1931 extern unsigned int halt_poll_ns_grow_start;
1932 extern unsigned int halt_poll_ns_shrink;
1935 const struct kvm_device_ops *ops;
1938 struct list_head vm_node;
1941 /* create, destroy, and name are mandatory */
1942 struct kvm_device_ops {
1946 * create is called holding kvm->lock and any operations not suitable
1947 * to do while holding the lock should be deferred to init (see
1950 int (*create)(struct kvm_device *dev, u32 type);
1953 * init is called after create if create is successful and is called
1954 * outside of holding kvm->lock.
1956 void (*init)(struct kvm_device *dev);
1959 * Destroy is responsible for freeing dev.
1961 * Destroy may be called before or after destructors are called
1962 * on emulated I/O regions, depending on whether a reference is
1963 * held by a vcpu or other kvm component that gets destroyed
1964 * after the emulated I/O.
1966 void (*destroy)(struct kvm_device *dev);
1969 * Release is an alternative method to free the device. It is
1970 * called when the device file descriptor is closed. Once
1971 * release is called, the destroy method will not be called
1972 * anymore as the device is removed from the device list of
1973 * the VM. kvm->lock is held.
1975 void (*release)(struct kvm_device *dev);
1977 int (*set_attr)(struct kvm_device *dev, struct kvm_device_attr *attr);
1978 int (*get_attr)(struct kvm_device *dev, struct kvm_device_attr *attr);
1979 int (*has_attr)(struct kvm_device *dev, struct kvm_device_attr *attr);
1980 long (*ioctl)(struct kvm_device *dev, unsigned int ioctl,
1982 int (*mmap)(struct kvm_device *dev, struct vm_area_struct *vma);
1985 void kvm_device_get(struct kvm_device *dev);
1986 void kvm_device_put(struct kvm_device *dev);
1987 struct kvm_device *kvm_device_from_filp(struct file *filp);
1988 int kvm_register_device_ops(const struct kvm_device_ops *ops, u32 type);
1989 void kvm_unregister_device_ops(u32 type);
1991 extern struct kvm_device_ops kvm_mpic_ops;
1992 extern struct kvm_device_ops kvm_arm_vgic_v2_ops;
1993 extern struct kvm_device_ops kvm_arm_vgic_v3_ops;
1995 #ifdef CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT
1997 static inline void kvm_vcpu_set_in_spin_loop(struct kvm_vcpu *vcpu, bool val)
1999 vcpu->spin_loop.in_spin_loop = val;
2001 static inline void kvm_vcpu_set_dy_eligible(struct kvm_vcpu *vcpu, bool val)
2003 vcpu->spin_loop.dy_eligible = val;
2006 #else /* !CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT */
2008 static inline void kvm_vcpu_set_in_spin_loop(struct kvm_vcpu *vcpu, bool val)
2012 static inline void kvm_vcpu_set_dy_eligible(struct kvm_vcpu *vcpu, bool val)
2015 #endif /* CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT */
2017 static inline bool kvm_is_visible_memslot(struct kvm_memory_slot *memslot)
2019 return (memslot && memslot->id < KVM_USER_MEM_SLOTS &&
2020 !(memslot->flags & KVM_MEMSLOT_INVALID));
2023 struct kvm_vcpu *kvm_get_running_vcpu(void);
2024 struct kvm_vcpu * __percpu *kvm_get_running_vcpus(void);
2026 #ifdef CONFIG_HAVE_KVM_IRQ_BYPASS
2027 bool kvm_arch_has_irq_bypass(void);
2028 int kvm_arch_irq_bypass_add_producer(struct irq_bypass_consumer *,
2029 struct irq_bypass_producer *);
2030 void kvm_arch_irq_bypass_del_producer(struct irq_bypass_consumer *,
2031 struct irq_bypass_producer *);
2032 void kvm_arch_irq_bypass_stop(struct irq_bypass_consumer *);
2033 void kvm_arch_irq_bypass_start(struct irq_bypass_consumer *);
2034 int kvm_arch_update_irqfd_routing(struct kvm *kvm, unsigned int host_irq,
2035 uint32_t guest_irq, bool set);
2036 bool kvm_arch_irqfd_route_changed(struct kvm_kernel_irq_routing_entry *,
2037 struct kvm_kernel_irq_routing_entry *);
2038 #endif /* CONFIG_HAVE_KVM_IRQ_BYPASS */
2040 #ifdef CONFIG_HAVE_KVM_INVALID_WAKEUPS
2041 /* If we wakeup during the poll time, was it a sucessful poll? */
2042 static inline bool vcpu_valid_wakeup(struct kvm_vcpu *vcpu)
2044 return vcpu->valid_wakeup;
2048 static inline bool vcpu_valid_wakeup(struct kvm_vcpu *vcpu)
2052 #endif /* CONFIG_HAVE_KVM_INVALID_WAKEUPS */
2054 #ifdef CONFIG_HAVE_KVM_NO_POLL
2055 /* Callback that tells if we must not poll */
2056 bool kvm_arch_no_poll(struct kvm_vcpu *vcpu);
2058 static inline bool kvm_arch_no_poll(struct kvm_vcpu *vcpu)
2062 #endif /* CONFIG_HAVE_KVM_NO_POLL */
2064 #ifdef CONFIG_HAVE_KVM_VCPU_ASYNC_IOCTL
2065 long kvm_arch_vcpu_async_ioctl(struct file *filp,
2066 unsigned int ioctl, unsigned long arg);
2068 static inline long kvm_arch_vcpu_async_ioctl(struct file *filp,
2072 return -ENOIOCTLCMD;
2074 #endif /* CONFIG_HAVE_KVM_VCPU_ASYNC_IOCTL */
2076 void kvm_arch_mmu_notifier_invalidate_range(struct kvm *kvm,
2077 unsigned long start, unsigned long end);
2079 #ifdef CONFIG_HAVE_KVM_VCPU_RUN_PID_CHANGE
2080 int kvm_arch_vcpu_run_pid_change(struct kvm_vcpu *vcpu);
2082 static inline int kvm_arch_vcpu_run_pid_change(struct kvm_vcpu *vcpu)
2086 #endif /* CONFIG_HAVE_KVM_VCPU_RUN_PID_CHANGE */
2088 typedef int (*kvm_vm_thread_fn_t)(struct kvm *kvm, uintptr_t data);
2090 int kvm_vm_create_worker_thread(struct kvm *kvm, kvm_vm_thread_fn_t thread_fn,
2091 uintptr_t data, const char *name,
2092 struct task_struct **thread_ptr);
2094 #ifdef CONFIG_KVM_XFER_TO_GUEST_WORK
2095 static inline void kvm_handle_signal_exit(struct kvm_vcpu *vcpu)
2097 vcpu->run->exit_reason = KVM_EXIT_INTR;
2098 vcpu->stat.signal_exits++;
2100 #endif /* CONFIG_KVM_XFER_TO_GUEST_WORK */
2103 * This defines how many reserved entries we want to keep before we
2104 * kick the vcpu to the userspace to avoid dirty ring full. This
2105 * value can be tuned to higher if e.g. PML is enabled on the host.
2107 #define KVM_DIRTY_RING_RSVD_ENTRIES 64
2109 /* Max number of entries allowed for each kvm dirty ring */
2110 #define KVM_DIRTY_RING_MAX_ENTRIES 65536